Division Manager Lobbying Power and the Allocation of Capital

Authors


  • The authors thank Cynthia J. Campbell, editor, an anonymous referee, Eric Powers, Greg Niehaus, and seminar participants at the 2004 FMA meetings for their helpful comments.

* Corresponding author: Furman University, 3300 Poinsett Highway, Greenville, SC 29613; Phone: (864) 294-3312; Fax: (864) 294-2990; E-mail: thomas.smythe@furman.edu

Abstract

We investigate whether division manager lobbying power affects the allocation of capital in multi-divisional firms. We find that firm-level disparities in division manager lobbying power (measured by tenure, time-in-position, board membership, and top executive status) affect capital allocation in expected ways. Managers with greater relative lobbying power compete for capital expenditures from a position of strength. Evidence is also provided which suggests that division manager ownership mitigates lobbying efforts. Furthermore, disparity in division manager lobbying power is associated with lower firm excess value. These results support the view that division manager influence activities impact the operation of internal capital markets.

1. Introduction

We examine the proposition that greater division manager lobbying power translates into greater capital expenditures for the manager's division. The empirical results support this proposition, which we call the Lobbying Power Hypothesis. Managers with greater lobbying power appear to compete for capital expenditures from a position of strength, based on regressions of segment level capital expenditures on lobbying power (controlling for division investment opportunities, size, age, and parent level capital constraints). It is estimated that more powerful managers garner 12.5% higher capital expenditures than other managers. If capital is allocated efficiently, capital expenditures should only be related to investment opportunities after controlling for firm-level capital constraints. In addition, univariate analysis provides evidence that divisions with poor investment opportunities receive more capital expenditures when managed by a powerful manager, while good divisions receive less when managed by a less powerful manager. In total, the results suggest that division manager influence activities hinder internal capital market efficiency (as proposed by Scharfstein and Stein, 2000; Wulf, 2002a, 2002b).

We acknowledge the possibility of endogeneity problems when regressing segment capital expenditures on manager lobbying power. Measures of segment investment opportunities are imperfect. It is possible that measures of lobbying power reflect differences in segment investment opportunities not captured by our investment opportunity measures. For example, managers with greater lobbying power could gravitate to divisions with better investment opportunities, or CEOs could place more powerful managers as heads of more important divisions. However, we also find evidence that division manager ownership in the firm weakens the association between capital allocation and manager lobbying power. While not conclusive, the ownership effect suggests that endogeneity does not drive the association between capital expenditures and lobbying power.

To address the endogeneity concern further, we use a variety of empirical approaches in an effort to ascertain the true direction of causality in our results. While none of the approaches by themselves fully address the problem of endogeneity, we believe the collective evidence provides strong support for the Lobbying Power Hypothesis. We first use a three-stage least squares approach by modeling both segment capital expenditures and our measure of lobbying power. Our results using this approach provide support for our hypothesis, although the results are sensitive to specification. We also examine changes in capital expenditures around the time of negative industry shocks. An exogenous shock to investment opportunities is expected to mitigate any endogenous correlation between lobbying power and investment opportunities, and thereby capital expenditures. Other studies that use an exogenous shock in a similar way to address endogeneity (in various settings) include Khanna and Tice (2001), Campello (2002), Lemmon and Lins (2003), and Goyal and Yamada (2004). The results are consistent with the baseline analysis. More powerful division managers experience smaller declines in capital allocation than less powerful managers. As a third test of whether the link between division manager lobbying power and capital expenditures is explained by endogeneity issues, we examine the association between manager lobbying power and firm value. If lobbying activities sufficiently disrupt the allocation of capital, firms with greater diversity in division manager lobbying power will have lower value. The results show that a one-standard-deviation increase in lobbying power diversity is associated with an 11.5% decrease in firm excess value. Based on the overall results, we conclude that division manager lobbying power adversely affects the capital allocation process.

In addition to extending the literature on the firm capital allocation process, this study has practical implications for firms and corporate governance. Our results demonstrate how manager characteristics that are readily observable can negatively influence the one process likely to have the greatest impact on shareholder wealth maximization, the capital allocation process. Additionally, we find evidence that higher levels of division manager ownership help to mitigate the problem, consistent with Jensen and Meckling (1976). Armed with this information, headquarters can take steps to mitigate the self-serving influence activities of higher ranking division managers through compensation contracting (Wulf, 2002a) and better monitoring of the capital allocation process. In addition, boards and other monitoring agents can prod headquarters to undertake these steps.

2. Literature review

In the literature, explanations for conglomerate internal capital market inefficiency largely rely on theorized agency problems at the division manager level arising from incomplete information, limited control over division managers, and incentive problems.1 For example, Harris and Raviv (1996) propose that headquarters reacts to division manager incentives to misrepresent investment opportunities by implementing spending limits that lead to over (under)-investment in low (high) growth divisions. Scharfstein and Stein (2000) construct a two-stage agency model where CEOs use capital expenditures to mitigate the influence activities of division managers. In their “diversity cost model,”Rajan, Servaes and Zingales (2000) posit that imperfect contracting motivates division managers to undertake self-serving “defensive” investments that enable their division to retain surpluses.

A number of studies propose that division managers derive private benefits from capital expenditures. Harris, Kriebel and Raviv (1982) theorize that capital spending reduces division level managerial effort needed to meet operating performance goals. Scharfstein and Stein (2000) suggest capital expenditures can be used to “bribe” division managers. Also, similar to CEOs, division managers can derive private benefits (such as greater pay, prestige, and enhanced job prospects) from managing larger operations.2 These private benefits motivate division managers to lobby headquarters for greater capital expenditures.

Despite the theoretical focus on division manager agency problems in models of internal capital markets, few empirical studies directly examine the role of division managers in the allocation of capital, likely due to the difficulty of obtaining detailed information about such managers. One exception is Wulf (2002a) who provides evidence that capital expenditures for a firm's smallest segment are less responsive to the segment's profitability when managerial compensation is more closely linked to overall firm performance. This evidence, she argues, indicates that headquarters places greater weight on “subjective” information from managers when manager interests are more closely aligned with shareholders. Wulf (2002b) proposes that division managers use their influence to misrepresent information about other divisions' investment prospects and shows that capital expenditures for a firm's smallest segment are less responsive to the segment's profitability when the largest and smallest segments are more closely related.

3. Sample

To test the Lobbying Power Hypothesis, we construct a sample of firms for which we are able to identify division managers for all or all but one of the firm's operating segments (i.e., segments assigned valid SIC codes in the Compustat database). We require division managers to be distinct from the CEO of the parent, since we are investigating agency issues between the division managers and headquarters. Obtaining detailed information for division managers involves close inspection of annual reports and proxy statements. We first select all firms listed in Compustat's business segment file in 1999, which are incorporated in the United States, whose largest segment provides no more than 90% of total sales, and whose total book value of assets is at least $100 million. The year 1999 is chosen since it is the first full fiscal year that accounting standards direct segment reporting to correspond to “…the way that management organizes the segments within the enterprise for making operating decisions and assessing performance” (FASB 131, paragraph 4). Under FASB 131, segment reporting is more likely to match organizational structure. We exclude smaller firms since it is less likely that annual reports and proxy statements will be available. Following the literature, firms with segments operating in the financial services industries are excluded due to the difficulty in measuring their investment opportunities. We exclude firms having one or more segments with only international operations, again because of the difficulty in measuring investment opportunities. We require the sum of segment reported assets be within 25% of firm assets to reduce problems associated with a firm's segment reporting. Finally, we exclude firms with missing capital expenditure data for any operating segment.

It is not possible to identify division managers for many firms since organizational structures sometimes differ from segment reporting (despite the new accounting standards) and many firms provide very limited descriptions of manager responsibilities. An inspection of roughly 590 firms yields a sample of 120 firms with 330 division managers, where each division manager is in place at least one quarter prior to the beginning of the fiscal year.3 To satisfy the last requirement, we shift the analysis of capital expenditures either one fiscal year forward to 2000 (for 29 firms) or back to 1998 (for 13 firms) from 1999.

For each division manager, we collect the following information as of the beginning of the fiscal year: tenure, time-in-position, whether the manager is listed as one of the firm's top five executives, and membership on the firm's board of directors. Tenure, time-in-position, and board membership are gleaned from 10 Ks, annual reports, proxy statements, and news reports from Lexis–Nexis. Top executive status comes from Compustat's Execucomp database and proxy statements.

As shown in Table 1, sample firms are relatively large with average sales of $2.9 billion and average assets of $2.4 billion. However, firm size varies considerably across the sample, with total assets ranging between $103 million and $24.4 billion. On average, sample firms report 2.8 operating segments (i.e., segments with assigned SIC codes) and operate in 1.8 industries (as defined at the two-digit SIC level). The segments are also large on average with mean sales of $1 billion and total assets of $804 million. Following Rajan, Servaes and Zingales (2000), we divide capital expenditures by beginning assets, which equals year-end book value of assets minus capital expenditures plus depreciation. The ratio of capital expenditures to beginning assets for business segments is winsorized at the sample's 5th and 95th percentiles. Based on the mean (0.017) and median (0.007) of the industry adjusted ratio of capital expenditures in our sample, segment capital expenditures exceed those of single segment firms operating in the same industry.

Table 1. 
Descriptive statistics
This table presents descriptive statistics for a sample of 120 multi-segment firms and their 330 segments with identified managers for one year (either 1998, 1999 or 2000 as determined by available data). Sales and assets are in millions of dollars. Capital expenditures is the ratio of capital expenditures to beginning-of-year assets, where beginning-of-year assets equal end-of-year assets minus capital expenditures plus depreciation. Segment level ratio of capital expenditures is winsorized at the 5th and 95th percentile for the sample. Segment level industry adjusted ratio of capital expenditures equals the segment's ratio of capital expenditures minus the industry median.
 MeanMedianQuartile 1Quartile 3MinimumMaximum
  1. ***Denotes statistical significance at the 0.01 level based on the t-test for the mean and the Wilcoxon rank-sum test for the median.

Parent level
Sales2,8871,1295202,2178633,702
Assets2,4079654582,81610324,366
Capital expenditures (ratio)0.0610.0500.0350.0760.0080.287
#segments2.783.002.003.002.006.00
#industry groups1.822.001.002.001.004.00
Segment level
Sales1,05836315489111.326,080
Assets804305122891412,048
Capital expenditures (ratio)0.0630.0520.0290.0840.0090.184
Capital expenditures (ratio), industry adjusted0.017***0.007***−0.0120.038−0.1100.162

The 330 segments in our sample represent 48 industries, defined at the two-digit SIC level. Industrial, Commercial Machinery and Computer Equipment is the most frequently represented industry with 44 segments (13.3% of the sample). The second most common industry is Chemicals and Allied Products with 27 segments or 8.2% of the sample.

Table 2 contains additional summary statistics but for segment managers. Forty division managers (12.1% of the sample) serve on their parent firm's board of directors, while 68% (225) are one of the parent's top five executives, as listed in the firm's proxy statement and ranked by salary before bonuses and other compensation. Just under 25% have tenure of four years or less with the parent firm, and 46.4% have been the segment head for two years or less.4

Table 2. 
Segment manager characteristics
This table presents descriptive statistics for a sample of 330 segment managers from 120 multi-segment firms for one year (either 1998, 1999 or 2000 as determined by available data). Tenure is the segment manager's number of years with the firm and Time-in-position is the number of years the manager has served as segment head. Director denotes membership on the firm's board of directors. Top Exec indicates that a segment manager is one of the top five executives in a firm based on the firm's proxy filing.
Years≤1234≥5N
Tenure
 Frequency 29261017248330
 Proportion 8.8% 7.9% 3.0% 5.2% 75.2% 
 Cumul. propor. 8.8%16.7%19.7%24.9%100.0% 
Time-in-position
 Frequency 98553434109330
 Proportion29.7%16.7%10.3%10.3% 33.0% 
 Cumul. propor.29.7%46.4%56.7%67.0%100.0% 
FrequencyProportion    
Director 4012.1%    
Top Exec22568.2%    

4. Research design and results

4.1. Measuring lobbying power

The lobbying power and lobbying efforts of division managers are not, practically speaking, directly observable. We estimate a manager's lobbying power based on four observable characteristics that likely reflect or play a part in determining such power: tenure, time-in-position, top executive status, and board membership. From these four characteristics, we construct a composite measure of relative lobbying power, which we call RPI (Relative Power Index). Before describing the RPI variable, we first briefly discuss the theoretical connection between lobbying power and each of the four characteristics, beginning with tenure.

Less tenured managers have had less opportunity to build relationships with headquarters' personnel and to learn how the firm's capital allocation process works (particularly its bureaucratic nuances). Furthermore, less tenured managers have a limited performance record within the firm that could lead headquarters to view their capital budgeting requests with greater skepticism, even when the less tenured manager heads a segment with relatively good investment opportunities. Regarding time-in-position, managers could require time to consolidate their power base as segment manager in the capital allocation process. In addition, similar to tenure, headquarters may tend to impose soft rationing on a segment until its manager has a track record of performance in that position.

Top executive status can confer lobbying power. U.S. securities law requires publicly traded firms to identify executive officers, defined as persons who perform a policy-making function. By identifying an employee as an executive officer, the firm is explicitly indicating that these individuals have greater authority over firm operations than other firm employees. Being higher up in the firm's power structure, top executive segment managers lobby from a position of strength when competing for capital expenditures against nontop executive segment managers.

Lastly, board membership arguably provides greater lobbying power to a segment manager. A primary responsibility of the board is to monitor CEO performance and take disciplinary action as needed. The segment manager who is also a director is thus in a position to pledge his loyalty to the CEO in exchange for side payments, such as greater capital allocation. In addition, board membership suggests that the manager has greater authority over the firm's operations than managers who are not board members.

From the four characteristics, tenure, time-in-position, top executive, and board membership, we construct a composite measure of a manager's overall lobbying power, denoted as Power Index. The construction of Power Index is presented in Equation (1).

image(1)

In Equation (1), Tenure≥ 5 equals one when a manager has tenure of five years or more. Time≥ 3 takes a value of one when a manager heads a segment for three years or more. Top Exec equals one when a manager has top five executive status, and Director equals one when a manager is on the firm's board of directors. Power Index ranges from zero to four with higher values indicating greater lobbying power.

The classification point for tenure (Tenure≥ 5) of five years is the upper limit of our tenure data. In many cases, tenure cannot be reliably ascertained beyond five years. Approximately one-quarter of our managers have tenure of four years or less, making five years a natural classification point. The classification point for time-in-position (Time≥ 3) is the sample median.

One possible shortcoming of Power Index is that it weights each characteristic equally; however, assigning varying weights to the characteristics would be arbitrary, as there is no theoretical guidance as to how weights should otherwise be assigned. Another possible concern is that the four characteristics chosen to construct Power Index are not independent but could be highly correlated. For the pair wise correlations of the variables making up the power index, all but one correlation coefficient is less than 0.20. The Tenure≥ 5, Time≥ 3 pair is approximately 0.41. For the most part, each characteristic seems to contribute a distinct dimension in the makeup of the Power Index, although there is some overlap in the tenure and time-in-position variables. For completeness, we also present evidence on how each relative manager characteristic affects the capital allocation process using separate regressions.

It is differences in lobbying power among a firm's managers that is expected to affect the allocation of capital, with more capital flowing to more powerful managers. RPI denotes the Relative Power Index and equals a manager's Power Index minus the average Power Index for all division managers in the firm, as shown in Equation (2) for division manager j.5

image(2)

Correlation between RPI and manager relative base salary (computed as manager salary divided by average salary for all division managers in the firm) is 0.364, significant at the 1% level. The correlation provides a degree of validation of RPI as a measure of manager power, assuming that more powerful managers negotiate bigger compensation packages for themselves. We do not incorporate base salary into RPI because salary is not available for many managers.

4.2. Segment level regressions of capital expenditures

We first test the Lobbying Power Hypothesis by estimating models of segment level capital expenditures and examining the explanatory power of RPI, our measure of relative division manager lobbying power. The regressions also include controls for the following factors that could help explain segment level capital expenditures: segment investment opportunities, segment size, segment age, industry effects, and parent level capital constraints.

The dependent variable in the regressions (CAPEX) is constructed as shown in Equation (3), following an approach similar to Rajan, Servaes and Zingales (2000). To control for parent level capital constraints and industry effects, the industry median ratio of capital expenditures and the firm's average ratio of industry adjusted capital expenditures for all of a firm's segments are subtracted from the ratio of segment level capital expenditures for segment j.

image(3)

In Equation (3), Ij denotes segment j's capital expenditures, Aj is the segment's beginning assets, Indj denotes the median for single segment firms in segment j's industry, and n is the number of firm segments.6 The bracketed summation term captures the firm's tendency to invest more or less than the industry for the year, reflecting the level of capital constraints imposed by headquarters and the firm's segment level investment opportunities relative to those of industry median single segment firms. Each segment is weighted equally when constructing the bracketed term, a departure from Rajan, Servaes and Zingales' (2000) asset weighted average. Asset weighted averages push a firm's largest segment capital expenditure measure toward zero.

As proxies for segment investment opportunities at the beginning of the fiscal year, we use three alternative measures: industry median q (Qind); lagged return on assets (ROA); and an estimated Tobin's q (Qest) similar to that of Billett and Mauer (2003). While industry median q (Qind) has been criticized as a poor proxy for segment investment opportunities (e.g., Whited, 2001), we still use it as one alternative since it is commonly employed in other studies (e.g., Shin and Stulz, 1998; Rajan, Servaes and Zingales, 2000). Qind is taken at the most detailed SIC level where at least five single segment firms are available for its computation.7 Lagged return on assets, denoted as ROA, equals segment operating income before depreciation divided by segment total assets, with each item coming from the prior year. Billett and Mauer (2003) use lagged return on assets as a measure of segment investment opportunities.

Qest is computed as follows. Observed values of Tobin's q, ROA′ (defined as contemporaneous EBITD/assets), ROANEG′ (equals ROA′ when ROA′ is negative and zero otherwise), SIZE (log of assets), and TURNOVER (sales/assets) for all single segment firms in the segment's two-digit SIC industry are used to estimate yearly regressions of Tobin's q.8 As shown in Equation (4), the coefficient estimates from the yearly regressions are then used to estimate the segment's q (i.e., Qest) as of the previous fiscal year-end, given the segment's ROA′, SIZE, and TURNOVER.

image(4)

Segment values of ROA′, SIZE, and TURNOVER are taken as of the previous fiscal year end. The above construction of Qest differs from that of Billet and Mauer (2003) by the incorporation of ROANEG′. Without the ROANEG′ term, coefficient estimates for ROA′ are negative for a number of industry years in the single segment firm regressions of Tobin's q, a result that is counter-intuitive.

Since the analysis examines the intra-firm allocation of capital, a segment's investment opportunities relative to the firm's other segments is expected to be important. We compute relative measures of Qest, ROA, and Qind by subtracting the average of the respective measure for all the firm's segments (as shown for Qest in Equation (5)). The relative versions are denoted as RelQest, RelROA, and RelQind.

image(5)

where n is the number of the firm's segments in our sample. Before computing the relative measures, Qest, ROA, and Qind are winsorized at our sample's 5th and 95th percentiles, reducing the potential impact of outlier observations on the relative measures of other segments. Descriptive statistics for the variables employed in the capital expenditure regressions appear in Table 3.

Table 3. 
Summary statistics for model of segment capital expenditure variables
This table presents summary statistics for a sample of 330 segments from 120 multi-segment firms for one year (either 1998, 1999 or 2000 as determined by available data). CAPEX is formulated as follows for segment j (similar to Rajan, Servaes and Zingales, 2000). inline imagewhere I equals segment capital expenditures, A is the segment's beginning of period assets, Ind is the industry median for single segment firms, and n is the firm's number of segments. Power Index measures segment manager lobbying power. Power Index=Tenure≥ 5 +Time≥ 3 +Top Exec+Director where Tenure≥ 5 = 1 for managers with tenure ≥ 5 years; Time≥ 3 = 1 for managers with time-in-position ≥ 3 years; Top Exec= 1 for managers who are a top five executive; and Director= 1 for managers on the board of directors. RPI is Relative Power Index and equals the managers Power Index minus the average Power Index of all the firm's segment managers. Qest is the estimated q for segment j as of the beginning of the fiscal year, formulated as follows (similar to Billett and Mauer, 2003). inline imagewhere ROA′ equals segment EBITD/assets, ROANEG′ is ROA′ when it is negative and zero otherwise, SIZE is the log of segment assets, and TURNOVER is segment sales/assets. Coefficient estimates in the calculation of Qest are obtained from regressions of q using samples of single segment firms. RelQest measures segment investment opportunities relative to other segments in the firm and equals the segment's Qest minus the average Qest of all the firm's segments. Qest is winsorized at the 5th and 95th percentiles before the calculation of RelQest. ROA is lagged segment return on assets. Qind is lagged median q for the segment's industry. RelROA and RelQind are relative versions of ROA and Qind constructed like RelQest. LOGTA is the log of segment beginning-of-year assets. RELSIZE is segment beginning-of-year assets divided by the sum of beginning-of-year assets for all of the firm's segments. SEG AGE equals the number of years a segment has operated in a firm, truncated at five years. Ownership is the segment manager's percentage ownership in the firm.
  Mean Median Quartile 1Quartile 3 MinimumMaximum
CAPEX −0.0001 −0.0004 −0.01710.0154 −0.09970.1207
Power Index2.092.001.003.00 0.004.00 
RPI0.000.00−0.50 0.50 −2.25 2.00 
Qest  2.2362  2.0453  1.51772.7361  0.85914.6210
RelQest  0.0021  0.0010 −0.32440.3257 −2.10312.5079
ROA  0.1857  0.1737  0.12040.2285  0.04850.4044
RelROA  0.0001  0.0008 −0.04170.0398 −0.17790.1837
Qind  1.5252  1.3229  1.18501.6625  0.97412.8626
RelQind  0.0007  0.0000 −0.09410.0876 −1.34841.1477
LOGTA  5.7713  5.6930  4.80166.8022  1.36109.3011
RELSIZE  0.3628  0.3258  0.18720.5249  0.01420.9054
SEG AGE  4.7758  5.0000  5.00005.0000  1.00005.0000
Ownership  0.3889  0.1914  0.08070.4045  0.00008.7000

Regressions of CAPEX, shown in Table 4, support the Lobbying Power Hypothesis. In models (1) through (3), the coefficient estimates for RPI are positive and significant, with p-values ranging from 0.030 to 0.088. Furthermore, the RPI coefficient estimates are relatively stable as the investment opportunity measure is alternated between RelQest, RelROA, and RelQind. In models (4) through (6), an indicator variable RPI > 0 replaces RPI in the regressions. RPI > 0 equals one when a division manager has greater lobbying power than the firm's average division manager (i.e., when RPI is positive, which is the case for 127 of our 330 managers). The positive and significant coefficient estimates for RPI > 0 indicate that greater capital flows to more powerful managers. The results show that a one-standard-deviation change in RPI is associated with an increase of 6.3% in the ratio of segment capital expenditures, while the coefficient estimates for RPI > 0 indicate the ratio of capital expenditures is 12.5% higher for managers with positive relative lobbying power.9

Table 4. 
Regressions of segment capital expenditures: Power Index
Segment capital expenditures are regressed on segment manager Relative Power Index (RPI), a measure of segment manager lobbying power. The dependent variable CAPEX is formulated as follows for segment j. inline imagewhere I denotes segment capital expenditures, A is the segment's beginning of period assets, Ind denotes the industry median for single segment firms, and n is the firm's number of segments. Power Index=Tenure≥ 5 +Time≥ 3 +Top Exec+Director where Tenure≥ 5 = 1 for managers with tenure ≥ 5 years; Time≥ 3 = 1 for managers with time-in-position ≥ 3 years; Top Exec= 1 for managers who are a top five executive; and Director= 1 for managers on the board of directors. RPI is Relative Power Index and equals the manager's Power Index minus the average Power Index of all the firm's segment managers. RPI > 0 equals 1 when RPI is positive and 0 otherwise. RelQest measures segment investment opportunities relative to all segments in the firm and equals the segment's estimated q (i.e., Qest) minus the average Qest of all the firm's segments. RelROA and RelQind are constructed like RelQest but using segment lagged ROA and lagged median q for the segment's industry, respectively. I/A, Qest, ROA, Qind are winsorized at the 5th and 95th percentiles prior to construction of the relative measures CAPEX, RelQest, RelROA, RelQind. LOGTA is the log of segment beginning-of-year assets. RELSIZE is segment beginning-of-year assets divided by the sum of beginning-of-year assets for all of the firm's segments. SEG AGE equals the number of years a segment has operated in a firm, truncated at five years. Ownership is segment manager percentage ownership in the firm. The regressions also include year dummies 1998 and 2000. p-values appear in parentheses below the coefficient estimates and are based on standard errors robust to heteroskedasticity and correlation among segments of a firm.
Investment opportunity measure  (1) RelQest  (2) RelROA  (3) RelQind  (4) RelQest  (5) RelROA  (6) RelQind  (7) RelQest  (8) RelQest
  1. ***, **, *Denote significance at the 1%, 5%, and 10% levels, respectively.

Intercept0.00290.00760.0055−0.00020.00470.00160.00910.0054
(0.785)(0.466)(0.614)(0.982)(0.651)(0.8682(0.538)(0.723)
Investment opportunities0.0118***0.0762**0.00540.0122***0.0798***0.00540.0130***0.0131***
(0.000)(0.015)(0.441)(0.000)(0.010)(0.449)(0.001)(0.000)
RPI0.0050*0.0048*0.0062**   0.0081** 
(0.067)(0.088)(0.030)   (0.033) 
RPI > 0   0.0078*0.0072*0.0087** 0.0099**
   (0.050)(0.086)(0.038) (0.030)
LOGTA−0.0003−0.0002−0.0004−0.0002−0.0002−0.0004−0.0005−0.0005
(0.529)(0.613)(0.275)(0.585)(0.674)(0.328)(0.677)(0.702)
RELSIZE−0.0031−0.0062−0.0115−0.0020−0.0052−0.0104−0.0077−0.0085
(0.714)(0.460)(0.177)(0.805)(0.528)(0.215)(0.469)(0.429)
SEG AGE−0.0001−0.00090.0002−0.0001−0.00100.0002−0.0007−0.0007
(0.981)(0.661)(0.918)(0.959)(0.635)(0.916)(0.769)(0.789)
Ownership      0.00110.0037
      (0.298)(0.123)
RPI (RPI > 0) ×Ownership      −0.0100**−0.0088***
      (0.018)(0.002)
R20.0830.0510.0290.0840.0510.0260.1120.113
Observations   330   330   330   330   330   330   230   230

The regressions also include controls for segment size and age, as well as year dummies for 1998 and 2000 (year dummies are not reported to save space). LOGTA for segment j equals the natural log of the segment's beginning assets. RELSIZE for segment j is the ratio of segment j's beginning assets to the sum of beginning assets for all the firm's segments. SEG AGE has a range of one to five, denoting the age of the segment within the firm. None of the control variables are significant in the regressions of capital expenditures. In addition, the results do not differ qualitatively when indicator variables for a firm's largest and smallest segment replace RELSIZE, an indicator variable for segment age less than five years replaces SEG AGE, or industry affects are controlled by including relative industry capital expenditures as an explanatory variable (as opposed to industry adjusting the dependent variable).

In columns 7 and 8 of Table 4, we repeat the analysis from columns (1) and (4), but include a measure of division manager ownership. Based on Jensen and Meckling (1976) and Jensen and Murphy (1990), division manager ownership can reduce the affect of lobbying power on capital allocation if influence activities tend to destroy firm value and thereby lower the manager's wealth. Ownership equals the percentage of firm shares owned by a division manager. The interaction term composed of RPI (RPI > 0) and Ownership is negative and significant in the regressions, suggesting that division manager ownership mitigates the agency problem of lobbying power.

The association between CAPEX and RPI support the proposition that division manager influence activities impact the allocation of capital. Univariate analysis of CAPEX shows that the treatment of “good” and “poor” divisions differs with division manager lobbying power in a manner suggestive of relative inefficient investment. We divide the sample into four classes based on division investment opportunities being relatively good (RelQest > 0) or poor (RelQest < 0) and division manager lobbying power being high (RPI > 0) or low (RPI < 0). As shown in Table 5, “good” divisions with “low power” managers have CAPEX levels that do not statistically differ from zero (bottom right quadrant), while “good” divisions with “high power” exceed zero. Similarly, CAPEX for “poor” divisions with “low power” is statistically negative, while CAPEX for “poor” divisions with “high power” does not differ from zero.

Table 5. 
Univariate analysis of segment capital expenditures
This table contains an analysis of CAPEX for segments classified by relative investment opportunities and manager lobbying power. RelQest < 0 denotes poor relative investment opportunities, RelQest > 0 good. RPI > 0 denotes high lobbying power, RPI < 0 low. p-values appear in parentheses below the sample means (t-test) and sample medians (Wilcoxon rank sum test).
   Investment opportunitiesTest for difference
Poor (RelQest < 0)Good (RelQest > 0)
  1. ***, **, * indicate statistical significance at the 0.01, 0.05 and 0.10 level, respectively.

  Mean−0.00090.0111*** 
 High (RPI > 0) (0.839)(0.009)(0.050)*
  Median−0.00240.0089*** 
   (0.414)(0.007)(0.019)**
Lobbying power n6265 
 Mean−0.0113***0.0012 
Low (RPI < 0) (0.004)(0.690)(0.012)**
 Median−0.0109***−0.0001 
  (0.002)(0.926)(0.017)**
 n6664 
Test for differenceMean(0.078)*(0.059)* 
 Median(0.114)(0.069)* 

One question that arises from our analysis of RPI is whether and how each of the components of the index (tenure, time-in-position, director status, and top executive status) independently affects capital expenditures. While we believe that the relative Power Index provides a more robust measure of influence, we also estimate regressions similar to model one presented previously in Table 4 but alternately replace RPI with the relative measure of each manager characteristic used to construct RPI in the regression. The results are presented in Table 6.

Table 6. 
Regressions of segment capital expenditures: Relative manager characteristics
Segment capital expenditures are regressed on segment manager characteristics, as single components of segment manager lobbying power. The dependent variable CAPEX is formulated as follows for segment j. inline imagewhere I denotes segment capital expenditures, A is the segment's beginning of period assets, Ind denotes the industry median for single segment firms, and n is the firm's number of segments. Relative measures of manager characteristics are interchanged in the regression on capital expenditures. RelTen, RelTime, RelDir, and RelTopExec are relative measures of division manager tenure, time-in-position, director status, and top executive status within the firm. RelTen equals Tenure≥ 5 minus the average of Tenure≥ 5 for all the firm's division managers where Tenure≥ 5 = 1 for managers with tenure ≥ 5 years. RelTime, RelDir, and RelTopExec are constructed in the same fashion except using Time≥ 3, Director, and TopExec where: Time≥ 3 = 1 for managers with time-in-position ≥ 3 years; Director= 1 for managers on the board of directors; and Top Exec= 1 for managers who are a top five executive. RelQest measures segment investment opportunities relative to all segments in the firm and equals the segment's estimated q (i.e., Qest) minus the average Qest of all the firm's segments. I/A and Qest are winsorized at the 5th and 95th percentiles prior to construction of the relative measures CAPEX and RelQest. LOGTA is the log of segment beginning-of-year assets. RELSIZE is segment beginning-of-year assets divided by the sum of beginning-of-year assets for all of the firm's segments. SEG AGE equals the number of years a segment has operated in a firm, truncated at five years. Ownership is segment manager percentage ownership in the firm. The regressions also include year dummies 1998 and 2000. p-values appear in parentheses below the coefficient estimates and are based on standard errors robust to heteroskedasticity and correlation among segments of a firm.
Manager characteristic(1) RelTen(2) RelTime(3) RelDir(4) RelTopExec
  1. *** and ** indicate statistical significance at the 0.01 and 0.05 level, respectively.

Intercept−0.00110.0016−0.00050.0013
(0.922)(0.880)(0.960)(0.904)
RelQest0.0127***0.0121***0.0124***0.0123***
(0.000)(0.000)(0.000)(0.000)
Manager characteristic−0.00240.0116**0.00980.0083
(0.719)(0.033)(0.452)(0.221)
LOGTA−0.0003−0.0001−0.0002−0.0004
(0.547)(0.753)(0.565)(0.349)
RELSIZE−0.00080.0002−0.0019−0.0044
(0.923)(0.976)(0.820)(0.607)
SEG AGE0.0006−0.00020.00060.0005
(0.784)(0.926)(0.804)(0.799)
R20.0700.0890.0740.078
Observations 330 330 330 330

As is evident, a manager's relative time-in-position plays a statistically significant role in explaining capital expenditures in a way that is consistent with the Lobbying Power Hypothesis. Managers that have been division managers longer receive higher allocations of capital. The results in the other columns suggest that on their own, neither tenure, director status, nor top executive status significantly impact capital allocation. However, at least the sign of the coefficient estimates for director and top five executive is consistent with the Lobbying Power Hypothesis.

4.3. Endogeneity concerns

Endogeneity is a potential concern when interpreting the regressions of capital expenditures on lobbying power. While we argue that certain division manager characteristics indicate a manager's ability to distort the capital allocation process, an alternative possibility is that CEOs intentionally assign managers with such characteristics to lead divisions with better investment opportunities and where plans call for greater capital expenditures to grow that part of the firm. Additionally, manager lobbying power and segment investment opportunities are difficult to measure. It could be argued that the results are due to correlation between our measures of lobbying power and true segment investment opportunities. The finding that division manager ownership affects the association between CAPEX and RPI suggests otherwise but is not conclusive on its own.

To address the endogeneity concern further, we first examine the association between RPI and our three measures of investment opportunities. If RPI simply proxies for unobserved segment investment opportunities in the capital expenditure regressions, one would expect RPI to be related to our measures of investment opportunities assuming they are reasonable measures of such opportunities. First, we examine the pair wise correlations between RPI and each investment opportunity measure. In Panel A of Table 7, we see that correlation coefficients range from 0.033 to 0.198. The highest correlation is between RelROA and RPI. In contrast, RPI's association with RelQest and RelQind is quite low.

Table 7. 
Association between measures of lobbying power and investment opportunities
Panel A: Correlations This panel presents the pairwise correlations between RPI and the three measures of investment opportunities RelQest, RelROA, and RelQind. p-values are in parentheses.
 RelQestRelROARelQind
  1. *** and * indicate statistical significance at the 0.01 and 0.10 level, respectively.

RPI0.09935*0.19822***0.03299
(0.0715)(0.0003)(0.5504)
Panel B: Regressions
This panel presents regressions of investment opportunities, defined as RelQest, RelROA, and RelQind on the RPI. p-values are presented in parentheses.
Investment opportunity measure(1) RelQest(2) RelROA(3) RelQind
Intercept0.00210.00010.0007
(0.613)(0.340)(0.632)
Power Index0.08640.0179***0.0128
(RPI)(0.161)(0.001)(0.667)
R20.0100.0390.001
Observations330330330

In Panel B, we report statistics from regressing each investment opportunity measure on RPI. In the models where RelQest and RelQind are the dependent variables, RPI is not statistically different from zero, and the models have little explanatory power. In the regression of RelROA, the coefficient estimate of RPI is positive and significant with an R2 of 0.039. In sum, the results indicate some correlation between RPI and RelROA, but very little with RelQest and RelQind. The more material association between RelROA and RPI is possibly due to lobbying powers' effect on segment ROA. That is, the greater capital allocation received by more powerful managers can result in better return on assets ex post, as the work of Harris, Kriebel and Raviv (1982) suggests. In our opinion, the results in Table 7 do not indicate that RPI is simply proxying for segment investment opportunities. Nevertheless, additional tests of the Lobbying Power Hypothesis are conducted to further address the endogeneity issue. We proceed by employing three additional empirical approaches: examining segment capital expenditures and lobbying power as simultaneous equations, investigating the effect of lobbying power on capital expenditures around the time of exogenous investment opportunity shocks, and examining the effect of lobbying power diversity on firm value.

4.4. Capital expenditures and lobbying power as simultaneous equations

As one way to address the potential endogeneity and causality issues in regressing segment capital expenditures on manager lobbying power, we use three-stage least squares to model capital expenditures and lobbying power as simultaneous equations. The base system of equations for the two dependent variables is as follows:

image
image(7)

The equation for segment capital expenditures is the same as the one used earlier except that we incorporate industry affects by including industry level capital expenditures as a right-hand-side variable (RelIndCapx), as opposed to industry adjusting the dependent variable. This change has the advantage of strengthening identification. LOGTA, RELSIZE, and SEG AGE are viewed as pre-determined variables and are, therefore, treated as exogenous in the system of equations.

As shown in Equation (7), we construct a model of RPI that consists of a number of segment characteristics that one could theorize to be associated with division manager lobbying power. Unlike capital expenditures, there are currently no conventional models of lobbying power on which to base the equation for RPI. Thus, one caveat to consider while interpreting the results of the empirical approach is that our model of lobbying power could be lacking. For example, we lack additional division manager attributes on which to model RPI due to limited data availability.

The results are contained in Table 8. In the lower portion of the table, the results show that RelROA and RELSIZE are positively associated with RPI, while capital expenditures do not appear to explain RPI. In the upper portion of the table, the results provide evidence in support of the Lobbying Power Hypothesis; however, the support is sensitive to model specification. RPI is positively associated with capital expenditures (regressions (1) and (3)) but only when RelROA is available as an excluded instrument for RPI in the capital expenditure equation (i.e., when RelROA is an explanatory variable for RPI but not for CAPEX2). When RelROA is an explanatory variable for CAPEX2 but not for RPI, neither RPI nor CAPEX2 have significant coefficient estimates (results untabulated). Furthermore, when the industry adjusted version of capital expenditures is used (i.e., CAPEX), RPI coefficient estimates approach but do not reach statistical significance (results untabulated).

Table 8. 
Simultaneous equations: Capital expenditures and lobbying power
This table contains results from estimating simultaneous equations for segment capital expenditures and division manager lobbying power using three-stage least squares. The structural equations for segment capital expenditures and division manager lobbying power are as follows. inline imageCAPEX2 is segment ratio of capital expenditures to beginning of period assets minus the average ratio of capital expenditures to beginning of period assets over all of a firm's segments. Inv. Opp. denotes segment relative investment opportunities, alternately measured by RelQest and RelQind. RelQest equals the segment's estimated q (i.e., Qest) minus the average Qest of all the firm's segments. RelQind and RelIndCapx are constructed like RelQest but using lagged industry median q (Qind) and the industry ratio of capital expenditures to beginning of period assets (IndCapx). RPI is Relative Power Index and equals the manager's Power Index minus the average Power Index of all the firm's segment managers. Power Index=Tenure≥ 5 +Time≥ 3 +Top Exec+Director where Tenure≥ 5 = 1 for managers with tenure ≥ 5 years; Time≥ 3 = 1 for managers with time-in-position ≥ 3 years; Top Exec= 1 for managers who are a top five executive; and Director= 1 for managers on the board of directors. LOGTA is the log of segment beginning-of-year assets. RELSIZE is segment beginning-of-year assets divided by the sum of beginning-of-year assets for all of the firm's segments. SEG AGE equals the number of years a segment has operated in a firm, truncated at five years. RelROA is segment lagged ROA minus the average lagged ROA over all a firm's segments. The ratio of capital expenditures to assets, Qest, Qind, ROA are winsorized at the 5th and 95th percentiles prior to construction of the relative measures CAPEX2, RelQest, RelQind, RelIndCapx, and RelROA. p-values appear in parentheses to the right of each coefficient estimate.
Inv. Opp.RelQestRelROARelQind
Coeff. est.p-valueCoeff. est.p-valueCoeff. est.p-value
  1. ***, **, * indicate statistical significance at the 0.01, 0.05 and 0.10 level, respectively.

Dependent variable is CAPEX2
Intercept0.0151(0.338)0.1388(0.661)0.0260(0.131)
Inv. Opp.0.0076**(0.027)−0.5069(0.697)−0.0010(0.897)
RPI0.0233*(0.097)0.2528(0.642)0.0386***(0.005)
RelIndCapx0.6775***(0.000)0.9716(0.384)0.6794***(0.000)
LOGTA−0.0003(0.843)−0.0004(0.950)−0.0003(0.848)
RELSIZE−0.0099(0.364)−0.1285(0.630)−0.0203*(0.074)
SEG AGE−0.0021(0.436)−0.0188(0.639)−0.0036(0.247)
Dependent variable is RPI
Intercept−0.5337**(0.037)−0.5337***(0.037)−0.5231**(0.041)
CAPEX2−1.9791(0.712)−1.9791(0.712)−1.7284(0.753)
RelROA2.3103***(0.003)2.3103***(0.003)2.5207***(0.002)
RELSIZE0.5211***(0.007)0.5211***(0.007)0.4961***(0.010)
RelQest0.0499(0.552)0.0499(0.552)  
RelQind    0.0783(0.573)
SEG AGE0.0721(0.171)0.0721(0.171)0.0718(0.174)
Observations330 330 330 

Thus, the simultaneous equations approach offers mixed support for the Lobbying Power Hypothesis. The results show no indication that capital expenditures explain lobbying power, while providing some evidence that lobbying power affects capital expenditures. Given the mixed results, more evidence is needed to evaluate the Lobbying Power Hypothesis.

4.5. Industry investment opportunity shocks

As another test of the Lobbying Power Hypothesis, we examine the change in capital expenditures around the time of exogenous industry investment opportunity shocks, both positive and negative, for our sample of 330 segments and their managers. Incorporating an exogenous shock is intended to mitigate any endogeneity problem. Ideally, one would consider an exogenous shock to division manager lobbying power, but we know of no such shock that can reasonably be observed. We can, however, identify exogenous shocks to segment investment opportunities and examine whether the firm's reaction to the shock (as measured by change in capital expenditures) is conditioned on division manager lobbying power. The exogenous shock to segment investment opportunities is expected to mitigate any endogenous correlation between division manager lobbying power and segment investment opportunities, and thereby, segment capital expenditures. This approach to addressing endogeneity is similar to that of Khanna and Tice (2001), Campello (2002), Lemmon and Lins (2003), and Goyal and Yamada (2004).10

In this analysis, we regress the change in capital expenditures on pre-shock division manager lobbying power, as well as other controls. Our focus is on how lobbying power affects the firm's reaction to the exogenous shock, as opposed to how a change in lobbying power affects capital expenditures. Thus, we use the pre-shock level of division manager lobbying power in our regressions.

The Lobbying Power Hypothesis predicts that managers with greater lobbying power are better able to forestall capital expenditure cuts following negative industry shocks than managers with weaker lobbying power. Predictions concerning positive shocks and lobbying power are less clear. If more powerful managers receive excess capital expenditures before a positive shock while less powerful managers experience capital rationing, it is unclear whether increases in capital flows will be greater for more or less powerful managers following a positive shock. For completeness, however, we examine both types of shocks.

We define negative shocks to industry investment opportunities as having occurred when year-to-year industry capital expenditures fall by 30% or more (i.e., the median percentage change in the ratio of capital expenditures to beginning assets for single segment firms in the industry is less than or equal to negative 30%). Similarly, a year-to-year 30% or more increase in industry capital expenditures defines a positive shock. The 30% threshold is intended to be a conservative demarcation of a shock. Shocks are defined at the three-digit SIC level, requiring at least five single segment firms, in an attempt to balance concerns about the number of single segment firms in an industry and broadness of the industry definition. To obtain a sufficient sample of shock events, we examine shocks occurring during a three year period centered on the year of our initial capital expenditure analysis (i.e., the year a segment appears in our sample of 330 managers) for negative shocks and for a five year period for positive shocks. Division manager information is updated when the shock occurs in a year other than the year in which the segment appears in our original sample. We require that managers of shocked segments be in place at least one fiscal year prior to the year of the shock. Based on these criteria, negative industry shocks for 61 segments and positive industry shocks for 44 segments are identified.

We examine two measures of change in capital expenditures for shocked segments: change in relative capital expenditures (ΔCAPEX) and the simpler change in the ratio of capital expenditures (ΔRatio of Capital Expenditures). CAPEX is parent and industry adjusted, as shown earlier in Equation (3). For ΔRatio of Capital Expenditures, the ratio of capital expenditures equals segment capital expenditures divided by beginning assets. Changes for capital expenditures are computed as values for the shock year minus values for the prior year. We control for change in segment investment opportunities (alternately based on Qest, ROA, and Qind), size, and age. To allow for the possibility that shocks occur after the beginning of the shock year, we compute change in investment opportunities from t− 2 to t, where t is the shock year. Segment size and age are computed as of the beginning of the shock year.

The results for negative shocks, shown in Table 9, support the Lobbying Power Hypothesis. In models (1) through (3), ΔCAPEX serves as the dependent variable. RPI has a positive and statistically significant coefficient estimate, indicating that more powerful managers are better at resisting cuts in capital allocation following negative industry shocks than weaker managers. This is not explained by more powerful managers heading divisions less susceptible to industry shocks. Univariate analysis shows that the mean change in investment opportunities for more and less powerful managers does not statistically differ based on any of the three measures (details not reported).

Table 9. 
Change in segment capital expenditures following a negative industry shock
Two measures of change in segment capital expenditures (ΔCAPEX and ΔRatio of capital expenditures) are regressed on segment manager RPI, a measure of segment manager lobbying power, for a sample of 61 segments operating in an industry experiencing a negative shock to investment opportunities. A negative industry shock is defined as a 30% or greater decline in industry capital expenditures. CAPEX is formulated as in Table 4. ΔCAPEX is the segment's CAPEX for the year of the industry shock minus the same for the prior year. Ratio of capital expenditures equals capital expenditures divided by beginning-of-year assets. ΔRatio of Capital Expenditures is the segment's ratio of capital expenditures for the year of the industry shock minus the segment's ratio of capital expenditures for the prior year. Power Index is formulated as in Table 4. RPI is Relative Power Index and equals the manager's Power Index minus the average Power Index of all the firm's segment managers. ΔRelQest is the change in segment relative estimated q, where Qest and RelQest are formulated as in Table 4. ΔRelROA is change in segment relative return on assets. ΔRelQind is the change in relative median q for the segment's industry. Change in RelQest, RelROA, RelQind, Qest, ROA, and Qind are computed over years t– 2 to t where t is the shock year. LOGTA is the log of segment beginning of shock year assets. RELSIZE is segment beginning of shock year assets divided by the sum of beginning-of-year assets for all of the firm's segments. SEG AGE equals the number of years a segment has operated in a firm, truncated at five years. p-values appear in parentheses below the coefficient estimates and are based on standard errors robust to heteroskedasticity and correlation among segments of a firm.
Dependent variable Investment opp. measure(1)(2)(3)(4)(5)(6)
ΔCAPEXΔRatio of Capital Expenditures
ΔRelQestΔRelROAΔRelQindΔQestΔROAΔQind
  1. ** and * indicate statistical significance at the 0.05 and 0.10 level, respectively.

Intercept−0.0711−0.0774−0.0797−0.0870−0.0826−0.0743
(0.452)(0.403)(0.399)(0.348)(0.318)(0.427)
Investment opportunities, change (t− 2 to t)0.00090.07350.00710.00370.1564**0.0019
(0.833)(0.142)(0.258)(0.541)(0.046)(0.868)
RPI0.0074*0.0076*0.0077*0.0135*0.0143*0.0131*
(0.078)(0.099)(0.062)(0.082)(0.068)(0.074)
LOGTA0.00140.00150.00170.00070.00080.0003
(0.385)(0.317)(0.310)(0.879)(0.845)(0.937)
RELSIZE0.02060.01950.01990.04540.03510.0435
(0.222)(0.271)(0.236)(0.135)(0.207)(0.130)
SEG AGE0.01200.01310.01340.01290.01290.0112
(0.523)(0.474)(0.474)(0.448)(0.416)(0.548)
Change in ratio of industry capital expenditures   0.15440.08330.1635
   (0.356)(0.546)(0.324)
R20.1160.1510.1280.0990.1870.095
Observations616161616161

In models (4) through (6), ΔRatio of Capital Expenditures for the shocked segments is regressed on RPI. The coefficient estimates for RPI are positive, significant, and economically material, again supporting the Lobbying Power Hypothesis. Based on the coefficient estimates, a one-standard-deviation increase in RPI results in a 14.3% increase in the ratio of capital expenditures to beginning assets.

In analogous regressions for the 44 segments operating in industries experiencing positive exogenous shocks, RPI is not significant (results not tabulated). Thus, the dynamics of division manager influence are not detected in the setting of positive shocks. We propose that such an outcome is not surprising due to the ambiguity of lobbying power's effect when there is a positive shock to investment opportunities.

4.6. Link to firm value

If diversity in division manager lobbying power hinders the efficient allocation of capital, then one might expect diversity in division manager lobbying power to be associated with lower firm value. The association between lobbying power and capital expenditures could also lower firm value by reducing managerial effort. Knowing that more resources (and possibly greater recognition) tend to flow to the more powerful manager, a less powerful manager could accept this fate and exert less managerial effort, allowing the more powerful manager to also exert less effort in the competition among managers. Evidence that diversity in lobbying power is associated with lower firm value would also further allay the endogeneity concern.

A more direct test of the proposition that lobbying power diversity hinders the efficient allocation of capital (outside our earlier univariate analysis of CAPEX) would be to regress some measure of allocational efficiency on lobbying power diversity. However, capital allocation efficiency is quite difficult to measure. The most widely cited metric is that of Rajan, Servaes and Zingales (2000) which focuses on over (under) investment in lower (higher) q segments. We do not find a significant relation between the Rajan, Servaes and Zingales (2000) measure and lobbying power diversity, and unfortunately, we know of no good way to directly test whether lobbying power diversity is associated with the full range of allocational inefficiency (i.e., over (under) investment in higher (lower) q segments as well as over (under) investment in lower (higher) q segments). Thus, we rely on an indirect test of lobbying power diversity's impact on capital allocation, and we examine the association between diversity in lobbying power and firm value. To conduct this examination, we regress firm excess value on diversity in division manager lobbying power.

Excess value is computed following Berger and Ofek (1995) as the natural log of observed value minus the natural log of the firm's imputed value. Imputed value is based on the firm's segment sales and the median ratio of value to sales for each segment's industry. Observed firm value equals market value of common equity, plus book values of debt and preferred stock. Diversity in division manager lobbying power equals the standard deviation of the Power Index (as defined previously in Equation (1)).

Controls are included for firm operating performance (EBITD/SALES), firm size (LOGSIZE), and growth opportunities (CAPX/SALES) following Berger and Ofek (1995). In addition, alternative controls are included for forms of diversity other than diversity in division manager lobbying power. The alternative controls for diversity are: number of segments (SEGS), Herfindhal index based on segment assets (HERF), number of two-digit SIC business groups (BIZGRPS), diversity in segment asset weighted investment opportunities (RSZ) based on Rajan, Servaes and Zingales (2000), and diversity in segment investment opportunities (SS) based on the intuition of Scharfstein and Stein (2000). Equations (8) and (9) show the construction of RSZ and SS, respectively.

image(8)
image(9)

All variables are winsorized at the 5th and 95th percentiles for the sample except for LOGSIZE, SEGS, HERF, and BIZGRPS.

Results are presented in Table 10. Of the diversity measures, lobbying power consistently has a negative and significant coefficient estimate, while SS has significance among the other diversity measures. A one-standard-deviation change in division manager lobbying power diversity is predicted to reduce firm value by 11.5%. Results are qualitatively similar if parent book value of assets replaces sales in the construction of EBITD/SALES, LOGSIZE, and CAPEX/SALES. Thus, the results provide evidence that diversity in division manager lobbying power is associated with lower firm value, consistent with the proposition that diversity in lobbying power disrupts the efficient allocation of capital.

Table 10. 
Excess value and diversity in lobbying power
This table reports regressions explaining excess values of the 120 sample firms for the year their segments appear in our sample, where excess value is defined similarly to Berger and Ofek (1995). Lobbying Power Diversity is the standard deviation of Power Index values for the firm's division managers. Power Index=Tenure≥ 5 +Time≥ 3 +Top Exec+Director where Tenure≥ 5 = 1 for managers with tenure ≥ 5 years; Time≥ 3 = 1 for managers with time-in-position ≥ 3 years; Top Exec= 1 for managers who are a top five executive; and Director= 1 for managers on the board of directors. EBITD/SALES is the firm's operating income before depreciation divided by sales. FIRM SIZE is the log of the firm's sales. CAPEX/SALES is the firm's capital expenditures divided by sales. SEGS equals the number of operating segments reported for the firm-year. HERF is a Herfindhal index constructed using segment assets. RSZ equals the standard deviation of segment asset-weighted Qest divided by firm average segment Qest, similar to the Rajan, Servaes and Zingales's (2000) measure of diversity of asset-weighted investment opportunities. SS equals the standard deviation of segment Qest based on the intuition of Scharfstein and Stein (2000). BIZGRPS equals the number of business groups within a firm defined at the two-digit SIC level. Excess value, Lobbying Power Diversity, RSZ, SS, EBITD/SALES, and CAPEX/SALES are winsorized at the 5th and 95th percentiles. p-values appear in parentheses below the coefficient estimates and are based on standard errors robust to heteroskedasticity.
Firm diversification measure(1) SEGS(2) HERF(3) RSZ(4) SS(5) BIZGRPS
  1. ** and * indicate statistical significance at the 0.05 and 0.10 level, respectively.

Intercept−0.7482*−1.0261**−0.8046**−0.9272**−0.7696**
(0.056)(0.040)(0.045)(0.021)(0.027)
Lobbying power diversity−0.2330**−0.2138*−0.2228*−0.2472**−0.2393**
(0.037)(0.060)(0.051)(0.022)(0.027)
EBITD/SALES2.4116**2.3176**2.3332**2.6236***2.4144**
(0.023)(0.027)(0.029)(0.009)(0.020)
FIRM SIZE0.03600.04740.03860.04260.0308
(0.426)(0.298)(0.375)(0.322)(0.497)
CAPEX/SALES0.06650.06720.1193−0.05740.0257
(0.967)(0.965)(0.940)(0.969)(0.987)
Firm diversification0.00430.43430.24510.1493**0.0417
(0.942)(0.331)(0.589)(0.030)(0.585)
R20.1210.1280.1230.1520.123
No. of observations119119119119119

5. Conclusion

Our study adds to the literature on internal capital markets by exploring the impact of divisional manager lobbying power on the allocation of capital. We generally find that more powerful division managers receive greater allocations of capital, where division manager lobbying power is based on a composite measure incorporating division manager tenure, time-in-position, top executive status, and membership on the board of directors. Recognizing the potential for the endogenous determination of our power index and capital expenditures, we spend considerable time using alternative empirical approaches to test the robustness of our primary findings. In general, our results hold true in regressions of segment capital expenditures and in regressions of change in segment capital expenditures following negative exogenous shocks to investment opportunities, while controlling for segment investment opportunities, segment size and age, and parent level capital constraints. Furthermore, we find that diversity in division manager lobbying power is associated with lower firm value. When we use a three-stage least squares approach to address endogeneity, our results are supportive of our primary finding, but are sensitive to model specification. Thus, the overall results support the proposition that division manager influence activities hinder the efficient operation of internal capital markets, as suggested by the works of Harris and Raviv (1996), Rajan, Servaes and Zingales (2000), and Scharfstein and Stein (2000), and that the influence activities lead to lower firm value. In addition, we find evidence supportive of the notion that division manager ownership mitigates these influence activities, consistent with results in Wulf (2002a) concerning compensation contracting. In total, however, it appears that monitoring and control mechanisms within conglomerates do not fully mitigate the influence activities of their division managers.

Footnotes

  • 1

    The efficiency of conglomerate internal capital markets remains a matter of debate. A partial listing of works concerning internal capital market efficiency includes: Lamont (1997); Stein (1997); Shin and Stulz (1998); Hubbard and Palia (1999); Chevalier (2000); Rajan, Servaes and Zingales (2000); Scharfstein and Stein (2000); Whited (2001); Maksimovic and Phillips (2002); Gertner, Powers and Scharfstein (2002); and Billett and Mauer (2003).

  • 2

    See Jensen (1986), Jensen and Murphy (1990), Stulz (1990), Morck, Shleifer and Vishny (1990), and Gibbons and Murphy (1992) for discussions of how firm size may provide private benefits to a CEO.

  • 3

    A total of 19 segments described in Compustat as “other,” or something similar, are reported by the 120 firms in the sample. These “other” segments are excluded because closer examination reveals that corporate items (including capital expenditures) are often included in their segment reporting and because of the difficulty of properly measuring investment opportunities for segments composed of assorted operations (as these “other” segments typically are). In addition, three segments with no identifiable managers are excluded from our analysis since manager characteristics are therefore not available, and a sub-sample of three precludes any meaningful analysis. These three segments are small relative to their firm's other segments, making up from 4.7% to 11.9% of aggregated segment assets.

  • 4

    The time-in-position data are noisy due to changes in segment names and reporting over time, the sometimes vague descriptions of past managerial responsibilities, and the scarcity of public announcements of segment manager appointments.

  • 5

    The analysis in Table 6 for individual manager characteristics is conducted in a similar manner. That is, we use relative measures of each characteristic constructed as in Equation (2).

  • 6

    Industry medians are taken from the most detailed SIC level where the median computation is based on at least five single-segment firms. Before calculating CAPEX, segment values of Ij/Aj are winsorized at the sample's 5th and 95th percentiles.

  • 7

    Tobin's q is computed as a market-to-book ratio following Gertner, Powers and Scharfstein (2002), Burch and Nanda (2003), etc. We compute the market-to-book ratio as (market value of equity + book value of assets − book value of equity)/book value of assets. Chung and Pruitt (1994) show that a similar estimate of q explains over 95% of the variation in the more complicated q as calculated in Lindenberg and Ross (1991).

  • 8

    We define single segment firms as firms whose largest segment makes up 95% or more of the firm's total sales. Single-segment firm values of Tobin's q, ROA, SIZE, and TURNOVER used in the yearly regressions are winsorized by year at the 5% and 95% levels. We require that each regression be based on at least 20 single-segment firms in the segment's two-digit SIC industry. If there are an insufficient number of single segment firms, we use the one-digit level but continue to require at least 20 single-segment firms for valid coefficient estimates.

  • 9

    We gauge the economic significance of RPI and RPI > 0 by focusing on the change in a segment's ratio of capital expenditures to beginning assets as a more intuitive setting than percentage change in CAPEX. We treat the industry and parent adjustments in CAPEX as constants which thus cancel out when computing the change in CAPEX, leaving the change in the ratio of capital expenditures to beginning assets. The change in the ratio of capital expenditures to beginning assets is divided by the sample's mean ratio of capital expenditures. Computing the percentage change in CAPEX evaluated at the mean values of the explanatory variables is not meaningful since the predicted value of CAPEX using these mean values is slightly less than zero (not surprising since CAPEX and investment opportunities are relative values).

  • 10

    Khanna and Tice (2001) and Campello (2002) examine whether the reaction to an exogenous shock is affected by firm-level diversification, as a way to assuage endogeneity issues tied to the diversification decision. Khanna and Tice (2001) consider a shock to investment opportunities (Wal-Mart store openings) and Campello (2002) makes use of a shock to funds availability (monetary policy changes). Lemmon and Lins (2003) examine the association between ownership structure and value around the Asian currency crisis of the late 1990s. Goyal and Yamada (2004) examine the association between cash flow and investment following an exogenous shock to asset prices and monetary policy in Japan.

Ancillary