Corporate Strategy, Analyst Coverage, and the Uniqueness ... · Corporate Strategy, Analyst...

Corporate Strategy, Analyst Coverage, and the Uniqueness Paradox

Lubomir P. Litov Olin Business School,

Washington University in St. Louis [email protected]

Patrick Moreton

Olin Business School, Washington University in St. Louis

[email protected]

Todd R. Zenger1 Olin Business School,

Washington University in St. Louis [email protected]

March 18, 2009

Abstract In this paper we argue that managers confront a paradox in selecting strategy. On the one hand, capital markets systematically discount uniqueness in the investment strategy choices of firms. Uniqueness in strategy heightens the cost of collecting and analyzing information to evaluate a firm’s future value. These greater costs in strategy evaluation discourage the collection and analysis of information regarding the firm, and result in a valuation discount. On the other hand, uniqueness in strategy is a necessary condition for creating economic rents and should, but for this information cost, be positively associated with firm value. We find empirical support for both propositions using a firm panel dataset between 1985 and 2007.

1 We thank Garrick Blalock, Richard Frankel, Dirk Hackbarth, Bart Hamilton, Rob Solomon, Bernard Yeung and seminar participants at University of Pennsylvania, INSEAD, London Business School, NYU, USC, Washington University in St. Louis, and Cornell University for helpful comments and suggestions in preparing this manuscript.

2

Corporate Strategy, Analyst Coverage, and the Uniqueness Paradox

Abstract

In this paper we argue that managers confront a paradox in selecting strategy. On the one hand, capital markets systematically discount uniqueness in the investment strategy choices of firms. Uniqueness in strategy heightens the cost of collecting and analyzing information to evaluate a firm’s future value. These greater costs in strategy evaluation discourage the collection and analysis of information regarding the firm, and result in a valuation discount. On the other hand, uniqueness in strategy is a necessary condition for creating economic rents and should, but for this information cost, be positively associated with firm value. We find empirical support for both propositions using a firm panel dataset between 1985 and 2007.

3

Introduction

In an efficient capital market, equity-based rewards encourage managers

to select high quality strategies—strategies that raise the value of the firm. Selecting a

strategy with high future returns involves searching for economic rents—rents that

fundamentally derive from scarcity or uniqueness in the bundles of complementary

assets, activities, and resources that the manager assembles (Bowman, 1974; Rumelt,

1984; Barney, 1986; Montgomery and Wernerfelt, 1988; Brandenburger and Stuart,

1996).2 If the manager selects a common strategy, then competition within product

markets and competition within factor markets for the resources to execute the selected

strategy, eliminate economic rents. Thus, if synergies or complementarities are common

or obvious, the market for business units—a strategic factor market (Barney, 1986)—

incorporates these synergies into the prices paid, leaving no avenue for value creation.

Only firms possessing unique asset positions or managers possessing unique foresight

about the value of asset combinations can predictably obtain complementary assets at less

than their value in their new foreseen use.3 Uniqueness is therefore a necessary condition

of high quality strategy and a necessary condition for value creation.

However, the effectiveness of equity-based rewards in prompting managers to

select value-creating strategies depends on the market’s capacity to efficiently evaluate

the expected future returns of strategies that managers select. In this regard, capital

markets confront a problem of information asymmetry (Akerlof, 1973; Hubbard, 1998);

managers possess knowledge about the quality or value of strategies that the market

2 Indeed, economic rents are defined as “reward[s] for scarcity of superior efficiency that …vanish as [they become] common….” (Becker, 1971:77). 3 Note that a firm’s strategy need not remain unique to generate value. Thus, even if other firms observe the value in a unique strategy and proceed to copy it, these imitating firms are more likely to pay prices for the required assets that are fully reflective of their value in this new, more valuable application.

4

lacks. If information about the “quality” of selected strategies is costly to obtain, the lack

of information elevates uncertainty in the market and diminishes equity values. Implicit

in most models of strategy choice is an assumption that the manager’s task is solely to

maximize the quality of strategy selected where quality is measured as the present value

of its future returns. The magnitude of the information problem that accompanies a

strategy is presumed an exogenous feature of the market. Yet, certainly this assumption

is flawed. Strategies differ significantly in the scope of the information problem that they

impose on the market. Some strategies are rather easily evaluated and examined, while

others are quite difficult and costly to evaluate and assess. Thus, in selecting a particular

strategy, the manager determines both the probable future returns and the scope of the

associated information problem.

Unique strategies are likely to impose the greatest information burden on the

market. Assessing the future value of uncommon combinations of assets and businesses

requires additional effort on the part of participants in the market. While clear incentives

exist to uncover this unique value, these incentives are tempered by the elevated effort

costs. The result is a manager who faces a paradox in strategy choice. Choosing a high

quality, unique strategy maximizes the expected value of future returns, but imposes

higher information costs on the market, resulting in a market discount. Choosing

transparent or common strategy reduces the information problem, but likely involves a

compromise on quality.

Our objective in this paper is to argue and empirically demonstrate that managers

confront a uniqueness paradox in selecting strategy. While unique strategies are more

highly valued, they are also more costly to evaluate. This heightened cost leads to

5

systematically less coverage by securities analysts and results in valuation discount of

what might otherwise be a premium for uniqueness. We empirically test propositions

relating to this uniqueness paradox using a 23-year panel dataset linking approximately

14,000 firms from Compustat’s Industrial and Segment files between 1985 and 2007 with

their corresponding analysts who appear in the I/B/E/S Detail History file. We find

results consistent with a discounted uniqueness premium: while unique strategies trade on

average at a premium to common strategies, more unique strategies receive less analyst

coverage, and this reduced analyst coverage creates a corresponding discount in market

valuation.

Our paper is organized as follows. In the next section we discuss a conceptual

model of the relationship between firm strategy and equity value. Consistent with prior

work in accounting and finance, we argue that analyst coverage positively influences

equity value (Hong, Lim, & Stein, 2000; Elgers, Lo, & Soffer, 2001) and that strategy

choice, particularly uniqueness and complexity, affects the costs of an analyst initiating

coverage. Our third section describes the data and methods used in modeling the effects

of strategy choice on coverage and coverage on valuation. It further presents empirical

results. The fourth section discusses robustness issues. The final section concludes with a

brief summary and some directions for future research.

II. Hypothesis Development

2.1. Firm Strategy, Analyst Coverage, and Equity Values

With incentives to increase the equity value of the firms that employ them (Jensen

& Murphy, 1990; Hall & Leibman, 1998), managers will evaluate the likely impact of

alternative strategy choices on equity values. We view this link between strategy choice

6

and market valuation as occurring in three stages. Managers first choose strategies which

differ in quality as measured by the future cash flows they generate and range from the

rather common or simple to the extremely unique or complex. Common strategies are

those observed frequently among firms in an industry, while unique strategies are those

rarely observed. Simple strategies are strategies constrained to a single business line,

while complex strategies involve a presence in multiple industries. Market participants,

including analysts, then observe the selected strategy and, based on an assessment of

costs and benefits, decide whether to invest in acquiring further information concerning

its value. We operationalize market participants’ choice of whether to invest in

information acquisition using analysts’ decisions regarding the initiation of coverage.

For those firms selected for coverage, analysts then collect and evaluate relevant

information and issue a report available to at least a subset of the investors in the market.

For each analyst, there is a positive probability that he or she will discover some

information that other analysts do not discover. Consequently, the more analysts who

cover a firm, the more informed is the market for that firm’s equity, i.e. the more accurate

is the information about the future prospects of the firm and the lower is the uncertainty

associated with investing in the firm. While a variety of factors may motivate analyst

coverage (see Lin & McNichols, Michaely & Womack, 1999), empirical evidence

generally confirms that increased analyst coverage increases the informational efficiency

of markets (Frankel, Kothari, and Weber, 2006; Hong, Lim, and Stein, 2000; Elgers, Lo,

and Pfeiffer, 2001; Lys & Sohn, 1990). In a final stage, investors in the market review

the available information about the firm, including reports issued by analysts, and then

submit their buy and sell orders to a market maker, who sets a market-clearing price.

7

This process in which publicly traded firms ‘manufacture’ equity investments for

sale to investors through retail brokers is not unlike the process used to sell other

complex, difficult-to-assess products, where managers seek and buyers pay a premium

for third party certification of quality or value. In this setting, investors pay a premium

for securities that receive third party evaluation by analysts. Moreover, these analysts

make decisions on the brokerage firms’ behalf regarding which equities to cover and thus

receive featured “shelf space.” Issuers of equity then have an incentive to cultivate the

attention of analysts who both increase the distribution of the firm’s equity and enhance

its value by reducing the amount of uncertainty that investors have about its performance

prospects.

A manager’s choice of strategy therefore plays two critical roles. First, it defines

the expected financial returns of the firm, including the mean and variance of the firm’s

future cash flows, which in a fully efficient market with no costs of collecting

information defines the firm’s equity price. Second, the firm’s strategy determines the

costs that an investor or market intermediary faces in accurately predicting this mean and

variance. When information is costly to obtain, information is likely to be limited in the

market, thus creating greater uncertainty about a firm’s value. When investors are risk-

averse and unable to eliminate all firm-specific risk through diversification strategies, a

feature of many models of rational trade in financial securities (e.g. Hellwig, 1980), the

market discounts the values of higher risk firms (Merton, 1987). Thus, informational

uncertainty about a firm’s strategy increases the investor’s firm-specific risk from owning

the firm’s stock and thereby diminishes the security’s price. Given this effect, the firm’s

management has an incentive to attract more coverage by analysts in order to increase the

8

amount of information possessed by investors and to increase the equity value of the firm

they manage.4

2.2. Uniqueness, Complexity, and the Information Costs of Coverage

A manager’s choice of strategy, particularly the uniqueness or complexity,

profoundly affects not only the future returns of the firm, but also the costs imposed on

analysts or the brokerage firms that employ them in providing coverage. Our assumption

is that the scope of coverage for a particular security is a function of the costs and

benefits of collecting information about that strategy. On the one hand, the adoption of a

unique or complex strategy could arguably attract analyst coverage, because investors

may place a high value on information precisely in those settings where information is

costly to obtain. However, this argument presumes that analysts and the firms that

employ them are strongly rewarded for satisfying investors’ preferences for information

gathering and analysis. At best, this is only partly true. The alternative hypothesis

which we empirically explore is that the costs of covering more unique or complex

strategy choices generally or frequently outweigh any added benefits to the analyst.

While analysts or their employers bear the full cost of information collection, the benefits

of accurate analysis are quite broadly shared. Consequently, as the costs of collecting

and analyzing information about a firm’s strategy rise, analyst coverage should decline.

The complexity and uniqueness of strategy choices shape the associated costs of

analysis. Thus, the number of distinct businesses in which the firm competes strongly 4 Empirically, Amir, Lev, and Sougiannis (1999) have shown that earnings forecasts by analysts explain between 12 per cent and 40 per cent of the above-normal returns earned by investors from publicly traded stocks, suggesting that analysts do indeed provide information that is valuable to the markets. Zuckerman (1999) has also shown that firms which receive coverage from a larger portion of the analysts focused on the industries represented in their portfolio are more highly valued, as measured by the Berger and Ofek excess value measure.

9

influences the complexity of the analyst’s task. Analysts specialize by industry,

presumably to economize on the costs of information gathering and analysis. If a firm

competes in a range of industries, this increases the costs of coverage, because it either

requires analysts to develop expertise in multiple industries or it requires coverage and

collaboration among analysts across multiple industries. Prior empirical work suggests

that an increase in the complexity of the analysts’ task reduces the accuracy of their

earnings forecasts (cites from Duru and Reeb, 2002; Brown, Richardson, and Schwager,

1987). Our argument here is that diversification increases the complexity of the analysts’

task and the costs of analysis, which in turn causes the level of coverage itself to decline.

Firms pursuing unique or less familiar strategies are also more costly for analysts to

evaluate. Like many other activities, securities analysis is an activity governed by

economies of scale. A firm adopting a common strategy, a strategy involving a

combination of assets and businesses that is common in the industry, imposes a small

incremental cost on the analyst who chooses to cover it, because the analyst has already

made substantial investments in understanding the strategy, as pursued by others. By

contrast, a firm with an unusual strategy may impose a large incremental cost on the

analyst. Assessing the value of a unique collection of businesses not only requires an

understanding of the separate industries in which each competes, but also an

understanding of any complementarities or synergies that are generated through the

combination. The more unique the combination assembled, the less likely it is that any

given analyst will be familiar with these synergies. By simplifying a complex strategy,

or adopting a more common strategy, managers lower analysts’ costs of coverage,

thereby attracting coverage and escalating information available about their security.

10

To summarize, we hypothesize:

H1: Analyst coverage is decreasing in the uniqueness of a corporate strategy.

H2: Analyst coverage is decreasing in the complexity of a firm’s strategy.

2.3. Strategy Choice, Equity Prices, and the Uniqueness Paradox

If, as noted above, uniqueness in strategy is a necessary condition for deliberate

value creation, yet uniqueness diminishes analyst coverage and the informational

efficiency of the market for the firm’s equity, then the manager faces a difficult dilemma.

Pressure by analysts to simplify or conform and thereby lower the costs of analysis may

be quite explicit. For instance, a 1999 analyst report from Paine Webber pushing for

Monsanto’s break-up as a life sciences company provides a surprisingly candid revelation

of these complexity- and uniqueness-related costs:

The life sciences experiment is not working with respect to our analysis or in reality. Proper analysis of Monsanto requires expertise in three industries: pharmaceuticals, agricultural chemicals and agricultural biotechnology. Unfortunately, on Wall Street, particularly on the sell-side, these separate industries are analyzed individually because of the complexity of each. This is also true to a very large extent on the buy-side. At PaineWebber, collaboration among analysts brings together expertise in each area. We can attest to the challenges of making this effort pay off: just coordinating a simple thing like work schedules requires lots of effort. While we are willing to pay the price that will make the process work, it is a process not likely to be adopted by Wall Street on a widespread basis. Therefore, Monsanto will probably have to change its structure to be more properly analyzed and valued.

The suggestion here is clear—that Monsanto should unbundle so as to reduce the

information costs that accompany this complex strategy and thereby promote more

extensive and precise analysis, and ultimately raise the aggregate valuation of these

assets. Empirical evidence supports this prescription, at least as it relates to reductions in

the complexity of a firm’s strategy. Gilson, Healy, Noe, and Palepu (2001) find that

11

focus-increasing transactions (spin-offs, carve-outs, and targeting stock offerings)

increase analyst coverage and increase the accuracy of analysis.5 Bhushan (1989) also

shows that analyst coverage increases as the firm reduces the number of SICs in which a

firm operates. Firms face similar pressures to unbundle unique asset combinations. For

instance, the Financial Times (Roberts, 2005) described Georgia Pacific as “trading at a

discount to the sum of its parts” because it was “an awkward mix of assets that are

difficult to evaluate together.”

Yet, clearly the manager’s primary path to value creation cannot be simply

lowering information costs. After all, economic rents are rewards for scarcity and these

rewards vanish as rent-generating assets become common (Becker, 1971). Thus, value

creation requires an insightful manager who uniquely sees value in particular asset

combinations and thus procures these at prices below their value in their new use

(Barney, 1986). If the value of particular asset combinations is widely known and thus

widely implemented by other firms in an industry, the firm will pay prices for these assets

that are fully reflective of this value. Thus, uniqueness in strategy is key to creating

economic rents and increasing value. As a consequence, the manager necessarily faces a

paradox in selecting strategy: uniqueness in strategy is essential to the discovery of

economic rents, but uniqueness elevates the information costs associated with evaluating

strategy. The simple relationship between uniqueness and firm value is therefore

uncertain. Instead, we expect to observe uniqueness having a positive direct effect on

firm value, but a negative indirect effect through analyst coverage. To summarize:

5 Similarly, Zuckerman (2000) finds evidence that that suggests firms reshape themselves to match the expertise of the analysts providing coverage.

12

H3: While uniqueness in strategy will reduce firm value through reduced analyst coverage, controlling for this effect, firm value is increasing in the uniqueness of strategy choice.

III. Empirical Design

3.1. Data

To analyze these questions, we construct a 23-year panel dataset of firms and their

analyst following between 1984 and 2007. We use the CRSP Monthly Files, the

combined CRSP-Compustat database and the I/B/E/S detailed history datasets in

constructing our dataset.6 We consider an analyst covering a firm in year t if that analyst

has issued annual earnings forecast for that firm’s fiscal period ending in year t.

INSERT TABLE 2

Our sample excludes firms in the financial industry and consists of 59,158

observations on 7,636 unique firms. The mean number of yearly observations per firm is

7.7. Fifty six percent of the firms in our sample are covered by at least one analyst (Table

2). Among covered firms, the mean number of analysts per firm is 8.9, with the

maximum number of analysts covering a firm being 64 (Compaq Computer in 1989).

Our primary interests are three-fold. First, we wish to examine whether a firm’s

strategy choice affects the level of coverage it receives. Second, we seek to confirm that

the amount of coverage that the firm receives affects the stock market’s valuation of the

firm, as well as its subsequent performance. Finally, we seek to examine how a firm’s

strategy choice affects firm value both directly and through analyst coverage.

6 I/B/E/S International Inc. is a unit of Thomson Financial and collects analysts’ estimates and research for institutional investors. It is available to researchers through the Wharton Research Data Services.

13

3.2. Measures

3.2.1. Analyst Coverage

We model the coverage level received by a firm using the following stylized

model of analyst behavior. Define Xi as a vector of firm i traits that are believed to affect

the costs and benefits to the analyst/brokerage firm from covering the firm. Define Dik as

the uniqueness of firm i’s corporate strategy relative to the strategies of the other firms

considered for coverage by analyst k. Finally, define

ikikiik uDU +++≡ 210 βββ X (1)

as the net benefit that analyst k obtains when covering firm i, in which uik is a disturbance

term that incorporates analyst k’s idiosyncratic benefit from covering firm i. With this

specification, analyst k covers firm i if Uik > 0. Assume for simplicity that,

and that the disturbance is uncorrelated both within and across indices. We can then take

the expectation of Uik across all relevant k to obtain the average benefit that an analyst

will obtain from covering firm i:

[ ] iikkii DU εβββ +++≡ E210 X (2)

If we assume that the firm-specific disturbance term εi is i.i.d. across i with some

known distribution F(εi), then the probability that firm i is covered by an analyst is:

[ ]( ) [ ]( )ikkiikkii DFD E1EPr 210210 ββββββε ++−=++> XX

If there are Aj analysts covering industry j, then we can use (2) to generate

[ ]( )( )ikkiji DFAa E1 210 βββ ++−= X , the number of analysts that find it attractive to

cover firm i. Dividing both sides by Aj, we obtain:

[ ]( )( )ikkij

i DFAa

E1 210 βββ ++−= X , (3)

as firm i’s share of the analysts covering industry j.

kiik vu += ε

14

With this simple specification and the additional assumption that F(εi) is the

standard normal distribution, it is relatively straightforward to estimate the parameters in

using the maximum likelihood method and the share of analysts

covering a firm as the dependent variable. Note, however, that there are a small number

of industries each year for which the firms in that industry receive no coverage. For

firms in these industries, we assume that [ ]( ) 0E210 ≅++ ikki DF βββ X so that our

dependent variable for coverage level becomes:

0 if 0 Adj. Coverage

/ if 0j

ii j j

A

a A A

=⎧⎪= ⎨ >⎪⎩

In robustness checks, we will utilize four other measures of analyst coverage.

First, we measure analyst coverage as a simple count of the number of analysts which

cover a particular security. The primary difficulty with this measure is the large

proportion of firms that receive no coverage whatsoever. While we use methods

appropriate for this type of dependent measure, we also present analyses using a simple

dichotomous measure coded as 1 when the firm receives any coverage whatsoever in a

given year, and 0 otherwise. Thirdly, we split adjusted coverage into two components.

Imputed adjusted coverage is the adjusted coverage (sales weighted) that a firm would

have received, if each of its component segments received analyst coverage as though it

were the median pure-play firm. This measure is appealing as we can use it to test if

lower coverage is due to a firm adopting business units in industries that are difficult to

analyze, rather than due to a firm adopting a complex strategy of interaction among these

segments (i.e., the corporate strategy). To capture the latter, we examine the excess

coverage that we define as the difference of the actual and imputed adjusted coverage.

( )210 ,, βββ=β

15

We refer to the imputed coverage as segment-specific coverage, and to the excess

coverage as strategy-specific coverage.

3.2.2. Firm strategy choice proxies

We compute measures of both strategy uniqueness and strategy complexity,

arguing that the choice to pursue either raises analysts’ coverage costs. Analysts

generally specialize by industry and thus consider for coverage only a subset of the

publicly traded firms. Firms which pursue common strategies should be more familiar

and more easily analyzed by industry specialized analysts. For this reason, we measure

the similarity of a firm’s strategy relative to other firms in its primary SIC.

For each firm we define the vector of their sales across all segments, si =

,1, , ,. .i t i n tSales Sales ′⎡ ⎤⎣ ⎦ . We then normalize this vector to unit length, by dividing all

vector elements by , ,j i j t

sales∑ , where i indexes the firm, and j indexes the set of

segment industries (n = 1,107 for 1984-2007), for a given year t.

We then define the primary industry for each firm-year as the industry with the

highest sales of the corporation. For example, GE’s primary industry is SIC 6153 in

2005, since GE Capital had the highest sales within GE in 2005. For each primary

industry, j* each year t, we define the industry vector of sales, as sj*, t =

1, ,. .

i ii n isales sales ′⎡ ⎤

⎣ ⎦∑ ∑ , where i indexes the firms in each of the n=1,107

segment industries in Compustat that are in primary industry j* in that year. We

16

normalize this vector to unit length by dividing with 1,107

,1

i i jj

sales=∑∑ (i indexes the firms, j

indexes 1,107 industries.)

Using the vectors of the firm and its primary industry distribution of sales across all

industries in the economy, we define the firm’s measure of conformity as , , *,i t i t j tconf s s′= .

This measure reflects the proximity of the sales “distribution” for each firm to the one of its

counterparts in the primary industry. Note that ,0 1i tconf≤ ≤ , and higher values of this

measure imply greater conformity to the corporate strategy of other firms in the same

primary industry. To capture the influence of analysts’ coverage on the choices of

corporate strategy we refine our conformity measure further as * *, , *,i t i t j tconf s s′= . The

vector **,j ts is defined as * ** *1, ,

. .i ii n i

sales sales ′⎡ ⎤⎣ ⎦∑ ∑ for primary industry j* in year t

where the vector elements summations include only firms that have any analyst coverage in

primary industry j* in year t. For convenience, we redefine these measures as 1-confi,t and

1-conf*i,t. We refer to these as measures of corporate strategy uniqueness (UNIQUE1 and

UNIQUE2, correspondingly), as higher values indicate lower level of conformity with the

strategy of industry peer group. In our analysis we examine both. For brevity, we show

results with the latter measure only.

The costs of analysis should also rise with the complexity that accompanies increased

diversification. As firms enter related or unrelated industries, analysis of the firm requires

either multiple analysts to collaboratively evaluate the firm or analysts to develop expertise

across multiple industries. In either case, analysis of the firm is more costly than analyzing

a single segment firm. We measure the total number of reported segments, using a series

17

of dummy variables (S1, S2, S3,…, S8) which are coded as 1 if the number of segments in

which the firm competes equals 1, 2, 3, or 8 or more, respectively, and coded as 0

otherwise.

3.2.3. Control variables

As control variables in our analysis, we include several firm-specific factors

identified either empirically or theoretically to affect the costs and benefits of coverage

from the analysts’ perspective. Following Barth, Kaznick and McNichols (2001) and

Bushan (1989) we include the variables for each firm’s 3 to 5 year compound annual

growth rate in sales, the log of its annual trading volume in its shares, the coefficient of

variation of its earnings over the last 3 to 5 years, the log of the number of common

shareholders, and the log common equity for the firm.7 We also control for the

intangible assets of the firm: R&D, advertising, recognized intangible assets, and

depreciation, the first two as a share from total expenses, and the latter two as a share

from total assets (following Barth et al). We also control for the issuance of debt or

equity in the prior year with a dummy variable that has value 1 if the company has issued

either stock or debt.8 Finally, we include variables measuring the firm’s return on average

equity (ROE), the sales-weighted average of the Herfindahl concentration index for all

the SICs in which the firm operates, and a sales-weighted average of its share of the sales

in each of the industries in which it operates, an indicator for regulated industry firms, an

indicator for S&P 600 membership, as well as an indicator for the introduction of

Regulation Fair Disclosure (Reg. F.D.) To confirm that our results are not driven by

7 Barth, Kaznick and McNichols (2001) include the log of market value (which is equivalent to including log(price)+log(shares)) in their regressions as well as several measures of the relative level of intangible assets possessed by the firm. We chose not to use market value in our regressions because the firm’s stock price is endogenous to the level of analyst coverage under the hypotheses we are examining in this paper. 8 In robustness checks we also use the log of amount of debt issued in the prior year. Results are unchanged. Note however that debt issuance is endogenous to analyst coverage, Chang, Dasgupta, Hilary (2006).

18

outliers, we winsorize all variables generated from accounting data at the 2.5 percentiles

in each tail.9 Table 1 provides concise description of the main variables. Table 2 Panel A

provides means, medians, and standard deviations for all variables.

INSERT TABLE 2

From Table 2, we see that 37.9% (100-62.1%) of all firms in our sample have

more than one segment. Moreover, in Panel C we find that fabricated products,

transportation, chemicals, printing and publishing, defense, precious metals and

shipbuilding industries are the industries in which firms have the highest levels of

strategy uniqueness. From Panel D, we also note that corporate strategy uniqueness, as

we measure it, has generally decreased from its 1980s levels.

In Panel E we show correlations of the main variables of interest. We note that

our measures of uniqueness are highly correlated with the number of segments. Hence, to

ascertain robustness of our results, we also examine alternative measures of uniqueness

that are orthogonal to the number of segments. Note further that all of our uniqueness

measures are highly persistent. To correct for the effect of this persistence on our results,

we cluster-adjust our standard errors at the firm level.

3.3. Results

3.3.1. Analyst coverage

Table 3 presents our main multivariate results on analyst coverage. Column 1

presents the estimates of a negative binomial regression of a simple count measure of the

number of analysts covering a firm. We use a conditional fixed effects negative binomial

regression method to model the effects of changes in our independent variables across

9 We also performed all analyses with un-winsorized variables. Our results remain.

19

time and within a firm on the number of analysts covering that firm. We include in this

model both year and industry effects and the full set of control variables described above.

The results support our fundamental hypothesis that costly-to-analyze securities receive

less analyst coverage. Controlling for other factors which may influence analyst

coverage, firms with more lines of business and more unique strategies receive less

analyst coverage. The array of dummy variables measuring diversification shows a

rather linear progression of increasingly negative coefficient values. The estimated

effects of our control variables are also generally consistent with our description of the

analyst coverage decision. Larger firms, as measured by their common equity and firms

that are traded more heavily receive more coverage.

The firm’s performance as measured by accounting profitability (ROE) and sales

growth also have positive effects on the level of coverage. The issuance of either equity

or debt in the prior fiscal year is positively related to the number of analysts providing

coverage. The measure of recognized intangibles in the firm is negatively related to

coverage, in line with Barth et al. (2001). Similarly, the R&D spending and the

depreciation are positively and significantly related to coverage, once again as in Barth et

al. (2001).

Column 2 of Table 3 presents a logit specification with a simple dichotomous

dependent variable, measuring whether or not the firm received analyst coverage of any

magnitude. The results are quite consistent with the negative binomial regression model.

Strategy uniqueness (UNIQUE1) is negatively related to the receipt of analyst coverage.

Similarly, the dummy variables measuring the scope of diversification show the same

very consistent pattern with prior specifications (although now the linear progression in

20

coefficient magnitudes is near monotonic.) Increases in uniqueness or complexity

significantly reduce the likelihood that the firm receives analyst coverage at any level.

Coefficients for the control variables are also similar in sign and significance with the

negative binomial models.

Table 3 also presents models of analyst coverage using the share of analyst

coverage received by the firm (adjusted coverage) as the dependent measure. Column 3

presents the results of an OLS regression of adjusted coverage on the strategy variables,

the diversification dummy variables, as well as the control variables described above.

The results are generally consistent with the results presented so far. The results suggest

that strategy uniqueness (UNIQUE1) and diversification (only at the number of segments

greater than seven) have significant negative effects on adjusted coverage. The sign and

significance of the control variables are also quite consistent across these specifications.

A concern with the OLS specifications of column 3 is that 43.4% of our

observations have a value of 0 for adjusted coverage, suggesting a left censoring

problem. To correct for this, column 4 is a model of adjusted coverage, but estimated by

the Tobit method to account for the left-tail censoring of the dependent variable at zero.10

Consistent with the OLS specification and the negative binomial and logit results, an

increase in corporate uniqueness or diversification of a firm’s strategy is associated with

a decrease in the share of analyst coverage that a firm receives. Note that while the series

of segment dummies do not show the orderly progression of increasingly negative

coefficient values as in columns 1 and 2, the trend of greater diversification diminishing

coverage remains apparent, especially if a firm reports more than three segments.

10Unfortunately a fixed effects Tobit specification would yield biased results (Honoré, 1993). We show it here for consistency, however in robustness we also examine the coefficients of a random effect Tobit method. Our results remain.

21

3.3.2. Strategy-specific and segment-specific analyst coverage

In columns 5 and 6 we attempt to split adjusted coverage into two components:

segment-specific coverage or the portion attributable to the analyst coverage that specific

segments receive and strategy-specific coverage or the portion attributable to the

combination of these segments. We seek to confirm that the effect of strategy

uniqueness in limiting analyst coverage is related to the combination of businesses and

not merely to entering segments that receive less coverage overall.11 We find support for

this hypothesis. Greater strategy uniqueness (UNIQUE1) is associated with less strategy-

specific coverage. However our results on diversification (complexity) are not sustained.

In column 6 we further relate segment-specific coverage to corporate strategy uniqueness.

Our results are similar: firms that have corporate strategies more similar to the centroid

strategy in their primary industry tend to receive more coverage. We note also that the

point estimates of corporate uniqueness are three-fold more negative in the strategy-

specific coverage model as compared to the segment-specific coverage model, indicating

that it is the corporate-strategy that is more difficult to analyze, as opposed to the

component industries for the focal firm.

3.3.3. Robustness

One concern with our measure of corporate strategy uniqueness is that it may

merely reflect the focal firm’s performance co-movement with its primary industry. To

address this, we replicate our results so far using industry fixed effects, instead of the

11 The number of observations in this regression decreases significantly, in order to be able to compute the strategy-specific coverage.

22

firm fixed effects. In columns (7)-(12) we replicate all of our model specifications with

industry fixed effects. Our results remain.

3.3.4. Changes in coverage

Our particular interest is in whether changes in a strategy’s uniqueness or

complexity from year to year are associated with changes in the level of analyst coverage.

We hence examine a differences regression for the models in Table 3. The results are

presented in Table 4.

INSERT TABLE 4

In column 1 we regress log(1+analyst coverage) on the same control variables as in Table

3. We obtain that firms whose strategies become more unique experience less analyst

coverage in the subsequent year. This result is also supported in column 2, where the

dependent variable is the adjusted coverage. When we split this into strategy-specific and

segment-specific coverage, we find that both measures decrease subsequent to increases

in the strategy uniqueness of the focal firm. Note further that the magnitude of decrease is

higher for the strategy-specific coverage, indicating that indeed difficulty to analyzing the

strategy, i.e. the combination of businesses, rather than component industries is

particularly responsible for this relation strategy uniqueness and coverage.

We conclude that there exists a causal link between the uniqueness of the

corporate strategy and the subsequent analyst coverage. We now turn to the firm

valuation implications of this fact.

3.5. Firm Value and Analyst Coverage

3.5.1. Tobin’s q

23

Our interest in examining the link between firm valuation, analyst coverage and

corporate strategy uniqueness is twofold. On the one hand, we have shown that the more

unique is a firm’s corporate strategy, the less coverage by analysts that it receives. We

hence expect that unique strategies will result in lower Tobin’s q, as analyst coverage is

positively associated with it (Doukas, Kim, and Pantzalis, 2005). On the other hand

though, strategy uniqueness is a necessary condition for the generation of economic rents,

and hence we expect the selection of a unique strategy to result in higher Tobin’s q. Thus,

the full effect of selecting a unique strategy on Tobin’s q is theoretically uncertain,

reflective of the fundamental paradox the manager faces in choosing strategy.

To examine the relationship between analyst coverage, corporate strategy

uniqueness and firm value, we regress Tobin’s q on analyst coverage measures, a set of

controls generally known to correlate with Tobin’s q, and measures of corporate

uniqueness. For each firm, our dependent variable is Tobin’s q, which we calculate as

the market value of the firm divided by its book value (Kaplan and Zingales (1997).

We also calculate and examine two industry-adjusted measures of Tobin’s q. The

first one is Tobin’s q measured as market-to-book ratio, less the sales-weighted market-

to-book ratios of its segments (where the segment market-to-book ratios are computed

annually, as the median industry ones based only on single-segment firms; we follow the

methodology of Berger and Ofek, 1995.) The second one is Tobin’s q less the imputed

Tobin’s q for its primary industry centroid strategy (a centroid strategy is the strategy of

business segments combinations at the primary industry level; we define it in detail in the

robustness section.) We compute the latter as the dot product of the primary-industry

normalized sales vector sj*, t and a vector of normalized annual median M/B ratios for

24

single-segment firms in each industry. We refer to that valuation measure as the

uniqueness premium/discount, as firms with more unique strategies may have valuations

that differ from those of their primary-industry centroid strategy.

INSERT TABLE 5

A concern with our analytic method is the joint determination of Tobin’s q,

corporate strategy, and analyst coverage, as outlined above. To address the endogeneity

problem, we run a two-stage least squares (2SLS) regression, where we instrument for

the analyst coverage with variables that capture the visibility of the company in financial

markets. Our choice of instruments is based on the intuition that market visibility would

influence analyst coverage for exogenous reasons, while influencing the choice of

corporate investment strategy. We therefore use as instrumental variables the log of

trading volume and the indicator for S&P 600 index membership.

Column 1 of Table 5 presents estimates from the second-stage equation of the

2SLS system of the market-to-book ratio. The coefficient for adjusted coverage is

positive and significant, as predicted. The result suggests that firms which receive more

coverage trade at a higher premium relative to firms receiving less coverage. In addition,

the coefficient for ROE is positive and statistically significant, while the coefficient for

log sales is negative and significant. These results are broadly consistent with other work

looking at Tobin’s q as a measure of firm value. Consistent with the diversification

discount literature (Lang & Stulz (1994)), our results also suggest that firms that are in

more segments have lower Tobin’s q. Consistent with the strategy literature’s argument

that uniqueness is a necessary condition for competitive advantage and value creation, we

find that the selection of unique strategy is positively related to Tobin’s q.

25

Column 2 presents a specification for the diversification discount—the measure of

firm value in excess of the pure play value of the constituent segments. The results are

generally consistent. The premium for the adoption of a unique strategy remains, though

it diminishes in magnitude. Consistent with the diversification literature we find that the

number of segments is associated with lower valuation, i.e. a diversification discount.

Note that the positive coefficients on the uniqueness measure in these specifications do

not preclude the possibility of a uniqueness discount that accompanies reduced analyst

coverage. Our argument is that uniqueness discourages coverage and reduced coverage

dampens Tobin’s q. Thus, while unique strategies receive a premium, that premium is

lower (i.e. discounted) relative to where it would be with greater analyst coverage.

Column 3 presents results for the uniqueness discount. The dependent variable

here is the difference in the focal firm’s valuation and the valuation of its primary

industry centroid strategy. The coefficient for adjusted coverage remains positive,

significant, and similar in magnitude. The sign, significance, and magnitude of our

control variable coefficients are also similar to the Column 1 results. Finally, the results

on strategy choice, both uniqueness and diversification, remain largely unchanged from

the Column 1 results.

Our results are robust to alternative controls for analyst coverage. When we

instead control for the log(1+analyst coverage) the coefficient on the uniqueness of

corporate strategy measure are:

For model (1): βUNIQUE1= -0.202 (t-stat= 5.18), βCOVERAGE = 1.18 (t-stat=22.4)

For model (2): βUNIQUE1= -0.13 (t-stat = 3.61), βCOVERAGE = 0.98 (t-stat=20.1)


26

Furthermore, the results obtain when we control for industry instead of firm fixed effects:

For model (1): βUNIQUE1= -0.202 (t-stat= 5.18), βCOVERAGE = 1.18 (t-stat=22.4)


For model (3): βUNIQUE1= -0.144 (t-stat = 4.03), βCOVERAGE = 0.995 (t-stat=20.6).

3.5.2. Sales growth and dynamics of value-strategy relationship

We have so far established that uniqueness in corporate strategy deters analyst

coverage, but attracts higher valuations. Our explanation for these counterintuitive results

is that there is a tension between the effects of strategy uniqueness on information

production and its direct effects on enterprise value, i.e., that strategy uniqueness in

investment choice should be positively associated with economic rents and firm growth.

While the reduced information production from strategic uniqueness appears to lower

valuations, but for this effect, strategic uniqueness should, in an efficient market, be

associated with higher valuations. We next verify this claim by examining the link of

uniqueness with subsequent sales growth.

INSERT TABLE 6

In Table 6 we study the link between the uniqueness of the corporate strategy and

the subsequent five years sales growth. Here our setup is similar to that adopted in Table

5. We use a 2SLS method, where the endogenous variable (analyst coverage) is once

again instrumented with log trading volume and S&P 600 index membership indicator. In

Model 1 we regress the one-year-ahead sales growth on strategy uniqueness, segment

dummies, adjusted analyst coverage, and controls. Our results are similar to those in

Table 5: firms that maintain unique investment strategies tend to have higher sales

27

growth. In columns 2-5 we further show the coefficient estimates for corporate strategy

uniqueness on sales growth for years two through five. There is almost near monotone

decline in the coefficient magnitude and statistical significance. This is consistent with

our interpretation of strategy uniqueness as a generator of economic rents that dissipate

over time, due to, among other factors, competition in the primary industry or rather

direct imitation of the strategy. The coefficient on analyst coverage has uniformly a

positive sign, indicating that firms that attract more analyst coverage tend to have higher

sales growth in the subsequent five years. Our results on the segment indicators reveal

that one-year sales growth is decreasing in the addition of up to three segments. However,

sales growth over 2-, 3- and 4- years is decreasing also in the addition of up to four

segments.

These results are also robust to using instead industry fixed effects in models (6)-

(10), although the dissipation over time in economic magnitude and statistical

significance of the coefficient on corporate strategy uniqueness is stronger. These results

are also robust to alternative proxies for analyst coverage. When we control for

log(1+analyst coverage) we receive the following estimates:

For model (1): βUNIQUE1= 0.122 (t-stat= 14.5), βCOVERAGE = 0.08 (t-stat=11.3)

For model (2): βUNIQUE1= 0.11 (t-stat = 8.6), βCOVERAGE = 0.11 (t-stat = 10.7)

For model (3): βUNIQUE1= 0.069 (t-stat = 4.62), βCOVERAGE = 0.11 (t-stat=9.1)



IV. Robustness

28

4.1. Alternative measures of strategy uniqueness

For robustness we consider an alternative measure of corporate strategy uniqueness.

We construct a measure of the “typical” strategy of all firms that list SIC j as their primary

SIC by creating an s-length vector for all firms listing j as their primary SIC. We refer to

this strategy as a centroid one. The elements 1, 2, 3…, s of this vector for each firm are

equal to the mean sales across all firms in industry j in each of the s SICs that appear

among the segment data for each of the firms in primary industry j. The uniqueness of

each firm’s strategy is then measured as the Euclidian distance, d, between the industry

centroid and an s-element vector constructed from the firm sales in each of the s SICs that

are observed in the industry. This Euclidian distance is then standardized in the following

way:

1

1 1

0 if 0 UNIQUE3

/ if 0

j

j l

iNi

ii i iN N

i i

d

d d d

⎧ =⎪= ⎨

≠⎪⎩

∑

∑ ∑

with Nj equal to the number of firms in industry j. Thus, the variable UNIQUE3i

is a measure of the uniqueness of a firm i’s strategy relative to the average distance of its

peers in primary industry j.

Note that the industry centroid strategy used in the strategy measure is not meant

to portray the most “familiar” strategy in the industry. Rather, it is simply a reference

point against which to compare the strategies of all the firms in the industry. A simple

example will best illustrate this point. Suppose that there are two firms, A and B, in

industry j, each of which has the same sales volume, sj in industry j. Suppose that both

firm A and B are also active in two other SICs, 1 and 2, and that they have sA1, sA2, sB1,

and sB2 in sales in these two industry. To further simplify the example assume that sA1 =

29

sA2 = sB1= sB2 = s so that the industry’s centroid strategy is (sj, s, s). Since this vector is

the same as the vectors characterizing each of the firm’s strategy, the Euclidian distance

between each firm and the centroid is equal to 0, and the mean distance is also equal to

zero, implying that UNIQUE3A = UNIQUE3B = 0. That is, the strategy’s of firms A and

B are commonplace in the extreme.

Now suppose that firm A has sA1 in sales in SIC 1 but none in SIC 2, and that firm

B has no sales in sic 1 but sB2 in sales in industry 2. Again, to simplify the example

assume that sA1 = sB2 = s. The centroid of the industry would then be the vector

(sj,s/2,s/2) and the Euclidian distance of each firm from this centroid would be

( )22/2 sd = . Since both firms are equidistant from the centroid, the mean distance

among all firms would also be d, and our measure of the uniqueness of each firm would

be DIST= d/d = 1. The implication is that both firms are somewhat unique, relative to the

firms in the first example. However, relative to each other, they are equally familiar from

the perspective of an observer of industry j. Therefore, the variable UNIQUE3 provides a

useful summary of the uniqueness of a firm’s strategy both within and across industries.

In Table 2 we present descriptive statistics for UNIQUE3. We note that it has

6.8% correlation with UNIQUE1. Similar to the latter, UNIQUE3 is also highly

persistent, and it has declined from its high levels in the 1980s.

When we use UNIQUE3 as the alternative measure of strategy uniqueness in

models (1)-(3) of Table 3 we obtain the following coefficient estimates:

Model (1): βUNIQUE3= -0.008 (t-stat= -1.70)

Model (2): βUNIQUE3= -0.054 (t-stat = -1.78)

Model (3): βUNIQUE3= -0.012 (t-stat = -3.68)

30

These are consistent with our main findings.

4.3. Multicollinearity

A further concern with UNIQUE1 is its high correlation with the number of

segments (64.9%). There is hence an issue of multicollinearity. To address this we re-

estimate our models in Table 3 without the segment indicators. Our results remain. For

example, the coefficient on the UNIQUE1 is -0.382, highly significant (t-stat is 15.1).

Second, we consider an adjusted measure of uniqueness, UNIQUE1*, which is the

residual of the regression of UNIQUE1 on the number of segments. Using that measure

instead, our results still obtain (for example the coefficient on UNIQUE* in model 1 of

Table 3 is -0.0247, with t-statistic of -7.99). Third, we note that our results using

UNIQUE1 are similar to those using UNIQUE3 and that the correlation of the former

with the number of segments is moderate, at -20.2%.

4.2. Alternative Sample Selection

In our selection procedure we follow the approach by Barth et al (2001). However

our results are robust to including financial companies.

Next, SFAS 131 changed the method of disclosure of segment information in

December 1997. Such change could materially affect our measure of corporate strategy

uniqueness. To verify that our results hold, we split our sample into pre-1997 and post-

1997. The coefficients estimates for model (1) in Table 3 across the two are as follows:

Model (1), pre-Dec 1997 sample: βUNIQUE3= -0.205 (t-stat= -4.71)

Model (1), post-Dec 1997 sample: βUNIQUE3= -0.176 (t-stat= -5.53)

Lastly our results are robust to controlling for the introduction of new regulation

on sell-side analysts’ research, in September 2002 (NASD Rule 2711, NYSE Rule 472,

31

and the Global Analyst Research Settlement) by including an indicator variable for it

(zero pre-Sept 2002 and one otherwise).

V. Discussion and Conclusions

Our empirical results are generally consistent with the relationships we

hypothesized. First, we find that firms which choose more unique strategies receive less

analyst coverage than those which choose more familiar strategies. We can reasonably

conclude that as firms move further away from the strategies of their peers in their

primary industry, they receive less coverage. Alternatively, as they move closer to the

strategy of their peers, analyst coverage increases. Second, this relationship appears to be

causal, as changes in the level of strategy uniqueness drive the changes in levels of

analyst coverage.

Third, firms that receive less coverage also have lower valuations than those that

receive more coverage. Moreover, the effect of a reduction in coverage after controlling

for the likelihood that both coverage and value are correlated with unobserved firm

effects is non-trivial in magnitude. Holding all other variables constant, an increase in

adjusted coverage by one standard deviation increases the firm’s Tobin’s q by

approximately 23%-48% above its mean, depending on the specification used.

Our results thus suggest that managers face a clear paradox in choosing their

investment strategy. While uniqueness in investment strategy choices is a necessary

condition for value creation, as measured by expected future operating performance,

uniqueness in strategy also mandates more costly expenditures by participants in the

market to evaluate such strategy. If analysts (or the investment banks which assign

analysts) are rewarded in ways that overcome these greater costs associated with

32

evaluating unique strategy, then capital markets will correctly (efficiently) evaluate

unique strategies. However, casual observations and our empirical results suggest that

this is not the case. The strong link between investment banking business and analyst

rewards suggests that at best analysts face a multi-tasking problem when allocating effort

to accuracy in analysis of strategies. While accurate and thorough analysis yields some

positive returns to analysts and investment banks, the capacity to draw investment

banking business or trade volume through coverage choices and analysis has been and

continues to be a stronger financial motivation. Because analysts are not directly

rewarded for effort, effort allocated to costly-to-analyze firms is likely to be less than

effort allocated to easily analyzed firms, all else equal. Consequently, capital markets

may systematically discount uniqueness, the most important element required for value

creation.

The result of the above is that managers make strategy choices that are, at the

margin, more common than they would be if the manager were simply choosing

strategies that maximize the discounted present value of expected long term operating

performance. In light of these results, the structure of CEO compensation may have an

important bearing on strategy choice. Rewards that are more weakly linked to present

market value may encourage more unique strategy choices, while rewards linked to

current market valuations encourage strategy choices that are more responsive to effort-

averse analysts and are thus more common and less complex.

33

References

Amihud, Y. and B. Lev, 1981, Risk Reduction as a Managerial Motive for Conglomerate Mergers, Bell

Journal of Economics, 12: 605–17.

Amir, E., B. Lev, and T. Sougiannis, 1999, What Value Analysts? Working Paper, Tel Aviv University.

Barney, J., 1986, Strategic Factor Markets: Expectations, Luck, and Business Strategy, Management

Science, 32: 1231-1241.

Becker, G., 1972, Economic Theory. Alfred A. Knopf: New York.

Berger, P. and Ofek, E., 1995, Diversification’s Effect on Firm Value, Journal of Financial Economics,

37: 39–65.

Berle, A. and G. Means, 1968, The Modern Corporation and Private Property. New York: Harcourt Brace

& World.

Bhushan, R., 1989, Firm Characteristics and Analyst Following, Journal of Accounting and Economics,

11/2/3: 255–74.

Bowman, E., 1974, Epistemology, corporate strategy, and academe. Sloan Management Review. 15: 35-

50.

Brandenburger, A. and J. Stuart, 1996, Value-Based Business Strategy, Journal of Economics and

Management Strategy, 5: 5–24.

Brown, L., G. Richardson, and S. Schwager, 1987, An Information Interpretation of Financial Analyst

Superiority in Forecasting Earnings, Journal of Accounting Research, 25: 49-67.

Donaldson, G., 1984, Managing Corporate Wealth. New York: Praeger.

Duru, A. and D. Reeb, 2002, International Diversification and Analyst Forecast Accuracy and Bias,

Accounting Review, 77: 415-433.

Elgers, P., M. Lo, and R. Pfeiffer, R., 2001, Delayed Security Price Adjustments to Financial Analysts’

Forecasts of Annual Earnings, Working paper, University of Massachusetts - Amherst.

Fama, E., 1970, Efficient Capital Markets: A Review of Theory and Empirical Work, Journal of Finance,

383-417.

Fishman, M. and K. Hagerty, 1989, Disclosure Decisions by Firms and the Competition for Price

Efficiency, Journal of Finance, 44: 633–46.

Frankel, R., S. Kothari, and J. Weber, 2006, Determinants of the Informativeness of Analyst Research,

Journal of Accounting and Economics, Vol. 41: 29-54.

Grossman, S., and J. Stiglitz, 1980, On the Impossibility of Informationally Efficient Markets, American

Economic Review, 70: 393-408.

34

Hall, B. and J. Liebman, 1998, Are CEOs Really Paid Like Bureaucrats?, Quarterly Journal of

Economics, 113: 653–91.

Hellwig, M., 1980, On the Aggregation of Information in Competitive Markets, Journal of Economic

Theory, 22: 477–98.

Holmstrom, B. and S. Kaplan, 2001, Corporate Governance and Merger Activity in the U.S.: Making

Sense of the 1980s and 1990s’. NBER working paper.

Hong, H., T. Lim, and J. Stein, 2000, Bad News Travels Slowly: Size, Analyst Coverage, and the

Profitability of Momentum Strategies, Journal of Finance, 55: 265-295.

Honoré, B., 1993, Orthogonality Conditions for Tobit Models with Fixed Effects and Lagged Dependent

Variables, Journal of Econometrics, Vol. 59: 35-61.

Jensen, M., 1986, Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers, American

Economic Review, 76: 323–9.

Jensen, M., and W. Meckling, 1976, Theory of the Firm: Managerial Behavior, Agency Costs and

Ownership Structure, Journal of Financial Economics, 3: 305–60.

Jensen, M., and K. Murphy, 1990, Performance Pay and Top-Management Incentives, Journal of Political

Economy, 98: 225–64.

Lang, L. and R. Stulz, 1994, Tobin’s Q, Corporate Diversification, and Firm Performance, Journal of

Political Economy, 102: 1248–80.

Lin, H., M. McNichols, 1998, Underwriting Relationships, Analysts’ Earnings Forecasts and Investment

Recommendations, Journal of Accounting and Economics, 25: 101-127.

Lys, T., and S. Sohn, 1990, The Association Between Revisions of Financial Analysts’ Earnings

Forecasts and Security Price Changes, Journal of Accounting and Economics, 13: 341-363.

MacDonald, G. and M. Ryall, 2002, How Do Value Creation and Competition Determine Whether a Firm

Appropriates Value? Working paper, Washington University in St. Louis.

Merton, R., 1987, A Simple Model of Capital Market Equilibrium with Incomplete Information, The

Journal of Finance, 42: 483–510.

Michaely, R., and K. Womack, 1999, Conflict of Interest and the Credibility of Underwriter Analyst

Recommendations, Review of Financial Studies, 12: 573-608.

Montgomery, C. and B. Wernerfelt, 1988, Diversification, Ricardian Rents, and Tobin’s Q. Rand Journal

of Economics, 19: 623-632.

Murphy, K., 1985, Corporate Performance and Managerial Remuneration: An Empirical Analysis,

Journal of Accounting and Economics,7: 11–42.

35

Palepu, K., 1985, Diversification Strategy, Profit Performance and the Entropy Measure, Strategic

Management Journal, 6: 239–55.

Penrose, E., 1959, The Theory of the Growth of the Firm. New York: Oxford University Press.

Rhoades, S., 1973, The Effect of Diversification on Industry Profit Performance in 241 Manufacturing

Industries: 1963, Review of Economics and Statistic, 55: 146-155.

Rhoades, S., 1974, A Further Evaluation of the Effect of Diversification on Industry Profit Performance,

Review of Economics and Statistic, 56: 557–559.

Roberts, D., Georgia Pacific Decides to Leave the Spotlight, The Financial Times, 15 November 2005, p.

29.

Rumelt, R., 1974, Strategy, Structure, and Economic Performance. Cambridge, MA, Harvard University

Press.

Rumelt, R., 1982, Diversification Strategy and Profitability, Strategic Management Journal, 3: 359–369.

Rumelt, R., 1984, Toward a Strategic Theory of the Firm, in R. Lamb (ed.), Competitive Strategic

Management. Englewood Cliffs, NJ: Prentice-Hall.

Schoar, A., 2002, Effects of Corporate Diversification on Productivity, Journal of Finance, 57: 2379–

2403.

Servaes, H., 1996, The Value of Diversification During the Conglomerate Merger Wave, The Journal of

Finance, 51: 1201–25.

Shleifer, A. and R. Vishny, 1989, Management Entrenchment: The Case of Manager-Specific

Investments, Journal of Financial Economics, 25: 123–39.

Teece, D., 1984, Economic Analysis and Strategic Management, California Management Review, 26: 87–

110.

Teece, D., R. Rumelt, S. Winter, and G. Dosi, 1994, Understanding Corporate Coherence: Theory and

Evidence, Journal of Economic Behavior & Organization, 23: 1–30.

Villalonga, B., 2003, The Diversification Discount, Cambridge, MA: Harvard Business School.

Wernerfelt, B., 1984, A Resource-Based View of the Firm, Strategic Management Journal, 5: 171-181.

Zuckerman, E., 1999, The Categorical Imperative: Securities Analysts and the Illegitimacy Discount,

American Journal of Sociology, 104: 1398-1438.

Zuckerman, E., 2000, Focusing the Corporate Product: Securities Analysts and De-Diversification,

Administrative Science Quarterly, 45: 590–621.

36

Table 1. Variables definitions

Key variables UNIQUE1

Measure of the uniqueness of the corporate strategy based on the dot product of the normalized vector of sales for the focal firm across 1,107 available four-digit SIC code defined industries, with the similar normalized vector for the sales of all firms in the primary industry, that are covered in the corresponding annual period by at least one analyst

UNIQUE2

Measure of the uniqueness of the corporate strategy based on the dot product of the normalized vector of sales for the focal firm across 1,107 available four-digit SIC code defined industries, with the similar normalized vector for the sales of all firms in the primary industry

UNIQUE3

Measure of deviation from the typical (centroid) strategy of all firms that list SIC j as their primary SIC code. We first create the centroid strategy as a vector with s elements for each firm, where the elements 1, 2, 3…, s of this vector are equal to the mean sales across all firms in industry j in each of the s SICs that appear among the segment data for each of the firms in primary industry j, in a given year. The uniqueness of each firm’s strategy is then measured as the Euclidian distance between the industry centroid and an s-element vector constructed from the firm sales in each of the s SICs that are observed in the industry.

Adjusted Coverage Analyst coverage scaled by the primary industry coverage for that year.

Analyst Coverage Number of analysts issuing EPS annual forecasts for this fiscal year. We consider the calendar year of fiscal year period end date as the coverage date.

Imputed (segment-specific) Coverage Adjusted coverage of the firm computed based on the median pure play coverage of all its segments. Sales-weighted. Excess (strategy-specific) Coverage

Actual adjusted coverage minus imputed analyst coverage. We require that the included coverage for pure-play median firms is available in order to be able to compute the excess coverage.

Uniqueness Premium

Actual M/B minus the dot product of sj*, t and a vector of annual median M/B ratios for public single-segment firms in each industry. We require at least three single-segment firms within an SIC code in order to compute the median M/B ratio. If less than three single-segment firms are available, we then use the median single segment firm at the 3-digit SIC code level. If we still have less than 3 available firms, we then retrieve the median M/B ratio for single-segment firms at the 2-digit SIC level.

Diversification Discount

Actual M/B minus the dot product of sj*, t and a vector of annual median M/B ratios for public single-segment firms in each industry.

Other control variables Total assets Compustat data # 6, in million US$, logarithm Trade volume Logarithm of the number of shares traded (in million) Firm growth 3- or 5-year compound annual sales growth Issuance Dummy Dummy variable with value of 1 if the firm has issues debt or equity in the preceding fiscal year, in US$ million Earnings variability Coefficient of variation of the earnings over the preceding five years. # of firms in industry Logarithm of the number of firms in the focal firm’s primary SIC in the current fiscal year ROE Return on average equity Number of shares The number of shares, in millions Number of shareholders

Logarithm of the number of common shareholders (#100)

Market equity Logarithm of the common equity of the firm (#25 * #199) Average Herfindahl Index

Sales-weighted Herfindahl index for all SIC codes in which the firm operates. Recorded in percent.

Average market share Sales-weighted average of firm’s share of sales in each of the industries in which it operates. Recorded in percent.

RD_F R&D spending divided by total operating expenses minus the same ratio computed for the firm’s primary industry.

See Barth et al. (2001) for details.

ADV_F Advertising expense divided by total operating expense minus the same ratio computed for the firm’s primary

industry. See Barth et al. (2001) for details.

INTANG_F Ratio of recognized intangibles assets to total assets minus the median such ratio in the firm’s primary industry. See

Barth et al. (2001) for details.

DPT_F Depreciation expense divided by total operating expense minus the median such ratio for the firm’s primary

industry. See Barth et al. (2001) for details.

37

Table 2. Panel A. Key dependent and explanatory variables descriptive statistics. The mean statistics for the segment dummy variables is in percent. Statistics Mean Median Min Max Std. Dev. N Coverage Dummy 0.56 1 0 1 0.50 59,158 Number of Analysts 5.02 1 0 23 7.01 59,158 Coverage Share 0.16 0.02 0.00 1.00 0.27 59,158 Strategy-specific Coverage 0.08 0.01 -0.26 0.83 0.21 26,137 Segment-specific Coverage 0.16 0.10 0.01 0.73 0.17 26,137 UNIQUE1 0.24 0.18 0.78 0 0.78 59,158 UNIQUE2 0.26 0.21 0.8 0 0.78 59,158 UNIQUE3 1.02 0.67 0.00 4.82 1.05 59,155 Log (sales, US$ mln) 5.36 5.40 -6.91 12.56 2.28 59,012 Log (assets, US$ mln) 1.59 1.68 -5.30 2.61 0.50 59,042 Diversification Discount 0.15 -0.03 -1.51 4.98 1.06 58,486 Uniqueness Premium 0.16 -0.04 -1.47 4.87 1.06 58,486 Segment 1 Dummy 62.1% 1 0 1 0.49 59,158 Segment 2 Dummy 19.1% 0 0 1 0.39 59,158 Segment 3 Dummy 11.2% 0 0 1 0.32 59,158 Segment 4 Dummy 4.6% 0 0 1 0.21 59,158 Segment 5 Dummy 1.9% 0 0 1 0.14 59,158 Segment 6 Dummy 0.7% 0 0 1 0.08 59,158 Segment 7 Dummy 0.2% 0 0 1 0.05 59,158 Segment 8+ Dummy 0.2% 0 0 1 0.04 59,158 Table 2. Panel B. Key controls descriptive statistics Statistics Mean Median Min Max Std. Dev. N Log trading volume 16.27 16.33 11.36 19.70 2.01 59,118 Sales Growth (past 3 years) 0.29 0.24 -1.05 1.87 0.53 59,013 Issuances in prior year indicator 0.89 1.00 0.00 1.00 0.32 59,158 Equity Issuances 0.66 0.74 -5.12 5.57 2.57 44,246 Debt Issuances 3.17 3.33 -3.08 7.66 2.56 32,646 Earnings Vol. 0.23 0.21 -3.78 3.90 1.17 59,158 # of firms in industry 49 23 1 437 66 59,158 ROE -0.01 0.09 -2.09 1.06 0.46 59,105 Log number of shareholders 0.83 0.67 -3.22 4.31 1.64 58,297 Log common equity 4.62 4.62 -0.02 8.58 2.05 58,063 Average HHI 0.20 0.16 0.02 0.78 0.16 59,006 Average market share 0.08 0.01 0.00 0.66 0.15 59,006 RD_F (Barth et al., 2001) 0.01 0.00 -0.08 0.28 0.06 59,114 ADV_F (Barth et al., 2001) 0.00 0.00 -0.05 0.05 0.02 59,114 INTANG_F (Barth et al., 2001) 0.04 0.00 -0.11 0.36 0.10 59,157 DPT_F (Barth et al., 2001) 0.01 0.00 -0.08 0.15 0.04 59,063 Number of segments 1.69 1.00 1.00 11.00 1.12 59,158

38

Table 2. Panel C. Fama-French primary industry distribution of key explanatory and dependent variables. Tabulations of main variables across the sample averages for Fama-French industry definitions. The industry tabulations for strategy- and industry- specific coverage are based on the intersection of the samples of available observations for the latter two variables, and are shown in bold. We show tabulates sorted on UNIQUE1.

FF Industry

Code Industry Name Coverage Dummy

Number of

AnalystsCoverage

Share

Strategy-Specific

Coverage

Segment-Specific

CoverageUNIQUE1

UNIQUE3

Number of

segmentsUniqueness

Premium Div.

Discount3 Candy & Soda 0.65 7.16 0.41 0.03 0.47 0.10 1.13 1.43 0.58 0.57

28 Non-Metallic and Industrial Metal Mining 0.58 9.07 0.19 0.11 0.22 0.11 1.41 1.25 0.21 0.25

43 Restaurants, Hotels, Motels 0.61 6.50 0.20 0.10 0.23 0.11 1.08 1.37 0.18 0.19 11 Healthcare 0.51 4.36 0.19 0.08 0.27 0.12 1.16 1.44 0.16 0.16 41 Wholesale 0.59 5.53 0.23 0.09 0.23 0.14 1.09 1.67 0.10 0.10 13 Pharmaceutical Products 0.59 5.68 0.05 0.04 0.03 0.15 1.04 1.27 0.41 0.35 9 Consumer Goods 0.62 4.50 0.32 0.14 0.29 0.16 1.08 1.54 0.21 0.21

35 Computers 0.55 4.54 0.11 0.07 0.10 0.16 1.03 1.48 0.16 0.16 36 Electronic Equipment 0.52 4.68 0.07 0.04 0.08 0.17 0.99 1.32 0.15 0.14 48 Other 0.09 0.47 0.02 -0.02 0.13 0.17 1.14 1.68 -0.07 -0.12 10 Apparel 0.67 5.26 0.36 0.08 0.45 0.17 1.01 1.57 0.22 0.22 5 Tobacco Products 0.86 8.82 0.65 0.18 0.37 0.18 1.03 2.18 1.90 1.94

16 Textiles 0.66 4.25 0.37 0.11 0.39 0.19 1.06 1.78 0.13 0.14 34 Business Services 0.70 4.96 0.49 0.04 0.56 0.20 1.04 1.73 0.09 0.07 12 Medical Equipment 0.56 4.22 0.06 0.05 0.06 0.21 0.98 1.28 0.24 0.25 7 Entertainment 0.51 4.95 0.21 0.16 0.24 0.22 1.21 1.54 0.05 0.06

30 Petroleum and Natural Gas 0.69 5.26 0.45 0.02 0.43 0.23 0.84 2.28 0.30 0.31 2 Food Products 0.65 6.40 0.38 0.11 0.40 0.24 1.03 1.80 0.28 0.28

18 Construction 0.50 3.55 0.20 0.10 0.25 0.24 1.02 1.98 0.08 0.08 42 Retail 0.55 3.92 0.23 0.12 0.23 0.25 1.12 1.90 0.12 0.12 1 Agriculture 0.67 4.84 0.59 0.01 0.68 0.25 0.81 2.11 0.94 0.95

32 Telecommunication 0.49 4.68 0.09 0.05 0.13 0.26 0.99 2.14 0.02 0.02 37 Measuring and Control

i0.53 5.06 0.07 0.06 0.07 0.27 0.87 1.45 0.16 0.13

38 Business Supplies 0.51 3.33 0.08 0.06 0.08 0.27 0.97 1.48 0.07 0.08 6 Recreation 0.46 2.84 0.17 0.14 0.19 0.27 1.1 1.68 0.01 -0.03

33 Personal Services 0.51 5.81 0.09 0.05 0.12 0.27 1.25 1.95 0.03 0.05 31 Utilities 0.46 5.66 0.07 0.05 0.08 0.27 0.88 1.75 0.15 0.14 23 Automobiles and Trucks 0.55 4.36 0.16 0.09 0.20 0.28 1.28 1.48 0.22 0.25 4 Beer & Liquor 0.45 5.40 0.29 0.30 0.34 0.30 1.13 2.07 0.08 0.11

19 Steel Works Etc 0.69 6.41 0.29 0.13 0.19 0.31 1.02 2.03 0.18 0.17 24 Aircraft 0.69 5.75 0.17 0.12 0.11 0.34 1.06 2.09 0.17 0.17 21 Machinery 0.59 4.59 0.19 0.11 0.16 0.34 1.05 1.89 0.14 0.10 22 Electrical Equipment 0.52 3.12 0.20 0.08 0.21 0.35 1.02 1.93 0.14 0.12 39 Shipping Containers 0.68 6.73 0.28 0.09 0.25 0.37 1.04 2.08 0.15 0.16 15 Rubber and Plastic

d0.47 2.25 0.17 0.13 0.15 0.37 1 1.87 0.26 0.21

17 Construction Materials 0.61 4.30 0.34 0.12 0.37 0.38 1.02 2.18 0.23 0.22 29 Coal 0.81 7.11 0.62 0.37 0.37 0.39 0.83 2.49 0.40 0.37 20 Fabricated Products 0.49 2.34 0.25 0.07 0.43 0.4 1.06 1.96 0.07 0.05 40 Transportation 0.63 5.05 0.44 0.31 0.36 0.4 0.97 2.33 0.11 0.13 14 Chemicals 0.67 6.86 0.23 0.11 0.17 0.42 1 2.14 0.08 0.04 8 Printing and Publishing 0.67 7.42 0.25 0.17 0.18 0.43 1.04 2.35 0.11 0.11

26 Defense 0.72 5.36 0.33 0.17 0.18 0.46 1.17 2.41 0.09 0.08 27 Precious Metals 0.67 4.95 0.54 0.36 0.41 0.52 0.97 3.01 0.65 0.64

25 Shipbuilding Equipment 0.59 7.17 0.21 0.15 0.15 0.55 1.18 2.71 0.05 0.04

39

Table 2. Panel D. Key explanatory and dependent variables over time The industry tabulations for strategy- and industry- specific coverage are based on the intersection of the samples of available observations for the latter two variables, and are shown in bold.

Year Coverage Dummy

Number of Analysts

Coverage Share

Strategy-Specific

Coverage

Segment-Specific

Coverage UNIQUE1 UNIQUE3 Number of segments

Uniqueness Premium

Div. Discount

1985 0.32 3.64 0.15 0.11 0.26 0.31 1.12 1.9 -0.02 -0.01 1986 0.37 4.01 0.16 0.09 0.24 0.27 1.1 1.9 0.02 0.03 1987 0.39 4.21 0.16 0.10 0.22 0.26 1.09 1.8 0.08 0.10 1988 0.39 3.99 0.15 0.11 0.21 0.26 1.06 1.7 0.06 0.07 1989 0.47 4.62 0.16 0.11 0.18 0.25 1.05 1.7 0.08 0.09 1990 0.48 4.62 0.16 0.10 0.19 0.24 1.03 1.7 0.09 0.11 1991 0.49 4.27 0.17 0.10 0.19 0.23 1.03 1.6 0.11 0.12 1992 0.51 4.23 0.16 0.11 0.18 0.23 1.04 1.6 0.12 0.13 1993 0.52 4.52 0.16 0.10 0.17 0.23 1.06 1.6 0.08 0.08 1994 0.52 4.42 0.15 0.08 0.16 0.22 1.05 1.6 0.09 0.09 1995 0.53 4.44 0.15 0.09 0.15 0.21 1.07 1.5 0.07 0.08 1996 0.55 4.51 0.14 0.09 0.14 0.21 1.04 1.5 0.10 0.11 1997 0.57 4.70 0.14 0.08 0.13 0.2 1.05 1.5 0.18 0.20 1998 0.58 4.79 0.14 0.07 0.13 0.27 1 1.8 0.27 0.27 1999 0.59 4.82 0.16 0.07 0.14 0.27 1.02 1.8 0.18 0.20 2000 0.60 4.87 0.16 0.07 0.14 0.26 1.01 1.8 0.27 0.27 2001 0.60 5.08 0.17 0.07 0.15 0.25 1 1.7 0.25 0.26 2002 0.63 5.44 0.17 0.06 0.15 0.25 0.98 1.8 0.21 0.22 2003 0.66 5.75 0.18 0.06 0.15 0.25 0.97 1.7 0.19 0.20 2004 0.71 6.22 0.18 0.06 0.14 0.25 0.97 1.7 0.19 0.21 2005 0.75 6.69 0.19 0.06 0.14 0.22 0.95 1.7 0.20 0.22 2006 0.77 7.31 0.20 0.07 0.14 0.23 0.95 1.6 0.18 0.19 2007 0.83 8.36 0.23 0.06 0.17 0.22 0.94 1.6 0.28 0.29

40

Table 2. Panel E. Pairwise correlations. The p-values for each of the pair wise correlations are shown below the estimate, in parentheses. The estimates in bold are computed based on the sample of available data of strategy-specific analyst coverage and segment-specific analyst coverage.

Coverage Dummy

Coverage Count Adj. Coverage

Strategy-specific

Coverage

Segment-specific

Coverage UNIQUE3 UNIQUE1

Coverage Count 68.2% 100.0% (0.00)

Adj. Coverage 52.2% 52.0% 100.0% (0.00) (0.00)

Strategy-specific Coverage12 - 39.7% 76.6% 100.0%

- (0.00) (0.00) Segment-specific

Coverage 3.1% 2.5% 41.5% -3.7% 100.0% (0.00) (0.00) (0.00) (0.00)

UNIQUE3 -9.4% -28.5% -16.9% -28.3% -4.2% 100.0% (0.00) (0.00) (0.00) (0.00) (0.00)

UNIQUE1 0.4% 4.3% 1.3% -11.2% -13.7% 6.8% 100.0% (0.14) (0.00) (0.00) (0.00) (0.00) (0.00)

Number of segments 5.7% 12.1% 14.9% 18.8% 6.6% -20.2% 64.9% (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)

Persistence of the measures of uniqueness: autocorrelations.

Variable Rho1 Rho2 Rho3

UNIQUE1 0.85 0.77 0.70 UNIQUE2 0.82 0.75 0.68 UNIQUE3 0.90 0.84 0.78

12 We cannot compute correlation of the coverage dummy with strategy-specific coverage as the former does not vary when the latter is available.

41

Table 3. Analyst coverage regressions. This table provides estimates of the effects of corporate strategy choice on the level of analyst coverage. Models (1) and (7) are negative binomial regressions of a simple count measure of analysts covering the firm analyst on strategy choice and various industry and firm controls. Models (2) & (8) examine coverage as a dichotomous variable using a logit specification. The dependent measure is coded as 1 if the firm receives any coverage in a given year and 0 otherwise. Models (3) and (9) examine adjusted coverage measured as the percentage of all analysts in an industry which cover the focal firm, by the OLS method. Models (4) and (10) do the same allowing for truncation of the dependent variable in a tobit model, correcting for the left truncation in the dependent measure. Models (5) and (11) examine the strategy-specific analyst coverage. Models (6) and (12) examine the segment-specific analyst coverage. All variables are defined in Table 1. The t-statistics (in parentheses) are based on robust standard errors cluster-adjusted at the firm level. The ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Regressions include year fixed effects, industry firm effects (models (7)-(12)) and firm fixed effects (models (1)-(12)).

Coverage Coverage Dummy

Adj. Coverage

Adj. Coverage

Strategy-specific

Coverage

Segment-specific

Coverage Coverage Coverage Dummy

Adj. Coverage

Adj. Coverage

Strategy-specific

Coverage

Segment-specific

Coverage Variable

(end of prior fiscal year) Neg. Bin. Logit OLS Tobit OLS OLS Neg. Bin. Logit OLS Tobit OLS OLS (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

Log Sales 0.171*** 0.208*** -0.001 0.015*** 0.007** -0.003 -0.001 0.414*** -0.001 0.019*** 0.000 0.000 (19.29) (13.82) (1.23) (7) (2.26) (1.55) (0.13) (8.62) (0.8) (7.92) (0.32) (0.25) Log trading volume 0.349*** 0.435*** 0.032*** 0.072*** 0.004* 0.002 0.193*** 0.684*** 0.017*** 0.05*** 0.006*** 0.001 (67.46) (37.17) (39.52) (44.96) (1.87) (1.61) (33.53) (25.06) (18.3) (27.57) (3.18) (0.76) Sales Growth (past 3 years) 0.189*** 0.21*** 0.005*** 0.022*** 0.019*** -0.001 0.138*** 0.059 0.004*** 0.01*** 0.032*** -0.001 (17.77) (10.59) (3.69) (7.78) (11.93) (0.91) (16.31) (1.5) (2.86) (4.03) (32.05) (1.12) Issuances in prior year 0.334*** 0.495*** 0.013*** 0.073*** 0.002 0.009*** 0.113*** 0.289*** 0.006*** 0.038*** 0.009** 0.009*** (16.35) (14.36) (5.33) (13.92) (0.37) (3.05) (6.21) (4.53) (2.65) (8.07) (2.2) (3.46) Earnings Vol. 0.005 0.013 -0.002*** -0.001 -0.001 0.0001 0.002 0.031* 0.0001 0.001 -0.002* -0.001** (0.93) (1.51) (3.91) (0.76) (0.65) (0.67) (0.54) (1.96) (0.07) (1.17) (1.93) (2.27) # of firms in industry -0.002*** 0.000 0.0001*** 0.0001*** 0.0001*** 0.0001*** -0.001*** -0.005*** 0.0001*** 0.0001*** 0.0001*** 0.0001*** (6.81) (0.04) (8.9) (5.08) (5.5) (3.53) (5.67) (4.9) (9.43) (10.28) (2.78) (5.34) ROE 0.294*** 0.331*** 0.016*** 0.045*** 0.004 0.0001 0.191*** 0.194*** 0.003* 0.019*** 0.005 0.0001 (15.35) (10.7) (9.49) (10.65) (1.26) (0.06) (12.85) (3.76) (1.74) (5.03) (1.55) (0.10) Log number of shareholders -0.084*** -0.108*** -0.003*** -0.011*** 0.0001 0.001 -0.087*** -0.021 -0.001 -0.006*** 0.000 0.000 (21.11) (11.02) (4.01) (9.3) (0.16) (0.6) (16.15) (0.63) (0.91) (3.66) (0.45) (0.54) Segment 2 -0.121*** -0.114*** 0.0001 -0.005 0.017*** 0.001 -0.082*** -0.034 0.003 0.003 0.015*** 0.003*** (7.94) (3.54) (0.16) (1.12) (7.08) (0.49) (5.93) (0.48) (0.99) (0.62) (9.75) (2.66) Segment 3 -0.12*** -0.081* 0.016*** 0.013** 0.13*** 0.078*** -0.098*** -0.107 0.008** 0.01* 0.147*** 0.13*** (6.04) (1.91) (4.81) (2.4) (10.49) (9.17) (5.55) (1.16) (2.31) (1.75) (14.37) (20.38) Segment 4 -0.219*** -0.258*** -0.007 -0.017** 0.307*** 0.072*** -0.116*** -0.014 0.000 0.000 0.526*** 0.056*** (7.54) (4.24) (1.35) (2.18) (19.24) (6.62) (4.92) (0.12) (0.08) (0.03) (49.71) (8.46) Segment 5 -0.286*** -0.40*** 0.001 -0.015 -0.003 0.003 -0.148*** -0.393** 0.007 0.002 -0.008** 0.008*** (7.54) (4.53) (0.13) (1.43) (0.69) (1.23) (4.81) (2.29) (1.19) (0.22) (2.5) (3.98) Segment 6 -0.389*** -0.418*** -0.002 -0.022 0.001 0.007* -0.236*** -0.699*** -0.002 -0.017 -0.002 0.012***

42

(6.72) (2.94) (0.17) (1.32) (0.15) (1.77) (5.38) (2.67) (0.17) (1.15) (0.45) (4.16) Segment 7 -0.726*** -1.38*** -0.062*** -0.134*** -0.025*** -0.001 -0.555*** -0.94*** 0.027** -0.005 -0.022*** 0.014*** (6.25) (6.65) (2.85) (4.8) (2.90) (0.10) (6.87) (2.78) (1.97) (0.21) (2.89) (2.89) Segment 8 -1.085*** -1.854*** -0.084*** -0.202*** -0.001 0.006 -0.866*** -1.369*** 0.029 -0.016 0.018 0.019** (7.23) (7.56) (3.08) (6.07) (0.08) (0.64) (7.43) (2.73) (1.66) (0.51) (1.36) (2.33) Log common equity 0.288*** 0.446*** 0.016*** 0.057*** 0.09*** -0.034* 0.156*** 0.393*** 0.012*** 0.041*** -0.010 0.024 (30.71) (26.54) (15.15) (24.58) (3.54) (1.94) (18.39) (9.93) (8.2) (16.43) (0.39) (1.57) Average HHI 0.436*** 0.517*** 0.201*** 0.384*** 0.074** -0.061*** 0.023 -0.638*** 0.148*** 0.293*** -0.008 -0.031 (8.86) (5.07) (18.64) (27.54) (2.18) (2.66) (0.5) (2.83) (18.05) (21.7) (0.22) (1.42) Average market share -0.133*** 0.939*** 0.762*** 0.82*** 0.173** -0.196*** 0.192*** 1.932*** 0.545*** 0.765*** 0.023 -0.08* (2.86) (7.5) (57.96) (55.65) (2.55) (4.22) (3.89) (5.93) (50.14) (46.17) (0.33) (1.78) RD_F (Barth et al., 2001) 1.774*** 2.462*** -0.027*** 0.177*** 0.021 0.018 0.42*** 2.495*** 0.031 0.207*** 0.022 -0.023** (17.18) (11.64) (2.82) (6.13) (0.65) (0.81) (3.56) (4.08) (1.42) (5.91) (1.21) (2.01) ADV_F -0.164 0.723 0.16*** 0.171* 0.125 0.028 -0.099 2.949 0.123* 0.174* -0.11* 0.061 (0.57) (1.09) (3.11) (1.94) (1.3) (0.43) (0.29) (1.62) (1.91) (1.67) (1.85) (1.63) INTANG_F -0.296*** -0.164 -0.033*** -0.085*** 0.017 -0.017* 0.333*** 2.055*** 0.026*** 0.010 -0.018* -0.011* (6.06) (1.41) (4.15) (5.79) (1.23) (1.81) (7.09) (7.35) (2.68) (0.67) (1.83) (1.82) DPT_F 0.617*** 0.922*** 0.028 0.135*** -0.061 0.033 -0.209 -0.549 -0.023 0.011 0.076** -0.05*** (3.75) (2.94) (1.44) (3.17) (1.36) (1.06) (1.28) (0.7) (0.85) (0.22) (2.52) (2.65) UNIQUE1 -0.249*** -0.588*** -0.084*** -0.144*** -0.043*** -0.029*** -0.248*** -0.695*** -0.054*** -0.111*** -0.069*** -0.026*** (6.64) (7.78) (13.53) (14.15) (5.17) (5.08) (8.02) (4.81) (10.26) (12.23) (9.07) (5.39) S&P 600 Index Membership 0.229*** 0.788*** 0.014*** 0.042*** -0.004 -0.002 0.041*** 0.249** -0.015*** -0.014*** -0.009*** 0.005*** (18.46) (17.36) (5.01) (8.93) (1.05) (0.65) (2.81) (2.41) (4.93) (2.92) (3.0) (2.7) Regulated industry Indicator -0.244* -0.552** -0.018 -0.07*** -0.1*** 0.029 -1.504*** -1.057* -0.047*** -0.066*** -0.007 -0.001 (1.73) (2.51) (1.44) (2.88) (2.73) (1.16) (34.33) (1.87) (2.93) (5.18) (0.22) (0.05) Reg F.D. dummy -0.768*** 1.239*** -0.038*** -0.013 -0.088*** -0.08*** -17.702 1.696*** 0.005 -0.045*** -0.11*** -0.078*** (17.76) (14.69) (5.99) (1.2) (9.79) (12.95) (0.05) (11.39) (1.15) (4.8) (11.88) (13.46) Observations 59,158 59,158 59,158 59,158 26,140 26,140 59,158 59,158 59,158 59,158 26,140 26,140 Chi-squared - 15030.8 - 24506.5 - - 11239.6 5629.8 - 10415.8 - - R-squared - - 53.4% - 70.7% 78.4% - - 78.7% - 49.5% 68.7% Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Industry Fixed Effects No No No No No No Yes Yes Yes Yes Yes Yes

Firm Fixed Effects Yes Yes Yes Yes Yes Yes No No No No No No

Table 4. Analyst coverage and corporate strategy conformity: first differences regressions. OLS regressions of various measures of coverage as in Table 3. All control variables are defined in Table 1. The t-statistics (in parentheses) are based on robust standard errors cluster-adjusted at the firm level. The ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

∆Log (1+Coverage

Count) ∆Adj. Coverage ∆Strategy-specific

Coverage ∆Segment-specific

Coverage ∆Log Sales 0.081*** 0.005*** 0.011*** -0.005** (11.64) (3.06) (3.53) (2.47) ∆Log trading volume 0.042*** 0.002** 0.004** 0.002 (9.36) (2.39) (2.03) (1.67) ∆Sales Growth (past 3 years) 0.053*** 0.004*** 0.005*** -0.001 (17.82) (5.58) (2.81) (0.77) ∆Issuances in prior year indicator 0.006 0.002 0.003 0.001 (1.3) (1.68) (0.79) (0.32) ∆Earnings Vol. 0.001 0.000 0.000 0.000 (1.15) (0.39) (0.66) (0.1) ∆(# of firms in industry) -0.001*** 0.001*** 0.001** 0.001*** (2.84) (6.6) (2.26) (6.42) ∆ROE -0.055*** -0.005*** -0.011*** 0.001 (10.25) (4.35) (3.92) (0.69) ∆Log number of shareholders 0.009* 0.002** 0.003* -0.001 (1.95) (2.45) (1.85) (0.84) ∆Segment 2 0.109*** 0.011*** 0.019*** -0.003 (16.37) (7.13) (7.52) (1.53) ∆Segment 3 -0.003 0.13*** 0.106*** 0.09*** (0.12) (9.65) (3.74) (5.25) ∆Segment 4 0.157*** 0.374*** 0.237*** 0.085*** (3.61) (14.17) (5.78) (3.46) ∆Segment 5 -0.012 0.019*** 0.03*** 0.002 (1.24) (6.66) (5.73) (0.5) ∆Segment 6 -0.012 0.029*** 0.046*** 0.003 (0.93) (7.51) (5.37) (0.51) ∆Segment 7 -0.021 0.038*** 0.055*** 0.006 (1.33) (6.72) (4.16) (0.79) ∆Segment 8 -0.038** 0.042*** 0.072*** 0.004 (1.98) (5.41) (3.85) (0.29) ∆Log common equity -0.066** 0.06*** 0.079*** 0.02* (1.99) (4.72) (2.87) (1.75) ∆Average HHI -0.084* 0.034 0.045** 0.044*** (1.8) (1.58) (2.11) (3.77) ∆Average market share -0.123* 0.023 0.050 0.054*** (1.76) (0.6) (0.96) (3.39) ∆RD_F 0.223*** 0.011 0.027* -0.008 (3.12) (1.07) (1.81) (1.08) ∆ADV_F 0.492** 0.099* -0.057 0.148 (2.18) (1.89) (0.45) (1.5) ∆INTANG_F 0.144*** 0.035*** 0.029* -0.019* (4.08) (3.75) (1.95) (1.74) ∆DPT_F 0.011 -0.011 0.013 0.011 (0.13) (0.61) (0.35) (0.52) ∆UNIQUE1 -0.037** -0.124*** -0.126*** -0.069*** (2.12) (16.33) (9.46) (9.6) ∆S&P 600 Index Membership Indicator 0.033** -0.002 -0.002 -0.001

44

(2.21) (0.52) (0.39) (0.18) ∆Regulated industry Indicator 0.057 -0.038* -0.040 0.013 (1.32) (1.71) (1.01) (0.63) ∆Reg F.D. dummy -0.02** 0.004 -0.007 0.012*** (2.11) (1.42) (1.61) (4.23) Observations 56,979 56,979 23,426 23,426 R-squared 3.1% 4.8% 2.7% 1.4%

45

Table 5. Corporate value, analyst adjusted coverage, and corporate strategy uniqueness. The table presents the second stage of the two-stage least squares (2SLS) of the effects of analyst coverage and corporate strategy uniqueness on Tobin’s q, the diversification discount, and the uniqueness premium. We measure analyst coverage as adjusted coverage. We treat adjusted analyst coverage as endogenous variable and instrument it with the log trading volume and a dummy variable for S&P 600 membership. All variables are defined in Table 1. All regressions include year- and firm fixed effects. The t-statistics (in parentheses) are based on robust standard errors cluster-adjusted at the firm level. The ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

M/B

Adj. M/B #1 (Diversification

Discount)

Adj. M/B #2 (Uniqueness

Premium) (1) (2) (3) Log Sales -0.107*** -0.11*** -0.101*** (6.68) (7.82) (7.17) Earnings Vol. -0.008 -0.009* -0.008 (1.31) (1.71) (1.38) ROE 0.158*** 0.128*** 0.135*** (7.67) (7.06) (7.41) Log number of shareholders -0.051*** -0.031*** -0.035*** (4.14) (2.87) (3.18) Log common equity -0.404*** -0.061** -0.065** (20.24) (2.41) (2.57) Average HHI -1.841*** -0.17*** -0.174*** (12.47) (4.92) (5.02) Average market share -5.36*** -0.172*** -0.181*** (13.06) (3.86) (4.04) Segment 2 -0.119*** -0.232*** -0.243*** (4.15) (3.79) (3.94) Segment 3 -0.239*** -0.27*** -0.279*** (6.08) (3.03) (3.11) Segment 4 -0.256*** -0.464*** -0.483*** (5.04) (3.4) (3.52) Segment 5 -0.338*** -0.491*** -0.436** (4.83) (2.74) (2.42) Segment 6 -0.449*** -0.343*** -0.351*** (4.41) (19.53) (19.84) Segment 7 -0.685*** -1.393*** -1.41*** (4.38) (10.75) (10.84) Segment 8 -0.64*** -4.383*** -4.537*** (3.12) (11.96) (12.32) Instrumented Adjusted Coverage 9.294*** 7.8*** 7.92*** (14.39) (13.6) (13.75) UNIQUE1 1.183*** 0.938*** 0.964*** (12.03) (10.9) (11.16) Regulated industry Indicator 0.072 0.234 0.298* (0.42) (1.55) (1.96) Reg F.D. dummy 0.467 0.685*** 0.669*** (0.52) (12.06) (11.72) Observations 59,158 59,158 59,158 Model Chi-squared statistic (p-value) 82542.8 (0.0) 1634.0 (0.0) 1717.5 (0.0)

46

Table 6. Panel A. Corporate value, analyst adjusted coverage, and corporate strategy conformity. The table presents the second stage of the two-stage least squares (2SLS) of the effects of adjusted coverage and corporate strategy uniqueness on measures of real performance: sales growth over the subsequent five years). We treat adjusted analyst coverage as endogenous and instrument it with the log trading volume and an indicator S&P 600 membership. We show only the main variables of interest, but control for all variables as in Table 5, Panel A. All variables are defined in Table 1. All regressions include year fixed effects. Models (1) through (6) include firm fixed effects, models (7) through (12) include industry effects (four-digit SIC). The t-statistics (in parentheses) are based on robust standard errors cluster-adjusted at the firm level. The ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

Sales growth: 1-year 2-years 3-years 4-years 5-years 1-year 2-years 3-years 4-years 5-years (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Segment 2 -0.016*** -0.041*** -0.058*** -0.062*** -0.05*** -0.012*** -0.043*** -0.062*** -0.072*** -0.063*** (3.12) (5.49) (6.4) (6.37) (4.34) (3.53) (6.58) (7.55) (7.64) (5.42) Segment 3 -0.018** -0.048*** -0.06*** -0.051*** -0.032* -0.02*** -0.044*** -0.058*** -0.053*** -0.034** (2.57) (4.72) (4.7) (3.62) (1.82) (4.39) (5.16) (5.17) (4.07) (2.01) Segment 4 -0.008 -0.03** -0.053*** -0.063*** -0.036* -0.012* -0.041*** -0.068*** -0.085*** -0.065*** (0.86) (2.26) (3.31) (3.61) (1.69) (1.83) (3.55) (4.72) (5.14) (3.12) Segment 5 -0.016 -0.041** -0.072*** -0.061** -0.048 -0.028*** -0.052*** -0.086*** -0.083*** -0.066** (1.32) (2.28) (3.23) (2.5) (1.61) (3.08) (3.24) (4.24) (3.59) (2.28) Segment 6 -0.006 -0.022 -0.035 -0.047 -0.031 -0.015 -0.021 -0.033 -0.049 -0.032 (0.35) (0.85) (1.13) (1.38) (0.75) (0.99) (0.88) (1.1) (1.47) (0.77) Segment 7 0.024 0.025 0.006 -0.026 -0.014 0.06** 0.035 0.017 -0.023 0.004 (0.88) (0.63) (0.12) (0.49) (0.22) (2.59) (0.95) (0.36) (0.45) (0.06) Segment 8 0.026 0.085* 0.133** 0.176*** 0.278*** 0.038 0.045 0.079 0.121** 0.21*** (0.77) (1.74) (2.33) (2.98) (4.04) (1.43) (0.98) (1.42) (2.03) (2.98) Instrumented Adjusted Coverage 0.914*** 1.274*** 1.387*** 1.057*** 1.121*** 0.559*** 0.972*** 1.066*** 0.873*** 0.839*** (8.06) (7.48) (6.5) (4.38) (3.81) (15.59) (8.86) (6.87) (4.38) (3.26) UNIQUE1 0.086*** 0.114*** 0.121*** 0.065* 0.066 0.039*** 0.059*** 0.052* 0.010 -0.007 (4.97) (4.37) (3.62) (1.72) (1.46) (4.05) (2.91) (1.88) (0.31) (0.17) Observations 59,158 54,615 48,592 43,153 38,138 59,158 54,615 48,592 43,153 38,138 Quasi R-squared 32.4% 46.5% 56.1% 63.2% 70.2% 4.3% 4.3% 5.7% 6.6% 7.8% Industry Fixed Effects No No No No No Yes Yes Yes Yes Yes Firm Fixed Effects Yes Yes Yes Yes Yes No No No No No

Corporate Strategy, Analyst Coverage, and the Uniqueness ... · Corporate Strategy, Analyst...

Documents

Transcript of Corporate Strategy, Analyst Coverage, and the Uniqueness ... · Corporate Strategy, Analyst...