The Importance of Industry Links in Merger Waves
KENNETH R. AHERN and JARRAD HARFORD⋆
ABSTRACT
We represent the economy as a network of industries connected through customer and supplier
trade flows. Using this network topology, we find that stronger product market connections lead
to a greater incidence of cross-industry mergers. Second, mergers propagate in waves across the
network through customer-supplier links. Merger activity transmits to close industries quickly
and to distant industries with a delay. Finally, economy-wide merger waves are driven by merger
activity in industries that are centrally located in the product market network. Overall, we show
that the network of real economic transactions helps to explain the formation and propagation of
merger waves.
Journal of Finance , forthcoming
⋆Kenneth Ahern is at the University of Southern California, Marshall School of Business. JarradHarford is at the University of Washington, Foster School of Business. We thank two anony-mous referees, an anonymous associate editor, Cam Harvey (editor), Sugato Bhattacharyya, HansDegryse, Ran Duchin, Gerard Hoberg, Jonathan Karpoff, Han Kim, Vojislav Maksimovic, DavidMcLean, Sara Moeller, Gordon Phillips, Ed Rice, Matthew Rhodes-Kropf, David Robinson, ShawnThomas, Karin Thorburn and seminar participants at the 2010 Texas Finance Festival, EuropeanSummer Symposium in Financial Markets (2010) First European Center for Corporate ControlStudies Workshop (2010), 2010 Frontiers in Finance Conference, 2011 Washington University Con-ference on Corporate Finance, 2011 AFA Meetings, 2011 UBC Winter Finance Conference, GeorgiaState University, University of Illinois, University of Maryland, University of Michigan, Universityof North Carolina – Chapel Hill, University of Pittsburgh, and University of Wisconsin for helpfulsuggestions. We also thank Jared Stanfield for excellent research assistance. This research waspartially completed while Kenneth Ahern was at the University of Michigan.
1
A growing body of evidence shows that industry characteristics affect many firm decisions, includ-
ing financial policy (MacKay and Phillips, 2005), internal capital markets (Lamont, 1997), and
corporate governance (Giroud and Mueller, 2010). This line of research emphasizes that strategic
interactions between firms and their industry rivals have important implications for fundamental
questions in financial economics. We broaden this analysis by making a simple, though consequen-
tial, observation: Industries do not exist in isolation, but rather are connected through a complex
network of customer-supplier relationships. This implies that whole industries may be affected
by shocks that are transmitted through the customer-supplier network. In this paper, we inves-
tigate how inter-industry relations affect the timing and incidence of one of the most important
phenomena in corporate finance: merger waves.
The industry network model of an economy has at least three new implications for merger waves.
First, industry-level economic shocks could lead to cross-industry vertical merger waves. Though
it is well documented that merger waves occur within industries (Mitchell and Mulherin, 1996;
Maksimovic and Phillips, 2001; Rhodes-Kropf, Robinson, and Viswanathan, 2005), vertical merger
waves may be just as common. Second, merger waves could propagate through customer and
supplier links without direct vertical integration. For instance, the reorganization of a supplier
industry could cause a customer industry to reorganize in response. Third, the structure of the
industry network could determine how industry-level M&A activity aggregates into an economy-
wide merger wave. These implications are important for understanding how economic fundamentals
at the industry-level influence economy-wide outcomes.
To test these three implications, we empirically model the product market network using input-
output data from the Bureau of Economic Analysis. These data provide trade flows between 471
industries accounting for all sectors in the economy. Using these industry definitions, we also create
a network representing cross-industry mergers over the period 1986 to 2010, where the strength of
the connection between two industries is proportional to the level of their cross-industry merger
activity. Thus, for a comprehensive set of industries, we define two different types of inter-industry
connections: input-output trade flows and cross-industry mergers.
We first characterize the product market and merger networks. We find that both networks
are sparse, but highly inter-connected through a relatively small set of centralized ‘hub’ industries.
2
To illustrate, more than 95% of industry-pairs in the product market network have almost no
customer-supplier relations. Similarly, all cross-industry mergers in our sample occur in just 6% of
all possible industry-pairs. This means that the average industry engages in mergers with a small
set of local industries that are closely related through customer-supplier links. We also find that
the product market and merger networks both exhibit small-world properties, where the average
industry is separated from most other industries by only two or three direct connections, even across
471 different industries. In addition, we find that an industry’s centrality in the product market
network is correlated with its centrality in the merger network, as are other network characteristics,
such as clustering and average distance. Thus, the structure of the merger network is highly similar
to the structure of the product market network.
This characterization of the product market and merger networks supports our first finding
that vertical mergers are common and highly clustered in a relatively small set of directly-linked
industry-pairs. Of the 51,002 mergers in the sample, 61% are inter-industry mergers. Prior research
identifies many reasons for vertical mergers.1 Neoclassical theory proposes that vertical mergers may
eliminate an existing inefficiency, such as double price markups in successive monopolies (Spengler,
1950; Perry, 1978b) or input substitution (Vernon and Graham, 1971; Schmalensee, 1973; Warren-
Boulton, 1974). Another neoclassical motive for vertical mergers is to prevent resale of an input
in downstream industries in order to allow price discrimination across different price elasticities of
demand (Perry, 1978a; Katz, 1987). An alternative to the neoclassical theory, transaction costs
may lead to vertical integration if the net benefits of internal transactions are larger than those of
transacting in a market (Coase, 1937; Williamson, 1979). The costs of market transactions and the
corresponding holdup problems increase with uncertainty and with relationship-specific investments
(Klein, Crawford, and Alchian, 1978). Thus, firms with complementary assets may merge with each
other to overcome incomplete contracts (Rhodes-Kropf and Robinson, 2008).
We find evidence consistent with transaction cost theories, while showing that input-output trade
flows predict cross-industry mergers. We estimate exponential random graph models (ERGM),
which are multivariate maximum likelihood regressions developed specifically to allow for simulta-
neous dependence relations between all nodes in a network. The ERGM results show that there
1Comprehensive surveys of the motives for vertical integration can be found in Tirole (1988) and Perry (1989).
3
are more inter-industry mergers between two industries when they have stronger customer-supplier
relations, controlling for industry valuation, scope, size, returns, concentration, and macroeconomic
shocks. We also find that cross-industry mergers are more likely when industries have greater R&D
expenditures and that R&D magnifies the effects of product market links. To the degree that R&D
proxies for incomplete contracts, these results are consistent with holdup problems. In addition, we
find evidence that cross-industry mergers are positively related to asset complementarity, following
Hoberg and Phillips (2010b). We are careful to note that we do not claim to separately identify
each motivation for vertical mergers, but instead, we provide evidence that shows that product
market trade flows have a first-order effect on the incidence of cross-industry mergers.
The relations between product market links and mergers are economically significant. Industry
pairs without a meaningful economic connection have, on average, 0.11 mergers between them over
the sample period. Those with a strong connection have an average of 12.5 mergers. This effect is
present in every year from 1986 to 2010 and is stronger during market booms and aggregate merger
waves. This implies that economic fundamentals are more, not less, important during merger waves.
The second implication of the industry network model is that merger waves could propagate
through customer and supplier links without direct vertical integration. Galbraith (1952) predicts
that industry consolidation in an upstream industry leads to consolidation in a downstream industry
to counteract the monopoly power created through the initial consolidation. More recent theoretical
industrial organization models predict that changes in the substitutability of products or changes
to the cost structure of one industry affect the incentives to merge for firms in vertically related
industries (Horn and Wolinsky, 1988; Inderst and Wey, 2003). Thus, merger activity could be
transmitted through economic links between industries, even without vertical integration.
Consistent with this, we find evidence that mergers propagate across the industry network fol-
lowing a wave-like pattern. We measure each industry’s exposure to merger activity in related
industries, not including mergers with the industry itself. We use graph theory techniques to iden-
tify which industries are close and which are distant in the product market network. Accounting for
a number of controls, including industry fixed effects and an industry’s own lagged merger activity,
we find that mergers in close industries have a strong positive effect on an industry’s own merger
activity after a one-year delay, while merger activity in distant industries has a positive impact
4
after a delay of two or three years. Thus, merger waves travel across customer-supplier links, even
without direct vertical integration. We also find that the impact of mergers in supplier industries is
larger and travels faster across the network, than does the impact of mergers in customer industries.
This likely reflects the fact that the supplier network is more densely connected.
In the last section of the paper, we investigate the third implication of the network perspective:
the structure of the industry network could determine how industry-level M&A activity aggregates
into an economy-wide merger wave. In vector autoregressions, we find that the industries that
experience merger waves during the height of overall economy-wide merger activity are the most
central industries in the product market network. This is a direct consequence of the highly skewed
distribution of inter-industry connections. As merger activity transmits across the network towards
more central industries, many overlapping industry waves occur, which produces an aggregate
merger wave. This evidence contradicts the idea that industry merger activity caused by random
shocks do not cluster in time, and therefore cannot explain economy-wide aggregate merger waves
(Shleifer and Vishny, 2003). Our evidence suggests that even if the initial industry shocks are
random, aggregate merger waves occur, in part, because of the structure of the industry network.
This paper makes two primary contributions to the literature. First, this paper is related to
recent research that investigates the role of industry relations in corporate finance. Bhattacharyya
and Nain (2011) study the price effects on suppliers and customers following horizontal mergers.
Becker and Thomas (2010) examine how changes in concentration in downstream industries affect
concentration in upstream industries. Fee and Thomas (2004) and Shahrur (2005) use vertical
relations to test the effects of horizontal mergers on market power, building from Eckbo (1983) and
Stillman (1983). Hertzel, Li, Officer, and Rodgers (2008) find that suppliers to firms that file for
bankruptcy suffer negative and significant wealth effects. Our paper is the first to focus on the
role of input-output connections for cross-industry mergers. Although it is generally accepted that
some mergers are motivated by vertical integration, very little about vertical mergers has actually
been documented. Fan and Goyal (2006) report that prior to their paper, even basic facts such as
the proportion of mergers that are vertical were unknown. Our paper is unique in that we study
the determinants of the incidence and timing of inter-industry mergers across all industries, rather
than the value implications of the mergers that do occur. Our paper is also related to a strain
5
of recent research on merger waves, including Maksimovic, Phillips, and Yang (2010), Duchin and
Schmidt (2012), Garfinkel and Hankins (2011), and Ovtchinnikov (2010).
The second contribution of this paper is to model the economy as a network of customer and
supplier relations. This approach is related to Hoberg and Phillips (2010a, 2010b), who use
network techniques to group firms based on textual product market descriptions. In our paper,
we exploit the input-output trade flows to model network ties based on exogenous real economic
trade flows between industries. The network approach provides key benefits over the analysis of
single connections between suppliers and customers. In particular, by considering all industries,
we alleviate selection bias caused by only considering industry-pairs directly involved in mergers.
Second, the network approach explicitly accounts for dependencies between all industries, including
higher-order connections and allows for tests of the propagation of industry-level shocks from one
industry to another, across the entire economy. We believe that this approach will have far-reaching
applications for understanding the interaction of corporate finance and industrial organization. For
the sake of brevity, we present only a fraction of the description of the product market network in
the paper, but provide a comprehensive report in the Internet Appendix, which may be useful for
future research.
The rest of this paper is organized as follows. Section I presents the industry and merger data
and describes the construction of the networks we analyze in the paper. Section II presents tests
that compare the industry input-output network to the merger network in a static setting. In
Section III, we present tests of the propagation of merger waves across the industry network over
time. Section IV presents tests of aggregate merger waves and network centrality. Section V
concludes.
I. Data Sources and Methods
A. Customer-Supplier Trade Network Data
Since 1947, the Bureau of Economic Analysis (BEA) has provided Input-Output (IO) accounts
of dollar flows between all producers and purchasers in the U.S. economy. Producers include
all industrial and service sectors as well as household production. Purchasers include industrial
6
sectors, households, and government entities. Thus, these data cover the entire economy, not just
manufacturing industries. The IO tables are based primarily on data from the Economic Census
and are updated every five years with a five-year lag. Since our merger data (described below)
cover the period 1986 to 2010, we use the IO tables from the years 1982, 1987, 1992, 1997, and
2002, the most recent report as of July 2012.
The BEA defines industries at two levels of aggregation, detailed and summary. The number of
detailed industries, excluding households and government sectors, ranges between 411 and 478 in
the different reports. This is slightly more narrow than the 416 three-digit 1987 SIC codes, but
substantially more coarse than the 1,005 four-digit SIC codes. The detailed IO industries are also
closer to the number of four-digit NAICS codes in 1997 (313), than to the number of five-digit
NAICS codes (721), or six-digit NAICS codes (1,179). The number of summary-level IO industries
ranges between 77 and 126, which is similar to two-digit SIC codes (83) and three-digit NAICS
codes (96).2 Thus, the coarseness of the IO industry definitions are roughly equivalent to two and
three-digit SIC codes, which have been used extensively in prior research.
In each report, the BEA updates the classifications used in the IO tables to reflect changes in
the economy. The classifications are designed to group firms into industries that best measure
customer and supplier relations, using the most recent standardized industry classifications. Prior
to 1997, the IO industries were defined based on 1977 and 1987 SIC codes. In 1997 and 2002, the
BEA based the IO industries on 1997 and 2002 NAICS codes, following the policy of most U.S.
government agencies to switch from SIC to NAICS codes. Concordance tables between NAICS and
SIC codes and IO industry codes are provided by the BEA.
Since our unit of observation is an industry-pair, to maintain consistency over the years in our
sample we cannot combine data from different BEA reports in the same analysis. Therefore, in
the main analysis, we present results using the 1997 detail-level IO definitions. We choose the
1997 report because 1997 splits our merger data into two approximately equal time periods. The
1997 report is also concurrent with the largest aggregate merger activity in our sample period. We
choose to focus on the detail-level industries in the main analysis, because it allows for a more
granular representation of the economy. Therefore, unless otherwise noted, the results presented
2Internet Appendix Table I reports the number of industries across SIC, NAICS, and BEA IO definitions for variousyears.
7
in the paper refer to the detail-level industries in 1997. However, for robustness, in the Internet
Appendix, we run our tests using both detailed and summary-level IO relations from the 1982,
1987, 1992, and 2002 reports. We will direct the reader to specific tables in the Internet Appendix
for each robustness check.
Each IO report defines ‘commodity’ outputs and producing ‘industries.’ A commodity, as defined
by the BEA, is any good or service that is produced. An industry may produce more than one
commodity, which means that more than one industry may produce the same good or service.
However, the output of an industry is typically dominated by one commodity. The ‘Make’ table
of the IO report records the dollar value of each commodity produced by the producing industry.
In the 1997 report, there are 480 commodities and 491 industries in the Make table. The ‘Use’
table defines the dollar value of each commodity that is purchased by each industry or final user.
There are 486 commodities in the Use table purchased by 504 industries or final users.3 Costs
are reported in both purchaser and producer costs (the differences are due to retail and wholesale
markups, taxes, and other transaction costs). Throughout the paper we use producers’ prices, but
using purchasers’ prices makes little difference.
From the Use and Make tables, we create matrices that record flows of inputs and outputs
between industries. Following Becker and Thomas (2010) we calculate SHARE, an I × C matrix
(Industry × Commodity) that records the percentage of commodity c produced by industry i. The
USE matrix is a C× I matrix that records the dollar value of industry i’s purchases of commodity
c as an input. The REV SHARE matrix is SHARE×USE and is the I× I matrix of dollar flows
from the customer industry on column j to supplier industry on row i. Finally, the CUST matrix is
REV SHARE divided by the sum of all sales for an industry. The SUPP matrix is REV SHARE
divided by the sum of all purchases, by industry. The CUST matrix records the percentage of
industry i’s sales that are purchased by industry j. The SUPP matrix records the percentage of
industry j’s inputs that are purchased from industry i. These two matrices describe the relative
trade flows between all industries in the economy.
3The six additional commodities that are in the Use table but not in the Make table are, noncomparable imports,used and secondhand goods, rest of world adjustment to final uses, compensation of employees, indirect business taxand nontax liability, and other value added. The thirteen industries or final users in the Use table that are not inthe Make table include personal consumption expenditures, private fixed investment, change in private inventories,exports and imports, and federal and state government expenditures.
8
The IO tables treat compensation of employees as a commodity input in production. However,
there is no corresponding industry that produces compensation. Because of this, employee compen-
sation gets dropped from the industry matrices. Therefore, we create an artificial labor industry
to make sure that we account for labor as an input in the industry matrices. Without including
labor costs, other inputs in labor-intensive industries will appear to be a larger component of total
inputs than they actually are, relative to capital-intensive industries. The additional labor industry
is used only to account for inputs, and we do not include labor as an industry or commodity in our
final sample. After excluding household and government industries, as well as exports and imports,
and making a few minor adjustments, we are left with 471 industries. A detailed description of the
data is reported in Section I of the Internet Appendix.
One of the important features of the input-output matrix is that it is largely exogenous to merger
activity. This is because the basic input requirements in the production of any good are determined
mainly by the good’s production function, not by the ownership structure of the firms that produce
the inputs.4 The exogeneity of the product market network, with respect to ownership, mitigates
concerns about reverse causality, where merger waves cause product market relations to change.
In addition, by using the 1982 IO reports in robustness tests, we ensure that the IO relations are
exogenous to merger activity from 1986 to 2010.
B. Merger Network Data
Merger data are from SDC Thomson Platinum database. We collect data for all mergers that
meet the following criteria: 1) Announcement dates between 1/1/1986 and 12/31/2010; 2) Both
target and acquirer are U.S. firms; 3) The acquirer buys 20% or more of the target’s shares; 4) The
acquirer owns 51% or more of the target’s shares after the deal; 5) Only completed mergers; and
6) Transaction values of at least $1 million. Since the focus of this study is merger activity, rather
than wealth effects, we do not restrict the legal form of organization of the target or acquirer. This
produces a sample of 51,002 observations. By not restricting our sample to public firms, we have
a much more complete sample than is typically used in existing merger research.
4It is possible that vertically integrated firms could use substitute inputs based on their ownership of certain suppliersegments. However, if input substitution leads to inefficient production, these firms are unlikely to survive, or,alternatively, the input substitution is not important. For this to affect our results, the input substitution would needto occur at an industry-level, rather than at a firm-level.
9
For each observation, we record the value of the deal, the date, and the NAICS codes of the
acquirer and target. Because SDC records 2007 NAICS codes we convert all NAICS codes from
SDC to 1997 NAICS codes to match to the IO data. Then for each deal we map the 1997 NAICS
to the appropriate 1997 IO industry. In the robustness tests that use IO reports from years other
than 1997, we match SIC codes from SDC. This means, for example, that in the tests that use
the 1982 IO reports, we first convert 1987 SIC codes reported in SDC to 1977 SIC codes to match
to the IO definitions. Section I of the Internet Appendix provides more details on the mapping
between industry classifications.
Next, we record merger activity both yearly and cross-sectionally for each directed IO industry-
pair of acquirer and target industries. This produces 4712 = 221, 841 unique pairs. Directed
industry pairs means that we differentiate between acquirer and target industries. For each time
window (yearly and cross-sectionally) we record the number and dollar value of mergers where the
acquirer is in industry i and the target is in industry j. This means we have separate observations
for deals involving acquirers in industry i that are buying targets in industry j and deals involving
acquirers in industry j that are buying targets in industry i. Since in inter-industry mergers, it
is likely that the acquirer could be in either industry, we also record the data in a non-directed
way between two industries. This yields 12 × 471 × (471 − 1) = 110, 685 unique industry pairs per
window of observation.
In the main analysis, we match firms to IO industries using their primary NAICS code. However,
this does not account for diversified firms. We address this concern in a few ways. First, as
mentioned previously, the IO industries are roughly as coarse as three-digit SIC codes. Firms
with multiple, but related, segments, will tend to get assigned to the same IO industry, regardless
of which segment’s 6-digit NAICS code is used. However, this does not account for firms with
multiple unrelated segments that would be assigned to different IO industries, depending upon
which industry is listed as its primary segment. Therefore, we use three alternative methods to
assign firms to IO industries.
In the first alternative method, we use all industry codes reported in SDC to identify a full set of
IO industries per firm. We then assign equal weight to merger counts and dollar volumes for each
of these IO industry codes. For instance, if an acquirer is in industries 1 and 2 and a target is in
10
industry 3, we assign 0.5 merger counts to the industry pair (1,3) and 0.5 counts to industry pair
(2,3). As a second alternative, we give greater priority to horizontal mergers, followed by vertical
mergers, and then unrelated mergers. For each pair of merging firms, we first identify horizontal
mergers as any overlaps in all possible IO industry codes. If there are any horizontal matches,
we assign an equal fraction of the merger count or dollar volume to the overlapping IO industry
codes. If there are no horizontal matches, but there is a vertical relation between any of the firms’
IO codes, we assign the deal activity equally to those IO industry codes. Vertical relations are
defined at two threshold levels. In particular, we record a vertical relation if two industries exceed
a threshold of one percent across any of the following four vertical relations: 1) acquirer industry
purchases from target, 2) target industry purchases from acquirer, 3) acquirer industry sells to
target, and 4) target industry sells to acquirer. We also create a third mapping using a five percent
threshold of vertical relations. If there are neither horizontal nor vertical industry relations, we
assign the deal equally across all of the unrelated industry pairs. This assignment method makes
sure that we do not count horizontal mergers as vertical mergers for integrated firms. We describe
the industry assignment in more detail in Section I of the Internet Appendix.5
C. Other Industry Characteristics
Our aim in this paper is to understand how mergers transmit across industries from a macroeco-
nomic perspective. Therefore, we do not claim to separately identify the various theories of vertical
integration empirically. To do so convincingly requires industry case studies. For instance, in a
famous paper, Masten (1984) tests for holdup problems using measures of the specificity of design
and location for 1,887 aerospace components. Similar papers provide evidence in various indus-
tries, such as coal (Joskow, 1985, 1987), aluminum (Stuckey, 1983), chemicals (Lieberman, 1991),
and paper industries (Ohanian, 1994), each with unique data. However, we include standardized
measures to provide high-level evidence of the importance of possible motives for vertical mergers.
First, as discussed previously, holdup problems associated with incomplete contracts could lead
to vertical mergers. To measure the likelihood of holdup problems between two industries, we
5Though a one percent threshold may seem unimportant, accounting for labor input makes intermediate goodsrelatively small. In particular, we show later in Table I that over 95% of the inputs in an average industry individuallyaccount for less than one percent of total inputs. Of those few industries that supply more than one percent, 46%supply less than two percent of total inputs.
11
record the maximum of industry R&D/Assets for each industry-pair. Industry R&D/Assets is the
median R&D/Assets for all firms in the IO industry, using data from Compustat. Though not
perfect, investment in R&D proxies for contracting problems because it is a measure of intangible
firm-specific knowledge. Larger R&D investments mean more assets that are prone to frictions
from incomplete contracts. Following this interpretation, R&D has been used by many papers in
finance and economics to proxy for difficulty in contracting. These include Denis, Denis, and Atulya
(1997), Allen and Phillips (2000), and Fee, Hadlock, and Thomas (2006). We use the maximum of
the industry-pair’s R&D because the presence of contracting problems in one industry is sufficient
to create a holdup problem.
Second, cross-industry mergers are more likely when there are asset complementarities between
merging firms. Rhodes-Kropf and Robinson (2008) present a search model of mergers based on the
property rights theory of the firm (Hart, 1995) where asset complementarities increase the synergy
gains of a merger. Hoberg and Phillips (2010b) provides empirical evidence that firms that are
more similar to each other, and also less similar to industry competitors, have greater increases in
cash flows and more new product introductions after merging. Furthermore, using a measure of
pair-wise similarity of product descriptions, Hoberg and Phillips show that SIC and NAICS codes
do not adequately capture firm similarity.
To test these theories, we create a measure of inter-industry asset complementarity, which we
denote ‘HP Similarity,’ based on the text-based similarity measure developed in Hoberg and Phillips
(2010a) and Hoberg and Phillips (2010b) (HP), and provided on Jerry Hoberg’s website. Hoberg
and Phillips’s text-based measure identifies firm-pairs that have similar product descriptions in
their 10-K filings for all firms on Compustat from 1996 to 2008. To aggregate the HP firm-level
data to IO industry-levels, we record the total number of firms in the HP database that are in a
given IO industry-pair. Thus, our measure of asset complementarity between two industries is the
total number of firm-pairs in a given IO industry-pair-year that HP identify as similar in a given
year. Because these data are only available for roughly half of our sample period, we do not include
these measures in the main results, but present them in robustness tests described below.
We also account for the size and scope of industries since both are likely related to merger
activity. To measure industry size, we would prefer to have the total number of firms operating in
12
each industry on a yearly basis. However, the available data either do not cover our time period, do
not have detailed industry classifications, or only cover a limited subset of firms (e.g., public firms
in CRSP and Compustat). Instead, we use precise data on the number of establishments from the
U.S. Census Bureau’s County Business Patterns (CBP) database. Establishments are defined as
single physical locations. Thus larger firms have more establishments. These data are based on the
Census Bureau’s Business Register, the most complete account of business activities available, and
cover the vast majority of industries, including manufacturing and service industries. The data are
reported at 4-digit SIC and 6-digit NAICS codes. An advantage of establishment-level data is that
industry classifications are more precise than firm-level data since establishments are more likely
to engage in activities that fall primarily in one industry classification. We aggregate these data to
IO industries following the mapping discussed above.
To account for the scope of industries, we record the percent of all NAICS or SIC codes (de-
pending on IO report year) that map to a particular IO industry. Since SIC and NAICS codes are
defined to be relatively equal in scope (Economic Classification Policy Committee, 1993; Gollop,
1994), this variable provides a measure of the variation of business activities of each IO industry. See
the Section I of the Internet Appendix for more details. We also control for industry concentration
using the 8-firm concentration ratio from the Economic Census of the U.S. Census Bureau. Like
the IO data, the Economic Census is conducted every five years, in years ending in two and seven.
With the exception of agriculture and public administration, concentration measures are reported
for all industries. Since these data cover firms of all sizes and the vast majority of industries, they
provide the most comprehensive concentration ratios available. In contrast, concentration ratios
calculated using Compustat sales are subject to both a severe size bias and a public-listing bias.
We map SIC and NAICS codes to IO industries yearly, using the most recent concentration ratios.
To account for valuation-driven mergers, we include various variables, including industry median
market-to-book, returns, and standard deviation of returns. We also calculate the difference of
these variables between two industries in each industry pair. Finally, we calculate an Industry
Economic Shock Index as in Harford (2005). For each industry, we compute the first principal
component of the medians of the absolute value of changes in cash flow, asset turnover, R&D,
capital expenditures, employee growth, return on assets, and sales growth for each firm in the
13
industry. We rank this principal component across industries and time and choose industry-years
in the top quartile as ‘shock’ years. This variable measures shocks to economic fundamentals at
the industry level. All variables are described in detail in Section I of the Internet Appendix.
D. Networks
The primary goal of this paper is to identify the relationship between the IO network and the
merger network. To provide a framework for the following analysis, we discuss how networks are
defined and measured.
Any network can be described by an N×N adjacency matrix, A, consisting of N unique ‘nodes,’
which are connected through ‘edges.’ Emphasizing the importance of edges in a network, nodes are
most generally defined as an endpoint of an edge. Each entry in the adjacency matrix A, denoted
aij , for row i and column j, records the strength of the connection between nodes i and j. A binary
matrix simply records a one if there is a connection and zero if no connection, but different values
may also be assigned in a weighted adjacency matrix to indicate the strength of the connection. In
addition, A is not restricted to be symmetric so that connections may be directional.
A primary innovation of this paper is to treat the industry input-output data and merger data as
networks. Specifically, each of the networks has the same set of nodes (i.e., the 471 industries from
the 1997 IO tables), but the connections between the nodes are either product market relationships
in the IO network, or inter-industry mergers in the merger network. This is easily accomplished
by simply treating the input-output matrices and the cross-industry merger matrix as adjacency
matrices. Thus for the same set of industries we record multiple connections, based either on
product market relations or merger activity. Though there is a natural fit between input-output
tables and network analysis, to the best of our knowledge, this is the first paper to make this
connection.
To illustrate the network concepts, Figure 1 presents representations of two simple input-output
networks of six industries in the timber sector. These networks are a subset of the entire IOFigure 1
industry network we use in later tests. Each network consists of six nodes that are connected
through directed weighted edges. Subfigure a presents the network of customers as an adjacency
matrix (from the CUST matrix) and subfigure b presents the network of suppliers as an adjacency
14
matrix (from the SUPP matrix). Subfigure c presents both the customer and supplier network in
a graphical representation.
Though input-output relations are often modeled as a linear chain, Figure 1 reveals that the
path from raw materials to finished goods is much more complex, even in this highly reduced
subset of the network. The forestry support industry provides inputs into the nurseries and logging
industries. Of all non-labor inputs in the forest nurseries industry, 64% are purchased from the
forestry support industry (a21 in subfigure b), though of all sales by the forestry support industry,
only 14% are purchased by the forest nurseries industry (a21 in subfigure a). Weighted asymmetric
network ties are evident throughout this sector. For example, the forest nursuries industry also
supplies to the logging and sawmill industries, though the connection to logging is stronger than
to sawmills. Pulp mills receive inputs from both the logging and sawmill industries. Finally the
sawmill industry supplies to the wood doors industry.
The complexity of networks is obvious even in such a simple subset of the data. Increasing
the number of nodes to 471 and increasing the number of connections exponentially provides an
extremely complex network of industry relations. Therefore, as stated previously, to analyze both
the IO and merger networks, we use techniques from graph theory and social networks. We briefly
discuss these techniques next, including the concepts of centrality, clustering, and average shortest
paths. Each of these is discussed in greater detail in Section II of the Internet Appendix.
Network centrality refers to how important one node in a network is relative to other nodes.
Importance is based on how many connections a node has and to which other nodes these connec-
tions are made. For our purposes, this means how important an industry is in the flow of inputs
and outputs between all industries, or in the number of cross-industry mergers. We employ two
measures of network centrality: degree centrality and eigenvector centrality. The degree centrality
of a given node in a network is simply the number of links that come from it. Formally, node i’s
degree centrality is the sum of its row in the network’s adjacency matrix where connections are
binary. If connections are weighted values, then the degree is referred to as strength. The other
centrality measure we consider is eigenvector centrality, formally defined by Bonacich (1972) as the
15
principal eigenvector of the network’s adjacency matrix. Intuitively, a node will be considered more
central if it is connected to other nodes that are themselves central.6
There are other measures of centrality, but we choose to focus on degree centrality and eigenvector
centrality because they best reflect how shocks would propagate through an economy. Borgatti
(2005) shows that these two measures capture a flow process across a network that is not restricted
by prior history (such as a viral infection like chicken pox would be, since a node is immune after
receiving the virus) and allows for a shock to spread in two different directions at the same time
(as opposed to a package that moves along a network which can only be in one place at one time).
Therefore, these measures of centrality allow an economic shock that flows to the same industry
from two different sources to have a larger impact than a single shock, and allows the shock to
spread in parallel to multiple industries simultaneously.
The second type of network measure we examine is clustering. Clustering refers to how embedded
a node is in the network, or in our case, how embedded an industry is in the economy. More
formally, we calculate the clustering coefficient of Watts and Strogatz (1998). Defining a node’s
neighborhood as the set of nodes to which a particular node is connected, the clustering coefficient is
the proportion of observed connections between the nodes in its neighborhood to the total possible
connections. Intuitively, the greater is the clustering coefficient of an industry in the customer-
supplier network, the more its customers and/or suppliers also trade with each other. In contrast,
the trading partners of industries with low clustering coefficients, trade little with each other. This
measure helps us understand how merger activity is likely to transmit across the IO network.
Finally, we measure each industry’s average path length. For a given industry, we calculate the
shortest path through the network to every other industry in the network using Dijkstra’s (1959)
algorithm. We then take the average of the path lengths for each industry. This measure presents
another indication of how connected an industry is. This again is important for understanding
network dynamics, since it measures the closeness of an industry to all others, on average, and also
at the network level it indicates how densely connected is the network, or in our case, the entire
economy. For more details, see Albert and Barabasi (2002).
6If we define the eigenvector centrality of node i as ci, then ci is proportional to the sum of the cj ’s for all other
nodes j 6= i: ci =1λ
∑
j∈M(i) cj = 1λ
∑N
j=1 Aijcj , where M(i) is the set of nodes that are connected to node i and λ
is a constant. In matrix notation, this is Ac = λc. Thus, c is the principal eigenvector of the adjacency matrix.
16
II. The Relation Between Product Market and Merger Networks
In this section of the paper, we test whether the IO network of customers and suppliers can
explain the merger network of cross-industry mergers in a cross-sectional setting.
A. Summary Statistics
Table I presents summary statistics of the 1997 input-output relationships. We divide the sampleTable I
into inter-industry pairs, intra-industry pairs, and inter-industry pairs that have substantial trade
relations. To identify industry pairs with a substantial relationship, we follow Fan and Goyal (2006)
and Ahern (2012) and require either (1) that a customer industry buys at least 1% of a supplier
industry’s total output (Customer %), or (2) that a supplying industry supplies at least 1% of the
total inputs of a customer industry (Supplier %). This is necessary since most industry-pairs have
almost zero trade relationships. As mentioned above, accounting for labor input reduces the share
of intermediate inputs considerably. Across all 110,685 inter-industry pairs, the mean percentage of
sales purchased by a customer is only 0.22%. Likewise, the percentage of inputs that one industry
supplies to another in an average industry-pair is only 0.27%. More than 95% of industry-pairs have
customer and supplier relationships less than 1%. Considering the breadth of the U.S. economy,
it is expected that industries do not have customer-supplier relations with most other industries.
For example, we would not expect that firms in the forestry support industry have substantial
trade relations with firms in the financial services industry. However, these results indicate that
customer-supplier relations are highly clustered in a very small set of industry-pairs.
In the inter-industry pairs with substantial trade flows, the average percentage of total sales
purchased is 5% and the median is 2.2%. The average percentage of total inputs supplied is 3.9%
and the median is 2.2%. Intra-industry pairs also exhibit trade flows. In this case the industry uses
a portion of its output as an input. For example, a firm that produces energy must also use energy
in its production process. The median supply and customer relationships are 1.1% and 1.4% and
close to 50% of industries have supplier and customer relationships less than 1%.
In Internet Appendix Table II, we provide the same statistics for each IO report year for both
detail and summary-level industry definitions. The statistics show that customer-supplier relations
remain stable over time from 1982 to 2002. This likely reflects that vertical relations are persistent
17
and also that the BEA updates its industry definitions to maintain a consistent measurement of
input-output relations. The statistics of the IO relations at the summary-industry level are surpris-
ingly similar to the detail-industry level, given that there are only 124 summary-level industries,
compared to 471 detail-level industries. In particular, across all industry-pairs, the average per-
centage of inputs supplied is roughly 0.80%, but the median is about 0.20%, compared to 0.01% in
the detail-level relations. These results are consistent with a product market network composed of
few key industries and many less important ones.
Turning to the merger data, Figure 2 summarizes the time series of aggregate merger activity
in our sample. This figure primarily establishes that our merger sample is similar to those usedFigure 2
in other studies of mergers and of clustering of merger activity in time. As is typical, the 1980s
merger wave is small in comparison to the activity in the mid to late 1990s. The most recent wave
that began in 2003 to 2004, ends in 2009 due to the financial crisis.
Table II describes the merger data at the industry-pair and industry level. Industry level ob-Table II
servations are aggregates of industry-pair observations. In the entire sample across all years, there
are a total of 51,002 mergers and acquisitions representing total deal value of $16.7 trillion in 2010
dollars. Of these, 19,962 are intra-industry, horizontal mergers, representing $6.6 trillion in deals.
The remaining 31,040 deals are inter-industry deals, accounting for $10.1 trillion.
Across all possible pairwise inter-industry combinations, the average industry pair has 0.28 merg-
ers over the 25 year sample period and 94% have no mergers at all. This means that though
inter-industry mergers are more common than intra-industry mergers in our sample, they are not
uniformly distributed across industry-pairs, but rather, are highly clustered. Only 6% of the 110,685
industry-pairs account for all 31,040 inter-industry deals. If economic fundamentals drive merger
waves, it is not surprising that they will cluster in a small set of industry-pairs given the clustering
in the product market network. Looking across all possible inter-industry pairings for any given
industry, the mean number of cross-industry mergers for a detail-level industry during our 25-year
sample period is 65.9 and the median is 15. This compares with an average of 42.4 and median
of 4.0 for intra-industry mergers. Eighteen percent of industries have no intra-industry mergers
during the sample period, compared with roughly one percent for inter-industry mergers.
18
Summary statistics of merger activity for each of the alternative IO report years (1982, 1987,
1992, and 2002), for both levels of industry aggregation (detail and summary-levels), and for each
of the alternative industry assignment methods are presented in Internet Appendix Table III. First,
the basic patterns of industry clustering and differences between inter- and intra-industry merger
activity is relatively stable over the different IO report years. Second, in the 124 summary-level
industries, we also find that inter-industry mergers are highly clustered in a few industry-pairs:
roughly 60% of industry-pairs have no mergers during the 25 year sample period and there are still
more inter-industry mergers (28,672) than intra-industry mergers (22,330). When we assign firms
to industries using all reported SIC or NAICS codes on SDC, the fraction of mergers that are cross-
industry mergers increases, and when we give greater priority to horizontal mergers, the opposite
holds. However, we still observe highly concentrated inter-industry mergers in all of the various
industry assignment procedures. In particular, even when giving priority to horizontal mergers and
in the broad summary-level industries, we still find that roughly a third of all mergers occur across
industries.
B. Comparing Merger and IO Networks: Univariate Evidence
We compare the merger and IO networks to each other in two ways. First, we compare the entire
structure of each network. Second, we compare the networks industry-by-industry.
In Figure 3, we present the degree distributions of the product market and merger networks.Figure 3
Recall that an industry’s degree is the number of connections between industries. The degree
distribution, p(k), is the proportion of industries with k direct connections. Since the merger
and IO networks appear highly clustered in the summary statistics, their degree distributions are
likely to be skewed. Gabaix (2009) shows that many phenomena in economics and other fields are
similarly clustered and many approximate a power law distribution, p(k) = ck−α. If the distribution
follows a power law, then the relation between the number of connections and the probability of
connections would follow a linear pattern in logarithmic scale. For reference, we plot the estimated
power law line using the maximum likelihood method of Clauset, Shalizi, and Newman (2009),
though it is not important for our purposes that the distribution is statistically a power law or not.
19
Figure 3 reveals that both inter-industry mergers and IO connections are characterized by many
industries with few connections and a few industries with many connections. The circles in the
lower right corners of the graphs represent the few rare industries with a very large number of direct
connections to other industries. For instance, in the supplier industries, only a half of one percent
of industries have over 200 connections. In comparison, the fraction of industries with at least 20
connections is about 20%. The degree distributions of the customer-supplier and merger networks
are similar. The estimate of α in p(k) = ck−α in the merger network is 3.3. In the supplier network,
it is 3.1. Using the more coarse summary-level definitions produces a similar pattern, as shown in
Internet Appendix Figure VI.
These patterns are important for a number of reasons. First, as mentioned above, if economic
fundamentals drive merger waves, it is expected that they will cluster in the relatively small set
of industries that are connected through IO relations. Second, it suggests that if merger activity
follows the industry network over time, we should not expect to see random unrelated merger waves,
but rather we would expect many merger waves occurring simultaneously. We discuss this point in
more detail in Section IV.
Next, in Table III we present averages and medians of industry-level network statistics for the
supplier, customer, and merger networks. The average (median) industry has about 22 (16) connec-Table III
tions to suppliers and 16 (13) connections to customers, where connections are substantial relations,
as defined previously. The average industry has cross-industry mergers with 27 different industries
in our sample period and the median is 18. The average shortest path across industries is about two
for all networks, which reveals the ‘small-world’ nature of these networks: across 471 industries, a
typical industry is only 2 to 2.5 connections away from every other industry. In fact, the maximum
shortest path length between any two industries, known as the diameter of the network, is three
in the supplier network, six in the customer network, and five in the merger network. These re-
sults indicate that though the networks are sparse, they are still highly connected through central
‘hub’ industries. Overall, we find that at an aggregate network level, the industries exhibit similar
features. The most notable difference is that the average industry in the merger network is more
clustered and more central than it is in the IO networks. Internet Appendix Table IV presents
20
these statistics for each IO report year, each of the different algorithms for assigning firms to IO
industries, and using the summary-level industry definitions.
Next, we examine industry-by-industry relationships between the IO and merger networks. In
Table IV, we present the 10 most central industries in the supplier, customer and merger networks
according to degree centrality. Many of the most central industries in the IO network are alsoTable IV
among the most central in the merger network, including wholesale and retail trade industries,
construction, motor vehicle parts and administrative support services. These are the industries
that have inter-industry mergers with the largest number of different industries, not necessarily the
most mergers overall, as well as many connections in the product market. The substantial overlap
indicates that industries that are economically central as customers or suppliers are also central in
the merger network.7 Internet Appendix Table V provides these same results for each IO report
year and at the summary-level.
In Table V, we present correlations of the industry-by-industry network characteristics. First,Table V
the centrality measures are correlated across networks, so that central industries in the IO networks
are likely to be central industries in the merger network. The correlation between the centrality of
the customer and merger industries is a significant 53.3%. Similarly, the correlation of an industry’s
average path length in the merger network with its average path length in the customer network is
a significant 24.1%. We find that the correlations are equivalent or stronger in the summary-level
networks (see Internet Appendix Tables VI and VII).
The univariate results in this section provide strong evidence that the industries that are impor-
tant in the IO network in terms of centrality, path lengths, and clustering, are also important in
the merger network. In the next section, we control for additional factors that could be related to
industry structure, such as market valuations, industry concentration, or industry size.
C. Comparing Merger and IO Networks: Multivariate Evidence
Our final cross-sectional analysis of the relation between the IO and merger networks uses a
network analysis technique called Exponential Random Graph Models (ERGM). Just as a logit
regression produces maximum likelihood estimates (MLE) of a single dependent variable, ERGM
7As discussed below, we drop retail and wholesale industries in robustness tests.
21
produces MLE estimates of the entire network, including node characteristics and the strength
of connections between nodes. Importantly, like a multivariate regression, ERGM allows multiple
variables to jointly explain the observed network. The key difference between ERGM and logit
or ordinary least squares (OLS) regressions is that ERGM explicitly accounts for higher order
dependence between industries by modeling the entire network outcome, rather than single industry
or industry-pair outcomes. Since the core of our argument is that industry connections affect
the incidence of mergers, ERGM is a necessary tool to account for dependency between industry
nodes. In Section III of the Internet Appendix, we provide a detailed description of the theory and
implementation of ERGM, as well as citations to the papers that originally developed ERGM.
The results of the ERGM analysis are presented in Table VI. The coefficient values are theTable VI
estimates of the marginal effect of the explanatory variable on the conditional log-odds that two
industries will have an additional inter-industry merger. Explanatory variables describe both intra-
industry characteristics, such as an industry’s concentration ratio, and inter-industry relations,
such as customer-supplier relations. For example, ‘Target Buys from Acquirer’ is the inter-industry
percentage of the target’s industry’s purchases from the acquirer’s industry. The coefficient estimate
of the variable ‘Number of Connections,’ is unique to ERGM and measures the marginal change
on the M&A network from adding a random connection.
The results in Table VI show that the IO networks significantly help to explain the merger
network. Each of the four IO network connections has a positive and significant effect on the
likelihood of merger connections, both separately in columns 1 through 4, and jointly in column
5, each incrementally contributing to an understanding of the occurrence and intensity of merger
activity between industries. This result implies that cross-industry merger activity is related to the
strength of the customer-supplier relationship between the industries.
In column 6 of Table VI, we add industry characteristics as additional explanatory variables to
account for possible holdup problems, overvaluation, and economic characteristics of the industries.
These additional control variables are defined at both the industry-pair level, including the absolute
differences in returns, volatility, M/B, and concentration, as well as the industry-level, including
industry median M/B, median R&D, and others described previously. The industry-level variables
22
describe characteristics of one industry node in the network, whereas the industry-pair variables
describe connections between industries.
We find that adding the additional controls does not affect the strong predictive power of the
IO connections. In fact, three out of four coefficient estimates of the IO connections increase after
adding the controls. This result is particularly strong since the data limitations of the control
variables reduce the size of the network in the analysis, and hence the expected predictive power
of the industry connection variables. These results show that a first-order determinant of cross-
industry mergers is the economic trade flows between industries.
We also find that greater differences in M/B and average returns between industries leads to a
lower likelihood of a cross-industry merger, consistent with Rhodes-Kropf and Robinson (2008).
However, we find that industries with greater median M/B and smaller variance in returns are less
likely to be involved in a merger. We also do not find any strong evidence of overvaluation-driven
mergers, based on our control variables.
The results in Table VI also show that greater R&D expenditures is related to increased merger
likelihood, consistent with holdup problems. In column 7, we include interactions of the maximum
R&D variable with the four IO relations variables to test whether stronger customer-supplier rela-
tions magnify holdup problems. The results indicate that when an acquirer is either an important
customer or supplier to the target, greater holdup problems, as proxied by R&D, are associated
with more mergers, on the margin. In contrast, when a target is an important customer or supplier
to the acquirer, greater holdup problems are associated with fewer mergers. An alternative way to
characterize these results is that when holdup problems are larger, the firm that has stronger IO
connections becomes the acquirer, rather than the target. This result is consistent with one of the
predictions of the property rights theory of the firm (Grossman and Hart, 1986; Hart and Moore,
1990; Hart, 1995). The theory predicts that of the two merging firms, the one that would have
greater investment distortions if the merger did not happen will be the acquirer (Grossman and
Hart, 1986). We find that in industry-pairs where holdup problems are larger, the acquiring firm
tends to be the firm that is a large customer or supplier and that the firm that is the target has
weaker IO relations. Assuming that potential distortions in investments are related to the strength
of the IO connections, these results provide further evidence that holdup problems not only help
23
to explain the occurrence of cross-industry mergers, but also which firm is the ultimate owner of
the combined assets.
The impact of stronger IO connections on the likelihood of cross-industry mergers is economically
important. Using the coefficient estimates from the specification with the most controls, we find
that for a one standard deviation increase in IO relations, the odds of an additional merger between
two industries increases by 6 to 20%. However, this calculation understates the effect because it
estimates the effect of changing one customer-supplier connection strength while holding the others
constant. In reality, the strengths of the connections are correlated so that when one is higher,
the other measures are higher as well. To estimate the economic significance of customer-supplier
relations on merger incidence, we compare the average incidence of mergers in industry pairs where
all IO measures are below 1% to those where there is a strong IO connection. We define a strong
connection to occur when industry i buys 5% or more of industry j’s output and industry j supplies
at least 5% of industry i’s total inputs. The average merger incidence for industry pairs with IO
measures below 1% is 0.11. The average for an industry pair with a strong connection is 12.5.
These results imply not only that vertical mergers occur, but more interestingly, that stronger IO
connections are associated with more vertical mergers.
In Internet Appendix Table VIII, we present ERGM tests using the summary-level industry def-
initions and using the different IO report years. We find a similar pattern in using each of the IO
report years from 1982 to 2002. In all specifications, the IO variables are positive and significant.
Additionally, the coefficient on the interaction between maximum R&D and the intensity of the
acquirer as a customer or supplier is positive and significant in the large majority of specifica-
tions. The summary-level results are largely consistent with the detail-level tests, where stronger
customer-supplier connections leads to a greater likelihood of cross-industry mergers. The results
for the control variables are less robust.
These results are robust to other controls. In tests reported in Internet Appendix Table IX, we
control for asset complementarities using the ‘HP Similarity’ variable described above. Our main
results are unchanged, with strong positive associations between the customer-supplier relations
and cross-industry merger volume. Consistent with Hoberg and Phillips (2010b), we find that
greater asset complementarities are associated with greater cross-industry merger flows. In Internet
24
Appendix Table X, we calculate ERGM tests as in our main tests, but match firms to industries
using each of the three alternative methods discussed previously: 1) using all industry codes,
2) giving priority to horizontal mergers then vertical mergers at the 1% relations-level, and 3)
giving priority to horizontal mergers then vertical mergers at the 5% level. Using these alternative
assignment procedure, we find results that are consistent with our main tests, showing a positive
relation between IO connections and merger activity.
To further ensure that our results are not driven by overvaluation-driven mergers, in Internet
Appendix Table XI we present correlations between the strength of customer-supplier relations
and average three-day abnormal announcement returns for the acquirer and for the size-weighted
combined returns of the acquirer and target. On average, we find positive, but small correlations,
for both the acquirer and the combined returns. In addition, we do not find any strong relations
between the average fraction of cash used in mergers and the IO relations. These results show
that the IO variables do not proxy for misvaluation, and if anything, stronger IO connections are
associated with greater acquirer gains and total synergies.8
Though these tests account for an average effect of the control variables, the interpretation of
their effect on merger activity is unclear since they may change over time. Therefore, we separately
estimate ERGMs for each year in the sample period. Figure 4 presents the t−statistics from each
of the four explanatory IO networks in ERGM tests which are run using yearly M&A network data.
The edge covariance coefficients are highly significant in each year, as in the overall sample.Figure 4
Although ERGM analysis is the best way to analyze our question, it is new to the literature. As
a check, we repeat our analysis with OLS regressions. Regressing the value and count of mergers
between industries on the four measures of their IO connectedness produces the same inferences:
IO connections are highly significant in explaining merger activity. In addition, for robustness, we
drop the wholesale and retail trade industries from our analysis, following Acemoglu, Johnson, and
Mitton (2009), and find our results are unchanged.
8Tests of announcement returns are complicated by truncated distributions since we only observe returns wheremergers actually occurred. Our focus in this paper is the pattern of merger incidence across industries. Therefore,we leave a more in-depth investigation of the value-implications of mergers across the IO network for future research.
25
D. Summary of Cross-Sectional Tests
In this section, we have presented a number of related results. First, we have shown that the
structure of the IO and merger networks are highly similar. Second, we have shown that industry-
level network characteristics are highly correlated between the IO and merger networks. Finally,
we have shown in multivariate tests that control for a host of variables, that IO relations predict
cross-industry merger activity and that these relations are more important when holdup problems
are more likely. In contrast, we could have found that many mergers were in unrelated industries,
or only occurred in high valuation industries or industries with more complementary assets, in-
dependent of the strength of the IO connection. Taken together, these results provide consistent
evidence that merger activity follows fundamental economic relations on multiple dimensions.
III. Propagation of Mergers Across the Industry Network
In this section, we consider the dynamic aspects of mergers across industries over time. In
particular, we predict that the likelihood that an industry experiences a merger wave will be
greater if its customer and/or supplier industries recently experienced merger waves. Thus, under
this hypothesis, merger waves beget merger waves across the IO network. The alternative hypothesis
is that merger activity in an industry is unrelated to its customer and supplier industries’ merger
activity, or instead is driven by some other forces, such as misvaluation or asset complementarity.
First, to illustrate how the diffusion of merger activity across related industries occurs, we present
an example from the forest industry we discussed above.
A. Diffusion of Mergers Across the Forest Industry
The forest industry is an ideal setting to illustrate merger diffusion because it experienced a
large external shock which led to an industry reorganization. In 1990, the Northern Spotted Owl
was listed as “threatened” under the Endangered Species Act. Further injunctions in 1991 and the
enactment of the Northwest Forest Plan in 1994 led to the protection of 24.4 million acres of federal
land in Washington, Oregon, and California, the historic home of the timber industry (Ferris, 2009).
At the time, much of the timber supply came from logging on federal land. Smaller sawmills and
26
logging companies that relied on the federal lands were squeezed out by larger suppliers that owned
private nurseries. In addition, the industry moved away from the Northwest and towards the South
where timber tracts were privately owned. However, the protection of the old-growth timber led to
a severe and long-lasting supply shock.
Subfigure a of Figure 5 presents the time-series of the volume and price of timber in Oregon
from 1986 to 2008. The volume of timber harvested dropped precipitously from about 8.5 billionFigure 5
board feet in 1989 to about 4 billion board feet in 1997. This supply shock caused the price index of
timber to rise from 6,155 in 1989 to 11,047 in 1993 and then decline to 7,913 in 1997. Though, these
data are from Oregon, it is indicative of the effect at the national level, since the forest industry
was concentrated in the Pacific Northwest.
The timber supply and price shock led to a large-scale consolidation in timber-related industries.
Recall from Figure 1, the timber sector is comprised of a number of industries that are inter-related
through trade. Subfigure b of Figure 5 presents the merger activity from 1990 to 2005 in the
following industries: 1) sawmills, 2) forest nurseries, forest products, and timber tracts, 3) logging,
and 4) pulp mills. To compare merger activity across the industries, for each industry-year, we
calculate the time-series percentile of the number of mergers involving firms in each industry over
the period 1986 to 2008. We then take the two-year moving-average of the percentile time-series.
First, the sawmill industry (indicated by the solid line in subfigure b of Figure 5) experienced a
large merger wave starting in 1994 and ending in 1999, its largest merger activity over the 23-year
sample period. Next, the forest nurseries industry (dashed line) experienced its largest merger wave
in our sample period from roughly 1996 to 2001. Following this, both logging (dotted line) and
pulp mills (circled line) experienced large merger waves, with merger activity peaking in 1999 and
2000, respectively. Overall, subfigure b shows a clear time sequence of industry waves in related
industries.
Subfigure c of Figure 5 presents the same industry time-series of merger activity where the
leading industries have been shifted back in time to match the timing of the sawmills industry
merger wave. Matching the one-period leading merger activity in the forest nurseries industry,
and the three-period leading activity in logging and pulp mills industries to the sawmill industry
merger wave presents a striking picture. The duration, intensity, and general shape of all four
27
industry-merger waves are highly comparable. In fact, though the figure shows only the 1990s, the
merger activity between the time-shifted industry series over the whole sample period 1986 to 2008
are significantly correlated. For instance, the correlation between the current merger activity in
the sawmill industry with the one-period leading merger activity in the forest nurseries industry is
72.8% (p−value< 0.001). The correlation between current activity in the sawmill industry and the
three-period leading activity in the pulp mills industry is 61.1% (p−value= 0.007).
The evidence presented on the timber-related industries lends support to the importance of
industry links in merger waves. A distinctive economic shock changed the fundamental economic
environment in the sawmill and logging industries. Each responded through mergers. This in
turn had an affect on forest nurseries and pulp mills, which also responded to the new environment
through an industry merger wave. We next test whether these results generalize to other industries.
B. Closeness-Weighted M&A Activity in Connected Industries
In this section, we present the results from tests of the diffusion of merger activity across the
industry network. We first create a measure of an industry’s exposure to merger activity that
does not include the industry itself. For each industry, we calculate a weighted exposure to high
levels of M&As in all other industries, where the weights are the inverse of the shortest distance
from the subject industry to each other industry. Specifically, we calculate the following for each
industry-year:
Closeness-Weighted M&A Activityit =∑
j 6=i
1
distij
∑
k 6=i
vjkt (1)
where distij is the shortest path between industry i and j, and vjkt is a measure of M&As between
industries j and k in year t. We compute the shortest path in two ways: the first uses the network
defined by customer links greater than one percent, and the second uses supplier links. This allows
us to differentiate the importance of exposure to mergers in upstream versus downstream industries.
To account for cross-sectional differences in average merger activity across industry-pairs, we
measure vjkt as an indicator variable for industry-pair-years with high merger activity. The indi-
cator takes the value of one if the log of the inflation-adjusted dollar volume between industries
28
j and k in year t is greater or equal to the 75th percentile of the industry-pair’s time series of
dollar volumes from 1986 to 2010. This controls for the fact that the volume of mergers between
two related industries is normally higher than between two unrelated industries. Thus,∑
k 6=i vjkt
records for each industry j, the total number of industry pairs that experience high merger activity
with industry j in year t, including intra-industry mergers in industry j itself, but excluding merger
activity between i and j directly. This sum for industry j is then weighted by the inverse of the
discrete number of customer or supplier connections between industries i and j. An industry that
is a direct customer or supplier to industry i is one step away. Intuitively, this measure captures
an industry’s exposure to high merger activity in more closely-connected industries, not counting
merger activity involving the industry itself.
Alternatively, an industry’s merger activity may be part of aggregate merger activity driven by
macroeconomic shocks or widespread technological changes. Additionally, an industry’s position
in the network will affect its distance to other industries, and hence its likelihood of exposure
to industries that are experiencing merger waves. Compared to an industry on the periphery of
the network, a central industry will naturally be exposed to more industries that are experiencing
merger waves, simply because it is more connected. Thus, if we observe that the central industry
also experiences more mergers, it could be caused by its centrality, rather than its exposure to
mergers in other industries.
To control for both time-varying economy-wide factors and industry-specific fixed factors, we
estimate the following equation:
log(1 + vi,t) = α+ ρ log(1 + vi,t−1) + βCloseness-Weighted M&A Activityi,t−1 + γi + τt + εi,t (2)
where vi,t =∑
j vijt is the number of industry pairs involving industry i that experience high
merger activity in year t, as above. The industry fixed effects, γi, control for all time-invariant
industry characteristics, such as centrality and scope. They also account for much of the cross-
sectional differences in industry size, valuation, and returns, which are persistent over time, though
we explicitly control for these variables in later tests. The year fixed effects, τt, capture any
macroeconomic shocks, such as the market return and the economy-wide availability of financing,
among other possible factors. To account for persistence in mergers, Equation 2 also includes lagged
29
merger activity in industry i. Thus, this empirical model isolates the impact of within-industry time-
series variation in exposure to mergers, while controlling for macroeconomic time-series changes and
persistence in merger activity.
To estimate Equation 2, we follow Arellano and Bond (1991) to account for the endogeneity
created by using lagged dependent variables in a fixed effects model. Table VII presents the
coefficient estimates. The first three columns define distance using customer relations. The lastTable VII
three columns use supplier relations. In column 1, we find a positive and significant effect of
customer-based closeness-weighted M&A activity in time t − 1 on industry merger activity at
time t. This implies that when the customers of an industry are engaged in more mergers, the
industry’s own merger activity increases the following year. In column 4, we find a similar positive
and significant effect for supplier connections. However, the magnitude of the effect is doubled,
indicating that exposure to mergers from supplier industries has a bigger impact on future merger
activity.
In columns 2 and 5, we include three additional lags of the closeness measure and the subject
industry’s own merger activity to control for delayed responses. The coefficient on the one-year lag
is unchanged for both customers and suppliers. The two-year lag is negative and significant and
the three- and four-year lags have small negative effects. This time-series pattern is consistent with
a large shock in year t− 1 leading to future merger activity in connected industries.
The prior results present predictive regressions, using only information available in prior years
to predict merger activity in the current year. A concern with this approach is that the one-year
lagged exposure to M&As may simply proxy for current activity because of persistence in M&A
activity. If current activity in connected industries has a positive impact, we could not distinguish
whether the diffusion of mergers happens within one year or if there was simply an omitted variable
driving both subject and connected industries’ mergers. To address this concern, in columns 3 and
6, we include concurrent exposure to merger activity. For customers, the coefficient is insignificant.
For suppliers, it is negative and significant. In both cases, the effect of one-year lagged exposure
remains positive and significant. These results provide further evidence that a wave-like peak
of M&As in connected industries occurs one year before the subject industries increased merger
activity. In Internet Appendix Table XII, we present the same analysis using the summary-level
30
industry definitions. We find a similar pattern as in the detailed-level industries. The one-year
lagged closeness-weighted M&A activity is positive and significant in all cases. This result is robust
to the inclusion of longer time lags and to concurrent M&A activity in connected industries, as in
the detail-level results.
While these tests show a positive effect on merger activity related to the one-year lagged merger
activity in connected industries, they also show negative effects in the concurrent year and in year
t − 2. This means that controlling for the large positive effect at year t − 1, the merger activity
in the related industries leads to fewer mergers in the subject industry. This may be caused by
multicollinearity from persistence in the Closeness-Weighted M&A Activity variable. To investigate
this, in Internet Appendix Table XIII we include each of the closeness-weighted activity variables
from t to t− 3 separately. We find the strongest effects at t− 1 and smaller positive or no effects
at t, t − 2, and t − 3. We only find small negative effects at t − 3 in a few specifications, once
we control for four lags of own-industry M&A activity. These tests provide further evidence in
support of the idea that merger waves travel across industries with the peak of M&As in connected
industries occurring one year in advance of the subject industry’s M&A wave.
The impact of M&A activity in close industries is economically meaningful. A one standard de-
viation change in the lagged customer closeness-weighted M&A activity implies an increase of 0.380
in the number of industry-pairs that are experiencing high merger activity. This is a substantial
increase compared to the average of 0.323. This means that the number of cross-industry merger
waves doubles following high merger activity in customer and supplier industries. Moreover, the
marginal effect for supplier links is even stronger, with double the impact of the customer links.
While the industry fixed effects account for time-invariant determinants of M&A activity, there
may be time-varying factors that are correlated with the closeness-weighted M&A activity variables,
such as variables measuring misvaluation, holdup problems, and asset complementarity. There-
fore, in Internet Appendix Table XIV, we run additional tests that include the industry economic
shock index, an indicator for a deregulatory shock, industry median market-to-book, median R&D,
and the yearly average stock return and standard deviation of returns for firms in the industry.9
9In these panel tests, we include industry fixed effects which subsume all time-invariant industry-level variables,including concentration, size, and scope (which are calculated at the IO year and hence fixed). However, we includethe time-varying rate spread and deregulatory shock variables because they are meaningful in the panel setting.
31
Though the sample size is reduced, we still find positive and significant coefficients on the lagged
closeness-weighted M&A activity variable, and negative coefficients for longer lags. Industry R&D
is positively related to M&A activity, though the other variables are insignificant. The insignifi-
cance likely occurs because the time-series variation in M/B and returns is small, relative to the
cross-sectional variation, which is captured by the industry fixed effects. In other unreported tests,
we find positive, though insignificant interactions between close merger activity and industry R&D
expenditures. We also verify that these results are robust to assigning firms to IO industries using
all listed NAICS and SIC codes in Internet Appendix Table XV.
Next, in Internet Appendix Table XVI, we study the effect of the closeness-weighted M&A
activity on a subject industry’s horizontal mergers, excluding cross-industry mergers. This tests
whether exposure to other industries that are experiencing heightened M&A activity causes internal
reallocation of assets within an industry, as opposed to vertical mergers. We find a positive and
significant effect on lagged closeness-weighted M&A activity in all specifications, consistent with
our main results. The economic significance relative to the average is the same or stronger than
the main results. These results indicate that the diffusion of M&As can happen indirectly as
industries move assets internally to more efficient ownership arrangements, rather than directly
through cross-industry mergers with other vertically integrating industries.
In a series of robustness checks, we control for the influence of mergers that flow across industries
through asset complementarities, rather than customer-supplier links. Similar to our definition of
Closeness-Weighted M&A Activity, for each industry we calculate the exposure to high levels of
M&As in industries that share asset complementarities, based on the measure of Hoberg and Phillips
(2010a) and Hoberg and Phillips (2010b). In particular, we calculate:
HP-Weighted M&A Activityit =∑
j 6=i
HPijt
∑
k 6=i
vjkt (3)
where HPijt is an indicator variable that equals one if two IO industries have any firms that
Hoberg and Phillips identify as similar in a given year, using their text-based similarity scores.10
High merger activity, vjkt, is recorded as above. Thus, this variable measures the amount of
10Similar results are obtained if we use a continuous variable to measure the number or fraction of firms that aresimilar according to HP.
32
merger activity in close industries, where closeness is based on asset complementarity, rather than
direct product market relations. In Internet Appendix Tables XVII, XVIII, and XIX, we run
robustness tests that include the HP-based measure as an explanatory variable, for both detail and
summary-level industry definitions, and controlling for additional variables such as R&D. We find
that merger activity in close industries, based on asset complementarity, has a positive, though not
always significant, effect on a subject industry’s merger activity. This means that merger activity
transmits across industries connected through similar product offerings. At the same time, we find
that the transmission of merger activity across the customer-supplier network remains strongly
significant and positive, as in our main tests. These results show that mergers propagate through
multiple economic links between industries.
To provide more evidence for the effect of misvaluation-driven mergers, we follow the same
methodology as in the previous tests to calculate closeness-weighted acquirer returns and lags of
own-industry returns. In Internet Appendix Table XX, we find that the one-year lagged closeness-
weighted returns are positively and significantly related to an industry’s own acquirer returns.
This means that acquirers are more likely to have high returns (above the 75th percentile of the
industry-pair time series) if acquirers in closely-connected industries had unusually high returns in
the prior year. Though we view this evidence as supportive of an efficiency motivation for mergers
that flow through the customer-supplier network, as we noted before, our focus is on the incidence
of mergers, not their value implications, which is beyond the scope of this paper.
C. The Effect of Close Versus Distant M&As
In this section, we investigate how the distance between two industries across the IO network
affects the diffusion of merger activity at different time lags. In contrast to the prior section,
we separately identify the timing of the impact of mergers in close versus distant industries. If
merger activity is diffusing across an economy, we would expect that the merger activity in closely
connected industries would have a greater impact on an industry’s merger activity in the near
future than would the merger activity in distantly connected industries. In addition, if merger
activity diffuses in a wave-like pattern, we expect to observe a positive relationship between time
and distance, such that M&A activity in more distantly-connected industries has a delayed positive
33
effect on a subject industry’s future merger activity. The following equation captures these effects:
log(1 + vi,t) = α+
4∑
s=1
ρs log(1 + vi,t−s) +
4∑
s=1
θsvClose,t−s +
4∑
s=1
φsvDistant,t−s + γi + τt + εi,t, (4)
where vi,t is as defined above, and vClose,t and vDistant,t−s is the aggregate merger activity in close
and distant industries, not including mergers with firms in industry i. Close industries are defined
as the industries that are directly connected to the subject industry through a customer or supplier
link above the one percent threshold. Distant industries are those that have the maximum shortest
path from the subject industry. In the supplier network, this is three connections away. In the
customer network, it is four connections.11
One concern with the empirical model is the potential for multicollinearity in the variables.
Internet Appendix Table XXI shows that there is persistence in merger activity in both time and
network distance. Therefore, we use only the closest and most distant industries, as opposed to
the full set of discrete distances from one to the maximum, to avoid multicollinearity. Thus, in
contrast to the weighted measure of closeness used previously, this model separately identifies the
impact of mergers in the closest industries compared to the most distant industries.
Table VIII presents the coefficient estimates of Equation 4. First, in columns 1 and 5, we includeTable VIII
only the merger activity in close industries to avoid multicollinearity in the explanatory variables.
Similar to the results in Table VII, we find a strong positive relation between lagged merger activity
in close industries and current merger activity in the subject industry.
In columns 2 and 6, we include only distant industry variables. When distance is measured
through customer relations in column 2, we find a positive effect for distant M&As that occurred
two years prior. For the supplier network, we find positive relations in distant industries at a lag
of three and four years. Columns 3 and 7 include close and distant in the same model, with little
change in results. These results are consistent with a wave-like diffusion process, where merger
activity in distant industries spills over into industries connected through the customer-supplier
network. These results could also shed light on the channel through which shocks spread through
11The maximum shortest distance is six in the customer network, but the numbers of mergers in industries that aremore than four steps away is orders of magnitude less than in the closer industries. This reflects that industries inthe periphery do not make many mergers. Thus, to make sure our results are not driven by outliers, we use industriesthat are four steps away as our ‘Distant’ industries. If anything, this will make our results biased towards zero.
34
customer-supplier relations. Though we imagine that the strongest effect of the supply chain follows
from reorganizations in the close industry, this is not the only way that reorganization in the supply
chain can matter. For example, mergers in distant industries could lead to changes in trade relations
of close industries, without industry restructuring, which could then lead to industry restructuring
in the subject industry. Alternatively, following merger waves in more distant industries, a subject
industry may anticipate changes to close industries and preemptively reorganize, without close
industries experiencing a merger wave. The degree to which distant M&A activity matters, after
controlling for close M&A activity, provides evidence that shocks travel through the IO network,
but not necessarily in the form of direct reorganization in the close industries.
Finally, as before, we include concurrent merger activity to account for any spurious correla-
tion caused by an omitted variable that drives current and lagged activity. For customer links,
the one-year lagged close merger activity remains positive and significant, though distant merger
activity becomes insignificant. For supplier links, both the one-year lagged close merger activity
and the three-year lag of distant activity remain positive and significant. In addition, we find that
distant activity at the one-year lag is also positive and significant. Because of the concerns of
multicollinearity, we view the positive effect of distant mergers with a three-year lag as the most
robust result for suppliers.
In Internet Appendix Table XXII, we present robustness tests using the summary-level industry
definitions. We find consistent results though with weaker significance levels. In Internet Appendix
Table XXIII, we control for time-varying industry characteristics and find similar or stronger results,
including a positive and significant effect of one-year lagged close M&A activity and positive and
significant effects for three- and four-year lagged distant M&A activity. R&D expenditures are
positively related to merger activity. Next, in Internet Appendix Table XXIV, we find a positive
and significant effect for lagged close M&A activity on horizontal mergers, but the impact of distant
M&As becomes insignificant. This suggests that the indirect effect of M&As in related industries
on internal industry reorganization is driven by mergers in close industries, not distant industries.
Finally, Internet Appendix Table XXV presents tests based on acquirer returns and finds patterns
similar to the patterns of merger incidence, where returns in close industries have a positive effect
after one year, and returns in distant industries have a positive effect after two years.
35
The results in this section provide evidence that mergers follow a wave-like pattern across the
input-output network topology. Similar to the illustration of the timber industry, high merger
activity in close industries affects the subject industry’s merger activity with a one-year delay,
whereas merger activity in distant industries takes a longer time to impact the subject industry.
It is also interesting to note that the effects of related industries mergers differs between customer
and supplier links. First, exposure to M&As through supplier links has a stronger effect than
through customer links. Second, the diffusion of mergers through supplier links occurs more rapidly
than through customer links. Consistent with prior research that indicates that economic shocks
are more likely to travel upward through a supply chain than downward (Hertzel, Li, Officer, and
Rodgers, 2008; Bhattacharyya and Nain, 2011), we find that merger shocks travel faster upward
through suppliers than they travel downward through customers. Our network approach identifies
one possible reason for this phenomenon. As we show in Table I, the supplier network is denser
and more interconnected than the customer network. Given this structural difference, it is not
surprising that shocks travel faster through the denser supplier network.
IV. Network Centrality and Aggregate Merger Waves
In this section, we investigate how industry-level diffusion of merger activity relates to the time
series of economy-wide aggregate merger waves. One possibility is that aggregate merger waves
occur as merger activity transmits towards central industries. Recall that industries in both the IO
and merger networks have highly skewed distributions of connections, which approximate power
law distributions. A few ‘hub’ industries have many direct connections, while many industries have
relatively few direct connections. Shocks that follow the IO network will move towards the center
of the network, branching out in parallel to other industries.
An alternative possibility is that aggregate merger waves are composed of industries on the
periphery of the network. While central industries have more product market connections to other
industries than do peripheral industries, their diversity of connections could reduce the impact of
any one shock in a connected industry. In addition, firms in central industries could have merger
options across a greater diversity of connected industries, raising the threshold for selecting a merger
partner. In this scenario, the limited, but relatively important local connections of peripheral
36
industries could lead to their participation in aggregate merger waves. Ultimately, it is an empirical
question whether having more connections or stronger connections determines whether an industry
will experience merger waves during economy-wide aggregate merger waves.
Figure 6 compares the time series of the percentage of total industries that are experiencing
a merger wave in a given year to the time series of the average input-output centrality of those
industries that are experiencing a merger wave. Merger waves are defined as above, as industry-Figure 6
years where the number of mergers of firms in an industry is greater than the 75th percentile of
the number of yearly mergers over 1986 to 2010. Thus, the greater is the fraction of industries that
are merging in a given year, the greater is the aggregate merger wave. During the peak in 1998,
65% of the 471 industries in our main sample experience a merger wave.
The figure shows a strong correlation between the two time series, indicating that firms in
central industries are more likely to merge during aggregate merger waves. In Table IX, we test
this relationship statistically. First, in Panel A, we run time-series regressions of the current andTable IX
lagged percentages of industries in a merger wave on the average centrality of the wave industries.
Standard errors are corrected for heteroskedasticity and autocorrelation following the procedure of
Newey and West (1987) and using the automatic lag selection model of Newey and West (1994).
We find a strong statistically significant relation between current aggregate merger activity and
centrality whether assigning firms to industries based on their primary NAICS codes or based on
all of their NAICS codes. These results show that more central industries experience merger waves
concurrently with aggregate merger waves.
In Panel B, we run predictive vector autoregressions (VARs) of the percentage of industries
in merger waves and their centrality. Standard errors are again corrected using the Newey-West
procedures. We find that lagged aggregate merger activity is positively related to current centrality.
The results in Panel B also show that an increase in aggregate merger activity Granger causes an
increase in the average centrality of the industries in merger waves. There is weak evidence of
reverse Granger causation as well. In Internet Appendix Tables XXVI, XXVII, and XXVIII, we
show that these results are robust to using the 1982 IO industry relations, the summary-level
industry definitions, to defining merger waves using a 50th percentile threshold, and to excluding
37
the tech bubble years 1997 through 2002. These results confirm that there is a strong positive
time-series relation between aggregate merger waves and the centrality of the firms in the waves.
The distinction between the tests in Panel A and Panel B is that Panel A includes concurrent
effects, whereas Panel B is a forward-looking predictive regression and also includes two equations
that are simultaneously estimated. In Panel A, we find that in years when more industries have a
merger wave, those industries are more central, relative to the average year. In Panel B, we find
that when there are more industries experiencing a merger wave in this year, the centrality of the
industries experiencing a merger wave next year is higher, relative to the average year. However, this
does not say that the centrality of industries this year is also not high compared to an average year,
as in Panel A. It is simply not tested in Panel B. More generally, the tests in Panel A of Table IX
show that aggregate merger waves are composed of central industries. The predictive tests in Panel
B show that waves propagate towards the center over time, rather than away from the center. In
addition, the evidence from these tests supports the idea that the number of connections is more
important for the transmission of merger shocks than the strength of connections.
These results provide a new explanation for why aggregate merger waves occur. A criticism of
prior research that argued that economic industry shocks produced merger waves was that random
industry shocks could not explain why overall merger activity in an economy was also not randomly
distributed over time. We argue that the initial economic industry shocks may be random, but
the subsequent ‘after-shocks’ follow the IO links, which are not random. Instead, the pattern of
aggregate merger waves is likely driven by the fat-tailed nature of product market connections. A
similar argument is made by Gabaix (2011) regarding productivity shocks. He argues that since
the distribution of firm sizes is also approximately a power law distribution, idiosyncratic shocks
to small firms do not average out idiosyncratic shocks to large firms.
If aggregate merger waves are a result of the structure of the IO network, then we can draw direct
relationships between networks and characteristics of merger waves. For instance, the speed with
which the central industries are affected depends on how highly connected the central industries
are to the rest of the network. In a dense network, the central industries are highly connected to
most of the other industries and will be affected quickly when any industry undergoes a merger
wave. In a sparser network, merger activity will propagate to the center more slowly. When the
38
shock spreads to a central industry, it will quickly spread to other central industries, creating the
observed jump pattern in average centrality. In addition, we have also shown in Tables I and VIII
that the supplier network is denser than the customer network and that mergers tend to diffuse
faster through supplier networks than through customer networks. This implies that aggregate
merger waves are more likely to be caused by the transmission of shocks through suppliers towards
central industries.
V. Conclusion
Using detailed data from the Bureau of Economic Analysis, this paper models the U.S. economy
as a network of industries connected through customer-supplier trade flows. Larger trade flows
represent stronger inter-industry connections. We hypothesize that economic shocks will travel
across the economy through this network in a predictable way. We investigate one type of economic
shock: merger waves. Neoclassical theory argues that mergers represent efficient reallocations of
resources. This suggests that the timing and incidence of cross-industry merger waves will be
influenced by the real economic linkages in the industry network. We test three hypotheses: 1)
inter-industry mergers will cluster in industry-pairs with strong trade flows; 2) merger waves will
propagate across industries through customer-supplier links; and 3) the structure of the customer-
supplier network will influence which industries are involved in economy-wide merger waves.
First, we find that cross-industry mergers are highly clustered in a small number of industry-
pairs. Out of all possible industry-pairs, 94% experience no mergers at all during 1986 to 2010.
This means every cross-industry merger in our sample occurs in just 6% of industry-pairs. This
pattern is almost identical for cross-industry trade flows, where 95% of industry-pairs have almost
zero or no trade flows at all. Using network techniques from graph theory, we find that the pattern
of cross-industry mergers and the customer-supplier network are similar in other ways. Both exhibit
small-world properties, where an average industry is just two or three links away from every other
industry. Industries that are more central and clustered in the customer-supplier network are also
more central and clustered in the network of inter-industry mergers.
Second, we find that industry merger activity travels in a wave-like pattern through customer-
supplier links. We measure the distance between all industries in the customer-supplier network,
39
where distance is calculated based on the strength of trade flows. For each industry, we measure the
intensity of merger activity in all other industries, excluding mergers with the industry itself. We
find that an industry that is exposed to mergers in close industries experiences increased merger
activity in the following year. Merger activity in distant industries leads to increased merger
activity in two or three years. Thus, mergers propagate through the customer-supplier network
in a predictable, wave-like pattern. We also find that these effects are stronger and the delay is
smaller when shocks travel through supplier links, compared to customer links. This likely reflects
that the network of suppliers is more densely connected than the network of customers.
Third, we show that the industries that experience merger waves in concert with an aggregate
merger wave are more central in the product market network. Our evidence suggests that merger
waves travel through the customer-supplier links towards central industries, which then diffuse
merger activity outward across many different industries. These results show that even if industry-
level merger waves are motivated by random industry shocks, the structure of the customer-supplier
network leads to aggregate economy-wide merger waves.
The results in this paper show that merger waves are driven, in part, by economic fundamentals
of product market relations. Our results are robust to proxies for overvaluation-driven mergers.
Even if merger waves are caused by the spread of misvaluation across product market relations, our
results imply that it still must be the case that the underlying exogenous product market relations
explain a large portion of merger activity across an economy over time. Finally, though we do
not claim to conduct rigorous tests of competing theories of vertical integration, we find evidence
consistent with an incomplete contracts model, where asset complementarity and holdup problems
motivate vertical mergers.
One of the primary innovations of this paper is to model merger waves in a network setting where
networks are defined by actual trade flows across industries. Using the well-developed techniques
from network and graph theory, we are able to analyze a much more complex dynamic process of
merger waves than has been done in prior research. To our knowledge, this is the first paper to
model inter-industry trade flows as a network. We believe that this approach will prove to have a
multitude of applications in economics, beyond merger waves.
40
REFERENCES
Acemoglu, Daron, Simon Johnson, and Todd Mitton, 2009, Determinants of vertical integration:
Financial development and contracting costs, Journal of Finance 64, 1251–1290.
Ahern, Kenneth R., 2012, Bargaining power and industry dependence in mergers, Journal of Fi-
nancial Economics 103, 530–550.
Albert, Reka, and Albert-Laszlo Barabasi, 2002, Statistical mechanics of complex networks, Review
of Moden Physics 74, 47–97.
Allen, Jeffrey W., and Gordon M. Phillips, 2000, Corporate equity ownership, strategic alliances,
and product market relationships, The Journal of Finance 55, 2791–2815.
Arellano, Manuel, and Stephen Bond, 1991, Some tests of specification for panel data: Monte
carlo evidence and an application to employment equations, The Review of Economic Studies
58, 277–297.
Becker, Mary J., and Shawn Thomas, 2010, The spillover effects of changes in industry concentra-
tion, Unpublished Working Paper, University of Pittsburgh.
Bhattacharyya, Sugato, and Amrita Nain, 2011, Horizontal acquisitions and buying power: A
product market analysis, Journal of Financial Economics 99, 97–115.
Bonacich, Philip, 1972, Factoring and weighting approaches to status scores and clique identifica-
tion, Journal of Mathematical Sociology 2, 113–120.
Borgatti, Stephen P., 2005, Centrality and network flow, Social Networks 27, 55–71.
Clauset, Aaron, Cosma R. Shalizi, and M.E.J. Newman, 2009, Power-law distributions in empirical
data, SIAM Review 51, 661–703.
Coase, R. H., 1937, The nature of the firm, Economica 4, 386–405.
Denis, David J., Diane K. Denis, and Atulya, 1997, Agency problems, equity ownership, and
corporate diversification, Journal of Finance 52, 135–160.
Dijkstra, E. W., 1959, A note on two problems in connexion with graphs, Numerishce Mathematik
1, 269–271.
Duchin, Ran, and Breno Schmidt, 2012, Riding the merger wave, Journal of Financial Economics,
forthcoming.
41
Eckbo, B. Espen, 1983, Horizontal mergers, collusion, and stockholder wealth, Journal of Financial
Economics 11, 241–273.
Economic Classification Policy Committee, 1993, Criteria for determining industries, Discussion
Paper 4 Economic Classification Policy Committee.
Fan, Joseph P.H., and Vidhan K. Goyal, 2006, On the patterns and wealth effects of vertical
mergers, Journal of Business 79, 877–902.
Fee, C. Edward, Charles J. Hadlock, and Shawn E. Thomas, 2006, Corporate equity ownership and
the governance of product market relationships, Journal of Finance 61, 1217–1251.
Fee, C. Edward, and Shawn Thomas, 2004, Sources of gains in horizontal mergers: Evidence from
customer, supplier, and rival firms, Journal of Financial Economics 74, 423–460.
Ferris, Ann, 2009, Environmental regulation and labor demand: The Northern Spotted Owl, Uni-
versity of Michigan, Working Paper.
Gabaix, Xavier, 2009, Power laws in economics and finance, Annual Review of Economics 1, 255–
293.
, 2011, The granular origins of aggregate fluctuations, Econometrica 79, 733–772.
Galbraith, John Kenneth, 1952, American Capitalism: The Concept of Countervailing Power (M.
E. Sharpe, Inc.).
Garfinkel, Jon A., and Kristine Watson Hankins, 2011, The role of risk management in mergers
and merger waves, Journal of Financial Economics 101, 515–532.
Giroud, Xavier, and Holder M. Mueller, 2010, Does corporate governance matter in competitive
industries?, Journal of Financial Economics 95, 312–331.
Gollop, Frank M., 1994, The heterogeneity index: A quantitative tool to support industrial classi-
fication, Discussion Paper 2 Economic Classification Policy Committee.
Grossman, Sanford J., and Oliver D. Hart, 1986, The costs and benefits of ownership: A theory of
vertical and lateral integration, Journal of Political Economy 94, 691–719.
Harford, Jarrad, 2005, What drives merger waves?, Journal of Financial Economics 77, 529–560.
Hart, Oliver, 1995, Firms, Contracts, and Financial Structure (Clarendon Press: Oxford, England).
, and John Moore, 1990, Property rights and the nature of the firm, Journal of Political
Economy 98, 1119–1158.
42
Hertzel, Michael G., Zhi Li, Micah S. Officer, and Kimberly J. Rodgers, 2008, Inter-firm linkages
and the wealth effects of financial distress along the supply chain, Journal of Financial Economics
87, 374–387.
Hoberg, Gerard, and Gordon Phillips, 2010a, Dynamic text-based industry classifications and en-
dogenous product differentiation, Unpublished working paper. University of Maryland, College
Park, MD.
, 2010b, Product market synergies and competition in mergers and acquisitions: A text-
based analysis, Review of Financial Studies 23, 3773–3811.
Horn, Henrick, and Asher Wolinsky, 1988, Bilateral monopolies and incentives for merger, The
RAND Journal of Economics 19, 408–419.
Inderst, Roman, and Christian Wey, 2003, Bargaining, mergers, and technology choice in bilaterally
oligopolistic industries, The RAND Journal of Economics 34, 1–19.
Joskow, Paul L., 1985, Vertical integration and long term contracts: The case of coal-burning
electric generating stations, Journal of Law, Economics, and Organization 1, 33–80.
, 1987, Contract duration and relationship specific investments, American Economic Review
77, 168–175.
Katz, Michael L., 1987, The welfare effects of third-degree price discrimination in intermediate
good markets, The American Economic Review 77, 154–167.
Klein, Benjamin, Robert G. Crawford, and Armen A. Alchian, 1978, Vertical integration, appropri-
able rents, and the competitive contracting process, Journal of Law and Economics 21, 297–326.
Lamont, Owen, 1997, Cash flow and investment: Evidence from internal capital markets, Journal
of Finance 52, 83–109.
Lieberman, Marvin, 1991, Determinants of vertical integration: An empirical test, Journal of
Industrial Economics, Special Issue on Vertical Relationships 39, 451–466.
MacKay, Peter, and Gordon M. Phillips, 2005, How does industry affect firm financial structure?,
Review of Financial Studies 18, 1433–1466.
Maksimovic, Vojislav, and Gordon Phillips, 2001, The market for corporate assets: Who engages
in mergers and asset sales and are there efficiency gains?, Journal of Finance 56, 2019–2065.
43
Maksimovic, Vojislav, Gordon M. Phillips, and Liu Yang, 2010, Private and public merger waves,
Journal of Finance, forthcoming.
Masten, Scott E., 1984, The organiation of production: Evidence from the aerospace industry,
Journal of Law and Economics 27, 403–417.
Mitchell, Mark L., and J. Harold Mulherin, 1996, The impact of industry shocks on takeover and
restructuring activity, Journal of Financial Economics 41, 193–229.
Newey, Whitney K., and Kenneth D. West, 1987, A simple, positive semi-definite, heteroskedasticity
and autocorrelation consistent covariance matrix, Econometrica 55, 703–708.
, 1994, Automatic lag selection in covariance matrix estimation, The Review of Economic
Studies 61, 631–653.
Ohanian, Nancy Kane, 1994, Vertical integration in the U.S. pulp and paper industry, 1900-1940,
The Review of Economics and Statistics 76, 202–207.
Ovtchinnikov, Alexei V., 2010, Merger waves following industry deregulation, Unpublished Working
Paper, Vanderbilt University.
Perry, Martin K., 1978a, Price discrimination and forward integration, The Bell Journal of Eco-
nomics 9, 209–217.
, 1978b, Vertical integration: The monopsony case, The American Economic Review 68,
561–570.
, 1989, Vertical integration: Determinants and effects, in Richard Schmalensee, and
Robert D. Willig, ed.: Handbook of Industrial Organization., vol. I . pp. 185–255 (North Holland:
Amsterdam,).
Rhodes-Kropf, Matthew, and David Robinson, 2008, The market for mergers and the boundaries
of the firm, Journal of Finance 62, 1169–1211.
Rhodes-Kropf, Matthew, David T. Robinson, and S. Viswanathan, 2005, Valuation waves and
merger activity: The empirical evidence, Journal of Financial Economics 77, 561–603.
Schmalensee, Richard, 1973, A note on the theory of vertical integration, The Journal of Political
Economy 81, 442–449.
Shahrur, Husayn, 2005, Industry structure and horizontal takeovers: Analysis of wealth effects on
rivals, suppliers, and corporate customers, Journal of Financial Economics 76, 61–98.
44
Shleifer, Andrei, and Robert W. Vishny, 2003, Stock market driven acquisitions, Journal of Finan-
cial Economics 70, 295–311.
Spengler, Joseph J., 1950, Vertical integration and antitrust policy, The Journal of Political Econ-
omy 58, 347–352.
Stillman, Robert, 1983, Examining antitrust policy towards horizontal mergers, Journal of Finan-
cial Economics 11, 225–240.
Stuckey, John A., 1983, Vertical Integration and Joint Ventures in the Aluminum Industry (Harvard
University Press: Cambridge).
Tirole, Jean, 1988, The Theory of Industrial Organization (The MIT Press: Cambridge, Mas-
sachusetts).
Vernon, John M., and Daniel A. Graham, 1971, Profitability of monopolization by vertical integra-
tion, The Journal of Political Economy 79, 924–925.
Warren-Boulton, Frederick R., 1974, Vertical control with variable proportions, The Journal of
Political Economy 82, 783–802.
Watts, Duncan J., and Steven H. Strogatz, 1998, Collective dynamics of ‘small-world’ networks,
Nature 393, 440–442.
Williamson, Oliver E., 1979, Transaction-cost economics: The governance of contractual relations,
Journal of Law and Economics 22, 233–261.
45
Forestry SupportForest NurseriesLoggingSawmillsPulp MillsWood Doors
0 0 0 0 0 014 0 0 0 0 06 48 21 0 0 00 30 38 9 0 00 0 2 1 1 00 0 0 3 0 0
(a) Adjacency Matrix Representation of the Timber Network (% of Sales Purchased)
Forestry SupportForest NurseriesLoggingSawmillsPulp MillsWood Doors
0 0 0 0 0 064 1 1 0 0 07 29 41 0 0 00 11 50 17 0 00 0 24 14 1 00 0 0 18 0 1
(b) Adjacency Matrix Representation of the Timber Network (% of Input Supplied)
ForestrySupport
ForestNurseries
Logging
Sawmills
PulpMills
WoodDoors
64%
14%
7%
6%
29%
48%
11%
30%
50%
38%
24%
2%
14%
1%
18%
3%
Legend
Seller
Buyer%
ofInp
ut
Supplied
%of
Sales
Purch
ased
(c) Graphical Representation of the Timber Network
Figure 1A Portion of the Timber Industry NetworkThis figure presents the adjacency matrices of subsets of the customer and supplier networks fromthe 1997 U.S. Bureau of Economic Analysis Input-Output tables. The column labels of the adjacencymatrices are the transpose of the row labels, and are omitted for brevity. Each entry of the adjacencymatrix in Panel (a) is the percentage of total sales of the column industry that is purchased by therow industry. Each entry in the adjacency matrix in Panel (b) is the percentage of total non-laborinput costs of the row industry that are purchased by the column industry. Panel (c) presents bothadjacency matrices in a graphical representation. For each industry-pair, the arrows point fromsuppliers to customers. The number of the top of the arrow gives the total inputs purchased by thecustomer from the seller, as a percentage of total inputs purchased from all sources, as reported inPanel (b). The number on the bottom of the arrow is the total sales purchased by the customerindustry, as a percetange of the supplier’s total sales to all customer industries, as reported in theadjacency matrix in Panel (a). The opposite customer-supplier relations exist (e.g., sawmills supplyto logging), but they are not reported in this figure.
46
Billions(2010Dollars) N
umber
ofMergers
Number of Mergers(right scale)
Value of Mergers(left scale)
1990 1995 2000 2005 2010
300
600
900
1200
1500
500
1000
1500
2000
2500
Figure 2Dollar Value and Number of Mergers, 1986–2010Aggregate merger volume in 2010 adjusted U.S. dollars and by the number of mergers. Merger datais from SDC.
47
bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bcbcbc
bcbc
bc
1 10 100 10000.1%
1%
10%
100%
Prob(x
≥X)
# of Inter-Industry Connections, X
(a) Supplier Network
bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bcbc bc bc bc bc
bcbcbc
bc
bc
1 10 100 10000.1%
1%
10%
100%
Prob(x
≥X)
# of Inter-Industry Connections, X
(b) Merger Network
Figure 3Degree Distribution of Merger and IO NetworksThese figures represents the distribution of degree centrality in log-log scale. Circles represent the degree centrality of industries,indicating how many direct connections an industry has to other industries. The dashed line is the from the estimate of the alpha termin the power distribution P (k) = ck−α. There are 471 detail-level industries using the 1997 IO Tables produced by the U.S. Bureauof Economic Analysis. Supplier network connections occur if an industry supplies more than 1% of the total inputs of a customerindustry. Merger network connections occur if there exist any cross-industry mergers. The merger data are over 1986–2010 from SDC.
48
t-statistic
Target Buys from Acquirer
Acquirer Buys from Target
Acquirer Sells to Target
bc bc Target Sells to Acquirer
bc
bc bc
bcbc
bc bcbc bc bc
bc bc bc bc bc bc
bcbc bc
bcbc bc
bc
bcbc
1990 1995 2000 2005 2010
0
5
10
15
Figure 4t−Statistics from Yearly ERGM TestsThis figure represents the t−statistic on each of the four IO networks (Target Buys from Acquirer,Acquirer Buys from Target, Acquirer Sells to Target, and Target Sells to Acquirer) from yearly expo-nential random graph model (ERGM) tests from 1986 to 2010. See Table VI for variable definitionsand data sources.
49
BillionsofBoardFeet
&LogPrice
Index Volume Harvested
Log Price Index
1988 1990 1992 1994 1996 1998 2000 2002 2004
3.5
5.0
6.5
8.0
9.5
11.0
(a) Oregon Timber Industry
Percentile
ofIndustry
Merger
Activity Sawmills
Forest Nurseries
Logging
bc bc Pulp Millsbc
bc
bc bc
bc
bc
bc bc
bc
bcbc bc
bc
bc
bc
bc
1990 1994 1998 2002
0.2
0.4
0.6
0.8
1.0
(b) Coincident Merger Activity
Percentile
ofIndustry
Merger
Activity Sawmillst
Forest Nurseriest+1
Loggingt+3
bc bc Pulp Millst+3bc
bc
bc bc
bc
bcbc bc
bc
bc
1990 1992 1994 1996 1998
0.2
0.4
0.6
0.8
1.0
(c) Leading Merger Activity
Figure 5Diffusion of Merger Activity in Timber-Related IndustriesPanel (a) presents the volume (in billions of board feet) and log price index for Oregon timber. Dataare from the Oregon Department of Forestry, Annual Timber Harvest Reports. Panel (b) presentsindustry merger activity in Bureau of Economic Analysis IO industry classifications: 1) Sawmills,2) Forest nurseries, forest products, and timber tracts, 3) Logging, and 4) Pulp Mills. For eachindustry-year, figures present the two-year moving-average of the percentile of the number of mergersinvolving firms in each industry over the period 1986 to 2008. Panel (c) uses the one-year leadingdata for Forest nurseries, and the three-year leading data for Logging and Pulp Mill mergers.
50
Mean
Degree
Centrality
PercentofIndustries
inWave
Mean Centrality(line)
% of Industries in Wave(bars)
1990 1995 2000 2005 2010
0.2
0.4
0.6
0.8
1.0
10%
20%
30%
40%
50%
60%
Figure 6Network Centrality and Industry Merger WavesVertical bars represent the percentage of all 471 industries in year t that are experiencing a mergerwave. A merger wave is when an industry has more mergers than the 75th percentile of mergers,relative to the industry’s time series of mergers from 1986 to 2010. The black line represents theaverage degree centrality of the industries that are experiencing a merger wave, where centrality iscomputed from the weighted and directed network of supplier links between industries in the 1997detailed-level input-output network. Merger firms are classified into industries based on their primaryNAICS code as reported in SDC.
51
Table IInput-Output Summary StatisticsThis table presents summary statistics of the Input-Output relationships of industries as defined by the 1997 Bureau of EconomicAnalysis Input-Output (IO) Detail-Level Industry classification (IO industries). Inter-industry pairs include all combinations of theIO industries (excluding own-industry pairs). Inter-industry pairs > 1% are only those observations where either Customer % orSupplier % is greater than 1%. Intra-industry observations include relations of firms that are in the same IO industry. Customer %is the percentage of industry i’s sales that are purchased by industry j. Supplier % is the percentage of industry i’s inputs that arepurchased from industry j. All numbers, except observations, are in percentages.
Customer % Supplier %
Inter-IndustryPairs
Inter-IndustryPairs> 1%
Intra-Industry
Inter-IndustryPairs
Inter-IndustryPairs> 1%
Intra-Industry
Mean 0.220 5.060 3.310 0.270 3.860 4.110Median 0.010 2.190 1.140 0.010 2.200 1.4005th percentile 0.000 1.060 0.000 0.000 1.050 0.00095th percentile 0.620 18.270 12.470 0.980 10.870 16.010
Frequency Percentage0–1% 96.568 — 47.346 95.131 — 42.6751–2% 1.567 45.644 12.527 2.231 45.816 14.4372–3% 0.615 17.926 6.582 0.810 16.627 5.9453–4% 0.329 9.582 4.246 0.456 9.371 4.2464–5% 0.190 5.528 5.945 0.339 6.959 4.459> 5% 0.732 21.321 23.355 1.034 21.228 28.238
52
Table IIMerger Summary StatisticsThis table presents summary statistics of the sample of mergers over the period 1986 to 2010 byindustry pairs. Merger data is from SDC. Industries are defined by the 1997 Bureau of EconomicAnalysis Input-Output (IO) Detail-Level Industry classifications (IO Industries). Inter-industry pairsinclude all combinations of the IO industries (excluding own-industry pairs). Industry-level observa-tions aggregate the industry-pair data to a single IO industry. Intra-industry observations includemergers of firms that are in the same IO industry. Inter-industry observations at the industry-levelincludes all inter-industry mergers across all other industries for each of the IO industries dividedby two, since each inter-industry merger is double-counted at the industry-level. 2010 millions of USdollars are reported in brackets.
Industry-Level
Inter-Industry Pairs Inter-Industry Intra-Industry
Observations 110,685 471 471
Total Mergers 31,040 31,040 19,962[$10,135,331] [$10,135,331] [$6,636,782]
Mean 0.28 65.90 42.38[$92] [$21,519] [$14,091]
Median 0.00 15.00 4.00[$0] [$2,867] [$244]
5th Percentile 0.00 1.50 0.00[$0] [$87] [$0]
95th Percentile 1.00 287.50 200.00[$8] [$75,990] [$53,046]
Maximum 1,008 3,320 3,118[$410,643] [$1,749,955] [$1,153,641]
Frequency Percentage
None 94.16 0.85 18.471 3.35 4.46 11.682–5 1.70 15.29 26.546–20 0.59 39.28 23.3521–50 0.13 20.38 7.86>50 0.07 19.75 12.10
53
Table IIIMean and Median Network Statistics by NetworkDegree centrality is an industry’s number of inter-industry connections. IO degree centrality ismeasured using the binary connections in the 1997 US BEA Input-Output Networks (Customer orSupplier) at the detail-level. A binary connection is defined as a connection where one industryeither supplies at least 1% of the connected industry’s inputs, or buys at least 1% of the connectedindustry’s output. Merger degree centrality is measured using the binary network of inter-industrymergers, where a binary connection is defined as any inter-industry mergers between two industriesover 1986 to 2010. See text for definitions of eigenvector centrality, average shortest path, clusteringcoefficient, and max(shortest distance). Top cells are means and bottom cells are medians, in brackets.
Network
Supplier Customer Merger
Degree Centrality 22.883 16.132 27.444[16.000] [13.000] [18.000]
Eigenvector Centrality 0.037 0.033 0.046[0.033] [0.025] [0.045]
Average Shortest Path 1.966 2.537 2.075[1.971] [2.467] [2.032]
Clustering Coefficient 0.461 0.275 0.422[0.462] [0.250] [0.400]
Max(Shortest Distance) 3.000 6.000 5.000
54
Table IVThe Most Central Industries in the IO and Merger NetworksDegree centrality is an industry’s number of inter-industry connections. IO degree centrality ismeasured using the binary connections in the Input-Output Network using data from the U.S. Bureauof Economic Analysis for 1997 at the detail-level. A binary connection is defined as a connectionwhere one industry either supplies at least 1% of the connected industry’s inputs, or buys at least 1%of the connected industry’s output. Merger degree centrality is measured using the binary network ofinter-industry mergers, where a binary connection is defined as any inter-industry mergers betweentwo industries over 1986 to 2010. Boldfont indicates a merger industry also in the top 10 IO industries.
Rank Supplier Network Customer Network Merger Network
1 Wholesale trade Construction Securities, commodity contracts,& investments
2 Mgmt. of companies & enterprises Wholesale trade Wholesale trade
3 Truck transp. Retail trade Retail trade
4 Power generation & supply Motor vehicle parts manuf. Construction
5 Real estate Real estate Funds, trusts, & other financialvehicles
6 Iron & steel mills Food srvcs. & drinking places Software reproducing
7 Paperboard container manuf. Hospitals Motor vehicle parts manuf.
8 Plastics plumbing fixtures & Telecommunications Information srvcs.all other plastics products
9 Monetary auth. & depository Iron & steel mills All other electronic componentcredit intermed. manufacturing
10 Lessors of nonfinancial Power generation & supply Mgmt. consulting srvcs.intangible assets
55
Table VCorrelations Between Industry Characteristics Across NetworksCentrality is an industry’s number of inter-industry connections. IO degree centrality is measured using the binary connections in the1997 US BEA Input-Output Networks (Customer or Supplier) at either the detail-level of industries. A binary connection is definedas a connection where one industry either supplies at least 1% of the connected industry’s inputs, or buys at least 1% of the connectedindustry’s output. Merger centrality is measured using the binary network of inter-industry mergers, where a binary connection isdefined as any inter-industry mergers between two industries over 1986 to 2010. See text for definitions of average shortest path andclustering coefficient p−values of are reported in parentheses. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗, for the 0.01, 0.05,and 0.10 levels.
CustomerCentrality
SupplierCentrality
MergerCentrality
CustomerAvg. Path
SupplierAvg. Path
MergerAvg. Path
CustomerClustering
SupplierClustering
Supplier Centrality 0.556∗∗∗
(< 0.001)
Merger Centrality 0.533∗∗∗ 0.393∗∗∗
(< 0.001) (< 0.001)
Customer Avg. Path −0.568∗∗∗ −0.316∗∗∗ −0.271∗∗∗
(< 0.001) (< 0.001) (< 0.001)
Supplier Avg. Path −0.447∗∗∗ −0.790∗∗∗ −0.163∗∗∗ 0.260∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001)
Merger Avg. Path −0.302∗∗∗ −0.225∗∗∗ −0.578∗∗∗ 0.241∗∗∗ 0.108∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (0.020)
Customer Clustering −0.225∗∗∗ −0.135∗∗∗ −0.129∗∗∗ −0.123∗∗∗ 0.104∗∗ 0.077∗
(< 0.001) (0.004) (0.005) (0.008) (0.026) (0.097)
Supplier Clustering −0.449∗∗∗ −0.528∗∗∗ −0.330∗∗∗ 0.342∗∗∗ 0.444∗∗∗ 0.234∗∗∗ 0.230∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Merger Clustering −0.207∗∗∗ −0.152∗∗∗ −0.293∗∗∗ 0.180∗∗∗ 0.137∗∗∗ −0.155∗∗∗ 0.008 0.164∗∗∗
(< 0.001) (0.001) (< 0.001) (< 0.001) (0.003) (< 0.001) (0.863) (< 0.001)
56Table VI
Exponential Random Graph Model to Explain the M&A NetworkThis table reports the coefficient estimates from exponential random graph models. The coefficient estimates are the marginal effect of the explanatory variable
on the conditional log-odds that two industries will have an additional inter-industry merger. The connections in the merger network are the dependentvariables, where the merger network is constructed as the number of inter-industry mergers between two industries using SDC merger data over 1986 to2010. The explanatory variables are the connections in the IO network constructed as in the text using data from the 1997 Input-Output tables from theU.S. Bureau of Economic Analysis. ‘Target Buys From Acquirer’ is the network where each connection is the percentage that the Target industry buys ofthe Acquirer industry’s output. The connections in ‘Target Sells to Acquirer’ are the percentage of inputs supplied by the Target industry to the Acquirerindustry. The coefficient on ‘Number of Connections’ is the marginal effect of an additional random connection on the conditional log-odds ratio of twoindustries having an additional merger in the merger network. | ∆ Variable | is the absolute difference between two industry nodes’ value of variable. AIC isthe Akaike’s Information Criterion. p−values are reported in parentheses. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗, for the 0.01, 0.05, and 0.10levels.
(1) (2) (3) (4) (5) (6) (7)
Number of Connections −3.519∗∗∗ −3.523∗∗∗ −3.545∗∗∗ −3.549∗∗∗ −3.596∗∗∗ −3.480∗∗∗ −3.468∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 6.564∗∗∗ 3.328∗∗∗ 5.239∗∗∗ 5.093∗∗∗
(<.001) (<.001) (<.001) (<.001)Acquirer Buys from Target 8.092∗∗∗ 4.680∗∗∗ 4.465∗∗∗ 3.763∗∗∗
(<.001) (<.001) (<.001) (<.001)Acquirer Sells to Target 16.208∗∗∗ 13.922∗∗∗ 21.301∗∗∗ 19.590∗∗∗
(<.001) (<.001) (<.001) (<.001)Target Sells to Acquirer 17.585∗∗∗ 14.790∗∗∗ 16.793∗∗∗ 18.423∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} 4.933∗∗∗ 4.871∗∗∗
(<.001) (<.001)| ∆ Industry M/B| −0.113∗∗ −0.115∗∗
(0.016) (0.015)| ∆ Industry Mean Returns| −0.697∗∗∗ −0.696∗∗∗
(<.001) (<.001)| ∆ Std Dev of Returns| −0.149 −0.149
(0.238) (0.240)| ∆ Concentration Ratio| 0.003∗∗ 0.003∗∗
(0.013) (0.014)T Buys from A×max{Industry R&D} 12.402
(0.557)A Buys from T×max{Industry R&D} 53.284∗∗
(0.024)A Sells to T×max{Industry R&D} 73.437∗∗
(0.020)T Sells to A×max{Industry R&D} −74.823∗∗∗
(0.008)Industry Economic Shock Index −0.241∗∗∗ −0.241∗∗∗
57
Table VI - Continued
(1) (2) (3) (4) (5) (6) (6)
(<.001) (<.001)Industry Median M/B −0.057∗ −0.057∗
(0.086) (0.086)Industry Median R&D 0.510 0.441
(0.538) (0.597)Industry Mean Return 0.253 0.253
(0.119) (0.119)Industry Std Dev of Returns 2.259∗∗∗ 2.259∗∗∗
(<.001) (<.001)Concentration Ratio −0.011∗∗∗ −0.012∗∗∗
(<.001) (<.001)Industry Size 0.000 0.000
(0.285) (0.276)Industry Scope 0.419 0.364
(0.855) (0.874)AIC 58,158 58,034 57,791 57,691 56,938 25,563 25,551Number of Industries 471 471 471 471 471 214 214
58Table VII
The Dynamic Impact of Mergers In Close IndustriesThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 fordetail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Thedependent variable is the log of one plus the number of industry pairs involving industry i that experience high merger activity in year t. Highmerger activity equals one if the log of the inflation-adjusted dollar volume between industries j and k in year t is greater or equal to the 75thpercentile of the industry-pair’s time series of dollar volumes from 1986 to 2010. ‘Closeness-Weighted M&A Activityt’ is
∑
j 6=i1
distij
∑
k 6=ivjkt,
where distij is the shortest path between industry i and j, and vjkt is an indicator variable for high merger activity in industry-pair-years, jkt, asdefined above. The shortest path is measured using either the network based on discrete customer links (columns 1–3) or supplier links (columns4-6), where links are defined as customer or supplier flows greater than 1%. ‘Own M&A Activityt−1’ is the lagged dependent variable at t − 1.‘Own Industry Fixed Effects’ are accounted for through first differencing. Lagged levels of the independent variables are used as instrumentsfor the endogenous first differences. Numbers in parantheses are p-values based on standard errors that are robust to general cross-section andtime-series heteroskedasticity and within-group autocorrelation. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6)
Closeness-Weighted M&A Activityt −0.024 −0.556∗∗∗
(0.665) (< 0.001)
Closeness-Weighted M&A Activityt−1 0.615∗∗∗ 0.680∗∗∗ 0.445∗∗∗ 1.230∗∗∗ 1.251∗∗∗ 1.011∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Closeness-Weighted M&A Activityt−2 −0.323∗∗∗ −0.259∗∗∗ −0.416∗∗∗ −0.442∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001)
Closeness-Weighted M&A Activityt−3 −0.051 −0.105 −0.080 −0.197∗
(0.425) (0.101) (0.406) (0.052)
Closeness-Weighted M&A Activityt−4 0.017 0.002 −0.091 −0.112(0.781) (0.974) (0.430) (0.296)
Own M&A Activityt−1 13.505∗∗∗ 20.035∗∗∗ 21.119∗∗∗ 14.930∗∗∗ 20.617∗∗∗ 23.589∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−2 9.066∗∗∗ 10.326∗∗∗ 7.890∗∗∗ 10.342∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−3 6.657∗∗∗ 7.283∗∗∗ 5.789∗∗∗ 7.423∗∗∗
(< 0.001) (< 0.001) (0.001) (< 0.001)
Own M&A Activityt−4 2.790∗ 3.122∗ 1.858 3.034∗
(0.090) (0.057) (0.260) (0.061)
59
Table VII - Continued
Dependent Variable: Industry Merger Activityt
Own Industry Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yesχ2 302.732 393.583 402.453 280.762 381.718 417.524p-value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 10,833 9,420 9,420 10,833 9,420 9,420
60Table VIII
The Diffusion of Merger Activity Across Close and Distant IndustriesThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 fordetail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Thedependent variable is the log of one plus the number of industry pairs involving industry i that experience high merger activity in year t. Highmerger activity equals one if the log of the inflation-adjusted dollar volume between industries j and k in year t is greater or equal to the 75thpercentile of the industry-pair’s time series of dollar volumes from 1986 to 2010. ‘Close M&A Activityt’ is aggregate merger activity in closeindustries, not including mergers with firms in industry i. Merger activity is defined by the number of industry-pairs that experience high-mergeractivity, as above. ‘Distant M&A Activityt’ is defined analogously. Close industries are defined as the industries that are directly connected tothe subject industry through a customer or supplier link above the one percent threshold. Distant industries are those that have the maximumshortest path from the subject industry. In the supplier network, this is three connections away. In the customer network, it is four connections.‘Own M&A Activityt−1’ is the lagged dependent variable at t − 1. ‘Own Industry Fixed Effects’ are accounted for through first differencing.Lagged levels of the independent variables are used as instruments for the endogenous first differences. Numbers in parentheses are p-valuesbased on standard errors that are robust to general cross-section and time-series heteroskedasticity and within-group autocorrelation. Statisticalsignificance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6) (7) (8)
Close M&A Activityt 0.432∗∗∗ 0.433∗∗∗
(<.0.001) (<.0.001)
Close M&A Activityt−1 0.470∗∗∗ 0.454∗∗∗ 0.111∗∗ 0.572∗∗∗ 0.588∗∗∗ 0.231∗∗∗
(<.0.001) (<.0.001) (0.046) (<.0.001) (<.0.001) (0.004)
Close M&A Activityt−2 −0.159∗∗∗ −0.153∗∗∗ −0.133∗∗ −0.228∗∗∗ −0.202∗∗∗ −0.202∗∗∗
(0.005) (0.008) (0.017) (<.0.001) (0.002) (0.002)
Close M&A Activityt−3 −0.072 −0.075 0.013 −0.057 −0.021 0.018(0.243) (0.237) (0.839) (0.328) (0.716) (0.761)
Close M&A Activityt−4 −0.076 −0.060 −0.019 −0.110∗ −0.088 −0.044(0.166) (0.288) (0.742) (0.072) (0.170) (0.509)
Distant M&A Activityt −0.010 −0.170∗∗∗
(0.816) (0.004)
Distant M&A Activityt−1 −0.236∗∗∗ −0.123∗∗ −0.015 −0.074 0.098∗ 0.118∗∗
(< 0.001) (0.032) (0.779) (0.201) (0.080) (0.021)
Distant M&A Activityt−2 0.081∗ 0.088∗∗ −0.014 0.061 0.068 −0.007(0.070) (0.048) (0.751) (0.257) (0.210) (0.889)
Distant M&A Activityt−3 0.040 −0.021 0.057 0.173∗∗∗ 0.096∗ 0.092∗
61
Table VIII - Continued
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6) (7) (8)
(0.477) (0.705) (0.315) (0.001) (0.059) (0.068)
Distant M&A Activityt−4 0.078 0.058 0.019 0.126∗∗ 0.086 0.006(0.239) (0.376) (0.770) (0.039) (0.164) (0.923)
Own M&A Activityt−1 18.808∗∗∗ 22.319∗∗∗ 18.803∗∗∗ 17.777∗∗∗ 18.635∗∗∗ 22.415∗∗∗ 18.257∗∗∗ 17.198∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−2 9.733∗∗∗ 11.758∗∗∗ 9.753∗∗∗ 9.230∗∗∗ 9.427∗∗∗ 11.947∗∗∗ 9.152∗∗∗ 8.641∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−3 6.955∗∗∗ 7.599∗∗∗ 6.966∗∗∗ 6.512∗∗∗ 6.878∗∗∗ 7.936∗∗∗ 6.768∗∗∗ 6.212∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−4 3.065∗ 2.938∗ 3.111∗ 2.635 2.837 3.370∗∗ 2.817 2.352(0.070) (0.066) (0.066) (0.125) (0.105) (0.036) (0.108) (0.178)
Own Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yesχ2 416.227 401.270 421.617 412.450 413.673 394.583 413.417 401.866p-value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 9,420 9,420 9,420 9,420 9,420 9,420 9,420 9,420
62
Table IXAggregate Merger Activity and CentralityThis table presents coefficients from time-series regressions from 1988 to 2010. The dependent variablein Panel A is ‘Centralityt,’ the average degree centrality of the industries that are experiencing amerger wave, where centrality is computed from the weighted and directed network of supplier linksbetween industries in the input-output network. ‘Industry Merger Waves (%)t’ is the percent of allindustries in year t that are experiencing a merger wave. A merger wave is when an industry hasmore mergers than the 75th percentile of mergers in a given year, relative to the industry’s timeseries of mergers from 1986 to 2010. Supplier networks are based on the 1997 BEA IO Report,using the detailed-level of industries. Merger firms are classified into industries based on eithertheir primary SIC or NAICS code, or using all industry codes reported in SDC, as indicated in thecolumn heading. Panel B present vector autoregressions with two endogenous variables: ‘Centralityt,’and ‘Merger Wavest,’ as defined above. p−value are reported in parentheses from standard errorscorrected for heteroskedasticity and autocorrelation using the procedure of Newey and West (1987)and the automatic lag selection of Newey and West (1994). Granger causality tests are reported of thenull hypothesis that centrality does not Granger cause market merger waves and that merger wavesdo not Granger cause average centrality. We report the χ2 and p−value for each test. Significance isindicated at 1%, 5%, and 10% levels by ∗∗∗, ∗∗, and ∗.
Firm-Level Industry Classification: Primary All
Panel A: Concurrent and Lagged Time-Series RegressionsDependent Variable: Average Degree Centrality of Wave Industryt
Industry Merger Waves (%)t 1.288∗∗∗ 1.160∗∗∗
(< 0.001) (< 0.001)
Industry Merger Waves (%)t−1 −0.256 −0.757∗∗∗
(0.262) (< 0.001)
Industry Merger Waves (%)t−2 0.205 0.429∗∗∗
(0.313) (< 0.001)
Constant 0.353∗∗∗ 0.699∗∗∗
(< 0.001) (< 0.001)
Adjusted R2 0.709 0.513Observations 23 23
63
Table IX - Continued
Firm-Level Industry Classification: Primary All
Panel B: Predictive Vector Autoregressions and Granger CausalityDependent Variable: Centralityt Wavest
Average Centralityt−1 0.244 0.262∗∗∗
(0.391) (0.009)
Average Centralityt−2 0.077 −0.033(0.783) (0.767)
Industry Merger Waves (%)t−1 1.231∗∗∗ 1.111∗∗∗
(0.009) (< 0.001)
Industry Merger Waves (%)t−2 −0.693 −0.627∗∗∗
(0.154) (< 0.001)
Constant 0.364∗∗ 0.015(0.031) (0.828)
Adjusted R2 0.544 0.851Observations 23 23
Granger CausalityH0: Centrality ; Wavesχ2 6.316∗∗
p−value (0.043)
H0: Waves ; Centralityχ2 7.958∗∗
p−value (0.019)
Internet Appendix for
“The Importance of Industry Links in Merger Waves”
Kenneth R. Ahern and Jarrad Harford∗
This Internet Appendix provides more details on the data used in the paper, a technical appendix
on exponential random graph models, and robustness tests discussed in the paper.
I. Data and Measures
This section provides additional details on the sources of data and how the data are constructed.
We first discuss the industry definitions of the input-output data we use to build our network. Then
we discuss the asset complementarity and concentration measures used in the paper.
A. Input-Output Relations
Since 1947, the Bureau of Economic Analysis (BEA) has provided Benchmark Input-Output (IO)
accounts for all producers and purchasers in the U.S. economy. Producers include all industrial
sectors as well as household production. Purchasers include industrial sectors, households, and
government entities. These tables are based primarily on data from the Economic Census and are
updated every five years with a five-year lag. As of January 2012, the most recent report available
is for 2002. In this paper, we use the reports in 1982, 1987, 1992, 1997, and 2002.
In each report, the BEA updates the set of industries used in the IO tables to reflect changes in
the economy. The industry definitions are designed to group firms into industries to best measure
customer and supplier relations. Prior to 1997, the IO industries were defined based on SIC codes.
The 1982 report is based on 1977 SIC codes, which are a minor revision of the 1972 SIC codes.
The 1987 and 1992 reports are based on 1987 SIC codes. In 1997, the BEA based the IO industries
on 1997 NAICS codes, following the policy of most U.S. government agencies to switch from SIC
to NAICS codes. The 2002 IO report is based on the revised 2002 NAICS classifications. The
∗Citation format: Kenneth R. Ahern and Jarrad Harford, YEAR: XXX, Internet Appendix to “The Importance ofIndustry Links in Merger Waves,” Journal of Finance [vol #XXX], [pages XXX], http://www.afajof.org/IA/[yearXXX].asp. Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting infor-mation supplied by the authors. Any queries (other than missing material) should be directed to the authors of thearticle.
2
policy to switch to NAICS codes was implemented to provide a consistent and updated industrial
classification system. The SIC system was first implemented in the 1930s and designed to classify
primarily manufacturing industries. The use of NAICS codes allowed for more service-oriented
industries to better reflect changes to the U.S. economy. In addition, NAICS industries are all
defined based on the economic concept of similar production processes, whereas some SIC codes
are based on production and others on demand. Concordance tables between the BEA-defined IO
industries and 4-digit SIC and 6-digit NAICS codes are reported with the tables in each year.1
For the industry classifications used in this paper, we follow the BEA’s IO industry classifications
with minor changes. In particular, we account for duplicates created when 4-digit SIC or 6-digit
NAICS codes are not as detailed as the IO codes. In some cases more than one IO industry code
is defined by an identical set of SIC or NAICS codes. For instance, in 1997, the BEA breaks down
the construction industry into different types of construction, such as manufacturing buildings,
commercial buildings, and sewer construction, though each of these industries maps to the same
NAICS codes. Since we can only identify merging firms at the NAICS level, merger activity in each
of the sub-industries of construction would be identical. To remove duplicates, we aggregate these
industries into one industry. Second, we only include industries with a concordance to an SIC or
NAICS code. This means we exclude government and household sectors.
In each report, the BEA defines industries at two levels of aggregation, detailed and summary.
Internet Appendix Table I reports the number of unique industries, not counting government or
household sectors, for each of the IO report years. For comparison, the number of SIC and NAICS
codes at different levels of aggregation are also presented.2 Internet Appendix Table I also presents
the number of unique industries in our data, after collapsing industries that are identical at the SIC
or NAICS level, as described above. The number of detailed industries we use ranges between 411
and 478. This is slightly more narrow than the 416 three-digit 1987 SIC codes, but substantially
more coarse than the 1,005 four-digit SIC codes. The number of our detailed industries is also closer
to the average number of four-digit NAICS in 1997 and 2002 (315), than to the average number
of five-digit NAICS codes (723). Again, our detailed industries are substantially more coarse than
1The 1982 files do not contain a concordance table, but concordance is available using the concordance table for thePersonal Consumption Expenditure tables from the U.S. Bureau of Labor Statistics.2The numbers of SIC and NAICS codes are taken from the concordance tables provided by the Census Bureau.
3
the 1,174 six-digit NAICS codes. The summary-level industries we use are similar to two-digit SIC
codes and three-digit NAICS codes. Thus, the coarseness of our industry definitions are roughly
equivalent to two and three-digit SIC codes, which have been used extensively in prior research.
To measure merger activity at the IO-code level, we must assign acquirer and target firms to
IO industry codes. This presents a number of challenges. First, we must map the SIC or NAICS
codes of a firm, as reported in SDC, to an IO industry. Since we run our tests using each of the
IO report industry definitions (1982, 1987, 1992, 1997, and 2002), we are forced to map across
two concordance tables: the concordance between different versions of SIC and NAICS codes, and
the concordance between SIC and NAICS codes and the IO industry codes. For instance, when
SDC reports a 1997 NAICS code, to map this to a 1987 IO industry, we must first convert the
1997 NAICS code to a 1987 SIC code, and then map the 1987 SIC code to the 1987 IO industry
code. In some cases, the mappings between different revisions of SIC and NAICS systems are not
one-to-one. The concordances between classification systems often either aggregate or separate
parts of industries in one system when mapping to the other. This sometimes increases or reduces
the number of SIC or NAICS industry codes in an IO industry. The 1982 IO report requires the
most distant mapping, as these industries are based on 1977 SIC codes. Thus, a firm’s 2002 NAICS
code must first be converted to a 1997 NAICS code, then to a 1987 SIC code, then to a 1977 SIC
code in order to assign the firm to a 1982 IO code. To increase the quality of industry matching,
we present analysis using the 1997 IO codes in our main tests, but run robustness checks using
alternative years.
Second, in a few cases, the BEA maps a single 4-digit SIC or 6-digit NAICS code to multiple IO
industry codes. To address this problem, we assign firms to only one of the matching IO industry
codes at random. This means that in the aggregate, each of the matching IO codes is equally
represented, and at the firm-level, a firm’s SIC or NAICS code is only matched to a single IO
industry code.
Finally, we address firms with multiple SIC or NAICS codes in a number of ways. The first
way is to use only the primary SIC or NAICS code, as reported in SDC. The second way is to
use all industry codes. In this case, we first map all of the SIC and NAICS codes to IO industry
codes. In most cases, a firm’s number of unique SIC or NAICS codes is mapped to a smaller
4
number of IO industry codes. We then assign equal weight to merger counts and dollar volumes
for each of these IO industry codes. For instance, if an acquirer is in industries 1 and 2 and a
target is in industry 3, we assign 0.5 merger counts to the industry pair (1,3) and 0.5 counts to
industry pair (2,3). As a third way to account for firms with multiple industry codes, we give
greater priority to horizontal mergers, followed by vertical mergers, and then unrelated mergers.
For each pair of merging firms, we first identify horizontal mergers as any overlaps in IO industry
codes. If there are any horizontal matches, we assign an equal fraction of the merger count or dollar
volume to the overlapping IO industry codes. If there are no horizontal matches, but a vertical
relation in any of the firms’ IO codes, we assign the deal activity equally to those IO industry
codes. Vertical relations are defined at two threshold levels. In particular, we record a vertical
relation if two industries exceed a threshold of 1 percent across any of the following four vertical
relations: 1) acquirer industry purchases from target, 2) target industry purchases from acquirer, 3)
acquirer industry sells to target, and 4) target industry sells to acquirer. We also create a secondary
mapping using a 5 percent threshold of vertical relations. If there are neither horizontal nor vertical
industry relations, we assign the deal equally across all of the unrelated industry pairs. We refer to
these industry assignments as HV1% and HV5%, to indicate the priority of horizontal and vertical
relations. Each of these different procedures is done identically at both the detail and summary IO
industry levels.
In sum, we consider a variety of different methods of mapping mergers to industry definitions.
In particular, we have distinct data for each of the five IO reporting years, the detail and summary
levels of industry aggregation, and four methods of mapping firms that operate in multiple industries
to the IO industry network. This provides 40 different combinations of industry mappings for
robustness.
B. Asset Complementarity
Our measure of asset complementarity is based on the text-based similarity measure developed
in Hoberg and Phillips (2010a) and Hoberg and Phillips (2010b) and provided on Jerry Hoberg’s
website. This text-based measure is constructed from yearly firm-by-firm pair-wise similarity scores
between the word vectors of the product descriptions from 10-K filings for all firms on Compustat
5
from 1996 to 2008. The dataset reports firm-pairs with similarities above a minimum threshold
of similarity. The threshold value is set to match the coarseness of three-digit SIC codes. To
map the firm-pairs to IO industry codes, we match the Compustat GVKEY identifiers of the firms
provided in the Hoberg and Phillips database to their historical SIC and NAICS codes, as reported
on Compustat. We match the historical SIC and NAICS codes to the IO industry codes, as detailed
above. Then, we record the total number of firms in the Hoberg and Phillips database that are in
a given IO industry-pair. Thus, our measure of asset complentarity between two IO industries is
the total number of firm-pairs in the IO industry pair that Hoberg and Phillips identify as similar
in a given year. We also record an analogous version where we record the sum of the book assets
of all firms in firm-pairs that match to a particular IO industry.
C. Industry Concentration
Industry concentration measures used in the paper are from the Economic Census of the U.S.
Census Bureau. Like the IO data, the Economic Census is conducted every five years, in years
ending in two and seven. The Economic Census reports the percentage of total industry sales
attributed to the 4, 8, 20, and 50 largest firms in an industry. With the exception of agriculture and
public administration, concentration measures are reported for all industries. For the manufacturing
sector alone, Herfindahl measures are reported as well. We use the 8-firm concentration ratio
because it has the least censoring at either 0% (more common for the 4-firm ratio) or 100% (more
common for the 20 and 50-firm ratios). Since these data cover firms of all sizes and the vast
majority of industries, they provide the most comprehensive concentration ratios available. In
contrast, concentration ratios calculated using Compustat sales are subject to both a severe size
bias and a public-listing bias.
Like the IO data, industries are classified by SIC codes before 1992 and by NAICS codes for
1997, 2002, and 2007. Therefore, we convert industry definitions using concordance tables between
each industry classification system in order to match the classification system employed by the IO
report. For instance, if we use the industry definitions provided by the 1982 IO report to build the
network, for the 2008 observation, we first convert industry definitions from the 2007 NAICS code
to 1987 SIC, then 1987 SIC to 1977 SIC codes. For each calendar year, we use the most recent
6
concentration ratios. For summary-level industry definitions, we do not have a corresponding
aggregation of concentration ratios. Therefore, we take the median concentration ratio across the
underlying detailed-level IO industry codes.
D. Industry Size and Scope
To account for industry size, we use data from the U.S. Census Bureau’s County Business
Patterns database. These data record the number of individual establishments at the 4-digit SIC
and 6-digit NAICS codes annually from 1986 to 2009. From the CBP webpage, an establishment
is a “single physical location at which business is conducted or services or industrial operations are
performed.” These data are based on the Business Register (BR), a database of all companies in the
U.S. collected by the U.S. Census Bureau. The BR is the most complete data available on businesses
in the U.S. Establishments are assigned to one industry code from data in the Economic Census or
other Census surveys, based on the primary activity of the establishment. The CBP data cover the
majority of industries in the U.S., including both manufacturing and service industries. It excludes
crop and animal production, rail transportation, pension funds, trusts, private households, and
public administration. Following the procedures for industry mappings detailed above, we aggregate
the SIC and NAICS-level CBP data to IO industry-levels to measure the size of economic activity
in each IO industry.
An alternative measure of industry size is the number of firms in an industry, rather than the
number of establishments. We choose to use the number of establishments for a few reasons. First,
data on the total number of firms across all industries is limited. Compustat data only includes
large, publicly-traded firms. Since our merger data is not limited to public firms, this would impose
a severe constraint. A number of alternatives to Compustat exist. The Statistics of U.S. Businesses
(SUSB), the Nonemployer Statistics (NE) program, and the Business Employment Dynamics (BED)
program each have data that aggregate establishments to firms. However, their sample periods or
types of firms covered are typically limited. The SUSB data with detailed industries only goes back
to 1998. The NE data only cover the smallest firms, and the BED program only reports major
industry sectors. Therefore, there is no publicly-available data on number of firms from 1986 to
2010 that cover all firms in all industries at a detailed classification level. Second, an advantage
7
of establishment-level data is that industry classifications are more precise than at the firm level.
Compared to firms, establishments are more likely to engage in activities that fall primarily in one
industry classification. This means our measure of firm size is appropriate for acquisitions and
divestitures of divisions within conglomerate firms, as well as focused firms. In the cross-sectional
tests, we use the average number of establishments over the entire sample period.
To account for industries that have more variation in the types of goods and services they
produce, we create a measure of industry scope. We record the percent of all NAICS or SIC codes
(depending on IO report year) that map to a particular IO industry. This variable is defined at
both the detailed and summary level. Underlying this measure’s usefulness is the assumption that
SIC and NAICS codes are defined to be relatively equal in scope. Government documents confirm
that this is the case.
According to a report by the Economic Classification Policy Committee (1993), two criteria
were used to determine the boundaries of SIC codes, including when new codes should be created
and obsolete codes should be combined with other codes. First, the economic significance of the
industry was historically defined as the size of a 4-digit SIC industry relative to its SIC Division (e.g.,
Manufacturing), based on five size variables, such as employees, payroll, and value added. Industries
that were too small relative to the Division size were combined with other industries. The second
criterion was based on specialization and coverage ratios that are designed to measure homogeneity
of economic activity within an SIC code. In particular, they were measures of concentration of the
output of the primary product for each industry. Thus, the goal was to keep industries size and
heterogeneity within limits. As such, the SIC codes were revised many times to reflect changes in
the size and specificity of industries.
When NAICS codes were developed, these issues were discussed and revised, as the NAICS codes
were designed to be based solely on production functions, not outputs, as were the SIC codes. In
particular, the boundaries of NAICS codes were designed to contain production processes that were
within a heterogeneity threshold. Heterogeneity is defined as a weighted average of differences in
production parameters describing technologies employed by firms in an industry (e.g., the output
elasticity on a particular factor input in a Cobb-Douglas production function). A more detailed
8
description is presented in Gollop (1994). Thus, though these measures are likely imperfect, SIC
and NAICS codes are defined to relatively equal in scope, which validates our measure of scope.
II. Network Statistics
In this section of the appendix, we provide more detailed examples of various network statistics,
focusing on those used in the paper. This section is only designed to give more detail to the concepts
we use in the main paper and does not try to provide a complete understanding of graph theory.
The interested reader should see Wasserman and Faust (1994) for a book-length study of formal
network analysis.
To illustrate these concepts, we will use the simplified example of the timber industry’s input-
output network presented in the main paper. In Internet Appendix Figures I and II, we present
the adjacency matrix and network figure of the timber industry from Figure 1 of the main paper,
showing only the supplier network:
In the terminology of graph theory, each industry is a node in the network, connected to other
nodes through edges, or ties. The entries in this matrix represent the percentage of a customer
industry’s (on the rows) total inputs purchased from the supplier industries (on the columns). For
instance, the Wood Doors industry purchased 18% of its total inputs from the Sawmill industry.
This means a connection is a supply relationship. Notice that this matrix is not symmetric and
entries take positive values between 0 and 100. In the context of a graph, this means the connections
between the edges are directional and weighted. For simplicity, graphs may be binary, where entries
only take the value of zero or one, to indicate the existence of a connection between two nodes.
A large literature has developed many methods to characterize and measure various attributes
of networks. We discuss three common categories of network statistics: centrality, closeness, and
clustering.
9
0 0 0 0 0 0
64 1 1 0 0 0
7 29 41 0 0 0
0 11 50 17 0 0
0 0 24 14 1 0
0 0 0 18 0 1
Forestry Support
Forestry Nurseries
Logging
Sawmills
Pulp Mills
Wood Doors
ForestrySupport
ForestryNurseries
Logging
Saw
mills
Pulp
Mills
WoodDoors
Suppliers
Custom
ers
Internet Appendix Figure IAdjacency Matrix Example: Supply Relations in the Timber Sector
ForestrySupport
ForestNurseries
Logging
Sawmills
PulpMills
WoodDoors
64%
7%
29%
11%
50%
24%
14%
18%
Internet Appendix Figure IINetwork Graph Example: Supply Relations in the Timber Sector
10
A. Centrality
The network concept of centrality is designed to capture the relative importance of a node or an
edge in a graph. It can be measured in various ways.
A.1. Degree Centrality and Strength
The simplest centrality measure is degree centrality. Degree centrality is the sum of binary
connections. In weighted graphs, its counterpart measure is the sum of weights, known as strength.
Each of these measures can be further divided into indegree and outdegree centrality, based on the
direction of the connections. Indegree (instrength) is the sum of connections (weights) directed
towards the node. Outdegree (outstrength) is the sum of connections (weights) originating from
the node.
In the timber example, the outstrength represents the importance of an industry as a supplier.
This is the sum over the rows. This produces an outstrength of 71 for Forestry Support, 41 for
Forest Nurseries, 116 for Logging, 49 for Sawmills, 1 for Pulp Mills, and 1 for Wood Doors. The
instrength is the sum over the columns. In the example, this produces an instrength of 0 for
Forestry Support, 66 for Forest Nurseries, 77 for Logging, 78 for Sawmills, 39 for Pulp Mills, and
19 for Wood Doors. Since these are fractions of total inputs, if the complete network was presented,
each of these would equal 100, so the instrength in our setting is not useful.
A.2. Eigenvector Centrality
The second measure of centrality we use in the paper is called eigenvector centrality (Bonacich,
1972). This measure considers a node to be more central if it is connected to other nodes that are
themselves more central. This is formalized as follows. If we define the eigenvector centrality of
node i as ci, then ci is proportional to the sum of the cj ’s for all other nodes j 6= i:
ci =1
λ
∑
j∈M(i)
cj =1
λ
N∑
j=1
Aijcj (IA.1)
11
where M(i) is the set of nodes that are connected to node i and λ is a constant. In matrix notation,
this is
Ac = λc (IA.2)
Thus, c is the principal eigenvector of the adjacency matrix.
Eigenvector centrality cannot always be applied to asymmetric adjacency matrices (Bonacich and
Lloyd, 2001). For simplicity, we make the adjacency matrix symmetric by taking the maximum
value of the upper and lower triangles, as presented in Internet Appendix Figure III. Taking the
eigenvector corresponding to the largest eigenvalue of this symmetric matrix as the eigenvector
centrality, we find centrality measures of 0.30 for Forestry Support, 0.42 for Forest Nurseries, 0.67
for Logging, 0.48 for Sawmills, 0.22 for Pulp Mills, and 0.08 for Wood Doors. Thus, Logging has
both the greatest eigenvector centrality and strength in this network. Wood Doors has the lowest
eigenvector centrality. These measures quantify what is obvious from inspection of the network in
Internet Appendix Figure II.
0 64 7 0 0 0
64 1 29 11 0 0
7 29 41 50 24 0
0 11 50 17 14 18
0 0 24 14 1 0
0 0 0 18 0 1
Internet Appendix Figure IIISymmetric Adjacency Matrix Example: Supply Relations in the Timber Sector
12
A.3. Betweenness Centrality
Betweenness centrality is designed to capture the network-wide importance of an industry as a
conduit or gatekeeper in the network. In this sense, betweenness centrality is closer to eigenvector
centrality than degree centrality. Betweenness centrality is defined for node k as the percentage
of shortest paths between two nodes i and j that include node k (Freeman, 1979). The shortest
path is also known as geodesic distance. For instance, ignoring the weights, but accounting for the
directionality in Internet Appendix Figure I, there are two shortest paths between Forestry Support
and Wood Doors. One path goes through Forest Nurseries, then Sawmills, then to Wood Doors.
The second goes through Logging, then Sawmills, then Wood Doors. Each of these paths has a
length of three. For calculating betweenness centrality, if a node k is on a shortest path between
two other nodes, but there are other paths of the same length that do not include node k, then
node k receives a fraction of a shortest path. For example, the Logging industry is a node on the
single shortest path from Forestry Support to Pulp Mills, but also on one of two shortest paths
between Forestry Support and Pulp Mills and one of two shortest paths between Forest Nurseries
and Pulp Mills. Logging’s between centrality is then two divided by twenty possible shortest paths
or 0.10.
A similar measure is edge betweenness centrality. This measure is calculating as betweenness
centrality, but for edges, rather than nodes. This quantifies how important a particular connection
is to the network flow. An obvious example is the importance of the Bay Bridge connecting San
Francisco and Oakland, which is an edge on many shortest paths between different locations in the
Bay Area.
A.4. Which Centrality Measure is Appropriate?
A criticism of social network research is that though it provides many statistical measures of
connections, the underlying assumptions that allow proper usage and interpretation are lacking.
Borgatti (2005) tackles this issue directly for common centrality measures. He differentiates possible
flow processes in networks into two dimensions: the mechanism of node-to-node transmission and
the trajectory of the flow. In particular, the mechanism of transmission includes parallel duplication,
where the flow can be replicated and passed to multiple nodes simultaneously (such as email), serial
13
duplication, where the flow can only occur one at a time (viral infection), or where the flow is not
reproduced but must itself be passed along (package delivery). The trajectory of the flow can be
defined as flows that follow the shortest path, a sequence of nodes that does not repeat any edge,
a sequence that does not repeat any node, or an unrestricted sequence.
Borgatti (2005) shows that betweenness centrality assumes that the traffic flow in a network is
not reproduced at each node and it always takes the shortest path, rather than spreading randomly.
Thus betweenness centrality probably is inadequate to describe how information flows, since in-
formation is copied, not transferred, or how infections spread, since they do not have a particular
target. Instead, it is more suited for delivery of a package. In contrast, eigenvector centrality as-
sumes traffic can flow randomly, rather than along a shortest path. In addition, it assumes the flow
can be duplicated and spread in parallel. This makes it consistent with a flow process of influence or
information. Finally, in contrast to the global property of eigenvector centrality, degree centrality
can be considered as a measure of local importance that allows for parallel duplication of traffic.
Our paper is concerned with economic shocks that transmit across industries. These shocks affect
the demand and cost functions of firms. Therefore, a shock to one industry can affect multiple
customer and supplier industries simultaneously, and the shocks can change as they transfer, much
like information passed from one person to another. Therefore, degree and eigenvector centrality
are best suited for measuring the centrality of industries in the input-output network.
B. Closeness
A related concept to centrality is average shortest path. This measure is designed to capture
how close a node is to the other nodes in the network, on average. In a binary network, this is
simple to compute. One simply counts the shortest distance from one node to all other nodes,
where distance is measured through a path of connected nodes. The average shortest path is the
average over these path lengths. For instance, the Forestry Support industry has a path length of
one to the Forest Nurseries and Logging industries, two to the Sawmills and Pulp Mills industries,
and three to the Wood Doors industry. Including a path length of zero to itself gives an average
shortest path of 1.5. The average shortest path in this binary network is 1.167 for Forest Nurseries,
1.0 for Logging and Sawmills, 1.333 for Pulp Mills, and 1.667 for Wood Doors. Thus, Logging and
14
Sawmills are the closest industries to an average industry, and Wood Doors is the furthest, again
confirming what observation would suggest.
In complex networks with hundreds of nodes, visual inspection is inadequate. In addition, finding
the shortest path between all nodes and accounting for the strength of the connection between ties
is computationally challenging. To accomplish these goals, we use Dijkstra’s (1959) algorithm. To
account for weighted connections, we take the inverse of the strength of a connection as a measure
of distance, following Newman (2002). In this case, we find that Logging has the shortest weighted
path (0.037), and Wood Doors has the longest (0.081).
Network-wide statistics of path lengths give a summary measure of how connected a network is
overall. In particular, diameter is the maximum path across all shortest paths, or Max(Shortest
Distance). For the binary network example, this is three, the path length from Forestry Support to
Wood Doors. Similar to this is network average shortest path, which is simply the average over the
entire list of shortest paths between all node-pairs. In the binary network example, this is 1.278.
These measures quantify the density of a network. Highly connected industries with low diameters
and small average path lengths indicate that transmission between nodes can occur quickly.
C. Clustering Coefficient
Clustering refers to how embedded a node is in the network, or in our case, how embedded
an industry is in the economy. More formally, we calculate the clustering coefficient of Watts
and Strogatz (1998). Defining a node’s neighborhood as the set of nodes to which a particular
node is connected, the clustering coefficient is the proportion of observed connections between the
nodes in its neighborhood to the total possible connections. Intuitively, the greater is the clustering
coefficient of an industry in the customer-supplier network, the more its customers and/or suppliers
also trade with each other. In contrast, the trading partners of industries with low clustering
coefficients, trade little with each other. This measure is designed to capture how closely connected
is the network at a local level. It is easy to see in the binary example, that the clustering coefficient
for Forestry Support is 1, since it is connected to two industries, Forest Nurseries and Logging, and
these two industries are connected. The clustering coefficient for Forest Nurseries is 0.667, since
it is connected to three industries, of which two are connected. The coefficient is 0.5 for Logging,
15
0.333 for Sawmills, 1 for Pulp Mills and 0 or undefined for Wood Doors. Thus, while Logging is the
most central industry in this example, it is in a local group of industries that are less interconnected
than are other industries that are less central.
III. Exponential Random Graph Models
Exponential random graph models (ERGM) were developed to understand the determinants of
stochastic networks of social interactions (Holland and Leinhardt, 1981; Frank and Strauss, 1986;
Wasserman and Pattison, 1996). They assume that ties between agents in the network are random
variables, that these random variables may be dependent upon other random ties, and that the
ties between agents occur with a probability based on an exponential function of network statistics.
This section of the appendix is designed to provide a more detailed understanding of ERGM and
its implementation than is presented in the main paper. It is not intended to replace the large
literature on ERGM, upon which we base this appendix. In particular, this section of the appendix
borrows heavily from a number of prior papers, including Robins, Pattison, Kalish, and Lusher
(2007), Robins and Morris (2007), Hunter, Handcock, Butts, Goodreau, and Morris (2008), Morris,
Handcock, and Hunter (2008), and Johannes and Polson (2009).
A. Theoretical Framework
ERGM are estimated using a maximum-likelihood estimator. Given a fixed set of N nodes, if we
let G denote a random graph on these nodes (i.e., a random set of connections), and let g denote
the observed graph on the N nodes, then the distribution of G is,
Pθ(G = g) =exp{θ′s(g)}
c(θ,G)(IA.3)
16
where
θ ≡ An unknown vector of parameters (IA.4)
s(g) ≡ A known vector of network statistics on g (IA.5)
c(θ,G) ≡ A normalizing constant =∑
all h ∈ G
exp{θ′s(h)} (IA.6)
The constant c is included to ensure that the probabilities are bound between zero and one. Similar
to the maximum-likelihood estimator of a limited-dependent variable model, an ERGM estimates
the unknown parameters θ (the coefficients on the s(g)) by finding the particular θ that provides
the closest network to the observed network g by maximizing the log-likelihood function. The s(g)
include both node (industry)-specific variables as well as variables that describe the edges that
connect the nodes. In the context of industries, this means the s(g) explanatory variables can
include both industry-level attributes such as median market-to-book ratio, number of firms, and
average returns, as well as industry-pair variables between two industries, such as input-output
relations, differences in market-to-book ratios, and the similarity of product offerings.
The key difference between ERGM and a common limited-dependent variable model, such as a
logistic regression, is that the objective function of the maximization problem in a common logistic
regression is a single outcome variable, whereas in ERGM, it is an entire network. To illustrate
this idea, assume that we could only estimate the relations between industry-level characteristics
and industry-level outcomes. For instance, we could regress the number of mergers involving firms
in a given industry on the number of firms in the same industry. In this relationship, we could not
determine cross-industry merger activity. Now consider a regression of the number of cross-industry
mergers between two industries on a set of variables that includes the size of each industry in the
pair and the absolute value of the difference in size between the industries. The unit of observation
in this case is an industry-pair. This specification accounts for cross-industry mergers and a limited
set of determinants. However, this specification does not account for industry relations at a higher
order of separation than the industry-pair. To do this, one must use the entire network as an
outcome variable, as in ERGM, and allow for dependency between nodes and connections.
17
Estimating all possible random graphs is computationally challenging, which makes calculating
the constant c in Equation IA.3 infeasible for all but very small networks. Following standard
methods, we use Markov chain Monte Carlo (MCMC) methods to overcome this computational
complexity, as developed in Geyer and Thompson (1992). Taking the log of Equation IA.3, produces
the loglikelihood function:
ℓ(θ) = θ′s(g) − log c(θ,G) (IA.7)
Now, subtract a function with the same form, but arbitrary θ0 to create the log-ratio of likelihood
values:
ℓ(θ)− ℓ(θ0) = (θ − θ0)′s(g)− log
[
c(θ,G)
c(θ0, G)
]
(IA.8)
Geyer and Thompson (1992) show that the Law of Large Numbers provides the following approxi-
mation:
ℓ(θ)− ℓ(θ0) ≈ (θ − θ0)′s(g)− log
[
1
m
m∑
i=1
exp{
(θ − θ0)′s(Gi)
}
]
(IA.9)
where G1, . . . , Gm is a random sample of graphs from the distribution Pθ,G.
This approximated likelihood function depends upon the observed graph, g, a series of random
graphs, G1, . . . , Gm, and the arbitrary vector of parameters, θ0. For large enough sample size m,
the exact specification of θ0 is irrelevant. However, convergence may be slow if θ0 is not chosen to
be close enough to the true parameter values, θ. Therefore, θ0 is chosen using a procedure called
pseudolikelihood (Beseg, 1974; Strauss and Ikeda, 1990). Pseudolikelihood estimates a standard
logistic regression for each node-pair, holding all other node-pairs fixed. In other words, this
estimate assumes connections between nodes are independent. This assumption means that the
parameters estimates are likely biased, but they provide a good starting point for θ0.
To draw m random graphs from the distribution Pθ,G, we use MCMC methods. The general
idea behind MCMC methods is to create a Markov chain on the set of nodes in G, where the
equilibrium distribution equals Pθ,G. This is implemented using the Metropolis-Hastings algorithm
which provides an iterative procedure to take random variables generated from a known distribution
18
and transform them into random variables generated from the desired distribution, Pθ,G. Once
the equilibrium distribution is reached through the iterative procedure, random draws can be
computed for them observations of G necessary to maximize the approximate loglikelihood function
in Equation IA.9.
B. Practical Implementation
We estimate the exponential random graph models using the Statnet suite of programming
packages (Handcock, Hunter, Butts, Goodreau, and Morris, 2003), which is available through
http://statnetproject.org and utilized in the R programming language. Statnet uses a Markov-
Chain Monte Carlo algorithm to compute various statistical estimates and tests related to social
networks. The software performs Monte Carlo simulations of networks to approximate a maximum
likelihood estimator of coefficients (i.e. for the log-odds of a tie occurring) in a network model
as well as computes measures of centrality. Detailed documentation of Statnet can be found at
statnetproject.org.
Standard ERGM implementation employs binary edges (two nodes are either connected or they
are not). We want to model the strength of the connection between two industries, recognizing that
many cross-industry mergers are more meaningful than one cross-industry merger. To do this, we
treat each additional merger as an additional edge, so that an industry pair with 10 cross-industry
mergers will have 10 merger edges between them. In addition to providing a way to model the
strength of the connection, this approach allows us to interpret the coefficient in the ERGM output
as the increase in the log-odds of one additional merger between the two industries for a unit
increase in the explanatory variable.
C. Goodness of Fit
As with any statistical model, it is important to measure the similarity between the predicted
network and the actual network. Based on theory in Brown (1986), Hunter, Handcock, Butts,
Goodreau, and Morris (2008) shows that the expected value of the network statistics vector s(g)
is equal to the observed statistics when the parameters θ are set equal to the maximum likelihood
estimates. This means that at a minimum, the probability mass of an ERGM is the same as the
19
observed network. However, we would like to quantify how similar an ERGM is to an observed
network in other ways, beyond the mean. In particular, degeneracy of the model, where probability
mass is placed on either a network that has no connections or a network where every node is
connected to every other node, is a potential problem. The expected value could equal the observed
expected values of network statistics, but clearly these degenerate networks are unlikely to be
realistic.
In ordinary least squares regressions, an R2 statistic indicates the fraction of total variance
around the mean that is explained by the estimated model. In maximum likelihood estimates,
no equivalent statistic is available. Instead, the Goodness of Fit (GoF) of maximum likelihood
estimators is typically measured using a comparison of the log-likelihoods from the empirical model
under consideration to an alternative model, in the form of a likelihood-ratio test. The choice of
the alternative model is left to the researcher. In practice, it is often a reduced model that only
includes an intercept term.
Similarly, the Akaike Information Criterion (AIC) is designed to provide a measure of the GoF
for MLEs. AIC is the difference between the number of parameters in the model and the maximized
log likelihood function. As in all MLEs, a particular value of an AIC has no clear benchmark, but
rather is useful in comparison to other values of AIC, with lower values indicating the preferred
model. Following standard practice, we report AIC values for the ERGM results in the main paper.
These estimates can be compared across model specifications to quantify comparisons in model fit.
Because no clear benchmarks are available for the GoF measures for MLEs, there is a concern that
even the most preferred model, based on a likelihood ratio test or AIC, does not produce estimates
that are similar to the true distribution. This is true for common MLEs, such as a logistic regression
model, and also for ERGM. However, the richness of the network-based approach of ERGM allows
additional GoF tests.
Hunter, Goodreau, and Handcock (2008) develops new GoF tests for ERGMs that compare prop-
erties of estimated networks to corresponding global network properties. First, using the estimated
parameters in the fitted model, a large number of networks are simulated. Then, network statistics
of the observed network are compared to statistics of the networks simulated from the fitted values.
The statistics used in the comparison are statistics that were not included as explanatory variables
20
in the empirical model. The choice of statistics provides various measures of GoF, across multiple
dimensions. Hunter, Goodreau, and Handcock (2008) study GoF using three sets of statistics:
degree distribution, edgewise shared partner distribution, and the geodesic distance distribution.
The degree distribution is the proportion of industries with k direct connections. (In the main
paper, we present the degree distributions for industry- and merger-based networks in Figure 3.)
The edgewise shared partner distribution is the proportion of edges whose endpoints both share
edges with exactly k other nodes. The geodesic distance distribution is the proportion of geodesic
distances in the network equal to a distance of k. Hunter, Goodreau, and Handcock (2008) argue
that each of these statistics have been shown to contain important information about the global
properties of a network, but are not typically directly related to empirical models. Thus, they
argue that this provides independent measures of goodness of fit across multiple dimensions. In
essence, this procedure is akin to comparing predicted to actual out-of-model characteristics, using
structural parameters estimated from the data.
As an example, in Internet Appendix Figure IV, we present the in-degree goodness of fit measures
for our ERGM coefficient estimates from Table VI of the main paper, using the 1997 detail-level
industry definitions. In this setting, an industry’s in-degree is the number of industry firms that are
targets in an acquisition. The in-degree distribution describes the frequency of industries that have
a particular in-degree. The solid line in the figure represents the observed in-degree distribution
from the data. There are many industries with a few target firms and relatively fewer industries
with many target firms. Thus, the distribution is right-skewed. Next to the observed distribution
are two simulated distributions. First, the dotted line uses the coefficient estimates from the
parsimonious model, which only includes the IO variables, to simulate random networks. Then
the degree distribution is calculated using the average in-degree over the simulated networks. The
dashed line is computed analogously, but uses the full specification with all explanatory variables.
Internet Appendix Figure IV shows that the full specification produces an in-degree distribution
that is more similar to the true distribution than the specification with IO variables only. In
particular, though the simulations that only use the IO variables produce networks that are right
skewed, the additional variables in the full specification generate simulations that are even more
right skewed, with fatter tails, more closely matching the true distribution. This figure shows that
21
our models do a relatively good job of describing an out-of-model network characteristic, though
the empirical models cannot match the high sparsity of the true data.
In Internet Appendix Figure V, we present the geodesic distance distribution from the true
data and from simulations using the estimated coefficients. As before, the empirical models do a
relatively good job of matching the network structure, but do not generate networks that are as
sparse as are the true data. In particular, in the true data there are many industry pairs in the
network that cannot be connected through inter-industry merger activity (the ∞ category in the
figure). The empirical estimates create networks that are more dense on average than the true
network.
These goodness-of-fit tests provide evidence that our ERGM estimates are reliable. If the simu-
lations had produced networks with no connections or with all nodes connected to all other nodes,
then our estimates would be spurious. In other words, the GoF tests show that our estimated
networks are not degenerate, but instead have similarities with the true network. Though the
simulations are not a perfect match to the true network, that should not be expected. Compared
to typical R2s of ten percent or less in cross-sectional OLS regressions of announcement returns, or
the notoriously low R2 in common asset pricing models (Roll, 1988), the GoF tests show that our
estimates generate networks with out-of-model network characteristics that are relatively similar
to the true data, though they cannot reproduce the sparsity of the true merger network.
22
Frequency
ofIn-D
egree
In-Degree
Observed M&A Network
Predicted (All Variables)
Predicted (Input-Output Variables)
0 10 20 30 40 50 60 70 80
0.02
0.04
0.06
0.08
0.10
Internet Appendix Figure IVGoodness-of-Fit of In-Degree DistributionThis figure presents the in-degree distribution of the observed M&A network using the 1997 BEA detailedindustry definitions, compared to the average over simulated random networks where networks are simulatedusing the estimated coefficients from the exponential random graph models with either only the four input-output network flows as explanatory variables, or with all explanatory variables, as reported in Table VI ofthe main paper. In-degree is the number of times an industry firm is a target in an acquisition.
23
Frequency
ofGeodesic
Distance
Geodesic Distance
Observed M&A Network
Predicted(All Variables)
Predicted(Input-Output Variables)
0.1
0.2
0.3
0.4
0.5
1 2 3 4 5 6 7 8 9 ∞
Internet Appendix Figure VGoodness-of-Fit of Geodesic Distance DistributionThis figure presents the geodesic distance distribution of the observed M&A network using the 1997 BEAdetailed industry definitions, compared to the average over simulated random networks where networks aresimulated using the estimated coefficients from the exponential random graph models with either only thefour input-output network flows as explanatory variables, or with all explanatory variables, as reported inTable VI of the main paper. Geodesic distance is the length of the shortest path joining two nodes (or infinityif no path exists).
24
IV. Robustness Tests and Additional Tables
This section presents figures and tables referred to in the paper, but that were not included for
the sake of brevity.
25
bc bc bc bc bc bc bc bc bc bc bc bcbc bc bc bc bcbc bc bc bc bc bcbc bc bc bc bc bc bc bc bc bc bc bc bc bc
bcbc
bcbc
bc
1 10 100 10000.1%
1%
10%
100%
Prob(x
≥X)
# of Inter-Industry Connections, X
(a) Summary-Level Supplier Network
bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bc bcbc bc bc bcbc bc bcbcbcbcbc
bc
1 10 100 10000.1%
1%
10%
100%
Prob(x
≥X)
# of Inter-Industry Connections, X
(b) Summary-Level Merger Network
Internet Appendix Figure VIDegree Distribution of Merger and IO Networks using Summary-Level Industry DefinitionsThese figures represents the distribution of degree centrality in log-log scale. Circles represent the degree centrality of industries,indicating how many direct connections an industry has to other industries. The dashed line is the from the estimate of the alphaterm in the power distribution P (k) = ck−α. There are 124 summary-level industries using the 1997 IO Tables produced by theU.S. Bureau of Economic Analysis. Supplier network connections occur if an industry supplies more than 1% of the total inputs of acustomer industry. Merger network connections occur if there exist any cross-industry mergers. The merger data are over 1986–2010from SDC.
26
Internet Appendix Table INumber of Unique Industries by Industrial Classification System and YearThis table presents the number of industries in the ‘Make’ and ‘Use’ tables in the 1982, 1987, 1992, 1997, and 2002 reports ofthe Bureau of Economic Analysis (BEA IO), not counting government or household sectors. Two levels of industry definitions areprovided: Detailed and Summary. The ‘Paper’ column reports the number of industries used in the main paper for each year andindustry coarseness. These are reduced from the BEA IO numbers because multiple BEA IO industries share the same SIC or NAICScode. The ‘SIC Codes’ and ‘NAICS Codes’ columns report the number of unique SIC or NAICS industries that are mapped into BEAIO industries, according to the concordance tables provided by the BEA. See the Data section of the Internet Appendix for moredetails.
BEA IO Paper SIC Codes NAICS Codes
Year Detailed Summary Detailed Summary 4-digit 3-digit 2-digit 6-digit 5-digit 4-digit 3-digit
1982 528 78 478 77 939 389 771987 470 88 465 87 1, 005 416 831992 486 89 475 88 1, 005 416 831997 483 127 471 124 1, 169 721 313 962002 417 128 411 126 1, 179 725 317 100
27
Internet Appendix Table IIInput-Output Summary Statistics by IO Report Year and Industry CoarsenessThis table presents summary statistics of the Input-Output relationships of industries as definedby the Bureau of Economic Analysis Input-Output (IO) Industry classifications. The IO industrydefinitions are based on input-output relations between industries as recorded by the BEA in fiveseparate reports (1982, 1987, 1992, 1997, and 2002), for two levels of industry definitions (Detailed andSummary). Panel A presents the detailed-level data and panel B presents the summary-level data.Inter-industry pairs include all combinations of the industries (excluding own-industry pairs). Inter-industry pairs > 1% are only those observations where either Customer % or Supplier % is greaterthan 1%. Intra-industry observations include relations of firms that are in the same IO industry.Customer % is the percentage of industry i’s sales that are purchased by industry j. Supplier %is the percentage of industry i’s inputs that are purchased from industry j. All numbers, are inpercentages.
Customer % Supplier %
Inter-IndustryPairs
Inter-IndustryPairs> 1%
Intra-Industry
Inter-IndustryPairs
Inter-IndustryPairs> 1%
Intra-Industry
Panel A: Detail-Level
IO Report Year: 1982
Mean 0.220 5.800 3.680 0.250 3.970 4.140Median 0.000 2.460 1.390 0.000 2.070 1.5605th percentile 0.000 1.070 0.000 0.000 1.060 0.00095th percentile 0.530 22.510 14.790 0.930 12.280 16.030
Frequency Percentage0–1% 96.836 — 44.979 95.285 — 43.9331–2% 1.307 41.309 13.598 2.263 48.000 10.8792–3% 0.522 16.496 6.276 0.804 17.042 7.5313–4% 0.319 10.092 5.858 0.403 8.540 6.4854–5% 0.196 6.182 5.858 0.282 5.991 4.603> 5% 0.820 25.922 23.431 0.963 20.428 26.569
IO Report Year: 1987
Mean 0.230 6.460 3.320 0.250 3.990 4.180Median 0.000 2.370 1.150 0.000 2.080 1.4305th percentile 0.000 1.080 0.000 0.000 1.060 0.00095th percentile 0.480 24.910 12.870 0.950 12.750 16.920
Frequency Percentage0–1% 97.028 — 46.882 95.226 — 44.3011–2% 1.259 42.358 12.688 2.291 47.981 10.3232–3% 0.525 17.654 7.527 0.820 17.185 7.3123–4% 0.279 9.389 6.667 0.398 8.330 6.4524–5% 0.164 5.521 3.226 0.290 6.078 5.161> 5% 0.745 25.078 23.011 0.975 20.427 26.452
IO Report Year: 1992
Mean 0.220 5.980 3.690 0.250 3.980 4.500Median 0.000 2.360 1.600 0.000 2.070 2.010
28
Internet Appendix Table II - Continued
Customer % Supplier %
Inter-IndustryPairs
Inter-IndustryPairs> 1%
Intra-Industry
Inter-IndustryPairs
Inter-IndustryPairs> 1%
Intra-Industry
5th percentile 0.000 1.070 0.000 0.000 1.060 0.00095th percentile 0.500 25.640 15.960 0.910 12.650 18.820
Frequency Percentage0–1% 96.899 — 40.000 95.426 — 36.8421–2% 1.303 42.022 15.158 2.191 47.893 12.8422–3% 0.560 18.046 9.263 0.768 16.780 9.8953–4% 0.274 8.823 6.947 0.418 9.128 6.3164–5% 0.161 5.185 5.895 0.287 6.273 6.105> 5% 0.804 25.924 22.737 0.911 19.926 28.000
IO Report Year: 1997
Mean 0.220 5.060 3.310 0.270 3.860 4.110Median 0.010 2.190 1.140 0.010 2.200 1.4005th percentile 0.000 1.060 0.000 0.000 1.050 0.00095th percentile 0.620 18.270 12.470 0.980 10.870 16.010
Frequency Percentage0–1% 96.568 — 47.346 95.131 — 42.6751–2% 1.567 45.644 12.527 2.231 45.816 14.4372–3% 0.615 17.926 6.582 0.810 16.627 5.9453–4% 0.329 9.582 4.246 0.456 9.371 4.2464–5% 0.190 5.528 5.945 0.339 6.959 4.459> 5% 0.732 21.321 23.355 1.034 21.228 28.238
IO Report Year: 2002
Mean 0.260 4.930 3.410 0.300 3.560 4.190Median 0.010 2.170 1.340 0.010 2.080 1.5505th percentile 0.000 1.070 0.000 0.000 1.050 0.00095th percentile 0.770 16.190 12.660 1.180 10.500 16.810
Frequency Percentage0–1% 95.967 — 45.499 94.081 — 41.8491–2% 1.867 46.292 13.139 2.856 48.245 13.3822–3% 0.650 16.127 8.273 0.983 16.603 8.7593–4% 0.387 9.594 4.623 0.560 9.465 4.6234–5% 0.243 6.033 4.866 0.429 7.239 4.623> 5% 0.885 21.954 23.601 1.092 18.448 26.764
Panel B: Summary-Level
IO Report Year: 1982
Mean 1.280 5.710 8.220 1.310 4.000 9.600Median 0.160 2.350 5.060 0.340 2.250 6.6205th percentile 0.000 1.070 0.330 0.000 1.090 0.37095th percentile 4.450 23.130 23.300 5.750 11.610 30.040
29
Internet Appendix Table II - Continued
Customer % Supplier %
Inter-IndustryPairs
Inter-IndustryPairs> 1%
Intra-Industry
Inter-IndustryPairs
Inter-IndustryPairs> 1%
Intra-Industry
Frequency Percentage0–1% 80.349 — 9.091 71.736 — 9.0911–2% 8.305 42.261 14.286 12.167 43.047 10.3902–3% 3.349 17.044 5.195 5.332 18.863 5.1953–4% 2.290 11.652 14.286 2.666 9.432 7.7924–5% 1.162 5.913 6.494 2.051 7.255 10.390> 5% 4.546 23.130 50.649 6.049 21.403 57.143
IO Report Year: 1987
Mean 1.130 4.920 7.640 1.120 3.790 9.820Median 0.180 2.150 4.670 0.320 2.150 5.8905th percentile 0.000 1.100 0.210 0.010 1.070 0.25095th percentile 4.020 17.640 20.600 5.050 10.740 29.540
Frequency Percentage0–1% 80.647 — 12.644 75.969 — 12.6441–2% 8.794 45.442 9.195 11.067 46.051 3.4482–3% 3.475 17.956 12.644 4.411 18.354 13.7933–4% 2.058 10.635 9.195 2.085 8.676 5.7474–5% 1.016 5.249 11.494 1.337 5.562 8.046> 5% 4.010 20.718 44.828 5.132 21.357 56.322
IO Report Year: 1992
Mean 1.090 4.680 8.020 1.120 3.860 9.960Median 0.170 2.180 5.480 0.330 2.160 7.2205th percentile 0.000 1.070 0.170 0.000 1.050 0.20095th percentile 3.970 15.540 22.020 5.060 10.510 26.170
Frequency Percentage0–1% 80.199 — 12.500 76.176 — 11.3641–2% 8.882 44.855 7.955 11.076 46.491 7.9552–3% 4.075 20.581 12.500 3.919 16.447 4.5463–4% 1.855 9.367 10.227 2.116 8.882 12.5004–5% 1.097 5.541 3.409 1.646 6.908 7.955> 5% 3.892 19.657 53.409 5.068 21.272 55.682
IO Report Year: 1997
Mean 0.740 4.050 6.130 0.830 3.470 7.840Median 0.140 1.890 3.080 0.200 2.060 4.2005th percentile 0.000 1.060 0.000 0.000 1.060 0.00095th percentile 2.560 13.790 21.240 3.590 9.630 25.580
Frequency Percentage0–1% 85.904 — 33.065 81.327 — 30.6451–2% 7.435 52.744 8.065 9.035 48.385 8.8712–3% 2.465 17.488 8.871 3.344 17.907 4.032
30
Internet Appendix Table II - Continued
Customer % Supplier %
Inter-IndustryPairs
Inter-IndustryPairs> 1%
Intra-Industry
Inter-IndustryPairs
Inter-IndustryPairs> 1%
Intra-Industry
3–4% 1.075 7.628 3.226 1.928 10.323 4.8394–5% 0.629 4.465 4.839 1.180 6.320 5.645> 5% 2.492 17.674 41.936 3.187 17.065 45.968
IO Report Year: 2002
Mean 0.750 4.190 5.220 0.820 3.450 6.650Median 0.130 1.970 2.620 0.200 2.110 3.1705th percentile 0.000 1.050 0.000 0.000 1.060 0.00095th percentile 2.690 14.230 19.490 3.640 10.140 23.260
Frequency Percentage0–1% 85.994 — 35.714 81.448 — 30.1591–2% 7.162 51.133 9.524 8.749 47.160 11.9052–3% 2.451 17.498 7.143 3.708 19.986 5.5563–4% 1.270 9.066 6.349 1.549 8.350 7.1434–5% 0.686 4.896 4.762 1.308 7.050 6.349> 5% 2.438 17.407 36.508 3.238 17.454 38.889
31
Internet Appendix Table IIIMerger Summary Statistics by IO Report Year, Industry Assignment, and Industry CoarsenessThis table presents summary statistics of the sample of mergers over the period 1986 to 2010 by industry pairs. Merger data is from SDC. Industriesare defined by the Bureau of Economic Analysis Input-Output (IO) Industry classification. Inter-industry pairs include all combinations of theindustries (excluding own-industry pairs). Industry-level observations are observations at the IO Industry level. Intra-industry observationsinclude mergers of firms that are in the same IO industry. Inter-industry observations at the industry-level includes all inter-industry mergersacross all other industries for each of the industries divided by two, since each inter-industry merger is double-counted at the industry-level. 2010millions of US dollars are reported in brackets. The IO industry definitions are based on input-output relations between industries as recordedby the BEA in five separate reports (1982, 1987, 1992, 1997, and 2002), for two levels of industry definitions (Detailed and Summary). Panel Apresents the detailed-level data and panel B presents the summary-level data. Merging firms are assigned to IO industries in four ways to createfour different merger networks. In the ‘Primary’ network, firms are assigned to an industry based on its primary SIC/NAICS code as reportedin SDC. The ‘All’ network assigns firms using all industry codes reported by SDC. The ‘HV1%’ and ‘HV5%’ assign firms based on the followingpriority. If there are any shared IO codes between merging firms, we assign the merger equally to those industries. If there are no horizontalmatches, but a vertical relation of 1% or 5% for any of the four vertical relations (acquirer sells to target, acquirer buys from target, and viceversa), we assign the merger activity to those industries. If neither horizontal or vertical relations exist, we assign the merger activity equally tothe unrelated industry pairs.
Primary All HV1% HV5%
Industry-Level Industry-Level Industry-Level Industry-Level
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Panel A: Detailed-Level
IO Report Year: 1982
Observations 114,003 478 478 114,003 478 478 114,003 478 478 114,003 478 478Total Mergers 29,306 29,306 21,618 36,415 36,415 14,607 16,329 16,329 34,693 16,289 16,289 34,733
[$9,605,108] [$9,605,108] [$7,143,409] [$13,313,451] [$13,313,451] [$3,458,693] [$4,069,701] [$4,069,701] [$12,702,443] [$4,072,281] [$4,072,281] [$12,699,863]Mean 0.26 61.31 45.23 0.32 76.18 30.56 0.14 34.16 72.58 0.14 34.08 72.66
[$84] [$20,094] [$14,944] [$117] [$27,852] [$7,236] [$36] [$8,514] [$26,574] [$36] [$8,519] [$26,569]Median 0.00 11.00 2.00 0.00 13.22 1.42 0.00 5.89 6.00 0.00 6.37 6.04
[$0] [$2,048] [$120] [$0] [$3,763] [$121] [$0] [$912] [$1,006] [$0] [$1,053] [$1,001]5th Percentile 0.00 0.50 0.00 0.00 0.63 0.00 0.00 0.20 0.00 0.00 0.28 0.00
[$0] [$1] [$0] [$0] [$37] [$0] [$0] [$6] [$0] [$0] [$5] [$0]95th Percentile 0.00 264.00 120.00 0.50 320.10 102.92 0.17 122.73 300.78 0.22 125.83 299.37
[$0] [$70,205] [$38,776] [$111] [$107,115] [$18,282] [$6] [$20,118] [$89,627] [$10] [$25,426] [$93,049]Maximum 1,092 3,015 3,599 1,869 4,268 2,151 502 1,851 4,716 428 1,769 4,721
[$437,688] [$1,195,848] [$1,283,749] [$555,367] [$1,842,659] [$613,167] [$235,034] [$672,008] [$2,055,075] [$168,868] [$654,270] [$2,062,331]
Frequency PercentageNone 95.01 4.81 29.29 76.76 4.18 10.67 92.84 4.18 10.04 90.37 4.60 10.461 2.85 7.74 13.18 21.53 5.23 48.54 6.16 20.29 16.32 8.57 18.20 15.902–5 1.49 17.99 23.22 0.94 14.23 13.60 0.57 21.55 20.08 0.62 21.13 19.876–20 0.45 35.77 19.04 0.54 39.54 15.48 0.33 32.64 28.24 0.34 33.05 28.6621–50 0.13 15.48 5.86 0.14 16.74 3.56 0.07 11.72 10.88 0.06 12.97 10.46> 50 0.06 18.20 9.41 0.09 20.08 8.16 0.04 9.62 14.44 0.04 10.04 14.64
IO Report Year: 1987
Observations 107,880 465 465 107,880 465 465 107,880 465 465 107,880 465 465Total Mergers 29,151 29,151 21,828 36,341 36,341 14,686 16,279 16,279 34,747 16,265 16,265 34,765
[$9,368,623] [$9,368,623] [$7,401,431] [$13,291,926] [$13,291,926] [$3,482,394] [$4,092,115] [$4,092,115] [$12,682,153] [$4,097,521] [$4,097,521] [$12,676,810]Mean 0.27 62.69 46.94 0.34 78.15 31.58 0.15 35.01 74.72 0.15 34.98 74.76
[$87] [$20,148] [$15,917] [$123] [$28,585] [$7,489] [$38] [$8,800] [$27,273] [$38] [$8,812] [$27,262]
32Internet Appendix Table III - Continued
Primary All HV1% HV5%
Industry-Level Industry-Level Industry-Level Industry-Level
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Median 0.00 11.50 3.00 0.00 14.24 1.46 0.00 6.04 6.40 0.00 6.62 6.37[$0] [$2,197] [$132] [$0] [$4,095] [$136] [$0] [$987] [$1,133] [$0] [$1,204] [$1,008]
5th Percentile 0.00 1.50 0.00 0.00 2.06 0.00 0.00 0.67 0.00 0.00 0.81 0.00[$0] [$43] [$0] [$0] [$233] [$0] [$0] [$42] [$0] [$0] [$70] [$0]
95th Percentile 1.00 235.00 145.00 0.54 308.11 98.33 0.20 112.09 254.90 0.25 125.18 256.77[$4] [$65,486] [$38,251] [$126] [$103,944] [$17,193] [$8] [$19,650] [$81,805] [$11] [$22,538] [$81,595]
Maximum 1,061 3,024 3,605 1,861 4,261 2,133 484 1,876 4,725 632 1,983 4,706[$314,244] [$1,200,973] [$1,267,852] [$546,485] [$1,848,575] [$614,602] [$238,807] [$679,217] [$2,031,521] [$165,733] [$740,785] [$2,068,549]
Frequency PercentageNone 94.66 1.08 26.45 75.04 0.00 5.81 92.51 0.65 6.02 90.06 0.43 5.591 3.10 7.53 15.27 23.17 4.73 51.61 6.41 20.65 17.42 8.83 19.35 17.852–5 1.58 20.65 23.44 0.97 16.77 15.70 0.64 23.23 20.86 0.66 22.15 21.086–20 0.47 35.27 18.49 0.59 39.78 14.62 0.33 32.47 30.11 0.34 34.84 29.8921–50 0.13 16.56 6.45 0.14 18.49 3.87 0.07 12.47 10.97 0.08 12.47 10.75> 50 0.07 18.92 9.89 0.09 20.22 8.39 0.04 10.54 14.62 0.03 10.75 14.84
IO Report Year: 1992
Observations 112,575 475 475 112,575 475 475 112,575 475 475 112,575 475 475Total Mergers 29,721 29,721 21,264 36,810 36,810 14,220 16,538 16,538 34,493 16,536 16,536 34,491
[$10,051,327] [$10,051,327] [$6,719,186] [$13,409,750] [$13,409,750] [$3,364,822] [$4,315,123] [$4,315,123] [$12,459,294] [$4,286,637] [$4,286,637] [$12,487,616]Mean 0.26 62.57 44.77 0.33 77.50 29.94 0.15 34.82 72.62 0.15 34.81 72.61
[$89] [$21,161] [$14,146] [$119] [$28,231] [$7,084] [$38] [$9,084] [$26,230] [$38] [$9,025] [$26,290]Median 0.00 12.00 3.00 0.00 14.56 1.48 0.00 6.14 6.70 0.00 6.71 6.58
[$0] [$2,296] [$134] [$0] [$4,076] [$138] [$0] [$934] [$1,200] [$0] [$1,051] [$1,142]5th Percentile 0.00 1.00 0.00 0.00 1.93 0.00 0.00 0.54 0.00 0.00 0.67 0.00
[$0] [$37] [$0] [$0] [$253] [$0] [$0] [$33] [$0] [$0] [$54] [$0]95th Percentile 1.00 239.50 144.00 0.54 289.51 89.86 0.23 105.69 241.17 0.25 117.51 239.50
[$4] [$69,464] [$39,654] [$125] [$105,258] [$17,245] [$8] [$27,699] [$78,954] [$12] [$27,347] [$79,954]Maximum 1,032 3,000 3,613 1,860 4,242 2,134 560 2,033 4,704 687 1,710 4,709
[$315,737] [$1,210,133] [$1,278,731] [$557,208] [$1,837,369] [$610,049] [$324,764] [$767,350] [$2,066,211] [$317,950] [$641,245] [$2,109,117]
Frequency PercentageNone 94.67 0.84 26.74 75.10 0.21 7.16 92.45 0.63 5.68 89.67 0.42 6.321 3.02 8.42 13.26 23.07 5.05 48.63 6.49 21.68 17.68 9.26 20.00 16.422–5 1.61 18.74 24.21 0.99 15.58 15.58 0.63 21.05 20.00 0.61 20.42 20.846–20 0.49 35.16 18.32 0.60 38.95 15.58 0.33 33.68 29.47 0.35 35.16 29.6821–50 0.16 16.63 7.37 0.15 18.95 4.21 0.07 12.42 11.16 0.08 11.79 10.53> 50 0.06 20.21 10.11 0.09 21.26 8.84 0.04 10.53 16.00 0.03 12.21 16.21
IO Report Year: 1997
Observations 110,685 471 471 110,685 471 471 110,685 471 471 110,685 471 471Total Mergers 31,040 31,040 19,962 39,357 39,357 11,682 18,686 18,686 32,353 18,706 18,706 32,332
[$10,135,331] [$10,135,331] [$6,636,782] [$14,060,482] [$14,060,482] [$2,714,367] [$5,167,009] [$5,167,009] [$11,607,845] [$5,148,956] [$5,148,956] [$11,625,840]Mean 0.28 65.90 42.38 0.36 83.56 24.80 0.17 39.67 68.69 0.17 39.72 68.65
[$92] [$21,519] [$14,091] [$127] [$29,852] [$5,763] [$47] [$10,970] [$24,645] [$47] [$10,932] [$24,683]Median 0.00 15.00 4.00 0.00 17.69 2.11 0.00 8.30 8.08 0.00 8.91 7.93
[$0] [$2,867] [$244] [$0] [$5,189] [$182] [$0] [$1,355] [$1,577] [$0] [$1,568] [$1,501]5th Percentile 0.00 1.50 0.00 0.00 2.70 0.03 0.00 0.85 0.25 0.00 1.17 0.25
[$0] [$87] [$0] [$0] [$329] [$0] [$0] [$76] [$2] [$0] [$96] [$6]95th Percentile 1.00 287.50 200.00 0.69 332.45 130.90 0.33 161.30 304.23 0.33 150.95 306.12
[$8] [$75,990] [$53,046] [$167] [$123,286] [$24,468] [$15] [$38,388] [$105,981] [$21] [$34,897] [$100,023]Maximum 1,008 3,320 3,118 987 2,305 1,083 365 1,684 2,702 435 1,766 2,723
[$410,643] [$1,749,955] [$1,153,641] [$276,274] [$1,331,363] [$256,247] [$207,000] [$860,549] [$958,644] [$224,891] [$922,087] [$969,073]
Frequency Percentage
33
Internet Appendix Table III - Continued
Primary All HV1% HV5%
Industry-Level Industry-Level Industry-Level Industry-Level
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
None 94.16 0.85 18.47 70.16 0.21 4.46 90.50 0.64 4.25 86.20 0.85 3.821 3.35 4.46 11.68 27.63 3.40 44.16 8.09 11.25 11.89 12.49 10.83 12.312–5 1.70 15.29 26.54 1.19 8.28 19.75 0.85 21.87 18.68 0.76 21.23 20.596–20 0.59 39.28 23.35 0.75 41.40 16.77 0.43 38.22 34.82 0.43 39.07 32.2721–50 0.13 20.38 7.86 0.17 21.66 6.16 0.09 14.01 12.10 0.09 13.38 13.16> 50 0.07 19.75 12.10 0.09 25.05 8.70 0.04 14.01 18.26 0.04 14.65 17.83
IO Report Year: 2002
Observations 84,255 411 411 84,255 411 411 84,255 411 411 84,255 411 411Total Mergers 31,250 31,250 19,754 39,541 39,541 11,498 18,677 18,677 32,363 18,710 18,710 32,329
[$10,077,807] [$10,077,807] [$6,694,439] [$14,079,896] [$14,079,896] [$2,694,908] [$5,188,623] [$5,188,623] [$11,586,235] [$5,110,727] [$5,110,727] [$11,664,122]Mean 0.37 76.03 48.06 0.47 96.21 27.98 0.22 45.44 78.74 0.22 45.52 78.66
[$120] [$24,520] [$16,288] [$167] [$34,258] [$6,557] [$62] [$12,624] [$28,190] [$61] [$12,435] [$28,380]Median 0.00 20.50 6.00 0.00 25.37 3.00 0.00 11.21 13.15 0.00 11.99 13.42
[$0] [$4,042] [$572] [$0] [$7,510] [$346] [$0] [$2,018] [$2,332] [$0] [$2,210] [$2,369]5th Percentile 0.00 2.00 0.00 0.00 3.59 0.08 0.00 1.50 0.83 0.00 1.69 0.83
[$0] [$120] [$0] [$0] [$480] [$2] [$0] [$96] [$7] [$0] [$135] [$8]95th Percentile 1.00 308.50 273.00 1.07 342.51 143.03 0.50 166.10 345.83 0.50 157.72 339.49
[$31] [$96,481] [$57,169] [$284] [$129,823] [$30,918] [$29] [$35,933] [$135,347] [$36] [$38,420] [$133,141]Maximum 1,015 3,318 3,102 977 2,298 1,082 320 1,821 2,707 515 1,831 2,703
[$380,938] [$1,746,083] [$1,158,745] [$278,555] [$1,327,928] [$278,950] [$158,628] [$1,019,052] [$1,142,664] [$244,370] [$983,992] [$1,143,063]
Frequency PercentageNone 92.35 0.24 14.36 64.57 0.24 2.68 88.59 0.73 2.92 83.05 0.73 2.431 4.29 3.41 9.49 32.35 2.19 35.52 9.59 5.84 9.73 15.21 5.60 10.462–5 2.27 11.92 25.06 1.71 5.60 23.36 1.06 19.95 15.82 0.99 16.79 15.096–20 0.82 34.06 27.01 1.03 33.82 22.14 0.58 38.69 33.09 0.58 42.34 32.1221–50 0.18 25.06 10.71 0.21 27.49 6.57 0.12 18.98 16.30 0.12 18.73 18.25> 50 0.08 25.30 13.38 0.13 30.66 9.73 0.06 15.82 22.14 0.05 15.82 21.65
Panel B: Summary-Level
IO Report Year: 1982
Observations 2,926 77 77 2,926 77 77 2,926 77 77 2,926 77 77Total Mergers 23,427 23,427 27,497 29,235 29,235 21,787 12,932 12,932 38,090 12,873 12,873 38,149
[$7,577,479] [$7,577,479] [$9,171,038] [$10,779,947] [$10,779,947] [$5,992,197] [$3,077,027] [$3,077,027] [$13,695,117] [$3,064,692] [$3,064,692] [$13,707,452]Mean 8.01 304.25 357.10 9.99 379.68 282.95 4.42 167.95 494.68 4.40 167.18 495.44
[$2,590] [$98,409] [$119,104] [$3,684] [$139,999] [$77,821] [$1,052] [$39,961] [$177,859] [$1,047] [$39,801] [$178,019]Median 0.00 110.50 52.00 0.65 130.25 26.88 0.00 50.07 91.83 0.08 50.10 92.42
[$00] [$30,007] [$8,107] [$135] [$33,197] [$5,104] [$00] [$8,420] [$22,540] [$01] [$9,157] [$22,716]5th Percentile 0.00 11.00 2.00 0.00 9.22 1.74 0.00 3.58 7.33 0.00 2.75 7.33
[$0] [$2126] [$14] [$0] [$3554] [$171] [$0] [$520] [$1663] [$0] [$370] [$1663]95th Percentile 27.00 1146.00 1408.00 32.73 1135.73 1232.53 15.40 504.08 1695.83 14.60 562.51 1694.92
[$6261] [$342,373] [$588,221] [$11543] [$613,860] [$336,431] [$2599] [$86,684] [$799,104] [$2622] [$90,079] [$784,026]Maximum 1,355 4,360 7,041 1,752 4,858 6,404 719 2,254 8,065 1,159 2,361 8,077
[$500,484] [$1,886,317] [$2,997,369] [$891,682] [$1,879,818] [$2,598,990] [$355,535] [$762,179] [$3,681,977] [$407,786] [$688,839] [$3,692,103]
Frequency PercentageNone 52.02 0.00 2.60 20.16 0.00 1.30 58.13 0.00 1.30 48.22 0.00 1.301 14.73 1.30 1.30 47.74 0.00 3.90 23.55 3.90 2.60 35.24 1.30 1.302–5 17.70 1.30 7.79 13.74 2.60 6.49 7.79 1.30 0.00 6.53 3.90 1.306–20 9.36 9.09 23.38 11.21 6.49 29.87 6.56 18.18 10.39 6.22 18.18 9.0921–50 3.28 16.88 12.99 3.86 11.69 18.18 2.19 25.97 23.38 1.78 25.97 24.68> 50 2.91 71.43 51.95 3.28 79.22 40.26 1.78 50.65 62.34 2.02 50.65 62.34
IO Report Year: 1987
34Internet Appendix Table III - Continued
Primary All HV1% HV5%
Industry-Level Industry-Level Industry-Level Industry-Level
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Observations 3,741 87 87 3,741 87 87 3,741 87 87 3,741 87 87Total Mergers 25,507 25,507 25,468 32,026 32,026 19,002 14,150 14,150 36,876 14,170 14,170 36,858
[$8,433,825] [$8,433,825] [$8,335,420] [$11,920,577] [$11,920,577] [$4,853,842] [$3,620,229] [$3,620,229] [$13,154,066] [$3,656,620] [$3,656,620] [$13,117,625]Mean 6.82 293.18 292.74 8.56 368.12 218.41 3.78 162.64 423.86 3.79 162.87 423.66
[$2,254] [$96,941] [$95,809] [$3,186] [$137,018] [$55,791] [$968] [$41,612] [$151,196] [$977] [$42,030] [$150,777]Median 0.00 122.00 74.00 0.77 138.29 44.11 0.00 55.89 120.28 0.17 57.84 117.21
[$00] [$34,487] [$13,462] [$148] [$43,224] [$7,120] [$00] [$11,815] [$28,819] [$03] [$10,174] [$27,416]5th Percentile 0.00 19.50 5.00 0.00 27.16 4.39 0.00 9.01 13.50 0.00 8.92 11.60
[$0] [$3104] [$810] [$0] [$10203] [$404] [$0] [$1633] [$2284] [$0] [$2527] [$1988]95th Percentile 24.00 1117.00 977.00 31.22 1353.92 741.85 14.20 517.54 1468.45 13.38 538.77 1467.87
[$5868] [$249,820] [$548,367] [$11328] [$601,129] [$146,358] [$2439] [$120,283] [$480,873] [$2719] [$92,778] [$474,371]Maximum 1,246 4,243 5,899 1,348 4,214 5,224 466 2,246 6,995 730 2,225 6,982
[$553,876] [$1,928,503] [$2,327,680] [$558,462] [$1,957,521] [$1,822,735] [$261,654] [$847,044] [$3,189,871] [$322,351] [$818,872] [$3,172,880]
Frequency PercentageNone 50.47 0.00 0.00 15.74 0.00 0.00 55.31 0.00 0.00 42.50 0.00 0.001 16.12 0.00 0.00 51.67 0.00 3.45 25.85 1.15 1.15 39.96 1.15 0.002–5 17.94 0.00 5.75 13.69 1.15 3.45 8.45 1.15 0.00 7.32 1.15 1.156–20 9.70 5.75 22.99 11.79 3.45 26.44 6.66 14.94 8.05 6.47 14.94 8.0521–50 3.10 17.24 12.64 3.98 8.05 21.84 2.30 28.74 19.54 2.30 26.44 18.39> 50 2.67 77.01 58.62 3.13 87.36 44.83 1.44 54.02 71.26 1.44 56.32 72.41
IO Report Year: 1992
Observations 3,828 88 88 3,828 88 88 3,828 88 88 3,828 88 88Total Mergers 25,469 25,469 25,518 31,924 31,924 19,104 14,127 14,127 36,900 14,111 14,111 36,919
[$8,410,189] [$8,410,189] [$8,360,575] [$11,874,012] [$11,874,012] [$4,900,357] [$3,639,017] [$3,639,017] [$13,135,318] [$3,673,525] [$3,673,525] [$13,100,868]Mean 6.65 289.42 289.98 8.34 362.78 217.09 3.69 160.53 419.32 3.69 160.35 419.53
[$2,197] [$95,570] [$95,007] [$3,102] [$134,932] [$55,686] [$951] [$41,352] [$149,265] [$960] [$41,745] [$148,873]Median 0.00 120.25 73.00 0.68 138.99 42.77 0.00 57.35 117.46 0.17 59.89 115.66
[$00] [$32,349] [$12,432] [$131] [$38,591] [$6,892] [$00] [$11,831] [$28,341] [$02] [$10,263] [$27,756]5th Percentile 0.00 19.00 5.00 0.00 27.20 4.18 0.00 8.62 13.00 0.00 8.62 12.77
[$0] [$3377] [$770] [$0] [$10668] [$553] [$0] [$2493] [$2454] [$0] [$2307] [$2382]95th Percentile 22.00 1111.50 976.00 29.39 1123.02 766.88 14.02 494.20 1481.92 13.02 567.24 1481.75
[$5709] [$261,090] [$541,909] [$11102] [$594,703] [$149,898] [$2386] [$120,674] [$498,719] [$2517] [$133,481] [$473,398]Maximum 1,243 4,248 5,912 1,340 4,205 5,235 455 2,257 7,011 679 2,135 6,998
[$580,957] [$1,898,798] [$2,358,694] [$576,630] [$1,962,963] [$1,842,090] [$224,461] [$837,100] [$3,172,680] [$452,920] [$864,832] [$3,198,057]
Frequency PercentageNone 51.99 0.00 0.00 16.33 0.00 0.00 57.45 0.00 0.00 44.20 0.00 0.001 15.67 0.00 0.00 52.40 0.00 3.41 23.88 1.14 1.14 39.05 1.14 1.142–5 16.90 0.00 5.68 12.85 1.14 4.55 8.44 1.14 0.00 7.00 1.14 0.006–20 10.06 5.68 22.73 11.60 3.41 25.00 6.56 13.64 6.82 6.40 13.64 7.9521–50 2.93 17.05 12.50 3.84 7.95 21.59 2.30 28.41 20.45 1.91 28.41 19.32> 50 2.46 77.27 59.09 2.98 87.50 45.45 1.38 55.68 71.59 1.44 55.68 71.59
IO Report Year: 1997
Observations 7,626 124 124 7,626 124 124 7,626 124 124 7,626 124 124Total Mergers 28,672 28,672 22,330 36,979 36,979 14,060 17,068 17,068 33,970 17,004 17,004 34,035
[$9,534,958] [$9,534,958] [$7,237,104] [$13,470,989] [$13,470,989] [$3,303,859] [$4,648,768] [$4,648,768] [$12,126,027] [$4,620,242] [$4,620,242] [$12,154,607]Mean 3.76 231.23 180.08 4.85 298.22 113.38 2.24 137.65 273.95 2.23 137.13 274.48
[$1,250] [$76,895] [$58,364] [$1,766] [$108,637] [$26,644] [$610] [$37,490] [$97,791] [$606] [$37,260] [$98,021]Median 0.00 105.25 68.50 0.44 127.81 41.07 0.00 58.84 103.92 0.11 53.83 100.56
[$00] [$29,687] [$10,474] [$60] [$41,402] [$5,406] [$00] [$11,542] [$25,389] [$02] [$10,690] [$25,720]5th Percentile 0.00 13.00 4.00 0.00 21.18 2.61 0.00 9.15 5.83 0.00 9.28 6.17
[$0] [$2155] [$139] [$0] [$4417] [$108] [$0] [$1970] [$421] [$0] [$1539] [$445]95th Percentile 14.00 832.50 627.00 18.51 1337.70 445.23 9.44 465.72 1056.40 9.20 408.93 1040.25
35
Internet Appendix Table III - Continued
Primary All HV1% HV5%
Industry-Level Industry-Level Industry-Level Industry-Level
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
Inter-IndustryPairs
Inter-Industry
Intra-Industry
[$2819] [$271,076] [$259,701] [$6671] [$475,470] [$127,545] [$1693] [$120,304] [$385,918] [$1765] [$111,675] [$392,451]Maximum 1,065 3,319 4,036 1,595 2,323 2,066 373 1,622 3,751 438 1,732 3,746
[$453,447] [$1,748,481] [$1,537,578] [$469,780] [$1,356,666] [$378,266] [$210,180] [$850,532] [$1,019,300] [$231,038] [$929,984] [$1,039,428]
Frequency PercentageNone 61.51 0.00 2.42 22.90 0.00 0.81 57.85 0.81 1.61 44.28 0.00 0.811 14.99 0.81 0.81 51.63 1.61 3.23 26.17 0.81 0.81 41.32 1.61 1.612–5 13.56 0.81 3.23 10.95 0.00 5.65 7.34 0.81 0.00 6.41 0.81 0.816–20 6.40 6.45 18.55 9.91 3.23 26.61 6.40 12.10 9.68 5.65 12.90 8.8721–50 2.16 19.35 16.94 2.66 11.29 20.16 1.61 31.45 17.74 1.57 30.65 17.74> 50 1.38 72.58 58.06 1.95 83.87 43.55 0.62 54.03 70.16 0.76 54.03 70.16
IO Report Year: 2002
Observations 7,875 126 126 7,875 126 126 7,875 126 126 7,875 126 126Total Mergers 28,598 28,598 22,405 37,038 37,038 14,001 16,888 16,888 34,151 16,912 16,912 34,127
[$9,424,031] [$9,424,031] [$7,348,084] [$13,468,766] [$13,468,766] [$3,306,088] [$4,619,114] [$4,619,114] [$12,155,735] [$4,625,816] [$4,625,816] [$12,149,033]Mean 3.63 226.97 177.82 4.70 293.96 111.12 2.14 134.03 271.04 2.15 134.22 270.85
[$1,197] [$74,794] [$58,318] [$1,710] [$106,895] [$26,239] [$587] [$36,660] [$96,474] [$587] [$36,713] [$96,421]Median 0.00 100.75 67.00 0.40 124.10 37.02 0.00 48.39 99.39 0.08 54.48 99.75
[$00] [$27,403] [$9,776] [$53] [$41,365] [$5,236] [$00] [$9,718] [$24,090] [$01] [$10,426] [$25,463]5th Percentile 0.00 13.00 3.00 0.00 20.18 2.61 0.00 8.73 6.33 0.00 9.24 6.00
[$0] [$1038] [$100] [$0] [$4435] [$108] [$0] [$1322] [$376] [$0] [$1319] [$415]95th Percentile 13.00 914.50 625.00 17.61 1396.22 429.17 8.53 477.93 1185.73 8.67 483.17 1173.85
[$2637] [$253,659] [$247,036] [$6477] [$506,908] [$134,371] [$1413] [$128,945] [$469,336] [$1651] [$128,817] [$424,611]Maximum 1,028 3,315 4,044 1,596 2,324 2,068 346 1,685 3,756 513 1,793 3,758
[$508,810] [$1,745,504] [$1,540,361] [$483,237] [$1,348,090] [$390,828] [$160,938] [$947,614] [$1,112,829] [$254,630] [$970,379] [$1,146,522]
Frequency PercentageNone 62.76 0.00 2.38 23.85 0.00 1.59 59.91 0.79 1.59 46.41 0.00 1.591 14.65 0.79 0.79 51.43 1.59 3.17 24.69 0.79 0.79 39.50 1.59 0.792–5 13.07 0.79 3.97 10.81 0.00 3.97 7.43 0.79 0.00 6.18 0.00 0.796–20 6.11 7.14 19.05 9.42 3.17 27.78 5.71 13.49 10.32 5.65 13.49 9.5221–50 2.10 20.63 18.25 2.69 11.11 23.81 1.54 34.92 18.25 1.52 32.54 18.25> 50 1.32 70.63 55.56 1.80 84.13 39.68 0.72 49.21 69.05 0.72 52.38 69.05
36
Internet Appendix Table IVSummary Statistics of Industry and Merger Networks by IO Report Year, IndustryAssignment, and Industry CoarsenessThis table reports means and medians (in brackets) for six different networks: Input-Output (IO)supply relations, IO customer relations, and four merger networks. The IO networks are based oninput-output relations between industries as recorded by the BEA in five separate reports (1982,1987, 1992, 1997, and 2002), for two levels of industry definitions (Detailed and Summary). Mergernetworks have connections equal to the number of mergers between two industries, where industrydefinitions follow the IO industries. Merging firms are assigned to IO industries in four ways tocreate four different merger networks. In the ‘Primary’ network, firms are assigned to an industrybased on its primary SIC/NAICS code as reported in SDC. The ‘All’ network assigns firms usingall industry codes reported by SDC. The ‘HV1%’ and ‘HV5%’ assign firms based on the followingpriority. If there are any shared IO codes between merging firms, we assign the merger equally tothose industries. If there are no horizontal matches, but a vertical relation of 1% or 5% for any of thefour vertical relations (acquirer sells to target, acquirer buys from target, and vice versa), we assignthe merger activity to those industries. If neither horizontal or vertical relations exist, we assign themerger activity equally to the unrelated industry pairs. ‘Degree Centrality’, ‘Eigenvector Centrality’,‘Average Shortest Path’, ‘Clustering Coefficient’, and ‘Max(Shortest Distance)’ are defined in theInternet Appendix.
Network: Supplier Customer Merger
Firm Industry Codes: Primary All HV1% HV5%
Panel A: Detail-Level
IO Report Year: 1982
Degree Centrality 22.490 15.092 23.812 110.837 34.146 45.921[15.000] [11.000] [14.000] [93.000] [20.000] [28.000]
Eigenvector Centrality 0.037 0.033 0.046 0.045 0.046 0.046[0.033] [0.026] [0.045] [0.044] [0.044] [0.044]
Average Shortest Path 1.979 2.709 2.114 1.744 1.955 1.907[1.977] [2.671] [2.053] [1.786] [1.966] [1.941]
Clustering Coefficient 0.441 0.219 0.456 0.603 0.564 0.568[0.437] [0.200] [0.446] [0.616] [0.562] [0.560]
Max(Shortest Distance) 3.000 7.000 5.000 3.000 4.000 4.000
IO Report Year: 1987
Degree Centrality 22.151 13.789 24.791 115.819 34.774 46.120[15.000] [11.000] [15.000] [95.000] [22.000] [29.000]
Eigenvector Centrality 0.037 0.032 0.046 0.046 0.046 0.046[0.033] [0.021] [0.045] [0.044] [0.045] [0.045]
Average Shortest Path 1.976 2.740 2.104 1.749 1.946 1.918[1.976] [2.694] [2.049] [1.791] [1.961] [1.946]
Clustering Coefficient 0.440 0.237 0.470 0.629 0.591 0.596[0.438] [0.205] [0.444] [0.624] [0.581] [0.576]
Max(Shortest Distance) 3.000 8.000 5.000 3.000 4.000 4.000
IO Report Year: 1992
Degree Centrality 21.680 14.699 25.272 118.021 35.806 48.977[15.000] [11.000] [15.000] [99.000] [22.000] [31.000]
37
Internet Appendix Table IV - Continued
Network: Supplier Customer Merger
Firm Industry Codes: Primary All HV1% HV5%
Eigenvector Centrality 0.037 0.032 0.046 0.046 0.046 0.046[0.033] [0.023] [0.045] [0.044] [0.044] [0.044]
Average Shortest Path 1.990 2.730 2.113 1.749 1.953 1.917[1.981] [2.688] [2.049] [1.788] [1.967] [1.941]
Clustering Coefficient 0.440 0.231 0.474 0.621 0.587 0.572[0.444] [0.207] [0.467] [0.614] [0.576] [0.546]
Max(Shortest Distance) 3.000 7.000 5.000 3.000 4.000 4.000
IO Report Year: 1997
Degree Centrality 22.883 16.132 27.444 140.246 44.633 64.862[16.000] [13.000] [18.000] [124.000] [30.000] [47.000]
Eigenvector Centrality 0.037 0.033 0.046 0.046 0.046 0.046[0.033] [0.025] [0.045] [0.045] [0.045] [0.044]
Average Shortest Path 1.966 2.537 2.075 1.698 1.920 1.868[1.971] [2.467] [2.032] [1.733] [1.941] [1.899]
Clustering Coefficient 0.461 0.275 0.422 0.626 0.548 0.540[0.462] [0.250] [0.400] [0.625] [0.533] [0.527]
Max(Shortest Distance) 3.000 6.000 5.000 3.000 4.000 4.000
IO Report Year: 2002
Degree Centrality 24.268 16.535 31.348 145.275 46.788 69.489[17.000] [14.000] [22.000] [132.000] [34.000] [53.000]
Eigenvector Centrality 0.041 0.036 0.049 0.049 0.049 0.049[0.036] [0.029] [0.048] [0.048] [0.048] [0.047]
Average Shortest Path 1.948 2.469 1.995 1.641 1.894 1.835[1.959] [2.390] [1.981] [1.672] [1.918] [1.869]
Clustering Coefficient 0.460 0.258 0.442 0.652 0.527 0.564[0.471] [0.255] [0.427] [0.646] [0.517] [0.559]
Max(Shortest Distance) 3.000 5.000 4.000 3.000 4.000 4.000
Panel B: Summary-Level
IO Report Year: 1982
Degree Centrality 21.481 14.935 36.468 60.675 31.818 39.351[17.000] [13.000] [39.000] [66.000] [32.000] [43.000]
Eigenvector Centrality 0.103 0.101 0.113 0.113 0.113 0.113[0.091] [0.090] [0.116] [0.118] [0.113] [0.116]
Average Shortest Path 1.695 1.922 1.500 1.186 1.561 1.463[1.753] [1.883] [1.468] [1.117] [1.558] [1.416]
Clustering Coefficient 0.635 0.444 0.736 0.902 0.696 0.763[0.675] [0.403] [0.728] [0.900] [0.680] [0.760]
Max(Shortest Distance) 2.000 4.000 2.000 2.000 2.000 2.000
IO Report Year: 1987
Degree Centrality 20.667 16.644 42.598 72.460 38.437 49.448[17.000] [16.000] [43.000] [77.000] [36.000] [52.000]
38
Internet Appendix Table IV - Continued
Network: Supplier Customer Merger
Firm Industry Codes: Primary All HV1% HV5%
Eigenvector Centrality 0.097 0.095 0.106 0.107 0.106 0.106[0.086] [0.095] [0.107] [0.110] [0.105] [0.109]
Average Shortest Path 1.745 1.915 1.487 1.144 1.535 1.409[1.782] [1.897] [1.483] [1.092] [1.563] [1.379]
Clustering Coefficient 0.554 0.426 0.700 0.904 0.680 0.757[0.567] [0.400] [0.689] [0.906] [0.686] [0.760]
Max(Shortest Distance) 3.000 4.000 2.000 2.000 2.000 2.000
IO Report Year: 1992
Degree Centrality 20.727 17.227 41.773 72.796 37.023 48.546[16.000] [16.500] [40.500] [77.000] [33.000] [52.000]
Eigenvector Centrality 0.096 0.093 0.106 0.106 0.106 0.106[0.086] [0.086] [0.106] [0.109] [0.103] [0.109]
Average Shortest Path 1.742 1.908 1.503 1.150 1.557 1.426[1.796] [1.886] [1.517] [1.102] [1.602] [1.386]
Clustering Coefficient 0.579 0.425 0.688 0.898 0.673 0.757[0.590] [0.400] [0.691] [0.904] [0.676] [0.759]
Max(Shortest Distance) 2.000 4.000 2.000 2.000 2.000 2.000
IO Report Year: 1997
Degree Centrality 22.968 17.339 47.339 94.839 51.839 68.532[18.000] [15.000] [44.000] [101.000] [48.500] [68.000]
Eigenvector Centrality 0.079 0.076 0.089 0.089 0.089 0.089[0.070] [0.071] [0.088] [0.092] [0.088] [0.089]
Average Shortest Path 1.801 2.029 1.605 1.219 1.559 1.431[1.846] [1.988] [1.629] [1.169] [1.585] [1.436]
Clustering Coefficient 0.527 0.363 0.628 0.866 0.659 0.752[0.550] [0.353] [0.632] [0.867] [0.669] [0.759]
Max(Shortest Distance) 3.000 4.000 3.000 3.000 2.000 3.000
IO Report Year: 2002
Degree Centrality 23.191 17.508 46.556 95.191 50.111 66.984[18.000] [16.000] [43.500] [100.000] [45.500] [66.000]
Eigenvector Centrality −0.079 0.075 0.088 0.089 0.088 0.088[−0.070] [0.069] [0.087] [0.091] [0.086] [0.088]
Average Shortest Path 1.805 2.056 1.618 1.229 1.580 1.453[1.848] [2.008] [1.639] [1.191] [1.616] [1.460]
Clustering Coefficient 0.534 0.394 0.624 0.857 0.650 0.741[0.535] [0.398] [0.628] [0.860] [0.656] [0.740]
Max(Shortest Distance) 3.000 4.000 3.000 2.000 2.000 3.000
39
Internet Appendix Table VThe Most Central Industries in the IO and Merger Networks: All IO Report Years, Detailed and Summary Level,Primary and All IndustriesIO degree centrality is measured using the binary connections in the Input-Output Network using data from the U.S. Bureau of Economic Analysis reportsin years 1982, 1987, 1992, 1997, and 2002. Panel A reports degree centrality using industries at the detailed level of industries in the BEA reports. Panel Breports the same at the summary level of industries. A binary connection is defined as a connection where one industry either supplies at least 1% of theconnected industry’s inputs (‘Supplier Network’), or buys at least 1% of the connected industry’s output (‘Customer Network’). Merger degree centrality ismeasured using the binary network of inter-industry mergers, where a binary connection is defined as any inter-industry merger between two industries over1986 to 2010. In the ‘Merger Network: Primary Industry’ column, we assign merging firms to IO industries using only the primary SIC/NAICS code reportedin SDC. In the ‘Merger Network: All Industries’ column, we assign merging firms to IO industries using all SIC/NAICS codes reported in SDC. Entries inboldface in the merger network columns indicate that the industry is also one of the top 15 most central industries in either the supplier or customer industrynetwork.
Rank Supplier Network Customer Network Merger Network: Primary Industry Merger Network: All Industries
Panel A: Detail-Level
IO Report Year: 1982
1 Wholesale trade Large construction Credit agencies other than banks Credit agencies other than banks2 Private gas & elec. srvcs. (utilities) Construction, misc. Security & commodity brokers Wholesale trade3 Petroleum refining Eating & drinking places Wholesale trade Security & commodity brokers4 Motor freight transp. & warehousing Wholesale trade Computer & data proc. srvcs. Mgmt. consulting srvcs. & testing & rsrch. labs5 Miscellaneous plastics products Retail trade Retail trade Retail trade6 Blast furnaces & steel mills Facilities construction Mgmt. consulting srvcs. & testing & rsrch. labs Apparel made from purchased mats. & dressed furs7 Paperboard containers & boxes Blast furnaces & steel mills Other business srvcs. Miscellaneous plastics products8 Banking Home Construction Miscellaneous plastics products Other business srvcs.9 Industrial organic chems. ex. gum & wood chems. Motor vehicles & car bodies Other electronic components Computer & data proc. srvcs.
10 Construction, misc. Private hospitals Apparel made from purchased mats. & dressed furs Manufacturing industries, n.e.c. ex. fur dressing11 Real estate Miscellaneous plastics products Mechanical measuring devices Insurance carriers12 Railroads & related srvcs. Real estate Radio & TV communication equip. Chemical preparations, n.e.c.13 Communications, ex. radio & TV Motor vehicle parts & accessories Motor vehicle parts & accessories Other electronic components14 Commercial printing Banking Drugs Radio & TV communication equip.15 Eating & drinking places Miscellaneous repair shops Aircraft & missile equip., n.e.c. Mechanical measuring devices
IO Report Year: 1987
1 Wholesale trade New construction & maintenance & repair Credit agencies other than banks Wholesale trade2 Elec. & gas srvcs. (utilities) Wholesale trade Security & commodity brokers Credit agencies other than banks3 Motor freight transp. & warehousing Eating & drinking places Wholesale trade Security & commodity brokers4 Miscellaneous plastics products, n.e.c. Motor vehicles & passenger car bodies Retail trade, ex. eating & drinking Mgmt. & consulting srvcs., testing & rsrch. labs5 Paperboard containers & boxes Retail trade, ex. eating & drinking Computer & data proc. srvcs. Retail trade, ex. eating & drinking6 New construction & maintenance & repair Blast furnaces & steel mills Mgmt. & consulting srvcs., testing & rsrch. labs Apparel made from purchased mats.7 Blast furnaces & steel mills Miscellaneous plastics products, n.e.c. New construction & maintenance & repair Miscellaneous plastics products, n.e.c.8 Real estate agents, mgrs., opers., & lessors Hospitals Other electronic components Computer & data proc. srvcs.9 Industrial inorganic & organic chems. Motor vehicle parts & accessories Apparel made from purchased mats. New construction & maintenance & repair
10 Banking Real estate agents, mgrs., opers., & lessors Miscellaneous plastics products, n.e.c. Manufacturing industries, n.e.c.11 Commercial printing Paper & paperboard mills Mechanical measuring devices Other business srvcs.12 Radio & TV broadcasting Banking Motor vehicle parts & accessories Insurance carriers13 Plastics mats. & resins Automotive rental & leasing, w.o. drivers Drugs Mechanical measuring devices14 Railroads & related srvcs. Mgmt. & consulting srvcs., testing & rsrch. labs Motor freight transp. & warehousing Other electronic components15 Automotive rental & leasing, w.o. drivers Miscellaneous repair shops Aircraft & missile equip., n.e.c. Industrial inorganic & organic chems.
IO Report Year: 1992
1 Wholesale trade Construction, large Credit agencies other than banks Credit agencies other than banks2 Trucking & courier srvcs., ex. air Other new construction & maintenance Wholesale trade Wholesale trade3 Elec. srvcs. (utilities) Wholesale trade Security & commodity brokers Security & commodity brokers4 Miscellaneous plastics products, n.e.c. New residential 1-unit structures, nonfarm Acctng., auditing & bookkeeping, & misc. srvcs., n.e.c. Miscellaneous equip. rental & leasing5 Real estate agents, mgrs., opers., & lessors Eating & drinking places Retail trade, ex. eating & drinking Retail trade, ex. eating & drinking6 Paperboard containers & boxes Retail trade, ex. eating & drinking Miscellaneous equip. rental & leasing Manufacturing industries, n.e.c.7 Blast furnaces & steel mills Hospitals Other electronic components Miscellaneous plastics products, n.e.c.8 Other new construction & maintenance Blast furnaces & steel mills Detective & protective srvcs. Acctng., auditing & bookkeeping, & misc. srvcs., n.e.c.9 Industrial inorganic & organic chems. Truck trailers Mechanical measuring devices Detective & protective srvcs.
10 Petroleum refining Real estate agents, mgrs., opers., & lessors Manufacturing industries, n.e.c. Industrial inorganic & organic chems.11 Banking Motor vehicle parts & accessories Industrial inorganic & organic chems. Insurance carriers12 Mgmt. & consulting srvcs. Miscellaneous plastics products, n.e.c. Motor vehicle parts & accessories Chemicals & chemical preparations, n.e.c.13 Railroads & related srvcs. Industrial inorganic & organic chems. Miscellaneous plastics products, n.e.c. Mechanical measuring devices14 Industrial & commercial mach. & equip., n.e.c. Banking Drugs Other electronic components15 Natural gas distribution Highways & streets Aircraft & missile equip., n.e.c. New residential 1-unit structures, nonfarm
IO Report Year: 1997
1 Wholesale trade Construction Securities, commodity contracts, invmts. Wholesale trade2 Mgmt. of companies & enterprises Wholesale trade Wholesale trade Securities, commodity contracts, invmts.3 Truck transp. Retail trade Retail trade Mgmt. consulting srvcs.4 Power generation & supply Motor vehicle parts manuf. Construction Retail trade5 Real estate Real estate Funds, trusts, & other financial vehicles Mgmt. of companies & enterprises6 Iron & steel mills Food srvcs. & drinking places Software reproducing Construction7 Paperboard container manuf. Hospitals Motor vehicle parts manuf. Funds, trusts, & other financial vehicles8 Plastics plumbing fixtures & all other plastics products Telecommunications Information srvcs. Motor vehicle parts manuf.9 Monetary auth. & depository credit intermed. Iron & steel mills All other electronic component manuf. Architectural & engineering srvcs.
10 Lessors of nonfinancial intangible assets Power generation & supply Mgmt. consulting srvcs. Scientific rsrch. & development srvcs.11 Other basic organic chemical manuf. Mgmt. of companies & enterprises Architectural & engineering srvcs. Plastics plumbing fixtures & all other plastics products12 Scientific rsrch. & development srvcs. Automobile & light truck manuf. Business support srvcs. All other electronic component manuf.13 Plastics pkg. mats., film & sheet Paper & paperboard mills Software publishers Other misc. chemical product manuf.14 Telecommunications Offices of physicians, dentists, & other health prctnrs. Pharmaceutical & medicine manuf. Other commercial & service industry mach. manuf.15 Petroleum refineries Other basic organic chemical manuf. Laboratory apparatus & furniture manuf. Laboratory apparatus & furniture manuf.
IO Report Year: 2002
1 Wholesale trade Nonresidential manuf. structures Securities, commodity contracts, invmts., & related acts. Securities, commodity contracts, invmts., & related acts.2 Mgmt. of companies & enterprises Retail trade Wholesale trade Wholesale trade3 Truck transp. Wholesale trade Retail trade Retail trade4 Real estate Food srvcs. & drinking places Nonresidential manuf. structures Mgmt., scientific, & technical consulting srvcs.
40Internet Appendix Table V - Continued
Rank Supplier Network Customer Network Merger Network: Primary Industry Merger Network: All Industries
5 Elec. power generation, transmission, & distribution Motor vehicle parts manuf. Funds, trusts, & other financial vehicles Mgmt. of companies & enterprises6 Monetary auth. & depository credit intermed. Telecommunications Software, audio, & video media reproducing Nonresidential manuf. structures7 Lessors of nonfinancial intangible assets Mgmt. of companies & enterprises Internet service providers & web search portals Funds, trusts, & other financial vehicles8 Iron & steel mills & ferroalloy manuf. Hospitals Mgmt., scientific, & technical consulting srvcs. All other chemical product & preparation manuf.9 Other plastics product manuf. Offices of physicians, dentists, & other health prctnrs. Motor vehicle parts manuf. Architectural, engineering, & related srvcs.
10 Paperboard container manuf. Other plastics product manuf. Business support srvcs. Motor vehicle parts manuf.11 Telecommunications Printing Architectural, engineering, & related srvcs. Scientific rsrch. & development srvcs.12 Employment srvcs. Iron & steel mills & ferroalloy manuf. Software publishers All other misc. manuf.13 Semiconductor & related device manuf. Light truck & utility vehicle manuf. Scientific rsrch. & development srvcs. Software publishers14 Plastics pkg. mats. & unlaminated film & sheet manuf. Semiconductor & related device manuf. Real estate Other plastics product manuf.15 Scientific rsrch. & development srvcs. Other basic organic chemical manuf. Other plastics product manuf. Surgical appliance & supplies manuf.
Panel B: Summary-Level
IO Report Year: 1982
1 Wholesale & retail trade Construction Banking & insurance Wholesale & retail trade2 Business srvcs. Wholesale & retail trade Wholesale & retail trade Banking & insurance3 Rail, water, & motor transp. Rail, water, & motor transp. Business srvcs. Business srvcs.4 Private utilities Chemicals Chemicals Construction5 Petroleum Rubber & plastics Construction Chemicals6 Rubber & plastics Iron & steel Manufacturing, misc. Instruments7 Banking & insurance Business srvcs. Rail, water, & motor transp. Manufacturing, misc.8 Primary metals Doctors, private hospitals, schools Instruments Drugs9 Chemicals Primary metals Rubber & plastics Rail, water, & motor transp.
10 Fabricated metal products Metal stamping Semiconductors Private utilities11 Iron & steel Fabricated metal products Aircraft Food preparations12 Construction Food preparations Drugs Publishing & printing13 Metal stamping Carburetors, pistons, rings, valves Nonmetallic minerals & concrete Rubber & plastics14 General industrial mach. Motor vehicles & car bodies Fabricated metal products Nonmetallic minerals & concrete15 Turbines & engines General industrial mach. Radio & TV equip. Primary metals
IO Report Year: 1987
1 Wholesale trade Construction Wholesale trade Wholesale trade2 Other business & prof. srvcs., ex. medical Other business & prof. srvcs., ex. medical Finance Retail trade, ex. eating & drinking3 Elec. srvcs. (utilities) Primary iron & steel manuf. Other business & prof. srvcs., ex. medical Finance4 Rubber & misc. plastics products Wholesale trade Retail trade, ex. eating & drinking Computer & data proc. srvcs.5 Motor freight transp. & warehousing Rubber & misc. plastics products Scientific & controlling instruments Other business & prof. srvcs., ex. medical6 Construction Paper & allied products, ex. containers Computer & data proc. srvcs. Construction7 Other fabricated metal products Industrial inorganic & organic chems. Construction Other printing & publishing8 Finance Motor vehicles & passenger car bodies Rubber & misc. plastics products Insurance9 Primary nonferrous metals manuf. Automotive repair & srvcs. Miscellaneous manuf. Apparel
10 Industrial inorganic & organic chems. Health srvcs. Apparel Industrial inorganic & organic chems.11 Legal, engineering, acctng., & related srvcs. Food & kindred products Industrial inorganic & organic chems. Rubber & misc. plastics products12 Primary iron & steel manuf. Primary nonferrous metals manuf. Aircraft & parts Scientific & controlling instruments13 Real estate & royalties Retail trade, ex. eating & drinking Motor freight transp. & warehousing Miscellaneous manuf.14 Automotive repair & srvcs. Finance Crude petroleum & natural gas Primary nonferrous metals manuf.15 Other printing & publishing Glass & glass products Primary nonferrous metals manuf. Motor freight transp. & warehousing
IO Report Year: 1992
1 Other business & prof. srvcs., ex. medical Construction Wholesale trade Wholesale trade2 Wholesale trade Wholesale trade Finance Finance3 Motor freight transp. & warehousing Other business & prof. srvcs., ex. medical Other business & prof. srvcs., ex. medical Computer & data proc. srvcs.4 Rubber & misc. plastics products Health srvcs. Scientific & controlling instruments Other business & prof. srvcs., ex. medical5 Elec. srvcs. (utilities) Rubber & misc. plastics products Computer & data proc. srvcs. Retail trade, ex. eating & drinking6 Legal, engineering, acctng., & related srvcs. Primary iron & steel manuf. Construction Insurance7 Construction Industrial inorganic & organic chems. Miscellaneous manuf. Construction8 Real estate & royalties Retail trade, ex. eating & drinking Retail trade, ex. eating & drinking Rubber & misc. plastics products9 Industrial inorganic & organic chems. Food & kindred products Industrial inorganic & organic chems. Scientific & controlling instruments
10 Finance Communications, ex. radio & TV Rubber & misc. plastics products Miscellaneous manuf.11 Primary iron & steel manuf. Primary nonferrous metals manuf. Other fabricated metal products Industrial inorganic & organic chems.12 Primary nonferrous metals manuf. Automotive repair & srvcs. Aircraft & parts Crude petroleum & natural gas13 Other fabricated metal products Paper & allied products, ex. containers Crude petroleum & natural gas Food & kindred products14 Screw machine products & stampings Other printing & publishing General industrial mach. & equip. Other printing & publishing15 Other printing & publishing Real estate & royalties Electronic components & accessories Legal, engineering, acctng., & related srvcs.
IO Report Year: 1997
1 Nondepository credit intermed. & related acts. Wholesale trade Securities, commodity contracts, invmts. Wholesale trade2 Wholesale trade Construction Wholesale trade Retail trade3 Mgmt. of companies & enterprises Retail trade Funds, trusts, & other financial vehicles Securities, commodity contracts, invmts.4 Truck transp. Mgmt. of companies & enterprises Retail trade Funds, trusts, & other financial vehicles5 Real estate Real estate All other administrative & support srvcs. Mgmt. & technical consulting srvcs.6 Semiconductor & electronic equip. manuf. Motor vehicle body, trailer, & parts manuf. Construction Mgmt. of companies & enterprises7 Plastics & rubber products manuf. Plastics & rubber products manuf. Electronic equip. manuf. Construction8 Other fabricated metal product manuf. Food manuf. Mgmt. & technical consulting srvcs. Information srvcs.9 All other administrative & support srvcs. Nondepository credit intermed. & related acts. Commercial & service industry mach. All other administrative & support srvcs.
10 Telecommunications Other fabricated metal product manuf. Magnetic media manuf. & reproducing Software publishers11 Power generation & supply Food srvcs. & drinking places Nondepository credit intermed. & related acts. Magnetic media manuf. & reproducing12 Iron & steel mills & manuf. from purchased steel Basic chemical manuf. Computer systems design & related srvcs. Newspaper, book, & directory publishers13 Employment srvcs. Iron & steel mills & manuf. from purchased steel Real estate Nondepository credit intermed. & related acts.14 Petroleum & coal products manuf. Telecommunications Other general purpose mach. manuf. Food manuf.15 Printing & related support acts. Semiconductor & electronic equip. manuf. Information srvcs. Plastics & rubber products manuf.
IO Report Year: 2002
1 Wholesale trade Construction Securities, commodity contracts, invmts., & related acts. Retail trade2 Mgmt. of companies & enterprises Retail trade Wholesale trade Wholesale trade3 Monetary auth., credit intermed. & related acts. Wholesale trade Retail trade Securities, commodity contracts, invmts., & related acts.4 Real estate Mgmt. of companies & enterprises Funds, trusts, & other financial vehicles Funds, trusts, & other financial vehicles5 All other administrative & support srvcs. Food srvcs. & drinking places All other administrative & support srvcs. Mgmt., scientific, & technical consulting srvcs.6 Other fabricated metal product manuf. All other administrative & support srvcs. Construction Mgmt. of companies & enterprises7 Truck transp. Plastics & rubber products manuf. Electronic instrument manuf. Construction8 Plastics & rubber products manuf. Motor vehicle body, trailer, & parts manuf. Internet service providers, web search portals, & data proc. Software publishers9 Semiconductor & other electronic component manuf. Other fabricated metal product manuf. Mgmt., scientific, & technical consulting srvcs. Internet service providers, web search portals, & data proc.
10 Elec. power generation, transmission, & distribution Telecommunications Manufacturing & reproducing magnetic & optical media Monetary auth., credit intermed. & related acts.11 Telecommunications Ambulatory health care srvcs. Real estate All other administrative & support srvcs.12 Employment srvcs. Food manuf. Commercial & service industry mach. manuf. Manufacturing & reproducing magnetic & optical media13 Lessors of nonfinancial intangible assets Monetary auth., credit intermed. & related acts. Audio, video, & communications equip. manuf. Insurance carriers & related acts.14 Insurance carriers & related acts. Mgmt., scientific, & technical consulting srvcs. Computer systems design & related srvcs. Electronic instrument manuf.15 Mgmt., scientific, & technical consulting srvcs. Newspaper, periodical, book, & directory publishers Medical equip. & supplies manuf. Scientific rsrch. & development srvcs.
41
Internet Appendix Table VICorrelations Between Customer and Merger Networks by IO Report Year, Industry Assignment, and IndustryCoarsenessThis table presents correlation coefficients and p−values (below each coefficient entry) across industry-level network statistics for a sample ofmergers over the period 1986 to 2010. Industry characteristics of the count-based merger network are on rows and industry characteristics ofthe customer network are on columns. Merger data is from SDC. Industries are defined by the Bureau of Economic Analysis Input-Output(IO) Industry classification. The IO industry definitions are based on input-output relations between industries as recorded by the BEA infive separate reports (1982, 1987, 1992, 1997, and 2002), for two levels of industry definitions (Detailed and Summary). Panel A presents thedetailed-level data and panel B presents the summary-level data. Merging firms are assigned to IO industries in four ways to create four differentmerger networks. In the ‘Primary Industry’ network, firms are assigned to an industry based on its primary SIC/NAICS code as reported inSDC. The ‘All Industries’ network assigns firms using all industry codes reported by SDC. The ‘HV1%’ and ‘HV5%’ assign firms based on thefollowing priority. If there are any shared IO codes between merging firms, we assign the merger equally to those industries. If there are nohorizontal matches, but a vertical relation of 1% or 5% for any of the four vertical relations (acquirer sells to target, acquirer buys from target,and vice versa), we assign the merger activity to those industries. If neither horizontal or vertical relations exist, we assign the merger activityequally to the unrelated industry pairs. ‘Degree Centrality’, ‘Eigenvector Centrality’, ‘Average Shortest Path’, and ‘Clustering Coefficient’ aredefined in the Internet Appendix. Significance is indicated at 1%, 5%, and 10% levels by a, b, and c.
Customer Network
Primary Industry All Industries HV1% HV5%
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Panel A: Detail-Level
IO Report Year: 1982
Degree Centrality 0.42a 0.39a −0.33a 0.00 0.48a 0.52a −0.47a 0.00 0.45a 0.44a −0.35a 0.00 0.42a 0.42a −0.36a −0.01
< 0.01 < 0.01 < 0.01 0.93 < 0.01 < 0.01 < 0.01 0.95 < 0.01 < 0.01 < 0.01 0.97 < 0.01 < 0.01 < 0.01 0.86
Eigenvector Centrality 0.42a 0.40a −0.34a −0.01 0.48a 0.53a −0.48a 0.00 0.46a 0.45a −0.36a 0.00 0.42a 0.43a −0.37a −0.01
< 0.01 < 0.01 < 0.01 0.91 < 0.01 < 0.01 < 0.01 0.96 < 0.01 < 0.01 < 0.01 0.96 < 0.01 < 0.01 < 0.01 0.83
Average Path −0.30a −0.33a 0.35a 0.01 −0.48a −0.52a 0.48a −0.01 −0.35a −0.35a 0.35a −0.03 −0.36a −0.37a 0.33a 0.00
< 0.01 < 0.01 < 0.01 0.91 < 0.01 < 0.01 < 0.01 0.89 < 0.01 < 0.01 < 0.01 0.49 < 0.01 < 0.01 < 0.01 0.93
Clustering Coefficient −0.22a −0.23a 0.16a 0.01 −0.39a −0.35a 0.41a −0.02 −0.42a −0.42a 0.33a 0.02 −0.39a −0.38a 0.36a 0.03
< 0.01 < 0.01 < 0.01 0.90 < 0.01 < 0.01 < 0.01 0.67 < 0.01 < 0.01 < 0.01 0.74 < 0.01 < 0.01 < 0.01 0.53
IO Report Year: 1987
Degree Centrality 0.46a 0.50a −0.31a 0.00 0.47a 0.54a −0.46a 0.02 0.49a 0.54a −0.32a −0.01 0.45a 0.51a −0.34a 0.00
< 0.01 < 0.01 < 0.01 0.94 < 0.01 < 0.01 < 0.01 0.73 < 0.01 < 0.01 < 0.01 0.80 < 0.01 < 0.01 < 0.01 0.97
Eigenvector Centrality 0.47a 0.51a −0.32a 0.00 0.47a 0.54a −0.47a 0.02 0.49a 0.55a −0.34a −0.01 0.45a 0.52a −0.35a 0.00
< 0.01 < 0.01 < 0.01 0.93 < 0.01 < 0.01 < 0.01 0.69 < 0.01 < 0.01 < 0.01 0.83 < 0.01 < 0.01 < 0.01 0.98
Average Path −0.32a −0.37a 0.32a 0.00 −0.47a −0.54a 0.46a −0.02 −0.39a −0.44a 0.31a −0.04 −0.37a −0.43a 0.30a −0.02
< 0.01 < 0.01 < 0.01 0.98 < 0.01 < 0.01 < 0.01 0.71 < 0.01 < 0.01 < 0.01 0.36 < 0.01 < 0.01 < 0.01 0.71
Clustering Coefficient −0.24a −0.23a 0.20a −0.02 −0.37a −0.38a 0.38a 0.01 −0.41a −0.42a 0.28a 0.10b −0.37a −0.39a 0.28a 0.05
< 0.01 < 0.01 < 0.01 0.62 < 0.01 < 0.01 < 0.01 0.82 < 0.01 < 0.01 < 0.01 0.03 < 0.01 < 0.01 < 0.01 0.28
IO Report Year: 1992
Degree Centrality 0.45a 0.50a −0.28a 0.07 0.49a 0.57a −0.39a 0.09c 0.49a 0.55a −0.30a 0.08c 0.46a 0.53a −0.30a 0.08c
42Internet Appendix Table VI - Continued
Customer Network
Primary Industry All Industries HV1% HV5%
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
< 0.01 < 0.01 < 0.01 0.11 < 0.01 < 0.01 < 0.01 0.06 < 0.01 < 0.01 < 0.01 0.09 < 0.01 < 0.01 < 0.01 0.09
Eigenvector Centrality 0.46a 0.50a −0.29a 0.07 0.49a 0.57a −0.40a 0.09c 0.50a 0.56a −0.31a 0.08c 0.46a 0.54a −0.31a 0.08c
< 0.01 < 0.01 < 0.01 0.11 < 0.01 < 0.01 < 0.01 0.06 < 0.01 < 0.01 < 0.01 0.08 < 0.01 < 0.01 < 0.01 0.09
Average Path −0.31a −0.35a 0.26a −0.09b −0.49a −0.57a 0.40a −0.09b −0.39a −0.43a 0.25a −0.10b −0.39a −0.45a 0.28a −0.10b
< 0.01 < 0.01 < 0.01 0.04 < 0.01 < 0.01 < 0.01 0.04 < 0.01 < 0.01 < 0.01 0.02 < 0.01 < 0.01 < 0.01 0.04
Clustering Coefficient −0.23a −0.25a 0.17a −0.13a −0.39a −0.41a 0.34a −0.09b −0.42a −0.44a 0.29a −0.03 −0.36a −0.38a 0.24a −0.05
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.04 < 0.01 < 0.01 < 0.01 0.54 < 0.01 < 0.01 < 0.01 0.24
IO Report Year: 1997
Degree Centrality 0.53a 0.53a −0.27a −0.13a 0.44a 0.49a −0.37a −0.13a 0.52a 0.54a −0.29a −0.12b 0.45a 0.49a −0.29a −0.13a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Eigenvector Centrality 0.53a 0.53a −0.27a −0.13a 0.43a 0.48a −0.37a −0.13a 0.51a 0.54a −0.29a −0.12a 0.44a 0.48a −0.29a −0.13a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Average Path −0.30a −0.32a 0.24a 0.08c −0.44a −0.49a 0.37a 0.13a −0.43a −0.46a 0.27a 0.08c −0.41a −0.46a 0.28a 0.11b
< 0.01 < 0.01 < 0.01 0.10 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.09 < 0.01 < 0.01 < 0.01 0.02
Clustering Coefficient −0.21a −0.19a 0.18a 0.01 −0.36a −0.37a 0.37a 0.06 −0.38a −0.38a 0.30a 0.10b −0.33a −0.35a 0.34a 0.05
< 0.01 < 0.01 < 0.01 0.86 < 0.01 < 0.01 < 0.01 0.17 < 0.01 < 0.01 < 0.01 0.03 < 0.01 < 0.01 < 0.01 0.27
IO Report Year: 2002
Degree Centrality 0.53a 0.54a −0.26a −0.03 0.44a 0.49a −0.38a 0.02 0.54a 0.57a −0.30a −0.05 0.45a 0.51a −0.30a 0.00
< 0.01 < 0.01 < 0.01 0.61 < 0.01 < 0.01 < 0.01 0.75 < 0.01 < 0.01 < 0.01 0.33 < 0.01 < 0.01 < 0.01 0.95
Eigenvector Centrality 0.52a 0.54a −0.26a −0.02 0.42a 0.48a −0.38a 0.02 0.53a 0.57a −0.30a −0.05 0.44a 0.50a −0.30a 0.00
< 0.01 < 0.01 < 0.01 0.65 < 0.01 < 0.01 < 0.01 0.69 < 0.01 < 0.01 < 0.01 0.36 < 0.01 < 0.01 < 0.01 1.00
Average Path −0.34a −0.36a 0.26a 0.00 −0.44a −0.49a 0.38a −0.02 −0.46a −0.50a 0.28a 0.05 −0.41a −0.46a 0.28a 0.01
< 0.01 < 0.01 < 0.01 1.00 < 0.01 < 0.01 < 0.01 0.71 < 0.01 < 0.01 < 0.01 0.36 < 0.01 < 0.01 < 0.01 0.85
Clustering Coefficient −0.30a −0.30a 0.29a 0.05 −0.40a −0.42a 0.39a −0.03 −0.39a −0.41a 0.31a 0.06 −0.36a −0.38a 0.32a −0.02
< 0.01 < 0.01 < 0.01 0.31 < 0.01 < 0.01 < 0.01 0.51 < 0.01 < 0.01 < 0.01 0.25 < 0.01 < 0.01 < 0.01 0.70
Panel B: Summary-Level
IO Report Year: 1982
Degree Centrality 0.53a 0.55a −0.54a −0.12 0.43a 0.45a −0.51a −0.22c 0.66a 0.69a −0.64a −0.18 0.51a 0.53a −0.50a −0.11
< 0.01 < 0.01 < 0.01 0.31 < 0.01 < 0.01 < 0.01 0.05 < 0.01 < 0.01 < 0.01 0.13 < 0.01 < 0.01 < 0.01 0.35
Eigenvector Centrality 0.53a 0.55a −0.54a −0.11 0.43a 0.44a −0.51a −0.22c 0.65a 0.68a −0.65a −0.17 0.50a 0.52a −0.50a −0.11
< 0.01 < 0.01 < 0.01 0.33 < 0.01 < 0.01 < 0.01 0.05 < 0.01 < 0.01 < 0.01 0.13 < 0.01 < 0.01 < 0.01 0.36
Average Path −0.53a −0.55a 0.54a 0.12 −0.43a −0.45a 0.51a 0.22c −0.66a −0.69a 0.64a 0.18 −0.51a −0.53a 0.50a 0.11
< 0.01 < 0.01 < 0.01 0.31 < 0.01 < 0.01 < 0.01 0.05 < 0.01 < 0.01 < 0.01 0.13 < 0.01 < 0.01 < 0.01 0.35
Clustering Coefficient −0.48a −0.50a 0.49a 0.18 −0.42a −0.44a 0.40a 0.17 −0.59a −0.60a 0.55a 0.33a −0.49a −0.49a 0.44a 0.19
< 0.01 < 0.01 < 0.01 0.12 < 0.01 < 0.01 < 0.01 0.14 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.10
IO Report Year: 1987
Degree Centrality 0.45a 0.46a −0.35a 0.00 0.29a 0.29a −0.31a −0.08 0.59a 0.62a −0.51a −0.08 0.46a 0.49a −0.37a 0.01
< 0.01 < 0.01 < 0.01 0.99 < 0.01 < 0.01 < 0.01 0.47 < 0.01 < 0.01 < 0.01 0.48 < 0.01 < 0.01 < 0.01 0.95
Eigenvector Centrality 0.44a 0.45a −0.34a 0.01 0.28a 0.28a −0.31a −0.08 0.59a 0.61a −0.50a −0.07 0.45a 0.48a −0.37a 0.01
< 0.01 < 0.01 < 0.01 0.93 < 0.01 < 0.01 < 0.01 0.45 < 0.01 < 0.01 < 0.01 0.53 < 0.01 < 0.01 < 0.01 0.91
Average Path −0.45a −0.46a 0.35a 0.00 −0.29a −0.29a 0.31a 0.08 −0.59a −0.62a 0.51a 0.08 −0.46a −0.49a 0.37a −0.01
43
Internet Appendix Table VI - Continued
Customer Network
Primary Industry All Industries HV1% HV5%
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
< 0.01 < 0.01 < 0.01 0.99 < 0.01 < 0.01 < 0.01 0.47 < 0.01 < 0.01 < 0.01 0.48 < 0.01 < 0.01 < 0.01 0.95
Clustering Coefficient −0.41a −0.40a 0.31a 0.14 −0.34a −0.36a 0.26b 0.02 −0.58a −0.58a 0.54a 0.25b −0.44a −0.45a 0.35a 0.15
< 0.01 < 0.01 < 0.01 0.20 < 0.01 < 0.01 0.02 0.83 < 0.01 < 0.01 < 0.01 0.02 < 0.01 < 0.01 < 0.01 0.18
IO Report Year: 1992
Degree Centrality 0.52a 0.51a −0.44a 0.05 0.32a 0.30a −0.30a 0.13 0.66a 0.66a −0.57a 0.02 0.46a 0.47a −0.39a 0.16
< 0.01 < 0.01 < 0.01 0.61 < 0.01 < 0.01 < 0.01 0.22 < 0.01 < 0.01 < 0.01 0.87 < 0.01 < 0.01 < 0.01 0.14
Eigenvector Centrality 0.51a 0.50a −0.44a 0.07 0.31a 0.29a −0.29a 0.13 0.65a 0.66a −0.57a 0.03 0.45a 0.46a −0.39a 0.17
< 0.01 < 0.01 < 0.01 0.54 < 0.01 < 0.01 < 0.01 0.23 < 0.01 < 0.01 < 0.01 0.78 < 0.01 < 0.01 < 0.01 0.12
Average Path −0.52a −0.51a 0.44a −0.05 −0.32a −0.30a 0.30a −0.13 −0.66a −0.66a 0.57a −0.02 −0.46a −0.47a 0.39a −0.16
< 0.01 < 0.01 < 0.01 0.61 < 0.01 < 0.01 < 0.01 0.22 < 0.01 < 0.01 < 0.01 0.87 < 0.01 < 0.01 < 0.01 0.14
Clustering Coefficient −0.43a −0.41a 0.36a 0.09 −0.37a −0.41a 0.25b −0.10 −0.58a −0.55a 0.52a 0.09 −0.46a −0.48a 0.40a −0.04
< 0.01 < 0.01 < 0.01 0.41 < 0.01 < 0.01 0.02 0.36 < 0.01 < 0.01 < 0.01 0.41 < 0.01 < 0.01 < 0.01 0.73
IO Report Year: 1997
Degree Centrality 0.46a 0.45a −0.37a −0.01 0.37a 0.38a −0.40a 0.15c 0.49a 0.50a −0.39a −0.02 0.39a 0.41a −0.34a 0.06
< 0.01 < 0.01 < 0.01 0.87 < 0.01 < 0.01 < 0.01 0.10 < 0.01 < 0.01 < 0.01 0.82 < 0.01 < 0.01 < 0.01 0.51
Eigenvector Centrality 0.46a 0.45a −0.37a −0.01 0.36a 0.38a −0.40a 0.16c 0.49a 0.50a −0.38a −0.01 0.39a 0.40a −0.34a 0.07
< 0.01 < 0.01 < 0.01 0.91 < 0.01 < 0.01 < 0.01 0.07 < 0.01 < 0.01 < 0.01 0.92 < 0.01 < 0.01 < 0.01 0.46
Average Path −0.46a −0.45a 0.37a 0.01 −0.37a −0.38a 0.40a −0.15c −0.49a −0.50a 0.39a 0.02 −0.39a −0.41a 0.34a −0.06
< 0.01 < 0.01 < 0.01 0.89 < 0.01 < 0.01 < 0.01 0.09 < 0.01 < 0.01 < 0.01 0.82 < 0.01 < 0.01 < 0.01 0.50
Clustering Coefficient −0.38a −0.37a 0.33a 0.02 −0.38a −0.41a 0.34a 0.01 −0.51a −0.51a 0.42a 0.05 −0.35a −0.36a 0.27a −0.06
< 0.01 < 0.01 < 0.01 0.87 < 0.01 < 0.01 < 0.01 0.89 < 0.01 < 0.01 < 0.01 0.56 < 0.01 < 0.01 < 0.01 0.52
IO Report Year: 2002
Degree Centrality 0.53a 0.52a −0.44a −0.10 0.43a 0.43a −0.44a −0.09 0.60a 0.61a −0.47a −0.10 0.48a 0.49a −0.42a −0.10
< 0.01 < 0.01 < 0.01 0.28 < 0.01 < 0.01 < 0.01 0.34 < 0.01 < 0.01 < 0.01 0.27 < 0.01 < 0.01 < 0.01 0.29
Eigenvector Centrality 0.52a 0.52a −0.45a −0.10 0.42a 0.42a −0.44a −0.08 0.59a 0.61a −0.48a −0.10 0.48a 0.49a −0.43a −0.10
< 0.01 < 0.01 < 0.01 0.27 < 0.01 < 0.01 < 0.01 0.36 < 0.01 < 0.01 < 0.01 0.29 < 0.01 < 0.01 < 0.01 0.29
Average Path −0.53a −0.52a 0.44a 0.10 −0.43a −0.43a 0.44a 0.09 −0.60a −0.61a 0.47a 0.10 −0.48a −0.49a 0.42a 0.10
< 0.01 < 0.01 < 0.01 0.28 < 0.01 < 0.01 < 0.01 0.34 < 0.01 < 0.01 < 0.01 0.27 < 0.01 < 0.01 < 0.01 0.29
Clustering Coefficient −0.41a −0.41a 0.32a 0.10 −0.45a −0.49a 0.37a 0.12 −0.57a −0.55a 0.44a 0.17c −0.42a −0.42a 0.34a 0.05
< 0.01 < 0.01 < 0.01 0.28 < 0.01 < 0.01 < 0.01 0.20 < 0.01 < 0.01 < 0.01 0.06 < 0.01 < 0.01 < 0.01 0.56
44Internet Appendix Table VII
Correlations Between Supplier and Merger Networks by IO Report Year, Industry Assignment, and Industry Coarse-nessThis table presents correlation coefficients and p−values (below each coefficient entry) across industry-level network statistics for a sample ofmergers over the period 1986 to 2010. Industry characteristics of the count-based merger network are on rows and industry characteristics ofthe supplier network are on columns. Merger data is from SDC. Industries are defined by the Bureau of Economic Analysis Input-Output (IO)Industry classification. The IO industry definitions are based on input-output relations between industries as recorded by the BEA in five separatereports (1982, 1987, 1992, 1997, and 2002), for two levels of industry definitions (Detailed and Summary). Panel A presents the detailed-leveldata and panel B presents the summary-level data. Merging firms are assigned to IO industries in four ways to create four different mergernetworks. In the ‘Primary Industry’ network, firms are assigned to an industry based on its primary SIC/NAICS code as reported in SDC. The‘All Industries’ network assigns firms using all industry codes reported by SDC. The ‘HV1%’ and ‘HV5%’ assign firms based on the followingpriority. If there are any shared IO codes between merging firms, we assign the merger equally to those industries. If there are no horizontalmatches, but a vertical relation of 1% or 5% for any of the four vertical relations (acquirer sells to target, acquirer buys from target, and viceversa), we assign the merger activity to those industries. If neither horizontal or vertical relations exist, we assign the merger activity equally tothe unrelated industry pairs. ‘Degree Centrality’, ‘Eigenvector Centrality’, ‘Average Shortest Path’, and ‘Clustering Coefficient’ are defined inthe Internet Appendix. Significance is indicated at 1%, 5%, and 10% levels by a, b, and c.
Supplier Network
Primary Industry All Industries HV1% HV5%
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Panel A: Detail-Level
IO Report Year: 1982
Degree Centrality 0.40a 0.34a −0.18a −0.30a 0.33a 0.31a −0.19a −0.36a 0.44a 0.38a −0.17a −0.34a 0.35a 0.29a −0.12b −0.34a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.01 < 0.01
Eigenvector Centrality 0.40a 0.34a −0.18a −0.31a 0.33a 0.30a −0.19a −0.36a 0.44a 0.37a −0.18a −0.35a 0.35a 0.29a −0.12a −0.34a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Average Path −0.23a −0.20a 0.11b 0.27a −0.33a −0.31a 0.19a 0.36a −0.32a −0.28a 0.14a 0.29a −0.30a −0.25a 0.12a 0.33a
< 0.01 < 0.01 0.02 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Clustering Coefficient −0.18a −0.19a 0.15a 0.17a −0.29a −0.29a 0.21a 0.26a −0.31a −0.30a 0.19a 0.32a −0.27a −0.25a 0.16a 0.28a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
IO Report Year: 1987
Degree Centrality 0.43a 0.37a −0.18a −0.23a 0.37a 0.35a −0.20a −0.28a 0.48a 0.42a −0.22a −0.27a 0.38a 0.32a −0.15a −0.25a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Eigenvector Centrality 0.43a 0.37a −0.18a −0.24a 0.36a 0.35a −0.20a −0.28a 0.48a 0.42a −0.22a −0.28a 0.38a 0.32a −0.15a −0.25a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Average Path −0.26a −0.22a 0.12a 0.21a −0.37a −0.35a 0.22a 0.28a −0.37a −0.33a 0.20a 0.22a −0.31a −0.27a 0.16a 0.22a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Clustering Coefficient −0.19a −0.20a 0.13a 0.12a −0.32a −0.30a 0.19a 0.25a −0.35a −0.32a 0.19a 0.32a −0.28a −0.26a 0.12a 0.25a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
IO Report Year: 1992
Degree Centrality 0.44a 0.38a −0.21a −0.25a 0.38a 0.35a −0.23a −0.35a 0.49a 0.42a −0.23a −0.29a 0.40a 0.33a −0.17a −0.27a
45
Internet Appendix Table VII - Continued
Supplier Network
Primary Industry All Industries HV1% HV5%
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Eigenvector Centrality 0.44a 0.38a −0.21a −0.25a 0.37a 0.35a −0.23a −0.35a 0.49a 0.42a −0.23a −0.29a 0.39a 0.33a −0.17a −0.28a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Average Path −0.24a −0.21a 0.14a 0.21a −0.38a −0.35a 0.25a 0.36a −0.37a −0.33a 0.25a 0.24a −0.32a −0.27a 0.17a 0.24a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Clustering Coefficient −0.21a −0.21a 0.12b 0.19a −0.31a −0.30a 0.22a 0.31a −0.33a −0.29a 0.14a 0.34a −0.28a −0.26a 0.16a 0.30a
< 0.01 < 0.01 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
IO Report Year: 1997
Degree Centrality 0.39a 0.32a −0.16a −0.33a 0.36a 0.34a −0.24a −0.35a 0.50a 0.42a −0.28a −0.34a 0.38a 0.31a −0.20a −0.32a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Eigenvector Centrality 0.39a 0.32a −0.16a −0.33a 0.35a 0.33a −0.23a −0.35a 0.49a 0.41a −0.28a −0.34a 0.37a 0.31a −0.19a −0.32a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Average Path −0.23a −0.19a 0.11b 0.23a −0.36a −0.34a 0.24a 0.35a −0.41a −0.35a 0.23a 0.26a −0.35a −0.29a 0.18a 0.29a
< 0.01 < 0.01 0.02 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Clustering Coefficient −0.15a −0.19a 0.14a 0.16a −0.32a −0.32a 0.25a 0.32a −0.34a −0.32a 0.26a 0.38a −0.29a −0.29a 0.22a 0.31a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
IO Report Year: 2002
Degree Centrality 0.42a 0.40a −0.33a −0.29a 0.35a 0.36a −0.30a −0.34a 0.56a 0.53a −0.43a −0.33a 0.41a 0.39a −0.31a −0.28a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Eigenvector Centrality 0.42a 0.40a −0.33a −0.29a 0.34a 0.35a −0.28a −0.33a 0.54a 0.52a −0.42a −0.33a 0.40a 0.38a −0.30a −0.27a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Average Path −0.27a −0.26a 0.21a 0.21a −0.35a −0.36a 0.30a 0.33a −0.48a −0.46a 0.38a 0.30a −0.37a −0.36a 0.29a 0.25a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Clustering Coefficient −0.22a −0.24a 0.20a 0.31a −0.33a −0.34a 0.28a 0.31a −0.37a −0.36a 0.29a 0.37a −0.33a −0.33a 0.27a 0.32a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Panel B: Summary-Level
IO Report Year: 1982
Degree Centrality 0.58a 0.60a −0.58a −0.49a 0.43a 0.47a −0.43a −0.40a 0.78a 0.78a −0.78a −0.68a 0.55a 0.57a −0.55a −0.46a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Eigenvector Centrality 0.56a 0.58a −0.56a −0.47a 0.41a 0.45a −0.41a −0.38a 0.76a 0.77a −0.76a −0.67a 0.53a 0.56a −0.53a −0.44a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Average Path −0.58a −0.60a 0.58a 0.49a −0.43a −0.47a 0.43a 0.40a −0.78a −0.78a 0.78a 0.68a −0.55a −0.57a 0.55a 0.46a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Clustering Coefficient −0.56a −0.56a 0.56a 0.57a −0.59a −0.57a 0.59a 0.59a −0.65a −0.64a 0.65a 0.66a −0.57a −0.56a 0.57a 0.57a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
IO Report Year: 1987
Degree Centrality 0.50a 0.51a −0.49a −0.23b 0.34a 0.37a −0.34a −0.12 0.66a 0.65a −0.65a −0.41a 0.46a 0.47a −0.45a −0.20c
< 0.01 < 0.01 < 0.01 0.03 < 0.01 < 0.01 < 0.01 0.29 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.07
Eigenvector Centrality 0.49a 0.51a −0.48a −0.22b 0.32a 0.36a −0.33a −0.10 0.64a 0.64a −0.63a −0.39a 0.45a 0.47a −0.44a −0.18c
< 0.01 < 0.01 < 0.01 0.04 < 0.01 < 0.01 < 0.01 0.34 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.09
Average Path −0.50a −0.51a 0.49a 0.23b −0.34a −0.37a 0.34a 0.12 −0.66a −0.65a 0.65a 0.41a −0.46a −0.47a 0.45a 0.20c
46Internet Appendix Table VII - Continued
Supplier Network
Primary Industry All Industries HV1% HV5%
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
Degre
e
Centrality
Eigenvecto
r
Centrality
Avera
ge
Path
Clu
stering
Coeffi
cient
< 0.01 < 0.01 < 0.01 0.03 < 0.01 < 0.01 < 0.01 0.29 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.07
Clustering Coefficient −0.45a −0.44a 0.45a 0.31a −0.44a −0.42a 0.44a 0.28a −0.61a −0.58a 0.60a 0.52a −0.47a −0.45a 0.47a 0.33a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
IO Report Year: 1992
Degree Centrality 0.54a 0.53a −0.54a −0.36a 0.32a 0.34a −0.32a −0.14 0.69a 0.66a −0.69a −0.50a 0.48a 0.47a −0.48a −0.27a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.19 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Eigenvector Centrality 0.52a 0.52a −0.52a −0.35a 0.31a 0.33a −0.31a −0.13 0.67a 0.65a −0.67a −0.48a 0.47a 0.46a −0.47a −0.26b
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.24 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.02
Average Path −0.54a −0.53a 0.54a 0.36a −0.32a −0.34a 0.32a 0.14 −0.69a −0.66a 0.69a 0.50a −0.48a −0.47a 0.48a 0.27a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.19 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
Clustering Coefficient −0.49a −0.47a 0.49a 0.40a −0.44a −0.40a 0.44a 0.31a −0.61a −0.58a 0.61a 0.49a −0.48a −0.45a 0.48a 0.34a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
IO Report Year: 1997
Degree Centrality 0.39a 0.40a −0.41a −0.24a 0.33a 0.38a −0.36a −0.14 0.50a 0.48a −0.50a −0.32a 0.35a 0.36a −0.37a −0.19b
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.13 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.04
Eigenvector Centrality 0.39a 0.40a −0.41a −0.23a 0.32a 0.37a −0.36a −0.12 0.49a 0.47a −0.50a −0.31a 0.35a 0.36a −0.36a −0.18b
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.18 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.05
Average Path −0.39a −0.40a 0.41a 0.24a −0.33a −0.37a 0.36a 0.14 −0.50a −0.48a 0.50a 0.32a −0.35a −0.36a 0.37a 0.19b
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.13 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.04
Clustering Coefficient −0.35a −0.35a 0.35a 0.26a −0.34a −0.28a 0.35a 0.31a −0.49a −0.45a 0.48a 0.40a −0.33a −0.30a 0.33a 0.25a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
IO Report Year: 2002
Degree Centrality 0.42a 0.42a −0.38a −0.26a 0.34a 0.36a −0.32a −0.19b 0.57a 0.58a −0.52a −0.35a 0.40a 0.42a −0.36a −0.20b
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.03 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.02
Eigenvector Centrality 0.41a 0.42a −0.37a −0.26a 0.33a 0.35a −0.31a −0.18b 0.56a 0.57a −0.51a −0.33a 0.39a 0.41a −0.36a −0.19b
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.04 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.03
Average Path −0.42a −0.42a 0.38a 0.26a −0.34a −0.36a 0.32a 0.19b −0.57a −0.58a 0.52a 0.35a −0.40a −0.42a 0.36a 0.20b
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.03 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.02
Clustering Coefficient −0.36a −0.34a 0.32a 0.30a −0.38a −0.39a 0.35a 0.22b −0.54a −0.53a 0.49a 0.43a −0.37a −0.37a 0.33a 0.28a
< 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 0.02 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
47
Internet Appendix Table VIIIExponential Random Graph Models to Explain the M&A Network: Robustness ChecksThis table reports the coefficient estimates from exponential random graph models. The coefficient estimates are the marginal effect of theexplanatory variable on the conditional log-odds that two industries will have an additional inter-industry merger. Merger data is from SDC.Industries are defined by the Bureau of Economic Analysis Input-Output (IO) Industry classification. The IO industry definitions are based oninput-output relations between industries as recorded by the BEA in five separate reports (1982, 1987, 1992, 1997, and 2002), for two levelsof industry definitions (Detailed and Summary). Panel A presents the detailed-level data and panel B presents the summary-level data. Theconnections in the merger network are the dependent variables, where the merger network is constructed as the number of inter-industry mergersbetween two industries. The explanatory variables are the connections in the IO network constructed as in the text. ‘Target Buys From Acquirer’is the network where each connection is the percentage that the Target industry buys of the Acquirer industry’s output. The connections in‘Target Sells to Acquirer’ are the percentage of inputs supplied by the Target industry to the Acquirer industry. The coefficient on ‘Number ofConnections’ is the marginal effect of an additional random connection on the conditional log-odds ratio of two industries having an additionalmerger in the merger network. | ∆ Variable | is the absolute difference between two industry nodes’ value of variable. AIC is the Akaike’sInformation Criterion. p−values are reported in parentheses. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗, for the 0.01, 0.05, and 0.10levels.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Panel A: Detail-Level
IO Report Year: 1982
Number of Connections −3.676∗∗∗ −3.680∗∗∗ −3.696∗∗∗ −3.694∗∗∗ −3.731∗∗∗ −1.970∗∗∗ −1.973∗∗∗ −2.131∗∗∗ −2.095∗∗∗ −2.265∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 5.121∗∗∗ 2.072∗∗∗ 6.430∗∗∗ 2.767∗∗∗
(<.001) (<.001) (<.001) (<.001)Acquirer Buys from Target 6.114∗∗∗ 3.783∗∗∗ 6.105∗∗∗ 2.996∗∗∗
(<.001) (<.001) (<.001) (<.001)Acquirer Sells to Target 12.725∗∗∗ 11.657∗∗∗ 14.970∗∗∗ 13.671∗∗∗
(<.001) (<.001) (<.001) (<.001)Target Sells to Acquirer 11.877∗∗∗ 10.139∗∗∗ 12.897∗∗∗ 11.475∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} −8.912∗∗∗ −8.972∗∗∗ −8.799∗∗∗ −8.821∗∗∗ −8.423∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Industry M/B| −0.113∗∗∗ −0.113∗∗∗ −0.128∗∗∗ −0.124∗∗∗ −0.140∗∗∗
(0.004) (0.004) (0.001) (0.002) (<.001)| ∆ Industry Mean Returns| −0.702∗∗∗ −0.693∗∗∗ −0.649∗∗∗ −0.655∗∗∗ −0.654∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Std Dev of Returns| −0.017 −0.020 −0.017 −0.017 −0.022
(0.355) (0.284) (0.368) (0.366) (0.256)| ∆ Concentration Ratio| 0.000 0.000 0.000 0.000 0.000
(0.931) (0.916) (0.921) (0.914) (0.952)Target Buys from Acquirer×max{Industry R&D} 31.460 −29.547
(0.286) (0.384)Acquirer Buys from Target×max{Industry R&D} 102.817∗∗∗ 67.136∗∗
(0.002) (0.044)Acquirer Sells to Target×max{Industry R&D} 155.695∗∗∗ 165.099∗∗∗
(<.001) (<.001)Target Sells to Acquirer×max{Industry R&D} 72.370∗∗ 39.931
(0.039) (0.308)Industry Economic Shock Index −0.327∗∗∗ −0.327∗∗∗ −0.310∗∗∗ −0.313∗∗∗ −0.300∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median M/B 0.216∗∗∗ 0.216∗∗∗ 0.228∗∗∗ 0.225∗∗∗ 0.240∗∗∗
48Internet Appendix Table VIII - Continued
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median R&D 15.789∗∗∗ 15.718∗∗∗ 15.450∗∗∗ 15.627∗∗∗ 15.057∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Mean Return −0.677∗∗∗ −0.677∗∗∗ −0.700∗∗∗ −0.701∗∗∗ −0.698∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Std Dev of Returns 0.422∗∗∗ 0.423∗∗∗ 0.424∗∗∗ 0.425∗∗∗ 0.422∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Concentration Ratio −0.010∗∗∗ −0.010∗∗∗ −0.009∗∗∗ −0.009∗∗∗ −0.009∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Size 0.000 0.000 0.000 0.000 0.000
(0.747) (0.748) (0.786) (0.769) (0.820)Industry Scope 31.728∗∗∗ 31.745∗∗∗ 32.863∗∗∗ 32.623∗∗∗ 33.602∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)AIC 53,101 53,020 52,777 52,838 52,300 25,963 25,945 25,722 25,840 25,477Number of Industries 478 478 478 478 478 289 289 289 289 289
IO Report Year: 1987
Number of Connections −3.606∗∗∗ −3.607∗∗∗ −3.628∗∗∗ −3.626∗∗∗ −3.664∗∗∗ −0.378∗∗∗ −0.371∗∗∗ −0.447∗∗∗ −0.455∗∗∗ −0.570∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (0.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 4.331∗∗∗ 2.519∗∗∗ 4.781∗∗∗ 3.053∗∗∗
(<.001) (<.001) (<.001) (<.001)Acquirer Buys from Target 4.416∗∗∗ 2.670∗∗∗ 3.920∗∗∗ 1.969∗∗∗
(<.001) (<.001) (<.001) (<.001)Acquirer Sells to Target 12.854∗∗∗ 11.488∗∗∗ 13.892∗∗∗ 12.355∗∗∗
(<.001) (<.001) (<.001) (<.001)Target Sells to Acquirer 12.478∗∗∗ 11.048∗∗∗ 15.014∗∗∗ 13.928∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} 3.740∗∗∗ 3.647∗∗∗ 3.580∗∗∗ 3.619∗∗∗ 3.304∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Industry M/B| −0.153∗∗∗ −0.150∗∗∗ −0.158∗∗∗ −0.160∗∗∗ −0.170∗∗∗
(0.001) (0.001) (<.001) (<.001) (<.001)| ∆ Industry Mean Returns| −0.546∗∗∗ −0.547∗∗∗ −0.540∗∗∗ −0.540∗∗∗ −0.509∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Std Dev of Returns| 0.004 0.005 0.004 0.004 −0.003
(0.903) (0.886) (0.903) (0.903) (0.936)| ∆ Concentration Ratio| 0.004∗∗∗ 0.004∗∗∗ 0.005∗∗∗ 0.005∗∗∗ 0.005∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer×max{Industry R&D} 71.015∗∗ 17.231
(0.023) (0.454)Acquirer Buys from Target×max{Industry R&D} 141.009∗∗∗ 80.617∗∗
(<.001) (0.017)Acquirer Sells to Target×max{Industry R&D} 108.901∗∗∗ 95.007∗∗
(0.002) (0.011)Target Sells to Acquirer×max{Industry R&D} 87.548∗∗ 51.532
(0.013) (0.173)Industry Economic Shock Index −0.569∗∗∗ −0.568∗∗∗ −0.555∗∗∗ −0.554∗∗∗ −0.542∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median M/B 0.064∗ 0.064∗ 0.069∗∗ 0.070∗∗ 0.087∗∗∗
(0.052) (0.053) (0.035) (0.033) (0.009)Industry Median R&D 5.877∗∗∗ 5.832∗∗∗ 5.939∗∗∗ 5.966∗∗∗ 6.067∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Mean Return −0.604∗∗∗ −0.604∗∗∗ −0.566∗∗∗ −0.564∗∗∗ −0.517∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Std Dev of Returns 0.504∗∗∗ 0.504∗∗∗ 0.493∗∗∗ 0.492∗∗∗ 0.480∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)
49
Internet Appendix Table VIII - Continued
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Concentration Ratio −0.014∗∗∗ −0.014∗∗∗ −0.014∗∗∗ −0.014∗∗∗ −0.015∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Size 0.000∗∗∗ 0.000∗∗∗ 0.000∗∗∗ 0.000∗∗∗ 0.000∗∗∗
(0.004) (0.004) (0.003) (0.003) (<.001)Industry Scope 11.951∗∗∗ 11.865∗∗∗ 12.002∗∗∗ 12.027∗∗∗ 11.956∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)AIC 52,953 52,944 52,643 52,671 52,090 30,272 30,277 30,114 30,086 29,769Number of Industries 465 465 465 465 465 283 283 283 283 283
IO Report Year: 1992
Number of Connections −3.609∗∗∗ −3.612∗∗∗ −3.637∗∗∗ −3.631∗∗∗ −3.675∗∗∗ −0.360∗∗∗ −0.359∗∗∗ −0.468∗∗∗ −0.455∗∗∗ −0.582∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 5.244∗∗∗ 1.893∗∗∗ 6.306∗∗∗ 2.571∗∗∗
(<.001) (<.001) (<.001) (0.002)Acquirer Buys from Target 6.138∗∗∗ 3.335∗∗∗ 5.880∗∗∗ 2.290∗∗∗
(<.001) (<.001) (<.001) (0.006)Acquirer Sells to Target 15.457∗∗∗ 14.401∗∗∗ 16.135∗∗∗ 15.050∗∗∗
(<.001) (<.001) (<.001) (<.001)Target Sells to Acquirer 13.494∗∗∗ 11.863∗∗∗ 15.211∗∗∗ 14.242∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} 4.911∗∗∗ 4.810∗∗∗ 4.567∗∗∗ 4.782∗∗∗ 4.484∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Industry M/B| −0.383∗∗∗ −0.384∗∗∗ −0.400∗∗∗ −0.397∗∗∗ −0.415∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Industry Mean Returns| −0.550∗∗∗ −0.538∗∗∗ −0.508∗∗∗ −0.515∗∗∗ −0.495∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Std Dev of Returns| −0.032∗ −0.035∗ −0.032∗ −0.032∗ −0.040∗∗
(0.092) (0.066) (0.089) (0.090) (0.038)| ∆ Concentration Ratio| 0.003∗∗∗ 0.003∗∗∗ 0.003∗∗∗ 0.003∗∗∗ 0.003∗∗∗
(0.004) (0.004) (0.003) (0.004) (0.003)Target Buys from Acquirer×max{Industry R&D} 38.671 −48.575
(0.138) (0.232)Acquirer Buys from Target×max{Industry R&D} 111.462∗∗∗ 73.474∗∗
(<.001) (0.012)Acquirer Sells to Target×max{Industry R&D} 191.049∗∗∗ 197.324∗∗∗
(<.001) (<.001)Target Sells to Acquirer×max{Industry R&D} 59.040∗ 23.083
(0.057) (0.499)Industry Economic Shock Index −0.526∗∗∗ −0.526∗∗∗ −0.515∗∗∗ −0.516∗∗∗ −0.503∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median M/B 0.062∗∗ 0.063∗∗ 0.076∗∗ 0.072∗∗ 0.090∗∗∗
(0.045) (0.042) (0.014) (0.019) (0.004)Industry Median R&D 6.439∗∗∗ 6.391∗∗∗ 6.376∗∗∗ 6.505∗∗∗ 6.431∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Mean Return 0.444∗∗∗ 0.444∗∗∗ 0.478∗∗∗ 0.466∗∗∗ 0.520∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Std Dev of Returns 0.185∗∗∗ 0.185∗∗∗ 0.176∗∗∗ 0.180∗∗∗ 0.166∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Concentration Ratio −0.015∗∗∗ −0.015∗∗∗ −0.015∗∗∗ −0.015∗∗∗ −0.015∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Size 0.000∗∗∗ 0.000∗∗∗ 0.000∗∗∗ 0.000∗∗∗ 0.000∗∗∗
(0.008) (0.007) (0.005) (0.007) (0.003)Industry Scope 21.295∗∗∗ 21.329∗∗∗ 21.598∗∗∗ 21.500∗∗∗ 21.888∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)AIC 55,187 55,112 54,684 54,836 54,098 32,035 32,011 31,701 31,834 31,387
50Internet Appendix Table VIII - Continued
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Number of Industries 475 475 475 475 475 290 290 290 290 290
IO Report Year: 1997
Number of Connections −3.519∗∗∗ −3.523∗∗∗ −3.545∗∗∗ −3.549∗∗∗ −3.596∗∗∗ −3.286∗∗∗ −3.268∗∗∗ −3.324∗∗∗ −3.329∗∗∗ −3.468∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 6.564∗∗∗ 3.328∗∗∗ 8.920∗∗∗ 5.093∗∗∗
(<.001) (<.001) (<.001) (<.001)Acquirer Buys from Target 8.092∗∗∗ 4.680∗∗∗ 6.853∗∗∗ 3.763∗∗∗
(<.001) (<.001) (<.001) (<.001)Acquirer Sells to Target 16.208∗∗∗ 13.922∗∗∗ 21.824∗∗∗ 19.590∗∗∗
(<.001) (<.001) (<.001) (<.001)Target Sells to Acquirer 17.585∗∗∗ 14.790∗∗∗ 20.364∗∗∗ 18.423∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} 5.133∗∗∗ 5.093∗∗∗ 4.871∗∗∗ 5.075∗∗∗ 4.871∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Industry M/B| −0.084∗ −0.083∗ −0.094∗∗ −0.090∗∗ −0.115∗∗
(0.060) (0.062) (0.038) (0.045) (0.015)| ∆ Industry Mean Returns| −0.741∗∗∗ −0.748∗∗∗ −0.751∗∗∗ −0.762∗∗∗ −0.696∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Std Dev of Returns| −0.123 −0.117 −0.100 −0.109 −0.149
(0.327) (0.350) (0.428) (0.383) (0.240)| ∆ Concentration Ratio| 0.003∗∗∗ 0.003∗∗∗ 0.003∗∗∗ 0.003∗∗∗ 0.003∗∗
(0.008) (0.007) (0.010) (0.007) (0.014)Target Buys from Acquirer×max{Industry R&D} 41.042∗ 12.402
(0.094) (0.557)Acquirer Buys from Target×max{Industry R&D} 75.596∗∗∗ 53.284∗∗
(0.004) (0.024)Acquirer Sells to Target×max{Industry R&D} 87.524∗∗∗ 73.437∗∗
(0.005) (0.020)Target Sells to Acquirer×max{Industry R&D} −55.674∗∗ −74.823∗∗∗
(0.033) (0.008)Industry Economic Shock Index −0.250∗∗∗ −0.250∗∗∗ −0.241∗∗∗ −0.241∗∗∗ −0.241∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median M/B −0.088∗∗∗ −0.091∗∗∗ −0.087∗∗∗ −0.090∗∗∗ −0.057∗
(0.008) (0.006) (0.010) (0.007) (0.086)Industry Median R&D 0.050 0.012 0.204 0.331 0.441
(0.951) (0.988) (0.805) (0.687) (0.597)Industry Mean Return 0.300∗ 0.302∗ 0.280∗ 0.286∗ 0.253
(0.061) (0.060) (0.083) (0.076) (0.119)Industry Std Dev of Returns 2.328∗∗∗ 2.327∗∗∗ 2.291∗∗∗ 2.311∗∗∗ 2.259∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Concentration Ratio −0.012∗∗∗ −0.012∗∗∗ −0.012∗∗∗ −0.012∗∗∗ −0.012∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Size 0.000 0.000 0.000 0.000 0.000
(0.259) (0.262) (0.349) (0.317) (0.276)Industry Scope −1.532 −1.618 −0.621 −0.876 0.364
(0.504) (0.480) (0.786) (0.702) (0.874)AIC 58,158 58,034 57,791 57,691 56,938 26,199 26,237 25,922 26,080 25,551Number of Industries 471 471 471 471 471 214 214 214 214 214
IO Report Year: 2002
Number of Connections −3.241∗∗∗ −3.249∗∗∗ −3.277∗∗∗ −3.272∗∗∗ −3.329∗∗∗ −2.984∗∗∗ −2.980∗∗∗ −3.107∗∗∗ −3.091∗∗∗ −3.293∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 6.816∗∗∗ 2.920∗∗∗ 8.338∗∗∗ 4.809∗∗∗
51
Internet Appendix Table VIII - Continued
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
(<.001) (<.001) (<.001) (<.001)Acquirer Buys from Target 9.476∗∗∗ 5.668∗∗∗ 8.034∗∗∗ 4.483∗∗∗
(<.001) (<.001) (<.001) (<.001)Acquirer Sells to Target 19.603∗∗∗ 17.266∗∗∗ 19.129∗∗∗ 16.291∗∗∗
(<.001) (<.001) (<.001) (<.001)Target Sells to Acquirer 18.164∗∗∗ 14.591∗∗∗ 19.709∗∗∗ 17.366∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} 4.085∗∗∗ 4.065∗∗∗ 3.796∗∗∗ 4.133∗∗∗ 3.996∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Industry M/B| −0.078∗ −0.077∗ −0.090∗∗ −0.090∗∗ −0.106∗∗
(0.066) (0.068) (0.040) (0.039) (0.019)| ∆ Industry Mean Returns| −1.043∗∗∗ −1.042∗∗∗ −1.045∗∗∗ −1.063∗∗∗ −0.992∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Std Dev of Returns| −0.160 −0.161 −0.141 −0.143 −0.194
(0.187) (0.187) (0.251) (0.242) (0.117)| ∆ Concentration Ratio| 0.002∗∗ 0.002∗∗ 0.002 0.002∗ 0.002
(0.040) (0.041) (0.113) (0.064) (0.164)Target Buys from Acquirer×max{Industry R&D} −11.059 −27.888∗
(0.414) (0.100)Acquirer Buys from Target×max{Industry R&D} 3.643 18.093
(0.817) (0.324)Acquirer Sells to Target×max{Industry R&D} 103.513∗∗∗ 118.986∗∗∗
(<.001) (<.001)Target Sells to Acquirer×max{Industry R&D} −89.690∗∗∗−123.147∗∗∗
(<.001) (<.001)Industry Economic Shock Index −0.251∗∗∗ −0.251∗∗∗ −0.242∗∗∗ −0.241∗∗∗ −0.245∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median M/B −0.121∗∗∗ −0.122∗∗∗ −0.108∗∗∗ −0.113∗∗∗ −0.077∗∗
(<.001) (<.001) (0.002) (0.001) (0.028)Industry Median R&D 0.424 0.409 0.271 0.519 0.395
(0.548) (0.563) (0.703) (0.460) (0.578)Industry Mean Return 0.308∗ 0.308∗ 0.213 0.256 0.141
(0.082) (0.083) (0.230) (0.149) (0.430)Industry Std Dev of Returns 2.155∗∗∗ 2.155∗∗∗ 2.179∗∗∗ 2.172∗∗∗ 2.199∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Concentration Ratio −0.011∗∗∗ −0.011∗∗∗ −0.010∗∗∗ −0.011∗∗∗ −0.010∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Size 0.000 0.000 0.000 0.000 0.000
(0.762) (0.763) (0.779) (0.775) (0.655)Industry Scope 2.284 2.291 2.452 2.317 2.768
(0.320) (0.318) (0.285) (0.313) (0.228)AIC 54,455 54,280 54,016 54,108 53,308 24,673 24,671 24,420 24,558 24,106Number of Industries 411 411 411 411 411 245 245 245 245 245
Panel B: Summary-Level
IO Report Year: 1982
Number of Connections −1.163∗∗∗ −1.194∗∗∗ −1.229∗∗∗ −1.217∗∗∗ −1.294∗∗∗ −1.168∗∗∗ −1.109∗∗∗ −1.146∗∗∗ −1.260∗∗∗ −1.213∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (0.005) (0.008) (0.006) (0.003) (0.004)Target Buys from Acquirer 1.296∗ −1.888∗ 1.140 0.018
(0.073) (0.075) (0.277) (0.988)Acquirer Buys from Target 5.467∗∗∗ 3.232∗∗∗ 3.306∗∗∗ 1.504
(<.001) (<.001) (0.005) (0.192)Acquirer Sells to Target 9.708∗∗∗ 10.345∗∗∗ 5.452∗∗∗ 5.555∗∗∗
(<.001) (<.001) (0.001) (0.002)
52Internet Appendix Table VIII - Continued
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Target Sells to Acquirer 8.197∗∗∗ 6.421∗∗∗ 7.804∗∗∗ 7.283∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} 4.579 4.454 4.080 5.002 4.690
(0.178) (0.193) (0.235) (0.141) (0.174)| ∆ Industry M/B| 0.156 0.142 0.149 0.150 0.130
(0.392) (0.439) (0.418) (0.410) (0.481)| ∆ Industry Mean Returns| −0.913∗∗ −0.782∗ −0.785∗ −0.901∗ −0.683
(0.049) (0.092) (0.091) (0.053) (0.144)| ∆ Std Dev of Returns| −0.793∗∗∗ −0.796∗∗∗ −0.806∗∗∗ −0.804∗∗∗ −0.827∗∗∗
(0.001) (0.001) (<.001) (<.001) (<.001)| ∆ Concentration Ratio| −0.005∗ −0.005∗ −0.005∗ −0.005∗ −0.004
(0.052) (0.075) (0.084) (0.086) (0.185)Target Buys from Acquirer×max{Industry R&D} −5.573 −140.631∗∗
(0.908) (0.050)Acquirer Buys from Target×max{Industry R&D} 107.353∗ 157.648∗∗
(0.058) (0.017)Acquirer Sells to Target×max{Industry R&D} 143.464∗∗ 198.586∗∗∗
(0.012) (0.005)Target Sells to Acquirer×max{Industry R&D} −122.821∗∗−174.505∗∗∗
(0.014) (0.007)Industry Economic Shock Index 0.052 0.043 0.053 0.060 0.054
(0.199) (0.295) (0.190) (0.142) (0.193)Industry Median M/B −0.404∗∗∗ −0.410∗∗∗ −0.411∗∗∗ −0.390∗∗∗ −0.397∗∗∗
(0.006) (0.006) (0.006) (0.008) (0.008)Industry Median R&D −4.987 −4.939 −4.705 −4.696 −4.441
(0.122) (0.128) (0.149) (0.145) (0.174)Industry Mean Return −1.363∗∗∗ −1.453∗∗∗ −1.405∗∗∗ −1.364∗∗∗ −1.469∗∗∗
(0.004) (0.002) (0.003) (0.004) (0.002)Industry Std Dev of Returns 2.180∗∗∗ 2.169∗∗∗ 2.081∗∗∗ 2.147∗∗∗ 2.054∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Concentration Ratio −0.001 −0.002 −0.001 −0.001 −0.002
(0.438) (0.358) (0.465) (0.443) (0.392)Industry Size 0.000∗∗∗ 0.000∗∗∗ 0.000∗∗∗ 0.000∗∗∗ 0.000∗∗∗
(0.010) (0.007) (0.009) (0.008) (0.005)Industry Scope −6.119∗∗ −6.627∗∗∗ −5.235∗∗ −5.891∗∗ −5.533∗∗
(0.014) (0.009) (0.036) (0.019) (0.030)AIC 6,450 6,407 6,392 6,407 6,336 5,229 5,210 5,195 5,207 5,163Number of Industries 77 77 77 77 77 68 68 68 68 68
IO Report Year: 1987
Number of Connections −1.128∗∗∗ −1.150∗∗∗ −1.176∗∗∗ −1.183∗∗∗ −1.245∗∗∗ −1.620∗∗∗ −1.574∗∗∗ −1.648∗∗∗ −1.693∗∗∗ −1.691∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 2.608∗∗∗ −0.830 2.266∗∗ 0.497
(0.001) (0.422) (0.020) (0.667)Acquirer Buys from Target 6.057∗∗∗ 2.666∗∗∗ 4.035∗∗∗ 1.165
(<.001) (0.009) (<.001) (0.315)Acquirer Sells to Target 9.811∗∗∗ 9.721∗∗∗ 5.443∗∗∗ 5.179∗∗∗
(<.001) (<.001) (<.001) (0.002)Target Sells to Acquirer 10.812∗∗∗ 8.663∗∗∗ 10.613∗∗∗ 9.872∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} 4.371 4.552 4.435 4.883 5.064
(0.175) (0.161) (0.172) (0.131) (0.122)| ∆ Industry M/B| −0.393∗∗∗ −0.376∗∗∗ −0.388∗∗∗ −0.399∗∗∗ −0.388∗∗∗
(0.002) (0.003) (0.002) (0.002) (0.003)| ∆ Industry Mean Returns| −0.974∗∗∗ −0.945∗∗∗ −0.916∗∗∗ −0.893∗∗ −0.815∗∗
53
Internet Appendix Table VIII - Continued
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
(0.005) (0.006) (0.008) (0.010) (0.019)| ∆ Std Dev of Returns| −0.465∗∗ −0.463∗∗ −0.472∗∗ −0.492∗∗ −0.510∗∗∗
(0.014) (0.015) (0.013) (0.010) (0.008)| ∆ Concentration Ratio| −0.006∗∗∗ −0.006∗∗∗ −0.006∗∗∗ −0.005∗∗ −0.005∗∗
(0.006) (0.006) (0.008) (0.013) (0.026)Target Buys from Acquirer×max{Industry R&D} −39.286 −124.138∗
(0.332) (0.055)Acquirer Buys from Target×max{Industry R&D} 87.554 131.171∗∗
(0.110) (0.048)Acquirer Sells to Target×max{Industry R&D} 79.404 148.116∗∗
(0.140) (0.031)Target Sells to Acquirer×max{Industry R&D} −84.994∗ −151.012∗∗
(0.098) (0.023)Industry Economic Shock Index −0.037 −0.037 −0.032 −0.033 −0.033
(0.287) (0.282) (0.354) (0.346) (0.339)Industry Median M/B −0.128 −0.133 −0.118 −0.112 −0.109
(0.281) (0.265) (0.322) (0.345) (0.365)Industry Median R&D −2.195 −2.594 −2.553 −2.261 −2.488
(0.473) (0.400) (0.407) (0.460) (0.423)Industry Mean Return −1.552∗∗∗ −1.575∗∗∗ −1.535∗∗∗ −1.581∗∗∗ −1.640∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Std Dev of Returns 2.036∗∗∗ 2.005∗∗∗ 1.980∗∗∗ 1.993∗∗∗ 1.954∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Concentration Ratio −0.002 −0.002 −0.002 −0.002 −0.002
(0.258) (0.206) (0.239) (0.234) (0.188)Industry Size 0.000 0.000 0.000 0.000 0.000
(0.270) (0.226) (0.252) (0.216) (0.159)Industry Scope −0.001 −0.100 0.307 0.022 −0.111
(1.000) (0.960) (0.877) (0.991) (0.956)AIC 8,369 8,333 8,320 8,309 8,252 7,483 7,458 7,461 7,438 7,407Number of Industries 87 87 87 87 87 82 82 82 82 82
IO Report Year: 1992
Number of Connections −1.172∗∗∗ −1.196∗∗∗ −1.229∗∗∗ −1.222∗∗∗ −1.298∗∗∗ −1.709∗∗∗ −1.674∗∗∗ −1.692∗∗∗ −1.779∗∗∗ −1.772∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 3.128∗∗∗ −1.017 2.898∗∗∗ 1.119
(<.001) (0.354) (0.007) (0.358)Acquirer Buys from Target 6.920∗∗∗ 3.234∗∗∗ 5.020∗∗∗ 1.451
(<.001) (0.003) (<.001) (0.238)Acquirer Sells to Target 11.421∗∗∗ 11.553∗∗∗ 6.247∗∗∗ 5.686∗∗∗
(<.001) (<.001) (<.001) (0.001)Target Sells to Acquirer 10.511∗∗∗ 8.242∗∗∗ 11.184∗∗∗ 10.644∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} 5.700∗ 6.096∗ 5.932∗ 6.094∗ 6.598∗∗
(0.080) (0.063) (0.072) (0.061) (0.047)| ∆ Industry M/B| −0.567∗∗∗ −0.555∗∗∗ −0.560∗∗∗ −0.576∗∗∗ −0.567∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Industry Mean Returns| −0.932∗∗∗ −0.890∗∗∗ −0.857∗∗ −0.844∗∗ −0.718∗∗
(0.007) (0.010) (0.013) (0.015) (0.038)| ∆ Std Dev of Returns| −0.557∗∗∗ −0.562∗∗∗ −0.574∗∗∗ −0.596∗∗∗ −0.635∗∗∗
(0.004) (0.004) (0.003) (0.002) (0.001)| ∆ Concentration Ratio| −0.008∗∗∗ −0.008∗∗∗ −0.008∗∗∗ −0.008∗∗∗ −0.008∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer×max{Industry R&D} −31.956 −190.303∗∗
(0.463) (0.013)
54Internet Appendix Table VIII - Continued
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Acquirer Buys from Target×max{Industry R&D} 70.438 193.540∗∗∗
(0.180) (0.010)Acquirer Sells to Target×max{Industry R&D} 112.640∗∗ 213.879∗∗∗
(0.046) (0.005)Target Sells to Acquirer×max{Industry R&D} −139.751∗∗∗−250.109∗∗∗
(0.003) (<.001)Industry Economic Shock Index −0.104∗∗∗ −0.106∗∗∗ −0.099∗∗∗ −0.102∗∗∗ −0.105∗∗∗
(0.002) (0.002) (0.004) (0.003) (0.002)Industry Median M/B 0.105 0.106 0.103 0.113 0.123
(0.358) (0.354) (0.368) (0.323) (0.288)Industry Median R&D −3.376 −3.942 −3.948 −3.049 −3.610
(0.274) (0.206) (0.206) (0.323) (0.252)Industry Mean Return −1.492∗∗∗ −1.512∗∗∗ −1.482∗∗∗ −1.532∗∗∗ −1.590∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Std Dev of Returns 1.971∗∗∗ 1.940∗∗∗ 1.903∗∗∗ 1.942∗∗∗ 1.884∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Concentration Ratio −0.002 −0.002 −0.002 −0.002 −0.002
(0.216) (0.175) (0.204) (0.227) (0.180)Industry Size 0.000 0.000 0.000 0.000 0.000
(0.213) (0.172) (0.215) (0.190) (0.122)Industry Scope −10.284∗∗∗−10.400∗∗∗−10.155∗∗∗−10.153∗∗∗−10.293∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)AIC 8,428 8,388 8,361 8,371 8,290 7,539 7,514 7,506 7,493 7,438Number of Industries 88 88 88 88 88 83 83 83 83 83
IO Report Year: 1997
Number of Connections −1.466∗∗∗ −1.475∗∗∗ −1.497∗∗∗ −1.511∗∗∗ −1.575∗∗∗ −1.223∗∗∗ −1.225∗∗∗ −1.211∗∗∗ −1.304∗∗∗ −1.370∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 7.202∗∗∗ 2.483∗∗ 10.519∗∗∗ 5.930∗∗∗
(<.001) (0.026) (<.001) (0.009)Acquirer Buys from Target 9.146∗∗∗ 3.529∗∗∗ 12.955∗∗∗ 3.526
(<.001) (0.003) (<.001) (0.136)Acquirer Sells to Target 12.150∗∗∗ 9.986∗∗∗ 8.381∗∗∗ 4.595∗∗
(<.001) (<.001) (<.001) (0.037)Target Sells to Acquirer 14.855∗∗∗ 12.277∗∗∗ 20.046∗∗∗ 17.404∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} −5.282∗∗ −5.417∗∗∗ −5.191∗∗ −4.673∗∗ −4.560∗∗
(0.011) (0.009) (0.012) (0.024) (0.029)| ∆ Industry M/B| −0.111 −0.095 −0.113 −0.113 −0.098
(0.187) (0.257) (0.180) (0.182) (0.245)| ∆ Industry Mean Returns| −0.573∗∗ −0.585∗∗ −0.558∗ −0.561∗ −0.590∗∗
(0.046) (0.042) (0.052) (0.052) (0.041)| ∆ Std Dev of Returns| 0.046 0.059 0.041 0.049 0.073
(0.834) (0.789) (0.853) (0.824) (0.741)| ∆ Concentration Ratio| −0.006∗∗∗ −0.006∗∗∗ −0.006∗∗∗ −0.005∗∗∗ −0.005∗∗∗
(<.001) (<.001) (<.001) (<.001) (0.002)Target Buys from Acquirer×max{Industry R&D} −85.523∗ −142.212∗∗
(0.058) (0.037)Acquirer Buys from Target×max{Industry R&D} 110.321∗ 161.696∗∗
(0.057) (0.016)Acquirer Sells to Target×max{Industry R&D} 20.398 79.569
(0.607) (0.104)Target Sells to Acquirer×max{Industry R&D} −99.616∗∗−146.321∗∗∗
(0.015) (0.006)Industry Economic Shock Index −0.109∗∗∗ −0.107∗∗∗ −0.111∗∗∗ −0.107∗∗∗ −0.101∗∗∗
55
Internet Appendix Table VIII - Continued
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median M/B −0.375∗∗∗ −0.372∗∗∗ −0.376∗∗∗ −0.368∗∗∗ −0.353∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median R&D 7.170∗∗∗ 6.791∗∗∗ 6.805∗∗∗ 6.832∗∗∗ 6.620∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Mean Return −1.589∗∗∗ −1.568∗∗∗ −1.580∗∗∗ −1.575∗∗∗ −1.551∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Std Dev of Returns 1.895∗∗∗ 1.868∗∗∗ 1.889∗∗∗ 1.884∗∗∗ 1.834∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Concentration Ratio −0.001 −0.001 −0.001 −0.001 −0.001
(0.435) (0.408) (0.430) (0.427) (0.393)Industry Size 0.000∗∗ 0.000∗∗ 0.000∗ 0.000∗∗ 0.000∗∗
(0.049) (0.049) (0.058) (0.034) (0.024)Industry Scope −6.724∗∗∗ −6.863∗∗∗ −6.870∗∗∗ −7.148∗∗∗ −7.103∗∗∗
(0.006) (0.005) (0.005) (0.004) (0.004)AIC 14,878 14,851 14,832 14,788 14,677 10,773 10,721 10,771 10,677 10,624Number of Industries 124 124 124 124 124 103 103 103 103 103
IO Report Year: 2002
Number of Connections −1.503∗∗∗ −1.518∗∗∗ −1.548∗∗∗ −1.552∗∗∗ −1.632∗∗∗ −0.758∗∗∗ −0.774∗∗∗ −0.763∗∗∗ −0.863∗∗∗ −0.939∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001) (0.005) (0.005) (0.005) (0.002) (<.001)Target Buys from Acquirer 6.214∗∗∗ 1.994∗∗ 8.751∗∗∗ 5.027∗∗
(<.001) (0.040) (<.001) (0.027)Acquirer Buys from Target 9.570∗∗∗ 4.693∗∗∗ 13.806∗∗∗ 3.875
(<.001) (<.001) (<.001) (0.101)Acquirer Sells to Target 14.215∗∗∗ 12.482∗∗∗ 7.846∗∗∗ 4.700∗∗
(<.001) (<.001) (<.001) (0.044)Target Sells to Acquirer 15.081∗∗∗ 12.284∗∗∗ 21.985∗∗∗ 19.914∗∗∗
(<.001) (<.001) (<.001) (<.001)max{Industry R&D} −8.144∗∗∗ −8.341∗∗∗ −8.065∗∗∗ −7.436∗∗∗ −7.621∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Industry M/B| −0.043 −0.011 −0.039 −0.048 −0.028
(0.623) (0.903) (0.656) (0.585) (0.752)| ∆ Industry Mean Returns| −0.045 −0.085 −0.038 −0.037 −0.097
(0.880) (0.777) (0.898) (0.903) (0.749)| ∆ Std Dev of Returns| −0.152 −0.131 −0.156 −0.159 −0.120
(0.503) (0.564) (0.491) (0.486) (0.603)| ∆ Concentration Ratio| −0.007∗∗∗ −0.007∗∗∗ −0.007∗∗∗ −0.006∗∗∗ −0.006∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer×max{Industry R&D} −114.345∗∗ −280.320∗∗∗
(0.022) (0.001)Acquirer Buys from Target×max{Industry R&D} 119.205∗∗ 218.436∗∗∗
(0.037) (0.002)Acquirer Sells to Target×max{Industry R&D} 79.537∗ 203.220∗∗∗
(0.075) (0.001)Target Sells to Acquirer×max{Industry R&D} −129.409∗∗∗−222.831∗∗∗
(0.004) (<.001)Industry Economic Shock Index −0.104∗∗∗ −0.102∗∗∗ −0.105∗∗∗ −0.100∗∗∗ −0.096∗∗∗
(0.001) (0.001) (0.001) (0.002) (0.003)Industry Median M/B −0.498∗∗∗ −0.499∗∗∗ −0.492∗∗∗ −0.485∗∗∗ −0.477∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median R&D 10.367∗∗∗ 9.949∗∗∗ 9.763∗∗∗ 10.050∗∗∗ 9.976∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Mean Return −1.061∗∗∗ −1.057∗∗∗ −1.055∗∗∗ −1.062∗∗∗ −1.056∗∗∗
(0.006) (0.007) (0.006) (0.006) (0.007)
56Internet Appendix Table VIII - Continued
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Industry Std Dev of Returns 1.774∗∗∗ 1.749∗∗∗ 1.753∗∗∗ 1.756∗∗∗ 1.723∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Concentration Ratio −0.002 −0.002 −0.002 −0.002 −0.002
(0.113) (0.113) (0.112) (0.107) (0.108)Industry Size 0.000∗∗ 0.000∗∗ 0.000∗∗ 0.000∗∗ 0.000∗∗
(0.038) (0.040) (0.036) (0.019) (0.016)Industry Scope −14.458∗∗∗−14.073∗∗∗−14.151∗∗∗−14.619∗∗∗−14.138∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)AIC 15,090 15,042 15,009 14,995 14,846 9,658 9,583 9,641 9,552 9,471Number of Industries 126 126 126 126 126 98 98 98 98 98
57
Internet Appendix Table IXExponential Random Graph Models to Explain the M&A Network: Asset ComplementarityThis table reports the coefficient estimates from exponential random graph models. The coefficient estimates are the marginal effect of theexplanatory variable on the conditional log-odds that two industries will have an additional inter-industry merger. Merger data is from SDCfrom 1996 to 2008. Industries are defined by the Bureau of Economic Analysis Input-Output (IO) Industry classification. The IO industrydefinitions are based on input-output relations between industries as recorded by the BEA in 1997 using the detailed-level industry definitions.The connections in the merger network are the dependent variables, where the merger network is constructed as the number of inter-industrymergers between two industries. The explanatory variables are the connections in the IO network constructed as in the text. ‘Target BuysFrom Acquirer’ is the network where each connection is the percentage that the Target industry buys of the Acquirer industry’s output. Theconnections in ‘Target Sells to Acquirer’ are the percentage of inputs supplied by the Target industry to the Acquirer industry. The coefficienton ‘Number of Connections’ is the marginal effect of an additional random connection on the conditional log-odds ratio of two industries havingan additional merger in the merger network. | ∆ Variable | is the absolute difference between two industry nodes’ value of variable. AIC is theAkaike’s Information Criterion. p−values are reported in parentheses. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗, for the 0.01, 0.05,and 0.10 levels.
(1) (2) (3) (4) (5)
Number of Connections −3.291∗∗∗ −3.179∗∗∗ −3.270∗∗∗ −3.344∗∗∗ −3.573∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 13.842∗∗∗ 9.223∗∗∗
(<.001) (<.001)Acquirer Buys from Target 21.725∗∗∗ 9.112∗∗∗
(<.001) (0.001)Acquirer Sells to Target 30.900∗∗∗ 29.343∗∗∗
(<.001) (<.001)Target Sells to Acquirer 36.825∗∗∗ 33.445∗∗∗
(<.001) (<.001)max{Industry R&D} 9.861∗∗∗ 9.772∗∗∗ 10.239∗∗∗ 9.683∗∗∗ 10.581∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Industry M/B| 0.360∗∗∗ 0.396∗∗∗ 0.368∗∗∗ 0.354∗∗∗ 0.319∗∗∗
(<.001) (<.001) (<.001) (<.001) (0.002)| ∆ Industry Mean Returns| −0.084 −0.186 −0.180 −0.218 −0.223
(0.771) (0.520) (0.536) (0.457) (0.457)| ∆ Std Dev of Returns| 1.100∗∗∗ 1.139∗∗∗ 1.056∗∗∗ 1.168∗∗∗ 0.978∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)| ∆ Concentration Ratio| 0.022∗∗∗ 0.021∗∗∗ 0.021∗∗∗ 0.023∗∗∗ 0.022∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer×max{Industry R&D} −51.598∗ −34.215
58Internet Appendix Table IX - Continued
(1) (2) (3) (4) (5)
(0.073) (0.240)Acquirer Buys from Target×max{Industry R&D} −56.253 0.399
(0.328) (0.993)Acquirer Sells to Target×max{Industry R&D} −162.921∗∗∗ −141.623∗∗∗
(0.001) (0.005)Target Sells to Acquirer×max{Industry R&D} 157.520∗∗ 157.912∗∗
(0.027) (0.027)HP Similarity 0.071∗∗∗ 0.071∗∗∗ 0.071∗∗∗ 0.066∗∗∗ 0.062∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Economic Shock Index −0.328∗∗∗ −0.307∗∗∗ −0.301∗∗∗ −0.317∗∗∗ −0.309∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median M/B −0.391∗∗∗ −0.422∗∗∗ −0.422∗∗∗ −0.397∗∗∗ −0.367∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Median R&D −1.006 −0.964 −1.150 −1.123 −1.555
(0.447) (0.465) (0.383) (0.404) (0.250)Industry Mean Return 0.406 0.590∗∗ 0.514∗ 0.527∗ 0.497
(0.169) (0.046) (0.083) (0.078) (0.102)Industry Std Dev of Returns 1.758∗∗∗ 1.724∗∗∗ 1.776∗∗∗ 1.759∗∗∗ 1.858∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Concentration Ratio −0.012∗∗∗ −0.013∗∗∗ −0.012∗∗∗ −0.012∗∗∗ −0.011∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Industry Size 0.000∗∗∗ 0.000∗∗ 0.000∗∗ 0.000∗∗∗ 0.000∗∗
(0.008) (0.014) (0.036) (0.007) (0.029)Industry Scope 26.533∗∗∗ 26.128∗∗∗ 20.849∗∗∗ 28.566∗∗∗ 19.343∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)AIC 9,386 9,370 9,322 9,217 9,022Number of Industries 289 289 289 289 289
59
Internet Appendix Table XExponential Random Graph Models to Explain the M&A Network: Alternative Industry AssignmentsThis table reports the coefficient estimates from exponential random graph models. The coefficient estimates are the marginal effect of theexplanatory variable on the conditional log-odds that two industries will have an additional inter-industry merger. Merger data is from SDC.Industries are defined by the Bureau of Economic Analysis Input-Output (IO) Industry classification. The IO industry definitions are basedon input-output relations between industries as recorded by the BEA in 1997 using the detailed-level industry definitions. In Panel A, we usethe ‘All’ network to assign firms using all industry codes reported by SDC, not just primary industry codes. In Panels B and C, we use the‘HV1%’ and ‘HV5%’ networks to assign firms to industries based on the following priority. If there are any shared IO codes between mergingfirms, we assign the merger equally to those industries. If there are no horizontal matches, but a vertical relation of 1% or 5% for any of thefour vertical relations (acquirer sells to target, acquirer buys from target, and vice versa), we assign the merger activity to those industries. Ifneither horizontal or vertical relations exist, we assign the merger activity equally to the unrelated industry pairs. The connections in the mergernetwork are the dependent variables, where the merger network is constructed as the number of inter-industry mergers between two industries.The explanatory variables are the connections in the IO network constructed as in the text. ‘Target Buys From Acquirer’ is the network whereeach connection is the percentage that the Target industry buys of the Acquirer industry’s output. The connections in ‘Target Sells to Acquirer’are the percentage of inputs supplied by the Target industry to the Acquirer industry. The coefficient on ‘Number of Connections’ is the marginaleffect of an additional random connection on the conditional log-odds ratio of two industries having an additional merger in the merger network.AIC is the Akaike’s Information Criterion. p−values are reported in parentheses. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗, for the0.01, 0.05, and 0.10 levels.
(1) (2) (3) (4) (5)
Panel A: Industry Assignments Using All Firm-Level Segment Industries (All)
Number of Connections −3.940∗∗∗ −3.942∗∗∗ −3.981∗∗∗ −3.967∗∗∗ −4.029∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 6.610∗∗∗ 2.738∗∗∗
(<.001) (<.001)Acquirer Buys from Target 7.130∗∗∗ 4.043∗∗∗
(<.001) (<.001)Acquirer Sells to Target 19.290∗∗∗ 17.383∗∗∗
(<.001) (<.001)Target Sells to Acquirer 15.675∗∗∗ 13.153∗∗∗
(<.001) (<.001)AIC 42,116 42,074 41,572 41,824 40,972Number of Industries 471 471 471 471 471
Panel B: Industry Assignments with Priority to Horizontal Mergers, then Vertical (HV1%)
Number of Connections −3.877∗∗∗ −3.882∗∗∗ −3.934∗∗∗ −3.910∗∗∗ −3.996∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 7.718∗∗∗ 3.052∗∗∗
60Internet Appendix Table X - Continued
(1) (2) (3) (4) (5)
(<.001) (<.001)Acquirer Buys from Target 9.048∗∗∗ 5.319∗∗∗
(<.001) (<.001)Acquirer Sells to Target 24.582∗∗∗ 22.448∗∗∗
(<.001) (<.001)Target Sells to Acquirer 18.581∗∗∗ 15.438∗∗∗
(<.001) (<.001)AIC 44,293 44,183 43,410 43,884 42,542Number of Industries 471 471 471 471 471
Panel C: Industry Assignment with Priority to Horizontal Mergers, then Vertical (HV5%)
Number of Connections −3.911∗∗∗ −3.914∗∗∗ −3.960∗∗∗ −3.947∗∗∗ −4.029∗∗∗
(<.001) (<.001) (<.001) (<.001) (<.001)Target Buys from Acquirer 8.168∗∗∗ 3.837∗∗∗
(<.001) (<.001)Acquirer Buys from Target 9.169∗∗∗ 5.280∗∗∗
(<.001) (<.001)Acquirer Sells to Target 23.025∗∗∗ 20.492∗∗∗
(<.001) (<.001)Target Sells to Acquirer 19.901∗∗∗ 16.747∗∗∗
(<.001) (<.001)AIC 43,166 43,083 42,455 42,700 41,476Number of Industries 471 471 471 471 471
61
Internet Appendix Table XICorrelation of Customer-Supplier Relations and Announcement Returns and Payment MethodThis table reports correlation coefficients between four input-output relations and acquirer abnormal announcement returns, acquirer and targetcombined abnormal announcement returns, and the percent of cash used as payment. The four input-output relations are ‘T Buys A’ (percentagethat the Target industry buys of the Acquirer industry’s output), ‘T Sells A’ (percentage of inputs supplied by the Target industry to theAcquirer industry), and ‘A Buys T’ and ‘A Sells T’, defined analogously. Acquirer abnormal announcement returns are the average of thethree-day abnormal returns for all acquirers in a particular industry-pair. Three-day abnormal returns are calculated as the sum of the acquirer’sreturn minus the valued-weighted CRSP index over the three days surrounding the merger announcement. ‘Acquirer & Target Combined CAR’is the value-weighted average of the acquirer’s and target’s three-day abnormal announcement returns. ‘Percent of Cash Used as Payment’ is theaverage percent of the total consideration paid in the merger that is composed of cash, across all mergers for each industry-pair. Correlationsare presented pooling all years in the sample and for each year individually. Sample sizes vary because most industry-pairs have no mergers, andhence no abnormal returns or cash fractions. For all years the sample size is 1,405 industry-pairs.
Acquirer CAR Acquirer & Target Combined CAR Percent of Cash Used as Payment
T Buys A A Buys T A Sells T T Sells A T Buys A A Buys T A Sells T T Sells A T Buys A A Buys T A Sells T T Sells A
All years 0.03 0.04 0.03 0.04 0.04 0.02 0.01 0.04 −0.05 0.00 −0.01 −0.03
1986 0.11 0.13 0.10 0.09 −0.05 −0.06 −0.07 −0.07 −0.01 −0.08 0.00 −0.151987 0.15 0.10 0.12 0.08 0.01 0.04 −0.01 0.03 −0.03 0.05 −0.04 0.021988 0.06 0.05 0.03 0.17 0.12 0.12 0.06 0.22 −0.15 −0.05 −0.06 −0.071989 −0.02 0.08 0.03 0.09 0.31 0.29 0.23 0.28 0.04 −0.04 0.01 −0.021990 0.25 0.19 0.17 0.27 0.37 0.29 0.28 0.39 −0.08 0.05 −0.09 0.051991 0.05 0.03 0.06 0.04 0.01 −0.03 −0.01 −0.04 −0.06 −0.06 −0.05 −0.021992 0.19 −0.24 0.04 −0.23 0.15 −0.24 0.01 −0.24 −0.11 −0.09 −0.03 −0.111993 0.02 0.09 0.05 0.08 −0.05 −0.04 −0.04 −0.06 −0.12 0.05 −0.02 0.021994 −0.03 −0.02 −0.02 −0.02 −0.02 0.01 −0.02 −0.01 0.01 0.00 0.04 −0.011995 0.12 0.04 0.09 0.04 0.08 0.11 0.05 0.12 −0.02 −0.02 −0.01 −0.051996 0.11 0.01 0.05 −0.03 0.05 0.11 0.08 0.01 −0.04 0.02 −0.03 0.011997 0.05 0.02 0.12 0.02 0.04 0.00 0.07 −0.02 0.04 −0.03 0.01 −0.021998 0.10 0.04 0.07 0.03 −0.03 −0.07 −0.06 −0.06 −0.05 0.03 0.01 0.011999 0.08 0.01 0.06 0.02 0.02 0.05 −0.11 0.05 0.01 −0.01 0.00 −0.032000 −0.05 0.04 −0.05 0.00 −0.04 0.01 −0.06 −0.02 0.00 −0.02 0.04 −0.052001 0.03 0.10 0.05 0.06 0.05 −0.03 0.03 0.01 0.03 0.04 0.01 0.032002 0.09 0.05 0.11 0.03 0.03 0.05 0.06 0.02 −0.07 −0.02 −0.07 −0.082003 0.25 −0.02 0.01 0.02 0.14 −0.12 −0.09 −0.10 −0.07 −0.10 −0.08 −0.132004 −0.02 −0.25 0.06 −0.12 −0.02 −0.22 −0.04 −0.16 −0.04 −0.03 −0.04 −0.072005 0.09 0.21 0.14 0.21 0.05 0.10 0.09 0.13 −0.03 0.03 −0.03 −0.042006 0.08 0.10 0.08 0.07 0.09 0.08 0.07 0.07 −0.03 0.01 −0.03 −0.022007 0.12 0.11 0.11 0.10 0.01 −0.06 0.02 −0.05 0.01 −0.03 0.04 −0.032008 0.08 0.11 0.07 0.20 −0.13 −0.11 −0.14 −0.02 0.01 0.01 0.02 −0.022009 0.11 0.01 0.08 0.01 0.06 0.04 0.01 0.00 0.00 −0.01 0.01 −0.032010 0.10 −0.07 0.04 0.00 0.04 −0.20 −0.01 −0.05 0.02 0.05 0.01 0.01
62Internet Appendix Table XII
The Dynamic Impact of Mergers In Close Industries: Summary-Level IndustriesThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 forsummary-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Theestimation procedure is as described in Table VII of the main paper. p−values are reported in parentheses. Numbers in parentheses are p-valuesbased on standard errors that are robust to general cross-section and time-series heteroskedasticity and within-group autocorrelation. Statisticalsignificance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6)
Closeness-Weighted M&A Activityt −0.791∗∗∗ −1.228∗∗∗
(0.000) (0.001)Closeness-Weighted M&A Activityt−1 0.535∗∗∗ 0.687∗∗∗ 0.813∗∗∗ 0.850∗∗∗ 0.939∗∗∗ 0.992∗∗∗
(0.001) (0.001) (< 0.001) (0.002) (0.001) (< 0.001)
Closeness-Weighted M&A Activityt−2 −0.156 −0.141 −0.313∗∗∗ −0.281∗∗∗
(0.207) (0.238) (0.003) (0.007)
Closeness-Weighted M&A Activityt−3 −0.149 −0.325∗∗∗ 0.134 −0.076(0.145) (0.002) (0.289) (0.556)
Closeness-Weighted M&A Activityt−4 0.088 −0.045 −0.212 −0.472∗∗∗
(0.481) (0.726) (0.185) (0.008)
Own M&A Activityt−1 9.091∗∗∗ 11.683∗∗∗ 10.632∗∗∗ 9.891∗∗∗ 12.605∗∗∗ 11.853∗∗∗
(0.001) (0.001) (0.003) (0.001) (0.001) (0.001)
Own M&A Activityt−2 3.778 3.299 3.404 2.440(0.162) (0.222) (0.199) (0.357)
Own M&A Activityt−3 2.291 1.256 3.930 2.730(0.365) (0.612) (0.130) (0.263)
Own M&A Activityt−4 1.092 −0.534 0.315 −1.827(0.663) (0.826) (0.903) (0.480)
Own Industry Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yesχ2 470.037 520.493 416.334 420.078 464.934 403.472p−value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 2,852 2,480 2,480 2,852 2,480 2,480
63
Internet Appendix Table XIIIThe Dynamic Impact of Mergers In Close Industries: Single Regressors to Control for MulticollinearityThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 fordetail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Thedependent variable is the log of one plus the number of industry pairs involving industry i that experience high merger activity in year t. Highmerger activity equals one if the log of the inflation-adjusted dollar volume between industries j and k in year t is greater or equal to the 75thpercentile of the industry-pair’s time series of dollar volumes from 1986 to 2010. ‘Closeness-Weighted M&A Activityt’ is
∑
j 6=i1
distij
∑
k 6=ivjkt,
where distij is the shortest path between industry i and j, and vjkt is an indicator variable for high merger activity in industry-pair-years, jkt, asdefined above. The shortest path is measured using either the network based on discrete customer links (columns 1–3) or supplier links (columns4-6), where links are defined as customer or supplier flows greater than 1%. ‘Own M&A Activityt−1’ is the lagged dependent variable at t − 1.‘Own Industry Fixed Effects’ are accounted for through first differencing. Lagged levels of the independent variables are used as instrumentsfor the endogenous first differences. Numbers in parantheses are p-values based on standard errors that are robust to general cross-section andtime-series heteroskedasticity and within-group autocorrelation. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10levels.
Dependent Variable: Industry Merger Activityt
(1) (2) (3) (4) (5) (6) (7) (8)Panel A: Closeness Through Customer Links
Closeness-Weighted M&A Activityi,t 0.284∗∗∗ 0.220∗∗∗
(< 0.001) (< 0.001)
Closeness-Weighted M&A Activityi,t−1 0.615∗∗∗ 0.522∗∗∗
(< 0.001) (< 0.001)
Closeness-Weighted M&A Activityi,t−2 0.163∗∗∗ 0.023(0.001) (0.635)
Closeness-Weighted M&A Activityi,t−3 0.036 −0.115∗∗
(0.497) (0.023)
Own M&A Activityi,t−1 14.052∗∗∗ 13.505∗∗∗ 15.876∗∗∗ 16.887∗∗∗ 20.155∗∗∗ 18.149∗∗∗ 22.457∗∗∗ 23.303∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityi,t−2 10.337∗∗∗ 7.929∗∗∗ 11.708∗∗∗ 12.605∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityi,t−3 6.653∗∗∗ 5.047∗∗∗ 7.393∗∗∗ 8.160∗∗∗
(< 0.001) (0.003) (< 0.001) (< 0.001)
Own M&A Activityi,t−4 2.181 1.289 2.603 3.505∗∗
(0.176) (0.437) (0.111) (0.032)
Own Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes
64Table XIII - Continued
Dependent Variable: Industry Merger Activityt
Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yesχ2 276.865 301.968 263.256 253.856 366.115 365.550 375.623 380.834
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 10,833 10,833 10,362 9,891 9,420 9,420 9,420 9,420
Panel B: Closeness Through Supplier Links
Closeness-Weighted M&A Activityi,t 0.077 −0.119(0.457) (0.299)
Closeness-Weighted M&A Activityi,t−1 1.230∗∗∗ 1.120∗∗∗
(< 0.001) (< 0.001)
Closeness-Weighted M&A Activityi,t−2 0.188∗ −0.103(0.074) (0.305)
Closeness-Weighted M&A Activityi,t−3 0.023 −0.240∗∗
(0.821) (0.012)
Own M&A Activityi,t−1 15.138∗∗∗ 14.930∗∗∗ 16.154∗∗∗ 16.903∗∗∗ 22.101∗∗∗ 18.664∗∗∗ 23.187∗∗∗ 23.334∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityi,t−2 11.419∗∗∗ 7.070∗∗∗ 12.089∗∗∗ 12.641∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityi,t−3 7.168∗∗∗ 4.150∗∗ 7.946∗∗∗ 7.957∗∗∗
(< 0.001) (0.022) (< 0.001) (< 0.001)
Own M&A Activityi,t−4 2.284 0.316 3.061∗ 3.657∗∗
(0.156) (0.854) (0.061) (0.026)
Own Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yesχ2 264.913 280.756 259.602 254.017 352.084 313.793 376.431 383.179
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Observations 10,833 10,833 10,362 9,891 9,420 9,420 9,420 9,420
65
Internet Appendix Table XIVThe Dynamic Impact of Mergers In Close Industries: Additional ControlsThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 fordetail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Theestimation procedure is as described in Table VII of the main paper. Additional variables are as defined in the main paper. p−values arereported in parentheses. Numbers in parentheses are p-values based on standard errors that are robust to general cross-section and time-seriesheteroskedasticity and within-group autocorrelation. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6)
Closeness-Weighted M&A Activityt −0.416∗∗∗ −1.096∗∗∗
(0.001) (< 0.001)
Closeness-Weighted M&A Activityt−1 0.848∗∗∗ 1.048∗∗∗ 0.989∗∗∗ 1.381∗∗∗ 1.645∗∗∗ 1.617∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Closeness-Weighted M&A Activityt−2 −0.416∗∗ −0.333∗ −0.443∗∗ −0.375∗
(0.017) (0.054) (0.024) (0.054)
Closeness-Weighted M&A Activityt−3 −0.070 −0.159 −0.128 −0.295(0.612) (0.251) (0.460) (0.104)
Closeness-Weighted M&A Activityt−4 −0.103 −0.170 −0.166 −0.327∗∗
(0.364) (0.143) (0.311) (0.045)
Industry Economic Shock Index −0.009 0.049 0.049 −0.012 0.041 0.041(0.578) (0.104) (0.101) (0.501) (0.170) (0.173)
Shock Index × C&I Rate Spread 0.029∗ −0.040 −0.042 0.026 −0.038 −0.041(0.074) (0.169) (0.144) (0.114) (0.186) (0.156)
Deregulatory Shock −0.073 −0.071 −0.070 −0.068 −0.055 −0.041(0.197) (0.297) (0.302) (0.220) (0.403) (0.523)
Industry Median M/B 0.009 0.010 0.011 0.007 0.008 0.018(0.744) (0.737) (0.701) (0.810) (0.774) (0.506)
Industry Median R&D 1.276∗∗ 3.358∗∗ 3.600∗∗ 1.291∗∗ 3.852∗∗∗ 4.196∗∗∗
(0.048) (0.020) (0.011) (0.045) (0.008) (0.006)
Industry Mean Return −0.049 −0.032 −0.029 −0.042 −0.026 −0.023(0.137) (0.310) (0.370) (0.216) (0.426) (0.493)
Industry Standard Deviation of Returns 0.035 0.028 0.026 0.028 0.022 0.019
66Internet Appendix Table XIV - Continued
Dependent Variable: Industry Merger Activityt
(0.139) (0.252) (0.293) (0.233) (0.373) (0.449)
Own M&A Activityt−1 13.441∗∗∗ 20.789∗∗∗ 21.230∗∗∗ 15.160∗∗∗ 21.607∗∗∗ 23.785∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−2 8.954∗∗∗ 9.637∗∗∗ 7.336∗∗∗ 9.265∗∗∗
(< 0.001) (< 0.001) (0.003) (< 0.001)
Own M&A Activityt−3 4.458∗ 4.416∗ 2.950 4.048∗
(0.054) (0.051) (0.204) (0.070)
Own M&A Activityt−4 2.394 2.297 1.118 1.813(0.283) (0.287) (0.615) (0.407)
Own Industry Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yesχ2 306.114 381.747 362.269 293.428 391.583 439.889p-value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 3,377 2,879 2,879 3,377 2,879 2,879
67
Internet Appendix Table XVThe Dynamic Impact of Mergers In Close Industries: Industry Assignment Based on All Firm DivisionsThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 fordetail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on all SIC or NAICS codes reported in SDC.The estimation procedure is as described in Table VII of the main paper. p−values are reported in parentheses. Numbers in parentheses arep-values based on standard errors that are robust to general cross-section and time-series heteroskedasticity and within-group autocorrelation.Statistical significance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6)
Closeness-Weighted M&A Activityt 0.015 −0.093∗∗
(0.488) (0.011)
Closeness-Weighted M&A Activityt−1 0.062∗∗∗ 0.066∗∗∗ 0.039 0.117∗∗∗ 0.141∗∗∗ 0.135∗∗∗
(< 0.001) (0.004) (0.161) (0.001) (< 0.001) (< 0.001)
Closeness-Weighted M&A Activityt−2 −0.004 −0.001 −0.035 −0.030(0.873) (0.983) (0.225) (0.280)
Closeness-Weighted M&A Activityt−3 −0.020 −0.017 −0.022 −0.029(0.412) (0.486) (0.484) (0.351)
Closeness-Weighted M&A Activityt−4 −0.011 −0.008 −0.019 −0.035(0.594) (0.710) (0.514) (0.302)
Industry Economic Shock Index −0.028 −0.004 −0.002 −0.028 −0.004 0.001(0.219) (0.933) (0.959) (0.222) (0.932) (0.987)
Shock Index × C&I Rate Spread 0.050∗∗ 0.022 0.021 0.049∗∗ 0.023 0.020(0.018) (0.576) (0.599) (0.021) (0.564) (0.618)
Deregulatory Shock −0.151∗∗ −0.113 −0.108 −0.151∗∗ −0.115 −0.117∗
(0.020) (0.112) (0.130) (0.019) (0.101) (0.091)
Industry Median M/B 0.069∗ 0.074∗∗ 0.079∗∗ 0.071∗∗ 0.076∗∗ 0.084∗∗
(0.051) (0.046) (0.035) (0.046) (0.043) (0.028)
Industry Median R&D 0.639 −2.657 −2.681 0.633 −2.621 −2.533(0.363) (0.188) (0.182) (0.374) (0.204) (0.216)
Industry Mean Return 0.059 0.072∗ 0.072∗ 0.061 0.077∗ 0.077∗
(0.131) (0.072) (0.076) (0.121) (0.059) (0.056)
Industry Standard Deviation of Returns −0.018 −0.022 −0.024 −0.020 −0.025 −0.025
68Internet Appendix Table XV - Continued
Dependent Variable: Industry Merger Activityt
(0.573) (0.502) (0.461) (0.538) (0.453) (0.439)
Own M&A Activityt−1 1.349 −2.093 −2.587 1.949 −1.260 −2.520(0.598) (0.608) (0.522) (0.448) (0.759) (0.541)
Own M&A Activityt−2 −4.087 −4.364 −4.173 −5.189∗
(0.135) (0.109) (0.125) (0.057)
Own M&A Activityt−3 1.163 0.873 1.142 0.092(0.657) (0.736) (0.658) (0.971)
Own M&A Activityt−4 −2.353 −2.618 −2.315 −3.002(0.357) (0.305) (0.373) (0.243)
Own Industry Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yesχ2 −10.805 −9.560 −9.517 −10.767 −9.545 −9.474p-value (−1.541) (0.063) (0.068) (−1.539) (0.067) (0.073)Observations 3,377 2,879 2,879 3,377 2,879 2,879
69
Internet Appendix Table XVIThe Dynamic Impact of Mergers In Close Industries: Horizontal Mergers OnlyThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 fordetail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Thedependent variable is one if industry i experiences high merger activity in year t. High merger activity equals one if the log of the inflation-adjusteddollar volume of intra-industry mergers in industry i in year t is greater or equal to the 75th percentile of the time series of dollar volumes from1986 to 2010. The estimation procedure is as described in Table VII of the main paper. p−values are reported in parentheses. Numbers inparentheses are p-values based on standard errors that are robust to general cross-section and time-series heteroskedasticity and within-groupautocorrelation. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6)
Closeness-Weighted M&A Activityt 0.007 −0.112∗
(0.819) (0.096)
Closeness-Weighted M&A Activityt−1 0.105∗∗∗ 0.114∗∗∗ 0.094∗∗ 0.152∗∗ 0.191∗∗∗ 0.183∗∗
(< 0.001) (0.001) (0.011) (0.014) (0.007) (0.010)
Closeness-Weighted M&A Activityt−2 −0.041 −0.035 −0.081 −0.083∗
(0.271) (0.349) (0.104) (0.098)
Closeness-Weighted M&A Activityt−3 −0.014 −0.015 −0.063 −0.081(0.751) (0.724) (0.352) (0.242)
Closeness-Weighted M&A Activityt−4 −0.020 −0.021 −0.063 −0.089(0.597) (0.582) (0.391) (0.224)
Own M&A Activityt−1 10.953∗∗∗ 12.834∗∗∗ 12.866∗∗∗ 11.263∗∗∗ 12.901∗∗∗ 13.026∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−2 7.075∗∗∗ 7.119∗∗∗ 6.973∗∗∗ 7.038∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−3 4.791∗∗∗ 4.819∗∗∗ 4.762∗∗∗ 4.728∗∗∗
(0.007) (0.007) (0.007) (0.008)
Own M&A Activityt−4 1.209 1.208 1.125 0.973(0.527) (0.527) (0.558) (0.611)
Own Industry Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yesχ2 178.128 233.241 233.621 170.738 220.336 216.694p-value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
70Internet Appendix Table XVI - Continued
Dependent Variable: Industry Merger Activityt
Observations 10,833 9,420 9,420 10,833 9,420 9,420
71
Internet Appendix Table XVIIThe Dynamic Impact of Mergers In Close Industries: Asset ComplementarityThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industriesconnected through customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from1996 to 2008 for detail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primarySIC or NAICS codes. ‘HP-Weighted M&A Activityt’ is
∑
j 6=i HPijt
∑
k 6=i vjkt, where HPijt is an indicator variable that equals one iftwo IO industries have any firms identified as similar in a given year, using their text-based similarity scores in Hoberg and Phillips(2010a) and Hoberg and Phillips (2010b). High merger activity, vjkt, equals one if the log of the inflation-adjusted dollar volumebetween industries j and k in year t is greater or equal to the 75th percentile of the industry-pair’s time series of dollar volumesfrom 1986 to 2010. Thus, this variable measures the amount of merger activity in close industries, where closeness is based on assetcomplementarity. The estimation procedure is as described in Table VII of the main paper. p−values are reported in parentheses.Numbers in parentheses are p-values based on standard errors that are robust to general cross-section and time-series heteroskedasticityand within-group autocorrelation. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6)
Closeness-Weighted M&A Activityt −0.257∗∗∗ −0.954∗∗∗
(0.003) (< 0.001)
Closeness-Weighted M&A Activityt−1 0.347∗∗∗ 0.370∗∗∗ 0.133 0.848∗∗∗ 0.736∗∗∗ 0.714∗∗∗
(< 0.001) (< 0.001) (0.197) (< 0.001) (0.001) (0.001)
Closeness-Weighted M&A Activityt−2 −0.154 −0.007 −0.248∗ −0.241(0.111) (0.950) (0.100) (0.123)
Closeness-Weighted M&A Activityt−3 −0.098 −0.134 0.009 −0.313∗∗
(0.284) (0.153) (0.946) (0.047)
Closeness-Weighted M&A Activityt−4 0.153 −0.266∗∗ 0.142 −0.455∗∗
(0.105) (0.016) (0.469) (0.025)
HP-Weighted M&A Activityt 0.092∗∗∗ 0.105∗∗∗
(< 0.001) (< 0.001)
HP-Weighted M&A Activityt−1 0.087∗∗∗ 0.067∗∗∗ −0.008 0.081∗∗∗ 0.061∗∗∗ −0.020(< 0.001) (< 0.001) (0.653) (< 0.001) (0.001) (0.291)
HP-Weighted M&A Activityt−2 −0.040∗∗ −0.028 −0.041∗∗∗ −0.028(0.014) (0.105) (0.010) (0.110)
HP-Weighted M&A Activityt−3 0.008 0.029∗ 0.006 0.025(0.613) (0.090) (0.691) (0.133)
72Internet Appendix Table XVII - Continued
Dependent Variable: Industry Merger Activityt
HP-Weighted M&A Activityt−4 −0.027 −0.018 −0.024 −0.019(0.110) (0.320) (0.156) (0.271)
Own M&A Activityt−1 14.598∗∗∗ 10.811∗∗∗ 12.312∗∗∗ 15.642∗∗∗ 11.115∗∗∗ 13.754∗∗∗
(< 0.001) (0.001) (0.001) (< 0.001) (0.001) (< 0.001)
Own M&A Activityt−2 1.066 2.795 −0.016 1.923(0.688) (0.298) (0.995) (0.464)
Own M&A Activityt−3 1.577 2.095 1.176 1.316(0.567) (0.460) (0.663) (0.632)
Own M&A Activityt−4 −0.539 −0.316 −0.914 −0.984(0.838) (0.908) (0.726) (0.711)
Own Industry Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yesχ2 245.347 163.394 95.751 233.854 163.787 101.528p-value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 5,652 4,239 3,768 5,652 4,239 3,768
73
Internet Appendix Table XVIIIThe Dynamic Impact of Mergers In Close Industries: Asset Complementarity and R&DThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industriesconnected through customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from1996 to 2008 for detail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primarySIC or NAICS codes. ‘HP-Weighted M&A Activityt’ is
∑
j 6=i HPijt
∑
k 6=i vjkt, where HPijt is an indicator variable that equals one iftwo IO industries have any firms identified as similar in a given year, using their text-based similarity scores in Hoberg and Phillips(2010a) and Hoberg and Phillips (2010b). High merger activity, vjkt, equals one if the log of the inflation-adjusted dollar volumebetween industries j and k in year t is greater or equal to the 75th percentile of the industry-pair’s time series of dollar volumesfrom 1986 to 2010. Thus, this variable measures the amount of merger activity in close industries, where closeness is based on assetcomplementarity. The estimation procedure is as described in Table VII of the main paper. p−values are reported in parentheses.Numbers in parentheses are p-values based on standard errors that are robust to general cross-section and time-series heteroskedasticityand within-group autocorrelation. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6)
Closeness-Weighted M&A Activityt −0.484∗∗ −1.246∗∗∗
(0.013) (< 0.001)
Closeness-Weighted M&A Activityt−1 0.631∗∗∗ 0.541∗∗ 0.453∗∗ 1.232∗∗∗ 1.082∗∗∗ 0.914∗∗∗
(< 0.001) (0.010) (0.033) (< 0.001) (0.001) (0.004)
Closeness-Weighted M&A Activityt−2 −0.150 −0.153 −0.204 −0.337(0.485) (0.473) (0.419) (0.174)
Closeness-Weighted M&A Activityt−3 −0.038 −0.199 0.267 −0.110(0.834) (0.276) (0.230) (0.651)
Closeness-Weighted M&A Activityt−4 −0.056 −0.326 0.196 −0.482(0.769) (0.108) (0.519) (0.150)
HP-Weighted M&A Activityt 0.063 0.095∗∗
(0.117) (0.021)
HP-Weighted M&A Activityt−1 0.012 −0.020 −0.037 0.002 −0.036 −0.052(0.680) (0.588) (0.278) (0.950) (0.322) (0.121)
HP-Weighted M&A Activityt−2 −0.056 −0.054 −0.059 −0.055(0.127) (0.144) (0.101) (0.140)
HP-Weighted M&A Activityt−3 0.054 0.066 0.045 0.053(0.190) (0.104) (0.277) (0.194)
74Internet Appendix Table XVIII - Continued
Dependent Variable: Industry Merger Activityt
HP-Weighted M&A Activityt−4 −0.095∗∗∗ −0.075∗∗ −0.105∗∗∗ −0.080∗∗
(0.010) (0.049) (0.005) (0.036)
Industry Economic Shock Index 0.081∗ 0.157∗ 0.157∗ 0.071∗ 0.147∗ 0.174∗∗
(0.059) (0.072) (0.070) (0.097) (0.084) (0.042)
Shock Index × C&I Rate Spread −0.075∗ −0.149∗ −0.152∗ −0.072∗ −0.148∗ −0.174∗∗
(0.065) (0.065) (0.059) (0.073) (0.065) (0.029)
Deregulatory Shock 0.059 0.099(0.618) (0.369)
Industry Median M/B 0.028 0.031 0.030 0.028 0.024 0.038(0.405) (0.562) (0.567) (0.385) (0.630) (0.453)
Industry Median R&D 2.736∗ 2.637 2.928 3.190∗ 3.077 3.634(0.086) (0.187) (0.142) (0.054) (0.140) (0.110)
Industry Mean Return −0.045 −0.074 −0.078 −0.038 −0.067 −0.077(0.288) (0.259) (0.242) (0.376) (0.325) (0.282)
Industry Standard Deviation of Returns 0.027 0.021 0.026 0.022 0.012 0.021(0.439) (0.701) (0.644) (0.517) (0.824) (0.707)
Own M&A Activityt−1 17.153∗∗∗ 13.862∗∗∗ 13.965∗∗∗ 19.224∗∗∗ 15.078∗∗∗ 15.103∗∗∗
(< 0.001) (0.007) (0.006) (< 0.001) (0.005) (0.004)
Own M&A Activityt−2 2.736 3.349 1.295 1.720(0.462) (0.371) (0.727) (0.634)
Own M&A Activityt−3 0.496 0.470 0.378 0.055(0.902) (0.907) (0.924) (0.989)
Own M&A Activityt−4 0.174 −0.379 −0.096 −1.684(0.962) (0.917) (0.979) (0.633)
Own Industry Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yesχ2 194.847 101.504 106.640 198.550 107.198 125.390p-value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 1,720 1,225 1,225 1,720 1,225 1,225
75
Internet Appendix Table XIXThe Dynamic Impact of Mergers In Close Industries: Asset Complementarity in Summary-Level IndustriesThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industriesconnected through customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from1996 to 2008 for summary-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primarySIC or NAICS codes. ‘HP-Weighted M&A Activityt’ is
∑
j 6=i HPijt
∑
k 6=i vjkt, where HPijt is an indicator variable that equals one iftwo IO industries have any firms identified as similar in a given year, using their text-based similarity scores in Hoberg and Phillips(2010a) and Hoberg and Phillips (2010b). High merger activity, vjkt, equals one if the log of the inflation-adjusted dollar volumebetween industries j and k in year t is greater or equal to the 75th percentile of the industry-pair’s time series of dollar volumesfrom 1986 to 2010. Thus, this variable measures the amount of merger activity in close industries, where closeness is based on assetcomplementarity. The estimation procedure is as described in Table VII of the main paper. p−values are reported in parentheses.Numbers in parentheses are p-values based on standard errors that are robust to general cross-section and time-series heteroskedasticityand within-group autocorrelation. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6)
Closeness-Weighted M&A Activityt −0.714∗∗∗ −1.195∗∗∗
(0.004) (0.002)
Closeness-Weighted M&A Activityt−1 0.472∗∗∗ 0.594∗∗∗ 0.455∗ 0.772∗∗∗ 0.556∗∗ 0.600∗∗
(0.005) (0.002) (0.068) (0.005) (0.028) (0.027)
Closeness-Weighted M&A Activityt−2 −0.099 0.245 −0.219 −0.015(0.637) (0.305) (0.205) (0.937)
Closeness-Weighted M&A Activityt−3 −0.297 −0.605∗∗∗ 0.236 −0.215(0.176) (0.006) (0.215) (0.288)
Closeness-Weighted M&A Activityt−4 0.366∗ −0.105 −0.121 −0.668∗∗
(0.071) (0.706) (0.550) (0.012)
HP-Weighted M&A Activityt 0.064 0.085∗
(0.140) (0.063)
HP-Weighted M&A Activityt−1 0.053 0.080∗ 0.015 0.049 0.078∗ −0.006(0.167) (0.065) (0.757) (0.200) (0.074) (0.898)
HP-Weighted M&A Activityt−2 −0.019 0.000 −0.012 0.021(0.654) (0.993) (0.772) (0.648)
HP-Weighted M&A Activityt−3 −0.056 −0.036 −0.075∗ −0.053(0.148) (0.336) (0.054) (0.155)
76Internet Appendix Table XIX - Continued
Dependent Variable: Industry Merger Activityt
HP-Weighted M&A Activityt−4 0.056 0.066 0.074∗ 0.076∗
(0.189) (0.128) (0.080) (0.083)
Own M&A Activityt−1 12.150∗∗∗ 8.060 5.687 12.964∗∗∗ 7.985 7.535(0.003) (0.146) (0.313) (0.002) (0.147) (0.172)
Own M&A Activityt−2 −4.650 −2.183 −5.184 −1.984(0.270) (0.603) (0.212) (0.630)
Own M&A Activityt−3 −2.875 −4.190 −0.686 −1.769(0.458) (0.291) (0.863) (0.668)
Own M&A Activityt−4 −6.463∗ −7.452∗∗ −7.699∗∗ −9.023∗∗
(0.073) (0.044) (0.036) (0.013)
Own Industry Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yesχ2 303.538 201.638 149.371 290.786 203.876 152.512p-value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 1,488 1,116 992 1,488 1,116 992
77
Internet Appendix Table XXThe Dynamic Impact of Mergers In Close Industries on Acquirer ReturnsThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 fordetail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Thedependent variable is the log of one plus the number of industry pairs involving industry i that experience high acquirer returns in year t. Highacquirer returns equals one if the three-day cumulative abnormal return (relative to the CRSP value-weighted index) of the acquirer in mergersbetween industries j and k in year t is greater or equal to the 75th percentile of the industry-pair’s time series of acquirer abnormal returnsfrom 1986 to 2010. ‘Closeness-Weighted Acquirer Returnst’ is
∑
j 6=i1
distij
∑
k 6=ivjkt, where distij is the shortest path between industry i and
j, and vjkt is an indicator variable for high acquirer returns in industry-pair-years, jkt, as defined above. The shortest path is measured usingeither the network based on discrete customer links (columns 1–3) or supplier links (columns 4-6), where links are defined as customer or supplierflows greater than 1%. ‘Own Acquirer Returnst−1’ is the lagged dependent variable at t − 1. ‘Own Industry Fixed Effects’ are accounted forthrough first differencing. Lagged levels of the independent variables are used as instruments for the endogenous first differences. Numbers inparantheses are p-values based on standard errors that are robust to general cross-section and time-series heteroskedasticity and within-groupautocorrelation. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Average Acquirer Returnst
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6)
Closeness-Weighted Acquirer Returnsi,t −0.813∗∗∗ −2.674∗∗∗
(0.001) (< 0.001)
Closeness-Weighted Acquirer Returnsi,t−1 2.006∗∗∗ 1.586∗∗∗ 1.048∗∗∗ 3.704∗∗∗ 3.390∗∗∗ 2.085∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Closeness-Weighted Acquirer Returnsi,t−2 −0.022 −0.079 0.578 0.068(0.901) (0.654) (0.142) (0.817)
Closeness-Weighted Acquirer Returnsi,t−3 0.199 0.190 0.012 −0.177(0.232) (0.251) (0.974) (0.590)
Closeness-Weighted Acquirer Returnsi,t−4 −0.339∗ −0.255 −0.323 −0.076(0.067) (0.163) (0.269) (0.784)
Own Acquirer Returnsi,t−1 12.795∗∗∗ 16.265∗∗∗ 18.390∗∗∗ 15.185∗∗∗ 17.311∗∗∗ 21.009∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own Acquirer Returnsi,t−2 7.222∗∗∗ 9.279∗∗∗ 6.018∗∗ 10.014∗∗∗
(0.003) (< 0.001) (0.016) (< 0.001)
Own Acquirer Returnsi,t−3 10.260∗∗∗ 12.002∗∗∗ 8.716∗∗∗ 11.833∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own Acquirer Returnsi,t−4 3.216 4.548∗∗ 2.413 5.026∗∗
(0.113) (0.028) (0.254) (0.013)
78Table XX - Continued
Dependent Variable: Industry Average Acquirer Returnst
Own Industry Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yesχ2 151.374 210.264 236.804 186.480 231.556 285.212
0.000 0.000 0.000 0.000 0.000 0.000Observations 10,833 9,420 9,420 10,833 9,420 9,420
79
Internet Appendix Table XXICorrelations Between Time Lags and Network DistanceThis table presents correlation coefficients of merger activity across different lengths of industry connections and time lags. Industries are definedby the Bureau of Economic Analysis Input-Output (IO) Industry classification using the 1997 reports for the Detailed-Industry level (Panel A)and the Summary-Level (Panel B). Correlations are based on either customer or supplier relations.
Panel A: Detail-Level
Customer Network
1stt−1 1stt−2 1stt−3 1stt−4 2ndt−1 2ndt−2 2ndt−3 2ndt−4 3rdt−1 3rdt−2 3rdt−3 3rdt−4 4tht−1 4tht−2 4tht−3 4tht−4
1st degreet−1 1.001st degreet−2 0.94 1.001st degreet−3 0.86 0.94 1.001st degreet−4 0.79 0.88 0.96 1.002nd degreet−1 0.55 0.52 0.45 0.38 1.002nd degreet−2 0.49 0.55 0.52 0.48 0.91 1.002nd degreet−3 0.39 0.49 0.55 0.54 0.75 0.91 1.002nd degreet−4 0.32 0.41 0.50 0.55 0.61 0.79 0.94 1.003rd degreet−1 −0.36 −0.38 −0.42 −0.44 −0.08 −0.13 −0.23 −0.30 1.003rd degreet−2 −0.38 −0.37 −0.39 −0.41 −0.13 −0.09 −0.14 −0.21 0.93 1.003rd degreet−3 −0.42 −0.38 −0.36 −0.38 −0.24 −0.14 −0.09 −0.12 0.82 0.93 1.003rd degreet−4 −0.44 −0.41 −0.37 −0.36 −0.30 −0.21 −0.11 −0.07 0.74 0.86 0.95 1.004th degreet−1 −0.26 −0.26 −0.26 −0.26 −0.37 −0.39 −0.40 −0.40 0.34 0.32 0.29 0.28 1.004th degreet−2 −0.26 −0.26 −0.26 −0.26 −0.38 −0.38 −0.39 −0.39 0.33 0.34 0.32 0.30 0.97 1.004th degreet−3 −0.26 −0.26 −0.26 −0.26 −0.40 −0.39 −0.38 −0.38 0.30 0.33 0.34 0.33 0.93 0.97 1.004th degreet−4 −0.27 −0.27 −0.26 −0.26 −0.40 −0.40 −0.38 −0.37 0.28 0.31 0.34 0.34 0.91 0.95 0.98 1.00
Supplier Network
1stt−1 1stt−2 1stt−3 1stt−4 2ndt−1 2ndt−2 2ndt−3 2ndt−4 3rdt−1 3rdt−2 3rdt−3 3rdt−4
1st degreet−1 1.001st degreet−2 0.92 1.001st degreet−3 0.82 0.92 1.001st degreet−4 0.72 0.84 0.93 1.002nd degreet−1 0.36 0.30 0.18 0.07 1.002nd degreet−2 0.27 0.36 0.30 0.22 0.86 1.002nd degreet−3 0.10 0.26 0.36 0.33 0.60 0.86 1.002nd degreet−4 −0.01 0.14 0.30 0.37 0.38 0.68 0.91 1.003rd degreet−1 −0.14 −0.15 −0.17 −0.19 0.17 0.13 0.06 0.02 1.003rd degreet−2 −0.17 −0.14 −0.15 −0.16 0.11 0.17 0.13 0.08 0.93 1.003rd degreet−3 −0.24 −0.18 −0.14 −0.14 −0.02 0.11 0.17 0.15 0.86 0.93 1.003rd degreet−4 −0.28 −0.23 −0.17 −0.13 −0.09 0.01 0.13 0.17 0.82 0.87 0.94 1.00
Panel B: Summary-Level
Customer Network
80Internet Appendix Table XXI - Continued
1stt−1 1stt−2 1stt−3 1stt−4 2ndt−1 2ndt−2 2ndt−3 2ndt−4 3rdt−1 3rdt−2 3rdt−3 3rdt−4 4tht−1 4tht−2 4tht−3 4tht−4
1st degreet−1 1.001st degreet−2 0.92 1.001st degreet−3 0.80 0.92 1.001st degreet−4 0.69 0.83 0.94 1.002nd degreet−1 0.46 0.40 0.27 0.15 1.002nd degreet−2 0.36 0.45 0.39 0.30 0.85 1.002nd degreet−3 0.21 0.36 0.45 0.42 0.60 0.85 1.002nd degreet−4 0.08 0.24 0.39 0.46 0.38 0.66 0.90 1.003rd degreet−1 −0.39 −0.41 −0.44 −0.46 0.09 0.05 −0.03 −0.09 1.00
3rd degreet−2 −0.41 −0.39 −0.41 −0.43 0.04 0.09 0.05 −0.01 0.95 1.003rd degreet−3 −0.44 −0.41 −0.39 −0.40 −0.06 0.04 0.10 0.07 0.88 0.95 1.003rd degreet−4 −0.46 −0.43 −0.40 −0.38 −0.12 −0.03 0.07 0.10 0.82 0.90 0.97 1.004th degreet−1 −0.19 −0.19 −0.20 −0.20 −0.11 −0.11 −0.13 −0.14 0.50 0.49 0.46 0.43 1.004th degreet−2 −0.19 −0.19 −0.19 −0.19 −0.11 −0.11 −0.11 −0.12 0.50 0.50 0.49 0.46 0.89 1.004th degreet−3 −0.19 −0.19 −0.19 −0.19 −0.13 −0.11 −0.10 −0.11 0.47 0.50 0.50 0.49 0.86 0.89 1.004th degreet−4 −0.20 −0.19 −0.19 −0.19 −0.14 −0.13 −0.11 −0.10 0.45 0.48 0.50 0.50 0.83 0.86 0.89 1.00
Supplier Network
1stt−1 1stt−2 1stt−3 1stt−4 2ndt−1 2ndt−2 2ndt−3 2ndt−4 3rdt−1 3rdt−2 3rdt−3 3rdt−4
1st degreet−1 1.001st degreet−2 0.92 1.001st degreet−3 0.78 0.92 1.00
1st degreet−4 0.67 0.82 0.94 1.002nd degreet−1 0.18 0.10 −0.04 −0.15 1.002nd degreet−2 0.09 0.17 0.10 0.00 0.85 1.002nd degreet−3 −0.08 0.08 0.18 0.13 0.59 0.84 1.002nd degreet−4 −0.20 −0.03 0.12 0.19 0.37 0.65 0.90 1.003rd degreet−1 −0.08 −0.09 −0.09 −0.09 0.04 0.03 0.02 0.01 1.003rd degreet−2 −0.08 −0.08 −0.09 −0.09 0.03 0.04 0.03 0.02 0.97 1.003rd degreet−3 −0.09 −0.09 −0.08 −0.08 0.02 0.03 0.04 0.03 0.93 0.97 1.003rd degreet−4 −0.09 −0.09 −0.08 −0.08 0.01 0.02 0.04 0.04 0.89 0.93 0.97 1.00
81
Internet Appendix Table XXIIThe Diffusion of Merger Activity Across Close and Distant Industries: Summary-Level IndustriesThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 forsummary-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Theestimation procedure is as described in Table VIII of the main paper. p−values are reported in parentheses. Numbers in parentheses are p-valuesbased on standard errors that are robust to general cross-section and time-series heteroskedasticity and within-group autocorrelation. Statisticalsignificance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6) (7) (8)
Close M&A Activityt −0.054 −0.362∗∗∗
(0.563) (0.001)
Close M&A Activityt−1 0.397∗∗∗ 0.399∗∗∗ 0.332∗∗∗ 0.397∗∗∗ 0.464∗∗∗ 0.403∗∗∗
(0.000) (0.000) (0.001) (0.000) (0.000) (0.000)
Close M&A Activityt−2 −0.089 −0.098 −0.086 −0.089 −0.050 −0.066(0.318) (0.274) (0.342) (0.318) (0.592) (0.488)
Close M&A Activityt−3 −0.063 −0.050 −0.068 −0.063 −0.035 −0.078(0.390) (0.498) (0.362) (0.390) (0.667) (0.334)
Close M&A Activityt−4 −0.065 −0.058 −0.058 −0.065 −0.127 −0.182∗
(0.480) (0.536) (0.549) (0.480) (0.204) (0.068)
Distant M&A Activityt 3.164∗ −0.580∗∗∗
(0.053) (0.000)
Distant M&A Activityt−1 −3.195∗ −2.610 −1.703 −0.088 0.120 0.281∗∗∗
(0.081) (0.151) (0.230) (0.232) (0.143) (0.001)
Distant M&A Activityt−2 2.944∗∗ 3.266∗∗ 3.736∗∗ 0.024 0.011 −0.054(0.045) (0.035) (0.016) (0.747) (0.895) (0.490)
Distant M&A Activityt−3 −0.607 −0.995 −0.383 0.114 0.087 0.056(0.767) (0.631) (0.861) (0.225) (0.385) (0.569)
Distant M&A Activityt−4 1.869 1.371 1.984 −0.061 −0.147 −0.242∗∗
(0.441) (0.576) (0.395) (0.519) (0.144) (0.018)
Own M&A Activityt−1 10.050∗∗∗ 10.725∗∗∗ 10.271∗∗∗ 10.048∗∗∗ 10.050∗∗∗ 9.831∗∗∗ 10.152∗∗∗ 10.665∗∗∗
(0.004) (0.002) (0.003) (0.003) (0.004) (0.005) (0.004) (0.002)
Own M&A Activityt−2 4.379∗ 4.748∗ 4.160 3.924 4.379∗ 4.753∗ 4.346 4.373
82Internet Appendix Table XXII - Continued
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6) (7) (8)
(0.099) (0.074) (0.116) (0.137) (0.099) (0.080) (0.103) (0.103)
Own M&A Activityt−3 3.001 3.516 3.258 2.937 3.001 3.244 2.989 3.532(0.229) (0.155) (0.193) (0.241) (0.229) (0.203) (0.244) (0.164)
Own M&A Activityt−4 0.953 0.796 0.853 0.455 0.953 0.762 0.419 0.962(0.713) (0.757) (0.744) (0.861) (0.713) (0.773) (0.874) (0.713)
Own Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yesχ2 533.773 456.754 568.414 569.984 533.773 458.917 528.344 562.471p-value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 2,480 2,480 2,480 2,480 2,480 2,480 2,480 2,480
83
Internet Appendix Table XXIIIThe Diffusion of Merger Activity Across Close and Distant Industries: Additional ControlsThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 fordetail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Theestimation procedure is as described in Table VIII of the main paper. p−values are reported in parentheses. Numbers in parentheses are p-valuesbased on standard errors that are robust to general cross-section and time-series heteroskedasticity and within-group autocorrelation. Statisticalsignificance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6) (7) (8)
Close M&A Activityt 0.373∗∗∗ 0.269∗∗∗
(< 0.001) (0.006)
Close M&A Activityt−1 0.407∗∗∗ 0.371∗∗∗ 0.073 0.540∗∗∗ 0.546∗∗∗ 0.284∗∗∗
(< 0.001) (< 0.001) (0.481) (< 0.001) (< 0.001) (0.010)
Close M&A Activityt−2 −0.138 −0.137 −0.139 −0.277∗∗ −0.263∗∗ −0.289∗∗
(0.213) (0.214) (0.201) (0.013) (0.023) (0.011)
Close M&A Activityt−3 −0.068 −0.079 0.023 −0.022 0.048 0.121(0.496) (0.433) (0.825) (0.821) (0.627) (0.254)
Close M&A Activityt−4 −0.129 −0.139 −0.114 −0.071 0.003 0.032(0.124) (0.113) (0.200) (0.437) (0.974) (0.758)
Distant M&A Activityt 0.256 −0.220∗
(0.296) (0.075)
Distant M&A Activityt−1 −0.539∗∗ −0.385∗ −0.231 −0.031 0.188 0.216∗
(0.015) (0.082) (0.273) (0.809) (0.142) (0.094)
Distant M&A Activityt−2 −0.171 −0.088 −0.214 −0.089 −0.077 −0.129(0.343) (0.638) (0.263) (0.476) (0.541) (0.304)
Distant M&A Activityt−3 0.154 0.054 0.139 0.289∗∗ 0.246∗∗ 0.233∗
(0.409) (0.781) (0.476) (0.016) (0.049) (0.054)
Distant M&A Activityt−4 0.059 −0.106 −0.137 0.319∗∗ 0.300∗∗ 0.245∗
(0.761) (0.614) (0.537) (0.015) (0.033) (0.083)
Industry Economic Shock Index 0.047∗∗∗ 0.047∗∗∗ 0.046∗∗∗ 0.043∗∗∗ 0.041∗∗∗ 0.041∗∗∗ 0.038∗∗∗ 0.038∗∗∗
(0.112) (0.124) (0.125) (0.163) (0.178) (0.179) (0.208) (0.213)
Shock Index × C&I Rate Spread −0.037∗∗∗ −0.037∗∗∗ −0.036∗∗∗ −0.033∗∗∗ −0.034∗∗∗ −0.032∗∗∗ −0.031∗∗∗ −0.031∗∗∗
84Internet Appendix Table XXIII - Continued
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6) (7) (8)
(0.197) (0.206) (0.212) (0.259) (0.245) (0.270) (0.285) (0.290)
Deregulatory Shock −0.069∗∗ −0.063∗∗ −0.068∗∗ −0.057∗∗ −0.056∗∗ −0.061∗∗ −0.055∗∗ −0.057∗
(0.311) (0.358) (0.318) (0.406) (0.402) (0.374) (0.409) (0.397)
Industry Median M/B 0.015 0.012 0.015 0.018 0.012 0.014 0.009 0.007(0.611) (0.684) (0.607) (0.536) (0.677) (0.634) (0.754) (0.794)
Industry Median R&D 3.384∗∗∗ 3.228∗∗∗ 3.329∗∗∗ 3.048∗∗∗ 3.712∗∗∗ 3.302∗∗∗ 3.791∗∗∗ 3.491∗∗∗
(0.019) (0.023) (0.021) (0.039) (0.009) (0.018) (0.008) (0.012)
Industry Mean Return −0.026∗∗∗ −0.023∗∗∗ −0.026∗∗∗ −0.025∗∗∗ −0.025∗∗∗ −0.024∗∗∗ −0.028∗∗∗ −0.029∗∗∗
(0.408) (0.459) (0.404) (0.432) (0.430) (0.438) (0.369) (0.342)
Industry Standard Deviation of Returns 0.025∗∗∗ 0.020∗∗∗ 0.025∗∗∗ 0.019∗∗∗ 0.022∗∗∗ 0.024∗∗∗ 0.027∗∗∗ 0.025∗∗∗
(0.318) (0.413) (0.314) (0.451) (0.368) (0.332) (0.267) (0.297)
Own M&A Activityt−1 19.537∗∗∗ 20.911∗∗∗ 19.413∗∗∗ 18.737∗∗∗ 19.586∗∗∗ 20.584∗∗∗ 18.363∗∗∗ 17.940∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−2 10.504∗∗∗ 11.418∗∗∗ 10.467∗∗∗ 9.950∗∗∗ 10.024∗∗∗ 11.336∗∗∗ 9.289∗∗∗ 8.822∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−3 4.997∗∗ 4.976∗∗ 4.946∗∗ 4.685∗∗ 4.740∗∗ 5.495∗∗ 4.570∗∗ 4.327∗
(0.027) (0.022) (0.030) (0.039) (0.039) (0.012) (0.047) (0.057)
Own M&A Activityt−4 2.827 2.578 2.972 2.809 2.413 3.408 2.466 2.183(0.204) (0.234) (0.186) (0.210) (0.297) (0.125) (0.291) (0.341)
Own Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yesχ2 391.013 387.499 401.028 437.706 447.683 386.016 463.715 462.266p-value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 2,879 2,879 2,879 2,879 2,879 2,879 2,879 2,879
85
Internet Appendix Table XXIVThe Diffusion of Merger Activity Across Close and Distant Industries: Horizontal Mergers OnlyThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 fordetail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Thedependent variable is one if industry i experiences high merger activity in year t. High merger activity equals one if the log of the inflation-adjusteddollar volume of intra-industry mergers in industry i in year t is greater or equal to the 75th percentile of the time series of dollar volumes from1986 to 2010. The estimation procedure is as described in Table VIII of the main paper. p−values are reported in parentheses. Numbers inparentheses are p-values based on standard errors that are robust to general cross-section and time-series heteroskedasticity and within-groupautocorrelation. Statistical significance is indicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6) (7) (8)
Close M&A Activityt 0.072∗∗ 0.040(0.029) (0.295)
Close M&A Activityt−1 0.097∗∗∗ 0.094∗∗∗ 0.063∗ 0.093∗∗ 0.109∗∗ 0.082∗
(0.004) (0.007) (0.071) (0.036) (0.015) (0.075)
Close M&A Activityt−2 −0.033 −0.036 −0.037 −0.042 −0.042 −0.047(0.356) (0.319) (0.316) (0.149) (0.174) (0.131)
Close M&A Activityt−3 0.000 0.004 0.020 −0.045 −0.040 −0.030(0.994) (0.911) (0.629) (0.271) (0.351) (0.481)
Close M&A Activityt−4 −0.043 −0.043 −0.028 −0.076∗ −0.080∗ −0.075∗
(0.175) (0.182) (0.399) (0.092) (0.080) (0.089)
Distant M&A Activityt −0.005 −0.083∗∗
(0.852) (0.026)
Distant M&A Activityt−1 −0.011 0.012 0.023 0.044 0.069∗∗ 0.072∗∗
(0.726) (0.714) (0.516) (0.208) (0.050) (0.026)
Distant M&A Activityt−2 −0.034 −0.034 −0.045 0.005 −0.002 −0.010(0.339) (0.341) (0.232) (0.860) (0.961) (0.757)
Distant M&A Activityt−3 0.031 0.021 0.029 0.022 −0.010 −0.020(0.471) (0.635) (0.507) (0.552) (0.777) (0.588)
Distant M&A Activityt−4 0.027 0.019 0.014 0.049 0.020 0.002(0.592) (0.704) (0.793) (0.206) (0.620) (0.962)
Own M&A Activityt−1 12.674∗∗∗ 13.233∗∗∗ 12.676∗∗∗ 12.423∗∗∗ 12.512∗∗∗ 13.241∗∗∗ 12.491∗∗∗ 12.389∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
86Internet Appendix Table XXIV - Continued
Dependent Variable: Industry Merger Activityt
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6) (7) (8)
Own M&A Activityt−2 7.028∗∗∗ 7.378∗∗∗ 7.045∗∗∗ 6.879∗∗∗ 6.950∗∗∗ 7.369∗∗∗ 6.913∗∗∗ 6.862∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own M&A Activityt−3 4.843∗∗∗ 4.865∗∗∗ 4.856∗∗∗ 4.818∗∗∗ 4.952∗∗∗ 4.888∗∗∗ 4.972∗∗∗ 5.035∗∗∗
(0.006) (0.006) (0.006) (0.007) (0.005) (0.006) (0.005) (0.004)
Own M&A Activityt−4 1.351 1.041 1.335 1.368 1.338 1.056 1.296 1.413(0.479) (0.587) (0.484) (0.473) (0.484) (0.582) (0.498) (0.459)
Own Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yesχ2 226.828 209.965 230.543 244.943 211.903 214.518 227.659 235.161p-value (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 9,420 9,420 9,420 9,420 9,420 9,420 9,420 9,420
87
Internet Appendix Table XXVThe Diffusion of Acquirer Returns Across Close and Distant IndustriesThis table reports results from Arellano and Bond (1991) GMM regressions to estimate the effect of mergers that occur in industries connectedthrough customer-supplier links, but do not include mergers that involve the subject industry itself. Observations are from 1986 to 2010 fordetail-level IO industries defined in the 1997 IO reports, where firms are matched to industries based on primary SIC or NAICS codes. Thedependent variable is the log of one plus the number of industry pairs involving industry i that experience high acquirer returns in year t. Highacquirer returns equals one if the three-day cumulative abnormal return (relative to the CRSP value-weighted index) of the acquirer in mergersbetween industries j and k in year t is greater or equal to the 75th percentile of the industry-pair’s time series of acquirer abnormal returns from1986 to 2010. ‘Close Acquirer Returnst’ is average acquire returns in close industries, not including mergers with firms in industry i. ‘DistantAcquirer Returnst’ is defined analogously. Close industries are defined as the industries that are directly connected to the subject industrythrough a customer or supplier link above the one percent threshold. Distant industries are those that have the maximum shortest path fromthe subject industry. In the supplier network, this is three connections away. In the customer network, it is four connections. ‘Own AcquirerReturnst−1’ is the lagged dependent variable at t− 1. ‘Own Industry Fixed Effects’ are accounted for through first differencing. Lagged levels ofthe independent variables are used as instruments for the endogenous first differences. Numbers in parentheses are p-values based on standarderrors that are robust to general cross-section and time-series heteroskedasticity and within-group autocorrelation. Statistical significance isindicated by ∗∗∗, ∗∗, and ∗ for the 0.01, 0.05, and 0.10 levels.
Dependent Variable: Industry Average Acquirer Returnst
Closeness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6) (7) (8)
Close Acquirer Returnsi,t 1.002∗∗∗ 0.869∗∗∗
(< 0.001) (< 0.001)
Close Acquirer Returnsi,t−1 0.501∗∗ 0.489∗∗ 0.188 0.772∗∗∗ 0.838∗∗∗ 0.560∗∗∗
(0.015) (0.019) (0.271) (< 0.001) (< 0.001) (0.001)
Close Acquirer Returnsi,t−2 0.181 0.179 0.097 −0.004 0.040 −0.057(0.234) (0.250) (0.523) (0.982) (0.802) (0.712)
Close Acquirer Returnsi,t−3 0.158 0.160 0.141 −0.209 −0.246 −0.249(0.244) (0.240) (0.293) (0.232) (0.168) (0.131)
Close Acquirer Returnsi,t−4 −0.369∗∗ −0.361∗∗ −0.323∗∗ −0.254 −0.245 −0.247(0.030) (0.036) (0.049) (0.131) (0.153) (0.147)
Distant Acquirer Returnsi,t −0.199∗∗ −0.007(0.049) (0.981)
Distant Acquirer Returnsi,t−1 −0.233∗∗ −0.124 0.051 0.128 0.269 0.367(0.032) (0.233) (0.591) (0.622) (0.316) (0.127)
Distant Acquirer Returnsi,t−2 0.018 0.089 0.089 0.461∗∗ 0.578∗∗∗ 0.381(0.843) (0.337) (0.319) (0.025) (0.009) (0.106)
Distant Acquirer Returnsi,t−3 −0.039 −0.017 −0.076 −0.168 −0.330 −0.311
88Table XXV - Continued
Dependent Variable: Industry Average Acquirer ReturnstCloseness Through Customer Links Closeness Through Supplier Links
(1) (2) (3) (4) (5) (6) (7) (8)
(0.762) (0.891) (0.546) (0.461) (0.157) (0.165)
Distant Acquirer Returnsi,t−4 0.044 0.008 −0.013 −0.286 −0.413 −0.336(0.640) (0.936) (0.890) (0.311) (0.143) (0.230)
Own Acquirer Returnsi,t−1 16.255∗∗∗ 17.824∗∗∗ 16.225∗∗∗ 14.527∗∗∗ 16.154∗∗∗ 18.125∗∗∗ 16.047∗∗∗ 14.414∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own Acquirer Returnsi,t−2 8.531∗∗∗ 9.827∗∗∗ 8.530∗∗∗ 7.106∗∗∗ 8.554∗∗∗ 10.143∗∗∗ 8.581∗∗∗ 7.489∗∗∗
(< 0.001) (< 0.001) (< 0.001) (0.003) (0.001) (< 0.001) (0.001) (0.002)
Own Acquirer Returnsi,t−3 10.970∗∗∗ 12.053∗∗∗ 10.955∗∗∗ 9.679∗∗∗ 11.438∗∗∗ 12.261∗∗∗ 11.393∗∗∗ 10.675∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)
Own Acquirer Returnsi,t−4 4.597∗∗ 5.282∗∗ 4.581∗∗ 3.993∗ 5.134∗∗ 5.466∗∗∗ 5.166∗∗ 4.799∗∗
(0.026) (0.012) (0.027) (0.060) (0.017) (0.008) (0.016) (0.027)
Own Industry Fixed Effects Yes Yes Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yesχ2 224.875 211.013 227.620 221.756 224.500 219.920 238.269 244.122
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Observations 9,420 9,420 9,420 9,420 9,420 9,420 9,420 9,420
89
Internet Appendix Table XXVITime-Series Regressions of Aggregate Merger Activity on Average Centrality: Robustness TestsThis table presents coefficients from time-series regressions from 1988 to 2010. The dependent variable is the average degree centralityof the industries that are experiencing a merger wave, where centrality is computed from the weighted and directed network of supplierlinks between industries in the input-output network. ‘Industry Merger Waves (%)t’ is the percent of all industries in year t that areexperiencing a merger wave. A merger wave is when an industry has more mergers than the 75th or 50th percentile of mergers in agiven year (indicated in column heading), relative to the industry’s time series of mergers from 1986 to 2010. Supplier networks arebased on either the 1982 or the 1997 BEA IO Report, indicated in the column heading. Merger firms are classified into industriesbased on either their primary SIC or NAICS code, or using all industry codes reported in SDC, as indicated in the column heading.Panel A uses the detailed-level of industries and Panel B uses the summary-level of industries in the BEA IO reports. p−value arereported in parentheses from standard errors corrected for heteroskedasticity and autocorrelation following Newey and West (1987)and the automatic lag selection of Newey and West (1994). Significance is indicated at 1%, 5%, and 10% levels by ∗∗∗, ∗∗, and ∗.
Dependent Variable: Average Degree Centrality of Industries in Merger Wave at time t
Firm-Level Industry Classification: Primary All
IO Relations Based on Year: 1982 1997 1982 1997
Industry M&A Activity Greater Than: 75th 50th 75th 50th 75th 50th 75th 50th
Panel A: IO Detailed Level
Industry Merger Waves (%)t 1.161∗∗∗ 0.681∗∗∗ 1.288∗∗∗ 0.608∗∗∗ 1.269∗∗∗ 0.916∗∗∗ 1.160∗∗∗ 0.569∗∗∗
(< 0.001) (0.003) (< 0.001) (0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Industry Merger Waves (%)t−1 −0.015 −0.258 −0.256 0.034 −0.904∗∗ −0.932∗∗∗ −0.757∗∗∗ −0.428∗∗∗
(0.956) (0.496) (0.262) (0.880) (0.024) (< 0.001) (< 0.001) (0.001)Industry Merger Waves (%)t−2 −0.298∗ 0.149 0.205 0.174 0.413∗ 0.518∗∗∗ 0.429∗∗∗ 0.303∗∗∗
(0.093) (0.309) (0.313) (0.235) (0.061) (< 0.001) (< 0.001) (< 0.001)Constant 0.374∗∗∗ 0.564∗∗∗ 0.353∗∗∗ 0.570∗∗∗ 0.612∗∗∗ 0.737∗∗∗ 0.699∗∗∗ 0.841∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Adjusted R2 0.793 0.446 0.709 0.474 0.569 0.498 0.513 0.266Observations 23 23 23 23 23 23 23 23
Panel B: IO Summary Level
Industry Merger Waves (%)t 0.793∗∗∗ 0.241∗∗∗ 0.699∗∗∗ 0.464∗∗∗ 0.339∗∗∗ 0.222∗∗∗ 0.468∗ 0.281∗
(< 0.001) (0.008) (< 0.001) (< 0.001) (0.001) (0.009) (0.073) (0.054)Industry Merger Waves (%)t−1 −0.353∗∗∗ −0.039 −0.303 −0.477∗∗∗ −0.279 −0.093 0.025 −0.190
(0.001) (0.726) (0.162) (< 0.001) (0.263) (0.453) (0.939) (0.543)Industry Merger Waves (%)t−2 0.071 0.021 0.187∗∗ 0.391∗∗∗ 0.245 0.005 −0.274∗∗ 0.068
(0.322) (0.860) (0.049) (< 0.001) (0.109) (0.955) (0.038) (0.736)Constant 0.569∗∗∗ 0.746∗∗∗ 0.574∗∗∗ 0.715∗∗∗ 0.723∗∗∗ 0.799∗∗∗ 0.721∗∗∗ 0.797∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001) (< 0.001)Adjusted R2 0.556 0.089 0.591 0.309 0.043 0.034 0.232 0.062Observations 23 23 23 23 23 23 23 23
90Internet Appendix Table XXVII
Vector Autoregressions and Granger Causality: Robustness TestsThis table presents coefficients from vector autoregressions of IO centrality and aggregate merger waves from 1988 to 2010. The twoendogenous variables are ‘Centralityt’: the average degree centrality of the industries that are experiencing a merger wave in year t,where centrality is computed from the weighted and directed network of supplier links between industries in the input-output network;and ‘Merger Wavest’: the percent of all industries in year t that are experiencing a merger wave. A merger wave is when an industryhas more mergers than the 75th percentile of mergers in a given year, relative to the industry’s time series of mergers from 1986 to2010. Supplier networks are based on either the 1982 or the 1997 BEA IO Report, indicated in the column heading. Merger firms areclassified into industries based on either their primary SIC or NAICS code, or using all industry codes reported in SDC, as indicatedin the column heading. Panel A uses the detailed-level of industries and Panel B uses the summary-level of industries in the BEA IOreports. p−value are reported in parentheses from heteroskedasticity and autocorrelation robust standard errors following Newey andWest (1987, 1994). Granger causality tests are reported of the null hypothesis that centrality does not Granger cause market mergerwaves and that merger waves do not Granger cause average centrality. We report the χ2 and p−value for each test. Significance isindicated at 1%, 5%, and 10% levels by ∗∗∗, ∗∗, and ∗.
Firm-Level Industry Classification: Primary All
IO Relations Based on Year: 1982 1997 1982 1997
Dependent Variable: Centralityt Wavest Centralityt Wavest Centralityt Wavest Centralityt Wavest
Panel A: IO Detailed Level
Centralityt−1 0.431 0.263∗ 0.244 0.262∗∗∗ −0.351 0.201 0.157 0.269∗∗
(0.114) (0.090) (0.391) (0.009) (0.266) (0.214) (0.505) (0.017)Centralityt−2 0.135 0.195 0.077 −0.033 0.013 0.243 −0.628∗∗ −0.045
(0.605) (0.188) (0.783) (0.767) (0.966) (0.129) (0.011) (0.714)Merger Wavest−1 0.946∗∗ 0.929∗∗∗ 1.231∗∗∗ 1.111∗∗∗ 1.214∗∗∗ 1.109∗∗∗ 1.057∗∗∗ 1.133∗∗∗
(0.023) (< 0.001) (0.009) (< 0.001) (0.009) (< 0.001) (0.004) (< 0.001)Merger Wavest−2 −0.921∗∗∗ −0.623∗∗∗ −0.693 −0.627∗∗∗ −0.537 −0.674∗∗∗ −0.067 −0.512∗∗∗
(0.003) (0.002) (0.154) (< 0.001) (0.119) (< 0.001) (0.840) (0.001)Constant 0.322∗∗∗ −0.030 0.364∗∗ 0.015 0.930∗∗∗ −0.206 1.092∗∗∗ −0.104
(0.009) (0.713) (0.031) (0.828) (0.007) (0.258) (< 0.001) (0.405)Adjusted R2 0.615 0.830 0.544 0.851 0.260 0.825 0.414 0.831Observations 23 23 23 23 23 23 23 23
Granger Causality Tests
H0: Centrality ; Wavesχ2 6.077∗∗ 6.316∗∗ 3.007 5.832∗
p−value (0.048) (0.043) (0.222) (0.054)H0: Waves ; Centralityχ2 8.840∗∗ 7.958∗∗ 6.855∗∗ 11.179∗∗∗
p−value (0.012) (0.019) (0.032) (0.004)
91
Internet Appendix Table XXVII - Continued
Firm-Level Industry Classification: Primary All
IO Relations Based on Year: 1982 1997 1982 1997
Dependent Variable: Centralityt Wavest Centralityt Wavest Centralityt Wavest Centralityt Wavest
Panel B: IO Summary Level
Centralityt−1 0.226 0.298 0.196 −0.040 0.122 −0.018 0.038 −0.062(0.366) (0.159) (0.426) (0.840) (0.603) (0.904) (0.857) (0.616)
Centralityt−2 −0.538∗∗ −0.195 −0.319 0.153 0.101 −0.154 −0.391∗ 0.098(0.031) (0.361) (0.194) (0.430) (0.668) (0.294) (0.057) (0.415)
Merger Wavest−1 0.560∗ 0.988∗∗∗ 0.565∗∗ 1.433∗∗∗ 0.168 1.343∗∗∗ 0.781∗∗∗ 1.534∗∗∗
(0.056) (< 0.001) (0.023) (< 0.001) (0.533) (< 0.001) (0.002) (< 0.001)Merger Wavest−2 −0.021 −0.257 −0.084 −0.735∗∗∗ 0.015 −0.533∗∗∗ −0.537∗∗ −0.759∗∗∗
(0.930) (0.211) (0.748) (< 0.001) (0.957) (0.001) (0.019) (< 0.001)Constant 0.786∗∗∗ 0.008 0.698∗∗∗ 0.008 0.574∗∗ 0.190 0.990∗∗∗ 0.033
(< 0.001) (0.962) (< 0.001) (0.958) (0.013) (0.184) (< 0.001) (0.778)Adjusted R2 0.303 0.717 0.429 0.820 −0.041 0.800 0.248 0.871Observations 23 23 23 23 23 23 23 23
Granger Causality Tests
H0: Centrality ; Wavesχ2 2.510 0.614 1.170 0.787p−value (0.285) (0.735) (0.557) (0.675)H0: Waves ; Centralityχ2 5.609∗ 7.921∗∗ 1.804 8.812∗∗
p−value (0.061) (0.019) (0.406) (0.012)
92Internet Appendix Table XXVIII
Aggregate Merger Activity and Centrality Excluding the Tech BubbleThis table presents coefficients from time-series regressions from 1988 to 2010, excluding years 1997 through 2002. The dependentvariable in Panel A is ‘Centralityt,’ the average degree centrality of the industries that are experiencing a merger wave, where centralityis computed from the weighted and directed network of supplier links between industries in the input-output network. ‘Industry MergerWaves (%)t’ is the percent of all industries in year t that are experiencing a merger wave. A merger wave is when an industry hasmore mergers than the 75th percentile of mergers in a given year, relative to the industry’s time series of mergers from 1986 to2010. Supplier networks are based on the 1997 BEA IO Report, either at the detailed- or summary-level of industries, as indicatedin the column heading. Merger firms are classified into industries based on either their primary SIC or NAICS code, or using allindustry codes reported in SDC, as indicated in the column heading. Panel B present vector autoregressions with two endogenousvariables: ‘Centralityt,’ and ‘Merger Wavest,’ as defined above. p−value are reported in parentheses from standard errors correctedfor heteroskedasticity and autocorrelation using the procedure of Newey and West (1987) and the automatic lag selection of Neweyand West (1994). Granger causality tests are reported of the null hypothesis that centrality does not Granger cause market mergerwaves and that merger waves do not Granger cause average centrality. We report the χ2 and p−value for each test. Significance isindicated at 1%, 5%, and 10% levels by ∗∗∗, ∗∗, and ∗.
IO Industry Level: Detail-Level Industries Summary-Level Industries
Firm-Level Industry Classification: Primary All Primary All
Panel A: Concurrent and Lagged Time-Series RegressionsDependent Variable: Average Degree Centrality of Wave Industryt
Industry Merger Waves (%)t 1.721∗∗∗ 1.716∗∗∗ 1.038∗∗∗ 0.929∗∗∗
(< 0.001) (< 0.001) (< 0.001) (< 0.001)
Industry Merger Waves (%)t−1 −0.606∗∗ −1.169∗∗∗ −0.491∗∗∗ −0.485(0.026) (< 0.001) (0.007) (0.108)
Industry Merger Waves (%)t−2 0.673∗∗∗ 1.022∗∗ 0.092 −0.910∗∗∗
(< 0.001) (0.032) (0.525) (< 0.001)
Constant 0.146 0.565∗∗∗ 0.553∗∗∗ 0.804∗∗∗
(0.209) (< 0.001) (< 0.001) (< 0.001)
Adjusted R2 0.739 0.551 0.534 0.485Observations 15 15 15 15
93
Table XXVIII - Continued
IO Industry Level: Detail-Level Industries Summary-Level Industries
Firm-Level Industry Classification: Primary All Primary All
Panel B: Predictive Vector Autoregressions and Granger CausalityDependent Variable: Centralityt Wavest Centralityt Wavest
Average Centralityt−1 0.727∗∗ 0.493∗∗∗ 0.115 −0.113(0.043) (< 0.001) (0.783) (0.687)
Average Centralityt−2 −0.307 0.007 −0.486 −0.034(0.316) (0.953) (0.181) (0.893)
Industry Merger Wavest−1 0.715 0.518∗ 0.800 1.508∗∗∗
(0.323) (0.057) (0.224) (0.001)
Industry Merger Wavest−1 −0.206 −0.711∗∗∗ 0.196 −0.446(0.756) (0.003) (0.742) (0.305)
Constant 0.291 0.042 0.766∗∗∗ 0.112(0.118) (0.552) (0.008) (0.561)
Adjusted R2 0.445 0.768 0.124 0.560Observations 15 15 15 15
Granger CausalityH0: Centrality ; Wavesχ2 1.061 2.700p−value (0.588) (0.259)
H0: Waves ; Centralityχ2 13.403∗∗∗ 0.179p−value (0.001) (0.916)
94
REFERENCES
Beseg, Julian, 1974, Spatial interaction and the statistical analysis of lattice systems, Journal of
the Royal Statistical Society. Series B (Methodological) 36, 192–236.
Bonacich, Philip, 1972, Factoring and weighting approaches to status scores and clique identifica-
tion, Journal of Mathematical Sociology 2, 113–120.
Bonacich, Phillip, and Paulette Lloyd, 2001, Eigenvector-like measures of centrality for asymmetric
relations, Social Networks 23, 191–201.
Borgatti, Stephen P., 2005, Centrality and network flow, Social Networks 27, 55–71.
Brown, Lawrence D., 1986, Fundamentals of Statistical Exponential Families (Institute of Mathe-
matical Statistics: Hayward, CA).
Dijkstra, E. W., 1959, A note on two problems in connexion with graphs, Numerishce Mathematik
1, 269–271.
Economic Classification Policy Committee, 1993, Criteria for determining industries, Discussion
Paper 4 Economic Classification Policy Committee.
Frank, Ove, and David Strauss, 1986, Markov graphs, Journal of the American Statistical Associ-
ation 81, 832–842.
Freeman, Linton C., 1979, Centrality in social networks: I. Conceptual clarification, Social Networks
1, 215–239.
Geyer, Charles J., and Elizabeth A. Thompson, 1992, Constrained Monte Carlo maximum likeli-
hood for dependent data, Journal of the Royal Statistical Society. Series B (Methodological) 54,
657–699.
Gollop, Frank M., 1994, The heterogeneity index: A quantitative tool to support industrial classi-
fication, Discussion Paper 2 Economic Classification Policy Committee.
Handcock, Mark S., David R. Hunter, Carter T. Butts, Steven M. Goodreau, and Martina Morris,
2003, statnet: Software tools for the statistical modeling of network data.
Hoberg, Gerard, and Gordon Phillips, 2010a, Dynamic text-based industry classifications and en-
dogenous product differentiation, Unpublished working paper. University of Maryland, College
Park, MD.
95
, 2010b, Product market synergies and competition in mergers and acquisitions: A text-
based analysis, Review of Financial Studies 23, 3773–3811.
Holland, Paul W., and Samuel Leinhardt, 1981, An exponential family of probability distributions
for directed graphs, Journal of the American Statistical Association 76, 33–50.
Hunter, David R., Steven M. Goodreau, and Mark S. Handcock, 2008, Goodness of fit of social
network models, Journal of the American Statistical Association 103, 248–258.
Hunter, David R., Mark S. Handcock, Carter T. Butts, Steven M. Goodreau, and Martina Morris,
2008, ergm: A package to fit, simulate and diagnose exponential-family models for networks,
Journal of Statistical Software 24.
Johannes, Michael, and Nicholas Polson, 2009, Handbook of Financial Time Series . chap. Markov
Chain Monte Carlo, pp. 1001–1013 (Springer-Verlag: Berlin).
Morris, Martina, Mark S. Handcock, and David R. Hunter, 2008, Specification of exponential-
family random graph models: Terms and computational aspects, Journal of Statistical Software
24, 1–24.
Newman, M.E.J., 2002, Scientific collaboration networks. II. Shortest paths, weighted networks,
and centrality, Physical Review E 64, 47–97.
Robins, Garry, and Martina Morris, 2007, Advances in exponential random graph (p⋆) models,
Social Networks 29, 169–172.
Robins, Garry, Pip Pattison, Yuval Kalish, and Dean Lusher, 2007, An introduction to exponential
random graph (p∗) models for social networks, Social Networks 29, 173–191.
Roll, Richard, 1988, R2, Journal of Finance 43, 541–566.
Strauss, David, and Michael Ikeda, 1990, Pseudolikelihood estimation for social networks, Journal
of the American Statistical Association 85, 204–212.
Wasserman, Stanley, and Katherine Faust, 1994, Social Network Analysis: Methods and Applica-
tions (Cambridge University Press: Cambridge).
Wasserman, Stanley, and Philippa Pattison, 1996, Logit models and logistic regressions for social
networks: I. An introduction to Markov graphs and p∗, Psychometrika 61, 401–425.
Watts, Duncan J., and Steven H. Strogatz, 1998, Collective dynamics of ‘small-world’ networks,
Nature 393, 440–442.
Top Related