The Impact of Employee Activism on the Capital Markets
218
The Impact of Employee Activism on the Capital Markets A thesis submitted to The University of Manchester for the degree of Doctor of Philosophy in the Faculty of Humanities 2019 Xianglong Chen Alliance Manchester Business School
Transcript of The Impact of Employee Activism on the Capital Markets
The Impact of Employee Activism on the
Capital Markets
A thesis submitted to The University of Manchester for the degree of
Doctor of Philosophy
in the Faculty of Humanities
2019
Xianglong Chen
Alliance Manchester Business School
2
Table of Contents
ABSTRACT ................................................................................................................ 6
DECLARATION ........................................................................................................ 7
COPYRIGHT STATEMENT .................................................................................... 7
ACKNOWLEDGEMENTS ........................................................................................ 9
CHAPTER 1. INTRODUCTION ............................................................................. 11
1.1 Motivation ........................................................................................................ 11
1.2. Thesis Structure .............................................................................................. 16
References .............................................................................................................. 18
CHAPTER 2. DOES EMPLOYEE OWNERSHIP REDUCE STRIKE RIKS?
EVIDENCE FROM UNION ELECTIONS ............................................................. 21
2.1 Introduction ..................................................................................................... 22
2.2 Literature Review and Hypothesis Development ........................................... 27
2.2.1 Literature Review ........................................................................................ 27
2.2.1.1 Labour Unions and Strike Risk............................................................. 27
2.2.1.2 Employee Ownership ........................................................................... 31
2.2.2 Hypothesis Development............................................................................. 34
2.3 Data and Research Design ............................................................................... 37
2.3.1 Data ............................................................................................................ 37
2.3.1.1 Union Election Data ............................................................................ 38
2.3.1.2 Employee Stock Options Data .............................................................. 39
2.3.1.3 Labour Strike Data .............................................................................. 41
2.3.2 Sample Construction ................................................................................... 41
2.3.3 Summary Statistics ...................................................................................... 44
2.3.4 Research Design.......................................................................................... 44
2.3.4.1 Identification Strategy .......................................................................... 44
2.3.4.2 Empirical Models ................................................................................ 46
2.4. Empirical Findings ......................................................................................... 48
2.4.1 Moderating Effect of ESO Incentives on Union Strike Probability .............. 48
2.4.2 ESO Incentives Granted in Response to Union Elections ............................. 50
2.4.2.1 Evidence from Regression Discontinuity Design (RDD) Analysis: Local
Linear Regressions .......................................................................................... 50
2.4.2.2 RD Plots .............................................................................................. 52
3
2.4.3 Right-to-Work (RTW) Laws ....................................................................... 53
2.4.4 Role of Labour Skills .................................................................................. 54
2.4.5 Placebo Test ................................................................................................ 56
2.5 Conclusion........................................................................................................ 57
References .............................................................................................................. 60
Supplementary Appendix ..................................................................................... 81
CHAPTER 3. DOES CORPORATE SOCIAL RESPONSIBILITY SPENDING
AFFECT STRIKE RISK? EVIDENCE FROM UNION ELECTIONS ................. 95
3.1 Introduction ..................................................................................................... 96
3.2 Literature Review and Hypothesis Development ......................................... 102
3.2.1 Literature Review ...................................................................................... 102
3.2.1.1 Union Strikes ..................................................................................... 102
3.2.1.2 CSR Spending .................................................................................... 103
3.2.2 Hypothesis Development........................................................................... 107
3.2.2.1 CSR and Union Strike Probability ...................................................... 107
3.2.2.2 CSR as a Strategic Tool ..................................................................... 110
3.3 Data and Research Design ............................................................................. 111
3.3.1 Data and Sample ....................................................................................... 111
3.3.1.1 Union Election Data .......................................................................... 112
3.3.1.2 CSR Data ........................................................................................... 112
3.3.1.3 Labour Strikes Data ........................................................................... 114
3.3.2 Sample Construction ................................................................................. 114
3.3.3 Summary Statistics .................................................................................... 116
3.3.4 Research Design........................................................................................ 116
3.3.4.1 Identification Strategy ........................................................................ 116
3.3.4.2 Empirical Models .............................................................................. 117
3.4 Empirical Findings ........................................................................................ 119
3.4.1 CSR Spending and Union Strike Probability ............................................. 119
3.4.1.1 Overall CSR Level ............................................................................. 119
3.4.1.2 Decomposition of CSR ....................................................................... 120
3.4.2 CSR Adjustment in Response to Unionisation ........................................... 122
3.4.2.1 Financial Constraints ........................................................................ 123
3.4.2.2 Sin Industries ..................................................................................... 124
3.4.2.3 Product Market Competition .............................................................. 126
4
3.4.3 Robustness Test: Propensity Score Matched Sample ................................. 127
3.5 Conclusion...................................................................................................... 129
References ............................................................................................................ 131
Appendix .............................................................................................................. 149
CHAPTER 4. DO FINANCIAL ANALYSTS PLAY A COMPLEMENTARY
ROLE OR SUBSTITUTIVE ROLE IN THE CORPORATE INFORMATION
ENVIRONMENT? EVIDENCE FROM ORGANISED LABOUR ...................... 151
4.1 Introduction ................................................................................................... 152
4.2. Related Literature and Hypothesis Development........................................ 159
4.2.1 Related Literature ...................................................................................... 159
4.2.1.1 Financial Analysts ............................................................................. 159
4.2.1.2 Labour Unions ................................................................................... 165
4.2.2 Hypothesis Development........................................................................... 168
4.2.2.1 Labour Unions and Financial Analysts: “Complementary Role” ....... 168
4.2.2.2 Labour Unions and Financial Analysts: “Substitutive Role” .............. 172
4.3 Data and Methodology .................................................................................. 176
4.3.1 Data Sources and Sample Construction ..................................................... 176
4.3.2 Main Variables .......................................................................................... 177
4.3.2.1 Labour Unionisation Rate .................................................................. 177
4.3.2.2 Analyst Forecast Variables ................................................................ 177
4.3.3 Summary Statistics .................................................................................... 178
4.3.4 Empirical Models ...................................................................................... 179
4.4 Empirical Findings ........................................................................................ 180
4.4.1 Baseline Results ........................................................................................ 180
4.4.1.1 Labour Unions and Forecast Accuracy .............................................. 180
4.4.1.2 Labour Unions and Forecast Dispersion............................................ 181
4.4.2 Verification of Channels: Uncertainty versus Financial Reporting Quality 182
4.4.2.1 Proxies for Financial Reporting Quality ............................................ 184
4.4.2.2 Incremental Effect of Union Representation ....................................... 185
4.4.3 Right-to-Work (RTW) Legislation ............................................................ 187
4.4.4 Role of Labour Skills ................................................................................ 188
4.4.5 Mitigating Role of Labour Costs Information ............................................ 190
4.4.6 Strategic Optimism Bias ............................................................................ 191
4.5 Conclusion...................................................................................................... 193
5
References ............................................................................................................ 196
Appendix .............................................................................................................. 213
CHAPTER 5. SUMMARY AND SUGGESTIONS FOR FUTURE RESEARCH
................................................................................................................................. 214
This thesis contains 61,485 words including title page, tables, and footnotes.
6
Abstract
The University of Manchester
Xianglong Chen
Doctor of Philosophy (PhD)
The Impact of Employee Activism on the Capital Markets
September 2019
In this thesis, I present three self-contained essays studying the impact of employee
activism, an emerging phenomenon, on the capital markets. My first two essays
examine whether interest alignment and resource competition influence union strike risk,
while in my third essay, I evaluate whether financial analysts enrich the information set
for investors in anticipation of such employee risk. Together, these studies shed light on
the significant impact of employee activism on corporate decisions and the information
environment in the capital markets.
In the first essay, I examine the impact of employee stock options (ESO) on union strike
risk. Consistent with ESO realigning the interests of organised labour with those of their
employers, I find that firms offering higher levels of equity incentives to their
employees are exposed to significantly lower strike risk following unionisation. I also
provide evidence that managers strategically grant more ESO incentives in response to
unionisation, as a way to proactively improve the interest alignment between employees
and firms. My findings have important implications for accounting standard setters and
policymakers. Despite the benefit of ESO, the current accounting treatment of equity-
based compensation inhibits the expansion of employee ownership. Thus, my study
calls for more policy support to promote employee ownership in the heavily unionised
industrial firms in the U.S. as well as other jurisdictions, such as Europe, where union
activism is also prevalent.
The second essay explores the effect of corporate social responsibility (CSR) spending
on union strike risk. I find that CSR expenditure in non-employee dimensions, such as
community and environment, exacerbates strike risk following unionisation, whereas
employee-related CSR spending mitigates union strike risk. These contrasting effects
suggest that a high level of non-employee CSR spending intensifies the resource
competition between employees and other stakeholders. My findings suggest that
managers should regularly review their relationships with different stakeholders, and
highlight the importance of a balanced approach to stakeholder management when
making decisions regarding CSR investments. Overall, this study sheds light on the
inter-stakeholder relationship through the lens of the employees.
In the third and final essay, I investigate the interaction between organised labour and
sell-side analysts, key financial information intermediaries in the capital markets. Using
a large U.S. sample, I document that the labour unionisation rate is associated with
lower forecast accuracy and higher forecast dispersion in analysts’ earnings forecasts,
implying that financial analysts predominantly serve a “complementary role” rather than
a “substitutive role” when firms are subject to heightened uncertainty in human capital.
Crucially, further analysis indicates that the availability of labour cost information
significantly mitigates unions’ negative impact on analysts’ forecast quality, confirming
analysts’ reliance on publicly disclosed information. Overall, this paper shows that the
influence of organised labour extends beyond the corporate boundary to a group of
sophisticated market participants, and highlights the value relevance of disclosure
specifically related to human capital, in terms of improving the information
environment of the capital markets.
7
Declaration
No portion of the work referred to in the thesis has been submitted in support of an
application for another degree or qualification of this or any other university or other
institute of learning.
Copyright Statement
i. The author of this thesis (including any appendices and/or schedules to this thesis)
owns certain copyright or related rights in it (the “Copyright”) and s/he has given The
University of Manchester certain rights to use such Copyright, including for
administrative purposes.
ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic
copy, may be made only in accordance with the Copyright, Designs and Patents Act
1988 (as amended) and regulations issued under it or, where appropriate, in accordance
with licensing agreements which the University has from time to time. This page must
form part of any such copies made.
iii. The ownership of certain Copyright, patents, designs, trademarks and other
intellectual property (the “Intellectual Property”) and any reproductions of copyright
works in the thesis, for example graphs and tables (“Reproductions”), which may be
described in this thesis, may not be owned by the author and may be owned by third
parties. Such Intellectual Property and Reproductions cannot and must not be made
available for use without the prior written permission of the owner(s) of the relevant
Intellectual Property and/or Reproductions.
iv. Further information on the conditions under which disclosure, publication and
commercialisation of this thesis, the Copyright and any Intellectual Property and/or
Reproductions described in it may take place is available in the University IP Policy
(see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=24420), in any relevant
Thesis restriction declarations deposited in the University Library, The University
Library’s regulations (see http://www.library.manchester.ac.uk/about/regulations/) and
in The University’s policy on Presentation of Theses.
8
This thesis is dedicated to my wife, Renee Wang, and my parents, Wujun Chen and
Fang Liu, for all their unconditional love, sacrifice and support!
9
Acknowledgements
This PhD journey has been undoubtedly the most challenging yet most rewarding
experience in my life. Having benefitted tremendously from countless people over the
years, I know that writing this section is going to be just as difficult as the thesis itself.
This thesis simply would not have been possible without the contribution and support of
the following people.
First and foremost, I am eternally indebted to my greatest supervisors: Prof.
Konstantinos Stathopoulos and Prof. Edward Lee. For the past four years, I have been
extremely fortunate to have the opportunity to pursue my PhD research under their
excellent supervision. Words cannot express my gratitude for their guidance, nurturing,
patience and encouragement throughout this challenging process. I have thoroughly
enjoyed the numerous intellectual discussions and research meetings we had together
over the years and it was such a fulfilling learning process for me. Looking back, I am
truly appreciative for their tolerance and continuous support, both academically and
mentally, during those difficult moments. I cannot thank them both enough for always
keeping their doors open to me whenever I wish to talk to them and hear their advice.
They always encouraged me to look on the bright side, motivated me to step out of my
comfort zone and constantly challenge myself. I would not have been able to overcome
all the difficulties and successfully complete this PhD marathon without their constant
attention, care and trust. I also like to thank them for giving me the academic freedom to
explore the research questions I am passionate about. Not only did they provide
valuable comments and guidance leading up to this thesis, they also helped me to grow
and mature as an individual. Their academic rigour, critical thinking, dedication,
enthusiasm have profoundly shaped me as an academic researcher and will continue to
inspire me in my future career. I feel immensely privileged and grateful to have Kostas
and Edward as my PhD supervisors, who I will forever respect and cherish as my life
mentors!
I must also thank Prof. Norman Strong and Prof. Neslihan Ozkan for accepting to be the
examiners of my thesis. It is truly an honour for me to have such distinguished
professors as my examiners. Special thanks go to Prof. Andrew Stark, who was the
chair of my PhD panel reviews, for his constructive comments and valuable advice. I
remember how much I enjoyed listening to his insightful and thought-provoking
discussions, which have helped shape my work.
I would also like to take this opportunity to extend my gratitude to other faculty
members who have provided me solid research training in the first year or given me
helpful comments and emotional support at various phases of my PhD research: Prof.
Martin Walker, Prof. Chris Humphrey, Prof. Marie Dutordoir, Dr. Christos Begkos, Dr.
Ning Gao, Dr. Thomas Schleicher, Dr. Wei Jiang, Dr. Simon Kim, Dr. Alice Xu and Dr.
Colin Zeng. My thesis has also greatly benefited from the helpful comments and
valuable discussions by the following scholars: Prof. Steven Young, Prof. Richard
Barker, Prof. Giovanna Michelon, Prof. Mark Clatworthy, Prof. Ajay Patel, Prof. Ethan
10
Rouen, Prof. James Ohlson, Prof. Theodore Christensen, Prof. Jaideep Shenoy, Prof.
Caroline Flammer, Prof. Dan Amiram, Dr. Mohamed Ghaly and Dr. Tuan Ho.
I would also like to express my special appreciation to Prof. Annita Florou and Dr.
Catherine Chen. Without their encouragement, advice and recommendations during my
MSc studies at King’s College London, I would not have started this amazing journey. I
am extremely grateful to Prof. Annita Florou for her continued support at different
stages of my PhD.
My special thanks go to my good friends, Chen Hua and Dr. Zhangfan Cao, for their
consistent support. Chen shared the same office with me for the past four years and I
greatly appreciate his technical and emotional support throughout this incredible
journey. I must also thank Zhangfan for frequently visiting me in Manchester to cheer
me up. I am so lucky to have such supportive and loyal friends around me and I will
always treasure our friendship and precious memories we had together in the UK. I
would also like to thank other PhD colleagues and friends who shared this journey with
me for their encouragement and company: Dr. Nikos Tsileponis, Dr. Mostafa Harakeh,
Dimitrios Christoforakis, Najeeba Alzaimoor, Marta Almeida, Perla Mardini, Wei Liu,
Kara Ng, Lei Ni, Yingyin Lin, Jihye Kim and Dhruba Borah, Dr. Xiao Jiang, Ola
Akintola, Hunter Cai and Xin Sun. I also wish to thank the lovely staff at PGR office:
Lynne Barlow-Cheetham, Mark Falzon, Madonna Fyne-Maguire and Kristin Trichler
for their excellent assistance with the administrative matters during my PhD studies.
I am deeply indebted to my beloved parents, Wujun Chen and Fang Liu, for bringing
me to this world, giving me the best possible education and encouraging me to pursue
my dream. I would not have been where I am today without their unconditional love,
sacrifice and support. I cannot thank them enough for everything they have sacrificed
for me in the past thirty years.
Last, but certainly not least, I owe my sincere gratitude to my wife, Renee Lei Wang,
who left her family and a promising job just for me. I am so blessed to have such a
caring, supportive and considerate wife and life partner. I must also thank my little
angel, Creamie, for her company on those sleepless nights and for the much-needed
comfort and endless joy she brings to me along this journey.
I gratefully acknowledge the generous financial support from Alliance Manchester
Business School Doctoral Scholarship and The University of Manchester President’s
Doctoral Scholar Award to fund my PhD research.
My heartfelt gratitude goes to all of you for being part of this truly memorable journey!
Steven Xianglong Chen
Manchester, September 2019
11
Chapter 1
Introduction
1.1 Motivation
Managers and investors have long been concerned about risk, which has direct and
fundamental implications for corporate value and shareholder wealth (Campbell 1996).
In the aftermath of the largest economic crisis since the Great Depression, risk
management has become one of the top priorities of managers and the focal point for
investors, regulators and academics (Carrel 2010; Pirson and Turnbull 2011). Yet, a
prominent source of risk that managers increasingly face is the risk of labour strikes,
amid the rising phenomenon of employee activism (Coulman 2019; Edgecliffe-Johnson
2019). In today’s knowledge-based economy, employees play an unprecedentedly
important role in creating value for businesses and fuelling the growth of economies
(Barro 2001; McCracken et al. 2017). Crucially, employees are not only human capital,
but also a key stakeholder within firms. Edward Freeman, the founding father of
stakeholder theory, argues that the 21st century is a century of “Managing for
Stakeholders” (Freeman et al. 2007). Against the backdrop of firms’ increasing reliance
on human capital (Zingales 2000) and complexity of employee relations (Agrawal
2012), managing employees remains a challenging and urgent task for companies.
The growing prominence of employee activism and companies’ urgent need for better
risk management against labour strikes has accelerated the integration of labour
economics into capital markets research by accounting and finance scholars. The
existing literature in this intersection can be broadly categorised into two themes. One
strand of literature has focused on the impact of employee power on firm performance
and its implications for investors. This line of research documents a largely adverse
12
effect of employee activism on firm performance and corporate value due to rent
extraction and interest misalignment (Clark 1984; DiNardo and Lee 2004; Chen et al.
2011; Agrawal 2012; Lee and Mas 2012; Bradley et al. 2017; Campello et al. 2018).
Another strand of literature examines the interactions between employees and managers
and shows that employees can influence a wide spectrum of corporate decisions such as
accounting choices (Liberty and Zimmerman 1986; Siu et al. 2009; Hamm et al. 2018),
capital structure (Klasa et al. 2009; Matsa 2010), executive compensation (Huang et al.
2017) and corporate disclosure (Chung et al. 2016).
Nevertheless, there is a lack of direct evidence and discussion on how to better manage
employees and alleviate the risk of labour strikes. Furthermore, surprisingly little is
known about the influence of employees beyond corporations in the financial markets.
Hence, the purpose of my thesis is to deepen our understanding of the behaviour as well
as the influence of employees, as both a powerful stakeholder and a valuable intangible
asset, and to shed light on the significant role they can play in the capital markets.
Specifically, in my thesis, I focus on a specific group of activist employees: unionised
employees. When represented by labour unions, acting as collective-bargaining units,
employees possess greater bargaining power and impose higher strike risk on their
employers (Ashenfelter and Johnson 1969; Becker and Olson 1986; Cramton et al.
1999). Thus, unionisation, representing an increase in the collective power of employees
and intrinsic risk within firms, offers me an ideal setting in which to examine the impact
of employee activism on market participants. This thesis consists of three self-contained
essays in Chapters 2, 3 and 4, respectively. While each essay is independent of the
others, there is a coherent theme, namely, the impact of organised labour, as a powerful
stakeholder, on corporate decisions and the information environment in the capital
markets. I will briefly introduce each of the chapters below.
13
Motivated by the proliferation of employee ownership schemes, in Chapter 2, I examine
the effect of employee stock options (ESO) on union strike risk. Prior literature suggests
that the predominant reason for offering equity-based incentives to rank-and-file
employees is to motivate and retain key talent (Bhagat et al. 1985; Blasi et al. 2002;
Kim and Ouimet 2014; Call et al. 2016), and documents an overall positive impact on
firms in terms of productivity and financial performance (Rosen and Quarrey 1987;
Beatty 1995; Hochberg and Lindsey 2010; Fang et al. 2015). However, previous studies
focus on the benefits of ESO predominantly in the high-tech sector, where equity-based
incentives are more prevalent. So far, little is known about the implications of employee
ownership in unionised industries, which tend to be labour-intensive yet strategically
important. By exploiting the setting of union elections in U.S. firms, I find that those
offering higher levels of equity incentives to their employees are exposed to a
significantly lower likelihood of post-unionisation strikes, consistent with ESO
improving the interest alignment between employees and firms (i.e., employers).
Importantly, further analysis of firms’ ESO-granting behaviour around union election
events indicates that, in response to unionisation, managers strategically grant more
ESO incentives, to mitigate the strike risk by proactively aligning employees’ and
shareholders’ interests. Overall, my paper presents novel evidence of how employee
ownership moderates the behaviour of organised labour and reshapes labour-
management relations. Specifically, unlike ESO in high-tech industries, which are used
to retain and incentivise talent, in the context of unionised industries, my findings
suggest that ESO can be used by firms to improve interest alignment and thus better
management against strike risk. This paper has important implications for accounting
standard setters and policymakers. Despite the evident benefit of ESO, my study reveals
that the current accounting treatment of equity-based compensation (i.e., FAS 123R)
14
creates barriers to the expansion of employee ownership schemes, which highlights the
urgent need for more policy support to promote employee ownership within the labour-
intensive industrial firms that are key to revitalising the U.S. economy.
Chapter 3 explores labour unions’ attitude towards corporate social responsibility (CSR)
spending, against the backdrop of the phenomenal growth in such spending and rising
emphasis on stakeholder management. Employees, along with the community,
customers and environmentalists, among others, fall under the CSR framework. In a
multi-stakeholder environment, I conjecture that firms’ inability to meet the demands of
all the stakeholders due to limited financial resources is likely to cause resource
competition amongst stakeholders. In line with this conjecture, I find that firms with
high levels of non-employee CSR spending face a significantly higher risk of union
strikes, while high levels of employee-related CSR spending significantly mitigate such
risk. These results suggest that spending a disproportionally high amount on CSR could
intensify the resource competition between employees and other stakeholders.
Consequently, labour unions are likely to impose more pressure on managers by
instigating strikes so as to extract more of the scarce resources and ensure they have
priority over other stakeholders. I also present supportive evidence that firms
strategically reduce non-employee CSR expenditure following unionisation, in order to
mitigate the increased strike risk, though that adjustment is less salient among firms that
have a greater reliance on CSR spending due to their greater need to signal quality. My
results reveal an unintended consequence of CSR spending, that is, resource
competition amongst stakeholders, thus implying that managers should take a balanced
approach to stakeholder management, and strategically adjust CSR spending based on
regular reviews of the firm’s relations with different stakeholders. Overall, this chapter
15
sheds light on the inter-stakeholder relationship through the lens of employees, a key
stakeholder.
In Chapter 4, I study the influence of organised labour on the information environment
of capital markets. Specifically, I investigate whether financial analysts, as sophisticated
information intermediaries, are affected by the context of unionised firms, for which
investors have a greater information demand due to increased uncertainty in human
capital. Previous literature suggests that labour unions bring uncertainty to businesses
and weaken the information environment (Hilary 2006; Chen et al. 2011; Bova 2013;
Chung et al. 2016), which may negatively affect analysts’ forecasts, while another
strand of literature argues that financial analysts produce valuable information through
dedicated research (Asquith et al. 2005; Barron et al. 2008; Bradshaw et al. 2017; Loh
and Stulz 2018; Jennings 2019). It is unclear how financial analysts perform in the case
of high uncertainty in human capital. Using a large U.S. sample, I find that the labour
unionisation rate is associated with lower forecast accuracy and higher forecast
dispersion, implying that financial analysts predominantly play a “complementary role”
rather than a “substitutive role” when firms are facing significant uncertainty in human
capital. Cross-sectional analyses indicate that the union impact on analysts’ forecasts is
more pronounced for firms in low-skilled industries and firms headquartered in the non-
Right-to-Work states, where labour unions are more powerful. Notably, I also find
evidence of financial analysts’ reliance on labour cost information and strategic
optimism, corroborating the argument that they do rely more on corporate disclosure
than generating first-hand information through original research. My findings highlight
the impact of employees beyond the corporate boundary, on a group of sophisticated
information intermediaries in the financial markets. In light of the declining usefulness
of financial statements, my paper sheds light on the value relevance of human capital
16
information, and calls for the disclosure of such information to supplement the existing
financial reporting system and improve the information efficiency of the capital markets.
Overall, my thesis enhances our understanding of employee activism and its influence
on managerial decisions as well as the information environment in the capital markets.
Together, the three empirical studies in this thesis consistently demonstrate that
employees are powerful stakeholders and value-relevant intangible assets, thus
deserving more attention from various market participants, such as managers,
information intermediaries and policymakers.
1.2. Thesis Structure
The thesis follows the journal format structure in accordance with the Presentation of
Thesis Policy at the Alliance Manchester Business School. This allows chapters to be
incorporated into a format suitable for submission and publication in peer-reviewed
academic journals. Therefore, this thesis is structured into three essays containing
original research, in Chapters 2, 3, and 4. The chapters are self-contained, each having a
separate literature review, answering unique and original questions, and employing a
distinctive analysis and its own dataset. The equations, footnotes, tables, and figures are
independent, and the numbering starts at the beginning of each chapter. Page numbers,
titles, and subtitles follow a sequential order throughout the thesis.
The thesis continues as follows. Chapter 2 examines the impact of ESO on the
behaviour of organised labour. Chapter 3 investigates labour unions’ attitude towards
CSR spending. Chapter 4 explores how financial analysts, as information intermediaries,
are affected by heightened strike risk introduced by labour unions. Chapter 5 concludes,
with limitations and suggested directions for future research.
17
In Chapters 2, 3 and 4, I use “we” rather than “I” as these chapters are in the form of
working papers co-authored with my supervisors.
18
References
Agrawal, A.K. (2012). Corporate Governance Objectives of Labor Union Shareholders:
Evidence from Proxy Voting. Review of Financial Studies, 25(1), pp.187–226.
Ashenfelter, O. and Johnson, G.E. (1969). Bargaining Theory, Trade Unions, and
Industrial Strike Activity. American Economic Review, 59(1), pp.35–49.
Asquith, P., Mikhail, M.B. and Au, A.S. (2005). Information Content of Equity Analyst
Reports. Journal of Financial Economics, 75(2), pp.245–282.
Barro, R.J. (2001). Human Capital and Growth. American Economic Review, 91(2),
pp.12–17.
Barron, O.E., Byard, D. and Yu, Y. (2008). Earnings Surprises that Motivate Analysts
to Reduce Average Forecast Error. The Accounting Review, 83(2), pp.303–325.
Beatty, A. (1995). The Cash Flow and Informational Effects of Employee Stock
Ownership Plans. Journal of Financial Economics, 38(2), pp.211–240.
Becker, B.E. and Olson, C.A. (1986). The Impact of Strikes on Shareholder Equity.
Industrial and Labor Relations Review, 39(3), pp.425–438.
Bhagat, S., Brickley, J.A. and Lease, R.C. (1985). Incentive Effects of Stock Purchase
Plans. Journal of Financial Economics, 14, pp.195–215.
Blasi, J., Kroumova, M., Kruse, D.L. and Sesil, J. (2002). Broad-Based Employee Stock
Options — A Union-Nonunion Comparison. Advances in Industrial & Labor Relations.,
11, pp.69–94.
Bova, F. (2013). Labor Unions and Management’s Incentive to Signal a Negative
Outlook. Contemporary Accounting Research, 30(1), pp.14–41.
Bradley, D., Kim, I. and Tian, X. (2017). Do Unions Affect Innovation? Management
Science, 63(7), pp.2251–2271.
Bradshaw, M., Ertimur, Y. and O’Brien, P. (2017). Financial Analysts and Their
Contribution to Well-Functioning Capital Markets. Foundations and Trends in
Accounting, 11(3), pp.119–191.
Call, A.C., Kedia, S. and Rajgopal, S. (2016). Rank and File Employees and the
Discovery of Misreporting: The Role of Stock Options. Journal of Accounting and
Economics, 62(2–3), pp.277–300.
Campbell, J.Y. (1996). Understanding Risk and Return. Journal of Political Economy,
104(2), pp.298–345.
Campello, M., Gao, J., Qiu, J. and Zhang, Y. (2018). Bankruptcy and the Cost of
Organized Labor: Evidence from Union Elections. Review of Financial Studie, 31(3),
pp.980–1013.
Carrel, P. (2010). The Handbook of Risk Management Implementing a Post-Crisis
Corporate Culture. Chichester: John Wiley & Sons.
Chen, H.J., Kacperczyk, M. and Ortiz-Molina, H. (2011). Labor Unions, Operating
Flexibility, and the Cost of Equity. Journal of Financial and Quantitative Analysis,
46(01), pp.25–58.
Chung, R., Lee, B.B.H., Lee, W.J. and Sohn, B.C. (2016). Do Managers Withhold Good
19
News from Labor Unions? Management Science, 62(1), pp.46–68.
Clark, K.B. (1984). Unionization and Firm Performance: The Impact on Profits, Growth,
and Productivity. American Economic Review, 74(5), pp.893–919.
Coulman, L. (2019). Why is Employee Activism on the Rise ? In Forbes. Available
from: https://www.forbes.com/sites/laurencoulman/2019/05/30/why-is-employee-
activism-on-the-rise/#3bfa003474b1. [Accessed: 1 August 2019]
Cramton, P., Gunderson, M. and Tracy, J. (1999). The Effect of Collective Bargaining
Legislation on Strikes and Wages. Review of Economics and Statistics, 81(3), pp.475–
487.
DiNardo, J. and Lee, D. (2004). Economic Impacts of New Unionization on Private
Sector Employers: 1984–2001. Quarterly Journal of Economics, 119(4), pp.1383–1441.
Edgecliffe-Johnson, A. (2019). Activist Employees Pose New Labour Relations Threat
to Bosses. In Financial Times. Available from: https://www.ft.com/content/c1167d4a-
9cb5-11e9-b8ce-8b459ed04726. [Accessed: 1 August 2019]
Fang, H., Nofsinger, J.R. and Quan, J. (2015). The Effects of Employee Stock Option
Plans on Operating Performance in Chinese Firms. Journal of Banking & Finance, 54,
pp.141–159.
Freeman, R.E., Harrison, J.S. and Wicks, A.C. (2007). Managing for Stakeholders:
Survival, Reputation, and Success. Yale University Press, New Haven
Hamm, S.J.W., Jung, B. and Lee, W.J. (2018). Labor Unions and Income Smoothing.
Contemporary Accounting Research, 35(3), pp.1201–1228.
Hilary, G. (2006). Organized Labor and Information Asymmetry In The Financial
Markets. Review of Accounting Studies, 11(4), pp.525–548.
Hochberg, Y. V. and Lindsey, L. (2010). Incentives, Targeting, and Firm Performance:
An Analysis of Non-Executive Stock Options. Review of Financial Studies, 23(11),
pp.4148–4186.
Huang, Q., Jiang, F., Lie, E. and Que, T. (2017). The Effect of Labor Unions on CEO
Compensation. Journal of Financial and Quantitative Analysis, 52(2), pp.553–582.
Jennings, J. (2019). The Role of Sell-Side Analysts after Accusations of Managerial
Misconduct. The Accounting Review, 94(1), pp.183–203.
Kim, E.H. and Ouimet, P. (2014). Broad-Based Employee Stock Ownership: Motives
and Outcomes. Journal of Finance, 69(3), pp.1273–1319.
Klasa, S., Maxwell, W.F. and Ortiz-Molina, H. (2009). The Strategic Use of Corporate
Cash Holdings in Collective Bargaining with Labor Unions. Journal of Financial
Economics, 92(3), pp.421–442.
Lee, D.S. and Mas, A. (2012). Long-Run Impacts of Unions on Firms: New Evidence
from Financial Markets, 1961-1999. Quarterly Journal of Economics, 127(1), pp.333–
378.
Liberty, S.E. and Zimmerman, J.L. (1986). Labor Union Contract Negotiations and
Accounting Choices. The Accounting Review, 61(4), pp.692–712.
Loh, R.K. and Stulz, R.M. (2018). Is Sell-Side Research More Valuable in Bad Times?
Journal of Finance, 73(3), pp.959–1013.
20
Matsa, D.A. (2010). Capital Structure as a Strategic Variable: Evidence from Collective
Bargaining. Journal of Finance, 65(3), pp.1197–1232.
McCracken, M., McIvor, R., Treacy, R. and Wall, T. (2017). Human Capital Theory:
Assessing the Evidence for the Value and Importance of People to Organisational
Success. Chartered Institute of Personnel and Development (CIPD). Available from:
https://www.cipd.co.uk/Images/human-capital-theory-assessing-the-evidence_tcm18-
22292.pdf. [Accessed: 1 August 2019]
Pirson, M. and Turnbull, S. (2011). Corporate Governance, Risk Management, and the
Financial Crisis: An Information Processing View. Corporate Governance: An
International Review, 19(5), pp.459–470.
Rosen, C. and Quarrey, M. (1987). How Well is Employee Ownership Working?
Harvard Business Review, 65(5), pp.126–129.
Siu, W., Leung, C., Li, O.Z. and Rui, O.M. (2009). Labor Union and Accounting
Conservatism. SSRN eLibrary, pp.1–54.
Zingales, L. (2000). In Search of New Foundations. Journal of Finance, 55(4),
pp.1623–1653.
21
Chapter 2
Does Employee Ownership Reduce Strike Risk?
Evidence from Union Elections
Abstract
This paper investigates the impact of employee stock options (ESO) on labour unions’
propensity to initiate strikes. By exploiting the unique setting of union elections in U.S.
firms, we employ a triple-differences specification and find that firms offering higher
levels of equity incentives to their employees are exposed to a significantly lower
likelihood of union strikes. We interpret this moderating effect of ESO incentives on the
post-unionisation strike risk as evidence consistent with ESO realigning the interests of
organised labour with those of their employers. Consistent with this conjecture, further
analyses using a regression discontinuity design (RDD) present causal evidence that
firms strategically grant more stock option incentives to employees in response to the
unionisation of the labour force. The increase in option incentives is more pronounced
among firms holding union elections in non-right-to-work states, where labour unions
enjoy stronger bargaining power, and firms in low-skill industries, where the strike risk
is perceived to be higher. Our findings have policy implications not only for the U.S.
context but also for other jurisdictions with more powerful labour union movements,
thus a greater need for risk management against strikes.
22
2.1 Introduction
We examine the impact of employee stock options (ESO) on labour strike risk. ESO
have proliferated in the last two decades due to their perceived status as an important
mechanism in motivating and retaining employees, which ultimately enhances firm
performance and value (Core and Guay 2001; Chang et al. 2015; Babenko and
Tserlukevich 2016). Around 36 percent of all employees in publicly listed firms hold
stocks or stock options (GSS 2014). Moreover, there has been an eightfold surge in the
number of employees holding stock options, from 1 million in the 1990s to an estimated
8.5 million in 2014, which accounts for 7.2 percent of the total labour force in the U.S.
private sector (NCEO 2017). Despite the prevalence of ESO amongst publicly listed
firms, to date, little is known about the role of ESO in aligning the interests of
employees and their firms in the presence of labour unions. In this paper, we empirically
investigate how employee equity-based incentives, specifically incentives created by
ESO,1 affect the behaviour of organised labour with respect to a highly disruptive
activity for a firm, that is, strikes. We conjecture that, when employees are awarded
equity-based incentives, labour unions are likely to behave in a more cooperative
manner as a result of the improved alignment of the interests of organised labour and
the employers.
The economic impact of a labour strike is detrimental, imposing immediate and
substantial costs on the employer (Schmidt and Berri 2004). Becker and Olson (1986)
find that a strike involving more than 1,000 workers destroys 4.1 percent of shareholder
1 ESO differ from other employee ownership-based incentive schemes, such as employee stock ownership
plans (ESOP), in that ESO offer strong medium-term incentives, whereas other employee ownership
schemes tend to work like, and share more features with, pension schemes. In particular, a fixed
percentage of salary is contributed to the typical employee ownership scheme and employees cannot
access the funds until retirement or on leaving the company.
23
wealth, equivalent to around 80 million in 1980 dollars. Recent anecdotal evidence2
appears to suggest that labour strikes have become even more costly in recent times.
Apart from the direct impact on the firm, the adverse effects of labour strikes also
disseminate across industries, along the supply chain (McHugh 1991; DiNardo and
Hallock 2002). In industries of great strategic importance, such as manufacturing and
utilities (Chen et al. 2011), a large-scale strike could have a contagious and material
effect on the productivity of the U.S. economy as a whole. According to the Bureau of
Labor Statistics (2017), a total of 1.54 million days were left idle during 2016 as a
consequence of 15 mass strikes involving over 99,000 workers in the U.S., causing
severe uncertainty and disruption to the businesses and society. While it is difficult to
quantify the social and economic costs of labour strikes, the damage caused by them is
likely to have been exacerbated under economic downturns and political turbulence of
the past decade.
Under the monopoly model, labour unions use their collective bargaining power to
extract economic rents, demanding wage increases and better welfare systems for their
union members, normally low-skilled workers in labour-intensive industries such as
manufacturing (Ashenfelter and Johnson 1969; Freeman and Medoff 1979; Liberty and
Zimmerman 1986; Vedder and Gallaway 2002). Labour unions’ efforts to pursue their
own agendas often lead to suboptimal corporate decisions that destroy shareholders’
wealth (Freeman and Medoff 1979; Chen et al. 2011; Lee and Mas 2012). In order to
improve their bargaining position and impose more pressure on employers to
compromise during contract negotiations, labour unions often engage in a range of
2 In 2008, a 58-day strike by 27,000 machinists at Boeing, the largest aircraft manufacturer in the world,
caused $100 million of losses per day in deferred revenue, and $2 billion in lost profits. The share price
also plummeted by 56 percent to a five-year low during the strike period (Reuters 2008). More recently,
in 2016, Verizon, the largest telecommunication provider in the U.S., suffered a major strike involving
more than 40,000 employees. It is estimated that the seven-week strike cost Verizon $343 million in
revenue (Wall Street Journal 2016).
24
disruptive and value-destroying collective bargaining activities, such as strikes, and are
proven to influence a wide spectrum of corporate decisions, ranging from executive
remuneration to corporate innovation (Atanassov and Kim 2009; Klasa et al. 2009;
Matsa 2010; Chyz et al. 2013; Chino 2016; Bradley et al. 2017; Huang et al. 2017). The
bargaining power, hence influence, of labour unions, depends on the size of their
membership base. Despite the decline in U.S. union membership rates over the last
several decades (DiNardo and Lee 2004),3 labour unions currently represent over eight
million private-sector workers; 33 percent of the largest industrial firms have a
unionised workforce (Campello et al. 2018).
To mitigate the strike risk and improve their bargaining position against labour unions,
firms strategically adjust their capital structure and financial position, for instance by
reducing cash holdings (Klasa et al. 2009) or increasing leverage (Bronars and Deere
1991; Matsa 2010), to shelter financial resources from being targeted by labour unions.
Meanwhile, managers tend to engage in impression management, by signalling a
negative outlook (Bova 2013) or withholding good news (Chung et al. 2016) during
labour contract negotiations. In light of the growing presence of equity-based
compensation in non-tech industries where the labour force is primarily unionised
(Kroumova and Sesil 2006; EY 2014; NCEO 2017), we argue that employee ownership
incentives can be an effective instrument in reducing a union’s motivation to initiate a
strike by realigning the interests of organised labour with those of shareholders. Prior
empirical work explores the effect of employee ownership on firm performance, and
documents largely positive effects (Chang 1990; Ittner et al. 2003; Ikäheimo et al. 2004;
3 Many commentators expect this decline to reverse in the future given (1) the current administration’s
attempts to revitalise the U.S. manufacturing sector (Reuters 2017), which has traditionally been fertile
ground for unions’ recruitment of members and (2) persistent calls for ‘gig’ economy workers to unionise
in order to get basic labour protections currently not afforded to them (Lobel 2017).
25
Hochberg and Lindsey 2010; Fang et al. 2015; O’Boyle et al. 2016). In addition,
employee ownership affects various corporate issues, such as investment (Bens et al.
2002), corporate governance (Faleye et al. 2006; Bova et al. 2015b), innovation (Chang
et al. 2015), earnings quality (Jiraporn 2007; Call et al. 2016) and corporate disclosure
(Bova et al. 2015a).
We study the unique setting of union elections in U.S. listed firms from 2004-2011.
Although labour unions are arguably more prominent in other parts of the world, e.g.,
European countries (Lipset and Katchanovski 2001), we have chosen the U.S. context
for two reasons. First, the availability and granularity of the union election data allow us
to apply a quasi-experimental identification strategy that helps us draw strong causal
inferences. Second, the focus on the U.S. should work against us finding a significant
relation between ESO and strike risk if the labour union movement has, on average,
limited power and influence given the declining membership rates in that country. Thus,
findings consistent with a significant ESO effect should be viewed as evidence of the
strength of this result. Arguably, one should expect to find even more pronounced
effects when studying this relation in countries with an even more widespread presence
of labour unions.
Based on a propensity-score-matched (PSM) sample, we use a triple-differences
specification to test whether firms with a higher proportion of ESO incentives
experiencing a unionisation event see partial mitigation of the union’s impact on (i.e.,
raising of) the strike risk. Our results show that firms with higher levels of stock option
incentives held by rank-and-file employees have significantly lower post-unionisation
strike risk than their low-ESO-incentive counterparts. We interpret this finding as
evidence consistent with the moderating effect of ESO incentives on a union’s
engagement in strike activities, resulting from the ESO-driven interest alignment
26
between unionised employees and the firm, which discourages the union from initiating
this value-damaging bargaining tactic.
Next, we test whether firms strategically grant more ESO incentives post-unionisation
in response to the increased labour power. Prior literature suggests that firms make
strategic decisions to improve their bargaining position against labour unions (Bronars
and Deere 1991; Klasa et al. 2009; Matsa 2010; Bova 2013; Chung et al. 2016). Given
the interest realignment role of ESO in mitigating strike risk, we expect firms to use
ESO incentives as a strategic device to undermine union power and mitigate strike risk
following unionisation. Based on a regression discontinuity design (RDD), comparing
the ‘marginal winners’ and ‘marginal losers’ in union elections, we document robust
and causal evidence that labour unionisation leads to a significant increase in ESO
incentives granted per employee, which is consistent with our interest realignment
conjecture. Additional subsample analysis exploiting exogenous state-level variation in
union power suggests that such a reaction is more pronounced in firms whose
establishments (branches) vote in favour of unionisations (i.e., vote to be unionised) in
states without right-to-work (RTW) legislation, where union power is stronger. A
further test shows that the adjustment in ESO incentives is greater for firms in low-skill
industries, where unions are typically stronger, thus strike risk is expected to be higher.
Overall, this evidence is consistent with the increase in ESO incentives being a strategic
firm response to unionisation, aimed at managing the increased strike risk.
Our study makes the following contributions to the extant literature. First, we present
novel empirical evidence on the effect of ESO incentives on the behaviour of a key firm
stakeholder, organised labour. In particular, we show that, when employees receive
significant equity-based incentives, the impact of unionisation on strike risk is partially
mitigated. Second, our paper contributes to the existing literature on firms’ strategic
27
decision to improve their bargaining position against labour unions (Bronars and Deere
1991; Klasa et al. 2009; Matsa 2010; Chung et al. 2016; He et al. 2016a; Chino 2016)
by identifying a new tool, ESO. Third, unlike ESO in the high-tech sector, which are
granted exclusively as a way of motivating employees and retaining key talent, we
provide evidence that, in the context of unionisation, ‘old economy’ firms could also
use ESO as a strategic tool to realign the interests of organised labour and shareholders,
thus reducing the potential strike risk. Finally, our study has important implications for
accounting standard setters and policymakers. In spite of the benefits of employee
ownership we report, in terms of improving the interest alignment between employees
and shareholders, the current regulatory framework relating to ESO (FAS 123R) creates
hurdles for the expansion of employee ownership schemes. We thus highlight the need
for more policy support to promote employee ownership within labour-intensive
industrial firms, which are key to the revitalisation of the U.S. economy.
The remainder of the paper is organised as follows. Section 2 reviews the extant
literature on labour economics and employee ownership, which is followed by the
development of our research hypotheses. Section 3 describes the data collection and
sampling processes, as well as our empirical design. Section 4 presents our main
empirical results. Section 5 summarises the empirical findings and contributions of our
study.
2.2 Literature Review and Hypothesis Development
2.2.1 Literature Review
2.2.1.1 Labour Unions and Strike Risk
Labour unions are established by workers for the purpose of pursuing their collective
interests and welfare (Clark 1984; Addison and Hirsch 1989). Union representation is
28
typically more prevalent in labour-intensive manufacturing and transportation sectors
and least popular in professional service industries such as accounting and IT (Chen et
al. 2011).
The existing literature offers two competing theories on labour unions: the monopoly
model and the collective-voice model. Several labour economists (Freeman and Medoff
1979; Lewis 1983; DiNardo et al. 1997; Singh and Agarwal 2002; Klasa et al. 2009)
have reached a consensus that labour unions have a monopolistic nature, with a
persistent track record of demanding higher wages from management and initiating
strikes, despite the decline in union membership over the last few decades in the U.S.
(DiNardo and Lee 2004). While it is certainly true that a strike is the most powerful
bargaining tool of labour unions, a union’s decision to strike is a complex one,
determined by a number of factors (Ashenfelter and Johnson 1969; Tracy 1986;
Cramton and Tracy 1994). In theory, a strike is only initiated when a union perceives
the expected benefits of the strike to outweigh the costs they expect to bear (Reder and
Neumann 1980). Prior literature has established that strike probability is linked to firm
profitability, capital structure and labour market conditions in the geographic region or
industry in question (Liberty and Zimmerman 1986; Tracy 1986; Klasa et al. 2009;
Myers and Saretto 2016).
However, another key function of a labour union is to serve as a channel through which
employee voice is directly communicated to the management and shareholders
(Freeman and Medoff 1979; Freeman and McVea 2001; Fauver and Fuerst 2006).
Unlike the monopoly model, which predicts a negative effect of labour unions, the
collective voice theory argues that credible information and valuable advice from the
employees can be highly conducive to higher productivity and a healthy relationship
29
between the managers and employees, ultimately leading to firm success (Freeman and
Medoff 1979).
Empirically, numerous studies on labour economics have investigated the performance
effect of labour unions, establishing that unions’ impact on businesses is largely adverse
(Clark 1984; Ruback and Zimmerman 1984; Chen et al. 2011; Lee and Mas 2012).
Specifically, Clark (1984) and Rubak and Zimmerman (1984) document significantly
lower profits and market returns in unionised than non-unionised firms. Addison and
Hirsch (1989) introduce the ‘union tax theory’, concluding that unions charge firms an
implicit ‘tax’ on firm profits by collectively bargaining for higher wages with no
guarantee of improvements in effort and productivity (Vedder and Gallaway 2002;
Banning and Chiles 2007). Although labour unions can be beneficial to employees and
even individual companies, the consensus in the literature appears to be that the
negative effects due to increased wages and disruption seem to outweigh the benefits of
voice communication and productivity that labour unions facilitate.
The negative effect of labour unions also extends to the operation of the firm. Chen et al.
(2011) show that the union presence constrains the operational flexibility of the business,
resulting in a higher cost of equity to compensate for the higher risk undertaken by the
investors. A more recent paper by Bradley et al. (2017) illustrates that unionisation has a
harmful effect on innovation. Interestingly, DiNardo and Lee (2004) find little evidence
of a unionisation effect on employment, production or business survival in the U.S.
private sector. Still, given the overwhelmingly negative evidence, labour unions are
perceived by shareholders and management as a threat to firm value and long-term
prosperity (Bronars and Deere 1991).
In response to unions’ bargaining power and their potentially negative impact on
businesses, firms take strategic decisions on various fronts. Firstly, firms strategically
30
adjust their financial position to shelter resources from being targeted by labour unions.
For example, Klasa et al. (2009) find that firms in unionised industries hold less cash.
Other studies show that firms increase leverage to improve their bargaining position
against labour unions (Bronars and Deere 1991; Matsa 2010). Secondly, managers
engage in impression management, by signalling to their employees that the business
has less positive prospects than is really the case (Bova 2013; Chung et al. 2016). Bova
(2013) documents that unionised firms are more likely to marginally miss analysts’
earnings forecasts, while Chung et al. (2016) reveal that managers tend to withhold
positive news during labour negotiations to secure more favourable bargaining positions.
In addition, He et al. (2016) suggest that firms strategically constrain corporate payouts
to maintain operating flexibility and mitigate cash flow risk due to union presence. The
strategic firm response to union presence also extends to other stakeholders, such as
competitors. By focusing on non-unionised firms operating in unionised industries,
Aobdia and Cheng (2018) show that these firms strategically increase disclosure when
their unionised competitors are engaging in labour contract renegotiations, in order to
undermine their unionised rivals.
In the meantime, the collective power of labour unions encourages employees to freely
express criticism and dissatisfaction with managerial decisions, as well as executive
compensation packages (Freeman and Medoff 1979). Consistent with agency theory,
labour unions are widely recognised as an additional governance mechanism that helps
rein in managerial power and deters managerial opportunistic behaviour, such as rent
extraction and short-termism (DiNardo et al. 1997; Singh and Agarwal 2002; Banning
and Chiles 2007; Huang et al. 2017). For example, Huang et al. (2017) report
significantly lower executive compensation in the presence of a labour union.
31
Meanwhile, Chyz et al. (2013) document that increased monitoring by a labour union
also significantly constrains a firm’s ability to engage in tax avoidance activities.
The recent adoption of the RDD methodology by accounting, finance and management
scholars has revitalised efforts to identify the causal effect of labour unionisation on
various corporate decisions (He et al. 2016; Tian and Wang 2016; Bradley et al. 2017;
Campello et al. 2018). Specifically, Tian and Wang (2016) report that a labour union
can serve as a takeover defence, considerably lowering the probability of receiving a bid.
Campello et al. (2018) and He et al. (2016a) find negative impacts of unionisation on
bond values and the dividend payout ratio, respectively. Labour unions can also
influence financial reporting decisions by inducing accounting conservatism among
firms (Siu et al. 2009). Paired with the empirical findings of lower cash holdings (Klasa
et al. 2009) and a less positive outlook (Bova 2013; Chung et al. 2016), these results
seem to collectively suggest that firms make strategic decisions to gain more favourable
bargaining positions against organised labour.
2.2.1.2 Employee Ownership
Another key strand of the literature pertaining to our study is that on the development
and implications of employee ownership (EO), which emerged and began to proliferate
in the 1990s in the U.S. (Blair et al. 2000). According to Rosen et al. (2005), there are
various categories of EO plans currently offered by companies: ESOP, ESO plans,
employee stock purchase plans (ESPP), restricted stock plans, and Section 401(k) plans.
ESO appear to be the most popular equity-based incentives among U.S. listed firms
(Rosen et al. 2005; NCEO 2017). EO plans are offered to employees for various reasons.
The primary rationale is to motivate employees and align their interests with those of
the shareholders (Bhagat et al. 1985; Blasi et al. 2002; Kim and Ouimet 2014; Call et al.
2016). When participating in EO schemes, employees tend to have more positive work
32
attitudes and higher productivity, which ultimately contribute to better performance
(Bhagat et al. 1985; Blasi et al. 2002; Oyer and Schaefer 2005). Moreover, as a result of
interest alignment, employees are more proactively involved in monitoring managerial
behaviour and enhancing corporate governance (Jiraporn 2007; Kim and Ouimet 2014).
Another well-documented reason for firms to award equity-based compensation to rank-
and-file employees is to substitute for cash-based wage increases (Core and Guay 2001;
Oyer 2004; Oyer and Schaefer 2006; Kim and Ouimet 2014). However, Hayes et al.
(2012) show that the change in the accounting treatment for equity-based compensation
following the implementation of FAS 123R in 2005 has made this option less
financially appealing for firms.
Several studies also suggest that EO plans act as an anti-takeover device, enabling firms
to secure higher price premia from bidders (Gordon and Pound 1990; Dhillon and
Ramírez 1994; Beatty 1995; Blair et al. 2000; Rauh 2006; Cramton et al. 2008;
Babenko and Tserlukevich 2016). Others argue that EO plans are adopted to take
advantage of tax benefits (Gale and Potter 2002; Babenko and Tserlukevich 2009; Bova
et al. 2015a). Finally, EO plans are considered a desirable source of equity financing,
given the lower information asymmetry between managers and employees than between
managers and external investors (Fama and French 2005; Garmaise 2008; Babenko et al.
2011; Babenko and Sen 2016).
Empirical studies on EO mainly focus on its economic impact on the firm and
implications for the employees and management. Since the adoption of EO has a direct
and potentially significant impact on the ownership structure of the company, and
inevitably dilutes the ownership of existing shareholders, a large body of literature
investigates the investors’ perception of EO. Prior work shows that the market reaction
33
to its adoption is mixed, and largely dependent on the motivation behind it (Bhagat et al.
1985; Chang 1990; Chang and Mayers 1992; Beatty 1995).
Meanwhile, the performance implications of EO have also attracted much academic
attention (Rosen and Quarrey 1987; Rosen 1990; Jones and Kato 1995; Blasi et al. 1996;
Iqbal and Hamid 2000; Park et al. 2004). Rosen and Quarrey (1987) and Park et al.
(2004) report higher business survival rates in firms with EO schemes. Positive effects
on productivity and operating performance are reported by Beatty (1995) and Blasi et al.
(2002). Notably, Jones and Kato (1995) reveal that it takes three years on average to
realise the productivity effect, based on a sample of 585 firms with ESOP. Crucially,
firms tend to grow much faster as a result of the synergy between EO and employee
participation in the corporate decision-making process (Rosen and Quarrey 1987; Rosen
1990; Blasi et al. 2016). As for ESO specifically, Hochberg and Lindsey (2010) and
Fang et al. (2015) document a positive effect on firm performance, using U.S. and
Chinese samples, respectively.
Without doubt, when employees become shareholders, the traditional labour-
management relation fundamentally transforms, with labour exerting a stronger
influence on managerial decisions (Dhillon and Ramírez 1994; Jiraporn 2007; Zhang
2011; Bova et al. 2015a; Bova et al. 2015b; Chang et al. 2015). The intense monitoring
by employee-shareholders leads to less earnings management (Jiraporn 2007), lower
executive compensation (Zhang 2011) and more voluntary disclosure (Bova et al.,
2015a). Meanwhile, EO suppresses managerial risk-taking behaviour (Bova et al.,
2015b). Overall, this evidence is consistent with EO playing a positive role in corporate
governance, as a result of the aligned interests of employees and shareholders.
A notable difference between ESO and other EO schemes relates to their design and
horizon. On the one hand, ESO offer strong mid-term incentives, at least during the
34
vesting period, which is typically 2-5 years (Core and Guay 2001; Babenko and
Tserlukevich 2009; Gopalan et al. 2014). On the other hand, other EO plans such as
ESOP can be set up in a similar fashion to pension schemes, and not accessible until an
employee’s retirement or the termination of employment (Yates 2006; Kim and Ouimet
2014). Earlier work suggests that ESO are only commonplace in high-tech firms, and
thus focuses predominantly on the ‘new economy’ sectors (Ittner et al. 2003). However,
recent evidence shows that ESO have become increasingly popular in non-tech
industries as well (Kroumova and Sesil 2006; EY 2014; NCEO 2017). Given the
heterogeneity in labour forces across industries, it is interesting to explore, and more
importantly understand, the potential implications of ESO in non-tech industries, which
is the primary focus of our study.
2.2.2 Hypothesis Development
Both theoretical and empirical work suggests that labour unions can be detrimental to
firm value as they use their collective bargaining power to extract economic rents from
firms at the expense of shareholders (Clark 1984; Ruback and Zimmerman 1984;
Liberty and Zimmerman 1986; Tracy 1986; Addison and Hirsch 1989; Cramton and
Tracy 1994; DiNardo and Lee 2004). A union derives its bargaining power from its
ability to initiate strikes, which could be disruptive and value-destroying for the firm
(Ashenfelter and Johnson 1969; Cramton and Tracy 1994). Thus, having a labour union
inevitably increases the strike risk of the firm, particularly during contract negotiations.
In response to the union’s power and the associated strike risk, prior studies find that
firms strategically adjust their capital structure to shelter rents from organised labour
and improve their bargaining position (Bronars and Deere 1991; Klasa et al. 2009;
Matsa 2010). As a result, the perceived potential benefit of engaging in a strike falls
35
dramatically, making striking a less attractive option for the union (Myers and Saretto
2016).
An alternative to reducing the perceived benefits of a strike is for firms to make striking
more costly for the employees. We argue that employee equity-based incentives reduce
a union’s motivation to strike by making it more costly for union members. When
employees’ wealth is sensitive to the stock price, for example through stock option
incentives, employees would also suffer, at least partially, from any financial damage
caused by a strike. Moreover, previous studies suggest that employee ownership tends
to have a positive effect on employees’ work attitudes, productivity and job satisfaction
(Klein 1987; Pierce et al. 1991; Kruse 1996; Oyer and Schaefer 2005; Blasi et al. 2016),
making it more difficult for unions to garner employee support for strike action.
Therefore, we argue that ESO incentives will change the behaviour of labour unions as
a result of improved interest alignment between organised labour and the firm (Wheeler
2002; Yates 2006). When employees are made minority shareholders, labour unions are
expected to behave more cooperatively and responsibly when negotiating with firms,
avoiding the use of strikes whenever possible.
Nevertheless, we concur that ESO incentives might not necessarily have any influence
on a union’s decision to strike. First, rank-and-file employees, who are typically risk-
averse, would normally prefer a wage increase today to an unguaranteed return from
ESO in the future (Ittner et al. 2003; Babenko and Sen 2014; Chang et al. 2015; Bova et
al. 2015b). Therefore, labour union members, despite holding a small fraction of the
equity, might be indifferent to a decline in shareholders’ wealth resulting from a strike if
there was a good chance the latter would result in higher wages. Second, both Faleye et
al. (2006) and Agrawal (2012) document empirical evidence suggesting that organised
36
labour uses the stronger voice gained from employee ownership to serve its own
interests at the expense of shareholder interests.
On balance, however, we predict that ESO incentives significantly improve the interest
alignment between organised labour and the firm, forging a less aggressive and more
cooperative relationship between the two parties.
Therefore, we propose the following as our main hypothesis:
Hypothesis 1 (H1): ESO incentives have a moderating effect on a union’s propensity to
strike.
Issuing stock options to rank-and-file employees is a decision made by the managers of
the firm.4 Knowing that labour unions could be detrimental to firm value (Clark 1984;
Ruback and Zimmerman 1984; Lee and Mas 2012), firms will have an incentive to align
their interests with those of their employees through equity ownership. Specifically,
once unionised, a firm will be exposed to higher strike risk in the foreseeable future,
relative to non-unionised firms. Assuming ESO incentives indeed create strong interest
alignment, and thus influence the behaviour of labour unions (H1), firms may
strategically grant more ESO incentives after unionisation, to reduce the probability of
strikes.
Still, we acknowledge that firms might sometimes be reluctant to grant ESO in response
to labour unionisation. First, there could be concerns about the potential ownership
dilution imposed on existing shareholders when more stock options are issued. If firms
will have to issue new equity in order for employees to exercise their options, the
4 We assume that the management represents the interests of the shareholders; hence, we do not
differentiate between managers and shareholders in this study. However, we do acknowledge that there
could be managerial incentives behind the granting of ESO, such as managerial entrenchment
(Chaplinsky and Niehaus 1994).
37
current shareholders will eventually own less of a stake in the firm, and the management
and shareholders will have less control over the firm, assuming there are voting rights
attached to the equity granted to employees (Bens et al. 2003). Second, there are
accounting implications5 for the income statement, as firms are required to expense all
stock options granted in a fiscal year, which will reduce the reported profitability
(FASB 2004; Choudhary et al. 2009; Hayes et al. 2012).
Yet, we argue that the benefits of granting more ESO incentives should outweigh the
costs for firms facing a significantly higher strike risk.
Hence, we propose the following as our second hypothesis:
Hypothesis 2 (H2): Labour unionisation leads to more ESO incentives being granted.
2.3 Data and Research Design
2.3.1 Data
Our study utilises data from multiple sources: (1) The National Labor Relations Board
(NLRB) election database for union election results; (2) Standard and Poor’s
Execucomp and CRSP/Compustat merged databases for all data points used to calculate
key ESO variables; (3) the U.S. Bureau of Labor Statistics (BLS) and Federal Mediation
and Conciliation Service (FMCS) for strike information; (4) other relevant financial
data and firm information are also accessed from the CRSP/Compustat merged database.
After cleaning the data and merging various databases, we obtain a base sample of 324
unique union election events, from 2004 to 2011, based on which we form two separate
samples, containing 1,368 (PSM sample) and 254 (RDD sample) observations, to test
5 A significant change in accounting rules regarding equity-based compensation (FAS 123R) became
effective after 15 June 2005.
38
our hypotheses H1 and H2, respectively.6 The following subsections explain the sample
construction in detail.
2.3.1.1 Union Election Data
The NLRB union election database contains detailed information on each representation
election from 1980 to 2011, in which eligible employees voted to determine whether to
certify a union as a collective bargaining representative.7 Specifically, we extract the
following key election information: number of valid votes, number of votes for
unionisation, number of votes against unionisation, election outcome and election date.
For identification purposes, we also gather useful information covering the election case
ID, employer name, location, union involved and industry code. In addition, we
construct the Vote Share variable, defined as the ratio of the number of votes for
unionisation to the number of valid votes. Figure 1 illustrates the number of union
elections held in each year from 1980 to 2011, showing a significant decline in the last
three decades, consistent with that reported by prior literature (DiNardo and Lee 2004;
Campello et al. 2018).
***Insert Figure 1 here***
6 There is a trade-off between precision and generalisability in any research design. In this study, we opt
for precision, which allows us to establish causal inferences on the impact of ESO incentives on strike risk in the presence of labour unions. Thus, we end up with a relatively small sample size. Still, the
strength of the reported effect in this sample and in the U.S. context creates optimism over the
generalisability of our findings to larger samples and alternative (international) contexts.
7 Following DiNardo and Lee (2004) and Campello et al. (2018), the 1977-1999 union election data are
accessed from Thomas Holmes’s website (http://users.econ.umn.edu/~holmes/data/geo_spill/index.html),
whilst the 2000-2011 data are directly obtained from the NLRB website
(https://catalog.data.gov/dataset/nlrb-cats-final-r-case-data-bulk-19990101-20110930-in-xml). Following
a system upgrade, NLRB discontinued compiling the union election database in 2011. Under the new
system, without the option to access data in batches, union election results after 2011 can only be
searched online based on the unique case number assigned to each election.
39
2.3.1.2 Employee Stock Options Data
Since there is no comprehensive and dedicated database for stock options and other
equity-based compensation awarded to non-executive employees, we follow the extant
literature on non-executive stock options and collect firm-level option data for all
employees as well as senior executives, from 2004 to 2015, from Compustat and
Execucomp8 (Core and Guay 2001; Hochberg and Lindsey 2010; Babenko et al. 2011;
Chang et al. 2015; Babenko and Tserlukevich 2016). Importantly, to make sure we
capture the incentives held specifically by rank-and-file employees, we subtract the
executive incentives (Execucomp) from the option incentives granted to all employees
within the firm (Compustat). We follow Core and Guay (2002) and compute employee
option incentives, which captures the change in employees’ wealth in dollar terms that
would occur under a 1 percent change in the stock price. Since larger firms are more
likely to grant more stock option incentives to their employees in total, we follow prior
literature (Core and Guay 2001; Chang et al. 2015) and divide the ESO incentives by
the total number of employees to account for differences due to firm size. Specifically,
we construct the following two main ESO measures: Incentives Outstanding Per
Employee and Incentives Granted Per Employee. As an alternative proxy, we use the
proportion of outstanding option incentives held by non-executive employees
(Incentives_Outstanding_Pct) to measure the overall level of employee equity
incentives within the firm.9 In other words, amongst all the outstanding ESO incentives
8 As in Babenko and Tserlukevich (2016), our starting year (2004) is motivated by the availability of data.
Pre-2004, the option data are only available for less than 1 percent of the Compustat universe. Since FAS
123R, issued in 2004 and applied in 2005, more data on stock options are disclosed and more than 70
percent of the Compustat universe have such information available.
9 To ensure data quality, we compared our original ESO incentives data with Core and Guay (2001) and
found the data to be very similar. For example, the mean of Log(Incentives Outstanding) is 12.80 in our
data, and 12.57 in Core and Guay (2001). Meanwhile, the mean of Log(Incentives Outstanding Per
Employee) is 4.45 in our data, and 4.09 in their paper. Finally, our fraction of options outstanding held by
employees is 67.41%, while theirs is 66.9%. We would expect our averages to be slightly higher because
ESO have become more prevalent since the 1994-1997 period used by Core and Guay (2001).
40
held by all the employees (including executives), a higher percentage held by non-
executive (i.e., rank-and-file) employees would imply that the firm offered an overall
higher level of equity incentives to rank-and-file employees and could be considered
more employee-oriented.
While these ESO incentive proxies measure the level of stock option incentives held by
employees, to better gauge the unionisation effect on ESO (i.e., our H2) in our event-
study setting, it is more appropriate to focus on the change in ESO incentives around the
union election event. Furthermore, we argue that the change in ESO Incentives Granted
Per Employee is a cleaner measure with regard to the causal unionisation effect than the
change in ESO Incentives Outstanding Per Employee, since the latter incorporates noise
due to the number of options granted and number of options exercised in the same year,
which relate to options granted in previous years, which are beyond the firm’s current
control. Instead, assuming a firm does react to unionisation, it would be the level of
incentives in the form of contemporaneous option grants that firms would be able to
adjust directly in response to unionisation. For robustness, in addition to the ESO
incentives (i.e., delta), we use the actual number of stock options granted to rank-and-
file employees (ESO Number Granted Per Employee) as an alternative proxy for the
level of ESO grants in our H2. Crucially, Lee and Mas (2012) show that the
unionisation effect typically takes 15-18 months to fully materialise. Thus, to make sure
we fully capture the unionisation effect and account for the potential delay in firms’
adjustments in ESO grants, we measure the change in ESO grants from t-1, the year
immediately before the union election, to t+2, that is, two years following the election
event in year t.
41
Hence, we calculate our two key dependent variables for ESO grants, (1)
ΔLog(Incentives Granted Per Employee) and (2) ΔLog(Number Granted Per Employee),
within the window (t-1, t+2)10 defined below, where t is the year of a union election:
(1) Change in ESO Incentives Granted Per Employee:
∆Log(Incentives Granted Per Employee)= Log(Incentives granted per employee)(t+2)
− Log(Incentives granted per employee)(t−1)
(2) Change in ESO Number Granted Per Employee:
∆Log(Number Granted Per Employee)
= Log(Number granted per employee)(t+2)
− Log(Number granted per employee)(t−1)
2.3.1.3 Labour Strike Data
We manually collect the strike data from the BLS and FMCS for the 324 unique
election events in our sample. Specifically, because a strike is an extreme bargaining
incident that occurs only occasionally in a small number of firms, we collect the strike
information within the window of (t-4, t+4), that is, from four years before to four years
after the union election year. For each event-year observation, we construct the
following two key variables: (1) Strike Dummy, which is equal to one if the firm
experiences a strike in the year, and zero otherwise; (2) Strike Risk, which is an ordinal
variable that captures different levels of strike risk at the event-year level, where 1=no
strike; 2=one strike; 3=multiple strikes.
2.3.2 Sample Construction
Since the NLRB union election data are compiled by a government agency and do not
include the conventional firm identifiers such as GVKEY or CUSIP, there is no
10 We also use alternative windows of (-1,1) and (-1,3) in our main RDD analyses. The results for all
three windows are presented in Table 3 and discussed in Section 4.
42
common unique identifier across datasets for the purpose of data merging, except for the
company name. This leads to a complex and time-consuming data-matching process.
We resort to a fuzzy matching technique, similar to the matching algorithm used by
DiNardo and Lee (2004) and Lee and Mas (2012), which effectively matches our union
election data to the ESO data based on the similarity of company names.
Prior to matching the NLRB and ESO datasets, we start our data processing with the
NLRB 1980-2011 dataset, because the union election information is most crucial to our
identification strategy. Firstly, we only keep union elections under the ‘RC’ type as it
refers to elections for union certification. We eliminate election cases yet to be finalised
that might be subject to change and re-election, and only keep elections classified under
‘Closed’ status. More importantly, all the union elections are held at the establishment
level and not the firm level, meaning that there are multiple union elections at different
branches of the same company.11 Following prior literature, we retain only the first
election observation for each of the three scenarios described in the footnote and drop
the duplicate observations (DiNardo and Lee 2004; Bradley et al. 2017; Huang et al.
2017). This is because the first election result is likely to be the most exogenous and is
not subject to the influence of the results of other union elections held at different
establishments within the same company.
The resulting dataset is then matched with the ESO dataset using fuzzy matching
algorithms based on company names. After performing the fuzzy matching, we
manually review all matches, to ensure that the company name of the matched
observation for the union election is the same as that in our ESO observation. Consistent
11 Multiple elections under the same company name can arise from the following scenarios: (1) multiple
observations under the same election case ID in the same year; (2) multiple elections for the same
company but with different election case IDs in the same year; (3) multiple observations under the same
election case ID and same company name but in different years.
43
with Campello et al. (2018), we keep only the larger elections with at least 50 votes,
which are believed to have greater impact.
Our sampling procedure results in 324 unique election events12, from 2004 to 2011,
which form the base sample from which we construct our final samples to test our two
hypotheses.13 Despite the small sample, which might limit the generalisability of our
results, the union election setting allows us to draw strong causal inferences by applying
quasi-experimental strategies.
In order to test the ESO incentive effect on unions’ decision to strike (H1), we manually
collect the strike information from t-4 to t+4 for each union election event, t being the
election year. Thus, we are able to construct a panel dataset at event-year level. Since
our election sample is not balanced between the treated (unionised) and control (non-
unionised) groups, we use a PSM approach to make sure the two groups are comparable
in terms of firm characteristics prior to the election at t-1, the year before the union
election. Our PSM sample consists of 152 union election events with 76 ‘Wins’ (i.e.,
treatment group) and 76 ‘Loses’ (i.e., control group). Lastly, we utilise the GVKEY
identifier to merge the NLRB/ESO/Strike dataset with financial information and firm
characteristics from Compustat and CRSP. After matching the panel dataset containing
strike information and firm characteristics from four years before to four years after the
election year (i.e., 9-year period), we obtain our final sample of 1,368 observations for
testing our H1.
12 This is our baseline sample. Our matched sample, though small, is highly comparable to prior literature
using the union election setting (Lee and Mas 2012; Huang et al. 2017; Campello et al. 2018) in terms of
the number of observations per year. The small number of observations for the union/ESO dataset is
attributable to two factors: (1) the majority of union elections are held in private firms, therefore no ESO
or financial information is available; (2) the limited availability of the ESO data, which results in a very
short overlapping period between 2004 and 2011. Since our analyses require data for firm characteristics,
including ESO information at the year of the union election, we end up with 254 observations for which
data are available from CRSP.
13 Using the same baseline sample ensures consistency throughout our empirical analyses.
44
2.3.3 Summary Statistics
Table 1 shows the summary statistics for the union elections, ESO, strike data and firm
characteristics used in our empirical analyses. The average for the variable Vote Share is
45.1 percent in our sample, with around 36.1 percent of the sample firms voting to
approve unionisation.14 In terms of the ESO level, 73.4 percent of all outstanding stock
option incentives are held by non-executive employees.
***Insert Table 1 here***
2.3.4 Research Design
2.3.4.1 Identification Strategy
To test our hypotheses, we exploit the exogenous increase in employees’ bargaining
power based on the quasi-experimental setting of union elections in the U.S. (DiNardo
and Lee 2004; Lee and Mas 2012; Huang et al. 2017; Campello et al. 2018). As
regulated by the NLRB, labour union elections follow a simple majority rule, whereby
firms are unionised if the vote share in favour of unionisation passes 50 percent. As a
result of this clear deterministic rule, the treatment effect is unambiguous. 15 More
importantly, because of the secret-ballot election system, the parties concerned are not
expected to be able to precisely manipulate the election results. Thus, from a natural
experiment perspective, once the share of votes for unionisation reaches 50 percent, the
firm receives a permanent treatment effect of unionisation, which generates a
discontinuous increase in employees’ bargaining power.
The underpinning assumption of hypothesis H1 is that firms experiencing unionisation
are exposed to higher strike risk than those failing to unionise, because strikes are
14 The statistics are similar to those reported by previous papers using union elections (He et al. 2016a;
Qiu and Shen 2017; Campello et al. 2018).
15 For a detailed description of the NLRB unionisation process, see DiNardo and Lee (2004).
45
unilaterally initiated by labour unions as a key bargaining tactic against firms. To
investigate the role of ESO incentives in the behaviour of labour unions, we compare
the differential effect on unions’ strike propensity between high-ESO-incentive firms
and low-ESO-incentive firms using a triple-differences strategy.
However, it is possible that firms that pass unionisation (i.e., treatment group) and those
that refuse to unionise (i.e., control group) may be systematically different in firm
characteristics, which could potentially confound our analysis. To reduce sample bias
and alleviate such concerns over potential confounding effects, we use PSM to match
the treated firms (i.e., Treated=1) to control firms (i.e., Treated=0) based on their firm
characteristics at t-1.16 Specifically, we use the same set of control variables that are
included in our main model presented later to generate the propensity score as the
benchmark for our matching. Then, we choose the nearest-neighbour without
replacement, with a caliper of 0.01. This approach minimises the concern about
observable confounding factors affecting our inferences, since it gives us pairs of
treated and control firms that are indistinguishable in terms of the firm characteristics
included in our PSM analysis. Hence, any difference in strike probability at time t
between the two groups can be attributed to the significant difference in ESO incentives
at t-1. We expect the sign of the triple-differences coefficient to be negative, consistent
with the notion that ESO incentives moderate the union effect on strike probability,
supporting our conjecture that the interest-alignment effect of ESO incentives influences
union behaviour.
To test hypothesis H2, we conduct an event-study analysis to examine firm reaction to
the unionisation event using a quasi-experimental RDD. Several recent studies have
16 Firm characteristics before and after PSM are presented in Appendix 3.
46
used RDD to evaluate the causal effect of labour unionisation on corporate performance
and decisions (DiNardo and Lee 2004; Lee and Mas 2012; He et al. 2016a; Qiu 2016;
Tian and Wang 2016; Bradley et al. 2017; Campello et al. 2018). Because of the secret-
balloting system, the union elections in the vicinity of the cutoff point (50%) can be
seen as a ‘locally’ randomised treatment assignment. In other words, ‘marginal losers’
(firms that only just vote against unionisation) and ‘marginal winners’ (firms that only
just vote in favour) of union elections should not be systematically different in the
absence of the treatment effect, making the ‘marginal losers’ ideal counterfactuals for
the treated ‘marginal winners’ (DiNardo and Lee 2004; Lee 2008; Bradley et al. 2017;
Campello et al. 2018).17 The RDD helps us estimate the local average treatment effect
by comparing the outcome variable, i.e., the change in ESO Incentives Granted Per
Employee, between ‘marginal losers’ and ‘marginal winners’ in union elections.
Because of what we have discussed above, any difference in the outcome variable is
likely to be caused by the treatment effect (Lee 2008). By applying this conceptually
intuitive approach, we can draw a strong causal inference whilst minimising the
conventional endogeneity concerns associated with OLS regressions.
Under our hypothesis H2, we expect the unionisation effect to be positive, with firms
strategically granting more ESO incentives in reaction to the unionisation event, to
mitigate the increased strike risk.
2.3.4.2 Empirical Models
• Triple-Differences Regression Approach
We run the probit model (Equation 1) below to study how ex-ante ESO incentives could
affect the union’s likelihood to strike after unionisation. The dependent variable is
17 We conduct a series of validity tests to verify this crucial continuity assumption. The results are
included in Appendix 5.
47
Strike Dummy. The variable of interest is the interaction term Treatedi×Posti,tESOi,t-1,
whose coefficient β1 captures the differential treatment effect (i.e., unionisation) on the
strike risk between high-ESO-incentive and low-ESO-incentive firms. Following Klasa
et al. (2009), we also control for the change (i.e., first difference) in a vector of firm
characteristics that would affect the probability of a strike. To reduce the potential for
reverse causality, all independent variables are lagged by one year. Following H1, we
expect β1 to be negative, which would imply that ESO incentives moderate the effect of
unionisation on strike probability. As a robustness check, we run an ordered probit
model by replacing the strike dummy with the ordinal variable, Strike Risk.
Strike Dummyi,t =α+β1(Treatedi×Posti,tESOi,t-1)+β2Treatedi×Posti,t +β3Treatedi×ESOi,t-1
+β4Posti×ESOi,t-1 +β5Treatedi+β6Posti,t+β7ESOi,t-1+β8RTWj,t +β9Cashi,t-1 +β10Leveragei,t-1
+β11Dividendi,t-1 +β12Incomei,t-1 +β13WorkCapi,t-1+β14ZScorei,t-1 +β15Market-to-Booki,t-1
+Year FE+ Industry FE+ ɛijt (1)
• Regression Discontinuity Design (RDD)
To better establish the causal impact of unionisation, we resort to an RDD, focusing on
‘marginal elections’ within a small bandwidth around the 50 percent vote share
threshold. Consistent with prior RDD-based studies (Bradley et al. 2017; Campello et al.
2018), we use local linear (Equation 2) regressions to ensure the reliability of our results.
∆ESO Granted Per Employee=αl+ƛ×Unionisation+(X-0.5)×β
l+Unionisation×(X-0.5)×(βright-βleft)+ɛ (2)
RDD does not require the inclusion of covariates, given the underlying assumption of
continuity of vote share and firm characteristics around the threshold (Imbens and
Lemieux 2008; Lee and Lemieux 2010), which we empirically verify in Appendix 5.
We use the change instead of the level of ESO Incentives Granted Per Employee as our
main dependent variable in the RDD analysis, which allows us to effectively estimate
48
the difference-in-differences treatment effect using the RDD estimator. Essentially, we
compare the pre-to-post-election change in ESO Incentives Granted Per Employee
between the ‘marginal treated group’ and ‘marginal control group’. For robustness, we
also use the change in Number of ESO Granted Per Employee as our alternative
outcome variable. Since Lee and Mas (2012) document that the effect of unionisation
tends to materialise in 18 months, to make sure we capture the change in terms of ESO
incentives granted to employees, we primarily focus on the change from (t-1, t+2).
In terms of bandwidth choice in the RDD estimation, we follow the existing literature
and cross-validate the results by applying multiple bandwidths (Imbens and Lemieux
2008; Imbens and Kalyanaraman 2012; He et al. 2016a; Bradley et al. 2017; Campello
et al. 2018). We primarily use the optimal bandwidth developed by Imbens and
Kalyanaraman (2012), which minimises the mean squared error (MSE). To make sure
our results are not sensitive to this particular optimal bandwidth, we follow Campello et
al. (2018) and also include results for 75% and 125% of the optimal bandwidth.
Consistent with prior studies that apply the RDD methodology, we use both triangular
and rectangular kernel functions to modify the weightings of observations18 in local
linear regressions.
2.4. Empirical Findings
2.4.1 Moderating Effect of ESO Incentives on Union Strike Probability
To formally test our hypothesis H1 on the effect of ESO incentives on unions’ decision
to strike, we use a triple-differences approach and interact Treated*Post with a dummy
variable, ESO, which is equal to one if the lagged level of Log(Incentives Outstanding
18 While Cameron and Trivedi (2009) and Imbens and Lemieux (2008) suggest that rectangular and
triangular kernel functions are likely to generate similar results, the triangular kernel is preferred in our
RDD setting because it assigns more weight to the observations around the critical cutoff point (Fan and
Gijbels 1996; Calonico et al. 2016). The rectangular kernel is used as well, for robustness.
49
Per Employee) is above the sample median, and zero otherwise (Columns 1-4). As an
alternative specification, we also partition our sample into high-ESO-incentive and low-
ESO-incentive firms based on the sample median of the lagged proportion of
outstanding option incentives held by non-executive employees
(Incentives_Outstanding_Pct) (Columns 5-8). The three-way interaction term
Treated*Post*ESO captures the differential unionisation effect on strike likelihood
between high- and low-ESO-incentive firms.
As is shown in Table 2, Treated*Post is consistently positive and statistically significant,
which verifies our underlying assumption behind H1 that unionisation leads to a higher
probability of labour strikes. In line with our H1, the interaction term of interest,
Treated*Post*ESO, is significantly negative at the 1 percent level across all
specifications, suggesting that ESO incentives moderate the unionisation effect on strike
probability. Notably, the economic magnitude of the moderating effect is non-trivial.
The marginal effect indicates that the strike probability after unionisation among high-
ESO-incentive firms is 56 percent 19 lower than among their low-ESO-incentive
counterparts. There are two possible explanations for the moderating effect of ESO
incentives: First, as Reder and Neumann (1980) suggest, unions’ decision to strike can
be seen as a classic cost-benefit analysis. Unlike leverage, which reduces the labour-
perceived benefit of a strike (Myers and Saretto 2016), we argue that ESO incentives
will increase the potential cost of a strike, given that employees’ expected wealth is now
sensitive to any loss in firm value caused by the strike. Second, in addition to the
immediate financial implications, we expect that the strong interest-alignment effect of
ESO incentives also fundamentally changes the attitude of labour unions, shifting their
behaviour from monopolistic to more cooperative. Instead of bargaining for suboptimal
19 Based on the marginal effect at the means in Column (2).
50
wage increases, organised labour’s collective-bargaining strategies are likely to be more
long-term oriented, placing more emphasis on firm value creation in the medium-to-
long term, which would ultimately increase employees’ expected wealth through their
ESO.
***Insert Table 2 here***
2.4.2 ESO Incentives Granted in Response to Union Elections
Having established that ESO incentives moderate the unionisation effect on strike
probability, we now investigate the firm reaction to union election outcomes with regard
to their ESO-granting behaviour, thus testing our hypothesis H2.
2.4.2.1 Evidence from Regression Discontinuity Design (RDD) Analysis: Local Linear
Regressions
Table 3 reports the results of local linear regressions using outcome variables calculated
within different windows. Generally, while our results based on a window of (-1,1) are
statistically insignificant, our variable of interest Unionisation becomes positively
significant when we use the change in ESO grants within the window (-1,2) as our
outcome variable. These results suggest that firms’ adjustments in ESO grants in
reaction to unionisation start predominantly in the second year following the union
election, which seems to corroborate Lee and Mas (2012)’s finding that the unionisation
effect typically takes 15-18 months to fully materialise.
Specifically, Panel A of Table 3 shows that our results are consistently positive and
statistically significant within the window (-1,2) across different bandwidths, under a
triangular kernel function that assigns more weight to observations closer to the
threshold and is therefore the preferred weighting function. The results under the
rectangular kernel remain statistically significant and qualitatively the same, as shown
51
in Panel B. The evidence from the local linear regressions using the window (-1,2) helps
substantiate that ‘marginal winners’ (treated group) experience a statistically significant
and higher increase in ESO incentives granted per employee than ‘marginal losers’
(control group). The economic magnitude is non-trivial: a firm marginally returning a
decision to unionise in an election experiences a 1.69420 times larger increase in ESO
Incentives Granted Per Employee than a firm marginally voting against unionisation. As
a further robustness check, we use ∆Log(Number Incentives Granted)(-1,2) as an
alternative outcome variable and the results remain positive and statistically significant
(Columns 4-6).
Interestingly, the unionisation effect remains significant even during the third year after
unionisation (window (-1,3)), although the magnitude and significance of the treatment
effect are arguably weaker.
***Insert Table 3 here***
Overall, our RDD results offer robust evidence that firms marginally voting for
unionisation in union elections grant significantly more ESO incentives than those
marginally voting against unionisation, lending support to hypothesis H2. In other
words, in reaction to the exogenous increase in labour power, the unionised firms
strategically increase ESO incentives relative to their counterparts, in the control group,
that marginally failed to unionise.
We note that the RDD results have weak external validity and therefore the documented
causal effect may not be generalisable to observations falling outside the bandwidth we
study here. However, the consistent picture we obtain using different bandwidths
20 This result is based on a local linear regression under optimal bandwidth using the change in ESO
incentives within the (-1,2) window.
52
suggests that our results are not sensitive to the choice of bandwidth, alleviating this
generalisability concern.
2.4.2.2 RD Plots
We further produce regression discontinuity (RD) plots so as to visually inspect the
‘discontinuity’ around the cutoff of the outcome variable, namely, the causal impact of
labour unionisation on the change in ESO incentives granted per employee.21
***Insert Figure 2 here***
Figure 2 illustrates the RD plots using the linear function and second-order polynomial.
Panels A and B are based on our main outcome variable: ∆Log(Incentives Granted Per
Employee)(-1,2). Both the linear and polynomial plots demonstrate a dramatic jump on
the right-hand side of the graph as the vote share passes 50 percent, confirming a
positive impact of unionisation on the change in ESO Incentives Granted Per Employee.
The clear discontinuity in the outcome variable around the threshold is indicative of a
significant treatment effect. Such discontinuity is equally visible in Panels C and D
where we use the change in the number of ESO incentives granted per employee
(∆Log(Number Incentives Granted)(-1,2)) as an alternative outcome variable. Thus, the
RD plots offer graphical evidence to further support our prediction in hypothesis H2,
adding assurance and credibility to our causal inference drawn from the RDD
estimations.
Overall, the RDD analyses provide consistent evidence in support of our hypothesis H2
that firms strategically grant more ESO incentives following a unionisation event. We
argue that they do so since they wish to minimise strike risk through interest alignment.
21 The outcome variables hereafter are based on the change within the window (-1,2).
53
2.4.3 Right-to-Work (RTW) Laws
A key underlying assumption of the unionisation effect we infer from our observations
is the increase in labour bargaining power. Thus, the unionisation effect should be more
pronounced when unions enjoy greater power and moderated when unions’ bargaining
ability is undermined. To test this conjecture, we utilise state-level variation introduced
by the RTW laws in the U.S. that allow non-union employees to enjoy the same benefits
and treatment as union employees without paying union dues, therefore weakening
union power (Ellwood and Fine 1987; Campello et al. 2018). By exploiting this
exogenous state-level variation in union bargaining power due to RTW legislation,22 we
are able to investigate how RTW moderates the unionisation effect. We split our sample
into elections in RTW and non-RTW states based on whether a union election is held in
a firm establishment (branch) based in a state that has (has not) enacted RTW legislation,
respectively, and conduct subsample analysis using local linear regressions.
Table 4 reports our findings. Consistent with our conjectures, the unionisation effect is
only significant in firms holding union elections in non-RTW states (Panel B). In
contrast, the coefficient for unionisation is smaller and insignificant in firms holding
union elections in RTW states (Panel A), where union power is weak. This result is
consistent with firms responding to unionisation events by increasing ESO incentives
only when the unions are expected to have significant bargaining power. This helps
strengthen our inference from Section 4.2.
***Insert Table 4 here***
22 There are 22 RTW states during our sample period. The data source is the U.S. Department of Labour.
54
2.4.4 Role of Labour Skills
We further study how the firms’ adjustments of ESO incentives in reaction to
unionisation vary depending on their reliance on skilled labour. Given the prevalence of
ESO in high-skill industries, one could argue that the strategic increase in ESO
incentives might be driven solely by firms relying on skilled labour and that firms
relying heavily on low-skilled employees may not necessarily grant more ESO
incentives following unionisation. We consider this scenario unlikely, since
unionisation is not prevalent in high-skill industries. Still, this analysis is interesting
since it helps us to directly showcase the use of ESO incentives in low-skill industries as
well as high-skill industries.
To better understand the role of labour skills in the context of unionisation, we revisit
the labour economics literature. Prior literature establishes a positive union effect on
both wage items and non-wage items such as fringe benefits (Freeman 1980; Freeman
1981; Freeman and Medoff 1984; Pencavel and Hartsog 1984; Card 2001; Koeniger et
al. 2007). More importantly, numerous studies suggest that low-skilled workers tend to
benefit most from labour unions in terms of both pay improvement and job security
(Farber and Saks 1980; Freeman 1980; Lewis 1986; Card 1996). First, low-skilled
workers, who are typically at the bottom of the earnings distribution within the firm, are
more likely to be unsatisfied with their current pay and consequently gain most from
labour unions’ efforts to improve wages and reduce intra-firm pay inequalities (Farber
and Saks 1980). Second, compared with high-skilled employees, low-skilled workers
are exposed to higher unemployment risk (Akerlof and Yellen 1988). Therefore, a
priority for a labour union in a low-skill firm is to guarantee job security for its union
members, who are less mobile and have less bargaining power than high-skilled
employees (Farber and Saks 1980; Cooke 1983).
55
Given the lower pay and higher unemployment risk, low-skilled workers are more
reliant on labour unions to safeguard their jobs and negotiate higher wages on their
behalf. To meet the higher expectations and demands from their union members, labour
unions representing low-skilled workers are more likely to engage in collective-
bargaining activities and to be aggressive in pursuing their goals, for example by
initiating large-scale strikes. Hence, we argue that, compared with firms in high-skill
industries, firms with a higher proportion of low-skilled employees are subject to
stronger collective bargaining and thus exposed to a higher strike risk following
unionisation. Therefore, we predict that firms’ adjustment of ESO incentives in
response to unionisation will be more pronounced in firms in low-skill industries, where
strike risk is perceived to be relatively higher.
To conduct our analysis, we partition our sample into high-skill and low-skill firms
using an industry-specific Labour Skill Index (LSI), following Ghaly et al. (2017).
Essentially, the LSI captures the weighted average skill level of the occupations within
an industry, based on data from the Occupational Employment Statistics (OES) and the
O*NET program compiled by the U.S. Department of Labor. Table 5 presents the
results for this subsample analysis. Compared to the insignificant change for firms
relying on high-skilled labour, we find that the strategic adjustment of ESO incentives
granted following unionisation is significant in firms that rely heavily on low-skilled
labour, which is in line with our prediction. The results hold when we apply various
bandwidths, alternative specifications of the dependent variable (Columns 1-3 vs.
Columns 4-6) and a rectangular kernel function.23
*** Insert Table 5 here***
23 For brevity, results using the rectangular kernel function are not reported.
56
To make sure our results are not driven by the use of any particular proxy for labour
skills, we also use the three-year average R&D expenditure as an alternative measure
for the level of firm reliance on skilled labour. The underlying assumption is that firms
with higher R&D expenditure rely more heavily on skilled employees. We find very
similar results across various specifications (Table 6) using this alternative proxy.
*** Insert Table 6 here***
Overall, these results are consistent with the main finding of this study. In particular, the
prominent increase in ESO incentives in low-skill firms lends additional support to our
interest-realignment conjecture, since unionised firms that rely heavily on low-skilled
labour are perceived to be more vulnerable to strikes, thus having greater incentives to
be more responsive to unionisation by proactively improving interest alignment to
mitigate the potentially higher strike risk.24 In contrast, the weaker adjustment of ESO
incentives by high-skill firms is consistent with prior literature (Sesil et al. 2002; Ittner
et al. 2003; Chang et al. 2015) showing that employees in high-skill industries or high-
R&D firms are typically offered significant ESO or other equity-based incentives as part
of their compensation packages, irrespective of union election outcomes.
2.4.5 Placebo Test
Our RDD analysis exploits the exogenous and discontinuous increase in labour power
once the vote share passes the critical threshold of 50% for assigning the treatment of
unionisation. To check the robustness of our RDD results, we conduct a placebo test by
repeating the analysis using artificial thresholds for the treatment assignment. If the
increase in ESO grants we document in Table 3 is indeed caused by the unionisation,
24 Given the well-documented wage premium in unionised firms (Freeman 1980; Freeman and Medoff
1984; Pencavel and Hartsog 1984; Card 2001; Koeniger et al. 2007), we presume the ESO are granted as
a supplement to rather than substitution for wages. Due to data limitations, we cannot observe
wage/salary data for rank-and-file employees at the firm-year level, so cannot verify this conjecture.
57
which is determined by the vote share in favour of unionisation exceeding 50% in the
union election, then we should not observe any discontinuity or systematic difference in
ESO grants when we arbitrarily impose artificial cutoff points for unionisation.
Table 7 reports the results for local linear regressions using the triangular kernel
function25 based on a selection of artificial thresholds (40%, 45%, 55% and 60%).
Consistent with our previous RDD analyses, we use both the change in the ESO
incentives granted per employee (ΔLog(Incentive Granted Per Employee)(-1,2)) and the
change in the number of ESO incentives granted per employee (∆Log(Number Granted
Per Employee)(-1,2)) as our outcome variables. For comparison, results based on the true
threshold at 50% are also presented.
In line with our expectation, it is reassuring to find that the results based on arbitrary
thresholds are statistically insignificant across all specifications, with most of them
having a negative sign, in contrast to the consistently positive and statistically
significant coefficients based on the true 50% threshold in Columns (3) and (8). The
fact that our results for both outcome variables are only significant when applying the
true threshold at 50% provides further assurance that the increase in ESO grants we
observe is attributable to the treatment effect of unionisation rather than a spurious
correlation.
*** Insert Table 7 here***
2.5 Conclusion
In this paper, we examine the influence of employee stock options (ESO) on the
propensity of labour unions to instigate strikes. By exploiting the unique setting of
union elections in U.S. listed firms in a propensity-score-matched (PSM) sample, we
25 The results are very similar when using the rectangular kernel function.
58
find that ESO incentives significantly moderate labour unions’ power/motivation to
initiate strikes. The economic magnitude is non-trivial: firms with high (ex-ante) ESO
incentives have, on average, a 56 percent lower post-unionisation strike probability, in
comparison with their low-ESO-incentive counterparts. We argue that this evidence is
consistent with ESO partially mitigating strike risk by improving the interest alignment
between organised labour and the firm.
Subsequent analyses using a quasi-experimental RDD approach provide consistent
evidence that firms strategically grant more ESO incentives to mitigate strike risk and
improve interest alignment in response to a unionisation event. Further subsample
analyses confirm that such strategic corporate reaction is more pronounced for firms
holding elections in non-right-to-work states where union power is stronger and for
firms in low-skill industries where strike risk is perceived to be higher.
Overall, our study adds to our understanding of the behaviour of labour unions,
specifically how ESO influence unions’ decision to strike. We also enrich the literature
on the strategic corporate decision to improve firms’ bargaining position and human
capital management, by identifying ESO as a strategic tool that helps moderate the
strike incentives of labour unions. Additionally, we show that interest alignment is an
important determinant of the adoption of ESO in labour-intensive non-tech industries,
which complements prior evidence on the use of ESO in new-economy firms.
Finally, our study has implications for accounting standard setters and policymakers.
Despite the evident benefits of ESO highlighted in this study, the change in the
accounting treatment of equity-based compensation brought about by FAS 123R has
created hurdles for the adoption and growth of ESO and other employee ownership
schemes, which could potentially impede firm value creation. While it is important that
regulators and standard setters prevent the abuse of equity-based compensation schemes,
59
our findings highlight the need for firms to be given sufficient policy support (e.g., tax
benefits) favouring employee ownership. More ESO-friendly policies could help
improve interest alignment between firms and one of their key stakeholders, their
employees, which could contribute to value creation even among the heavily unionised
manufacturing industries that are the current focus of efforts to revitalise the U.S.
economy. Last but not least, our evidence from the U.S. could serve as a reference for
policymakers in other jurisdictions, such as Europe, where the labour union movement
is historically stronger and plays a more prominent role in the economy.
60
References
Addison, J.T. and Hirsch, B.T. (1989). Union Effects on Productivity, Profits, and
Growth: Has the Long Run Arrived? Journal of Labor Economics, 7(1), pp.72-105.
Agrawal, A.K. (2012). Corporate Governance Objectives of Labor Union Shareholders:
Evidence from Proxy Voting. Review of Financial Studies, 25(1), pp.187–226.
Akerlof, G.A. and Yellen, J.L. (1988). Fairness and Unemployment. American
Economic Review, 78(2), pp.44–49.
Aobdia, D. and Cheng, L. (2018). Unionization, Product Market Competition, and
Strategic Disclosure. Journal of Accounting and Economics, 65(2–3), pp.331–357.
Ashenfelter, O. and Johnson, G.E. (1969). Bargaining Theory, Trade Unions, and
Industrial Strike Activity. American Economic Review, 59(1), pp.35–49.
Atanassov, J. and Kim, E.H. (2009). Labor and Corporate Governance: International
Evidence from Restructuring Decisions. Journal of Finance, 64(1), pp.341–374.
Babenko, I., Lemmon, M. and Tserlukevich, Y. (2011). Employee Stock Options and
Investment. Journal of Finance, 66(3), pp.981–1009.
Babenko, I. and Sen, R. (2014). Money Left on the Table: An Analysis of Participation
in Employee Stock Purchase Plans. Review of Financial Studies, 27(12), pp.3658–3698.
Babenko, I. and Sen, R. (2016). Do Nonexecutive Employees Have Valuable
Information? Evidence from Employee Stock Purchase Plans. Management Science,
62(7), pp.1878–1898.
Babenko, I. and Tserlukevich, Y. (2009). Analyzing the Tax Benefits from Employee
Stock Options. Journal of Finance, 64(4), pp.1797–1825.
Babenko, I. and Tserlukevich, Y. (2016). Will I Get Paid? Employee Stock Options and
Mergers and Acquisitions. SSRN Electronic Journal, pp.1–47.
Banning, K. and Chiles, T. (2007). Trade-Offs in the Labor Union-CEO Compensation
Relationship. Journal of Labor Research, 28(2), pp.347–357.
Beatty, A. (1995). The Cash Flow and Informational Effects of Employee Stock
Ownership Plans. Journal of Financial Economics, 38(2), pp.211–240.
Becker, B.E. and Olson, C.A. (1986). The Impact of Strikes on Shareholder Equity.
Industrial and Labor Relations Review, 39(3), pp.425–438.
Bens, D.A., Nagar, V., Skinner, D.J. and Wong, M.H.F. (2003). Employee Stock
Options, EPS Dilution, and Stock Repurchases. Journal of Accounting and Economics,
36(1–3), pp.51–90.
Bens, D.A., Nagar, V. and Wong, M.H.F. (2002). Real Investment Implications of
Employee Stock Option Exercises. Journal of Accounting Research, 40(2), pp.359–393.
Bhagat, S., Brickley, J.A. and Lease, R.C. (1985). Incentive Effects of Stock Purchase
Plans. Journal of Financial Economics, 14, pp.195–215.
Blair, M.M., Kruse, D.L. and Blasi, J. (2000). Employee Ownership: An Unstable Form
or a Stabilizing Force? In The New Relationship: Human Capital in the American
Corporation, eds. Blair, M. & Kochan, T., Brookings Institution Press. Washington, DC,
pp. 241–298.
61
Blasi, J., Conte, M. and Kruse, D. (1996). Employee Stock Ownership and Corporate
Performance among Public Companies. Industrial and Labor Relations Review, 50(1),
pp.60–79.
Blasi, J., Freeman, R. and Kruse, D. (2016). Do Broad-Based Employee Ownership,
Profit Sharing and Stock Options Help the Best Firms Do Even Better? British Journal
of Industrial Relations, 54(1), pp.55–82.
Blasi, J., Kroumova, M., Kruse, D.L. and Sesil, J. (2002). Broad-Based Employee Stock
Options — A Union-Nonunion Comparison. Advances in Industrial & Labor Relations.,
11, pp.69–94.
Bova, F. (2013). Labor Unions and Management’s Incentive to Signal a Negat ive
Outlook. Contemporary Accounting Research, 30(1), pp.14–41.
Bova, F., Dou, Y. and Hope, O.-K. (2015a). Employee Ownership and Firm Disclosure.
Contemporary Accounting Research, 32(2), pp.639–673.
Bova, F., Kolev, K., Thomas, J.K. and Zhang, X.F. (2015b). Non-Executive Employee
Ownership and Corporate Risk. The Accounting Review, 90(1), pp.115–145.
Bradley, D., Kim, I. and Tian, X. (2017). Do Unions Affect Innovation? Management
Science, 63(7), pp.2251–2271.
Bronars, S. and Deere, D. (1991). The Threat of Unionization, the Use of Debt, and the
Preservation of Shareholder Wealth. Quarterly Journal of Economics, 106(1), pp.231–
254.
Bureau of Labor Statistics. (2017). Major Work Stoppages in 2016.
Call, A.C., Kedia, S. and Rajgopal, S. (2016). Rank and File Employees and the
Discovery of Misreporting: The Role of Stock Options. Journal of Accounting and
Economics, 62(2–3), pp.277–300.
Calonico, S., Cattaneo, M.D. and Farrell, M.H. (2016). Rdrobust: Software for
Regression Discontinuity Designs. The Stata Journal, (ii), pp.1–30.
Cameron, A.C. and Trivedi, P.K. (2009). Micreconometrics Using Stata. College
Station, TX: Stata Press.
Campello, M., Gao, J., Qiu, J. and Zhang, Y. (2018). Bankruptcy and the Cost of
Organized Labor: Evidence from Union Elections. Review of Financial Studies, 31(3),
pp.980–1013.
Card, D. (1996). The Effect of Unions on the Structure of Wages: A Longitudinal
Analysis. Econometrica, 64(4), pp.957–979.
Card, D. (2001). The Effect of Unions on Wage Inequality in the U.S. Labor Market.
Industrial and Labor Relations Review, 54(2), pp.296–315.
Cattaneo, M.D., Jansson, M. and Ma, X. (2016). Rddensity: Manipulation Testing
Based on Density Discontinuity. Stata Journal, (ii), pp.1–18.
Chang, S. (1990). Employee Stock Ownership Plans and Shareholder Wealth: An
Empirical Investigation. Financial Management, 19(1), pp.48–58.
Chang, S. and Mayers, D. (1992). Managerial Vote Ownership and Shareholder Wealth.
Evidence from Employee Stock Ownership Plans. Journal of Financial Economics,
32(1), pp.103–131.
62
Chang, X., Fu, K., Low, A. and Zhang, W. (2015). Non-Executive Employee Stock
Options and Corporate Innovation. Journal of Financial Economics, 115(1), pp.168–
188.
Chaplinsky, S. and Niehaus, G. (1994). The Role of ESOPs in Takeover Contests.
Journal of Finance, 49(4), pp.1451–1470.
Chen, H.J., Kacperczyk, M. and Ortiz-Molina, H. (2011). Labor Unions, Operating
Flexibility, and the Cost of Equity. Journal of Financial and Quantitative Analysis,
46(01), pp.25–58.
Chino, A. (2016). Do Labor Unions Affect Firm Payout Policy?: Operating Leverage
and Rent Extraction Effects. Journal of Corporate Finance, 41, pp.156–178.
Choudhary, P., Rajgopal, S. and Venkatachalam, M. (2009). Accelerated Vesting of
Employee Stock Options in Anticipation of FAS 123-R. Journal of Accounting
Research, 47(1), pp.105–146.
Chung, R., Lee, B.B.-H., Lee, W.-J. and Sohn, B.C. (2016). Do Managers Withhold
Good News from Labor Unions? Management Science, 62(1), pp.46–68.
Chyz, J.A., Ching Leung, W.S., Zhen Li, O. and Meng Rui, O. (2013). Labor Unions
and Tax Aggressiveness. Journal of Financial Economics, 108(3), pp.675–698.
Clark, K.B. (1984). Unionization and Firm Performance: The Impact on Profits, Growth,
and Productivity. American Economic Review, 74(5), pp.893–919.
Cooke, W.N. (1983). Determinants of the Outcomes of Union Certification Elections.
Industrial and Labor Relations Review, 36(3), pp.402–414.
Core, J. and Guay, W. (2001). Stock Option Plans for Non-Executive Employees.
Journal of Financial Economics, 61(2), pp.253–287.
Core, J. and Guay, W. (2002). Estimating the Value of Employee Stock Option
Portfolios and their Sensitivities to Price and Volatility. Journal of Accounting Research,
40(3), pp.613–640.
Cramton, P., Mehran, H. and Tracy, J. (2008). ESOP Fables: The Impact of Employee
Stock Ownership Plans on Labor Disputes. Federal Reserve Bank of New York Staff
Report, 347(September), pp.1–38.
Cramton, P. and Tracy, J. (1994). The Determinants of U.S. Labor Disputes. Journal of
Labor Economics, 12(2), pp.180–209.
Dhillon, U.S. and Ramírez, G.G. (1994). Employee Stock Ownership and Corporate
Control: An Empirical Study. Journal of Banking & Finance, 18(1), pp.9–25.
DiNardo, J. and Hallock, K. (2002). When Unions “Mattered”: The Impact of Strikes on
Financial Markets, 1925–1937. ILR Review, 55(2), pp.219–233.
DiNardo, J., Hallock, K. and Pischke, J.-S. (1997). Unions and Managerial Pay.
National Bureau of Economic Research, 6318, pp.1–40.
DiNardo, J. and Lee, D. (2004). Economic Impacts of New Unionization on Private
Sector Employers: 1984–2001. Quarterly Journal of Economics, 119(4), pp.1383–1441.
Ellwood, D.T. and Fine, G. (1987). The Impact of Right-To-Work Laws on Union
Organizing. Journal of Political Economy, 95(2), pp.250–273.
EY. (2014). EY Stock Based Incentive Survey 2014. Available at:
63
https://www.ey.com/in/en/services/tax/ey-stock-based-incentive-survey-the-results
[Accessed: 01 July 2017]
Faleye, O., Mehrotra, V. and Morck, R. (2006). When Labor Has a Voice in Corporate
Governance. Journal of Financial and Quantitative Analysis, 41(03), pp.489-510.
Fama, E.F. and French, K.R. (2005). Financing Decisions: Who Issues Stock? Journal
of Financial Economics, 76(3), pp.549–582.
Fan, J. and Gijbels, I. (1996). Local Polynomial Modelling and its Applications:
Monographs on Statistics and Applied Probability. Boca Raton, FL: CRC Press.
Fang, H., Nofsinger, J.R. and Quan, J. (2015). The Effects of Employee Stock Option
Plans on Operating Performance in Chinese Firms. Journal of Banking & Finance, 54,
pp.141–159.
Farber, H.S. and Saks, D.H. (1980). Why Workers Want Unions: The Role of Relative
Wages and Job Characteristics. Journal of Political Economy, 88(2), pp.349–369.
FASB. (2004). Statement of Financial Accounting No.123: Share-Based Payment.
Available
at:https://www.fasb.org/jsp/FASB/Document_C/DocumentPage?cid=1218220124271&
acceptedDisclaimer=true [Accessed: 01 July 2017]
Fauver, L. and Fuerst, M.E. (2006). Does Good Corporate Governance Include
Employee Representation? Evidence from German Corporate Boards. Journal of
Financial Economics, 82(3), pp.673–710.
Freeman, R. (1980). Unionism and the Dispersion of Wages. Industrial and Labor
Relations Review, 34(1), pp.3–23.
Freeman, R. (1981). The Effect of Unionism on Fringe Benefits. Industrial and Labor
Relations Review, 34(4), pp.489-509.
Freeman, R. and McVea, J. (2001). A Stakeholder Approach to Strategic Management.
SSRN Electronic Journal, pp.1–32.
Freeman, R. and Medoff, J. (1979). Two Faces of Unionism. Public Interest, 57, pp.69–
93.
Freeman, R. and Medoff, J. (1984). What Do Unions Do? New York: Basic Books.
Gale, W.G. and Potter, S.R. (2002). An Economic Evaluation of the Economic Growth
and Tax Relief Reconciliation Act of 2001. National Tax Journal, 55(1), pp.133–177.
Garmaise, M.J. (2008). Production in Entrepreneurial Firms: The Effects of Financial
Constraints on Labor and Capital. Review of Financial Studies, 21(2), pp.543–577.
Ghaly, M., Dang, V.A. and Stathopoulos, K. (2017). Cash Holdings and Labor
Heterogeneity: The Role of Skilled Labor. Review of Financial Studies, 30(10),
pp.3636–3668.
Gopalan, R., Milbourn, T., Song, F. and Thakor, A.V. (2014). Duration of Executive
Compensation. Journal of Finance, 69(6), pp.1–50.
Gordon, L.A. and Pound, J. (1990). ESOPs and Corporate Control. Journal of Financial
Economics, 27(2), pp.525–555.
GSS (2014). General Social Survey 2014.
Hayes, R.M., Lemmon, M. and Qiu, M. (2012). Stock Options and Managerial
64
Incentives for Risk Taking: Evidence from FAS 123R. Journal of Financial Economics,
105(1), pp.174–190.
He, J., Tian, X. and Yang, H. (2016). Payout Policy under Enhanced Labor Power:
Evidence from a New Approach. SSRN Electronic Journal, August, pp.1–58.
Hochberg, Y.V. and Lindsey, L. (2010). Incentives, Targeting, and Firm Performance:
An Analysis of Non-Executive Stock Options. Review of Financial Studies, 23(11),
pp.4148–4186.
Huang, Q., Jiang, F., Lie, E. and Que, T. (2017). The Effect of Labor Unions on CEO
Compensation. Journal of Financial and Quantitative Analysis, 52(2), pp.553–582.
Ikäheimo, S., Kjellman, A., Holmberg, J. and Jussila, S. (2004). Employee Stock Option
Plans and Stock Market Reaction: Evidence from Finland. The European Journal of
Finance, 10(2), pp.105–122.
Imbens, G. and Kalyanaraman, K. (2012). Optimal Bandwidth Choice for the
Regression Discontinuity Estimator. The Review of Economic Studies, 79(3), pp.933–
959.
Imbens, G.W. and Lemieux, T. (2008). Regression Discontinuity Designs: A Guide to
Practice. Journal of Econometrics, 142(2), pp.615–635.
Iqbal, Z. and Hamid, S.A. (2000). Stock Price and Operating Performance of ESOP
Firms: A Time-Series Analysis. Quarterly Journal of Business and Economics, 39(3),
pp.25–47.
Ittner, C.D., Lambert, R.A. and Larcker, D.F. (2003). The Structure and Performance
Consequences of Equity Grants to Employees of New Economy Firms. Journal of
Accounting and Economics, 34(1–3), pp.89–127.
Jiraporn, P. (2007). ESOPs and Earnings Management: An Empirical Note. Applied
Financial Economics Letters, 3(5), pp.287–293.
Jones, D.C. and Kato, T. (1995). The Productivity Effects of Employee Stock-
Ownership Plans and Bonuses: Evidence from Japanese Panel Data. American
Economic Review, 85(3), pp.391–414.
Kim, E.H. and Ouimet, P. (2014). Broad-Based Employee Stock Ownership: Motives
and Outcomes. Journal of Finance, 69(3), pp.1273–1319.
Klasa, S., Maxwell, W.F. and Ortiz-Molina, H. (2009). The Strategic Use of Corporate
Cash Holdings in Collective Bargaining with Labor Unions. Journal of Financial
Economics, 92(3), pp.421–442.
Klein, K. (1987). Employee Stock Ownership and Employee Attitudes: A Test of Three
Models. Journal of Applied Psychology, 72(2), pp.319–332.
Koeniger, W., Leonardi, M. and Nunziata, L. (2007). Labor Market Institutions and
Wage Inequality. ILR Review, 60(3), pp.340–356.
Kroumova, M.K. and Sesil, J.C. (2006). Intellectual Capital, Monitoring, and Risk:
What Predicts the Adoption of Employee Stock Options? Industrial Relations, 45(4),
pp.734–752.
Kruse, D.L. (1996). Why Do Firms Adopt Profit-Sharing and Employee Ownership
Plans? British Journal of Industrial Relations, 34(4), pp.515–538.
65
Lee, D.S. (2008). Randomized Experiments from Non-Random Selection in U.S. House
Elections. Journal of Econometrics, 142(2), pp.675–697.
Lee, D.S. and Lemieux, T. (2010). Regression Discontinuity Designs in Economics.
Journal of Economic Literature, 48(2), pp.281–355.
Lee, D.S. and Mas, A. (2012). Long-Run Impacts of Unions on Firms: New Evidence
from Financial Markets, 1961-1999. The Quarterly Journal of Economics, 127(1),
pp.333–378.
Lewis, H.G. (1983). Union Relative Wage Effects: A Survey of Macro Estimates.
Journal of Labor Economics, 1(1), pp.1–27.
Lewis, H.G. (1986). Union Relative Wage Effects: A Survey. Chicago: University of
Chicago Press.
Liberty, S.E. and Zimmerman, J.L. (1986). Labor Union Contract Negotiations and
Accounting Choices. The Accounting Review, 61(4), pp.692–712.
Lipset, S.M. and Katchanovski, I. (2001). The Future of Private Sector Unions in the
U.S. Journal of Labor Research, 22(2), pp.229–244.
Lobel, O. (2017). The Gig Economy & The Future of Employment and Labor Law.
University of San Francisco Law Review, 51, pp.51–74.
Matsa, D.A. (2010). Capital Structure as a Strategic Variable: Evidence from Collective
Bargaining. Journal of Finance, 65(3), pp.1197–1232.
McCrary, J. (2008). Manipulation of the Running Variable in the Regression
Discontinuity Design: A Density Test. Journal of Econometrics, 142(2), pp.698–714.
McHugh, R. (1991). Productivity Effects of Strikes in Struck and Nonstruck Industries.
Industrial and Labor Relations Review, 44(4), pp.722-732.
Myers, B.W. and Saretto, A. (2016). Does Capital Structure Affect the Behavior of
Nonfinancial Stakeholders? An Empirical Investigation into Leverage and Union
Strikes. Management Science, 62(11), pp.3235–3253.
NCEO. (2017). A Statistical Profile of Employee Ownership. Available at:
https://www.nceo.org/articles/employee-ownership-by-the-numbers [Accessed: 01 July
2017]
O’Boyle, E.H., Patel, P.C. and Gonzalez-Mule, E. (2016). Employee Ownership and
Firm Performance: A Meta-Analysis. Human Resource Management Journal, 26(4),
pp.425–448.
Oyer, P. (2004). Why Do Firms Use Incentives That Have No Incentive Effects?
Journal of Finance, 59(4), pp.1619–1650.
Oyer, P. and Schaefer, S. (2005). Why Do Some Firms Give Stock Options to All
Employees?: An Empirical Examination of Alternative Theories. Journal of Financial
Economics, 76(1), pp.99–133.
Oyer, P. and Schaefer, S. (2006). Costs of Broad-Based Stock Option Plans. Journal of
Financial Intermediation, 15(4), pp.511–534.
Park, R., Kruse, D. and Sesil, J. (2004). Does Employee Ownership Enhance Firm
Survival? In Advances in the Economic Analysis of Participatory and Labor-Managed
Firms. Emerald Publishing Limited
66
Pencavel, J. and Hartsog, C.E. (1984). A Reconsideration of the Effects of Unionism on
Relative Wages and Employment in the United States, 1920-1980. Journal of Labor
Economics, 2(2), pp.193–232.
Pierce, J., Rubenfeld, S.A. and Morgan, S. (1991). Employee Ownership: A Conceptual
Model of Process and Effects. Academy of Management Review, 16(1), pp.121–144.
Qiu, Y. (2016). Debt Structure as a Strategic Bargaining Tool. ECGI Finance Working
Paper, (November).
Qiu, Y. and Shen, T. (2017). Organized Labor and Loan Pricing: A Regression
Discontinuity Design Analysis. Journal of Corporate Finance, 43, pp.407–428.
Rauh, J.D. (2006). Own Company Stock in Defined Contribution Pension Plans: A
Takeover Defense? Journal of Financial Economics, 81(2), pp.379–410.
Reder, M.W. and Neumann, G.R. (1980). Conflict and Contract: The Case of Strikes.
Journal of Political Economy, 88(5), pp.867–886.
Reuters. (2008). Boeing Set to Announce New 787 Delays. Available at:
https://www.reuters.com/article/us-boeing-idUSTRE4B384G20081204 [Accessed: 01
July 2017]
Reuters. (2017). Trump Vows to Protect ‘Made In America’ Products. Available at:
https://www.reuters.com/article/us-usa-trade-manufacturing/trump-vows-to-protect-
made-in-america-products-idUSKBN1A2293 [Accessed: 01 July 2017]
Rosen, C. (1990). The Record of Employee Ownership. Financial Management, 19(1),
pp.39–47.
Rosen, C., Case, J. and Staubus, M. (2005). Every Employee an Owner. Really.
Harvard Business Review, 83(6), pp.122–130.
Rosen, C. and Quarrey, M. (1987). How Well is Employee Ownership Working?
Harvard Business Review, 65(5), pp.126–129.
Ruback, R.S. and Zimmerman, M.B. (1984). Unionization and Profitability: Evidence
from the Capital Market. Journal of Political Economy, 92(6), p.1134.
Schmidt, M.B. and Berri, D.J. (2004). The Impact of Labor Strikes on Consumer
Demand: An Application to Professional Sports. American Economic Review, 94(1),
pp.344–357.
Sesil, J.C., Kroumova, M.K., Blasi, J.R. and Kruse, D.L. (2002). Broad-Based
Employee Stock Options in US ‘New Economy’ Firms. British Journal of Industrial
Relations, 40(2), pp.273–294.
Singh, P. and Agarwal, N.C. (2002). Union Presence and Executive Compensation: An
Exploratory Study. Journal of Labor Research, 23(4), pp.631–646.
Siu, W., Leung, C., Li, O.Z. and Rui, O.M. (2009). Labor Union and Accounting
Conservatism. SSRN eLibrary, pp.0–54.
Tian, X. and Wang, W. (2016). Hard Marriage with Heavy Burdens: Labor Unions as
Takeover Deterrents. SSRN Electronic Journal, pp.1–43.
Tracy, J.S. (1986). An Investigation into the Determinants of U.S. Strike Activity.
American Economic Review, 76(3), pp.423–436.
Vedder, R. and Gallaway, L. (2002). The Economic Effects of Labor Unions Revisited.
67
Journal of Labor Research, 23(1), pp.105–130.
Wall Street Journal. (2016). Verizon CEO: Strike May Hit Results. Available at:
https://www.wsj.com/articles/verizon-ceo-strike-may-hit-results-1464101544 [Accessed:
01 July 2017]
Wheeler, H.N. (2002). The Future of the American Labor Movement. Cambridge, UK:
Cambridge University Press.
Yates, J. (2006). Unions and Employee Ownership: A Road to Economic Democracy?
Industrial Relations, 45(4), pp.709–733.
Zhang, W. (2011). The Tradeoff between Employee Ownership and CEO
Compensation: Evidence from Defined Contribution Pension Plans. SSRN Electronic
Journal, pp.1–58.
68
Table 1. Descriptive Statistics
This table provides summary statistics of our sample. Panel A reports union election statistics collected from the National Labour Relations Board (NLRB). Panel B reports employee stock options (ESO) data collected from the Execucomp and CRSP databases. Panel C reports the descriptive statistics for all
variables included in the strike analyses. All variables are defined in Appendix 1.
Variable Mean 25% Median 75% SD N
Panel A: Union Elections
Union Year 2006.065 2004 2006 2008 2.329 324 Outcome 0.361 0.000 0.000 1.000 0.481 324
Vote Share 0.451 0.306 0.414 0.569 0.206 324
Vote Total 173.515 72.500 106.000 181.000 202.673 324
Vote For 81.435 28.000 50.000 87.000 134.879 324
Vote Against 92.080 33.500 54.000 99.500 121.320 324
Panel B: ESO Grants Variables
Log (Incentives Granted Per Employee) 2.012 0.596 2.043 3.090 1.468 248
Log(Number Granted Per Employee) 3.312 2.408 3.685 4.528 1.728 234
∆Log(Incentives Granted Per Employee)(-1,1) -0.058 -0.688 -0.044 0.348 1.384 142 ∆Log(Number Granted Per Employee)(-1,1) -0.264 -0.729 -0.113 0.216 1.376 137
∆Log(Incentives Granted Per Employee)(-1,2) -0.208 -0.940 -0.132 0.271 1.506 137
∆Log(Number Granted Per Employee)(-1,2) -0.383 -0.940 -0.186 0.204 1.621 132
(continued on next page)
69
Panel C: Strike Test Sample
Log(Incentives Outstanding Per Employee) 3.890 3.210 3.815 4.739 1.289 1026
Incentives_Outstanding_Pct 0.734 0.641 0.767 0.878 0.188 1035 Treated 0.500 0.000 0.500 1.000 0.500 1368
Post 0.444 0.000 0.000 1.000 0.497 1368
RTW 0.270 0.000 0.000 1.000 0.444 1368 Strike Dummy 0.070 0.000 0.000 0.000 0.256 1368
Strike Risk 1.091 1.000 1.000 1.000 0.351 1368
ΔCash 0.015 -0.009 0.003 0.030 0.065 1368
ΔLeverage -0.030 -0.154 -0.008 0.146 0.984 1368 ΔDividend -0.009 0.000 0.000 0.013 0.199 1368
ΔIncome -0.001 -0.010 0.003 0.013 0.032 1368
ΔWorkCap 0.008 -0.014 0.007 0.034 0.053 1368 ΔZScore -0.006 -0.217 0.006 0.286 0.631 1368
ΔMarket-to-Book -0.039 -0.167 0.006 0.105 0.253 1368
70
Table 2. Moderating Effect of ESO on Union Strike Risk
The table below reports the results for the moderating effect of ESO on a union’s propensity to strike. The variable of interest is Treated*Post*ESO. In Columns (1)-(4), ESO is a dummy variable, equal to 1 if the proportion of number of outstanding stock options held by employees relative to the total number
of outstanding stock options at t-1 (Log(Incentives Outstanding Per Employee)t-1) is above the sample median. As an alternative proxy, we rerun our analyses
in Columns (5)-(8) using the proportion of outstanding ESO incentives held by employees at t-1 relative to the total outstanding ESO incentives (Incentives_Outstanding_Pct)t-1 to construct the ESO dummy. The dependent variable is Strike Dummy, equal to 1 if there is a strike during the fiscal year.
For robustness, we also use Strike Risk (1=no strike; 2=one strike; 3=multiple strikes) and run ordered probit regressions. P-values are displayed in
parentheses with standard errors clustered at the firm level. ***, ** and * denote significance levels of 1%, 5% and 10%, respectively. All variables are
defined in Appendix 1.
ESO Conditioner Log(Incentives Outstanding Per Employee)t-1 (Incentives_Outstanding_Pct)t-1
Probit Ordered Probit Probit Ordered Probit
(1) (2) (3) (4) (5) (6) (7) (8)
Strike Dummy Strike Dummy Strike Risk Strike Risk Strike Dummy Strike Dummy Strike Risk Strike Risk
Treated*Post*ESO -5.863*** -5.608*** -6.299*** -6.038*** -3.812*** -3.324*** -4.193*** -3.895***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Treated*Post 5.433*** 5.344*** 5.624*** 5.544*** 4.404*** 4.076*** 4.477*** 4.378***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Treated*ESO 1.675*** 1.967*** 1.521*** 1.861*** 0.945 1.105 0.989* 1.194
(0.001) (0.004) (0.001) (0.003) (0.131) (0.168) (0.092) (0.121)
Post*ESO 5.248*** 5.224*** 5.606*** 5.645*** 3.781*** 3.613*** 4.072*** 4.161***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Treated -0.578 -0.931 -0.350 -0.694 -0.309 -0.437 -0.271 -0.425
(0.237) (0.162) (0.404) (0.222) (0.600) (0.564) (0.616) (0.554)
Post -5.126*** -5.151*** -5.334*** -5.393*** -4.311*** -4.183*** -4.433*** -4.554***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
ESO -1.235*** -1.234*** -1.083*** -1.122*** 0.323 0.581 0.264 0.488
(0.000) (0.002) (0.000) (0.000) (0.430) (0.170) (0.497) (0.283)
RTW -0.812*** -0.627 -0.776*** -0.544 -0.730** -0.410 -0.714** -0.391
(0.009) (0.120) (0.005) (0.147) (0.030) (0.359) (0.018) (0.368)
(continued on next page)
71
ΔCash_lag -6.308* -6.643* -3.879 -4.800*
(0.074) (0.050) (0.184) (0.082)
ΔLeverage_lag 0.297 0.281* 0.483*** 0.485***
(0.102) (0.075) (0.010) (0.004)
ΔDividend_lag -0.076 0.499 2.226 2.568
(0.973) (0.810) (0.307) (0.255) ΔIncome_lag -0.748 -0.728 1.437 1.652
(0.885) (0.888) (0.817) (0.801)
ΔWorkCap_lag 4.924 5.979* 3.612 4.378
(0.157) (0.086) (0.320) (0.225) ΔZScore_lag -0.487 -0.618 -0.420 -0.648
(0.295) (0.187) (0.467) (0.271)
ΔMarket-to-Book_lag 1.105 1.542 1.500 2.066**
(0.314) (0.152) (0.143) (0.037)
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes
Pseudo R2 0.298 0.331 0.412 0.437 0.307 0.344 0.419 0.448 N 428 428 1026 1026 428 428 1035 1035
72
Table 3. Impact of Unionisation on ESO Grants: Regression Discontinuity Design
This table presents local linear regression results using (a) change in Log(Incentives Granted Per Employee) in Columns (1)-(3) and (b) change in
Log(Number Granted Per Employee) in Columns (4)-(6) as the dependent variables. For each dependent variable, we use the change in ESO grants within
different windows: (t-1, t+1), (t-1, t+2) and (t-1, t+3) with t being the year of the union election. In Panel A, we run local linear regressions under a triangular kernel using the optimal bandwidth defined by Imbens and Kalyanaraman (2012). Following Campello et al. (2018), we also use 75% and 125% of the
optimal bandwidth as a robustness check. As a further robustness test, we repeat the RDD analyses using a rectangular kernel in Panel B. The variable of
interest is Unionisation. P-values are displayed in parentheses with standard errors clustered by firm. ***, ** and * indicate significance at the 1%, 5% and
10% level, respectively. All variables are defined in Appendix 1.
Panel A: Triangular Kernel
Window (-1,1)
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3)
(4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal Optimal
75%
Optimal
125%
Optimal
Unionisation
0.155 -0.049 0.357 0.666 0.667 0.807
(0.801) (0.941) (0.514)
(0.288) (0.295) (0.186)
Window (-1,2)
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3)
(4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal Optimal
75%
Optimal
125%
Optimal
Unionisation
1.694*** 1.683*** 1.516*** 1.511** 1.362** 1.593**
(0.004) (0.009) (0.007)
(0.025) (0.047) (0.014)
(continued on next page)
73
Panel B: Rectangular Kernel
Window (-1,1)
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3)
(4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal Optimal
75%
Optimal
125%
Optimal
Unionisation -0.005 -0.208 0.687 0.562 0.720 0.516
(0.995) (0.778) (0.255) (0.414) (0.300) (0.424)
(continued on next page)
Window (-1,3)
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3)
(4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal Optimal
75%
Optimal
125%
Optimal
Unionisation 1.343** 1.215* 1.397** 1.186* 1.179 1.107*
(0.028) (0.071) (0.016) (0.098) (0.161) (0.092)
74
Window (-1,2)
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3)
(4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal Optimal
75%
Optimal
125%
Optimal
Unionisation 1.556** 1.690** 1.766*** 1.842** 1.298 1.491**
(0.037) (0.036) (0.003) (0.031) (0.100) (0.038)
Window (-1,3)
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3)
(4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal Optimal
75%
Optimal
125%
Optimal
Unionisation 0.953 1.060 1.650*** 0.987 1.346 1.388**
(0.196) (0.186) (0.007) (0.165) (0.146) (0.044)
75
Table 4. Subsample Analysis: Right-to-Work (RTW) Law
This table presents local linear regression results for subsamples of firms holding elections in RTW states (Panel A) and non-RTW states (Panel B) using a triangular kernel with a selection of bandwidths: the optimal bandwidth defined by Imbens and Kalyanaraman (2012) as well as 75% and 125% of the optimal
bandwidth following Campello et al. (2018). The variable of interest is Unionisation and the dependent variables are (a) change in ESO Incentives Granted
Per Employee within the (-1, 2) window in Columns (1)-(3) and (b) change in ESO Number Granted Per Employee within the (-1, 2) window in Columns
(4)-(6). P-values are displayed in parentheses with standard errors clustered by firm. ***, ** and * indicate significance at the 1%, 5% and 10% level,
respectively. All variables are defined in Appendix 1.
RTW States vs Non-RTW States
Panel A: RTW Subsample
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3)
(4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal Optimal
75%
Optimal
125%
Optimal
Unionisation 0.681 0.856* 0.565 0.195 0.278 0.319
(0.197) (0.084) (0.365) (0.852) (0.778) (0.760)
Panel B: Non-RTW Subsample
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3)
(4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal
Optimal
75%
Optimal
125%
Optimal
Unionisation 1.970** 2.042** 1.907*** 2.036** 1.745* 2.076**
(0.024) (0.037) (0.008) (0.042) (0.090) (0.022)
76
Table 5. Subsample Analysis: Labour Skills
This table presents local linear regression results separately for firms that rely on high-skilled labour (Panel A) and low-skilled labour (Panel B). We use the
sample median of the Labour Skill Index (LSI) (Ghaly et al. 2017) to partition our sample into a High-Skill and a Low-Skill group. The variable of interest is Unionisation and the dependent variables are (a) change in ESO Incentives Granted Per Employee within the (-1, 2) window in Columns (1)-(3) and (b)
change (-1,2) in ESO Number Granted Per Employee within the (-1, 2) window in Columns (4)-(6). P-values are displayed in parentheses with standard errors
clustered by firm. ***, ** and * indicate significance at the 1%, 5% and 10% level, respectively. All variables are defined in Appendix 1.
High-Skill vs Low-Skill Industries
Panel A: High-Skill Subsample
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3)
(4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal Optimal
75%
Optimal
125%
Optimal
Unionisation 0.483 0.861 0.291 0.530 0.468 0.428
(0.517) (0.134) (0.715)
(0.563) (0.620) (0.624)
Panel B: Low-Skill Subsample
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3)
(4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal
Optimal
75%
Optimal
125%
Optimal
Unionisation 2.070** 1.777 2.403*** 2.070*** 1.934*** 2.054***
(0.039) (0.107) (0.003)
(0.007) (0.006) (0.007)
77
Table 6. Robustness Test: R&D Expenditure as an Alternative Proxy for Labour Skill
This table presents local linear regression results separately for firms that rely on high-skilled (Panel A) and low-skilled (Panel B) labour. We use the three-year-average R&D expenditure as an alternative proxy for a firm’s reliance on skilled labour and partition our sample into high-R&D (High_RD=1) and low-
R&D (High_RD=0) firms based on whether the three-year-average R&D expenditure is above zero. The variable of interest is Unionisation and the
dependent variables are (a) change (-1.2) in ESO Incentives Granted Per Employee within the (-1, 2) window in Columns (1)-(3) and (b) change in ESO
Number Granted Per Employee within the (-1, 2) window in Columns (4)-(6). P-values are displayed in parentheses with standard errors clustered by firm.
***, ** and * indicate significance at the 1%, 5% and 10% level, respectively. All variables are defined in Appendix 1.
High-R&D vs Low-R&D Firms
Panel A: High R&D Subsample
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee)
(1) (2) (3) (4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal Optimal
75%
Optimal
125%
Optimal
Unionisation 1.356 1.378* 1.462 1.382 1.217 1.362
(0.133) (0.076) (0.123) (0.131) (0.195) (0.122)
Panel B: Low R&D Subsample
∆Log(Incentives Granted Per Employee) ∆Log(Number Granted Per Employee) (1) (2) (3) (4) (5) (6)
Bandwidth Optimal 75%
Optimal
125%
Optimal Optimal
75%
Optimal
125%
Optimal
Unionisation 2.565*** 2.433** 1.984*** 1.908** 2.310** 1.680**
(0.007) (0.026) (0.006) (0.043) (0.026) (0.040)
78
Table 7. Placebo Test: RDD using Artificial Treatment Assignment Thresholds
This table presents local linear regression results based on artificial thresholds for unionisation (40%, 45%, 55% and 60%) using optimal bandwidth defined
by Imbens and Kalyanaraman (2012). Following Campello et al (2018), we also use 75% and 125% of the optimal bandwidth as robustness checks. The
variable of interest is Unionisation and the dependent variables are (a) change in ESO Incentives Granted Per Employee within the (-1,2) window in Columns (1)-(5) and (b) change in ESO Number Granted Per Employee within the (-1,2) window in Columns (6)-(10). For comparison, results based on the true
threshold of 50% are also presented in Column (3) and Column (8). P-values are displayed in parentheses with standard errors clustered by firm. ***, ** and
* indicate significance at 1%, 5% and 10% level, respectively.
∆Log(Incentives Granted Per Employee)(-1,2)
∆Log(Number Granted Per Employee)(-1,2)
(1) (2) (3) (4) (5)
(6) (7) (8) (9) (10)
Artificial
Threshold 40% 45% 50% 55% 60% 40% 45% 50% 55% 60%
Optimal
Bandwidth
-1.122 -0.748 1.694*** -0.678 -1.242
-1.181 -0.170 1.511** -0.608 0.538
(0.109) (0.409) (0.004) (0.378) (0.148)
(0.162) (0.798) (0.025) (0.415) (0.348)
75% Optimal
Bandwidth
-1.095 -0.970 1.683*** -0.838 -1.685
-1.040 -0.446 1.362** -0.336 0.342
(0.122) (0.338) (0.009) (0.271) (0.128)
(0.236) (0.521) (0.047) (0.696) (0.581)
125% Optimal
Bandwidth
-0.978 -0.364 1.516*** -0.557 -0.900
-1.111 0.260 1.593** -0.549 0.490
(0.150) (0.656) (0.007) (0.463) (0.225) (0.167) (0.670) (0.014) (0.410) (0.352)
79
Figure 1. Time Trend of Union Elections
This figure describes the time-series variation in the occurrence and outcomes of union elections
from 1980 to 2011. The red solid line represents the median percentage of votes in favour of
unionisation (Vote Share for Union) in the elections in a given year; the blue dashed line
represents the total number of union elections (# Elections) held in a given year. Union election
data are collected from the National Labour Relations Board (NLRB).
80
Figure 2. RD Plots for the Unionisation Effect on ESO Incentives Granted
This figure illustrates the regression discontinuity plots using fitted linear and
quadratic functions. The horizontal axis represents the vote share for unionisation and the vertical axis represents the change over the window (-1, 2) in ESO Incentives Granted Per
Employee (Figures A-B) and ESO Number Granted Per Employee (Figures C-D). The dot
depicts the average outcome variables in each of the evenly-spaced bins, using the default
setting of rdplot command in Stata developed by Calonico et al. (2016).
(A) (B)
(C) (D)
81
Supplementary Appendix
Appendix 1. Definition of Variables
Variable Definition
Book-to-Market Book value of assets divided by sum of book value of liabilities and market value of equity
Cash Cash and short-term investments scaled by total assets
Dividend Dividend for common stock divided by earnings before interest and tax
ESO Dummy variable equal to one if (1) Log(Incentives Outstanding Per Employee)t-1 or (2)
(Incentives_Outstanding_Pct)t-1 is above the sample median, zero otherwise
ESO Incentives Granted Per Employee Sensitivity of the total value of stock options granted to non-executive employees during the fiscal year to a 1% change in stock price, divided by the number of employees
Log(Incentives Granted Per Employee) Logarithm of ESO Incentives Granted Per Employee
∆Log(Incentives Granted Per Employee) Log(Incentives Granted Per Employee)t+2 -Log(Incentives Granted Per Employee)t-1, t being the union election year
ESO Incentives Outstanding Per Employee Sensitivity of the total value of outstanding stock options held by non-executive employees during the
fiscal year to a 1% change in stock price, divided by the number of employees ESO Number Granted Per Employee Number of employee stock options granted per employee
Log(Number Granted Per Employee) Logarithm of ESO Number Granted Per Employee
∆Log(Number Granted Per Employee) Log(Number Granted Per Employee)t+2 -Log(Number Granted Per Employee)t-1, t being the union
election year
Log(ESO Incentives Outstanding Per Employee) Logarithm of ESO Incentives Outstanding Per Employee
Incentives_Outstanding_Pct Percentage of all outstanding stock option incentives held by non-executive employees during the fiscal
year
High_RD Dummy variable equal to one if R&D Expense is positive, zero otherwise
High_Skill Dummy variable equal to one if LSI is above the sample median, zero otherwise
Income Earnings before interest and tax
Interest Burden Interest expense scaled by operating income before depreciation
(continued on next page)
82
Leverage Book value of long-term debt divided by total market value of the firm
LSI Labour skill index measuring the level of reliance on skilled workers for each SIC 3-digit industry as in Ghaly et al. (2017)
Market-to-Book Market value over book value of total assets
Post Dummy variable equal to 1 if fiscal year is later than election year, zero otherwise
ProfitMargin Net income divided by total revenue
R&D Expense Three-year average of research and development expense scaled by total assets
ROE Net income divided by total equity
RTW Dummy variable equal to 1 if union election is held in a state with right-to-work legislation, zero otherwise
Total Sales Logarithm of total revenue.
Sales Growth (Total Salest-Total Salest-1)/Total Salest-1
StockReturn Percentage return on the firm’s stock in the fiscal year
Strike Dummy Dummy variable equal to 1 if there is a strike in fiscal year t, zero otherwise
Strike Risk Ordinal variable equal to 1 if there is no strike, 2 if there is one strike and 3 if there are multiple strikes in fiscal year t
Total Assets Logarithm of total assets
Total Employees Logarithm of the number of employees
Treated Dummy variable equal to 1 if vote share in favour of unionisation>50%, zero otherwise
Unionisation Dummy variable equal to 1 if vote share in favour of unionisation>50%, zero otherwise
Vote For Number of votes in favour of unionisation
Vote Share Number of votes for unionisation divided by total number of votes
Vote Total Total number of votes in the union election
WorkCap Working capital scaled by total assets
ZScore 1.2(working capital/total assets) +1.4(retained earnings/total assets) + 3.3(EBIT/total assets) + 0.6(market
value of equity/book value of total liabilities) + (sales/total assets)
83
Appendix 2. Industry Distribution
SIC2 Freq. Percent Industry
13 1 0.31 Oil and Gas Extraction
14 1 0.31 Mining and Quarrying of Nonmetallic Minerals, except Fuels
16 5 1.55 Heavy Construction other than Building Construction Contractors
20 26 8.07 Food and Kindred Products
23 2 0.62 Apparel and other Finished Products Made from Fabrics and Similar Materials
24 4 1.24 Lumber and Wood Products, except Furniture
25 7 2.17 Furniture and Fixtures
26 17 5.28 Paper and Allied Products
27 3 0.93 Printing, Publishing, and Allied Industries
28 20 6.21 Chemicals and Allied Products
29 7 2.17 Petroleum Refining and Related Industries
30 7 2.17 Rubber and Miscellaneous Plastics Products
32 3 0.93 Stone, Clay, Glass, and Concrete Products
33 13 3.42 Primary Metal Industries
34 2 0.62 Fabricated Metal Products, except Machinery and Transportation Equipment
35 7 2.17 Industrial and Commercial Machinery and Computer Equipment
36 5 1.55 Electronic and other Electrical Equipment and Components, except Computer Equipment
37 15 4.66 Transportation Equipment
38 13 4.04 Measuring, Analyzing, and Controlling Instruments; Photographic, Medical and Optical Goods; Watches and Clocks
39 1 0.31 Miscellaneous Manufacturing Industries
41 10 3.11 Local and Suburban Transit and Interurban Highway Passenger Transportation
42 2 0.62 Motor Freight Transportation and Warehousing
48 13 4.04 Communications
49 26 8.07 Electric, Gas and Sanitary Services
(continued on next page)
84
50 7 2.17 Wholesale Trade - Durable Goods
51 12 3.73 Wholesale Trade - Nondurable Goods
52 5 1.55 Building Materials, Hardware, Garden Supply, and Mobile Home Dealers
53 12 3.73 General Merchandise Stores
54 5 1.55 Food Stores
56 2 0.62 Apparel and Accessory Stores
57 3 0.93 Home Furniture, Furnishings, and Equipment Stores
59 12 3.73 Miscellaneous Retail
63 1 0.31 Insurance Carriers
67 11 3.42 Holding and other Investment Offices
70 1 0.31 Hotels, Rooming Houses, Camps, and other Lodging Places
73 10 3.11 Business Services
75 5 1.55 Automotive Repair, Services, and Parking
79 5 1.55 Amusement and Recreation Services
80 18 5.59 Health Services
82 1 0.31 Educational Services
83 2 0.62 Social Services
87 1 0.31 Engineering, Accounting, Research, Management, and Related Services
99 1 0.31 Non-classifiable Establishments
Total 324 100
85
Appendix 3. Firm Characteristics in Unmatched and Matched Samples
Unmatched (U) Mean T-test
Variable Matched (M) Treated Control t p>|t|
ΔCash_lag U 0.007 0.009 -0.25 0.799 M 0.014 0.016 -0.20 0.841
ΔLeverage_lag U -0.005 -0.165 0.77 0.443 M -0.030 -0.031 0.01 0.995
ΔDividend_lag U -0.019 -0.103 0.56 0.577 M -0.023 0.005 -0.88 0.379
ΔIncome_lag U -0.001 0.001 -0.26 0.799 M -0.001 -0.001 0.00 0.997
ΔWorkCap_lag U -0.007 -0.001 -0.65 0.514 M 0.005 0.011 -0.65 0.519
ΔZScore_lag U -0.276 -0.138 -0.69 0.489 M -0.134 0.122 -2.54 0.012
ΔMarket-to-Book_lag U -0.062 -0.086 0.49 0.622 M -0.071 -0.006 -1.59 0.114
Sample Mean Bias Median Bias
Unmatched 7.1 8.1
Matched 6.8 3.0
86
Appendix 4. Firm Adjustment of ESO Grants after Unionisation: Global Polynomial Regressions
This table presents global polynomial regression results with 2nd, 3rd and 4th orders of polynomial. The dependent variables are (a) ∆Log(Incentives Granted
Per Employee)(-1,2) in Columns (1)-(3) and (b) ∆Log(Number Granted Per Employee)(-1,2) in Columns (4)-(6) as the dependent variables. The variable of
interest is Unionisation. In Panel A, we run global polynomial regressions using a triangular kernel. As a further robustness test, we repeat the analyses using a rectangular kernel in Panel B. P-values are displayed in parentheses with standard errors clustered by firm. ***, ** and * indicate significance at the 1%, 5%
and 10% level, respectively. All variables are defined in Appendix 1.
Panel A: Triangular Kernel
∆Log(Incentives Granted Per Employee)(-1,2) ∆Log(Number Granted Per Employee)(-1,2)
(1) (2) (3) (4) (5) (6)
Polynomial
Order 2nd 3rd 4th 2nd 3rd 4th
Unionisation 1.844** 1.945** 1.654** 1.487* 1.599** 1.469*
(0.01) (0.01) (0.02) (0.05) (0.03) (0.05)
N 137 137 137 132 132 132
Panel B: Rectangular Kernel
∆Log(Incentives Granted Per Employee)(-1,2) ∆Log(Number Granted Per Employee)(-1,2)
(1) (2) (3) (4) (5) (6)
Polynomial
Order 2nd 3rd 4th 2nd 3rd 4th
Unionisation 1.678** 1.941** 1.963** 1.433* 1.348* 1.821**
(0.02) (0.02) (0.02) (0.09) (0.10) (0.02)
N 137 137 137 132 132 132
87
Appendix 5. Validity Tests for Regression Discontinuity Design (RDD)
1. Continuity of Forcing Variable
Our RDD methodology relies heavily on the continuity assumption of the forcing
variable, that is, the vote share (in favour of unionisation) in a union election. A
discontinuous distribution of this vote share around the 50 percent cutoff point is
regarded as a sign of manipulation of election results, which would fundamentally
violate the ‘locally randomised treatment’ setting and ultimately invalidate our RDD
results. Following recent studies (Bradley et al. 2017; Campello et al. 2018), we conduct
two types of diagnostic validity tests on this critical assumption for all union elections
in our sample period (2004-2011) using both graphical verification and formal statistical
manipulation tests.
A direct and intuitive starting point for verifying the continuity assumption is through a
graphical inspection of the distribution of the vote share. If there was systematic
manipulation within the small window around the cutoff point, the distribution of the
vote share would be discontinuous, exhibiting either a jump or a drop as the vote share
exceeded the 50 percent threshold. Figure A1 plots a histogram of the vote share
distribution of all the union elections in our sample period (2004-2011)26 from the
National Labour Relations Board (NLRB) database, across 20 equally spaced bins. As
the graph indicates, the vote share distribution is generally smooth and there is no clear
discontinuity around the cutoff. Thus, there is no compelling evidence of systematic and
precise manipulation of the vote share in union elections organised by the NLRB.
***Insert Figure A1 here***
Since the visual inspection of the smoothness of the vote share around the threshold is
very subjective and open to personal interpretation, we statistically test whether there is
a systematic difference in the density of the vote share within a close vicinity of the 50
percent threshold in our sample. There are two main manipulation tests in the existing
26 Due to the limited number of observations in our own sample, we use all the union election results from
NLRB between 2004 and 2011to plot the distribution of vote share and check whether the union elections,
as regulated by NLRB, are subject to systematic manipulations that would ‘alter’ the treatment
assignment. We believe using all the election results and thus a larger sample size allows us to inspect the
vote share distribution in a meaningful way at a systematic level, given the larger number of observations
around the 50% threshold. However, in the subsequent formal manipulation tests (Figure 2A and Table
1A), we use the 324 union elections in our sample to test whether the ‘continuity assumption’ holds in our
sample.
88
literature: the McCrary (2008) test and the Cattaneo et al. (2016) test. Essentially, both
tests estimate the density of the vote share around the 50 percent threshold and assess
whether there is any discontinuity in the density. While the McCrary (2008) test is
designed to detect manipulation of the forcing variable based on pre-binned data, the
Cattaneo et al. (2016) test avoids pre-binning the data, removing exposure to the risk of
additional arbitrary bias from that artificial data modification. For robustness, we
conduct both tests to ensure the ‘continuity assumption’ is satisfied. Figure A2 presents
the density of the vote share under the McCrary test. The insignificant t-statistic of
0.884 suggests there is no sign of manipulation across the 50 percent threshold in our
sample.
***Insert Figure A2 here***
Similarly, there is no statistically significant evidence of a discontinuity in the vote
share around the 50 percent mark based on the Cattaneo et al. (2016) test, reported in
Table A1. Consistent with prior union election literature (DiNardo and Lee 2004; Lee
and Mas 2012; He et al. 2016a; Qiu and Shen 2017; Campello et al. 2018), we conclude
that our forcing variable, Vote Share, is continuous and there is no evidence to suggest
there is precise manipulation around the 50 percent cutoff point in the union elections
within our sample.
***Insert Table A1 here***
2. Covariate Balance Test
We also test the continuity of the firm characteristics prior to the union election event,
following He et al. (2016) and Campello et al. (2018). This is also known as the
covariate balance test, and basically examines whether the predetermined firm
characteristics are continuous around the 50 percent threshold prior to union elections
(we measure them one year prior). Crucially, if there were a discontinuity, i.e., a
systematic diffference in one of those observable firm characteristics around the cutoff,
the treatment effect we observe could not reliably be attributed to unionisation, as it
might be a result of the change in that particular firm covariate before the union election.
A discontinuity in firm characteristics would also suggest that the treated firms on the
right side of the cutoff point, i.e., ‘marginal winners’, were significantly different from
those on the left side of the cutoff, i.e., ‘marginal losers’. In other words, the ‘marginal
89
winners’ and ‘marginal losers’ would not be similar in terms of pre-treatment firm
characteristics, making the ‘marginal losers’ (control group) weak counterfactuals for
the ‘marginal winners’ (treated group). On the other hand, if the continuity assumption
on the predetermined firm characteristics were satisfied, it would indicate that firms just
above and below the 50 percent threshold exhibited similar firm characteristics, such as
size, profitability and growth opportunities, before the election. Therefore, any
differences observed in the outcome variable, the change in Incentives Granted Per
Employee, would be attributable to the treatment effect of unionisation.
Table A2 reports the results of the covariate balance test performed using local linear
regressions. The insignificant results on a range of firm characteristics confirm that the
treated and control firms are similar prior to the union elections. More importantly, the
insignificant results for the ESO variables in Panel B suggest that there is no significant
difference in the ESO proxies, including our key dependent variable, Incentives Granted
Per Employee, between the treated and control firms prior to treatment.
***Insert Table A2 here***
To complement the results reported in Table A2, Figure A3 visually demonstrates the
continuous distributions of a selection of predetermined firm characteristics around the
50 percent cutoff, satisfying the continuity assumption regarding pre-election firm
covariates. We conclude that the unionisation effect on ESO incentives granted to rank-
and-file employees we report in this study is not due to confounding factors, since there
are no ex-ante differences in a range of firm characteristics, nor in the ESO incentive
levels, between the treated and control groups.
***Insert Figure A3 here***
90
Table A1. Manipulation Test
The table below presents the results of the density test developed by Cattaneo et al. (2016), performed to validate the underlying continuity assumption of the
forcing variable, i.e., Vote Share. The null hypothesis is that the forcing variable is continuous, indicating no precise manipulation of the vote share in union
elections. Panel A tests the continuity of the vote share for our sample under 2nd, 3rd and 4th order polynomials using a triangular kernel function based on a
combination of bandwidth choices: (1) mean squared error (MSE) of sum of densities, (2) MSE of difference in densities, (3) MSE of each density defined by Cattaneo et al. (2016). Panel B tests the continuity of the vote share using alternative bandwidths at 5%, 10% and 15% using the 2nd order polynomial
function. T-statistics are robust-adjusted and bias-corrected with p-values displayed in parentheses. ***, ** and * indicate significance at the 1%, 5% and
10% level, respectively.
Panel A: Different Polynomial Orders using Combined Bandwidth
Order 2nd 3rd 4th
Vote Share -0.315 -0.514 -1.265 (0.752) (0.608) (0.206)
N 324 324 324
Panel B: Alternative Bandwidths
Bandwidth 5% 10% 15%
Vote Share -0.477 -1.385 0.334
(0.634) (0.166) (0.738)
Order 2nd 2nd 2nd
N 324 324 324
91
Table A2. Balance Test of Firm Characteristics in the Pre-Election Year
This table presents local linear regression results under a triangular kernel function based on the optimal bandwidth from Imbens and Kalyanaraman (2012), performed to test the continuity of firm characteristics prior to the union election events. In Panel A, the dependent variables are a selection of firm
characteristics for the pre-election year (i.e., the year prior to the union election year). In Panel B, the dependent variables are the pre-election levels of ESO-
related variables. All variables are defined in Appendix 1.
Panel A: Continuity of Firm Characteristics in the Pre-election Year
Variable Coefficient Z-Statistic
Total Sales -0.243 -0.61 Total Assets -0.544 -1.51
Leverage 0.029 0.49
Interest Burden -0.017 -0.30
ROE -0.032 -0.97 Profit Margin 0.045 0.56
Book-to-Market -0.150 -1.23
Sales Growth -0.031 -0.60 R&D Expense 0.003 0.67
Stock Return 0.032 0.18
Dividend 0.074 0.91
Panel B: Continuity of ESO-related Variables in the Pre-election Year
Variable Coefficient Z-Statistic
Log(Incentives Granted Per Employee) -0.184 -0.43 Log(Number Granted Per Employee) 0.249 0.32
Log(Incentives Outstanding Per Employee) -0.270 -0.42
Incentives_Outstanding_Pct 0.048 0.47
92
Figure A1. Distribution of Vote Share
This figure presents a histogram of the distribution of the vote share across 20 equally spaced
bins. Union election results are obtained from the National Labour Relations Board (NLRB) for
the period 2004-2011.
93
Figure A2. Density Test
This figure plots the density of the union vote share (i.e., the percentage of votes in an election
in favour of unionisation) for our union election sample following McCrary (2008). The x-axis
represents the vote share. The dots represent the density estimate for each chosen bin and the bold line is the fitted density function of union vote shares with a surrounding 95% confidence
interval.
(T=0.884)
94
Figure A3. Continuity of Firm Characteristics
These figures show the continuity of firm characteristics in the pre-election year. The horizontal
axis represents the Vote Share in favour of unionisation and the vertical axis represents firm
characteristics for a given Vote Share. Figures (A) through (F) show the distributions for Total
Sales, Total Assets, Book-to-Market ratio, Leverage, Log(Incentives Granted Per Employee) and Log(Number Granted Per Employee), respectively. The dots represent the sample-average
firm characteristics within the bin. The black lines represent fitted quadratic polynomial
functions of these characteristics, fitted over each Vote Share bin. The grey lines represent the
95% confidence level.
(A) (B)
(C) (D)
(E) (F)
95
Chapter 3
Does Corporate Social Responsibility Spending Affect Strike Risk?
Evidence from Union Elections
Abstract
This paper investigates the effect of corporate social responsibility (CSR) spending on
labour unions’ propensity to initiate strikes. Given limited financial resources, we posit
that firms’ inability to satisfy the demands from all of their stakeholders leads to
resource competition among those stakeholders. By exploiting the unique setting of
union elections in U.S. firms as plausibly exogenous shocks to labour power, we
employ a triple-differences specification and find that firms with high levels of (non-
employee) CSR spending are exposed to a significantly higher risk of union strikes. We
interpret this finding as evidence consistent with CSR spending intensifying resource
competition between employees and other stakeholders. We also document evidence
that firms strategically curtail CSR expenditure in non-employee dimensions in
response to unionisation in order to mitigate the increased strike risk. Such downward
adjustment in CSR spending, however, is less pronounced in financially constrained
firms, firms in “sin” industries and firms facing high levels of product market
competition, due to their strong incentives to signal quality through CSR spending.
Overall, our study sheds light on the inter-stakeholder relationship through the lens of a
powerful stakeholder, i.e., organised labour, and has important implications for
managers facing severe labour risk.
96
3.1 Introduction
We examine the impact of corporate social responsibility (CSR) spending on labour
strike risk. In the past decade, CSR has attracted enormous interest from businesses,
government agencies, non-governmental organisations (NGOs) and academics across
the globe, evolving from what initially appeared to be a controversial managerial
decision (Friedman 1970) into a commonplace practice that firms voluntarily commit to
on a regular basis. Notably, Fortune 500 companies alone spend more than 15 billion
dollars per year and a considerable amount of time on numerous CSR initiatives
(Financial Times 2014), while more than 90% of the 250 largest firms in the world issue
standalone CSR reports annually (KPMG 2015). Despite the growing emphasis on CSR
issues, due to limited resources, one would expect managers to have to exercise
personal discretion in selecting CSR projects. We argue that such managerial discretion
on CSR spending could lead to resource competition amongst different stakeholders,
which would likely cause tensions within firms.
As powerful stakeholders in companies, labour unions use their collective bargaining
power to push for higher wages and to influence a wide range of corporate decisions
(Klasa et al. 2009; Matsa 2010; Chen et al. 2012; Chyz et al. 2013; Chung et al. 2016;
Bradley et al. 2017; Huang et al. 2017; Campello et al. 2018). While several of the
concerns of labour unions, such as employee benefits, working conditions and
workplace equality, generally fit under the CSR framework (Preuss et al. 2006; Compa
2008; Preuss 2008; Dawkins 2010; Sobczak and Havard 2015; Dawkins 2016), to date,
labour unions’ attitude towards the unprecedented CSR spending, particularly that on
other stakeholders (e.g., the environment and society), is underexplored and thus
remains unclear. In this paper, from a multi-stakeholder perspective, we aim to fill the
97
gap by examining the effect of CSR spending27 on the behaviour of organised labour, as
a key non-financial stakeholder.
Prior literature has established multiple rationales behind CSR spending and the
predominant view suggests that firms tend to use CSR as a device for enhancing their
reputation and brand image by showing commitment to the well-being of stakeholders
such as employees and customers (Barnett 2007; Renneboog et al. 2008; Harrison et al.
2010). Socially responsible firms are therefore expected to benefit from higher customer
loyalty, greater employee satisfaction and harmonious relationships with various
stakeholders, all of which will give firms a competitive advantage that will help them
create value for their shareholders in the long run (Porter and Kramer 2006; Choi and
Wang 2009; Carroll and Shabana 2010; Edmans 2012; Flammer 2015a; Liang and
Renneboog 2017a). However, a number of studies have raised suspicions about the
underlying motivations behind CSR, arguing that CSR spending is a reflection of the
agency problem whereby managers expropriate firm resources to fulfil their individual
social commitments and enhance their personal reputations at the cost of shareholders’
wealth (Jensen 2001; Cheng et al. 2013; Krüger 2015; Masulis and Reza 2015;
Davidson et al. 2016).
Unlike other stakeholders such as customers and the community, employees constitute a
powerful primary stakeholder that not only exists internally within the firm but also has
a significant long-term contractual claim in the form of wages and pensions (Sobczak
and Havard 2015; Helmig et al. 2016; Campello et al. 2018). Increasing employees’
bargaining power, labour unions serve as collective bargaining units that safeguard the
27 As explained, employee relations are generally considered one of the dimensions of the CSR framework. Therefore, it is intuitive that labour unions will welcome and promote more employee-related
CSR. For the purpose of our study, we differentiate employee-related CSR from CSR towards other
stakeholders, and mainly focus on unions’ stance towards CSR spending on non-employee stakeholders
such as communities and the environment.
98
interests of the employees, particularly in contract negotiations with their employers
(Freeman and Medoff 1979; DiNardo and Lee 2004; Campello et al. 2018). To achieve
their wage agenda, labour unions engage in a range of collective bargaining tactics, with
strikes being the most powerful (Ashenfelter and Johnson 1969; Myers and Saretto
2016). Given companies’ increasing investment and efforts in CSR activities aimed at
promoting the interests of stakeholders including employees, it is surprising to note that
employees still have to fight for fairer pay and basic necessities such as healthcare and
pension provisions, as has been seen in several high-profile labour strikes within the
recent revival of the labour union movement28. Noticeably, the strikers have included
employees of some of the largest (and presumably richest) companies in the world29.
Labour strikes are detrimental to employers and can have far-reaching yet serious
consequences for the wider economy and society. They not only directly cause
significant financial damage30 to the employers (Becker and Olson 1986; Schmidt and
Berri 2004) but also indirectly affect businesses along the supply chain (McHugh 1991;
DiNardo and Hallock 2002). Since labour unionisation is more prevalent in strategically
important industries such as manufacturing and transportation (Chen et al. 2011), large-
scale strikes in such industries may be more likely, and if they do occur could cause
severe disruption to the economy and uncertainty to society as a whole.
Prior literature offers mixed guidance regarding the relation between CSR expenditure
and union behaviour. On the one hand, drawing from stakeholder theory (Freeman
28 In 2016, 1.54 million working days were left idle as a result of 15 mass strikes involving more than
99,000 workers in the United States (Bureau of Labour Statistics 2017).
29 In the past decade, many multinational corporations, including several household names, have suffered
labour strikes, such as AT&T, Amazon, British Airways, Boeing, BP, McDonald’s, General Motors,
Verizon, Walmart and others.
30 In 2008, a 58-day strike by 27,000 machinists at Boeing, the largest aircraft manufacturer in the world,
caused $100 million of losses per day in terms of deferred revenue, and $2 billion in lost profits. The
company’s share price also plummeted by 56% to a five-year low during the strike period (Reuters 2008).
99
1984), by devoting resources to CSR projects, a company will earn a good reputation as
a “stakeholder-oriented” corporate citizen and foster harmonious relationships with its
various stakeholders, including employees (Barnett 2007; Cheng et al. 2014; Cuypers et
al. 2016; Lins et al. 2017). More specifically in terms of employee relations, a number
of studies suggest that, by being “socially responsible”, firms can attract and retain
talent (Albinger and Freeman 2000; Greening and Turban 2000), boost employee
morale (Balakrishnan et al. 2011; Flammer and Luo 2017), enhance job satisfaction
(Valentine and Fleischman 2008; Edmans 2012) and improve productivity and
innovation (Flammer 2015a; Flammer and Kacperczyk 2016). Therefore, a high-CSR
firm may benefit from better relationships with its employees and higher job satisfaction
among them, leading to more cooperative union behaviour.
On the other hand, from a multi-stakeholder perspective, a high level of non-employee
CSR expenditure would imply that the firm prioritises external stakeholders such as
environmentalists over its employees, causing serious tension between employees and
managers and escalating the conflict of interests amongst different stakeholders. When
competing for the limited firm resources against other stakeholders, labour unions are
more likely to engage in extreme collective-bargaining activities such as strikes to put
more pressure on managers to shift more resources to employees. Furthermore, labour
unions may perceive CSR spending as a managerial misappropriation of corporate
resources that should be invested in the employees (Moser and Martin 2012; Krüger
2015; Masulis and Reza 2015). Therefore, investing in non-employee CSR projects may
be perceived as a misuse of financial resources, provoking labour unions into extracting
rent through collective bargaining (Klasa et al. 2009; Barnea and Rubin 2010; Myers
and Saretto 2016).
In this paper, we empirically study this contentious question by exploiting the
100
exogenous variation in labour power resulting from union elections in the United States
between 2002 and 201131. While labour unions have historically been more prominent
in other parts of the world, such as Europe, the granularity of the union election data for
U.S. firms offers us an ideal setting in which to apply quasi-experimental identification
strategies that can help us draw strong causal inferences.
Employing a triple-differences empirical strategy, we find that firms with high levels of
non-employee CSR spending on environmental or societal dimensions are exposed to a
significantly higher post-unionisation strike risk than their low-CSR counterparts, in
line with the “resource competition” conjecture. In contrast, the unionisation effect on
the strike risk is significantly mitigated in the presence of high levels of CSR spending
on employee-related issues (e.g., employment quality). We argue that high levels of
CSR expenditure in non-employee dimensions can exacerbate the conflicts of interests
between employees and other stakeholders and intensify the resource competition
amongst the various stakeholders. Consequently, given the limited financial resources
firms have, labour unions are more likely to resort to extreme bargaining strategies such
as strikes to ensure the employees have priority over other stakeholders.
We further test whether firms strategically adjust their CSR expenditure levels to
mitigate the strike risk in response to unionisation. Consistent with prior literature
suggesting that firms make strategic decisions to improve their bargaining position
against labour unions, we find that firms significantly cut CSR spending in non-
employee dimensions to mitigate the strike risk following the event of unionisation.
However, such adjustments are less pronounced in financially constrained firms, firms
operating in “sin” industries and firms facing high levels of product market competition,
31 This is the period of overlap between our major data sources, the union election database that runs from
1980 to 2011 and the ASSET4 ESG database that covers 2002 to 2016.
101
due to their stronger incentives to signal quality through CSR spending.
Our study contributes to the literature in multiple ways. First, unlike previous CSR
studies which tend to treat stakeholders as a homogeneous group and predominantly
focus on the stakeholder-shareholder relationship, to our best knowledge, we provide
the first empirical evidence on the interplay amongst stakeholders and reveal an
unintended consequence of CSR spending, i.e., the resource competition among
different stakeholders. Second, by showing that firms strategically adjust their CSR
spending to mitigate the strike risk in response to unionisation, our study contributes to
the emerging literature on strategic corporate decisions in the context of strong labour
power (Klasa et al. 2009; Matsa 2010; Chung et al. 2016; Chino 2016; Huang et al.
2017) and adds to the ongoing debate on human capital management (Ghaly et al. 2015;
Chen et al. 2016; Fauver et al. 2018). Third, our study enriches the understanding of
unions’ behaviour and their decision to pursue an extreme collective-bargaining tactic,
namely, labour strikes. Our study has important managerial implications, suggesting
that firms should regularly review their relationships with various stakeholders and take
a balanced approach to stakeholder management. Our findings have implications not
only for the U.S. context but also for other jurisdictions, such as Europe, where unions
have historically been more active and thus there is a greater need for risk management
against strikes. Overall, our paper sheds light on the inter-stakeholder relationship
through the lens of a powerful stakeholder of a business, organised labour.
The remainder of the paper is organised as follows. Section 2 reviews the extant
literature on labour unions and CSR, followed by the development of our research
hypotheses. Section 3 describes the data collection and sampling processes as well as
our empirical design. Section 4 presents our main empirical results. Section 5
summarises the empirical findings and contributions of our study.
102
3.2 Literature Review and Hypothesis Development
3.2.1 Literature Review
3.2.1.1 Union Strikes
Labour unions are powerful primary stakeholders within businesses, influencing various
corporate decisions and ultimately firm performance (Klasa et al. 2009; Matsa 2010;
Chen et al. 2012; Chyz et al. 2013; Chung et al. 2016; Bradley et al. 2017; Huang et al.
2017; Campello et al. 2018). Prior literature documents a largely negative union effect
on firm performance and shareholders’ value (Clark 1984; Ruback and Zimmerman
1984; Lee and Mas 2012). The negative impact of labour unions also extends to
debtholders, with Campello et al. (2018) demonstrating that labour unionisation is
detrimental to bond values. Given the negative union effect on business operations and
the wealth of shareholders and debtholders, unionised firms have to pay a price
premium in order to access capital from both the equity and debt markets (Chen et al.
2011; Cheng 2017). In addition, Bradley et al. (2017) suggest that labour unions inhibit
firm innovation, by presenting the evidence that unionisation leads to a significant
reduction in both patent quality and quantity. Despite the overwhelmingly negative
view of labour unions, their scrutiny of management can improve corporate governance.
For example, they can significantly curb executive compensation (Huang et al. 2017)
and deter managers from engaging in tax-sheltering activities (Chyz et al. 2013).
To undermine union power, firms proactively make a range of strategic corporate
decisions. Primarily, previous literature shows that firms strategically adjust their
capital structures, in ways such as reducing cash holding (Klasa et al. 2009) and
increasing leverage (Bronars and Deere 1991; Matsa 2010) to shelter financial resources.
Secondly, several recent studies have shown that firms engage in “downward”
impression management in the presence of labour unions (Bova 2013; Chung et al.
103
2016). Specifically, Bova (2013) reveals that unionised firms have a higher propensity
to narrowly miss analysts’ earnings forecasts so as to manage unions’ expectations. In a
similar vein, Chung et al. (2016) present empirical evidence that unionised firms
strategically withhold good news during labour negotiations to undermine unions’
desire to extract economic rent.
A labour union’s power lies in its ability to initiate large-scale labour strikes, which are
evidently disruptive to firms’ operations and damaging to shareholders’ wealth
(Ashenfelter and Johnson 1969; Myers and Saretto 2016). To fulfil its wage agenda, a
labour union will often employ a wide range of collective-bargaining tools, including
strikes, to put more pressure on the employer (Tracy 1986; Cramton and Tracy 1994).
Inevitably, unionised firms are exposed to considerably higher strike risk, especially
during contract negotiations. However, whether to pursue this extreme bargaining
strategy can be viewed as a rational economic decision made by a labour union based on
a cost-benefit analysis (Ashenfelter and Johnson 1969). Therefore, a labour union is
more likely to call a strike when the perceived benefits, such as a potential pay rise, are
much higher than the perceived costs, such as the loss of wages during the strike period,
of engaging in the strike. To mitigate strike risk, firms make strategic decisions aimed at
reducing the perceived benefit of a labour strike, by sheltering financial resources away
from the organised labour (Bronars and Deere 1991; Klasa et al. 2009; Matsa 2010;
Myers and Saretto 2016).
3.2.1.2 CSR Spending
While the literature on CSR originally stems from the business ethics and sociology
disciplines, CSR has already attracted unprecedented interest from accounting and
finance scholars, and the research area has seen substantial growth in the last decade
104
(Huang and Watson 2015). So far, the literature has predominantly concentrated on the
drivers and financial outcomes of CSR activities.
The overarching theories behind CSR investment are (1) shareholder value
enhancement and (2) managerial utility maximisation (Krüger 2015; Liang and
Renneboog 2017b). Consistent with the value creation argument, it is widely agreed that
firms engage in CSR activities to improve their corporate reputation and enhance their
brand image (Fombrun and Shanley 1990; Porter and Kramer 2006; Renneboog et al.
2008). Others suggest that firms treat CSR as part of their advertising campaigns, aimed
at enhancing customer awareness and loyalty (Fisman et al. 2006; Pivato et al. 2008;
Servaes and Tamayo 2013). CSR is also used to signal a firm’s quality as a “socially
responsible” employer, helping it to attract and retain talent (Albinger and Freeman
2000; Greening and Turban 2000; Edmans 2012; Flammer and Kacperczyk 2019). In
light of the growing popularity of Socially Responsible Investment (SRI) funds, firms
are expected to behave in a more socially responsible manner in order to appeal to this
particular type of investor (Renneboog et al. 2008; Crane et al. 2009; Edmans 2011).
Additionally, Husted et al. (2016) and Cao et al. (2017) show that CSR engagement can
be triggered by firms in close geographic proximity and by peer pressure from
competitors to a firm. Irrespective of the motives for and beneficiaries of CSR
initiatives, the competitive advantage and social capital generated will ultimately lead to
the creation of shareholder wealth (Porter and Kramer 2006; Renneboog et al. 2008).
In contrast, drawing from agency theory, many scholars are sceptical about the
underlying incentives for CSR expenditure, arguing that CSR is a reflection of agency
problems (Friedman 1970; Jensen 2001; Bénabou and Tirole 2010). Friedman (1970)
fundamentally rejects the idea of CSR and maintains that the only responsibility of
corporate executives is to maximise profit for the shareholders, rather than spending
105
shareholders’ money to fulfil individual commitments to social values. Jensen (2001)
supports this view by pointing out that stakeholder theory inevitably induces agency
costs because the performance measures would become unclear if managers strive to
satisfy all stakeholders instead of just the shareholders. A number of empirical studies
document evidence in line with the managerial self-serving view proposed by Friedman
(1970) and Jensen (2001). Specifically, both Barnea and Rubin (2010) and Cheng et al.
(2013) find that managerial ownership is negatively related to CSR investment.
Meanwhile, monitoring from internal governance and external creditors curtails CSR
investment (Brown et al. 2006; Cheng et al. 2013). Furthermore, Marquis and Lee (2013)
document that CEOs with shorter tenures make considerably higher corporate donations,
suggesting that managers misuse corporate resources to enhance their personal
reputations and advance their careers. Despite being voluntary managerial decisions,
Tang and Tang (2018) argue that CSR decisions are also partly shaped by the CSR
orientations of the various stakeholders with which firms interact.
Another strand of the CSR literature has focused on the effect of CSR investment on
financial performance and firm value, reporting mixed results. An overwhelming
majority of the studies document a positive impact of CSR investment on financial
performance, supporting the value-enhancing theory (Renneboog et al. 2008; Artiach et
al. 2010; Schreck 2011; Servaes and Tamayo 2013; Eccles et al. 2014; Flammer 2015a;
Cuypers et al. 2016; Liang and Renneboog 2017a). Based on a meta-analysis of 162
articles, Margolis et al. (2009) conclude that the relation between CSR performance and
financial performance is weakly positive. Furthermore, Choi and Wang (2009) reveal
that good stakeholder relations contribute to the persistence of superior financial
performance and enable quicker recoveries from poor performance. Meanwhile, Lins et
al. (2017) show that CSR is particularly conducive to better performance in the context
106
of financial crises, as high-CSR firms enjoy more social capital and are trusted more by
their stakeholders under adverse economic circumstances. In addition, Cai et al. (2012)
find that CSR in “sin” industries has a positive impact on firm value. In contrast, Baird
et al. (2012) and Krüger (2015) report a negative CSR effect on firm performance and a
negative market reaction to positive CSR news, implying that CSR activities reflect
agency problems. Interestingly, McWilliams and Siegel (2000) discover a neutral
relation between CSR and firm performance, and attribute the discrepancies across
various CSR studies to different data sources, measurement errors, and
misspecifications in empirical designs.
In addition to firm performance, CSR has important implications for many other aspects
of business. A number of studies suggest that CSR reduces the cost of capital, thereby
improving firms’ access to finance (Dhaliwal et al. 2011; Goss and Roberts 2011; El
Ghoul et al. 2011; Cheng et al. 2014; Dhaliwal et al. 2014). Moreover, it is beneficial
for innovation, with Flammer and Kacperczyk (2016) providing evidence that
stakeholder orientation catalyses firm innovation by enhancing employees’ innovative
productivity. Admittedly, as pointed out earlier, if CSR spending is perceived by
employees as a managerial misappropriation of resources, it could hinder their
innovation. More recently, Flammer (2018) finds that firms with good CSR ratings are
more likely to win government procurement contracts, implying that firms use CSR as a
quality-signalling device to differentiate themselves from their competitors.
Previous studies have also indicated that CSR plays an important role in improving
corporate governance. Empirical evidence reveals that socially responsible firms are
less likely to engage in earnings management (Kim et al. 2012) and tax avoidance
activities (Hoi et al. 2013). Furthermore, Flammer et al. (2016) show that incorporating
CSR criteria into executive compensation contracts significantly reduces managerial
107
myopia. Focusing on “sin” industries, Jo and Na (2012) argue that CSR can effectively
mitigate the risk to which firms in controversial industries are typically exposed.
So far, the previous studies tend to treat stakeholders as a collective concept and analyse
CSR spending at an aggregate level without differentiating between the interests of
different stakeholders. Our study aims to fill this gap by decomposing CSR
commitments into multiple individual dimensions, and focusing specifically on the
interplay between employees and other stakeholders of a business.
3.2.2 Hypothesis Development
3.2.2.1 CSR and Union Strike Probability
A natural starting point for exploring the relation between CSR and union strike risk is
to understand unions’ attitudes towards CSR activities (Sobczak and Havard 2015).
Unions are established to safeguard the various interests of employees: wages, working
conditions, training, job security and other employee-related issues, all of which fall
into the employee relation dimension under the CSR framework (Preuss et al. 2006).
Given unions’ agenda of maximising employee wellbeing, it is intuitive that a labour
union would invariably welcome and favour a high level of labour-related CSR efforts
from the firm. As a result of such efforts, by directly providing a higher employment
standard and proactively building a harmonious labour-management relationship, firms
will be less prone to labour strikes32.
Nevertheless, labour unions’ stance towards CSR spending in non-employee
dimensions such as the community and the environment remains obscure. Informed by
the conflicting views on CSR expenditure established in the literature, we formulate two
32 Given that better employee treatment will naturally lead to a lower strike probability, in our study, we
differentiate employee-related CSR from other CSR initiatives and mainly focus on the impact of non-
employee CSR expenditure on union behaviour.
108
competing conjectures on the influence of non-employee CSR spending on union
behaviour: (1) resource competition and (2) quality signalling.
• Resource Competition
To begin with, consistent with the agency view of CSR expenditure that a manager
invests in CSR activities to fulfil individual social commitments and enhance his
personal reputation at the expense of the shareholders’ wealth, the labour union would
perceive CSR as a waste of financial resources that could otherwise be invested in the
employees, to improve pay or working conditions, especially if the cause of a CSR
project were very distant from the employees’ interests (Friedman 1970; Jensen and
Meckling 1976; Jensen 2001; Moser and Martin 2012; Cheng et al. 2013; Krüger 2015).
Moreover, like any other investment, CSR activities cost companies a significant
amount in financial resources (Russo and Perrini 2010; Lys et al. 2015)33. Therefore,
even if CSR is a genuine effort by the firm to serve its stakeholders, spending
voluntarily on non-employee CSR initiatives to serve external stakeholders would leave
the labour union with the impression that the company has excessive resources, which
might induce it to try to extract rents through collective bargaining (Barnea and Rubin
2010; Krüger 2015). Previous studies provide empirical evidence that firms are exposed
to higher strike risk in the presence of excessive cash holding (Klasa et al. 2009), high
executive compensation (Huang et al. 2017) and low leverage (Myers and Saretto 2016).
We argue that the impression of surplus resources created by CSR spending would also
provoke labour unions to strike, as the perceived benefit of doing so would evidently be
higher.
33 It is estimated that a Fortune 500 company, on average, spends more than 30 million dollars each year,
which is a material proportion of its income notwithstanding the size of such firms (Financial Times
2014).
109
Most crucially, from a multi-stakeholder perspective, a high level of CSR spending on
the society or environment would imply that the company prioritises external
stakeholders such as the community and environmentalists before even its employees,
which would be likely to trigger extreme dissatisfaction amongst employees (Donaldson
and Preston 1995; Helmig et al. 2016). Hence, we argue that a high level of CSR
spending on non-employee issues would intensify the conflicts of interests between the
employees and other stakeholders. In order to compete against other stakeholders for the
limited corporate resources, labour unions in high-CSR firms would be more likely to
engage in extreme bargaining activities such as strikes. Consistent with the “resource
competition” conjecture, we propose our first hypothesis H1a:
H1a: The positive union effect on strike risk is exacerbated in the presence of a high
level of (non-employee) CSR spending.
• Quality Signalling
Alternatively, drawing from stakeholder theory, a firm can use a high level of CSR
engagement to signal its quality to not only the investors but also its stakeholders,
giving the firm a competitive advantage (Porter and Kramer 2006). By earning a
reputation as being a “stakeholder-orientated” business and thus trust from various
stakeholders, a firm can build harmonious relationships with its stakeholders, including
its employees (Barnett 2007; Cuypers et al. 2016; Lins et al. 2017).
Furthermore, prior literature suggests that CSR has a positive impact on employees in
multiple ways, which could in turn influence the behaviour of organised labour (Moser
and Martin 2012; Cuypers et al. 2016). First of all, a number of studies show that CSR
is an effective way to attract and retain talent (Albinger and Freeman 2000; Greening
and Turban 2000; Bhattacharya et al. 2008; Bode et al. 2015; Carnahan et al. 2017;
110
Flammer and Kacperczyk 2019). Meanwhile, prior literature suggests that CSR
activities can boost employee morale and engagement, as employees take great pride in
working for companies committed to the society (Carroll and Shabana 2010;
Balakrishnan et al. 2011; Flammer and Luo 2017; Block et al. 2017; Colombo et al.
2019). As a result, high-CSR firms tend to enjoy higher employee satisfaction
(Valentine and Fleischman 2008; Edmans 2012) and consequently higher innovative
productivity from their employees (Flammer 2015a; Flammer and Kacperczyk 2016).
Therefore, we argue that amicable relationships with the employees and strong
employee satisfaction gained from working for a “socially responsible” employer would
induce a positive attitude towards the employer and attract more cooperative union
behaviour, thus significantly reducing the likelihood of a labour strike. Based on the
“quality signalling” story, we offer a competing hypothesis H1b:
H1b: The positive union effect on strike risk is mitigated in the presence of a high level
of (non-employee) CSR spending.
3.2.2.2 CSR as a Strategic Tool
It is well established in the labour union literature that companies make strategic
decisions to mitigate strike risk and improve their bargaining position against labour
unions (Bronars and Deere 1991; Klasa et al. 2009; Matsa 2010; Bova 2013; Chung et
al. 2016; Huang et al. 2017). We argue that firms strategically adjust their CSR
spending for the same reason.
On the one hand, following the “resource competition” story, from a multiple-
stakeholder perspective, in response to the increased strike risk and pressure from
labour unions, managers would significantly cut CSR spending on non-employee issues
to prevent rent extraction by labour unions and, more crucially, avoid giving their
111
employees the impression that they give priority to other stakeholders, such as
communities and environmental NGOs, over their employees (Ullmann 1985;
Donaldson and Preston 1995; Sobczak and Havard 2015; Helmig et al. 2016). Another
potential reason for a firm to reduce its CSR expenditure following unionisation relates
to the precautionary saving motive against any future labour disputes or wage increase
demands from the labour union (Han and Qiu 2007; He et al. 2016).
On the other hand, according to the “quality signalling” story, companies might
strategically increase their CSR efforts to improve their relationships with various key
stakeholders, including their employees. Intuitively, faced with powerful labour, firms
are more likely to proactively engage in CSR activities, particularly employee-related
CSR, to signal their “quality” as “socially responsible” employers, with a view to both
enhancing their employees’ job satisfaction (Valentine and Fleischman 2008; Edmans
2012) and forming good relationships with the labour unions (Helmig et al. 2016).
Overall, we conjecture that firms use CSR as a strategic instrument to undermine union
power and mitigate strike risk following the event of unionisation. Based on the
contrasting views that labour unions may have on CSR spending, we propose the
following competing hypotheses for H2:
H2a: Firms reduce (non-employee) CSR spending following unionisation.
H2b: Firms increase (non-employee) CSR spending following unionisation.
3.3 Data and Research Design
3.3.1 Data and Sample
To test our hypotheses, we collect our data from various sources: (1) the National Labor
Relations Board (NLRB) for union election results; (2) the Thomson Reuters ASSET4
ESG database for CSR data; (3) the Bureau of Labor Statistics (BLS) and the Federal
112
Mediation and Conciliation Service (FMCS) for strike data; (4) the CSRP/Compustat
merged database for other relevant financial data and firm characteristics. Following
data cleaning and matching, we obtain a base sample of 138 unique election firms from
2002 to 2011, based on which we construct two separate samples containing 563 and
1343 firm-year observations respectively to test our main hypothesis H1 and additional
hypothesis H2.
3.3.1.1 Union Election Data
Union election data from 1980 to 2011 are obtained from the NLRB, the governing
body for labour relations in the United States. According to the NLRB Act, eligible
employees at each establishment have to vote in the union election organised by the
NLRB to make a democratic decision on whether to certify a union as a collective
bargaining unit, following a simple majority rule34. For each union election, we extract
the following information: total number of valid votes (Vote Total), number of votes in
favour of unionisation (Vote For), number of votes against unionisation (Vote Against),
election outcome (Unionisation) and election date. In addition, we construct a new
variable Vote Share, defined as the number of votes for unionisation (Vote For) divided
by the total number of valid votes (Vote Total). For data-matching purposes, we also
gather the employer’s name, city, state and industry classification SIC code.
3.3.1.2 CSR Data
To directly answer our research question, it would be ideal to use the level of actual
CSR spending on each individual dimension. However, despite the increasing trend in
CSR reporting, firms are not legally required to disclose their CSR engagement in
monetary terms and, so far, there is no database reporting the level of CSR expenditure
34 For more details on the NLRB unionisation process, see DiNardo and Lee (2004)
113
at the firm-year level. Therefore, we have to rely on CSR performance as a proxy for the
level of CSR spending35 (Servaes and Tamayo 2013; Lys et al. 2015). Specifically, we
obtain CSR data for the period 2002-2016 from the Thomson Reuters ASSET4 ESG
database36, which assigns CSR scores based on percentile ranks, mainly for three CSR
dimensions: Society, Environment and Corporate Governance. Consistent with existing
CSR literature (Cheng et al. 2013; Flammer 2015a; Krüger 2015; Liang and Renneboog
2017b), we exclude the Corporate Governance category from our analysis as corporate
governance is a mechanism that serves the interests of shareholders and does not
necessarily incur monetary expenses, making it fundamentally different from CSR
initiatives aimed at addressing social problems and serving a wide range of stakeholders.
To measure the overall level of CSR spending, we construct a composite index,
CSRScore, an equally weighted average of the Society and Environment pillars. Because
ASSET4 includes employee-related issues under the Society pillar, which would
introduce noise and contaminate our empirical analyses, we also remove the two key
employee-related data points, (1) Employment Quality and (2) Training & Development,
and recalculate the score for Society without them. Meanwhile, we also construct an
35 Crucially, we assume a monotonic relation between CSR spending and CSR performance, i.e., that higher spending on CSR initiatives will lead to higher CSR performance. This assumption is theoretically
reasonable and consistent with Lys et al. (2015). Nevertheless, the authors admit that this is a data
limitation.
36 This is one of the two most popular CSR databases used in the existing CSR studies, the other being the
KLD database (Huang and Watson 2015). However, since the CSR data from KLD are effectively drawn
from the number of strengths and concerns on each dimension, they tend to be static, with limited
variation, due to the binary scoring system (Barnea and Rubin 2010; Schreck 2011). On the other hand,
ASSET4 is based on percentile rank, and the CSR score for a firm is relative to all the other firms.
Therefore, the data are more dynamic and a firm has to genuinely spend more money and resources than
other firms to improve its CSR rating. The ASSET4 database has been validated in prior CSR studies
(Ioannou and Serafeim 2012; Cheng et al. 2014; Lys et al. 2015; Hawn and Ioannou 2016). Particularly,
Lys et al. (2015) use the CSR score as a proxy for the level of CSR expenditure. Therefore, we believe that the ASSET4 database is more appropriate for our analyses. A more detailed description of the
ASSET4 ESG database can be found in Cheng et al. (2014) and the official website
(https://financial.thomsonreuters.com/en/products/data-analytics/company-data/esg-research-data.html).
114
Employee score by combining those two employee-related data points37. Thus, in our
empirical analyses, we mainly focus on the following key CSR variables: CSRScore,
CSR_noemp38, Environment, Society and Employee.
3.3.1.3 Labour Strikes Data
We manually collect the strike data from the BLS and the FMCS39 for the 138 unique
union election firms matched with CSR data. Specifically, since a strike is an extreme
bargaining incident that occurs only occasionally to a small number of firms, we collect
the strike information within the window (t-4, t+4), that is, from four years before to
four years after the union election year t. For each firm-year observation, we construct
the following two key variables: (1) Strike Dummy, which is equal to one if the firm
experiences a strike during the fiscal year, and zero otherwise; (2) Strike Risk, which is
an ordinal variable that captures different levels of strike risk at the firm-year level,
where 0=no strike; 1=one strike; 2=multiple strikes.
3.3.2 Sample Construction
Prior to merging the union election dataset with the CSR dataset, we start our data
processing with the NLRB dataset because the union election information is crucial to
our identification strategy. Firstly, following the routine process used in previous union
studies (DiNardo and Lee 2004; Lee and Mas 2012; Bradley et al. 2017; Campello et al.
2018), we only keep union elections classified under the “RC” type, which refers to
37 We assume Employment Quality (SOEQ) and Training & Development (SOTD) carry equal weight and
use their average to measure the overall level of employee welfare. Some argue that health and safety
could also be relevant to employees working in dangerous conditions, such as in the mining industry
(Christensen et al. 2017). Our results remain robust to the inclusion of the Health and Safety data point in
our Employee score.
38The variable CSRScore effectively captures Environment, Employee and Society, while CSR_noemp
captures the non-employee CSR performance (i.e., Environment and Society) after removing the
employee-related CSR data points.
39 The BLS documents information for large strikes involving more than 1000 people, whereas the FMCS
records information for strikes with less than 1000 people.
115
elections held for union certification, and those classified as “Closed”, to eliminate
unratified election cases that are subject to change or re-election. More importantly,
since all the union elections are held at the establishment level and not the firm level, it
is possible for multiple union elections to be taking place at different branches of the
same company. In line with the prior literature, we retain only the first election observed
for a given firm (DiNardo and Lee 2004; Bradley et al. 2017; Huang et al. 2017). This is
because the first election result is believed to be the most exogenous as it is not subject
to the influence of the results of other union elections from different branches of the
same company. Consistent with the union literature (Lee and Mas 2012; Campello et al.
2018), we only keep the larger elections, with at least 50 votes, which are believed to
have a material impact on firm decisions. This is then matched with the ASSET4 ESG
dataset using fuzzy matching algorithms40 based on company names, due to the lack of
unique identifiers such as GVKEY or CUSIP in the NLRB union election dataset,
followed by manual verifications (DiNardo and Lee 2004; Lee and Mas 2012).
Our sampling procedure results in 138 unique union election firms from 2002 to 2011.
In order to test the CSR effect on a union’s decision to strike (H1), for each of the 138
election firms, we have the strike information from t-4 to t+4, and thereby construct a
panel dataset at the firm-year level. Lastly, we utilise the GVKEY identifier to merge
the NLRB/CSR/Strike dataset with the financial information and firm characteristics
from Compustat and CRSP. After dropping observation with missing values, our final
sample for our main hypothesis consists of 563 firm-year observations. Similarly, to
examine the firm adjustment in CSR spending following unionisation (H2), we match
40 For more information on the fuzzy matching process, see Lee and Mas (2012) and Wasi and Flaaen
(2015).
116
the 138 election firms with the ASSET4 database and obtain a panel sample of 1343
observations.
3.3.3 Summary Statistics
Table 1 presents the descriptive statistics of the variables used in our empirical
analyses41. In terms of union election data, according to Panel A, the average vote share
is 40.7%, with the unions winning 26% of the elections. These statistics are similar to
those reported in previous union studies (He et al. 2016; Campello et al. 2018). Panel B
reports summary statistics for the CSR data in our main sample. Both the CSR index
and the individual pillars of Society, Environment and Employees have means around 50,
which is very similar to the figures in Cheng et al. (2014) and indicative of the
representativeness of our sample, as the CSR scores from ASSET4 are essentially
percentile ranks of CSR performance. Interestingly, the average score for the employee
dimension is lower than those for the other dimensions, suggesting that firms with
relatively less CSR in the employee dimension tend to have union elections.
***Insert Table 1 here***
3.3.4 Research Design
3.3.4.1 Identification Strategy
To test our main hypothesis H1, we exploit the unique quasi-experimental setting of
union elections, which generate exogenous variation in employees’ bargaining power42,
and employ a triple-differences strategy to establish the causal impact of CSR
41 Variable definitions are included in Appendix 1.
42 As regulated by the NLRB, union elections are organised under a secret-ballot election system and
follow a simple majority rule. Thus, once the vote share passes 50%, firms are subject to the treatment of
unionisation and experience a discontinuous increase in employees’ bargaining power.
117
spending 43 on the post-unionisation strike probability. Essentially, we compare the
unions’ propensity to strike between high-CSR and low-CSR firms.
To test our second hypothesis H2, we take an event-study approach to examine the firm
adjustment in CSR spending in reaction to the event of a union election using a
difference-in-differences (DID) identification strategy. Essentially, the DID regression
analysis compares the pre-to-post change in CSR spending for the treatment group, i.e.,
firms whose employees decide to form a union, to the change experienced by the
control group, i.e., firms refusing to unionise. Any difference between the two groups
should be attributable to the treatment of unionisation.
3.3.4.2 Empirical Models
To test our H1, we run the probit model shown in Equation (1) below to study how the
level of CSR spending, as proxied by the CSR scores, affects the union’s propensity to
strike. The dependent variable is Strike Dummy, which is equal to one if there is a strike
at the firm-year level. The variable of interest is the interaction term
Treatedi×Posti,t×CSRi,t where CSR is an indicator equal to one if the corresponding
CSR expenditure is above the sample median and zero otherwise. Therefore, the
coefficient β1 captures the differential treatment (i.e., unionisation) effect on strike risk
between high-CSR and low-CSR firms under a triple-differences strategy. Following
Klasa et al. (2009), we also control for the change (i.e., first difference) in a vector of
firm characteristics that would affect unions’ decision to strike. To address the concern
of reverse causality, all the control variables are lagged by one year. In addition, year
and industry (two-digit SIC code) fixed effects are included in the regressions to
account for macroeconomic conditions across years and unobservable time-invariant
43 In our empirical analysis, we focus on the aggregate level of CSR spending as well as the individual
CSR dimensions of Employee, Society and Environment.
118
industry factors. When testing the effect of an individual CSR dimension on strike risk,
we also control for the scores in other dimensions provided in the ASSET4 ESG
database in additional specifications44. Furthermore, we run an ordered probit model by
replacing Strike Dummy with an ordinal variable Strike Risk, which categorises the risk
into three levels (0=no strike; 1=one strike; 2=multiple strikes) as a robustness check.
All standard errors are clustered at the firm level.
Strike Dummy=α+β1(Treatedi×Posti,t×CSRi,t)+β2Treatedi×Posti,t +β3Posti,t ×CSRi,t
+β4 Treatedi × CSRi,t+ β5Posti,t +β6Treatedi +β7CSRi,t+β8RTWj,t +β9Cashi,t-1
+β10Leveragei,t-1 + β11Dividendi,t-1 + β12Incomei,t-1 + β13WorkCapi,t-1
+β14ZScorei,t-1 + β15Market-to-Booki,t-1+Year FE+ Industry FE+ ɛijt (1)
To test our H2, in our DID analysis, we run the following regression model shown in
Equation (2) to test firms’ CSR adjustment in response to unionisation. The dependent
variable is the CSR spending and the variable of interest is the interaction term
Treatedi×Posti,t, which represents the DID unionisation effect. Following prior literature
(Artiach et al. 2010; Ferrell et al. 2016; Liang and Renneboog 2017b), we also control
for a vector of firm characteristics as well as year and industry (two-digit SIC code)
fixed effects. To alleviate the concern of reverse causality, the control variables are
lagged by one year.
CSRi,t= α+β1(Treatedi×Posti,t)+β2Treatedi +β3Posti,t+β4Cashholdingi,t-1+β5FCFti,t-1+β6Leveragei,t-1
+ β7 InterestCoveri,t-1 + β8 Sizei,t-1 + β9 ROAi,t-1 + β10TobinQi,t-1+ β11Dividendi,t-1
+ β12 CapitalExpi,t-1 + β13 BlockOwnershipi,t-1+Year FE+ Industry FE+ɛit (2)
44 Controlling for the level of CSR expenditure in other dimensions is particularly important in our study because it allows us to study unions’ behaviour and attitude towards CSR expenditure in one dimension
(e.g., environment) whilst taking into account the levels of CSR expenditure in other dimensions (e.g.,
employees and society). In other words, we test whether labour unions’ strike propensity is affected by
both the absolute level and the relative level of CSR expenditure.
119
3.4 Empirical Findings
3.4.1 CSR Spending and Union Strike Probability
3.4.1.1 Overall CSR Level
To formally test our hypothesis H1, we begin our analysis by focusing on how the
overall level of CSR spending affects a union’s propensity to strike. Table 2 presents the
regression results from the triple-differences specifications45. Specifically, we interact
Treated*Post with a CSR indicator, which is equal to one if the overall CSR expenditure
is above the sample median and zero otherwise. The variable of interest is the three-way
interaction term Treated*Post*CSR, which effectively captures the differential
unionisation effect on strike probability between the high-CSR and low-CSR firms. As
illustrated in Columns (1)-(4), in line with our expectation, Treated*Post is consistently
positive and statistically significant, which satisfies our underlying assumption that
unionisation leads to higher strike risk. More importantly, Treated*Post*CSR is
positively significant, suggesting that a high level of overall CSR spending exacerbates
the union effect on strike likelihood. Economically, the marginal effect indicates that
high-CSR firms are exposed to a 44.28%46 higher strike probability than their low-CSR
counterparts. In the robustness checks, in Columns (5)-(8), we obtain very similar
results after removing the employee-related components from the overall CSR
expenditure. These results lend support to our “resource competition” conjecture (H1a)
that a high level of CSR spending intensifies the resource competition between
employees and other stakeholders.
***Insert Table 2 here***
45 Since strike events are rare, the number of observations in probit models (e.g., Columns 1 and 3) is
much smaller due to perfect predictions.
46 The marginal effect is based on the estimates from Column (3) in Table 2.
120
3.4.1.2 Decomposition of CSR
Having established that a high level of overall CSR spending exacerbates the union
effect on strike risk, in this section, we conduct further analyses at a more granular level
by decomposing the overall CSR score into the Employee, Society and Environment
dimensions to determine which CSR dimensions are the main drivers behind the above
results.
• Employee-related CSR
Starting with the Employee dimension, which captures both Employment Quality and
Training & Development, as illustrated in Table 3, we find that Treated*Post*CSR is
negatively significant, which is consistent with our prediction and intuition that a high
level of overall employee welfare will naturally lead to a lower post-unionisation strike
risk due to greater job satisfaction being perceived by the employees. Since the overall
Employee score is a combination of Employment Quality and Training & Development,
we repeat the analyses to explore which of these two employee issues is more important
to labour unions’ strike decision. Columns (5) to (8) present the results for Employment
Quality, which covers most of the labour unions’ concerns: wages, benefits, pay
disparity, job security and working conditions. Consistent with our expectation, we find
negative and highly significant results at the 1% level across the various specifications,
confirming that a high level of Employment Quality can significantly mitigate the strike
risk following unionisation. In comparison, the results in Columns (9) to (12) indicate
that Training & Development plays a relatively weaker role in mitigating strike risk.
Overall, we find consistent evidence that the unionisation effect on the strike probability
is significantly mitigated by a high level of spending on employee-related CSR
(Employee), predominantly driven by spending on Employment Quality rather than
Training & Development.
121
***Insert Table 3 here***
• Non-Employee CSR
In contrast to the picture for employee-related CSR spending, the results in Table 4
show that high levels of CSR expenditure on Environment and Society exacerbate the
strike risk following the unionisation of the labour force. The explanation for this is
two-fold. Primarily, from a multiple-stakeholder perspective, we argue that a high level
of non-employee CSR spending would suggest to the labour union that the firm
prioritises external stakeholders over the employees, a key internal stakeholder of the
firm, intensifying the perceived conflict of interests amongst the various stakeholders.
As a result, in the presence of high levels of CSR spending on non-employee
dimensions, labour unions will engage in extreme collective-bargaining tactics to
compete for the limited financial resources against the other stakeholders. In addition, a
high level of non-employee CSR expenditure would imply there are surplus resources,
which might provoke the labour union to attempt rent extraction through collective
bargaining (Klasa et al. 2009; Krüger 2015; Myers and Saretto 2016).
***Insert Table 4 here***
A comparison of the results in Tables 3 and 4 reveals that unions do react differently
based on the relevance of the CSR spending to employee interests, and further supports
the idea that labour unions use their bargaining power to try to pressurise firms to divert
more resources away from other stakeholders such as society and the environment, and
towards the employees. Taken together, the results in Section 4.1 present consistent
evidence in support of hypothesis H1a, in line with our “resource competition” story.
122
3.4.2 CSR Adjustment in Response to Unionisation
Now that we have established that non-employee CSR expenditure exacerbates the
unionisation effect on the strike probability, we then examine whether firms
strategically adjust their CSR spending in response to unionisation, which is our
hypothesis H2. Informed by previous literature on strategic firm decisions to undermine
union power and mitigate strike risk, we predict that firms will strategically cut their
CSR spending in non-employee dimensions following the unionisation of their labour
force. As shown in Table 5, the coefficient on the variable of interest Treated*Post,
which represents the DID treatment effect of unionisation on CSR spending, is
insignificant47.
***Insert Table 5 here***
The insignificant unconditional results imply that, when making CSR adjustment
decisions, managers face a difficult tradeoff between mitigating strike risk and
signalling quality. Previous literature has argued that certain firms tend to have a high
level of dependence on CSR to signal their quality to the market (McWilliams and
Siegel 2001; Kotchen and Moon 2012; Cheng et al. 2014; Oh et al. 2017; Flammer
2018). In other words, the quality signalling effect of CSR is so crucial to the viability
of these firms that they prioritise the need to signal quality over mitigating the strike
risk. Thus, these firms are reluctant to reduce their CSR spending significantly
following unionisation, despite the potentially higher strike risk. In the following
47 In an untabulated analysis, we estimate the treatment effect of unionisation on the change in CSR
spending (t-1, t+1) using a regression discontinuity design by focusing on the “marginal treated” and
“marginal control” firms with vote shares within a small bandwidth around the 50% threshold. We find that, relative to the “marginal control” firms, the “marginal treated” ones significantly cut non-employee
CSR expenditure (CSRnoemp), which is consistent with our “resource competition” story. However, due
to the small sample size of 52 observations, we are very cautious about making claims regarding the
generalisability of these results.
123
section, we explore such heterogeneities by partitioning our sample based on (1)
financial constraints, (2) “sin” industries and (3) product market competition.
3.4.2.1 Financial Constraints
Previous literature has shown that CSR plays an important role in corporate financing
because it enhances communication and transparency. This enables firms with superior
CSR performance to enjoy a lower cost of equity (Dhaliwal et al. 2011; Goss and
Roberts 2011; El Ghoul et al. 2011; Cheng et al. 2014; Dhaliwal et al. 2014), while
those with CSR concerns are penalised by higher costs of bank loans (Goss and Roberts
(2011). Similarly, Cheng et al. (2014) show that high-CSR firms enjoy better access to
capital and are significantly less likely to face financial constraint, attributing the
improved financing access to better stakeholder relationships and enhanced
transparency. In light of the rapid expansion of SRI funds, maintaining a decent level of
CSR performance is becoming increasingly vital, especially for financially constrained
firms wishing to secure sustainable access to capital (Renneboog et al. 2008).
Table 6 presents the results for the unionisation effect on CSR spending in various
dimensions, conditional on financial constraints 48 . Consistent with our “resource
competition” story, the negative and significant results for Treated*Post suggest that
unconstrained firms strategically reduce investment in non-employee CSR, such as
Environment and Society, to mitigate the increased strike risk following unionisation,
which supports our H2a. Interestingly enough, unconstrained firms seem to be reluctant
to proactively increase employee-related CSR as this could undermine their bargaining
position in contract negotiations with labour unions.
48 Consistent with Cheng et al. (2014), we use the Kaplan-Zingales (KZ) index lagged by one year and
define financially constrained firms as those whose KZ index is in the top quartile and unconstrained
firms as those whose KZ index is in the bottom quartile.
124
Constrained firms, as indicated by Treated*Post*Constraint, however, tend to maintain
much higher CSR levels than unconstrained firms, lending support to the view that
firms use CSR to signal quality and improve capital financing (Dhaliwal et al. 2011;
Goss and Roberts 2011; El Ghoul et al. 2011; Cheng et al. 2014; Dhaliwal et al. 2014).
We explain these different adjustments in CSR spending by the two groups as follows:
To begin with, unlike unconstrained firms, financially constrained firms are exposed to
a much lower strike risk as they have few resources that can be targeted by labour
unions (Myers and Saretto 2016). More importantly, constrained firms have strong
incentives to maintain their CSR at satisfactory levels in order to signal their quality to
the market so as to secure sustainable capital, which is key to their survival (Dhaliwal et
al. 2011; Goss and Roberts 2011; El Ghoul et al. 2011; Cheng et al. 2014; Dhaliwal et al.
2014).
***Insert Table 6 here***
3.4.2.2 Sin Industries
Another important heterogeneity lies in the controversial nature of the so-called “sin”
industries (e.g., tobacco, alcohol, gambling, the military, etc.). Under close scrutiny
from stakeholders, firms in “sin” industries engage intensively in CSR activities to
counterbalance the social immorality and negative externalities of their businesses (Cai
et al. 2012; Jo and Na 2012; Kotchen and Moon 2012; Oh et al. 2017). Therefore, given
the vital role CSR plays in “sin” industries, we postulate that a “sin” firm49 should have
a fundamentally different CSR strategy to a “non-sin” firm, placing greater emphasis on
the “quality signalling” role of CSR investment following a unionisation event.
49 Following the definitions of “sinful” businesses in previous studies (Hong and Kacperczyk 2009; Cai et
al. 2012), we identify the following as “sin” industries: alcohol, tobacco, gambling, weapons, oil and
nuclear power.
125
***Insert Table 7 here***
As shown in Table 7, following unionisation, rather than reducing CSR spending to
mitigate union power, “sinful” firms strategically increase their CSR spending in non-
employee dimensions in spite of the potentially higher strike risk. We interpret this
finding as evidence that such firms seek to enhance the impression external stakeholders
have of them, and to minimise any negative perceptions held by the market in the
context of labour unionisation, which supports our “quality signalling” story. This
finding also confirms the imperative role of CSR in counterbalancing negative
externalities and protecting fragile reputations of firms in “sin” industries. Surprisingly,
we find no significant adjustment of employee-related CSR spending, suggesting that
firms avoid engaging in unrequested spending on employees to preserve their
bargaining position, and engage in precautionary saving in preparation for potential
collective-bargaining activities by the labour unions in the near future (He et al. 2016).
In contrast with “sin” firms, in line with our H2a, firms in “non-sinful” industries, as
captured by Treated*Post*NonSin, maintain significantly lower levels of non-employee
CSR spending (in both Environment and Society dimensions) relative to “sin” firms, to
mitigate the strike risk and undermine union power. These contrasting CSR adjustments
in response to unionisation support our postulation that “sin” firms have a
fundamentally different CSR strategy from “non-sin” firms. It appears that the former
are constantly conscious of their stigmatised image and the controversy surrounding
their businesses, making them extremely hesitant to cut CSR spending even though, by
definition, CSR is meant to be a voluntary commitment to the society. Nevertheless,
these findings collectively support our conjecture that firms use CSR as a strategic
instrument in response to the unionisation event.
126
3.4.2.3 Product Market Competition
Finally, we explore the cross-sectional variation in product market competition to
further examine the heterogeneity of the unionisation effect on CSR spending. Prior
CSR literature has established that firms in highly competitive industries rely more on
CSR to differentiate themselves from their competitors and gain a competitive
advantage (McWilliams and Siegel 2001; Porter and Kramer 2006; Fernández-Kranz
and Santaló 2010; Zhang et al. 2010; Flammer 2015a; Flammer 2015b; Flammer 2018).
For example, the recent study by Flammer (2018) finds that the positive effect of CSR
on securing procurement contracts is more pronounced in competitive industries. Given
the stronger incentive to signal their quality to the market through CSR engagement, we
predict that firms facing higher levels of product market competition are less likely to
cut their CSR spending significantly in reaction to unionisation.
As a proxy for product market competition, we use a firm-specific measure of product
similarity developed by Hoberg and Phillips (2016)50. Intuitively, a higher level of
product similarity indicates higher market competition and thus a greater need for a firm
to differentiate itself from its rivals through CSR spending. To condition our results on
product market competition, we construct an indicator, PMC, equal to one if the level of
product similarity is in the top tercile and zero if it is in the bottom tercile. To conduct
our cross-sectional analysis of product market competition, we interact PMC with
Treated*Post, forming our variable of interest, whose coefficient is predicted to be
positive.
50 Unlike industry-level proxies based on static industry classifications such as SIC or NAICS codes, the
product similarity data are measured at the firm level, giving a more accurate reflection of the market
competition faced by each individual firm, given its product portfolio. Another appealing feature is that
the data are dynamic in the sense of being recalculated yearly to reflect changes in product market
conditions (Hoberg and Phillips 2016; Mattei and Platikanova 2017; Aobdia and Cheng 2018). The data
and detailed information regarding their construction are publicly available on Hoberg and Phillips’
website at http://hobergphillips.tuck.dartmouth.edu/industryconcen.htm.
127
Table 8 presents the unionisation effect on CSR spending conditional on product market
competition. In line with our conjecture, Treated*Post*PMC has a positively significant
effect at the 1% level on both the overall level of CSR spending in Columns (1) and (2)
and CSR spending in the Society dimension in Columns (7) and (8), whereas
Treated*Post has a consistently negative and significant effect in the corresponding
regressions. We interpret the contrasting effects of unionisation on CSR spending as
follows. While firms facing low levels of product market competition significantly
reduce their non-employee CSR spending following unionisation events, to undermine
union power and mitigate the strike risk (supporting H2a), firms facing high levels of
product market competition prioritise their need to signal quality. As a result, that latter
are reluctant to significantly reduce their non-employee CSR spending, particular in the
Society dimension, which would arguably be more relevant to consumers and thus more
effective in terms of differentiation from competitors (Pivato et al. 2008; Öberseder et al.
2013).
***Insert Table 8 here***
3.4.3 Robustness Test: Propensity Score Matched Sample
In this section, for robustness, we repeat our analyses for H2 using a propensity-score-
matched (PSM) sample. Despite the natural experimental setting of union elections,
there could still be some differences in firm characteristics between the treated and
control firms that might be confounding our results. To reduce sample bias, we match
our treatment firms (i.e., Treated=1) to control firms (i.e., Treated=0) using the PSM
approach to make sure that the two groups are comparable in terms of firm
characteristics at t-1, that is, the year before the union election. Specifically, we use the
same set of control variables as in Equation (2) to generate a propensity score as the
128
benchmark for our matching. Doing so not only ensures consistency across our analyses
but also importantly minimises the concern about observable confounding factors. In
other words, if all the observable factors likely to determine CSR spending are very
similar between the treatment group and the control group, then any differences in CSR
spending can plausibly be attributed to the treatment, i.e., unionisation. Specifically, we
construct our PSM sample with replacement using a common support51. Table 9 Panel
A shows the results of the covariate balance test, which compares the mean values of
the pre-treatment firm characteristics that are likely to affect CSR investment decisions.
Compared with the unmatched sample, in our PSM sample, there is no significant
difference between the treatment group and the control group for any of the firm
characteristics, which confirms the matching quality and alleviates the concern
regarding confounding effects52.
Panel B presents the results based on our PSM sample. Similarly to earlier, in Table 5,
the unconditional results remain insignificant in Columns (1) and (2). However, as
shown in Columns (3)-(8), the estimates on the key interaction terms
Treated*Post*Constraint, Treated*Post*NonSin and Treated*Post*PMC are all
significant and retain their respective signs, consistent with our results in Tables 6-8. In
summary, our results are robust to using a PSM sample after we have made sure that the
treatment and control groups are comparable, which offers additional assurance
regarding our empirical findings.
51 For the sake of sample size, we do not impose restrictions such as the same year, same industry or
nearest neighbour. In other words, we allow a treated firm to be matched with a control firm on the basis
of the firm characteristics alone. In the PSM sample, we have 26 treated firms and 57 control firms.
52 As indicated at the bottom of Panel A, there is an overall reduction in sample bias in terms of both
mean and median, despite the increase in bias for some variables, such as interest coverage. In addition,
the variable Tobin’s Q, which was marginally significant in the unmatched sample, becomes insignificant
after the propensity score matching.
129
***Insert Table 9 here***
3.5 Conclusion
In this paper, we empirically investigate the role of CSR expenditure in unions’
propensity to strike. In support of our “resource competition” story, we find that CSR
spending in non-employee dimensions such as environment and society amplifies the
unionisation effect on strike risk, while employee-related CSR significantly mitigates
such risk. The economic magnitude is nontrivial: firms with high levels of CSR
spending are exposed to a 44.28% higher post-unionisation strike probability, in
comparison with their low-CSR counterparts. We argue that high levels of CSR
spending on non-employee dimensions exacerbate the conflicts of interests between
employees and other stakeholders, and intensify unions’ efforts in collective-bargaining
activities as they compete for the limited firm resources against other stakeholders.
Subsequent analyses provide evidence that, in response to the unionisation event, firms
strategically reduce their CSR spending in non-employee dimensions in order to
mitigate the strike risk and improve their bargaining position against the labour unions.
However, such strategic downward adjustments in CSR expenditure are less
pronounced for financially constrained firms, firms in “sin” industries and firms facing
high levels of product market competition, due to their strong incentives to signal
quality through high levels of CSR spending.
Our study makes multiple contributions to the literature. First, it provides original
evidence on unions’ attitudes towards CSR spending on other stakeholders, revealing an
unintended consequence of high CSR expenditure, resource competition amongst
stakeholders, against the backdrop of the unprecedented CSR phenomenon. Second, we
contribute to the growing literature on strategic corporate decisions in the presence of
130
labour unions, by showing that firms tend to use CSR as a strategic instrument. Our
study also has important managerial implications. Rather than treating CSR as a box-
ticking exercise, firms should regularly review their relationships with different
stakeholders, and balance their interests through comprehensive strategic planning when
making CSR investment decisions. Finally, our study adds to the understanding of
union behaviour and the decision to initiate a labour strike. Thus, our evidence could
serve as a reference for managers and policymakers in both the United States and in
other jurisdictions, such as Europe, where the labour union movement plays a
prominent role in the economy, necessitating risk management against the threat of
labour strikes. Overall, our study sheds light on the inter-stakeholder relationship
through the lens of organised labour, a key primary stakeholder within a business.
131
References
Albinger, H.S. and Freeman, S.J. (2000). Corporate Social Performance and
Attractiveness as an Employer to Different Job Seeking Populations. Journal of
Business Ethics, 28(3), pp.243–253.
Aobdia, D. and Cheng, L. (2018). Unionization, Product Market Competition, and
Strategic Disclosure. Journal of Accounting and Economics, 65(2–3), pp.331–357.
Artiach, T., Lee, D., Nelson, D. and Walker, J. (2010). The Determinants of Corporate
Sustainability Performance. Accounting and Finance, 50(1), pp.31–51.
Ashenfelter, O. and Johnson, G.E. (1969). Bargaining Theory, Trade Unions, and
Industrial Strike Activity. American Economic Review, 59(1), pp.35–49.
Baird, P.L., Geylani, P.C. and Roberts, J.A. (2012). Corporate Social and Financial
Performance Re-Examined: Industry Effects in a Linear Mixed Model Analysis.
Journal of Business Ethics, 109(3), pp.367–388.
Balakrishnan, R., Sprinkle, G.B. and Williamson, M.G. (2011). Contracting Benefits of
Corporate Giving: An Experimental Investigation. The Accounting Review, 86(6),
pp.1887–1907.
Barnea, A. and Rubin, A. (2010). Corporate Social Responsibility as a Conflict between
Shareholders. Journal of Business Ethics, 97(1), pp.71–86.
Barnett, M.L. (2007). Stakeholder Influence Capacity and the Variability of Financial
Returns to Corporate Social Responsibility. Academy of Management Review, 32(3),
pp.794–816.
Becker, B.E. and Olson, C.A. (1986). The Impact of Strikes on Shareholder Equity.
Industrial and Labor Relations Review, 39(3), pp.425–438.
Bénabou, R. and Tirole, J. (2010). Individual and Corporate Social Responsibility.
Economica, 77(305), pp.1–19.
Bhattacharya, C.B., Sen, S. and Korschun, D. (2008). Using Corporate Social
Responsibility to Win the War for Talent. MIT Sloan Management Review, 49(2),
pp.37–44.
Block, E.S., Glavas, A., Mannor, M.J. and Erskine, L. (2017). Business For Good? An
Investigation into the Strategies Firms Use to Maximize the Impact of Financial
Corporate Philanthropy on Employee Attitudes. Journal of Business Ethics, 146(1),
pp.167–183.
Bode, C., Singh, J. and Rogan, M. (2015). Corporate Social Initiatives and Employee
Retention. Organization Science, 26(6), pp.1702–1720.
Bova, F. (2013). Labor Unions and Management’s Incentive to Signal a Negat ive
Outlook. Contemporary Accounting Research, 30(1), pp.14–41.
Bradley, D., Kim, I. and Tian, X. (2017). Do Unions Affect Innovation? Management
Science, 63(7), pp.2251–2271.
Bronars, S. and Deere, D. (1991). The Threat of Unionization, the Use of Debt, and the
Preservation of Shareholder Wealth. Quarterly Journal of Economics, 106(1), pp.231–
254.
132
Brown, W.O., Helland, E. and Smith, J.K. (2006). Corporate Philanthropic Practices.
Journal of Corporate Finance, 12(5), pp.855–877.
Bureau of Labor Statistics. (2017). Major Work Stoppages in 2016. Available at:
https://www.bls.gov/news.release/archives/wkstp_02092017.pdf [Accessed: 1 August,
2017]
Cai, Y., Jo, H. and Pan, C. (2012). Doing Well while Doing Bad? CSR in Controversial
Industry Sectors. Journal of Business Ethics, 108(4), pp.467–480.
Campello, M., Gao, J., Qiu, J. and Zhang, Y. (2018). Bankruptcy and the Cost of
Organized Labor: Evidence from Union Elections. Review of Financial Studie, 31(3),
pp.980–1013.
Cao, J., Liang, H. and Zhan, X. (2017). Peer Effects of Corporate Social Responsibility.
SSRN Electronic Journal, pp.1–47.
Carnahan, S., Kryscynski, D. and Olson, D. (2017). When Does Corporate Social
Responsibility Reduce Employee Turnover? Evidence from Attorneys before and after
9/11. Academy of Management Journal, 60(5), pp.1932–1962.
Carroll, A.B. and Shabana, K. (2010). The Business Case for Corporate Social
Responsibility: A Review of Concepts, Research and Practice. International Journal of
Management Reviews, 15(1), pp.1–18.
Chen, C., Chen, Y., Hsu, P.H. and Podolski, E.J. (2016). Be Nice to Your Innovators:
Employee Treatment and Corporate Innovation Performance. Journal of Corporate
Finance, 39, pp.78–98.
Chen, H., Kacperczyk, M. and Ortiz-Molina, H. (2012). Do Nonfinancial Stakeholders
Affect the Pricing of Risky Debt? Evidence from Unionized Workers. Review of
Finance, 16(2), pp.347–383.
Chen, H.J., Kacperczyk, M. and Ortiz-Molina, H. (2011). Labor Unions, Operating
Flexibility, and the Cost of Equity. Journal of Financial and Quantitative Analysis,
46(01), pp.25–58.
Cheng, B., Ioannou, I. and Serafeim, G. (2014). Corporate Social Responsibility and
Access to Finance. Strategic Management Journal, 35, pp.1–23.
Cheng, I., Hong, H. and Shue, K. (2013). Do Managers do Good with Other People’s
Money? National Bureau of Economic Research Working Paper Series, pp.1–40.
Cheng, L. (2017). Organized Labor and Debt Contracting: Firm-Level Evidence from
Collective Bargaining. The Accounting Review, 92(3), pp.57–85.
Chino, A. (2016). Do Labor Unions Affect Firm Payout Policy? Operating Leverage
and Rent Extraction Effects. Journal of Corporate Finance, 41, pp.156–178.
Choi, J. and Wang, H. (2009). Stakeholder Relations and the Persistence of Corporate
Financial Performance. Strategic Management Journal, 30(8), pp.895–907.
Christensen, H.B., Floyd, E., Liu, L.Y. and Maffett, M. (2017). The Real Effects of
Mandated Information on Social Responsibility in Financial Reports: Evidence from
Mine-Safety Records. Journal of Accounting and Economics, 64(2–3), pp.284–304.
Chung, R., Lee, B.B.-H., Lee, W.-J. and Sohn, B.C. (2016). Do Managers Withhold
Good News from Labor Unions? Management Science, 62(1), pp.46–68.
133
Chyz, J.A., Ching Leung, W.S., Zhen Li, O. and Meng Rui, O. (2013). Labor Unions
and Tax Aggressiveness. Journal of Financial Economics, 108(3), pp.675–698.
Clark, K.B. (1984). Unionization and Firm Performance: The Impact on Profits, Growth,
and Productivity. American Economic Review, 74(5), pp.893–919.
Colombo, S., Guerci, M. and Miandar, T. (2019). What do Unions and Employers
Negotiate Under the Umbrella of Corporate Social Responsibility? Comparative
Evidence from the Italian Metal and Chemical Industries. Journal of Business Ethics,
155(2), pp.445–462.
Compa, L.A. (2008). Corporate Social Responsibility and Workers’ Rights.
Comparative Labor Law and Policy Journal, 30(1), pp.1–10.
Cramton, P. and Tracy, J. (1994). The Determinants of U.S. Labor Disputes. Journal of
Labor Economics, 12(2), pp.180–209.
Crane, A., McWilliams, A., Matten, D., Moon, J. and Siegel, D. (2009). The Corporate
Social Responsibility Agenda. In The Oxford Handbook of Corporate Social
Responsibility. Oxford University Press, pp. 1–15.
Cuypers, I.R.P., Koh, P.-S. and Wang, H. (2016). Sincerity in Corporate Philanthropy,
Stakeholder Perceptions and Firm Value. Organization Science, 27(1), pp.173–188.
Davidson, R.H., Dey, A. and Smith, A. (2016). CEO Materialism and Corporate Social
Responsibility. Chicago Booth Research Paper, 16–11(137), pp.1–69.
Dawkins, C. (2010). Beyond Wages and Working Conditions: A Conceptualization of
Labor Union Social Responsibility. Journal of Business Ethics, 95(1), pp.129–143.
Dawkins, C. (2016). A Test of Labor Union Social Responsibility: Effects on Union
Member Attachment. Business & Society, 55(2), pp.214–245.
Dhaliwal, D., Li, O.Z., Tsang, A. and Yang, Y.G. (2014). Corporate Social
Responsibility Disclosure and the Cost of Equity Capital: The Roles of Stakeholder
Orientation and Financial Transparency. Journal of Accounting and Public Policy,
33(4), pp.328–355.
Dhaliwal, D.S., Li, O.Z., Tsang, A. and Yang, Y.G. (2011). Voluntary Nonfinancial
Disclosure and the Cost of Equity Capital: The Initiation of Corporate Social
Responsibility Reporting. The Accounting Review, 86(1), pp.59–100.
DiNardo, J. and Hallock, K. (2002). When Unions “Mattered”: The Impact of Strikes on
Financial Markets, 1925–1937. ILR Review, 55(2), pp.219–233.
DiNardo, J. and Lee, D. (2004). Economic Impacts of New Unionization on Private
Sector Employers: 1984–2001. Quarterly Journal of Economics, 119(4), pp.1383–1441.
Donaldson, T. and Preston, L.E. (1995). The Stakeholder Theory of the Corporation.
The Academy of Management Review, 20(1), pp.65–91.
Eccles, R.G., Ioannou, I. and Serafeim, G. (2014). The Impact of Corporate
Sustainability on Organizational Processes and Performance. Management Science,
60(11), pp.2835–2857.
Edmans, A. (2011). Does the Stock Market Fully Value Intangibles? Employee
Satisfaction and Equity Prices. Journal of Financial Economics, 101(3), pp.621–640.
Edmans, A. (2012). The Link between Job Satisfaction and Firm Value, with
134
Implications for Corporate Social Responsibility. Academy of Management Journal, 26,
pp.1–19.
El Ghoul, S., Guedhami, O., Kwok, C.C.Y. and Mishra, D.R. (2011). Does Corporate
Social Responsibility Affect the Cost of Capital? Journal of Banking & Finance, 35(9),
pp.2388–2406.
Fauver, L., McDonald, M.B. and Taboada, A.G. (2018). Does it Pay to Treat Employees
Well? International Evidence on the Value of Employee-Friendly Culture. Journal of
Corporate Finance, 50, pp.84–108.
Fernández-Kranz, D. and Santaló, J. (2010). When Necessity Becomes a Virtue: The
Effect of Product Market Competition on Corporate Social Responsibility. Journal of
Economics & Management Strategy, 19(2), pp.453–487.
Ferrell, A., Liang, H. and Renneboog, L. (2016). Socially Responsible Firms. Journal of
Financial Economics, 122(3), pp.585–606.
Financial Times. (2014). Fortune 500 Companies Spend More than $15bn on Corporate
Responsibility. Available at: https://www.ft.com/content/95239a6e-4fe0-11e4-a0a4-
00144feab7de [Accessed: 31 August, 2017]
Fisman, R., Heal, G. and Nair, V.B. (2006). A Model of Corporate Philanthropy.
Working Paper, Columbia University.
Flammer, C. (2015a). Does Corporate Social Responsibility Lead to Superior Financial
Performance ? A Regression Discontinuity Approach. Management Science, 61(11),
pp.2549–2568.
Flammer, C. (2015b). Does Product Market Competition Foster Corporate Social
Responsibility? Evidence from Trade Liberalization. Strategic Management Journal,
36(10), pp.1469–1485.
Flammer, C. (2018). Competing for Government Procurement Contracts: The Role of
Corporate Social Responsibility. Strategic Management Journal, 39(5), pp.1299–1324.
Flammer, C., Hong, B. and Minor, D. (2016). Corporate Governance and the Rise of
Integrating Corporate Social Responsibility Criteria in Executive Compensation. SSRN
Working Paper, pp.1–42.
Flammer, C. and Kacperczyk, A. (2016). The Impact of Stakeholder Orientation on
Innovation: Evidence from a Natural Experiment. Management Science, 62(7),
pp.1982–2001.
Flammer, C. and Kacperczyk, A. (2019). Corporate Social Responsibility as a Defense
against Knowledge Spillovers: Evidence from the Inevitable Disclosure Doctrine.
Strategic Management Journal, 40, pp.1243–1267.
Flammer, C. and Luo, J. (2017). Corporate Social Responsibility as an Employee
Governance Tool: Evidence from a Quasi-Experiment. Strategic Management Journal,
38, pp.163–183.
Fombrun, C. and Shanley, M. (1990). What’s in a Name? Reputation Building and
Corporate Strategy. Academy of Management Journal, 33(2), pp.233–258.
Freeman, R. and Medoff, J. (1979). Two Faces of Unionism. Public Interest, 57, pp.69–
93.
Freeman, R.E. (1984). Strategic Management: A Stakeholder Approach. Boston:
135
Pittman-Ballinger.
Friedman, M. (1970). The Social Responsibility of Business is to Increase its Profits.
The New York Times Magazine, 32 (September), p.13.
Ghaly, M., Dang, V.A. and Stathopoulos, K. (2015). Cash Holdings and Employee
Welfare. Journal of Corporate Finance, 33, pp.53–70.
Goss, A. and Roberts, G.S. (2011). The Impact of Corporate Social Responsibility on
the Cost of Bank Loans. Journal of Banking and Finance, 35(7), pp.1794–1810.
Greening, D.W. and Turban, D.B. (2000). Corporate Social Performance as a
Competitive Advantage in Attracting a Quality Workforce. Business & Society, 39(3),
pp.254–280.
Han, S. and Qiu, J. (2007). Corporate Precautionary Cash Holdings. Journal of
Corporate Finance, 13(1), pp.43–57.
Harrison, J.S., Bosse, D.A. and Phillips, R.A. (2010). Managing for Stakeholders,
Stakeholder Utility Functions, and Competitive Advantage. Strategic Management
Journal, 31(1), pp.58–74.
Hawn, O. and Ioannou, I. (2016). Mind the Gap: The Interplay between External and
Internal Actions in the Case of Corporate Social Responsibility. Strategic Management
Journal, 37(13), pp.2569–2588.
He, J., Tian, X. and Yang, H. (2016). Payout Policy under Enhanced Labor Power:
Evidence from a New Approach. SSRN Electronic Journal, August, pp.1–58.
Helmig, B., Spraul, K. and Ingenhoff, D. (2016). Under Positive Pressure: How
Stakeholder Pressure Affects Corporate Social Responsibility Implementation. Business
& Society, 55(2), pp.151–187.
Hoberg, G. and Phillips, G. (2016). Text-Based Network Industries and Endogenous
Product Differentiation. Journal of Political Economy, 124(5), pp.1423–1465.
Hoi, C.K., Wu, Q. and Zhang, H. (2013). Is Corporate Social Responsibility (CSR)
Associated with Tax Avoidance? Evidence from Irresponsible CSR Activities. The
Accounting Review, 88(6), pp.2025–2059.
Hong, H. and Kacperczyk, M. (2009). The Price of Sin: The Effects of Social Norms on
Markets. Journal of Financial Economics, 93(1), pp.15–36.
Huang, Q., Jiang, F., Lie, E. and Que, T. (2017). The Effect of Labor Unions on CEO
Compensation. Journal of Financial and Quantitative Analysis, 52(2), pp.553–582.
Huang, X.B. and Watson, L. (2015). Corporate Social Responsibility Research in
Accounting. Journal of Accounting Literature, 34, pp.1–16.
Husted, B.W., Jamali, D. and Saffar, W. (2016). Near and Dear? The Role of Location
in CSR Engagement. Strategic Management Journal, 37(10), pp.2050–2070.
Ioannou, I. and Serafeim, G. (2012). What Drives Corporate Social Performance? The
Role of Nation-Level Institutions. Journal of International Business Studies, 43(9),
pp.834–864.
Jensen, M.C. (2001). Value Maximization, Stakeholder Theory, and the Corporate
Objective Function. Journal of Applied Corporate Finance, 14(3), pp.8–21.
Jensen, M.C. and Meckling, W.H. (1976). Theory of the Firm: Managerial Behavior,
136
Agency Costs and Ownership Structure. Journal of Financial Economics, 3(4), pp.305–
360.
Jo, H. and Na, H. (2012). Does CSR Reduce Firm Risk? Evidence from Controversial
Industry Sectors. Journal of Business Ethics, 110(4), pp.441–456.
Kim, Y., Park, M.S. and Wier, B. (2012). Is Earnings Quality Associated with
Corporate Social Responsibility? The Accounting Review, 87(3), pp.761–796.
Klasa, S., Maxwell, W.F. and Ortiz-Molina, H. (2009). The Strategic Use of Corporate
Cash Holdings in Collective Bargaining with Labor Unions. Journal of Financial
Economics, 92(3), pp.421–442.
Kotchen, M. and Moon, J.J. (2012). Corporate Social Responsibility for Irresponsibility.
The B.E. Journal of Economic Analysis & Policy, 12(1). pp.1–21
KPMG. (2015). KPMG Survey of Corporate Responsibility Reporting 2015. Available
at: https://assets.kpmg.com/content/dam/kpmg/pdf/2016/02/kpmg-international-survey-
of-corporate-responsibility-reporting-2015.pdf [Accessed: 31 August, 2017]
Krüger, P. (2015). Corporate Goodness and Shareholder Wealth. Journal of Financial
Economics, 115(2), pp.304–329.
Lee, D.S. and Mas, A. (2012). Long-Run Impacts of Unions on Firms: New Evidence
from Financial Markets, 1961-1999. The Quarterly Journal of Economics, 127(1),
pp.333–378.
Liang, H. and Renneboog, L. (2017a). Corporate Donations and Shareholder Value.
Oxford Review of Economic Policy, 33(2), pp.278–316.
Liang, H. and Renneboog, L. (2017b). On the Foundations of Corporate Social
Responsibility. Journal of Finance, 72(2), pp.853–910.
Lins, K.V., Servaes, H. and Tamayo, A. (2017). Social Capital, Trust, and Firm
Performance: The Value of Corporate Social Responsibility during the Financial Crisis.
Journal of Finance, 72(4), pp.1785–1824.
Lys, T., Naughton, J.P. and Wang, C. (2015). Signaling through Corporate
Accountability Reporting. Journal of Accounting and Economics, 60(1), pp.56–72.
Margolis, J.D., Elfenbein, H.A. and Walsh, J.P. (2009). Does it Pay to be Good...and
Does it Matter? A Meta-Analysis of the Relationship between Corporate Social and
Financial Performance. SSRN Electronic Journal, 50, pp.1–68.
Marquis, C. and Lee, M. (2013). Who is Governing Whom? Executives, Governance,
and the Structure of Generosity in Large U.S. Firms. Strategic Management Journal,
34(2), pp.483–497.
Masulis, R.W. and Reza, S.W. (2015). Agency Problems of Corporate Philanthropy.
Review of Financial Studies, 28(2), pp.592–636.
Matsa, D.A. (2010). Capital Structure as a Strategic Variable: Evidence from Collective
Bargaining. Journal of Finance, 65(3), pp.1197–1232.
Mattei, M.M. and Platikanova, P. (2017). Do Product Market Threats Affect Analyst
Forecast Precision? Review of Accounting Studies, 22(4), pp.1628–1665.
McHugh, R. (1991). Productivity Effects of Strikes in Struck and Nonstruck Industries.
Industrial and Labor Relations Review, 44(4), pp.722-732.
137
McWilliams, A. and Siegel, D. (2000). Corporate Social Responsibility and Financial
Performance. Strategic Management Journal, 21(5), pp.603–609.
McWilliams, A. and Siegel, D. (2001). Corporate Social Responsibility: A Theory of
the Firm Perspective. Academy of Management Review, 26(1), pp.117–127.
Moser, D.V. and Martin, P.R. (2012). A Broader Perspective on Corporate Social
Responsibility Research in Accounting. The Accounting Review, 87(3), pp.797–806.
Myers, B.W. and Saretto, A. (2016). Does Capital Structure Affect the Behavior of
Nonfinancial Stakeholders? An Empirical Investigation into Leverage and Union
Strikes. Management Science, 62(11), pp.3235–3253.
Öberseder, M., Schlegelmilch, B.B. and Murphy, P.E. (2013). CSR Practices and
Consumer Perceptions. Journal of Business Research, 66(10), pp.1839–1851.
Oh, H., Bae, J. and Kim, S.J. (2017). Can Sinful Firms Benefit from Advertising Their
CSR Efforts? Adverse Effect of Advertising Sinful Firms’ CSR Engagements on Firm
Performance. Journal of Business Ethics, 143(4), pp.643–663.
Pivato, S., Misani, N. and Tencati, A. (2008). The Impact of Corporate Social
Responsibility on Consumer Trust: The Case of Organic Food. Business Ethics: A
European Review, 17(1), pp.3–12.
Porter, M.E. and Kramer, M.R. (2006). The Link between Competitive Advantage and
Corporate Social Responsibility. Harvard Business Review, 84(12), pp.78–92.
Preuss, L. (2008). A Reluctant Stakeholder? On the Perception of Corporate Social
Responsibility among European Trade Unions. Business Ethics: A European Review,
17(2), pp.149–160.
Preuss, L., Haunschild, A. and Matten, D. (2006). Trade Unions and CSR: A European
Research Agenda. Journal of Public Affairs, 6(1), pp.256–268.
Renneboog, L., Ter Horst, J. and Zhang, C. (2008). Socially Responsible Investments:
Institutional Aspects, Performance, and Investor Behavior. Journal of Banking and
Finance, 32(9), pp.1723–1742.
Reuters. (2008). Boeing Set to Announce New 787 Delays. Available at:
https://www.reuters.com/article/us-boeing/boeing-set-to-announce-new-787-delays-
idUSTRE4B384G20081204 [Accessed: 31 August, 2017]
Ruback, R.S. and Zimmerman, M.B. (1984). Unionization and Profitability: Evidence
from the Capital Market. Journal of Political Economy, 92(6), pp.1134-1157.
Russo, A. and Perrini, F. (2010). Investigating Stakeholder Theory and Social Capital:
CSR in Large Firms and SMEs. Journal of Business Ethics, 91(2), pp.207–221.
Schmidt, M.B. and Berri, D.J. (2004). The Impact of Labor Strikes on Consumer
Demand: An Application to Professional Sports. American Economic Review, 94(1),
pp.344–357.
Schreck, P. (2011). Reviewing the Business Case for Corporate Social Responsibility:
New Evidence and Analysis. Journal of Business Ethics, 103(2), pp.167–188.
Servaes, H. and Tamayo, A. (2013). The Impact of Corporate Social Responsibility on
Firm Value: The Role of Customer Awareness. Management Science, 59(5), pp.1045–
1061.
138
Sobczak, A. and Havard, C. (2015). Stakeholders’ Influence on French Unions’ CSR
Strategies. Journal of Business Ethics, 129(2), pp.311–324.
Tang, Z. and Tang, J. (2018). Stakeholder Corporate Social Responsibility Orientation
Congruence, Entrepreneurial Orientation and Environmental Performance of Chinese
Small and Medium-Sized Enterprises. British Journal of Management, 29(4), pp.634–
651.
Tracy, J.S. (1986). An Investigation into the Determinants of U.S. Strike Activity.
American Economic Review, 76(3), pp.423–436.
Ullmann, A.A. (1985). Data in Search of a Theory: A Critical Examination of the
Relationships among Social Performance, Social Disclosure, and Economic
Performance of U.S. Firms. Academy of Management Review, 10(3), pp.540–557.
Valentine, S. and Fleischman, G. (2008). Ethics Programs, Perceived Corporate Social
Responsibility and Job Satisfaction. Journal of Business Ethics, 77(2), pp.159–172.
Wasi, N. and Flaaen, A. (2015). Record Linkage Using Stata: Preprocessing, Linking,
and Reviewing Utilities. The Stata Journal, 15(3), pp.672–697.
Zhang, R., Zhu, J., Yue, H. and Zhu, C. (2010). Corporate Philanthropic Giving,
Advertising Intensity, and Industry Competition Level. Journal of Business Ethics,
94(1), pp.39–52.
139
Table 1. Descriptive Statistics
This table provides summary statistics for our main sample. Panel A reports union election statistics
collected from the National Labor Relations Board (NLRB). Panel B reports CSR data acquired from the
ASSET4 ESG databases. Panel C reports strike data obtained from the Bureau of Labor Statistics (BLS)
and the Federal Mediation and Conciliation Service (FMCS). Panel D reports data for firm characteristics
collected from the Compustat/CRSP Merged database. All variables are defined in Appendix 1.
Variable Mean 25% Median 75% SD N
Panel A: Union Election
Election Year 2004.942 2003 2004 2006 2.610 138
Unionisation 0.261 0 0 1 0.441 138
Vote Share 0.407 0.268 0.367 0.521 0.200 138
Vote Total 220.978 77 120 202 369.152 138
Vote For 82.645 29 50 91 124.574 138
Vote Against 131.442 43 68.5 118 270.049 138
Panel B: CSR Scores
CSRScore 55.514 30.910 59.425 79.375 26.076 563
CSR_noemp 54.796 32.964 57.258 75.491 22.954 563
Environment 54.230 21.130 56.060 84.310 30.455 563
Society 55.362 41.218 57.568 70.233 19.980 563
Employee 51.978 31.145 52.740 72.725 23.947 563
Employment Quality 53.629 30.620 53.330 78.010 27.524 563
Training&Development 50.327 18.960 49.750 78.450 29.034 563
Panel C: Strike Occurrences
Strike Dummy(0,1) 0.037 0 0 0 0.190 563
Strike Risk(0,1,2) 0.048 0 0 0 0.259 563
Treated 0.249 0 0 0 0.433 563
Post 0.629 0 1 1 0.484 563
RTW 0.329 0 0 1 0.470 563
Panel D: Firm Characteristics
ΔCash 0.005 -0.014 0.002 0.031 0.061 554
ΔLeverage 0.019 -0.175 -0.016 0.118 2.321 559
ΔDividend 0.001 -0.008 0.000 0.020 1.053 557
ΔIncome 0.000 -0.012 0.002 0.014 0.032 559
ΔWorking Capital 0.001 -0.022 0.000 0.029 0.060 555
ΔZScore -0.058 -0.254 0.041 0.323 1.316 537
ΔMarket-to-Book -0.042 -0.146 0.020 0.150 0.581 559
Cash 0.074 0.023 0.053 0.106 0.068 563
Leverage 1.432 0.422 0.739 1.492 3.047 563
Dividend 0.138 0.013 0.133 0.244 0.857 563
Income 0.100 0.058 0.092 0.137 0.058 563
Working Capital 0.113 0.015 0.089 0.203 0.124 563
ZScore 2.588 0.953 1.726 3.322 2.713 541
Market-to-Book Ratio 1.035 0.492 0.816 1.374 0.762 563
Log(Total Assets) 9.370 8.469 9.187 10.148 1.190 563
Log(Sales) 9.451 8.534 9.320 10.212 1.151 563
140
Table 2. Effect of Overall CSR Spending on Union Strikes
The table below reports the results for the effect of the overall CSR level on unions’ decision to strike. The variable of interest is Treated*Post*CSR. CSR
is a dummy variable, equal to 1 if CSRScore in the fiscal year t is above the sample median, in Columns (1)-(4). As a robustness test, we rerun our analyses in Columns (5)-(8) using CSR_noemp, from which employee-related CSR is removed. The dependent variable is Strike Dummy, equal to 1 if there is a
strike during the fiscal year. For robustness, we also use Strike Risk (0=no strike; 1=one strike; 2=multiple strikes) and run ordered probit regressions. P-
values are displayed in parentheses with standard errors clustered at the firm level. ***, ** and * denote significance levels of 1%, 5% and 10%,
respectively. All variables are defined in Appendix 1.
CSR Scores: CSRScore CSR_noemp
(1) (2) (3) (4) (5) (6) (7) (8)
Strike Dummy Strike Risk Strike Dummy Strike Risk Strike Dummy Strike Risk Strike Dummy Strike Risk
Treated*Post*CSR 2.551** 3.162*** 3.283** 3.597*** 2.493** 2.857*** 2.684* 2.880**
(0.038) (0.004) (0.027) (0.004) (0.031) (0.006) (0.058) (0.020)
Treated*Post 1.591 1.544* 1.840* 1.792* 1.816* 1.938** 2.470** 2.665**
(0.101) (0.080) (0.097) (0.056) (0.065) (0.034) (0.048) (0.018)
Post*CSR 0.468 0.452 0.395 0.476 1.134* 1.191** 1.343** 1.447***
(0.483) (0.474) (0.549) (0.448) (0.079) (0.044) (0.035) (0.010)
Treated*CSR -4.043*** -4.572*** -4.876*** -5.036*** -4.247*** -4.823*** -5.112*** -5.536***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Post -0.569 -0.435 -0.818 -0.683 -0.991* -0.900* -1.488** -1.374**
(0.293) (0.404) (0.251) (0.305) (0.070) (0.073) (0.041) (0.037)
Treated -0.410 -0.431 -0.040 -0.201 -0.465 -0.460 -0.305 -0.434
(0.501) (0.437) (0.948) (0.741) (0.461) (0.413) (0.656) (0.493)
CSR -0.266 -0.261 -0.167 -0.201 -0.347 -0.346 -0.081 -0.129
(0.539) (0.512) (0.706) (0.596) (0.458) (0.432) (0.857) (0.731)
Controls N N Y Y N N Y Y
Year FE Y Y Y Y Y Y Y Y
Industry FE Y Y Y Y Y Y Y Y
Pseudo R2 0.150 0.314 0.307 0.416 0.157 0.326 0.326 0.439
N 201 563 179 530 201 563 179 530
141
Table 3. Effect of Employee-related CSR Spending on Union Strikes
The table below reports the results for the effect of employee-related CSR on unions’ decision to strike. The variable of interest is Treated*Post*CSR. CSR is a dummy
variable, equal to 1 if the Employee score in the fiscal year t is above the sample median. As an alternative proxy for employee-related CSR, we rerun our analyses using
the scores for Employment Quality in Columns (5)-(8) and the scores for Training & Development in Columns (9)-(12). The dependent variable is Strike Dummy, equal to 1 if there is a strike during the fiscal year. For robustness, we also use Strike Risk (0=no strike; 1=one strike; 2=multiple strikes) and run ordered probit regressions. P-
values are displayed in parentheses with standard errors clustered at the firm level. ***, ** and * denote significance levels of 1%, 5% and 10%, respectively. All
variables are defined in Appendix 1.
CSR Dimension Employee Employment Quality Training & Development
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Strike
Dummy
Strike
Risk
Strike
Dummy
Strike
Risk
Strike
Dummy
Strike
Risk
Strike
Dummy
Strike
Risk
Strike
Dummy
Strike
Risk
Strike
Dummy
Strike
Risk
Treated*Post*CSR -2.345* -2.067* -2.091* -1.969* -4.503*** -3.777*** -4.717*** -4.072*** -2.192* -2.230* -2.387** -2.678**
(0.054) (0.066) (0.090) (0.077) (0.000) (0.003) (0.000) (0.002) (0.089) (0.069) (0.038) (0.022)
Treated*Post 2.390** 2.359** 2.517** 2.530** 3.343*** 3.058*** 3.697*** 3.447*** 2.366** 2.340** 2.639** 2.705**
(0.018) (0.017) (0.012) (0.012) (0.003) (0.002) (0.008) (0.007) (0.030) (0.026) (0.013) (0.012)
Controls Y Y Y Y Y Y Y Y Y Y Y Y
Other Dimensions N N Y Y N N Y Y N N Y Y
Year FE Y Y Y Y Y Y Y Y Y Y Y Y
Industry FE Y Y Y Y Y Y Y Y Y Y Y Y
Pseudo R2 0.305 0.421 0.334 0.448 0.342 0.432 0.384 0.468 0.300 0.416 0.336 0.450
N 179 530 179 530 179 530 179 530 179 530 179 530
142
Table 4. Effect of Non-Employee CSR Spending on Union Strikes
The table below reports the results for the effect of the non-employee CSR dimensions on unions’ decision to strike. Columns (1)-(6) and Columns (7)-(12) present the results for CSR in the Environment and Society dimensions, respectively. The variable of interest is Treated*Post*CSR. CSR is a dummy variable,
equal to 1 if the corresponding CSR dimension in fiscal year t is above the sample median. The dependent variable is Strike Dummy, equal to 1 if there is a
strike during the fiscal year. For robustness, we also use Strike Risk (0=no strike; 1=one strike; 2=multiple strikes) and run ordered probit regressions. P-values are displayed in parentheses with standard errors clustered at the firm level. ***, ** and * denote significance levels of 1%, 5% and 10%, respectively. All
variables are defined in Appendix 1.
CSR Dimension Environment Society
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Strike Dummy
Strike Risk
Strike Dummy
Strike Risk
Strike Dummy
Strike Risk
Strike Dummy
Strike Risk
Strike Dummy
Strike Risk
Strike Dummy
Strike Risk
Treated*Post*CSR 2.462** 2.922*** 2.113 2.455** 2.906* 3.384** 3.908*** 4.445*** 4.073** 4.226*** 2.821 3.062*
(0.026) (0.003) (0.123) (0.028) (0.088) (0.030) (0.003) (0.000) (0.013) (0.003) (0.127) (0.063)
Treated*Post 1.684* 1.711** 2.149* 2.165** 2.344* 2.396** 1.191 1.145 1.387 1.414 2.099 2.209*
(0.074) (0.043) (0.051) (0.018) (0.057) (0.030) (0.361) (0.324) (0.320) (0.255) (0.150) (0.083)
Controls N N Y Y Y Y N N Y Y Y Y
Other Dimensions N N N N Y Y N N N N Y Y
Year FE Y Y Y Y Y Y Y Y Y Y Y Y
Industry FE Y Y Y Y Y Y Y Y Y Y Y Y
Pseudo R2 0.150 0.316 0.300 0.416 0.389 0.481 0.161 0.320 0.318 0.425 0.388 0.472
N 201 563 179 530 179 530 201 563 179 530 179 530
143
Table 5. Labour Unionisation and CSR (Unconditional Results)
The table below reports the difference-in-differences (DID) regression results for the unionisation effect on CSR spending at the aggregate as well as dimensional levels. The variable of interest is Treated*Post and the dependent variables include CSRScore in Columns (1) and (2), CSR_noemp in Columns
(3) and (4), Employee in Columns (5)-(7), Environment in Columns (8)-(10), and Society in Columns (11)-(13). P-values are displayed in parentheses with
standard errors clustered at the firm level. ***, ** and * denote significance levels of 1%, 5% and 10%, respectively. All variables are defined in Appendix 1.
Dimension CSRScore CSR_noemp Employee Environment Society
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
Treated*Post -2.995 2.995 -3.179 1.178 -3.665 5.157 5.248 0.551 5.613 4.594 -6.909 -3.258 -6.353
(0.583) (0.584) (0.536) (0.819) (0.444) (0.383) (0.375) (0.932) (0.390) (0.331) (0.163) (0.527) (0.133)
Post 7.990** 6.747 7.396** 7.018* 5.665* 2.422 -1.165 9.419** 8.215* 5.638 5.372* 5.821* 3.705
(0.021) (0.103) (0.016) (0.055) (0.082) (0.530) (0.707) (0.021) (0.093) (0.166) (0.077) (0.091) (0.141)
Treated 10.418 -1.272 9.932* 0.786 1.729 -11.483* -11.132* 11.526* 0.493 5.789 8.338 1.078 5.185
(0.100) (0.843) (0.077) (0.890) (0.736) (0.074) (0.066) (0.084) (0.944) (0.209) (0.127) (0.848) (0.287)
Controls N Y N Y N Y Y N Y Y N Y Y
Other Dimensions N N N N N N Y N N Y N N Y
Year FE Y Y Y Y Y Y Y Y Y Y Y Y Y
Industry FE Y Y Y Y Y Y Y Y Y Y Y Y Y
R-squared 0.335 0.541 0.344 0.544 0.238 0.431 0.588 0.360 0.519 0.703 0.263 0.473 0.676
N 1343 1019 1343 1019 1343 1019 1019 1343 1019 1019 1343 1019 1019
144
Table 6. Labour Unionisation and CSR conditional on Financial Constraints
The table below reports the results for the unionisation effect on CSR in different dimensions conditional on financial constraints. We use the Kaplan-
Zingales (KZ) index as a proxy for financial constraints. The variable of interest is Treated*Post*Constraint, where Constraint is a dummy variable equal to
1 if the lagged KZ index is in the top quartile and equal to 0 if the lagged KZ index is in the bottom quartile. P-values are displayed in parentheses with
standard errors clustered at the firm level. ***, ** and * denote significance levels of 1%, 5% and 10%, respectively. All variables are defined in Appendix 1.
Financially Constrained vs Non-Constrained
Dimension CSRScore CSR_noemp Employee Environment Society
(1) (2) (3) (4) (5) (6) (7) (8)
Treated*Post*Constraint 31.467** 28.193** 14.304 -7.082 26.787* 10.768 29.599** 19.440**
(0.018) (0.023) (0.313) (0.537) (0.086) (0.347) (0.016) (0.037)
Treated*Post -22.137** -23.405** 1.487 18.206* -25.807** -18.738** -21.002* -16.500**
(0.044) (0.027) (0.908) (0.074) (0.034) (0.041) (0.053) (0.034)
Controls Y Y Y Y Y Y Y Y
Other Dimensions N N N Y N Y N Y
Year FE Y Y Y Y Y Y Y Y
Industry FE Y Y Y Y Y Y Y Y
R-squared 0.659 0.672 0.538 0.711 0.668 0.806 0.570 0.742
N 520 520 520 520 520 520 520 520
145
Table 7. Labour Unionisation and CSR Conditional on “Sin” Industries
The table below reports the results for the unionisation effect on CSR in different dimensions, conditional on whether the firms are in “sin” industries. The
variable of interest is Treated*Post*NonSin, where NonSin is a dummy variable equal to 1 if the firm does not belong to any of the “sin” industries: alcohol, tobacco, gambling, weapons, oil and nuclear power (Hong and Kacperczyk 2009). P-values are displayed in parentheses with standard errors clustered at the
firm level. ***, ** and * denote significance levels of 1%, 5% and 10%, respectively. All variables are defined in Appendix 1.
Sin Industries vs Non-Sin Industries
Dimension CSRScore CSR_noemp Employee Environment Society
(1) (2) (3) (4) (4) (5) (6) (7)
Treated*Post*NonSin -28.679*** -32.951*** -7.500 8.494 -36.583*** -28.959*** -29.318*** -23.046***
(0.000) (0.000) (0.473) (0.426) (0.001) (0.003) (0.000) (0.006)
Treated*Post 30.147*** 32.770*** 11.425 -3.800 41.209*** 33.085*** 24.330*** 15.240**
(0.000) (0.000) (0.179) (0.667) (0.000) (0.000) (0.000) (0.027)
Controls Y Y Y Y Y Y Y Y
Other Dimensions N N N Y N Y N Y
Year FE Y Y Y Y Y Y Y Y
Industry FE Y Y Y Y Y Y Y Y
R-squared 0.546 0.549 0.438 0.592 0.523 0.704 0.481 0.681
N 1019 1019 1019 1019 1019 1019 1019 1019
146
Table 8. Labour Unionisation and CSR Conditional on Product Market Competition
The table below reports the results for the unionisation effect on CSR in different dimensions conditional on product market competition. The variable of
interest is Treated*Post*PMC, where PMC is a dummy variable equal to 1 if the firm’s product similarity (Hoberg and Phillips 2016) in year t is in the top tercile and zero if it is in the bottom tercile. P-values are displayed in parentheses with standard errors clustered at the firm level. ***, ** and * denote
significance levels of 1%, 5% and 10%, respectively. All variables are defined in Appendix 1.
High Product Competition vs Low Product Competition
CSR Dimension: CSRScore CSR_noemp Employee Environment Society
(1) (2) (3) (4) (5) (6) (7) (8)
Treated*Post*PMC 36.236*** 36.321*** 1.118 -23.593* 31.836** 8.185 40.806*** 31.245***
(0.009) (0.003) (0.942) (0.074) (0.037) (0.451) (0.001) (0.000)
Treated*Post -15.963 -18.242** 3.988 17.348* -12.295 -0.408 -24.189*** -21.531***
(0.128) (0.040) (0.746) (0.066) (0.250) (0.949) (0.003) (0.000)
Controls Y Y Y Y Y Y Y Y
Other Dimensions N N N Y N Y N Y
Year FE Y Y Y Y Y Y Y Y
Industry FE Y Y Y Y Y Y Y Y
R-squared 0.623 0.635 0.470 0.605 0.606 0.735 0.559 0.720
N 602 602 602 602 602 602 602 602
147
Table 9. Robustness Test: Propensity-Score-Matched Sample
This table reports the results using a propensity-score-matched (PSM) sample as a robustness check of our results in Tables 5-8. The treatment group contains
the firms that vote for unionisation in their elections (i.e., with a vote share of more than 50%), and the control group contains firms for which the vote share is below the 50% threshold. We match the treatment firms to control firms with replacement and common support based on a vector of observable firm
characteristics at time t-1, that is, one year before the union election event. Panel A tabulates the means of the firm characteristics at t-1 for the treatment
group and the control group, before and after the propensity score matching. Panel B reports the results for CSR adjustments following unionisation, based on the PSM sample, using the same specifications as in Tables 5-8. P-values are displayed in parentheses with standard errors clustered at the firm level. ***, **
and * denote significance levels of 1%, 5% and 10%, respectively. All variables are defined in Appendix 1.
Panel A: Covariate Balance Test of PSM Sample
Variable Unmatched (U) Mean T-Test
Matched (M) Treated Control t p>|t|
Cashholdingt-1 U 0.076 0.058 1.41 0.160
M 0.069 0.057 0.63 0.534
FCFt-1 U 0.037 0.047 -1.12 0.263
M 0.041 0.041 0.00 0.997
Leveraget-1 U 0.276 0.267 0.28 0.781
M 0.284 0.285 -0.02 0.980
InterestCovert-1 U 13.547 13.601 -0.01 0.990
M 14.332 8.188 1.09 0.282
Sizet-1 U 8.797 8.723 0.24 0.809
M 8.669 8.893 -0.64 0.528 ROAt-1 U 0.033 0.045 -1.18 0.239
M 0.031 0.032 -0.04 0.965
TobinQt-1 U 1.041 1.279 -1.73* 0.085
M 1.049 0.996 0.35 0.729
CapitalExpt-1 U 0.041 0.048 -1.10 0.275
M 0.042 0.038 0.51 0.613
Dividendt-1 U 0.010 0.017 -1.16 0.248
M 0.008 0.009 -0.38 0.707
(continued on next page)
148
BlockOwnershipt-1 U 0.222 0.184 1.32 0.190
M 0.211 0.217 -0.13 0.897
Sample Mean Bias Median Bias
Unmatched 19.8 23.7
Matched 8.9 6.3
Panel B: Robustness analyses based on PSM Sample
DID Constrained Sin PMC
CSR Dimension CSRScore CSR_noemp CSRScore CSR_noemp CSRScore CSR_noemp CSRScore CSR_noemp
(1) (2) (3) (4) (5) (6) (7) (8)
Treated*Post*Constraint 27.344** 24.575*
(0.046) (0.061)
Treated*Post*NonSin -27.573*** -33.426***
(0.000) (0.000)
Treated*Post*PMC 21.376* 25.349**
(0.071) (0.021)
Treated*Post 1.036 -0.984 -16.968 -17.448 27.152*** 31.029*** -14.419 -17.540*
(0.838) (0.842) (0.193) (0.168) (0.000) (0.000) (0.188) (0.073)
Controls Y Y Y Y Y Y Y Y
Year FE Y Y Y Y Y Y Y Y
Industry FE Y Y Y Y Y Y Y Y
R-squared 0.593 0.593 0.752 0.744 0.596 0.597 0.644 0.647
N 882 882 446 446 882 882 558 558
149
Appendix
Definition of Variables
Variable Definition
CapitalExp Capital expenditure scaled by total assets.
Cash Cash and short-term investments scaled by total assets.
Constraint Dummy variable equal to one if lagged Kaplan-Zingales index is in the uppermost quartile (above the 75th percentile) and zero if it is in the lowest quartile (below the 25th percentile).
CSR Dummy variable equal to one if the corresponding CSR score is above the sample median and zero otherwise.
CSRScore Equally weighted average of the scores for environment, society and employee-related CSR. CSR_noemp Equally weighted average of the scores for environment and society, excluding employee-related CSR.
Dividend Dividend for common stock divided by total sales.
Employee Score for overall employee welfare, which is the equally weighted average of the scores for Employment Quality
and Training & Development. Employment Quality Score for employment quality.
Environment CSR score for environment dimension, covering resource reduction, emission reduction and product innovation.
FCF (Operating income before depreciation – interest expense – income taxes – capital expenditures)/total assets. Income Earnings before interest and tax.
InterestCover Operating income before depreciation divided by interest expenses.
Kaplan-Zingales Index −1.002 × Cash flow over lagged assets + 0.283 × Tobin’s q + 3.139 × Leverage –39.368 × Dividends – 1.315 × Cash holdings over lagged assets.
Leverage Book value of long-term debt divided by total assets.
Market-to-Book Ratio Market value over book value of total assets.
NonSin Dummy variable equal to one if the firm does not operate in one of the “sin” industries as defined by Hong and Kacperczyk (2009) and zero otherwise.
BlockOwnership Total percentage of share ownership held by blockholders.
PMC Dummy variable equal to one if the product similarity is in the top tercile during the fiscal year and zero if it is in the bottom tercile.
Post Dummy variable equal to one if the fiscal year is after the election year of the firm in question and zero otherwise.
Product Similarity A text-based, firm-specific measure of product similarity developed by Hoberg and Phillips (2016). (continued on next page)
150
ROA Net income divided by total assets.
RTW Dummy variable equal to one if the union election is held in a state with right-to-work legislation and zero
otherwise. Size The logarithm of total assets.
Society CSR score for the society dimension, covering community, health & safety, diversity & opportunity and product
responsibility. Strike Dummy Dummy variable equal to one if there is a strike in the fiscal year t.
Strike Risk Ordinal variable equal to zero if there is no strike, one if there is one strike and two if there are multiple strikes in
the fiscal year t.
TobinQ (Market value of equity+ Total debt)/ Book value of total assets. Training & Development Score for training and development for employees.
Treated Dummy variable equal to one if vote share>50% and zero otherwise.
Unionisation Dummy variable equal to one if vote share>50% and zero otherwise. Vote For Number of votes in favour of unionisation.
Vote Share Number of votes for unionisation divided by total number of votes.
Vote Total Total number of votes in the union election. Working Capital Working capital scaled by total assets.
ZScore 1.2(Working capital/Total assets) +1.4(Retained earnings/Total assets) +3.3(EBIT/Total assets) + 0.6(Market value
of equity/Book value of total liabilities)+(Sales/Total assets).
151
Chapter 4
Do Financial Analysts Play a Complementary or Substitutive Role in the
Corporate Information Environment? Evidence from Organised Labour
Abstract
This paper explores the primary role of financial analysts in the context of unionised
firms, where investors have greater information demand. Previous literature suggests
that labour unions create substantial uncertainty in firms and undermine the information
environment, while another strand of literature argues that analysts devote more effort
to generating valuable information through original research in the case of heightened
uncertainty or information asymmetry. To date, it is unclear whether financial analysts,
as professional information intermediaries, are affected by organised labour. Using a
large U.S. sample over the period of 1983-2015, we find that the labour unionisation
rate is associated with lower forecast accuracy and higher forecast dispersion,
suggesting that financial analysts predominantly play a “complementary role” rather
than a “substitutive role” when firms are facing significant uncertainty in human capital.
Overall, our study has important implications for managers, financial analysts and
regulators, by highlighting the value and hence necessity of non-financial information
disclosure specific to a key intangible asset of firms, i.e., their employees.
152
“…the strike potentially cost shareholders $0.05 to $0.07 in earnings per share for the
second quarter…” - Verizon CFO (Reuters 2016)
4.1 Introduction
It is widely accepted that financial analysts play two important roles as key information
intermediaries in the capital markets: (1) a complementary role 53 , facilitating the
dissemination of publicly available information from firms to investors and (2) a
substitutive role 54 , generating value-relevant information through their original and
specialised research that would not otherwise be available to the markets, as a
substitution for corporate disclosure (Lang and Lundholm 1996; Asquith et al. 2005;
Beyer et al. 2010; Bradshaw et al. 2017). While financial analysts’ contribution to the
information environment as information intermediaries has been extensively discussed
and recognised in the extant accounting and finance literature (Beyer et al. 2010;
Bradshaw et al. 2017), whether and how financial analysts might be affected by
organised labour remains an open question. In light of the increasingly important role of
human capital in today’s corporate environment, in this paper, we explore the behaviour
of financial analysts in the context of organised labour.
Unlike other stakeholders, organised labour constitutes a powerful stakeholder that not
only exists internally within the firm but also has a significant financial claim in the
form of wages and pensions (Faleye et al. 2006; Campello et al. 2018). Prior union
literature has mainly focused on the labour-management interplay and established that
53 Some studies also term this role “information dissemination” (Kross et al. 1990; Bradshaw et al. 2017)
or “information interpretation” (Chen et al. 2010; Livnat and Zhang 2012; Huang et al. 2018). For
consistency, we use the term “complementary role” hereafter in this paper.
54 Some papers term this “information provision”(Lang and Lundholm 1996), the “informational role”
(Bradshaw et al. 2017; Jennings 2019) or “information discovery” (Chen et al. 2010; Livnat and Zhang
2012; Huang et al. 2018). For consistency, we use the term “substitutive role” hereafter in this paper.
153
labour unions not only use their collective bargaining power to safeguard employees’
interests but also exert influence over a wide range of corporate decisions (Klasa et al.
2009; Matsa 2010; Chyz et al. 2013; Chung et al. 2016; Bradley et al. 2017c; Huang et
al. 2017; Hamm et al. 2018). In response, to mitigate labour risk and undermine union
power, managers strategically obfuscate information to improve their bargaining
position against the labour unions (Hilary 2006; Bova 2013; Chung et al. 2016). Thus,
given the greater uncertainty in human capital and information asymmetry, investors
and other market participants are likely to have an increased demand for information
and consequently analyst services. While recent papers (Loh and Stulz 2018; Jennings
2019) attempt to study the role of financial analysts using more extreme settings, such
as economic recessions or managerial misconduct lawsuits, we believe that focusing on
organised labour, an internal stakeholder that directly participates in day-to-day
business operations, is likely to help us tease out the primary role of financial analysts,
with greater generalisability. Hence, in this paper, we specifically investigate how
financial analysts perform when making forecasts for firms facing uncertainty in human
capital.
Informed by the two roles financial analysts play in the capital markets, we develop two
competing hypotheses regarding the direction of the unions’ influence on analysts’
earnings forecasts: (1) a “complementary role” and (2) a “substitutive role”. On the one
hand, if financial analysts primarily serve a “complementary role” by distributing and
interpreting existing information disclosed by firms to investors, we would predict that
both the direct and indirect impact of labour unions on the information environment
would lead to lower forecast quality. Directly, labour unions can affect analyst forecasts
by introducing substantial uncertainty into business operations and future earnings
prospects (Clark 1984; Ruback and Zimmerman 1984; Connolly et al. 1986; Chen et al.
154
2011). To be specific, given the wage agenda, labour unions engage in collective-
bargaining strategies to push for higher wages and better welfare for the employees,
thereby causing significant uncertainty in labour costs, which constitute a major
component of a firm’s annual expenditure55. Consequently, union presence can have
substantial implications for profit margins, and ultimately analysts’ ability to predict
future earnings, if financial analysts predominantly rely on readily available information
in the public domain. On top of the uncertainty in labour costs, labour unions have the
ability to unilaterally initiate large-scale labour strikes, which can be extremely
disruptive to firms’ operations and detrimental to financial performance (Ashenfelter
and Johnson 1969; Becker and Olson 1986; Myers and Saretto 2016).56 Therefore, firms
facing strong unions are exposed to significantly higher strike risk. Thus, from the
perspective of financial analysts acting as information disseminators, we postulate that
financial analysts, ex ante, can neither predict an occurrence nor quantify the economic
consequence of a strike when forecasting future earnings. Additionally, labour unions
can cause uncertainties and inflexibilities in the implementation of corporate strategies,
especially when those strategies are expected to have negative implications for the
employees (Atanassov and Kim 2009; Chen et al. 2011). While financial analysts
closely follow corporate strategies, which are highly relevant to future economic
prospects, it is unlikely that companies will voluntarily disclose the extent to which their
own employees might or might not cooperate in the delivery of such corporate strategies.
On top of the uncertainty labour unions bring to businesses, building on the argument
55 For instance, the total payroll and benefits in 2008 in the manufacturing sector, where labour unions are
more prevalent and active, were $784 billion, more than four times the total capital expenditure, at $166
billion, in the same year (Hamm et al. 2018).
56 This is also supported by the anecdotal evidence. Following a high-profile strike involving more than
40,000 Verizon employees in 2016, the CFO of Verizon at the time estimated that “the strike potentially
cost shareholders $0.05 to $0.07 in earnings per share” (Reuters 2016), while the Wall Street Journal
(2016) reported that the seven-week Verizon strike had cost the largest U.S. telecommunication provider
$343 million in revenue.
155
that managers strategically preserve information asymmetry to obfuscate their financial
position in the presence of labour unions (Liberty and Zimmerman 1986; Hilary 2006;
Chung et al. 2016), financial analysts may also be indirectly affected by labour unions
due to the deteriorated information environment in unionised firms. Taking these
arguments together, consistent with financial analysts primarily engaging in a
“complementary role”, we hypothesise that analysts’ earnings forecast quality will be
lower for firms with a high degree of labour unionisation.
On the other hand, if financial analysts perceive themselves more as information
providers and thus primarily play a “substitutive role”, we would expect them to devote
more resources to and exert more efforts towards generating more value-relevant
information through original research in the context of unionised firms. In response to
the increased uncertainty in human capital and deterioration in the information
environment due to union presence, investors will have a greater demand for
information, particularly from financial analysts, who are uniquely positioned to
produce high-quality information for investors, regarding a firm’s future economic
prospects (Lang and Lundholm 1996; Jennings 2019). In fact, financial analysts are
ideally suited to generate valuable information for investors. First of all, they have
privileged access to information from multiple sources (Huang et al. 2018). When
managerial disclosure is limited or less credible, financial analysts can still access
valuable information from other channels, such as original research, and private
interactions with senior management, rank-and-file employees, customers, and
competitors (Soltes 2014; Huang et al. 2018). In addition to the information advantage,
financial analysts have the skills and expertise to process and aggregate financial
information from various sources to generate more informative analyst reports for
investors (Healy and Palepu 2001; Bradshaw 2011; Huang et al. 2018). Thirdly,
156
financial analysts possess industry-specific knowledge that is proven to be extremely
valuable to investors, given that analysts typically follow firms within a particular sector
and may also have gained industry insights through their professional experience
(Bradley et al. 2017a; Jennings 2019). Empirically, the “substitutive role” of financial
analysts has also been documented in the prior literature (Chen et al. 2010; Bradley et al.
2017a; Huang et al. 2018; Jennings 2019). Therefore, assuming financial analysts are
primarily committed to generating first-hand information through individual research
(substitutive role), given the greater information asymmetry and uncertainty in
unionised firms, we would expect financial analysts to be more diligent in producing
original and value-relevant information for firms facing collective bargaining, in order
to meet the information demand from the investors. Hence, analysts’ forecast quality is
likely to be higher for unionised firms.
Hence, in this paper, we aim to disentangle the dual role of financial analysts by
investigating the analysts’ forecast quality in the presence of strong employee power.
Using a large panel dataset over the period of 1983-2015, we find that the labour
unionisation rate is associated with significantly lower forecast accuracy and higher
forecast dispersion, suggesting that the “complementary role” of financial analysts
dominates the “substitutive role”. 57 Further tests controlling for financial reporting
quality confirm that there is an incremental union effect that cannot be fully explained
by managerial obfuscation in unionised firms (Hilary 2006; Chung et al. 2016).
57 It should be noted that we are not at all suggesting that financial analysts are not generating new
information for investors and the capital markets. In fact, the premise of our research is that financial
analysts do perform both functions, which is supported in the prior literature. The conclusion we draw
from the empirical analyses is that, on aggregate, the “complementary role” seems to dominate, meaning
that financial analysts engage more in disseminating information that is publicly available than in
generating original and valuable information. In other words, if financial analysts overwhelmingly
perceived themselves as information providers, and were diligently conducting research on unionised
firms, we would not have documented such a negative association between labour unionisation and
forecast quality.
157
Therefore, we interpret this incremental effect as evidence consistent with labour union
representation incorporating inherent uncertainties that are difficult for financial
analysts to capture precisely in their earnings forecasts.
Subsample analyses indicate that the union impact on analyst forecast properties is more
pronounced for (1) firms headquartered in non-RTW (right-to-work) states, where
unions enjoy greater bargaining power, and (2) firms operating in low-skilled industries,
where unions play a more active role. Crucially, we find that the disclosure of labour-
related expenses can effectively mitigate the union effect on analysts’ earnings forecast
quality, confirming that the financial analysts predominantly rely on publicly disclosed
information, instead of generating such information through original research. This
analysis also lends additional support to our “uncertainty channel”, whereby labour
unions affect financial analysts’ ability to forecast future earnings by creating
substantial uncertainties in labour costs. Last but not least, we show that financial
analysts are more likely to issue optimistic forecasts for unionised firms, consistent with
financial analysts’ strategic optimism in response to the higher uncertainty in human
capital, and lower earnings predictability, for unionised firms.
Our study contributes to the literature in multiple ways. First of all, our study
contributes to the ongoing debate on the primary role of financial analysts in the capital
markets, against the backdrop of innovation in information technology and
transformation of the information environment (Lang and Lundholm 1996; Altınkılıç et
al. 2013; Loh and Stulz 2018; Schantl 2018; Huang et al. 2018; Jennings 2019).
Different from Loh and Stulz (2018) and Jennings (2019), who examine the role of
financial analysts under extreme circumstances (e.g., economic recessions and
managerial lawsuits), we study analysts’ performance in the context of strong power of
employees, a common key stakeholder playing an increasingly important role in today’s
158
economy. By exploiting the uncertainty in human capital within unionised firms, our
study suggests that financial analysts, on aggregate, predominantly engage in
information dissemination over information discovery. Thus, our study provides
additional insights into analysts’ behaviour and performance in the context of strong
employee power.
Secondly, our study extends the understanding of the influence of the employees as an
internal stakeholder, and more specifically its role in capital markets. While previous
studies have focused on unions’ influence on firm decisions (Klasa et al. 2009; Matsa
2010; Chung et al. 2016; Chino 2016; Bradley et al. 2017c; Huang et al. 2017; Hamm et
al. 2018), our results suggest employees’ influence extends well beyond the firms to a
group of sophisticated market participants, i.e., financial analysts, and can potentially
affect the information environment of the capital markets. While previous studies argue
that managers strategically preserve information asymmetry in order to improve their
bargaining position against labour unions (Hilary 2006; Bova 2013; Chung et al. 2016),
our study offers a more direct “uncertainty channel” by showing that the presence of
union representation itself constitutes an inherent uncertainty that materially undermines
financial analysts’ ability to predict future earnings.
Last but not least, our study has important implications for financial analysts, managers
and policymakers. Our paper echoes the chronic concern on the usefulness of financial
report information and the call for more relevant disclosure of non-financial information
to complement financial reporting regimes (Amir and Lev 1996; Aboody and Lev 1998;
Francis and Schipper 1999; Lev 2018). Similarly to Amir and Lev (1996) and Dhaliwal
et al. (2012), we highlight that human capital information, typically considered
secondary to conventional financial statements, can be highly relevant to investors.
Therefore, when making earnings forecasts, financial analysts should place more
159
emphasis on non-financial information related to human capital, which is a valuable
intangible asset of a firm (Amir et al. 2003). Our study also encourages managers to
make more specific disclosures on their human capital to transparently communicate
information about labour-management relations to their investors, thus signalling
quality to the market. Given the growing importance of employees in today’s
knowledge-intensive economic environment, regulators and standard setters may also
consider mandatory disclosure on human capital investment, to supplement the existing
financial reporting systems and enhance the overall information environment of the
capital markets (Amir and Lev 1996; Amir et al. 2003; Lev 2018). Overall, our study
sheds light on the primary role of financial analysts by examining the interplay between
financial analysts and a powerful internal stakeholder, the employees.
The remainder of the paper is organised as follows. Section 2 reviews the extant
literature and develops our research hypotheses. Section 3 describes our sample and the
empirical design we use in our analyses. Section 4 presents the main empirical results.
Section 5 summarises the findings and contributions of this study.
4.2. Related Literature and Hypothesis Development
4.2.1 Related Literature
4.2.1.1 Financial Analysts
Financial analysts are key information intermediaries who bridge the informational gap
between companies and investors in the capital markets (Bradshaw et al. 2017).
Through their collection and research of value-relevant information, financial analysts
make earnings forecasts and stock recommendations to investors, playing a central role
in facilitating the information flows and efficient functioning of capital markets (Lang
and Lundholm 1996; Healy and Palepu 2001; Beyer et al. 2010).
160
Much of the analyst literature focuses on the influence of financial analysts on capital
markets. First and foremost, empirical evidence confirms that financial analysts, as
information intermediaries, play an effective role in significantly reducing the
information asymmetry between companies and investors (Beyer et al. 2010; Mansi et
al. 2011; Bradshaw et al. 2017). In addition to facilitating the dissemination of value-
relevant information that exists in the markets (complementary role), financial analysts,
prior literature suggests, can also generate original and informative research, thus
increasing the supply of useful information in the capital markets (substitutive role)
(Huang et al. 2018; Jennings 2019). Specifically, prior studies suggest that analysts’
level of experience and skills (Clement 1999; Hashim and Strong 2018) and pre-analyst
industry expertise (Bradley et al. 2017a) are conducive to their generation of first-hand,
valuable outputs that are provided to information users such as investors. Meanwhile,
cultural diversity also enhances the quality of analysts’ earnings forecasts, as Merkley et
al. (2017a) find that a high level of cultural diversity amongst a group of sell-side
analysts, typically within the same brokerage firm, significantly improves the accuracy
and reduces the optimism bias and dispersion of the consensus forecasts.
In addition to improving the information environment, prior literature has shown that
financial analysts perform an external governance role by carrying out strong
monitoring of managerial behaviours (Yu 2008; Irani and Oesch 2013; Chen et al. 2015;
Bradley et al. 2017b; Chen and Lin 2017; Chen et al. 2018). Specifically, previous
studies present evidence that analyst coverage can improve financial reporting quality
(Irani and Oesch 2013; Bradley et al. 2017b) and deter opportunistic managerial
behaviours such as expropriation of shareholders’ wealth (Chen et al. 2015), earnings
management (Yu 2008; Bradley et al. 2017b) and tax avoidance (Chen and Lin 2017;
Chen et al. 2018). Recent studies suggest that financial analysts have a positive
161
influence on corporate investment efficiency (Chen et al. 2017) and the quality of
investment decisions (To et al. 2018).
Notwithstanding their contribution to the efficient functioning of capital markets,
financial analysts do inevitably rely on the information environment at the same time, in
order to fulfil their roles in the capital markets (Lang and Lundholm 1996; Healy and
Palepu 2001; Hope 2003a). Amongst various sources of information, a crucial and
essential type comes from financial reporting, predominantly financial statements and
annual reports (Healy and Palepu 2001). While financial information certainly plays a
central role in financial analysts’ careers, Dhaliwal et al. (2012) find that the
introduction of CSR reports improves financial analysts’ forecast accuracy, suggesting
that financial analysts also refer to non-financial information. Another main source of
information that analysts pay active attention to is corporate disclosure. Such disclosure,
though issued on a voluntary basis, is evidently informative to them and is associated
with higher earnings forecast accuracy (Lang and Lundholm 1996; Healy and Palepu
2001; Hope 2003a; Hope 2003b).
Alongside information availability, the quality of the information itself is equally
pivotal to financial analysts. One of the most important factors in information quality is
the institutional environment within which the information is disseminated (Hope 2003a;
Lang et al. 2003; Tan et al. 2011; Horton et al. 2013). For example, Lang et al. (2003)
show that firms that are cross-listed on U.S. stock exchanges enjoy greater analyst
coverage and higher forecast accuracy, attributable to the better information
environment in the U.S. Similarly, using an international sample, Hope (2003a) finds
that earnings forecasts are more accurate in countries with strong enforcement of
accounting standards. In addition to regulation and enforcement, accounting standards
can have a material impact on financial reporting quality, and consequently the
162
precision of financial analysts’ earnings forecasts. Previous studies document that
forecast accuracy increases significantly following the adoption of International
Financial Reporting Standards (IFRS), which significantly enhance the transparency and
comparability of financial information (Tan et al. 2011; Horton et al. 2013; Petaibanlue
et al. 2015). As well as institutional factors, auditors can have a direct impact on the
usefulness of financial information. Behn et al. (2008) find that analysts’ earnings
forecasts are more accurate for firms audited by Big 5 auditors, suggesting that higher
audit quality significantly improves the credibility and informativeness of financial
statements. De Franco et al (2011) find that financial statement comparability also
enhances analysts’ earnings forecasts. Meanwhile, other studies suggest that the
disclosure of accounting policies (Hope 2003b) and corporate governance (Bhat et al.
2006; Byard et al. 2006) is incrementally useful information, leading to more accurate
earnings forecasts from analysts.
In addition to the reliance on the existing information environment, financial analysts,
despite being professionally qualified experts, are likely to be affected by the inherent
complexity and uncertainty of the underlying businesses (Barron and Stuerke 1998;
Barron et al. 2002; Zhang 2006a; Mattei and Platikanova 2017; Amiram et al. 2018).
When uncertainty about future earnings is high, analysts’ earnings forecasts tend to be
less accurate and more dispersed, albeit issued in a more timely manner (Zhang 2006a;
Amiram et al. 2018). In other words, business complexity and uncertainty regarding
future earnings are difficult for financial analysts to capture precisely in their earnings
forecasts (Duru and Reeb 2002; Barron et al. 2002; Mattei and Platikanova 2017).
Specifically, Barron et al. (2002) find that earnings forecasts tend to be less accurate and
more dispersed for firms with high levels of intangible assets, while Duru and Reeb
(2002) show that analysts issue less accurate and more optimistic earnings forecasts for
163
firms with higher levels of international diversification. Recently, Mattei and
Platikanova (2017) have documented evidence that product market competition is
associated with lower precision in earnings forecasts, due to the increased uncertainty
regarding future cash flows.
Perhaps a fundamental scepticism about the credibility of analysts’ forecasts is that their
forecasting behaviours are ultimately a product of personal judgements and incentives
(Das et al. 1998; Clement 1999; Hong et al. 2000; Lim 2001; Chan et al. 2007; Bradley
et al. 2017a; Horton et al. 2017; Merkley et al. 2017b). Prior literature argues and
provides evidence implying that earnings forecasts are subject to optimism bias and
herding behaviour on the part of financial analysts.
Since forecast accuracy is profoundly important to the career success of financial
analysts (Mikhail et al. 1999; Lim 2001; Hong and Kubik 2003), analysts need to access
as much information relevant to future earnings as possible, from various sources, to
improve the quality of their earnings forecasts and consequently their career outcomes
over the long run. Arguably, the most direct and relevant information source is the
management, who have the most privileged access to all the first-hand information
relevant to the future earnings of the company. Thus, prior literature has established that
sell-side analysts have strong incentives to issue more optimistic earnings forecasts in
order to retain their access to private information from the management (Das et al. 1998;
Hong and Kubik 2003). Interestingly, Hong and Kubik (2003) find that, controlling for
accuracy, analysts who are more optimistic relative to their peers are more likely to
experience favourable career outcomes. Therefore, Lim (2001) argues that, for analysts,
trading off management access against forecast accuracy, it is an optimal and rational
strategy to issue positively biased earnings forecasts, and suggests that the magnitude of
the bias is determined by the firm’s information environment. In the context of the
164
banking industry, Horton et al. (2017) reveal that financial analysts specialising in the
banking sector are more likely to issue positively biased earnings forecasts for banks
that could be prospective employers for them. Other sources of conflicts of interests that
could bias analysts’ earnings forecasts upward include their incentives to attract
investment banking clients and generate trading commissions (Michaely and Womack
1999; Chan et al. 2007).
In parallel with the optimistic bias in the earnings forecasts, sell-side analysts also
engage in herding behaviour due to career concerns (Hong et al. 2000; Clement and Tse
2005). Since the performance of financial analysts is reviewed on a relative basis,
according to Hong and Kubik (2003), analysts with relatively more accurate forecasts
are more likely to enjoy career progression. Consistent with the career-concern
argument, Hong et al. (2000) find financial analysts, particularly those with limited
experience and information access, are more likely to herd with other analysts by
issuing earnings forecasts that are close to the consensus forecasts amongst their peers.
In an investigation into the consequences of analysts’ herding behaviours, Clement and
Tse (2005) find that herding forecasts are less accurate than bold forecasts, suggesting
that the latter convey more private and relevant information to investors.
Despite the aforementioned positive impact on corporate governance quality, analysts’
intense monitoring may also create excessive pressure on managers, leading to
suboptimal managerial decisions. For example, He and Tian (2013) document causal
evidence that analyst coverage hinders firm innovations because managers feel
pressured to meet short-term targets by cutting innovative projects, even if they could be
value-enhancing in the long run.
The interplay between analysts and managers is further complicated by managers’
strategic impression management (Cotter et al. 2006; Hilary 2006). On the one hand,
165
managers have strong incentives to meet or beat analysts’ earnings forecasts in order to
earn higher stock returns and signal strong future economic prospects to investors
(Bartov et al. 2002). On the other hand, Cotter et al. (2006) find that, to counterbalance
the analysts’ optimism in earnings forecasts, managers issue explicit earnings guidance
in an attempt to manage the expectations of financial analysts and lead them towards
more achievable earnings targets.
In the context of labour unions, Hilary (2006) suggests that, in order to improve the
management’s bargaining position against organised labour, managers purposefully
preserve information asymmetry between employees and employers.
4.2.1.2 Labour Unions
Labour unions constitute a powerful primary stakeholder that resides internally within
firms, exerting a strong influence over managerial decisions as well as external
stakeholders such as creditors (Bronars and Deere 1991; Matsa 2010; Chen et al. 2012;
Chyz et al. 2013; Chung et al. 2016; Bradley et al. 2017c; Huang et al. 2017; Cheng
2017; Campello et al. 2018; Hamm et al. 2018). The collective-bargaining power of
labour unions crucially lies in their ability to initiate labour strikes, which can be
extremely disruptive to firms’ operations and costly to the employers (Ashenfelter and
Johnson 1969; Schmidt and Berri 2004; Myers and Saretto 2016). The core agenda of
labour unions is to use their collective-bargaining power to safeguard employees’
interests and demand better welfare from the employers on behalf of individual
employees (Freeman and Medoff 1979).
Theoretically, the seminal work by Freeman and Medoff (1979) proposed that there
were “two faces” of labour unions. On the one hand, consistent with the “monopoly
model”, labour unions use collective-bargaining strategies, such as strikes, to extract
166
economic rent by suboptimally pushing up wages and benefits (Lewis 1964; Freeman
and Medoff 1979; Freeman 1981; Clark 1984; Tracy 1986). On the other hand, the
“collective voice” view argues that labour unions can serve as an effective channel
through which the employees can express their opinions and perform monitoring of the
management (Freeman and Medoff 1979; Chyz et al. 2013; Lin et al. 2018).
Empirically, previous studies have found evidence supporting both views. Consistent
with the “monopoly model”, prior literature has established a positive union effect on
both wages, and non-wage items such as fringe benefits, consistent with rent extraction
through collective bargaining (Lewis 1964; Freeman 1981; Freeman and Medoff 1984;
Pencavel and Hartsog 1984; Card 2001). Meanwhile, several studies show that labour
unions can deter opportunistic managerial behaviour through strong scrutiny (Chyz et al.
2013; Huang et al. 2017). For example, Chyz et al. (2013) find that labour unions
significantly undermine managers’ ability to engage in tax avoidance activities, while
Huang et al. (2017) document that executive compensation is significantly curtailed in
the presence of labour unions, suggesting that organised labour can improve corporate
governance.
Despite the empirical support for both views on labour unions, prior literature has yet to
reach a consensus on their aggregate economic effect (Clark 1984; Ruback and
Zimmerman 1984; DiNardo and Lee 2004; Lee and Mas 2012). Specifically, the earlier
work by Clark (1984) and Ruback and Zimmerman (1984) shows a negative union
effect on firm performance and shareholders’ wealth. By contrast, the latter two studies,
by exploiting the natural experimental setting of union elections, suggest that the
economic impact of labour unionisation on firm value is close to zero.
Nevertheless, the disruptions and uncertainties caused by labour unions are materially
detrimental, particularly when they engage in extreme collective-bargaining activities,
167
such as large-scale strikes. Becker and Olson (1986) find that a strike involving more
than 1,000 workers destroys 4.1 per cent of shareholders’ value, on average, which is
equivalent to around $80 million in 1980 prices58. The negative effect also extends to
the debtholders, as Campello et al. (2018) find that labour unionisation leads to a
decline in bond values. As a result, unionised firms face a higher cost of capital in both
the equity and debt markets (Chen et al. 2011; Cheng 2017). In addition, prior literature
finds that union representation creates operational inflexibilities (Atanassov and Kim
2009; Chen et al. 2011) and hinders firm innovation (Bradley et al. 2017c).
To improve their bargaining position and mitigate strike risk, firms proactively make
strategic corporate decisions. It is well documented that firms strategically adjust their
capital structures by reducing cash holdings (Klasa et al. 2009) and increasing leverage
(Bronars and Deere 1991; Matsa 2010; Myers and Saretto 2016) to essentially shelter
financial resources from organised labour. Furthermore, several studies find that
unionised firms engage in “downward” impression management, disseminating less
positive economic prospects by narrowly missing analysts’ forecasts (Bova 2013) and
strategically withholding good news (Chung et al. 2016) in order to undermine unions’
desire to extract economic rent during labour contract negotiations. Therefore, Hilary
(2006) argues that stronger labour power is associated with a higher degree of
information asymmetry in the capital markets because managers have strong incentives
to obfuscate information to preserve their bargaining position against labour unions.
58 Recent anecdotal evidence suggests that labour strikes have become even more costly in more recent times.
In 2008, a 58-day strike by 27,000 machinists at Boeing, the largest aircraft manufacturer in the world, caused $100 million of losses per day in deferred revenue, and $2 billion in lost profits. The share price also plummeted, by 56 per cent, to a five-year low during the strike period (Reuters 2008). For a more recent example, see footnote 4 for a description of the cost of the 2016 strike suffered by Verizon.
168
4.2.2 Hypothesis Development
While both the “complementary” and “substitutive” roles of financial analysts are
empirically supported in previous studies, it is unclear which role they primarily play in
the context of organised labour. In this section, guided by the two roles that financial
analysts play in the capital markets, we formulate competing hypotheses regarding
analyst forecast quality in the presence of labour unions.
4.2.2.1 Labour Unions and Financial Analysts: “Complementary Role”
Assuming financial analysts primarily serve a “complementary role” in the context of
unionised labour, labour unions could affect analyst forecast quality both directly
through the “uncertainty” channel and indirectly through the “financial reporting”
channel.
Directly, labour unions can bring substantial uncertainties to firms, on multiple fronts.
To begin with, labour costs constitute a significant proportion of companies’ total
expenditure. For instance, in the manufacturing sector, where labour unions are more
prevalent and active, the total payroll and benefits in 2008 were $784 billion, more than
four times of the total capital expenditures at $166 billion, in the same year (Hamm et al.
2018). Knowing that labour expense is a sizeable component on the income statement,
managers simply cannot afford to accept labour unions’ wage demands in their entirety
and will instead bargain with them to seek concessions with respect to wages (Klasa et
al. 2009). As a result, it is impossible for either party, let alone external stakeholders
such as financial analysts, to precisely predict, ex ante, what the labour costs will be in
the future. Therefore, it is reasonable to argue that the collective bargaining of a labour
union creates uncertainty in a firm’s labour costs, which can have a material, if not
substantial, impact on its profitability and ultimately bottom-line earnings per share
169
(EPS). Consequently, the strong collective bargaining engaged in by organised labour
makes it more difficult for financial analysts to predict future labour costs precisely,
leading to less accurate and more dispersed earnings forecasts for unionised firms.
In addition to the uncertainty in labour costs, labour unions’ ability to initiate large-scale
labour strikes creates further uncertainties in businesses (Ashenfelter and Johnson 1969;
Reder and Neumann 1980; Myers and Saretto 2016). Historically, 15 per cent of labour
contract negotiations have ended in strikes (Tracy 1986). Prior economic and finance
studies suggest that the financial position of the employers and labour market conditions
significantly affect unions’ strike decisions (Reder and Neumann 1980; Tracy 1986;
Cramton and Tracy 1992; Klasa et al. 2009). Despite much scholarly effort looking into
the determinants of labour strikes, however, predicting strike events is extremely
challenging for financial analysts. This is because the decision to strike is not solely
determined by the financial position of the employer, which analysts are typically good
at evaluating. Reder and Neaumann (1980) argue that another factor is the bargaining
styles of the negotiating parties, which again may vary across different industries. While
it is true to say that a strike will normally take place in the middle of a labour dispute,
calling a strike is not the only collective bargaining strategy organised labour can
employ. Employees could continue to work under an expired contract during labour
contract negotiations, which is known as a holdout (Cramton and Tracy 1992; Gu and
Kuhn 1998). Therefore, even if there is serious tension between the employees and
employers during negotiations, it is not a foregone conclusion that the employees will
strike, which makes it even harder for financial analysts to predict such events.
Furthermore, assuming analysts did have privileged access to private information that
suggested a strike was imminent, the exact timing of the strike event would still be
unknown and arguably random, at least from the perspective of the analysts. Moreover,
170
even if analysts could precisely predict the timing, it would be unrealistic to assume
they could, ex ante, accurately quantify and fully capture the economic consequences in
their earnings forecasts for the company. While there is no doubt that strikes are
extremely detrimental for employers, gauging the magnitude, in monetary terms, of the
disruption and damage caused is challenging even ex post. Generally, the cost of a strike
is a function of the number of employees involved, the duration of the stoppage, and the
final settlement reached by the two sides (Becker and Olson 1986). The reality that
financial analysts cannot possibly foresee any of the information regarding a potential
strike implies that the uncertainty posed by a labour union is unlikely to be precisely
accounted for in earnings forecasts, thus leading to lower quality in earnings forecasts.
Moreover, the existence of labour unions can be an obstacle that complicates the
implementation of corporate policies and strategies, even though these decisions may
well be value-creating for the shareholders. For example, organised labour can be very
resistant to firms’ restructuring decisions and cost-cutting strategies, which typically
involve plant closures and labour layoffs (Atanassov and Kim 2009; Chen et al. 2011).
Specifically, Atanassov and Kim (2009) demonstrate that strong unions intervene in the
restructuring process and can effectively avert large-scale layoffs and plant closures,
thus creating operational inflexibilities in the implementation of restructuring decisions.
Labour unions can also affect firms’ innovation strategies. Bradley et al. (2017c)
suggest that labour unions undermine firms’ efforts at research and development,
hindering their innovation. Therefore, we argue that the operating inflexibilities brought
about by labour unions cause uncertainties in the implementation of key corporate
strategies, which can have profound implications for firms’ prospects and shareholders’
wealth (John et al. 2015). Although financial analysts pay close attention to companies’
strategies and policies through corporate disclosure and announcements, the extent to
171
which employees will cooperate with the management in the delivery of those strategies
is arguably difficult for financial analysts to gauge as an external party.
Meanwhile, labour unions could also indirectly affect analysts’ forecast quality through
the financial reporting channel, due to managerial obfuscation. To improve their
bargaining position and undermine union power, managers proactively engage in a
range of strategic corporate decisions to shelter financial resources and obfuscate their
true financial position and earnings prospects from organised labour (DeAngelo and
DeAngelo 1991; Hilary 2006; Klasa et al. 2009; Matsa 2010; Bova 2013; Chung et al.
2016). Specifically, to shelter financial resources, firms strategically adjust their capital
structure by holding less cash (Klasa et al. 2009) and increasing leverage (Matsa 2010;
Myers and Saretto 2016). In addition to altering the data regarding their current
financial position, managers also proactively engage in impression management to
project a less positive view of their future earnings by manipulating earnings (DeAngelo
and DeAngelo 1991; Hamm et al. 2018), narrowly missing analysts’ forecasts (Bova
2013) and withholding good news (Chung et al. 2016). It is worth mentioning that
Hamm et al. (2018) argue that managers, facing a trade-off between sheltering resources
from employees and signalling job security to employees, choose to smooth earnings
optimally. Irrespective of the direction of the earnings management (i.e., whether
deflated or smoothed), the managers are artificially manipulating the actual earnings,
which incorporates distortion and noise into the accounting information. Therefore, the
aforementioned strategic corporate reactions generally lead to a poorer information
environment (Hilary 2006), and project a misleading and opaque image to financial
analysts regarding the firms’ prospects. Assuming financial analysts predominantly play
a “complementary role” (Lang and Lundholm 1996; Altınkılıç et al. 2013), we would
expect them to make their earnings forecasts based on all the publicly available
172
information disclosed by managers. Given the poorer information environment in the
presence of labour unions, we conjecture that analysts’ earnings forecasts will be less
accurate and more dispersed in firms facing strong union representation.
Critically, while we admit that the “uncertainty” and “financial reporting” channels are
not mutually exclusive, and could simultaneously affect financial analysts’ earnings
forecasts in the same direction, the financial reporting channel is ultimately contingent
on managers’ discretion and efforts to preserve their bargaining position against
organised labour. Irrespective of the managerial efforts to preserve information
asymmetry, we argue that the presence of a labour union itself is an inherent source of
uncertainty, difficult for financial analysts to capture fully in their earnings forecasts.
Taking these arguments together, assuming financial analysts primarily play a
“complementary role”, they are likely to be affected by the poor information
environment and significant uncertainty regarding human capital in unionised firms
(Zhang 2006b; Amiram et al. 2018). Hence, we propose our main hypothesis:
Hypothesis 1a (Complementary Role): Labour unionisation is negatively (positively)
associated with earnings forecast accuracy (dispersion).
4.2.2.2 Labour Unions and Financial Analysts: “Substitutive Role”
Alternatively, assuming financial analysts primarily play a “substitutive role”, they
should proactively engage in original research in the firms they are following, and
produce new and value-relevant information that would not otherwise be available to
the investors (Asquith et al. 2005; Barron et al. 2008; Barron et al. 2017; Bradshaw et al.
2017).
The informational role financial analysts play in capital markets is well recognised in
the analyst literature (Chen et al. 2010; Bradshaw et al. 2017; Huang et al. 2017).
173
Focusing on the content of analyst reports, Asquith et al. (2005) present evidence that
they are informative to investors, particularly in the case of downgrades. In an
investigation into analysts’ behaviour after significant forecast failures, Barron et al.
(2008) show that analysts are motivated to dedicate more effort to original research, and
have the ability to generate private information to improve their future earnings
forecasts. Recent studies suggest that financial analysts continue to produce valuable
information for the market, even under adverse circumstances. When uncertainty
increases or the information environment deteriorates, analysts put more focus on
information discovery, and their outputs become more informative, because investors
have a greater demand for high-quality information about firms’ future earnings
prospects (Loh and Stulz 2018; Jennings 2019).
If we assume analysts predominantly engage in the “substitutive role”, in response to
the heightened uncertainty in the workforce and deterioration of the information
environment in unionised firms, we would expect them to devote more resources and
effort to generating first-hand information about the underlying economics of unionised
firms, knowing that such information will be greatly valued by the investors (Loh and
Stulz 2018; Jennings 2019). By putting more effort into producing original reports,
financial analysts may also be rewarded in terms of reputation enhancement and career
progression.
Not only do financial analysts have incentives to provide new information in the context
of organised labour, but they are also capable of and ideally positioned to produce
valuable information highly relevant to a firm’s future performance (Loh and Stulz 2018;
Huang et al. 2018; Jennings 2019). Firstly, financial analysts have exclusive access to
information from multiple channels other than public disclosure (Huang et al. 2018).
When the information disclosed by managers is less credible, or limited, they can
174
generate valuable information from sophisticated research and private interactions with
employees, customers and competitors, in order to form a comprehensive and consistent
picture regarding the underlying economics of the company in question (Soltes 2014;
Huang et al. 2018)59. For example, when there is greater uncertainty about firms’ human
capital, and more complicated employee-employer relations, financial analysts may
obtain valuable insights into employee wellbeing through interactions with rank-and-file
employees. Secondly, financial analysts can use their skills and financial expertise to
analyse and aggregate financial and non-financial information from multiple sources,
providing highly informative outputs to investors (Healy and Palepu 2001; Bradshaw
2011; Dhaliwal et al. 2012; Huang et al. 2018). In addition, financial analysts tend to
specialise in a number of firms within a particular industry, making them experts of a
certain industry. Therefore, they are likely to possess firm or industry-specific
knowledge and insights that may help them to generate valuable information for
investors, with respect to the underlying economics and predicted profitability of
particular firms (Bradley et al. 2017a; Jennings 2019).
Previous literature has also produced empirical evidence consistent with financial
analysts providing incremental information to the capital markets (Chen et al. 2010;
Altınkılıç et al. 2013; Bradley et al. 2017a; Loh and Stulz 2018; Huang et al. 2018;
Jennings 2019). By exploiting the setting of economically bad times, Loh and Stulz
(2018) show that financial analysts are able to provide more valuable information and
accurate forecasts amid heightened uncertainty. In a similar vein, Jennings (2019) finds
that financial analysts generate more informative analyst research following accusations
59 Despite the passing of Regulation Fair Disclosure (RegFD) by the Securities and Exchange
Commission (SEC), which was specifically designed to tackle concern over “offline” interaction between
financial analysts and management, Soltes (2014) finds that financial analysts continue to access material
information from management privately in the post-RegFD period.
175
of managerial misconduct, which significantly undermine the credibility of
management-provided information and drive up the demand for analyst services. In
other words, financial analysts are capable of producing new information, even when
managerial disclosures are less credible or when uncertainty is systematically higher.
Similarly, knowing that investors have a greater information demand in the context of
increased uncertainty regarding human capital, we argue that analysts have greater
incentives to devote more resources and effort to generating more value-relevant
information for unionised firms specifically, in order to meet investors’ information
demands. As a result, we predict that financial analysts’ earnings forecast quality may
be higher for unionised firms due to the increased and dedicated efforts made by the
sell-side analysts. Hence, we propose a competing hypothesis H1b below.
Hypothesis 1b (Substitutive Role): Labour unionisation is positively (negatively)
associated with earnings forecast accuracy (dispersion).
Despite the potentially poorer financial transparency and information environment in
unionised firms, it is plausible that financial analysts, as sophisticated information users
and experts in the industries they specialise in, may well be capable of detecting
earnings manipulation and deciphering the underlying earnings prospects. For example,
Yu (2008) argues that financial analysts have the financial expertise to detect earnings
management, and finds that analyst coverage significantly reduces earnings
management. Focusing on non-GAAP earnings reporting, and comparing that of
managers and analysts, Bentley et al. (2018) reveal that financial analysts scrutinise
managers’ non-GAAP metrics and filter out earnings components that are deemed less
relevant. This evidence implies that financial analysts have the ability to assess and
distinguish the quality of information supplied by managers. Therefore, given analysts’
176
financial sophistication and firm/industry expertise, we propose a null hypothesis that
financial analysts’ earnings forecasts are not affected by labour unions60.
Hypothesis 1null: Labour unionisation is not associated with earnings forecast
accuracy/dispersion.
4.3 Data and Methodology
4.3.1 Data Sources and Sample Construction
Our study uses data from multiple sources. We obtain the labour unionisation data for
the period of 1983-2015 from the Union Membership and Coverage Database (UMCD)
maintained by Hirsch and Macpherson (2003)61. Our sample period starts in 1983, the
first year in which industry-level unionisation data were reported. We access all the
analyst earnings forecast data from the Detailed History File of the Institutional Brokers’
Estimate System (I/B/E/S), for 1983 to 2015. Consistent with prior analyst studies
(Zhang 2006a; Dhaliwal et al. 2012), we use the Detailed History File instead of the
Summary History File to mitigate concerns over stale forecasts and rounding errors
(Diether et al. 2002). Additional firm-level financial information and stock return data
are collected from Compustat and the Centre for Research in Security Prices (CRSP).
After merging the different databases, our baseline sample consists of 93,530 firm-year
observations from 12,744 unique firms, spanning over 30 years from 1983 to 2015.
60 It is also possible that the net effect of the “dual roles” is close to zero, in which case the incremental
value of original research is offset by the overall poorer information environment and uncertainties in
unionised firms, resulting in no systematic difference in forecast quality between unionised and non-
unionised firms.
61 The union data are downloaded from http://www.unionstats.com. Data on the unionisation rates are
drawn from the Current Population Survey and compiled annually by Hirsch and Macpherson (2003). For
more information regarding the construction of this comprehensive database, please see Hirsch and
Macpherson (2003).
177
4.3.2 Main Variables
4.3.2.1 Labour Unionisation Rate
A common challenge in union studies is the lack of a comprehensive firm-level
unionisation database since it is not mandatory for firms to disclose such information
(Klasa et al. 2009; Chen et al. 2011). Following prior literature, we obtain the industry-
level unionisation rate from the UMCD, as mentioned above, as a proxy for the union
strength at the firm level (Hilary 2006; Klasa et al. 2009; Matsa 2010; Chen et al. 2011;
Chen et al. 2012; Chyz et al. 2013; Huang et al. 2017; Hamm et al. 2018).62 Specifically,
the labour unionisation rate (UNION) is measured as the percentage of workers who are
represented by labour unions through collective-bargaining agreements within a three-
digit Census Industry Classification (CIC) industry63 in a given year. The unionisation
rate (UNION) across all CIC industries over the period of 1983-2015 is 11.82 per cent,
which is highly comparable to prior literature (Chen et al. 2011; Huang et al. 2017).
4.3.2.2 Analyst Forecast Variables
In this study, we focus on two of the most common earnings forecast properties in the
financial analyst literature, forecast error and forecast dispersion (Lang and Lundholm
1996; Chen et al. 2017; Mattei and Platikanova 2017). Consistent with previous studies
(Lang and Lundholm 1996; Clement 1999; Duru and Reeb 2002; Dhaliwal et al. 2012),
we use all the earnings forecasts issued by financial analysts in the fiscal year for a
given company to calculate these two variables. Specifically, following Dhaliwal et al.
62The use of the same union database (1) allows us to study the research question based on a larger sample
of firms from the whole spectrum of industries, hence increasing the generalisability of our results, and
(2), more importantly, provides consistency of union data, enabling direct comparison of our findings
with prior studies (Hilary 2006; Chen et al. 2011; Bova 2013; Hamm et al. 2018).
63 We use the crosswalk provided by the U.S. Census Bureau to convert the CIC industry codes into SIC
codes and thereby merge datasets from other sources. The crosswalk file can be accessed at
https://www.census.gov/topics/employment/industry-occupation/guidance/code-lists.html.
178
(2012), forecast error (FERROR) is defined as the average of the absolute errors of all
forecasts scaled by the share price:
𝐹𝐸𝑅𝑅𝑂𝑅𝑖,𝑡 =1
𝑁∑
|𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡𝑒𝑑 𝐸𝑃𝑆𝑖,𝑡,𝑗−𝐴𝑐𝑡𝑢𝑎𝑙 𝐸𝑃𝑆𝑖,𝑡|
𝑆ℎ𝑎𝑟𝑒 𝑃𝑟𝑖𝑐𝑒𝑖,𝑡
𝑁𝑗=1 (1)
where subscripts i, t, and j denote firm i, year t, and forecast j, respectively. Similarly,
consistent with Lang and Lundholm (1996), forecast dispersion (FDISPER) is computed
as the standard deviation of all the forecasts, deflated by the share price:
𝐹𝐷𝐼𝑆𝑃𝐸𝑅𝑖,𝑡 =𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝐷𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 𝑜𝑓 𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡𝑒𝑑 𝐸𝑃𝑆𝑖,𝑡
𝑆ℎ𝑎𝑟𝑒 𝑃𝑟𝑖𝑐𝑒𝑖,𝑡 (2)
where subscripts i and t again denote firm i and year t, respectively. In calculating the
standard deviation, we include companies that are followed by at least two financial
analysts in the year in question (Chen et al. 2017). Since both measures are scaled by
the share price, to avoid extremely small values in the denominator and to make sure
our results are not driven by small stocks, we exclude observations with a share price
below one dollar (Hope 2003b; Horton et al. 2017).
4.3.3 Summary Statistics
Table 1 presents the summary statistics for our baseline sample. The mean (median)
unionisation rate (UNION) is 9.6 (4.8) per cent, which is slightly lower than that in the
original union dataset. This is because our sample is essentially made up of I/B/E/S
firms that are covered by financial analysts, who are less likely to follow and make
earnings forecasts for unionised firms (Hilary 2006). The mean (median) value of
forecast error (FERROR) is 0.052 (0.007) and the mean (median) value of forecast
dispersion (FDISPER) is 0.031 (0.006). The descriptive statistics for both analyst
forecast measures are highly comparable to those reported in Lang and Lundholm
(1996). For example, the mean and median values of forecast error, which is the inverse
179
measure of forecast accuracy, in Lang and Lundholm (1996) are 0.042 and 0.008, which
are similar to the values of 0.052 and 0.007 in our sample. The mean (median) value of
analyst coverage (ANALYST_NUM) is 29.882 (18.000), suggesting that each firm-year
observation is followed by almost 30 financial analysts on average. The variable
definitions are provided in the appendix.
***Insert Table 1 here***
4.3.4 Empirical Models
To examine the influence of labour unions on the quality of analysts’ forecasts, we
estimate the following model:
FORECASTit=α +β1UNIONjt +β2SIZEit +β3MTBit +β4LOSSit +β5EARNSURPit +β6LEVit
+β7RD_EXPit +β8AGEit +β9ZSCOREit +β10SD_INCOMEit +β11SD_STKit
+β12ANALYST_NUMit +Firm FE +Industry×Year FE +State FE + ɛit (3)
We run Model (3) separately for each of the forecast quality measures, i.e., forecast
error (FERROR) and forecast dispersion (FDISPER), as our dependent variable. The
variable of interest is the unionisation rate (UNION). Following Mattei and Platikanova
(2017), we also control for a vector of firm characteristics that may affect analysts’
forecast quality. Apart from conventional firm characteristics such as size and financial
position, we also control for analyst coverage (ANALYST_NUM), as a proxy for the
general information environment (Hilary 2006; Chang et al. 2006; Tan et al. 2011;
Armstrong et al. 2012; Amiram et al. 2016). All variables are defined in the appendix.
Since our key variable, UNION, is measured at the industry level, our estimates are
unlikely to be driven by reverse causality, because there is little economic reason to
believe that the properties of analyst forecasts at the firm level would affect the
unionisation of the workforce across the industry. While reverse causality is less of a
180
concern, it is still possible that our estimates may suffer from omitted variable bias.
Therefore, we include a series of fixed effects to alleviate endogeneity concerns. To this
end, we include firm fixed effects to control for time-invariant firm characteristics that
may affect analyst forecast properties in all specifications. In addition, since our
variable of interest is measured at the industry-year level, we include industry-year
fixed effects to control for time-varying industry factors that may be correlated with our
key variable, the unionisation rate (UNION). This specification ensures that our results
are not confounded or spuriously driven by changes in other unobservable industry-
level factors. Finally, we include state fixed effects to account for state-level economic
and legal conditions, such as RTW legislation, which can seriously undermine unions’
bargaining power (Ellwood and Fine 1987; Chen et al. 2011). Consistent with prior
literature (Chen et al. 2011; Chino 2016; Huang et al. 2017), standard errors are
clustered at the CIC industry level, which is considered more conservative than
clustering at the firm level. Chen et al. (2011) suggest that clustering at the industry
level not only addresses the concern of serial correlation within a firm, but also within
industry groupings, important given that our variable of interest (UNION) is at the
industry level. For robustness, we also cluster standard errors at both industry and year
levels to address potential serial correlations within industry as well as year groups
(Petersen 2009; Chyz et al. 2013).
4.4 Empirical Findings
4.4.1 Baseline Results
4.4.1.1 Labour Unions and Forecast Accuracy
Table 2 presents the results of our baseline regressions on the influence of labour unions
on analyst forecast accuracy. We find consistent evidence that the labour unionisation
181
rate is associated with higher (lower) forecast error (accuracy), which points to sell-side
analysts playing the “complementary role”. The estimates of the control variables are
generally in line with prior literature in terms of having the expected signs. For example,
SIZE is negatively associated with forecast error, consistent with larger firms, in general,
having better information environments, while LOSS is positively significant since it is
more difficult to estimate future earnings for loss-making firms. In addition to our
control variables, to alleviate endogeneity concerns, we include firm fixed effects, to
control for unobservable firm characteristics that may affect analysts’ ability to
accurately forecast earnings, across all OLS specifications. In addition, we include year
fixed effects (Columns 1 and 4), industry-year fixed effects (Columns 2, 3, 5 and 6) and
state fixed effects (Columns 3 and 6), to mitigate the concern of omitted variable bias.
Since the variable of interest, UNION, is an industry-level variable, we cluster standard
errors at the industry level to address serial correlation at that level, across all OLS
regressions. For robustness, standard errors are clustered at both the industry and the
year level in Columns 4-6 (Petersen 2009; Chyz et al. 2013). UNION remains positive
and statistically significant across all specifications.
***Insert Table 2 here***
4.4.1.2 Labour Unions and Forecast Dispersion
We repeat our analysis using forecast dispersion (FDISPER) as the dependent variable,
and we document a positive relationship between unionisation rates and analysts’
forecast dispersion, across all specifications. Given the greater ex-ante uncertainty in
unionised firms and potentially poorer information environment, financial analysts,
primarily playing a “complementary role” in the markets, would be less likely to reach a
consensus with regard to the firms’ future economic performance (Imhoff and Lobo
182
1992; Lang and Lundholm 1996). In other words, had financial analysts, on aggregate,
engaged more in the “substitutive role” by conducting original research into these
unionised firms, as sophisticated information users, they should have been able to gather
the relevant intelligence and gain a better idea of what the future earnings were likely to
be, at least within a reasonable range.
***Insert Table 3 here***
In support of our Hypothesis 1a, the results in Tables 2 and 3 indicate that the labour
unionisation rate is associated with lower forecast accuracy and higher forecast
dispersion, suggesting that financial analysts are affected by the presence of labour
unions. We interpret these results as evidence consistent with financial analysts
predominantly playing a “complementary” rather than “substitutive” role in the capital
markets.
4.4.2 Verification of Channels: Uncertainty versus Financial Reporting Quality
Figure 2 Labour Unions and Analyst Forecasting: Plausible Channels
So far, our baseline results suggest that financial analysts are negatively affected by
labour unions, in the form of lower forecast accuracy and higher forecast dispersion.
While this result is consistent with our H1a, it is unclear whether this effect occurs
through the “financial reporting channel” or the “uncertainty channel”.
183
Prior union literature suggests that unionised firms engage in downward earnings
management and information obfuscation in order to mitigate strike risk and improve
their bargaining position against labour unions (Liberty and Zimmerman 1986; Hilary
2006; Bova 2013; Chung et al. 2016). Therefore, one may reasonably argue that the
union effect on analysts’ forecast quality we document in our baseline models could be
attributable to poorer financial reporting quality due to managerial obfuscation, rather
than “uncertainty” brought about by labour unions. Nonetheless, we argue that the two
channels are not mutually exclusive and predict that the two channels could
simultaneously impact the analyst forecast properties.
To disentangle the two channels and, more importantly, make sure that the labour union
effect on analysts’ forecasting is not solely driven by a poor information environment,
we further test the relationship between labour unions and analyst forecasting by
controlling for financial reporting quality, given that financial analysts rely heavily on
financial information to forecast future earnings. We predict that, after controlling for
financial reporting quality, the key variable, UNION, will remain positive and
statistically significant.
The rationale behind this additional test is that, assuming financial reporting quality is
the only channel through which labour unions can affect analysts’ earnings forecast
properties, controlling for financial reporting quality will essentially make the UNION
variable insignificant, since the entire union effect will be subsumed by the additional
control variables for financial reporting quality. In other words, if the UNION variable
is persistently significant after controlling for financial reporting quality, this will imply
that labour unions have an incremental effect on analysts’ forecast quality, on top of the
unions’ adverse influence on the information environment. Effectively, we disentangle
184
the two plausible channels and can, therefore, interpret with reasonable confidence that
this incremental effect is due to union presence imposing inherent uncertainties on firms.
4.4.2.1 Proxies for Financial Reporting Quality
To account for the financial reporting channel, in our further analysis, we include three
variables to capture the quality of financial reporting: accrual-based earnings
management (Kothari et al. 2005), real earnings management (Roychowdhury 2006;
Cohen et al. 2008) and financial statement comparability (De Franco et al. 2011).
Intuitively, both accrual-based and activities-based earnings management essentially
reduce earnings quality, and therefore the informativeness of accounting information,
whereas financial statement comparability benefits information users by lowering the
costs of processing financial information, and enhancing understanding of financial
information across comparable peers, which is particularly useful for financial analysts
(De Franco et al. 2011; Kim et al. 2016).
Specifically, to capture the level of accrual-based earnings management, we follow
Kothari et al. (2005) and estimate the absolute value of the performance-matched
discretionary accruals (ABS_DA). Following Roychowdhury (2006), we estimate the
abnormal levels of cash flow from operations (R_CFO), discretionary expenses
(R_DISX) and production costs (R_PROD) to proxy for the magnitude of real activities
manipulation. Then, consistent with Cohen et al. (2008), we construct a comprehensive
measure to capture the overall level of real earnings management (Combined_RAM) by
combining the three individual variables (R_CFO, R_PROD and R_DISX). Finally, as
the third proxy for financial reporting quality, we use the financial statement
185
comparability score developed by De Franco et al. (2011) 64 , which measures the
closeness of the accounting systems of two firms. The underlying rationale behind this
measure is that, for any given set of economic events, firms are more likely to produce
similar financial statements if they have more comparable accounting systems.
Essentially, the main comparability measure (CompAcct4)65 is defined as the average
comparability score of the four most comparable peers of a particular firm in a
particular year.
4.4.2.2 Incremental Effect of Union Representation
Table 4 reports the results after controlling for the three abovementioned proxies for
financial reporting quality, first of all separately: (1) accrual-based earnings
management (Columns 1-3), (2) real earnings management (Columns 4-6) and (3)
financial statement comparability (Columns 7-9), and then jointly (Columns 10-12),
using different sets of fixed effects66 . In addition, we include all the other control
variables from the original model described in Equation (3), for consistency.
Panel A presents the relationship between the labour unions and analyst forecast
accuracy (FERROR) after controlling for financial reporting quality. It is critical to note
that our variable of interest, UNION, remains positive and statistically significant after
controlling for the quality of financial reporting to take into account the effect of
managerial obfuscation in the presence of labour unions. In line with our prediction that
earnings management will distort accounting information and lower the informativeness
64 The dataset can be accessed at Rodrigo Verdi’s personal website
(http://mitmgmtfaculty.mit.edu/rverdi/). Detailed descriptions and construction of the data are presented
in De Franco et al. (2011).
65 Our results are robust to an alternative comparability measure (CompAcct10) based on the average
comparability value of the 10 most comparable firms.
66 After matching our baseline sample with all three financial reporting quality variables, our sample size
drops to 27,380. Standard errors are clustered at both the industry level and the year level, which is
considered more rigorous.
186
of accounting information (Dechow et al. 2010), both accrual-based earnings
management (ABS_DA) and real earnings management (Combined_RAM) are
significant and positively associated with forecast errors. In contrast, financial statement
comparability (CompAcct4) is associated with significantly lower earnings forecast
errors, which is consistent with the finding of De Franco et al. (2011). All three
financial reporting quality variables are consistently and highly significant at the 1 per
cent level, with expected signs across all specifications, confirming that labour unions
can affect analysts’ forecasts through the “information channel”. Importantly, the
persistent significance of our variable of interest UNION, after taking into account the
financial reporting quality in unionised firms (Hilary 2006), suggests that labour unions
do have an incremental effect on financial analysts’ forecast precision. Similarly, as
demonstrated in Panel B, we document a significantly positive union effect on earnings
forecast dispersion (FDISPER) after including the additional controls for financial
reporting quality.
While the information channel is indeed a plausible channel whereby managers
strategically preserve information asymmetry to improve their bargaining position
against union representation, our results suggest that managerial information
obfuscation is not the only channel through which labour unions can influence analysts’
forecast quality. We interpret this incremental union effect as evidence consistent with
labour union representation inducing significant uncertainty in businesses, thus
supporting our “uncertainty” conjecture. We argue that the very existence of a labour
union itself constitutes a material source of uncertainty in human capital, which cannot
be precisely modelled in analysts’ earnings forecasts. Therefore, in light of the added
uncertainties created by labour unions, analysts’ earnings forecasts tend to be less
accurate and more dispersed. Unlike the information channel, which relies on managers’
187
efforts at strategic obfuscation of their financial position (Liberty and Zimmerman 1986;
Hilary 2006; Bova 2013; Chung et al. 2016), our results offer a parallel and yet more
direct channel, whereby union presence itself is a source of uncertainty that can directly
affect analysts’ forecast quality. Importantly, unlike Loh and Stulz (2018) and Jennings
(2019), our results suggest that, in the context of union representation, financial analysts
predominantly use the publicly available information, instead of exerting much-needed
effort for unionised firms, where human capital uncertainty is inherently higher.
***Insert Table 4 here***
4.4.3 Right-to-Work (RTW) Legislation
An underpinning assumption behind the union effect is that our results are driven
primarily by the enhanced bargaining power of the employees. Following this logic, the
union effect on financial analysts should be more pronounced when union power is
stronger, and moderated when unions’ bargaining ability is undermined. Following prior
union literature (Chen et al. 2011; Campello et al. 2018), we exploit the exogenous
variation in union power at the state level due to RTW legislation in the U.S., which
seriously undermines labour unions’ bargaining power (Ellwood and Fine 1987).
Specifically, we partition our sample into RTW firms and non-RTW firms based on
whether their headquarters are located in a state that has enacted the RTW law. We
expect the relationship between labour unionisation and analysts’ forecast quality to be
stronger for firms based in non-RTW states, where unions enjoy greater bargaining
power.
Table 5 reports the findings of the subsample analysis comparing RTW and non-RTW
states. In line with our prediction, we find that the union effects on both forecast error
(FERROR) and forecast dispersion (FDISPER) are positively significant in the non-
188
RTW group (Columns 1-3 and Columns 7-9). In contrast, the insignificant results for
the UNION variable, for both forecast properties, in the RTW group (Column 4-6 and
Columns 10-12) suggest that labour unions have little influence on analyst forecast
quality in RTW states where unions’ power is profoundly weakened. Our subsample
analysis based on this exogenous variation in labour unions’ bargaining power supports
our prediction that financial analysts’ forecast quality is significantly affected only
when labour unions possess substantive bargaining power. This is because greater union
power is likely to introduce more uncertainty and trigger more managerial efforts to
preserve information asymmetry. If financial analysts are indeed primarily playing the
role of “information disseminators”, their forecast quality is likely to be impacted to a
greater extent. Therefore, this cross-sectional analysis also lends further assurance that
our main results are indeed driven by the bargaining power of the labour unions rather
than anything else.
***Insert Table 5 here***
4.4.4 Role of Labour Skills
We further study how labour skills may affect the relationship between organised labour
and analysts’ forecast quality. Prior literature in labour economics argues that low-
skilled workers tend to benefit most from labour unions, in terms of both pay
improvement and job security (Farber and Saks 1980; Freeman 1980; Lewis 1986; Card
1996). Compared with high-skilled employees, low-skilled employees are typically at
the bottom of the earnings distribution within the firm, and are exposed to significantly
higher unemployment risk (Farber and Saks 1980; Akerlof and Yellen 1988). Given the
lower pay and higher unemployment risk, low-skilled workers are more reliant on
labour unions to safeguard their jobs and negotiate higher wages on their behalf. To
189
meet the higher expectations and demands from their union members, labour unions
representing low-skilled workers are more likely to engage in collective-bargaining
activities and pursue their agenda aggressively, for example, by initiating large-scale
strikes. Thus, we predict that the union impact on analyst forecast quality should be
stronger in low-skilled industries, where labour unions are expected to play a greater
role and strike risk is perceived to be higher.
To conduct our analysis, we partition our sample into high-skill and low-skill firms
based on labour skills. To proxy for the level of labour skills, we use an industry-
specific Labor Skill Index (LSI), following Ghaly et al. (2017). Essentially, the LSI
captures the weighted average skill level of the occupations within an industry, based on
data from the Occupational Employment Statistics (OES) and O*NET program
compiled by the U.S. Department of Labor. Table 6 presents the results of this
subsample analysis.
In contrast to the insignificant results for firms in high-skilled industries, we find that
the union effect on forecast accuracy is statistically significant in firms that rely heavily
on a low-skilled workforce, which is consistent with firms in low-skilled industries
facing stronger collective bargaining and being more prone to strike threats. As for
forecast dispersion, we do find a significant result for the low-skilled subgroup in
Column 7 and an insignificant result for the high-skilled group with the same
specification (Column 10). However, the UNION variable is insignificant for the low-
skilled subgroup in Columns 8 and 9, under the alternative specifications. One possible
reason for the unsystematic difference in terms of forecast dispersion between the low-
skilled and high-skilled industries is that high-skilled industries tend to have higher
asset intangibility, which may also lead to larger discrepancies in analysts’ forecasts due
to the difficulty of evaluating intangible assets.
190
***Insert Table 6 here***
4.4.5 Mitigating Role of Labour Costs Information
Because of unions’ agenda of pushing for higher salaries for employees, unionised firms
are inevitably exposed to considerable uncertainties in labour costs, which significantly
undermine financial analysts’ ability to predict those costs, which represent a major
expenditure component of the income statement and ultimately the bottom-line earnings
figure in their forecasts. If financial analysts do indeed predominantly rely on readily
available information in the markets (complementary role) rather than proactively
collecting new information from original research (substitutive role), it would be
reasonable to assume that they do not obtain wage data unless companies voluntarily
disclose such information.
Therefore, we argue that information on labour costs would be extremely valuable and
particularly relevant in the context of unionised firms, setting a good benchmark for the
prediction of future labour costs. We predict that the availability of labour cost
information would significantly improve analysts’ capability to predict future labour
costs and hence mitigate the union effect on analyst forecast quality. We thus collect
information on labour-related expense (XLR) on Compustat and create a dummy
variable XLR_Dummy, equal to one for the observations where the variable XLR is
available and zero where it is missing 67 . Thus, we split our sample based on the
availability of labour costs.
Table 7 reports the results for the two subgroups: (1) firms disclosing labour costs
(Columns 1-3 and 7-9) and (2) firms not disclosing labour costs (Columns 4-6 and 10-
67 Since it is not mandatory for firms to disclose information on labour costs such as wages and salaries,
in our final sample about 7 per cent of the observations have wage information (XLR), which is similar to
Hamm et al. (2018).
191
12). While the union effect on both analyst forecast quality proxies persists for firms
that do not disclose labour cost information to the market, we find that the UNION
variable becomes insignificant for firms that do disclose such information, in all
specifications, which supports our conjecture that the availability of wage data will
significantly mitigate the union effect on forecast quality by improving analysts’ ability
to predict future labour costs, a major component of expenses and value-relevant
information for unionised firms.
The interpretation of this finding is two-fold. Firstly, this result further supports the
notion of a predominantly “complementary role” being played by sell-side financial
analysts. By confirming the crucial role of labour costs, our results imply that financial
analysts do rely on publicly disclosed information such as labour costs, as opposed to
generating such information through proactive research, even though, intuitively, labour
expenses would be very informative in the context of unionised firms. Secondly, the
mitigating effect of labour cost information is also consistent with our argument that
labour unions affect analysts’ forecast quality by causing significant uncertainty in
labour costs, therefore lending additional support to our “uncertainty channel”68.
***Insert Table 7 here***
4.4.6 Strategic Optimism Bias
So far, our analyses have mainly focused on how organised labour affects analysts’
forecasts in terms of accuracy and dispersion. Yet, another key dimension of analysts’
behaviour that is worth investigating is their optimism bias. Prior literature has
68 Since one may rightly argue that the disclosure of labour costs is an additional piece of information for
financial analysts, this could also be considered evidence in favour of the “information channel”. We
admit that this test cannot completely disentangle the two channels and therefore serves only as
suggestive evidence of the “uncertainty channel”.
192
established that they are more likely to issue positively biased forecasts when there is
significant uncertainty regarding a firm’s future profitability (Das et al. 1998; Lim 2001;
Zhang 2006a; Bradshaw 2011). By issuing more favourable earnings forecasts, financial
analysts can maintain access to private information from management69. Meanwhile,
financial analysts may also be motivated to be positively biased due to career concerns
(Hong and Kubik 2003; Horton et al. 2017). For example, Hong and Kubik (2003)
discover that optimistic analysts are more likely to achieve career advancement, after
controlling for forecast accuracy.
Building on this view, we argue that financial analysts will behave more strategically
with respect to their earnings forecasts in response to heightened uncertainty in human
capital. Knowing that their forecasts are more likely to be inaccurate due to the
uncertainty and complexity within unionised firms, they will rationally choose to issue
more optimistic forecasts in order to maintain access to private information and mitigate
their career concerns at the same time (Das et al. 1998; Lim 2001). In other words, if
financial analysts are indeed primarily playing a “complementary role”, we would
expect their earnings forecasts to be, on average, more optimistically biased for
unionised firms, whose earnings are less predictable.
To test our conjecture, we construct an indicator variable, Optimism_Bias, which takes
the value of one if the estimated EPS is larger than the actual EPS for the firm-year
observation, and zero otherwise. Specifically, we run probit and logit models 70 with
Optimism_Bias as our dependent variable and UNION as our key independent variable,
69 Both Mayew (2008) and Soltes (2014) suggest that financial analysts continue to access value-relevant
information through private interactions with management, even after the enforcement of RegFD in 2000.
70 Since our dependent variable (Optimism_Bias) is a binary variable, we estimate our probit/logit models
without firm fixed effects (Wooldridge 2010). Instead, we use industry-year fixed effects to account for
time-varying industry-specific characteristics. For brevity, the results of the logit models are not reported.
193
to test the relation between labour unionisation and analysts’ propensity to issue
optimistic forecasts. For robustness and easier interpretation, we repeat our analysis
with an alternative variable of interest, High_UNION, a dummy variable equal to one if
UNION is above sample median.
As presented in Table 8, the coefficients on both UNION (Columns 1 and 2) and
High_UNION (Columns 3 and 4) are consistently positive and statistically significant,
suggesting that financial analysts are more likely to issue optimistic forecasts to
unionised firms, where uncertainty in human capital is higher and the information
environment is poorer. Economically, financial analysts have around a 3 per cent 71
higher propensity to make optimistic earnings forecasts for firms in highly unionised
industries (High_UNION=1), relative to their counterparts in less unionised industries
(High_UNION=0). These results are consistent with financial analysts exhibiting
strategic behaviour by issuing optimistic forecasts in response to the substantial
uncertainty in human capital created by collective-bargaining power.
Collectively, the evidence of strategic optimism and reliance on corporate disclosure of
labour costs, along with the lower quality of analysts’ forecasts, presents a consistent
picture of financial analysts playing a predominantly “complementary role” in the
capital markets in the presence of high uncertainty in human capital.
***Insert Table 8 here***
4.5 Conclusion
In this paper, we examine the primary role of financial analysts in the context of
unionised firms, where investors have a greater information demand and a higher
71 The marginal effects for Column 3 and 4 are 0.035 and 0.026, respectively.
194
reliance on analysts’ research. Using a large U.S. panel dataset over a long sample
period of 1983-2015, we document evidence consistent with financial analysts primarily
playing a “complementary” rather than a “substitutive” role, when firms are subject to
heightened uncertainty in human capital. In line with our argument that labour unions
affect analysts’ forecasting by bringing significant uncertainty into labour costs, we
document that the availability of labour cost information significantly mitigates this
effect on analysts’ earnings forecast quality, in terms of both accuracy and dispersion.
Crucially, the mitigating effect of labour cost information also confirms that financial
analysts rely more on readily available information disclosed by management than on
original information obtained through their own sophisticated research, even though
such information can be extremely relevant and valuable for unionised firms.
Our study adds to the ongoing debate on the primary role of financial analysts in the
information environment of capital markets (Lang and Lundholm 1996; Altınkılıç et al.
2013; Loh and Stulz 2018; Schantl 2018; Huang et al. 2018; Jennings 2019) by offering
new insights into the interplay between financial analysts and an internal stakeholder. In
addition, our paper reveals that employees’ influence extends beyond the company
boundary to a group of sophisticated market participants, i.e., financial analysts, thus
potentially affecting the information environment of capital markets. Thirdly, consistent
with the argument that accounting information is losing value relevance (Lev 2018), our
study suggests that non-financial information on human capital, often neglected or
considered secondary to conventional financial information by analysts and investors, is
informative and would complement the existing financial reporting system. Lastly,
given that such information is highly relevant to investors, regulators and standard
setters may also consider making disclosure on human capital mandatory.
195
This study is subject to some limitations. First, we use the industry-level unionisation
rate to proxy for the firm-level unionisation rate. While this ensures consistency with
previous union studies (Klasa et al. 2009; Chen et al. 2011; Chino 2016; Huang et al.
2017) and greater generalisability of our results, more recent union papers (Bradley et al.
2017c; Campello et al. 2018) exploit the setting of union elections in the U.S. and apply
the quasi-experimental regression discontinuity design to establish the causal impact of
unionisation. Second, in this paper, we mainly focus on the properties of analysts’
earnings forecasts to infer analysts’ primary role, but do not consider other analyst
outputs, such as analyst revisions, target prices or stock recommendations. Third, the
scope of this paper and hence our main finding is limited to the context of organised
labour, and we acknowledge that, in other settings or contexts, analysts may be more
motivated to produce new information rather than disseminating and interpreting public
information. Given the growing awareness of and necessity for stakeholder management,
it is equally important to examine the role of other stakeholders such as suppliers or
customers in the information environment, and how financial analysts interact with
these other important stakeholders.
Overall, our study sheds light on the primary role of financial analysts by focusing on
the interactions between financial analysts and a key stakeholder within businesses, and
highlights the value relevance of an important intangible asset, human capital.
196
References
Aboody, D. and Lev, B. (1998). The Value Relevance of Intangibles: The Case of
Software Capitalization. Journal of Accounting Research, 36(36), pp.161–191.
Akerlof, G.A. and Yellen, J.L. (1988). Fairness and Unemployment. American
Economic Review, 78(2), pp.44–49.
Altınkılıç, O., Balashov, V.S. and Hansen, R.S. (2013). Are Analysts’ Forecasts
Informative to the General Public? Management Science, 59(11), pp.2550–2565.
Amir, E. and Lev, B. (1996). Value-Relevance of Nonfinancial Information: The
Wireless Communications Industry. Journal of Accounting and Economics, 22(1–3),
pp.3–30.
Amir, E., Lev, B. and Sougiannis, T. (2003). Do Financial Analysts Get Intangibles?
European Accounting Review, 12(4), pp.635–659.
Amiram, D., Landsman, W.R., Owens, E.L. and Stubben, S.R. (2018). How Are
Analysts’ Forecasts Affected by High Uncertainty? Journal of Business Finance &
Accounting, 45(3–4), pp.295–318.
Amiram, D., Owens, E. and Rozenbaum, O. (2016). Do Information Releases Increase
or Decrease Information Asymmetry? New Evidence from Analyst Forecast
Announcements. Journal of Accounting and Economics, 62(1), pp.121–138.
Armstrong, C.S., Balakrishnan, K. and Cohen, D. (2012). Corporate Governance and
the Information Environment: Evidence from State Antitakeover Laws. Journal of
Accounting and Economics, 53(1–2), pp.185–204.
Ashenfelter, O. and Johnson, G.E. (1969). Bargaining Theory, Trade Unions, and
Industrial Strike Activity. American Economic Review, 59(1), pp.35–49.
Asquith, P., Mikhail, M.B. and Au, A.S. (2005). Information Content of Equity Analyst
Reports. Journal of Financial Economics, 75(2), pp.245–282.
Atanassov, J. and Kim, E.H. (2009). Labor and Corporate Governance: International
Evidence from Restructuring Decisions. Journal of Finance, 64(1), pp.341–374.
Barron, O.E., Byard, D., Kile, C. and Riedl, E.J. (2002). High-Technology Intangibles
and Analysts’ Forecasts. Journal of Accounting Research, 40(2), pp.289–312.
Barron, O.E., Byard, D. and Yu, Y. (2008). Earnings Surprises that Motivate Analysts
to Reduce Average Forecast Error. The Accounting Review, 83(2), pp.303–325.
Barron, O.E., Byard, D. and Yu, Y. (2017). Earnings Announcement Disclosures and
Changes in Analysts’ Information. Contemporary Accounting Research, 34(1), pp.343–
373.
Barron, O.E. and Stuerke, P.S. (1998). Dispersion in Analysts’ Earnings Forecasts as a
Measure of Uncertainty. Journal of Accounting, Auditing & Finance, 13(3), pp.245–270.
Bartov, E., Givoly, D. and Hayn, C. (2002). The Rewards to Meeting or Beating
Earnings Expectations. Journal of Accounting and Economics, 33(2), pp.173–204.
Becker, B.E. and Olson, C.A. (1986). The Impact of Strikes on Shareholder Equity.
Industrial and Labor Relations Review, 39(3), pp.425–438.
Behn, B.K., Choi, J. and Kang, T. (2008). Audit Quality and Properties of Analyst
197
Earnings Forecasts. The Accounting Review, 83(2), pp.327–349.
Bentley, J.W., Christensen, T.E., Gee, K.H. and Whipple, B.C. (2018). Disentangling
Managers’ and Analysts’ non-GAAP Reporting. Journal of Accounting Research, 56(4),
pp.1039–1081.
Beyer, A., Cohen, D.A., Lys, T.Z. and Walther, B.R. (2010). The Financial Reporting
Environment: Review of the Recent Literature. Journal of Accounting and Economics,
50(2–3), pp.296–343.
Bhat, G., Hope, O.K. and Kang, T. (2006). Does Corporate Governance Transparency
Affect the Accuracy of Analyst Forecasts? Accounting and Finance, 46(5), pp.715–732.
Bova, F. (2013). Labor Unions and Management’s Incentive to Signal a Negative
Outlook. Contemporary Accounting Research, 30(1), pp.14–41.
Bradley, D., Gokkaya, S. and Liu, X. (2017a). Before an Analyst Becomes an Analyst:
Does Industry Experience Matter? Journal of Finance, 72(2), pp.751–792.
Bradley, D., Gokkaya, S., Liu, X. and Xie, F. (2017b). Are All Analysts Created Equal?
Industry Expertise and Monitoring Effectiveness of Financial Analysts. Journal of
Accounting and Economics, 63(2–3), pp.179–206.
Bradley, D., Kim, I. and Tian, X. (2017c). Do Unions Affect Innovation? Management
Science, 63(7), pp.2251–2271.
Bradshaw, M., Ertimur, Y. and O’Brien, P. (2017). Financial Analysts and Their
Contribution to Well-Functioning Capital Markets. Foundations and Trends in
Accounting, 11(3), pp.119–191.
Bradshaw, M.T. (2011). Analysts’ Forecasts: What do we Know after Decades of Work?
SSRN, (June), pp.1–55.
Bronars, S. and Deere, D. (1991). The Threat of Unionization, the Use of Debt, and the
Preservation of Shareholder Wealth. Quarterly Journal of Economics, 106(1), pp.231–
254.
Byard, D., Li, Y. and Weintrop, J. (2006). Corporate Governance and the Quality of
Financial Analysts’ Information. Journal of Accounting and Public Policy, 25(5),
pp.609–625.
Campello, M., Gao, J., Qiu, J. and Zhang, Y. (2018). Bankruptcy and the Cost of
Organized Labor: Evidence from Union Elections. Review of Financial Studies, 31(3),
pp.980–1013.
Card, D. (1996). The Effect of Unions on the Structure of Wages: A Longitudinal
Analysis. Econometrica, 64(4), pp.957–979.
Card, D. (2001). The Effect of Unions on Wage Inequality in the U.S. Labor Market.
Industrial and Labor Relations Review, 54(2), pp.296–315.
Chan, L.K.C., Karceski, J. and Lakonishok, J. (2007). Analysts’ Conflicts of Interest
and Biases in Earnings Forecasts. Journal of Financial and Quantitative Analysis,
42(04), pp.893-914.
Chang, X., Dasgupta, S. and Hilary, G. (2006). Analyst Coverage and Financing
Decisions. Journal of Finance, 61(6), pp.3009–3048.
Chen, H., Kacperczyk, M. and Ortiz-Molina, H. (2012). Do Nonfinancial Stakeholders
198
Affect the Pricing of Risky Debt? Evidence from Unionized Workers. Review of
Finance, 16(2), pp.347–383.
Chen, H.J., Kacperczyk, M. and Ortiz-Molina, H. (2011). Labor Unions, Operating
Flexibility, and the Cost of Equity. Journal of Financial and Quantitative Analysis,
46(01), pp.25–58.
Chen, N.X., Chiu, P.-C. and Shevlin, T. (2018). Do Analysts Matter for Corporate Tax
Planning? Evidence from a Natural Experiment. Contemporary Accounting Research,
35(2), pp.794-829.
Chen, T., Harford, J. and Lin, C. (2015). Do Analysts Matter for Governance? Evidence
from Natural Experiments. Journal of Financial Economics, 115(2), pp.383–410.
Chen, T. and Lin, C. (2017). Does Information Asymmetry Affect Corporate Tax
Aggressiveness? Journal of Financial and Quantitative Analysis, 52(5), pp.2053–2081.
Chen, T., Xie, L. and Zhang, Y. (2017). How Does Analysts’ Forecast Quality Relate to
Corporate Investment Efficiency? Journal of Corporate Finance, 43, pp.217–240.
Chen, X., Cheng, Q. and Lo, K. (2010). On the Relationship between Analyst Reports
and Corporate Disclosures: Exploring the Roles of Information Discovery and
Interpretation. Journal of Accounting and Economics, 49(3), pp.206–226.
Cheng, L. (2017). Organized Labor and Debt Contracting: Firm-Level Evidence from
Collective Bargaining. The Accounting Review, 92(3), pp.57–85.
Chino, A. (2016). Do Labor Unions Affect Firm Payout Policy? Operating Leverage
and Rent Extraction Effects. Journal of Corporate Finance, 41, pp.156–178.
Chung, R., Lee, B.B.-H., Lee, W.-J. and Sohn, B.C. (2016). Do Managers Withhold
Good News from Labor Unions? Management Science, 62(1), pp.46–68.
Chyz, J.A., Ching Leung, W.S., Zhen Li, O. and Meng Rui, O. (2013). Labor Unions
and Tax Aggressiveness. Journal of Financial Economics, 108(3), pp.675–698.
Clark, K.B. (1984). Unionization and Firm Performance: The Impact on Profits, Growth,
and Productivity. American Economic Review, 74(5), pp.893–919.
Clement, M.B. (1999). Analyst Forecast Accuracy: Do Ability, Resources, and Portfolio
Complexity Matter? Journal of Accounting and Economics, 27(3), pp.285–303.
Clement, M.B. and Tse, S. (2005). Financial Analyst Characteristics and Herding
Behavior in Forecasting. Journal of Finance, 60(1), pp.307–341.
Cohen, D.A., Dey, A. and Lys, T.Z. (2008). Real and Accrual-Based Earnings
Management in the Pre- and Post-Sarbanes-Oxley Periods. The Accounting Review,
83(3), pp.757–787.
Connolly, R.A., Hirsch, B.T. and Hirschey, M. (1986). Union Rent Seeking, Intangible
Capital, and Market Value of the Firm. The Review of Economics and Statistics, 68(4),
pp.567–577.
Cotter, J., Tuna, I. and Wysocki, P.D. (2006). Expectations Management and Beatable
Targets: How do Analysts React to Explicit Earnings Guidance? Contemporary
Accounting Research, 23(3), pp.593–628.
Cramton, P. and Tracy, J.S. (1992). Strikes and Holdouts in Wage Bargaining: Theory
and Data. American Economic Review, 82(1), pp.100–121.
199
Das, S., Levine, C.B. and Sivaramakrishnan, K. (1998). Earnings Predictability and
Bias in Analysts’ Earnings Forecasts. The Accounting Review, 73(2), pp.277–294.
DeAngelo, H. and DeAngelo, L. (1991). Union Negotiations and Corporate Policy. A
Study of Labor Concessions in the Domestic Steel Industry During the 1980s. Journal
of Financial Economics, 30(1), pp.3–43.
Dechow, P., Ge, W. and Schrand, C. (2010). Understanding Earnings Quality: A
Review of the Proxies, Their Determinants and Their Consequences. Journal of
Accounting and Economics, 50(2–3), pp.344–401.
De Franco, G., Kothari, S.P. and Verdi, R.S. (2011). The Benefits of Financial
Statement Comparability. Journal of Accounting Research, 49(4), pp.895–931.
Dhaliwal, D.S., Radhakrishnan, S., Tsang, A. and Yang, Y.G. (2012). Nonfinancial
Disclosure and Analyst Forecast Accuracy: International Evidence on Corporate Social
Responsibility Disclosure. The Accounting Review, 87(3), pp.723–759.
Diether, K., Malloy, C.J. and Scherbina, A. (2002). Differences of Opinion and the
Cross-Section of Stock Returns. Journal of Finance, 57(5), pp.2113–2141.
DiNardo, J. and Lee, D. (2004). Economic Impacts of new Unionization on Private
Sector Employers: 1984–2001. Quarterly Journal of Economics, 119(4), pp.1383–1441.
Duru, A. and Reeb, D.M. (2002). International Diversification and Analysts’ Forecast
and Bias. The Accounting Review, 77(2), pp.415–433.
Ellwood, D.T. and Fine, G. (1987). The Impact of Right-to-Work Laws on Union
Organizing. Journal of Political Economy, 95(2), pp.250–273.
Faleye, O., Mehrotra, V. and Morck, R. (2006). When Labor Has a Voice in Corporate
Governance. Journal of Financial and Quantitative Analysis, 41(03), pp.489-510.
Farber, H.S. and Saks, D.H. (1980). Why Workers Want Unions: The Role of Relative
Wages and Job Characteristics. Journal of Political Economy, 88(2), pp.349–369.
Francis, J. and Schipper, K. (1999). Have Financial Statements Lost Their Relevance?
Journal of Accounting Research, 37(2), pp.319-352.
Freeman, R. (1980). Unionism and the Dispersion of Wages. Industrial and Labor
Relations Review, 34(1), pp.3–23.
Freeman, R. (1981). The Effect of Unionism on Fringe Benefits. Industrial and Labor
Relations Review, 34(4), pp.489-509.
Freeman, R. and Medoff, J. (1979). Two Faces of Unionism. Public Interest, 57, pp.69–
93.
Freeman, R. and Medoff, J. (1984). What do Unions do? New York: Basic Books.
Ghaly, M., Dang, V.A. and Stathopoulos, K. (2017). Cash Holdings and Labor
Heterogeneity: The Role of Skilled Labor. The Review of Financial Studies, 30(10),
pp.3636–3668.
Gu, W. and Kuhn, P. (1998). A Theory of Holdouts in Wage Bargaining. American
Economic Review, 88(3), pp.428–449.
Hamm, S.J.W., Jung, B. and Lee, W.-J. (2018). Labor Unions and Income Smoothing.
Contemporary Accounting Research, 35(3), pp.1201–1228.
Hashim, N.A. and Strong, N.C. (2018). Do Analysts’ Cash Flow Forecasts Improve
200
Their Target Price Accuracy? Contemporary Accounting Research, 35(4), pp.1816–
1842.
He, J. and Tian, X. (2013). The Dark Side of Analyst Coverage: The Case of Innovation.
Journal of Financial Economics, 109(3), pp.856–878.
Healy, P.M. and Palepu, K.G. (2001). Information Asymmetry, Corporate Disclosure,
and the Capital Markets: A Review of the Empirical Disclosure Literature. Journal of
Accounting and Economics, 31(1–3), pp.405–440.
Hilary, G. (2006). Organized Labor and Information Asymmetry in the Financial
Markets. Review of Accounting Studies, 11(4), pp.525–548.
Hirsch, B.T. and Macpherson, D.A. (2003). Union Membership and Coverage Database
from the Current Population Survey (CPS): Note. Industrial and Labor Relations
Review, 56(2), pp.349–354.
Hong, H. and Kubik, J.D. (2003). Analyzing the Analysts: Career Concerns and Biased
Earnings Forecasts. Journal of Finance, 58(1), pp.313–351.
Hong, H., Kubik, J.D. and Solomon, A. (2000). Security Analysts’ Career Concerns and
Herding of Earnings Forecasts. The RAND Journal of Economics, 31(1), pp.121–144.
Hope, O.-K. (2003a). Disclosure Practices, Enforcement of Accounting Standards, and
Analysts’ Forecast Accuracy: An International Study. Journal of Accounting Research,
41(2), pp.235–272.
Hope, O.-K. (2003b). Accounting Policy Disclosure and Analysts’ Forecasts.
Contemporary Accounting Research, 20(2), pp.295–321.
Horton, J., Serafeim, G. and Serafeim, I. (2013). Does Mandatory IFRS Adoption
Improve the Information Environment? Contemporary Accounting Research, 30(1),
pp.388–423.
Horton, J., Serafeim, G. and Wu, S. (2017). Career Concerns of Banking Analysts.
Journal of Accounting and Economics, 63(2–3), pp.231–252.
Huang, A.H., Lehavy, R., Zang, A.Y. and Zheng, R. (2018). Analyst Information
Discovery and Interpretation Roles: A Topic Modeling Approach. Management Science,
64(6), pp.2833–2855.
Huang, Q., Jiang, F., Lie, E. and Que, T. (2017). The Effect of Labor Unions on CEO
Compensation. Journal of Financial and Quantitative Analysis, 52(2), pp.553–582.
Imhoff, E.A. and Lobo, G.J. (1992). The Effect of Ex Ante Earnings Uncertainty on
Earnings Response Coefficients. The Accounting Review, 67(2), pp.427–439.
Irani, R.M. and Oesch, D. (2013). Monitoring and Corporate Disclosure: Evidence from
a Natural Experiment. Journal of Financial Economics, 109(2), pp.398–418.
Jennings, J. (2019). The Role of Sell-Side Analysts after Accusations of Managerial
Misconduct. The Accounting Review, 94(1), pp.183–203.
John, K., Knyazeva, A. and Knyazeva, D. (2015). Employee Rights and Acquisitions.
Journal of Financial Economics, 118(1), pp.49–69.
Kim, J.B., Li, L., Lu, L.Y. and Yu, Y. (2016). Financial Statement Comparability and
Expected Crash Risk. Journal of Accounting and Economics, 61(2–3), pp.294–312.
Klasa, S., Maxwell, W.F. and Ortiz-Molina, H. (2009). The Strategic Use of Corporate
201
Cash Holdings in Collective Bargaining with Labor Unions. Journal of Financial
Economics, 92(3), pp.421–442.
Kothari, S.P., Leone, A.J. and Wasley, C.E. (2005). Performance Matched
Discretionary Accrual Measures. Journal of Accounting and Economics, 39(1), pp.163–
197.
Kross, W., Ro, B. and Schroeder, D. (1990). Earnings Expectations: The Analysts’
Information Advantage. The Accounting Review, 65(2), pp.461–476.
Lang, M.H., Lins, K.V. and Miller, D.P. (2003). ADRs, Analysts, and Accuracy: Does
Cross Listing in the United States Improve a Firm’s Information Environment and
Increase Market Value? Journal of Accounting Research, 41(2), pp.317–345.
Lang, M.H. and Lundholm, R.J. (1996). Corporate Disclosure Policy and Analyst
Behavior. The Accounting Review, 71(4), pp.467–492.
Lee, D.S. and Mas, A. (2012). Long-Run Impacts of Unions on Firms: New Evidence
from Financial Markets, 1961-1999. Quarterly Journal of Economics, 127(1), pp.333–
378.
Lev, B. (2018). The Deteriorating Usefulness of Financial Report Information and how
to Reverse it. Accounting and Business Research, 48(5), pp.465–493.
Lewis, H.G. (1964). Relative Employment Effects of Unionism. American Economic
Review, 54(3), pp.123–132.
Lewis, H.G. (1986). Union Relative Wage Effects: A Survey. Chicago: University of
Chicago Press.
Liberty, S.E. and Zimmerman, J.L. (1986). Labor Union Contract Negotiations and
Accounting Choices. The Accounting Review, 61(4), pp.692–712.
Lim, T. (2001). Rationality and Analysts’ Forecasts Bias. Journal of Finance, 56(1),
pp.369–385.
Lin, C., Schmid, T. and Xuan, Y. (2018). Employee Representation and Financial
Leverage. Journal of Financial Economics, 127(2), pp.303–324.
Livnat, J. and Zhang, Y. (2012). Information Interpretation or Information Discovery:
Which Role of Analysts do Investors Value More? Review of Accounting Studies, 17(3),
pp.612–641.
Loh, R.K. and Stulz, R.M. (2018). Is Sell-Side Research More Valuable in Bad Times?
Journal of Finance, 73(3), pp.959–1013.
Mansi, S.A., Maxwell, W.F. and Miller, D.P. (2011). Analyst Forecast Characteristics
and the Cost of Debt. Review of Accounting Studies, 16(1), pp.116–142.
Matsa, D.A. (2010). Capital Structure as a Strategic Variable: Evidence from Collective
Bargaining. Journal of Finance, 65(3), pp.1197–1232.
Mattei, M.M. and Platikanova, P. (2017). Do Product Market Threats Affect Analyst
Forecast Precision? Review of Accounting Studies, 22(4), pp.1628–1665.
Mayew, W.J. (2008). Evidence of Management Discrimination among Analysts during
Earnings Conference Calls. Journal of Accounting Research, 46(3), pp.627–659.
Merkley, K., Michaely, R. and Pacelli, J. (2017a). Cultural Diversity on Wall Street:
Evidence from Sell-Side Analysts’ Forecasts. SSRN Electronic Journal (July).
202
Merkley, K., Michaely, R. and Pacelli, J. (2017b). Does the Scope of the Sell-Side
Analyst Industry Matter? An Examination of Bias, Accuracy, and Information Content
of Analyst Reports. Journal of Finance, 72(3), pp.1285–1334.
Michaely, R. and Womack, K.L. (1999). Conflict of Interest and the Credibility of
Underwriter Analyst Recommendations. Review of Financial Studies, 12(4), pp.653–
686.
Mikhail, M.B., Walther, B.R. and Willis, R.H. (1999). Does Forecast Accuracy Matter
to Security Analysts? The Accounting Review, 74(2), pp.185–200.
Myers, B.W. and Saretto, A. (2016). Does Capital Structure Affect the Behavior of
Nonfinancial Stakeholders? An Empirical Investigation into Leverage and Union
Strikes. Management Science, 62(11), pp.3235–3253.
Pencavel, J. and Hartsog, C.E. (1984). A Reconsideration of the Effects of Unionism on
Relative Wages and Employment in the United States, 1920-1980. Journal of Labor
Economics, 2(2), pp.193–232.
Petaibanlue, J., Walker, M. and Lee, E. (2015). When Did Analyst Forecast Accuracy
Benefit from Increased Cross-Border Comparability Following IFRS Adoption in the
EU? International Review of Financial Analysis, 42, pp.278–291.
Petersen, M.A. (2009). Estimating Standard Errors in Finance Panel Data Sets:
Comparing Approaches. Review of Financial Studies, 22(1), pp.435–480.
Reder, M.W. and Neumann, G.R. (1980). Conflict and Contract: The Case of Strikes.
Journal of Political Economy, 88(5), pp.867–886.
Reuters. (2008). Boeing Set to Announce New 787 Delays. New York. Available at:
https://www.reuters.com/article/us-boeing-idUSTRE4B384G20081204 [Acessed: 1
August, 2018]
Reuters. (2016). Verizon, Unions Agree to Pay Raises, New Jobs to End Strike.
Available at: https://www.reuters.com/article/us-verizon-strike-deal/verizon-unions-
agree-to-pay-raises-new-jobs-to-end-strike-idUSKCN0YL1E7 [Acessed: 1 August,
2018]
Roychowdhury, S. (2006). Earnings Management through Real Activities Manipulation.
Journal of Accounting and Economics, 42(3), pp.335–370.
Ruback, R.S. and Zimmerman, M.B. (1984). Unionization and Profitability: Evidence
from the Capital Market. Journal of Political Economy, 92(6), pp.1134-1157.
Schantl, S.F. (2018). Analyst Information Acquisition and the Relative Informativeness
of Analyst Forecasts and Managed Earnings. Accounting and Business Research, 48(1),
pp.62–76.
Schmidt, M.B. and Berri, D.J. (2004). The Impact of Labor Strikes on Consumer
Demand: An Application to Professional Sports. American Economic Review, 94(1),
pp.344–357.
Soltes, E. (2014). Private Interaction between Firm Management and Sell-Side Analysts.
Journal of Accounting Research, 52(1), pp.245–272.
Tan, H., Wang, S. and Welker, M. (2011). Analyst Following and Forecast Accuracy
after Mandated IFRS Adoptions. Journal of Accounting Research, 49(5), pp.1307–1357.
To, T.Y., Navone, M. and Wu, E. (2018). Analyst Coverage and the Quality of
203
Corporate Investment Decisions. Journal of Corporate Finance, 51(June), pp.164–181.
Tracy, J.S. (1986). An Investigation into the Determinants of U.S. Strike Activity.
American Economic Review, 76(3), pp.423–436.
Wall Street Journal. (2016). Verizon CEO: Strike May Hit Results. Available at:
https://www.wsj.com/articles/verizon-ceo-strike-may-hit-results-1464101544 [Accessed:
1 August, 2018]
Wooldridge, J.M. (2010). Econometric Analysis of Cross Section and Panel Data.
Second Ed. MIT Press, Cambridge, Massachusetts.
Yu, F. (2008). Analyst Coverage and Earnings Management. Journal of Financial
Economics, 88(2), pp.245–271.
Zhang, X.F. (2006a). Information Uncertainty and Analyst Forecast Behavior.
Contemporary Accounting Research, 23(2), pp.565–590.
Zhang, X.F. (2006b). Information Uncertainty and Stock Returns. Journal of Finance,
61(1), pp.105–137.
204
Table 1: Descriptive Statistics
This table presents the descriptive statistics for variables used in our baseline analysis. Variables
are defined in the appendix.
Variable Mean p25 Median p75 SD N
UNION 0.096 0.019 0.048 0.138 0.113 93530
ANALYST_NUM 29.882 7.000 18.000 39.000 34.170 93530
FERROR 0.052 0.002 0.007 0.023 0.222 92259
FDISPER 0.031 0.002 0.006 0.018 0.115 92003
SIZE 6.278 4.905 6.123 7.483 1.899 93387
MTB 3.951 1.215 1.906 3.207 160.473 93339
LOSS 0.242 0.000 0.000 0.000 0.428 93530
EARNSURP 1.158 0.027 0.226 0.659 72.344 88746
LEV 0.222 0.039 0.179 0.341 0.219 92845
RD_EXP 0.044 0.000 0.000 0.043 0.115 93530
AGE 2.224 1.609 2.398 2.996 0.985 93530
ZSCORE 6.172 2.139 3.562 5.946 102.764 72734
SD_INCOME 0.062 0.011 0.026 0.061 0.179 70491
SD_STK 0.031 0.019 0.027 0.038 0.017 93511
205
Table 2: Labour Unionisation Rate and Analysts’ Forecast Accuracy
This table reports the results for the effect of labour unions on analysts’ forecast accuracy. The
dependent variable is forecast error (FERROR). The variable of interest is the unionisation rate
in the firm’s CIC industry (UNION). P-values are displayed in parentheses, with standard errors
clustered at the CIC industry level, in Columns 1-3. For robustness, standard errors are clustered at both the industry and the year level in Columns 4-6. ***, ** and * indicate significance at
1%, 5% and 10%, respectively. All variables are defined in the appendix.
Pooled OLS
(1) (2) (3) (4) (5) (6)
FERROR FERROR FERROR FERROR FERROR FERROR
UNION 0.043** 0.046* 0.059** 0.043*** 0.046** 0.059**
(0.037) (0.091) (0.046) (0.006) (0.035) (0.015)
SIZE -0.035*** -0.036*** -0.034*** -0.035*** -0.036*** -0.034***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
MTB 0.000 0.000 0.000 0.000 0.000 0.000
(0.365) (0.681) (0.897) (0.255) (0.646) (0.887)
LOSS 0.051*** 0.049*** 0.047*** 0.051*** 0.049*** 0.047***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
EARNSURP 0.000 0.000 0.000 0.000 0.000 0.000
(0.314) (0.395) (0.469) (0.287) (0.357) (0.440)
LEV 0.068*** 0.069*** 0.071*** 0.068*** 0.069*** 0.071***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
RD_EXP 0.079 0.079 0.090* 0.079 0.079 0.090*
(0.128) (0.143) (0.096) (0.131) (0.138) (0.095)
AGE 0.019*** 0.018*** 0.019*** 0.019*** 0.018*** 0.019***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
ZSCORE 0.000*** 0.000*** 0.000** 0.000*** 0.000*** 0.000**
(0.000) (0.000) (0.021) (0.000) (0.000) (0.031)
SD_INCOME 0.052* 0.052* 0.061** 0.052* 0.052* 0.061*
(0.050) (0.067) (0.043) (0.070) (0.086) (0.057)
SD_STK 2.040*** 2.110*** 2.146*** 2.040*** 2.110*** 2.146***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
ANALYST_NUM 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Firm FE Y Y Y Y Y Y
Year FE Y N N Y N N
Industry×Year
FE N Y Y N Y Y
State FE N N Y N N Y
Clustered by ind. Y Y Y Y Y Y
Clustered by year N N N Y Y Y
R2 0.521 0.541 0.540 0.521 0.541 0.540
N 52634 52460 48388 52634 52460 48388
206
Table 3: Labour Unionisation Rate and Analysts’ Forecast Dispersion
This table reports the results for the effect of labour unions on analysts’ forecast accuracy. The
dependent variable is forecast dispersion (FDISPER). The variable of interest is the unionisation
rate in the firm’s CIC industry (UNION). P-values are displayed in parentheses with standard
errors clustered at the CIC industry level in Columns 1-3. For robustness, standard errors are clustered at both the industry and year levels in Columns 4-6. ***, ** and * indicate
significance at 1%, 5% and 10%, respectively. All variables are defined in the appendix.
Pooled OLS
(1) (2) (3) (4) (5) (6)
FDISPER FDISPER FDISPER FDISPER FDISPER FDISPER
UNION 0.017 0.031** 0.035** 0.017 0.031** 0.035**
(0.143) (0.030) (0.026) (0.131) (0.021) (0.018)
SIZE -0.022*** -0.022*** -0.022*** -0.022*** -0.022*** -0.022***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
MTB -0.000 -0.000 -0.000 -0.000 -0.000 -0.000
(0.923) (0.250) (0.174) (0.916) (0.220) (0.165)
LOSS 0.024*** 0.023*** 0.023*** 0.024*** 0.023*** 0.023***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
EARNSURP 0.000 0.000 0.000 0.000 0.000 0.000
(0.668) (0.765) (0.955) (0.649) (0.744) (0.954)
LEV 0.027*** 0.027*** 0.029*** 0.027*** 0.027*** 0.029***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
RD_EXP 0.018 0.017 0.021* 0.018** 0.017** 0.021**
(0.115) (0.141) (0.092) (0.025) (0.040) (0.038)
AGE 0.010*** 0.010*** 0.010*** 0.010*** 0.010*** 0.010***
(0.000) (0.000) (0.000) (0.001) (0.001) (0.000)
ZSCORE 0.000 -0.000 0.000** 0.000 -0.000 0.000*
(0.794) (0.870) (0.037) (0.801) (0.889) (0.055)
SD_INCOME 0.018 0.019 0.025 0.018 0.019 0.025
(0.394) (0.394) (0.242) (0.403) (0.396) (0.240)
SD_STK 1.025*** 1.050*** 1.060*** 1.025*** 1.050*** 1.060***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
ANALYST_NUM 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Firm FE Y Y Y Y Y Y
Year FE Y N N Y N N
Industry×Year FE N Y Y N Y Y
State FE N N Y N N Y
Clustered by ind. Y Y Y Y Y Y
Clustered by year N N N Y Y Y
R2 0.576 0.592 0.590 0.576 0.592 0.590
N 52573 52392 48336 52573 52392 48336
207
Table 4: Controlling for Financial Reporting Quality
This table presents the relationship between labour unions and analysts’ forecasting properties, after controlling for financial reporting quality: accrual-based earnings
management (ABS_DA), real earnings management (Combined_RAM) and financial statement comparability (CompAcct4). Panel A reports the results for forecast accuracy (FERROR). Panel B reports the results for forecast dispersion (FDISPER). The variable of interest is the unionisation rate in the firm’s CIC industry (UNION). P-values are
displayed in parentheses with standard errors clustered at both the CIC industry and year levels. ***, ** and * indicate significance at 1%, 5% and 10%, respectively. All
variables are defined in the appendix.
Panel A: Analysts’ Forecast Accuracy
Accrual-Based
Earnings Management (EM)
Real Activities Manipulation
(RAM)
Financial Statement Comparability
(FSC) EM+RAM+FSC
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
FERROR FERROR FERROR FERROR FERROR FERROR FERROR FERROR FERROR FERROR FERROR FERROR
UNION 0.058*** 0.039* 0.046** 0.067*** 0.040** 0.047** 0.071*** 0.045** 0.053*** 0.064*** 0.039** 0.047**
(0.006) (0.053) (0.027) (0.002) (0.042) (0.020) (0.001) (0.021) (0.009) (0.003) (0.039) (0.015)
ABS_DA 0.102*** 0.109*** 0.106*** 0.104*** 0.113*** 0.109***
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Combined_RAM 0.015*** 0.018*** 0.018*** 0.018*** 0.023*** 0.023***
(0.001) (0.003) (0.006) (0.000) (0.001) (0.003)
CompAcct4 -0.012*** -0.012*** -0.016*** -0.011*** -0.011*** -0.016***
(0.001) (0.001) (0.000) (0.001) (0.001) (0.001)
Other controls Y Y Y Y Y Y Y Y Y Y Y Y
Firm FE Y Y Y Y Y Y Y Y Y Y Y Y
Year FE Y N N Y N N Y N N Y N N
Industry×Year FE N Y Y N Y Y N Y Y N Y Y
State FE N N Y N N Y N N Y N N Y
Clustered by ind. Y Y Y Y Y Y Y Y Y Y Y Y
Clustered by year Y Y Y Y Y Y Y Y Y Y Y Y
R2 0.518 0.534 0.526 0.516 0.532 0.525 0.519 0.534 0.528 0.521 0.537 0.531
N 25097 25063 23634 25097 25063 23634 25097 25063 23634 25097 25063 23634
208
Panel B: Analysts’ Forecast Dispersion
Accrual-Based
Earnings Management (EM)
Real Activities Manipulation
(RAM)
Financial Statement Comparability
(FSC) EM+RAM+FSC
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
FDISPER FDISPER FDISPER FDISPER FDISPER FDISPER FDISPER FDISPER FDISPER FDISPER FDISPER FDISPER
UNION 0.030** 0.033** 0.032** 0.035*** 0.034** 0.032** 0.035*** 0.035*** 0.034*** 0.033** 0.033** 0.032***
(0.020) (0.019) (0.016) (0.007) (0.016) (0.012) (0.006) (0.010) (0.006) (0.012) (0.016) (0.009)
ABS_DA 0.043*** 0.044*** 0.040*** 0.045*** 0.048*** 0.044***
(0.002) (0.001) (0.002) (0.001) (0.001) (0.001)
Combined_RAM 0.009** 0.011** 0.011** 0.011*** 0.013** 0.013**
(0.016) (0.030) (0.026) (0.005) (0.013) (0.013)
CompAcct4 -0.004*** -0.003*** -0.005*** -0.004*** -0.003** -0.005***
(0.006) (0.010) (0.006) (0.010) (0.013) (0.008)
Other controls Y Y Y Y Y Y Y Y Y Y Y Y
Firm FE Y Y Y Y Y Y Y Y Y Y Y Y
Year FE Y N N Y N N Y N N Y N N
Industry×Year FE N Y Y N Y Y N Y Y N Y Y
State FE N N Y N N Y N N Y N N Y
Clustered by ind. Y Y Y Y Y Y Y Y Y Y Y Y
Clustered by year Y Y Y Y Y Y Y Y Y Y Y Y
R2 0.604 0.618 0.605 0.603 0.617 0.604 0.603 0.617 0.605 0.605 0.619 0.606
N 25065 25031 23601 25065 25031 23601 25065 25031 23601 25065 25031 23601
209
Table 5: Subsample Analysis: RTW States versus Non-RTW States
This table presents the results of a subsample analysis between firm-year observations in Right-to-Work (RTW) and non-RTW states, based on whether a firm’s
headquarters are located in a state that has passed the RTW legislation. Columns 1-6 report the results for forecast accuracy (FERROR). Columns 7-12 report the results
for forecast dispersion (FDISPER). P-values are displayed in parentheses, with standard errors clustered at both the CIC industry and year levels. ***, ** and * indicate
significance at 1%, 5% and 10%, respectively. All variables are defined in the appendix.
Forecast_Error (FERROR) Forecast_Dispersion (FDISPER)
Non-RTW States RTW States Non-RTW States RTW States
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
UNION 0.079*** 0.054** 0.061** 0.027 -0.048 -0.048 0.044*** 0.042** 0.039** 0.005 0.000 0.000
(0.007) (0.028) (0.013) (0.464) (0.467) (0.467) (0.009) (0.021) (0.020) (0.827) (0.994) (0.994)
All controls Y Y Y Y Y Y Y Y Y Y Y Y
Firm FE Y Y Y Y Y Y Y Y Y Y Y Y
Year FE Y N N Y N N Y N N Y N N
Industry×Year FE N Y Y N Y Y N Y Y N Y Y
State FE N N Y N N Y N N Y N N Y
Clustered by ind. Y Y Y Y Y Y Y Y Y Y Y Y
Clustered by year Y Y Y Y Y Y Y Y Y Y Y Y
R2 0.515 0.536 0.528 0.545 0.579 0.579 0.588 0.605 0.585 0.658 0.693 0.693
N 18396 18334 16905 6679 6603 6603 18369 18311 16881 6674 6598 6598
210
Table 6: Subsample Analysis: Low-Skilled Industries versus High-Skilled Industries
This table presents the results of a subsample analysis between firm-year observations in low-skilled and high-skilled groups, based on whether their labour skill index (LSI)
(Ghaly et al. 2017) is below the sample median for the year. Columns 1-6 report the results for forecast accuracy (FERROR). Columns 7-12 report the results for forecast dispersion (FDISPER). P-values are displayed in parentheses with standard errors clustered at both the CIC industry and year levels. ***, ** and * indicate significance at
1%, 5% and 10%, respectively. All variables are defined in the appendix.
Forecast_Error (FERROR) Forecast_Dispersion (FDISPER)
Low-Skilled Industries High-Skilled Industries Low-Skilled Industries High-Skilled Industries
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
UNION 0.074* 0.068** 0.082** 0.060 -0.125 -0.098 0.036** 0.029 0.026 0.072 0.041 0.040
(0.068) (0.045) (0.036) (0.576) (0.275) (0.461) (0.029) (0.186) (0.266) (0.306) (0.535) (0.585)
All Controls Y Y Y Y Y Y Y Y Y Y Y Y
Firm FE Y Y Y Y Y Y Y Y Y Y Y Y
Year FE Y N N Y N N Y N N Y N N
Industry×Year FE N Y Y N Y Y N Y Y N Y Y
State FE N N Y N N Y N N Y N N Y
Clustered by ind. Y Y Y Y Y Y Y Y Y Y Y Y
Clustered by year Y Y Y Y Y Y Y Y Y Y Y Y
R2 0.537 0.565 0.550 0.546 0.554 0.548 0.647 0.669 0.646 0.659 0.665 0.648
N 8378 8362 7767 7624 7607 6967 8368 8352 7759 7615 7598 6956
211
Table 7: Subsample Analysis: Labour Costs Channel
This table presents results for a subsample analysis between firms that disclose labour costs (XLR_Dummy=1) and firms that do not (XLR_Dummy=0), based on
whether labour-related expense (XLR) is reported in the Compustat database. Columns 1-6 report the results for forecast accuracy (FERROR). Columns 7-12 report the results for forecast dispersion (FDISPER). P-values are displayed in parentheses, with standard errors clustered at both the CIC industry and year
levels. ***, ** and * indicate significance at 1%, 5% and 10%, respectively. All variables are defined in the appendix.
Forecast_Error (FERROR) Forecast_Dispersion (FDISPER)
(XLR_Dummy=0) (XLR_Dummy=1) (XLR_Dummy=0) (XLR_Dummy=1)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
UNION 0.071*** 0.059*** 0.064*** -0.049 -0.065 -0.055 0.038*** 0.044*** 0.042*** -0.029 -0.069 -0.051
(0.001) (0.008) (0.003) (0.373) (0.311) (0.410) -0.005 -0.003 -0.002 -0.211 -0.237 -0.416
All controls Y Y Y Y Y Y Y Y Y Y Y Y
Firm FE Y Y Y Y Y Y Y Y Y Y Y Y
Year FE Y N N Y N N Y N N Y N N
Industry×Year FE N Y Y N Y Y N Y Y N Y Y
State FE N N Y N N Y N N Y N N Y
Clustered by ind. Y Y Y Y Y Y Y Y Y Y Y Y
Clustered by year Y Y Y Y Y Y Y Y Y Y Y Y
R2 0.523 0.538 0.535 0.611 0.630 0.630 0.600 0.614 0.608 0.741 0.727 0.727
N 23333 23293 22280 1719 1558 1134 23304 23264 22252 1717 1556 1130
212
Table 8. Labour Unions and Analyst Optimism
This table presents results for the relation between the labour unionisation rate and analysts’
propensity for issuing optimistic earnings forecasts, based on a probit model. The dependent
variable is Optimism_Bias, which takes the value of one if the estimated EPS issued by the
analysts is larger than the actual EPS, and zero otherwise. The variable of interest is UNION in Columns 1-2. For robustness, in Columns 3-4, we use a dummy variable, High_UNION,
which is equal to one if the labour unionisation rate is above the sample median, and zero
otherwise. All regression models include industry-year fixed effects. P-values are displayed in parentheses, with standard errors clustered at the CIC industry level. ***, ** and *
indicate significance at 1%, 5% and 10%, respectively. All variables are defined in the
appendix.
Optimism_Bias Optimism_Bias Optimism_Bias Optimism_Bias
(1) (2) (3) (4)
UNION 0.636*** 0.448**
(0.003) (0.017) High_UNION 0.095*** 0.078**
(0.009) (0.014)
SIZE -0.182*** -0.182***
(0.000) (0.000)
MTB -0.000 -0.000
(0.855) (0.836)
LOSS 0.727*** 0.727***
(0.000) (0.000) EARNSURP -0.000** -0.000**
(0.037) (0.038)
LEV 0.310*** 0.313***
(0.000) (0.000) RD_EXP 0.060 0.073
(0.772) (0.722)
AGE 0.057*** 0.057***
(0.005) (0.005)
ZSCORE -0.003 -0.003
(0.183) (0.189) SD_INCOME -0.307*** -0.305***
(0.004) (0.004)
SD_STK -3.223** -3.263**
(0.018) (0.017) ANALYST_NUM 0.005*** 0.005***
(0.000) (0.000)
ABS_DA -0.379*** -0.372***
(0.008) (0.010)
Combined_RAM 0.189*** 0.189***
(0.000) (0.000) CompAcct4 -0.017 -0.017
(0.216) (0.211)
Industry×Year FE Y Y Y Y
Pseudo-R2 0.070 0.139 0.070 0.139 N 27016 25669 27016 25669
213
Appendix
Definition of Variables
Variable Definition
UNION Industry-level unionisation rate, defined as the percentage of employees
represented by labour unions in a specific industry.
ANALYST_NUM Number of financial analysts following the firm
FERROR Forecast error, defined as the average absolute value of the difference between estimated and actual EPS for all the earnings forecasts made for
the firm within the 12 months of the earnings announcement, scaled by the
share price at year t
FDISPER Forecast dispersion, defined as the standard deviation of all the earnings forecasts made for the firm within the 12 months of the earnings
announcement, scaled by the share price at year t
SIZE The logarithm of a firm’s market value of equity
MTB The market value of equity divided by the book value of equity LOSS An indicator variable equal to one for negative actual earnings per share
before extraordinary items and zero otherwise
EARNSURP Earnings surprises, defined as the absolute difference between income before extraordinary items at time t and income before extraordinary items
at time t-1, divided by income before extraordinary items at time t-1
LEV Total debt divided by total assets
RD_EXP Research and development expense divided by total assets
AGE Firm age, measured as the logarithm of the difference between the current year and the year when the firm appeared in CRSP for the first time
ZSCORE Altman Z Score=1.2(working capital/total assets) +1.4(retained
earnings/total assets) + 3.3(EBIT/total assets) + 0.6(market value of
equity/book value of total liabilities) + (sales/total assets) SD_INCOME Standard deviation of return on assets over the past five years
SD_STK Standard deviation of return over a 365-day period prior to the fiscal year-
end
ABS_DA Absolute value of performance-matched discretionary accruals, computed using the Modified Jones Model (Kothari et al. 2005)
Combined_RAM The sum of the standardized three real earnings management proxies, i.e.,
abnormal levels of cash flow from operations (R_CFO), discretionary expenses (R_DISX) and production costs (R_PROD) (Cohen et al. 2008)
CompAcct4 Firm-specific financial statement comparability score, measured as the
average score of the four peer firms with the highest comparability scores
(De Franco et al. 2011)
RTW Dummy variable equal to one if the firm is headquartered in a state that has passed Right-to-Work legislation
LowSkill Dummy variable equal to one if the industry-level labour skills index
developed by Ghaly et al. (2017) is below the sample median for the year, and zero otherwise
XLR_Dummy Dummy variable equal to one if the labour-related expense variable (XLR)
is available, and zero otherwise
Optimism_Bias Dummy variable equal to one if the average estimated EPS is larger than the actual EPS for all the earnings forecasts made for the firm within the 12
months of the earnings announcement, and zero otherwise
High_UNION Dummy variable equal to one if the industry’s labour unionisation rate is
above the sample median, and zero otherwise
214
Chapter 5
Summary and Suggestions for Future Research
This thesis explores the role of employee activism in the financial markets. The purpose
of this thesis is to enhance our understanding of the impact of employee activism and to
shed light on potential strategies to mitigate such risk. The overall conclusion is that
employees, as both a powerful stakeholder and a valuable intangible asset, exert
significant influence on corporate decisions and the information environment of the
capital markets. Hence, managers should make greater efforts to manage the
increasingly complex employee relations and associated risks, while financial analysts
and investors should pay more attention to disclosure that is specifically related to
employee risks. Below, I provide a summary of the key findings of the three chapters in
this thesis, along with implications and suggested directions for future research.
In Chapter 2, I investigate the impact of employee stock options (ESO) on union strike
risk. By exploiting the exogenous variation in labour power resulting from union
elections, I find that firms offering high levels of ESO incentives to employees are less
likely to be subject to strikes after unionisation, relative to their low-ESO counterparts. I
interpret this moderating role of ESO on unions’ strike propensity as evidence
consistent with ESO realigning the interests of employees and firms. Subsequent
analysis indicates that firms strategically grant more ESO incentives in reaction to
unionisation events. This strategic adjustment is more salient for firms facing stronger
union power, and hence having a greater need to manage labour risk by improving
interest alignment. My paper implies that, in the context of labour-intensive industrial
firms, employee ownership has the potential to fundamentally transform the labour-
management relationship, and can be adopted as an effective tool for management
against strike risk. Finally, my study suggests that the current accounting treatment (i.e.,
215
FAS 123R) creates a barrier to the expansion of employee ownership schemes, and thus
calls for favourable policies that will promote employee ownership in the highly
unionised manufacturing sector and thereby facilitate the much-needed revitalisation of
this strategically important sector.
However, in this paper, I focus on just one, albeit popular, type of employee ownership
scheme. Following the implementation of FAS 123R, many companies are choosing to
grant more restricted stock units (RSU), a hybrid of stock options and restricted stock,
to circumvent the unfavourable accounting treatment. Further research could look at
whether other employee ownership plans such as RSU have similar effects on
employees. In addition, my analysis is limited to a U.S. sample, yet employee
ownership schemes have gained prevalence in other parts of the world. It would,
therefore, be interesting to examine whether the effect of employee ownership varies
across different institutional environments, and extend the analysis into an international
sample.
In Chapter 3, I study the interplay amongst stakeholders through the lens of organised
labour, in the light of the notable corporate social responsibility (CSR) phenomenon in
the past decade. Specifically, I explore organised labour’s attitude towards firms’
spending on CSR projects. I find that firms with high levels of non-employee CSR
spending are exposed to a significantly higher risk of union strikes, while those with
high levels of employee-related CSR spending are less likely to incur strikes. These
opposing attitudes of organised labour towards employee and non-employee CSR
spending suggest that such spending can exacerbate resource competition between
employees and other stakeholders. I also show that firms strategically cut CSR
expenditure in non-employee dimensions following unionisation, in order to preserve
their bargaining position against labour unions, though such downward adjustments are
216
less pronounced in firms with strong incentives to signal their quality through CSR
spending. Overall, this study reveals an unintended consequence of CSR spending,
namely, resource competition amongst stakeholders, and sends an alarming message to
managers and regulators, amid the growing demand for stakeholder management.
Importantly, instead of treating different stakeholders as a homogeneous group,
managers would be advised to carefully review their relationships with different
stakeholders and take a balanced approach to stakeholder management. Finally, my
study is also relevant to policymakers, highlighting the urgent need for greater
standardisation and regulation on CSR disclosure to enhance the transparency and
scrutiny of managers’ CSR spending decisions.
However, a common challenge in CSR studies is data limitations. Since firms are not
legally required to report the amount they spend on CSR in dollar terms, I have had to
use data on CSR performance as a proxy. While I did collect the CSR ratings from one
of the largest and most credible data providers, and one commonly used in the CSR
literature, inevitably, using CSR performance as a proxy for CSR spending assumes a
monotonic and positive linear relationship, which could introduce bias and
measurement error into the analysis. As an alternative, further study in this area could
use philanthropic donations as a proxy for firms’ financial commitment to stakeholders.
The recent surge in CSR reporting regulations in a number of countries provides a good
setting in which to explore many important questions on the impact of mandatory
adopting of CSR reporting on employee issues. It would be worth probing whether
particular features (e.g., tone or length) of CSR reports have implications for employees’
work attitudes and productivity.
In Chapter 4, I examine whether financial analysts, as professional information
intermediaries, are affected by the collective bargaining power of organised labour. I
217
present evidence that labour unions have a negative impact on analyst forecast quality,
as measured by forecast accuracy and dispersion. I interpret this evidence as being
consistent with financial analysts primarily serving a complementary as opposed to a
substitutive role in the context of unionised firms, where investors have a greater
demand for informative analyst output. Further analysis indicates that the disclosure of
labour costs significantly mitigates unions’ negative impact on analysts’ forecasts,
suggesting that financial analysts do rely more on public disclosure than original
research. This evidence also implies that variability in employee salaries is one
dimension of uncertainty that labour unions bring to firms. Last but not least, I present
evidence that analysts are more likely to issue optimistic forecasts for firms in highly
unionised industries, as a strategic response to the substantial uncertainty in human
capital. This study adds to the ongoing discussion on the primary role of financial
analysts, amid the technological advances in information systems and the financial
media. My findings provide important insights into the impact employees have, beyond
their firms, on a group of sophisticated market participants, financial analysts. This
paper also suggests that human capital information, typically considered secondary to
financial statements, can be value-relevant and useful to investors. Therefore, this
research calls for more disclosure regarding human capital to improve the information
environment of the capital markets.
My findings are limited to unionised employees, typically in labour-intensive industrial
firms. To complete our understanding of the influence employees have on the
information environment, further studies could look at whether highly skilled
employees in the so-called high-tech industries have an even greater impact on analyst
forecasts. Additionally, this study focuses on the quality of the analyst forecasts, yet it
218
would also be relevant to focus on other activities or outputs of financial analysts, such
as revisions, target price and analyst reports.
Taken together, the three studies in this thesis present a consistent picture that
employees are influential participants in the capital markets, and they have important
and timely implications for managers, investors, financial analysts, regulators and
policymakers. While the growing dependence on human capital, particularly highly
skilled employees, poses many new challenges to companies, it opens up many exciting
opportunities for future research. In today’s complex business environment full of
market competition, political uncertainty and technological change, the traditional
“shareholder value maximisation” model appears to be obsolete and no longer
appropriate for businesses in the 21st century. Stakeholders such as customers, suppliers
and the community are becoming increasingly powerful. Understanding their potential
impact on corporate decisions, and their interactions with market participants, would
also be of great interest.
