The Effect of Social Pressures on CEO Compensation 1

Transcription

1 The Effect of Social Pressures on CEO Compensation 1 James Ang Florida State University Tallahassee, Florida jang@cob.fsu.edu Gregory Nagel Middle Tennessee State University Murfreesboro, TN gnagel@mtsu.edu Jun Yang Indiana University Bloomington, Indiana jy4@indiana.edu 1 We thank Nancy Acker, Lucy Ackert, Alex Borisov, Alex Butler, Randall Campbell, Melanie Cao, Michael Faulkender, Gerry Garvey, Eitan Goldman, Paul Grimes, Jeff Fisher, Byoung-Hyoun Hwang, Edwards Lazear, Cassandra Marshall, Ron Masulis, Todd Milbourn, Laura Starks, Irina Stefanescu, Ralph Walkling, David Yermack, Scott Yonker, Julie Zhu, Richard Mahoney (retired CEO from Monsanto Co.), the referee (anonymous), and seminar participants at Erasmus University, Indiana University, Mississippi State University and Washington University in St. Louis, and session participants at the 2009 China International Conference in Finance, the 2008 Financial Management Association meetings, and the FMA European meetings in Prague. We thank Scott Yonker for sharing the CEO home data and the Council for Community and Economic Research for providing us with the data on cost of living index; and Hannah Bolte and Jaden Falcone for editorial help.

2 The Effect of Social Pressures on CEO Compensation Abstract We analyze the effect of social pressures on CEO compensation via interacting with other CEOs, Forbes 400 people, and social elites in the local area; attending industry, alumni, and charitable events; and comparing luxury homes. Each venue is an independent source of social pressures that elevate CEO pay to a level not explained by local economic conditions, firm performance and characteristics, and corporate governance. Social premiums in CEO pay are greater at firms with young, non-ivy League and non-narcissistic CEOs and at firms with directors who are likely to understand and conform to social norms (well governed or locally rooted). Classification Code: G3, J31, J33 Keywords: Corporate governance; CEO compensation; social interactions; reference groups; social pressures.

3 The Effect of Social Pressures on CEO Compensation I think that what Larry Ellison and Bill Gates have is phenomenal wealth, Netscape cofounder Jim Clark once remarked. I'm just a two-bit billionaire. 2 The increases in the level and dispersion of CEO compensation since the early 1990s have attracted much attention from the media, activist shareholders, regulators, and financial economists. Much progress has been made in understanding CEO compensation how CEOs should be paid (the pay for performance relationship) 3 and whether their interests could ever be aligned with those of shareholders. In recent years, entrenched CEOs and lax boards of directors have often been blamed as culprits of the observed pattern of CEO pay. 4 Still, that which might have caused highly-paid CEOs to expect even higher pay remains far from fully understood. To shed light on this heated debate from a new perspective, we investigate the effect of social interactions and pressures from social peers on CEO compensation. In particular, we show that CEO compensation contains an element (referred to as the social premium) that is positively linked to social pressures. The social premium cannot be explained by performance, firm characteristics, CEO characteristics, and governance factors previously shown to affect CEO compensation. In addition, the social premium remains after controlling for the economic condition in the local area. As is well documented in the sociology and economics literature, one s happiness (and thus utility) at least in part depends on the income of one s reference group, after controlling for one s own income. Azar (2007) attributes the origin of this effect to Weber s Law, written in the early 17th century. Arthur Pigou (1920) quotes John Stuart Mill s observation that men do not desire to be rich, but richer than other men. 5 Seidl, Traub, and Morone (2006) document the effect of relative income in experimental studies, and Hagerty (2000) and McBride (2001) do so in empirical studies. Hamermesh (1975) formally models the influence of relative wages on efforts and incentives. 2 Globe and Mail, March 10, See Murphy (1999) for a comprehensive review of the literature on executive compensation. 4 Bebchuk and Fried (2004) exemplify the criticisms of the economic model s ability to explain executive pay. 5 This is also quoted by Graham and Pettinato (2002) and Luttmer (2005). 1

4 Veblen (1934) and Frank (2000) further show that a consumption arms race or conspicuous consumption could occur if one must consume more to keep up with the consumption of one s comparison group. To sustain the high consumption needed for retaining or improving social standing, one needs to receive pay higher than one s peers in the reference group. Social interactions provide the opportunities to collect information on the pay level necessary for achieving high social status. As the number of social peers increases, pressures for greater pay intensify. The resulting continual demand for higher pay is known as the hedonic treadmill hypothesis (Firebaugh and Tach 2005). To examine the effect of social comparisons and social pressures, it is critical to define the reference group, i.e., social peers. Luttmer (2005) documents that one s neighbors are often one s reference group. In a happiness survey conducted on 9,200 households in rural China, Knight, Song, and Ramani (2009) confirm that 70 percent of individuals consider their village as the reference group. In addition, there is evidence that reference groups are often people of similar age and educational background (Melenberg 1992). CEOs, like many other people, interact with their social peers by attending various social and charitable events. Inevitably, CEOs will interact with other CEOs from their industry (Bizjak, Lemmon, and Naveen 2008) or in close proximity. Likewise, CEOs retain their school ties (Cohen, Frazzini, and Malloy 2008; Cohen, Malloy, and Frazzini 2010; Shue 2011) via reunions and private events at exclusive alumni clubs as well as interact with people who serve on the same boards of non-profit organizations. Our primary definition of a CEO s social peers is other CEOs of firms whose corporate headquarters are located within 60 miles (100-kilometer) of the headquarters of the CEO s firm. We use the locations of corporate headquarters to define social circles for several reasons. First, business-related social activities of CEOs often occur in the communities where corporate headquarters are located. Second, the median distance between the corporate headquarters of S&P 500 firms and the main residences of their CEOs is 13.6 miles (Liu and Yermack 2007), and hence most non business-related social activities of CEOs also take place near corporate headquarters. Third, the location of corporate headquarters is determined by factors largely exogenous to current CEO compensation, such as the origin of the founding family, infrastructure, local taxes and costs, and availability of human capital, as well as the proximity to raw materials, suppliers, and customers. It is hard to imagine a board relocating the firm s 2

5 headquarters simply to obtain favorable social circles for the CEO. Thus, using locations of corporate headquarters helps avoid the reverse causality between the choice of social circles and the determination of CEO compensation. Finally, the choice of the 60-mile distance is based on many studies in sociology, economics, and finance. 6 In secondary tests, we expand the social circles of CEOs to include (1) CEOs from the same industry; 7 (2) Forbes 400 people in the local area; (3) prominent alumni of the CEOs; and (4) people serving on the same non-profit boards with the CEOs. We find that each social circle has an effect on the elevation of CEO pay, but there are few interactions between these social circles. Social circles can affect CEO pay through social comparison and social pressures. Social premium is measured by the portion of CEO compensation linked to social circle size but not explained by economic, governance, or location-specific variables previously shown to affect CEO compensation. Economic variables include firm size, market-to-book, stock performance, accounting performance, and firm risk. Governance variables include whether a CEO serves as the chairman of the board, CEO tenure, the percentage of shares held by blockholders, institutions and insiders, respectively; the number of directors, the percentage of inside directors, the Gompers, Ishii, and Metrick (2003) anti-takeover index (GIM), and the number of previous connections between the CEO and directors, which is a proxy for the social dependence of the board (Core, Holthausen, and Larcker 1999; Gomper, Ishii, and Metrick 2003; Hwang and Kim 2009). Location-specific variables include stock returns of local firms in excess of market returns and the cost of living index for professionals in the local area. Our study focuses on the S&P 1500 companies during and yields the following findings. First, CEO compensation contains a social premium. Using the value of exante total pay (the ExecuComp variable TDC1, expressed in 2005 dollars), we find that the 6 Watts (2004) shows the importance of geography in people s social network. Kleinberg (2001) suggests defining a set of geographic groups by centering groups of various geographic radii at each person in the network. Urry (2007) documents that an average American traveled about 30 miles per day in the 2000s. Moreover, the 60-mile distance is practical for attending social events, which occur fairly frequently but not every day. It has also been used in numerous studies in economics and finance such as Kedia and Rajgopal (2009), Malloy (2005), and Coval and Moskowitz (2001). Alternative distances were also used in the literature: Ivkovic and Weisbenner (2007) use 50 miles in defining whether investors are in the neighborhood of the firms in which they invest. 7 Interactions with CEOs from the same industry may not be socially driven because they compete in the product market and labor market, and often benchmark against each other on performance and executive compensation. Industry peers, especially ones of similar size, are definitely relevant for pay comparison (Bizjak, Lemmon and Naveen, 2008). We control for industry peers when showing the social premium related to social peers from the local area. 3

6 average pay for a CEO increases by $550,000 as the number of local CEOs increases from 15 to 79 (moving from the 25 th to 75 th percentile of the sample), all else equal. We show that social premiums exist for various pay measures, and retain after controlling for state fixed effects and a non-linear effect of the cost of living in the local area. Moreover, the social premium result holds if we exclude the largest social circles. Prior literature shows the importance of geography in people s social network (Watts 2004; Kleinberg 2001). Mok, Carrasco, and Wellman (2009) demonstrate that the frequency of face-to-face interactions decreases with geographic distance. As a result, the strength of social pressures and their influence on CEO pay should also decrease with distance. We test the effect of geographic distance on social premiums and find that the strongest effect exists in social circles within 30 miles of the headquarters of the CEO s firm; the effect is weakened by 45% in social circles of 30 to 60 miles away but remains statistically significant at better than 1%. The effect of social circles on CEO pay disappears beyond 60 miles. While the 60-mile distance is manageable for attending social events that occur fairly frequently, the 30-mile distance is more practical for socializing on a regular basis. This evidence could help us address the potential concern that local social peers are picking up the effect of unspecified local variables such as culture, access to local amenities and infrastructure, and proximity to suppliers and customers; those local factors do not change as dramatically as social interactions when the distance increases from 30 to 60 miles. Social pressures for greater pay can only be transformed into pay increases when the board of directors and the CEO agree on social premiums. How one responds to social pressures is likely to depend on one s experience and personal traits. Thus, we investigate how social premiums vary with CEO age, the status of the CEO s alma mater, and the CEO s narcissistic traits. We find that young CEOs and CEOs who did not graduate from prestigious universities 8 are more sensitive to social pressures, likely because they are eager to acquire their social status via high pay in the absence of pedigree. Interestingly, narcissistic CEOs receive higher total pay than other CEOs but significantly lower social premiums. Narcissists believe in their superior ability and thus expect to receive high pay as a proper recognition. At the same time, they shun 8 Prestigious universities include: Ivy Leagues (Brown University, Columbia University, Cornell University, Dartmouth College, Harvard University, Princeton University, University of Pennsylvania, Yale University), Stanford University, and MIT. 4

7 socializing with people deemed inferior and are thus less likely to be exposed to social pressures. 9 Given that social premiums are pay in excess of firm performance, it is useful to identify which boards award social premiums to their CEOs. We examine how firms with different governance characteristics respond to social pressures and the corresponding effect on the magnitude of social premiums. We find mixed evidence using eight conventional corporate governance measures: high social premiums are granted to CEOs at firms with high block ownership, smaller boards, and low GIM index (all of which are proxies for good corporate governance). On the other hand, high social premiums also exist at firms with a high fraction of inside directors (a proxy for poor corporate governance). These seemingly conflicting governance mechanisms could be reconciled because a board with more insiders has strong local roots. More importantly, a board with more connections to the CEO does not grant the CEO a higher social premium. Overall, we show that social premiums are not higher at firms with weak corporate governance. As modeled by Acharya and Volpin (2010), in a labor market with scarce managerial talent, even well-governed firms have to conform to the social norm in compensating the CEOs if competing firms are doing the same. One could argue that the link between CEO pay and the number of social peers may be driven by factors omitted from our empirical specifications. To address this concern, all of our empirical specifications include year fixed effects to capture time trend and industry fixed effects to capture time-invariant and industry-specific characteristics. Further, our results hold when we examine the link between the change in the social circle size and the subsequent change in CEO pay. Moreover, we show that the effect of social pressures on CEO pay exists in various social circles. Our research contributes to the finance literature by applying concepts in sociology, especially the dependence of happiness on income relative to the reference group and the influence of social pressures, to research on executive compensation. Our research complements (1) Kedia and Rajgopal (2009), who document the effect of local companies on the grants of stock options to rank and file workers due to labor market competition; (2) Hwang and Kim (2009), who show that firms with both conventionally and socially independent boards exhibit lower CEO pay, higher pay performance sensitivity, and stronger turnover performance 9 Walter Isaacson discusses the anti-social trait of narcissists in his book Steve Jobs. 5

8 sensitivity; 10 (3) Bizjak, Lemmon, and Naveen (2008), who focus on the influence of industrysize peers on CEO compensation; and (4) Faulkender and Yang (2010) and Bizjak, Lemmon, and Nguyen (2011), who show the selection bias of compensation peer companies and its influence on CEO compensation. Our research differs from concurrent papers by Bouwman (2009), who suggests that CEO pay regresses toward the average pay in the local area and attributes this to envy; Francis, Hasan, John, and Waisman (2008), who examine the effect of geographic locations (urban, small city, and rural) on pay for performance of CEOs and attribute the variation in pay performance sensitivity to monitoring costs and competition in the local labor market; 11 and Knyazeva, Knyazeva, and Masulis (2012), who study the effects of local director labor market on the board structure of nearby firms. Looking into the social lives of CEOs, we examine the effect of social pressures from various groups of social peers on CEO compensation. Social pressures could provide an explanation as to why highly paid CEOs believe they deserve even higher pay and why strong boards endorse it, rather than simply attributing the behavior to greed. The paper proceeds as follows. Section 1 describes the data and develops the empirical strategy. Section 2 presents the results of multivariate regressions. Section 3 analyzes alternative sources of social pressures. Section 4 examines the effect of CEO personal traits and corporate governance on social premiums and Section 5 concludes. 1. Data, preliminary analysis, and empirical strategy In this section, we describe the data, conduct a preliminary analysis, and state the main empirical strategy for the multivariate analysis. Our sample contains the Standard and Poor s (S&P) 1500 companies between 1994 and The S&P 1500 companies are comprised of the S&P 500, S&P Mid Cap 400, and S&P Small Cap 600 companies. We use the historical S&P 1500 indices to identify sample firms Previous studies on the effect of social comparison on executive pay are mainly concerned with the directors network; see, for example, Larcker, Richardson, Seary, and Tuna (2005), Kovacevic (2005), O Reilly, III., Main, and Crystal (1988), Barnea and Guedj (2007), Hwang and Kim (2009), and Engelberg, Gao, and Parsons (2012). 11 Yonker (2011) stresses the geographic preference of CEOs and the difference in performance and pay between local and non-local CEOs. Even though the labor market for rank and file workers is segmented by geographic locations; as shown in Kedia and Rajgopal (2009), we believe the labor market for CEOs is not. 12 Our results are robust to using the S&P 500, S&P Mid Cap 400 and S&P Small Cap 600 index components as defined by Standard & Poor s in

9 1.1 Variable descriptions Our pay determination model includes the following pay determinants identified by existing research: (1) stock and accounting performance; (2) complexity and risks of managerial tasks (size, market-to-book, growth, and risks); (3) corporate governance (whether the CEO serves as the chairman of the board, CEO tenure, the percentage ownership of blockholders, institutions and insiders; the number of directors, the percentage of inside directors, the GIM index, and prior connections between the CEO and directors); (4) local economic environment (returns of local stocks in excess of market returns and the cost of living index at the Metropolitan Statistical Area, MSA, level); and (5) social variables (the number of S&P 1500 CEOs, Forbes 400 people, and social elites in the local area; the luxury home value in the MSA; the number of the CEO s prominent alumni; the number of non-profit organization boards on which the CEO serves; and the CEO s narcissism score). We refer to the variables in groups (1) and (2) as the economic variables. Compensation variables are from the ExecuComp database: the ex-ante total pay (TDC1) includes salary, bonuses, other annual compensation, total value of stock options and restricted stock granted during the year, long-term incentive payout, and other compensation. Share price information is from the University of Chicago s Center for Research in Security Prices (CRSP). Company financial and accounting information is from the Standard and Poor s Compustat database. Historical locations of corporate headquarters are found using historical zip codes, provided by Compact Disclosure. These zip codes are then linked to the latitudes and longitudes at The ACCRA cost of living indexes of each year are provided by the Council for Community and Economic Research ( The sources for governance and director variables include the Investor Responsibility Research Center (IRRC), Corporate Proxy, and Compact Disclosure. The number of the CEO s prominent alumni, non-profit board seats, and prior connections to directors via work, non-profit and education are derived from the BoardEx database. Individuals included in the Forbes 400 list in each year of our sample period are assigned latitude and longitude positions based on the state and city information provided by Forbes; then each Forbes 400 individual is assigned to the 60-mile radius of the corporate headquarters. IRS top wealth holder data are provided by the IRS in 1998 ( Home locations of social elites are found via zip 7

10 codes of all people listed in the Social Register, 2004 Edition. Luxury home values by MSA are provided by a private source that also provided the data to Business Week. 1.2 Preliminary analysis Figure 1 plots the average level of CEO pay (TDC1 in the ExecuComp database) in a social circle against the size of the circle measured by the number of local CEOs. Even though there are fluctuations in the average CEO pay as the size of the social circle increases, the positive correlation between the two is clearly visible. Table 1 provides descriptive statistics for variables used in the analysis. Panel A summarizes compensation variables. The average and median of total annual compensation for our sample CEOs are $5.009 million and $2.649 million, respectively. Panel B lists nine variables regarding CEO social circles. Column 1 describes our primary measure of the social circle size: the number of S&P 1500 firms headquartered with 60 miles of the firm s headquarters (local CEOs). The count includes the firm itself. The average number of local CEOs is 59.9, the median is 45, and the 25 th and 75 th percentiles are 15 and 79, respectively. The largest social circle is in the MSA of New York-Northern New Jersey-Long Island (NY-NJ-PA), which contains 157 of the S&P 1500 firms. 13 There are 262 CEOs who have no peer CEOs within the 60-mile radius (for example, both Bismarck, ND and Tupelo, MS have only one of the S&P 1500 firms). Column 2 shows the year-to-year changes of the number of local CEOs due to adjustments of the S&P index components (including the addition of the Small Cap 600 firms to the S&P index in October, 2004, as well as addition and deletion of firms due to changes in market capitalization and liquidity, mergers, acquisitions, bankruptcies, and privatizations) and relocations of corporate headquarters. 14 Column 3 of Panel B reports the number of Forbes 400 people who live within 60 miles of the firms headquarters. Column 4 describes the number of prominent alumni who attended the same college with the CEO around the same time and are present in the BoardEx database as In our sample of S&P 1500 firms during , only 157 firms moved their corporate headquarters more than 30 miles. Of the 157 firms, 90 firms went through mergers and acquisitions within a year. Out of the remaining 67 firms, 20 appointed new CEOs in the year of the headquarters relocation. 8

11 officers or directors of public firms, private companies or non-profit organizations. Column 5 counts the number of non-profit organizations on which the CEO severs as a director or officer. 15 Column 6 of Panel B reports the number of social elites who live within 60 miles of the company s headquarters. Social elites include those who inherited wealth, top executives and former top executives, of whom very few are current CEOs of S&P 1500 firms; see the 2004 Social Register published by the Social Register Association in New York. Column 7 reports the number of top wealth holders in the state of the firm s headquarters. It is provided by the IRS in 1998 (count of individuals with wealth above $1 million). 16 Column 8 reports the value of luxury homes (the 99 th percentile of home values in the CEO s MSA). 17 This variable, different from the cost of living index, is more relevant for social comparisons especially those with people in the local area and is thus closely related to social premiums in CEO pay. Column 9 of Panel B describes the narcissism score of a CEO based on the number of persons in a photo and the size of the photo included in the annual report (Chatterjee and Hambrick 2007). The narcissism score is 4 if the CEO is the only person in a photo that covers a whole page in the annual report; 3 if the CEO has a solo photo that covers less than a page; 2 if other officers or directors are present in the same photo with the CEO or in other photo(s) on the same page; and 1 if the CEO does not have a photo in the annual report. For our sample CEOs, the median narcissism score is 2.5, in between Michael Dell (score of 2) and Bill Gates (score of 3). Narcissists believe that they deserve high pay as a proper recognition of their superior ability. At the same time, they are somewhat anti-social and are thus less likely to be exposed to social pressures. A few of the nine social variables are highly correlated. 18 Thus in the empirical specifications, we orthogonalize different social variables to capture the incremental effect on CEO compensation of each social venue. 15 For the number of alumni and the number of non-profit boards, we replace missing counts by zero in our regression analysis. Results are similar if we use the subsamples with non-missing values. 16 Our results are similar if we use the number of top wealth holders by state in In our sample, the mean and median of luxury home values in all MSAs are $1.10 million and $0.87 million, respectively; these values are $1.41 million and $1.06 million, respectively, in These values are lower than $2.3 million, the median market value of the main residences for S&P 500 CEOs in late 2006 (See Liu and Yermack, 2007). Considering that our sample includes CEOs of S&P 500, Mid Cap 400, as well as Small Cap 600 firms and that the price in the housing market went up during our sample period, we believe luxury home values in the MSA could serve as a proxy for the values of luxury homes in the local area to which CEOs, social elites, and their spouses pay attention. Moreover, for a subsample of 523 CEOs (Cronqvist, Makhija, and Yonker, 2012) that we have the purchase prices of their homes, the mean and median values of CEO homes are $1.68 million and $1.11 million, respectively. 18 All but one correlation among social variables are lower than 0.6. The correlation between the number of S&P 1500 CEOs and the number of Forbes 400 people in the local area is the exception. 9

12 Panel C of Table 1 summarizes traditional pay determinants such as firm performance, risk, and complexity of business. It also contains economic conditions such as the cost of living index given by MSA and excess returns of local stocks, as measured by the value-weighted return (TRS1YR) of all companies headquartered within 60 miles of the firm s headquarters less the CRSP value-weighted monthly market return (VWRETD). In our regressions on social premiums, we control for these two local variables to show that social premiums in CEO compensation do not merely reflect pay adjustments for different living standards in different areas. Panel D describes eight corporate governance measures previously shown to affect CEO compensation (Core, Holthausen, and Larcker 1999; Bebchuk and Cohen 2005). We also add the total number of connections between the CEO and directors of the company via education, work, or services for non-profit organizations (Engelberg, Gao, and Parsons 2012; Nguyen 2011; Hwang and Kim 2009) Empirical strategy Our multivariate analyses examine the effect of social pressures on social premiums in CEO compensation. The baseline model has two groups of variables: the size of the social circle, a proxy for social pressures; and firm characteristics. Ln (TDC1) = f (Ln(number of local CEOs), market-to-book, (ROA), (stock return), Ln(sales), ROA, lagged ROA, stock return, lagged stock return). This specification is in line with the economic model for executive pay (Core, Holthausen, and Larcker 1999; and Murphy 1999) in which CEOs are compensated for stock and accounting performance, for managing complex operations, and for taking risks and generating growth. We winsorize CEO compensation and the number of local CEOs at the 1 st and 99 th percentiles, then take a log transformation of each to overcome the skewness in the data. Economically, the coefficient estimate of Ln(number of local CEOs) measures the elasticity of CEO pay to social circle size. 19 All missing counts on CEO and director connections are replaced with zero in the regression analyses. Our results do not change if we use the subsample with non-missing values. 10

13 We then add corporate governance variables and local economic variables, as described in Section 1.1, to this baseline model. It is critical to filter out the portion of CEO compensation adjusted for the local living standard before attributing CEO compensation to pressures from local social peers. We include indicators for the Fama-French 49 industry classifications in all regressions. Because the sample contains panel data over 12 years, we cluster standard errors at the firm level (Petersen 2009) and add year dummy variables in all regressions. In the robustness tests, we add state fixed effects to account for the effect of time-invariant and state-specific characteristics that are omitted in the specifications. 2. Empirical Results In this section, we focus on the effect of one social circle: local CEOs. We first show that the positive link between CEO pay and the number of local CEOs depicted in Figure 1 continues to hold in multivariate regression analyses. The number of S&P 1500 companies whose headquarters are located within 60 miles of the headquarters of the CEO s firm measures the size of the social circle and serves as a proxy for social pressures. Table 2 summarizes our main empirical findings: CEO compensation increases with the number of local CEOs after controlling for other pay determinants; that is, we show that the social premium exists. There are four specifications, each of which uses the ex-ante total annual pay (Ln(TDC1)) as the dependent variable and includes an expanded set of explanatory variables. Column 1 reports the results using the number of local CEOs and a set of economic variables as explanatory variables. The coefficient estimate of Ln(number of local CEOs) has the predicted positive sign and is statistically significant at better than 1%. Not surprisingly, CEOs of larger firms and firms with higher risks, higher growth and better performance receive higher pay. Column 2 adds eight corporate governance variables, seven of which are statistically significant at better than 10%. CEO compensation is higher at firms in which the CEO chairs the board, institutional shareholders have higher ownership, the board is larger, and the GIM index is higher. CEO compensation is lower at firms in which block holders and insiders have higher ownership, and the CEO has been at the post longer. These findings are consistent with the existing literature on executive compensation such as Core, Holthausen, and Larcker (1999), Hartzell and Starks (2003), and Bizjak, Lemmon, and Naveen (2008). 11

14 Two local variables added in Column 3 measure local economic conditions that may affect the level of CEO compensation: the cost of living index in the MSA and the excess returns of local stocks. The coefficient of the former is positive and significant at the 5% level. This indicates that CEO compensation is adjusted for local economic conditions. 20 More importantly, Ln(number of local CEOs) is significant at better than 1%, suggesting the social premium goes beyond compensating CEOs for different living standards in different geographic areas. Economically, the average social premium for a CEO in a social circle with 79 CEOs (the 75 th percentile of social circles) is $0.555 million higher than the average social premium for a CEO in a social circle with 15 CEOs (the 25 th percentile of social circles). 21 This pay increase corresponds to 11% and 21%, respectively, of the mean and median of the total annual pay for our sample CEOs. 22 Column 4 adds indicators for firms in the S&P 500 index and firms in the S&P Mid Cap 400 index as well as their interactions with Ln(number of local CEOs). This specification is designed to investigate whether CEOs of large firms are under greater social pressures. We find that social premiums appear to be higher for CEOs at the S&P 500 firms, but the difference is not statistically significant. We then rerun the regression of CEO pay on the number of local CEOs using three alternative pay measures: salary, salary and bonuses, and the ex-post total pay (TDC2, which is same as TDC1 except we replace the value of options granted with the value of options exercised during the year). 23 As shown in Table 3, the social premium exists for all components of CEO compensation and is stronger for equity-based pay. 20 Our results are robust to using housing price indices by MSA, provided by the Office of Federal Housing Enterprise Oversight (OFHEO). 21 The average pay for S&P 1500 CEOs in a social circle with 15 CEOs is $3.908 million. The predicted average pay for S&P 1500 CEOs in a social circle with 79 CEOs is calculated as follows: Ln(pay(79)) Ln(pay(15)) = *(Ln(79) Ln(15)), where the value of is obtained from Column 3 of Table 2. Thus, pay(79) = 3.908*Exp(0.0799*Ln(79/15)) = $4.463 million. This is higher than the average compensation for CEOs in circles with 15 CEOs by = $0.555 million, all else equal. 22 The social premium also exists if we use the number of local CEOs in the previous year as the main explanatory variable. Our interpretation is as follows. A CEO learns from either public sources or face-to-face interactions with other local CEOs about what level of pay is needed to maintain or improve social ranking. In those cases, the desired pay packages are implemented the following year. In many other cases, the CEO could form the pay expectation early on through either communicating directly with local peer CEOs or shared compensation consulting firms. In those cases, CEOs can influence their own pay packages in the contemporaneous year. 23 Even though the board has direct influence over the level of ex-ante total pay (TDC1), the ex-post total pay (TDC2) is highly correlated with the ex-ante one. In addition, CEOs and their spouses may also pay attention to the money pocketed and then consumed by their social peers. Thus, social premiums also exist when CEO compensation is measured ex post. 12

15 Next, we examine how geographic distance between CEOs and their social peers affects the frequency of social interactions and thus the intensity of social pressures. We expect the strength of social pressures to decline as the geographic distance exceeds that for a practical day trip. According to Urry (2007), the average distance of daily travels for Americans is about 30 miles. Therefore, we count respectively the number of peer CEOs in the 30 mile, mile, and mile radius. The results presented in Table 4 show that the impact of social pressures on CEO compensation is the highest for social circles within 30 miles, is much weaker (reduced by 45%) for social circles between 30 and 60 miles, and disappears completely beyond 60 miles. These results are consistent with the premise that CEOs attend social events within a practical distance on a regular basis and thus are under greater influences from peers in these close circles. The decreasing magnitude of social premiums over geographic distance helps us further address the potential issue of omitted local variables such as weather; culture; proximity to suppliers, customers, and prestigious universities; and access to the airport, seaport, and major highways, because those local factors do not change as dramatically when the distance increases from 30 to 60 miles. The social premium associated with local CEOs retains under various additional specifications; see results reported in Table 5. First, social premiums are not merely reflecting the non-linear impact of costs of living, as reported in Column 1; and are robust to defining performance relative to the industry median; see Column 2. In addition, the social premium result survives when state fixed effects are added into the regression; see Column More importantly, the social premium result is not driven by the largest social circles, such as those in New York. When social circles with more than 79 local CEOs (the 75 th percentile of our sample) are excluded from the analysis, social premiums do not change; see Column 4. Column 5 reports the results using the number of S&P 1500 CEOs in a CEO s MSA. Column 6 reports the influence of the average CEO pay in the CEO s local area (self-exclusive) on subsequent CEO pay (Bouwman 2009). 25 In summary, the empirical findings reported in Tables 2 5 are consistent with the hypothesis that social pressures from nearby CEOs affect CEO compensation. 24 In an unreported regression, social premium related to local CEO peers exists in a specification with firm-fixed effects. 25 The correlation between the number of local CEOs and the average of lagged pay for CEOs in the local area (selfexclusive) is When we keep both variables in the same regression, only the number of local CEOs retains its statistical significance. 13

16 3. Other sources of social pressures In the previous section, we establish the link between CEO compensation and the number of local CEOs after controlling for economic, governance, and local factors. In this section, we present other venues of social interactions that may generate pressures for elevating CEO pay. We show that each of those social venues has its own influence on CEO pay, and local CEOs have an influence on CEO pay beyond the influence of those other sources. One natural question regarding the effect of local CEOs is whether locations of corporate headquarters represent industry clustering in location choices. In addition, CEOs compare their pay with other CEOs in the same industry regardless of whether their firms are located in close proximity. Bizjak, Lemmon, and Naveen (2008) show that CEOs whose pay was below the median pay of the industry-size peers in the previous year receive higher pay raises and attribute this to competitive benchmarking in the labor market. Thus, in all empirical specifications, we control for time-invariant and industry-specific characteristics using Fama-French 49 industry fixed effects. Further, we directly test how the change in the number of local CEOs affects the change in CEO pay, controlling for the effect of industry peers. We first use the empirical specification of Bizjak, Lemmon, and Naveen (2008), adding to our regression the change in the number of local CEOs and its interaction with an indicator for lagging CEO pay (pay below the industry-size median level in the prior year). Next, we modify the baseline specification in Table 2 to a change-on-change regression, incorporating the indicator for lagging CEO pay and its interaction with ΔLn(number of local CEOs) and keeping Ln(lagged sales) as specified by Bizjak, Lemmon, and Naveen (2008). The results reported in Table 6 confirm that a CEO whose pay was below industry-size peers receives a higher pay raise as the indicator variable of lagging CEO pay is positive and significant. Moreover, a CEO with more peers in the local area also has greater pay increases. Thus, both industry peers and local CEOs affect CEO pay. Those two sources of pressures do not interact with each other, as indicated by the insignificant loading on the interaction term. Beyond other CEOs of large companies in the local area, a CEO may also socialize with other nearby prominent and wealthy people. One such social circle is the superrich people included in the list of the Forbes 400 who live within 60 miles of the company s headquarters. Some of the Forbes 400 people are themselves CEOs or former CEOs. We first examine the 14

17 individual effect of each social circle (local CEOs vs. local Forbes 400 people), then orthogonalize one social circle to the other; finally, we interact the two circles by examining whether a CEO with more surrounding Forbes 400 people faces more intense the pressure from local CEOs. The results reported in Table 7 show that the number of local Forbes 400 people affects CEO pay; see Column 2. The influence on CEO pay of local CEOs dominates that of nearby Forbes 400 people, as indicated by the insignificant loading of the residual of the number of Forbes 400 people on the number of CEOs in the local area. 26 In addition to socializing with prominent local people and comparing pay with CEOs from the same industry, a CEO may also socialize with his college classmates via activities at local alumni associations, reunions, and private events at exclusive alumni clubs such as Harvard Club of New York City. A recent study by Shue (2011) shows the influence of MBA classmates at Harvard Business School on executive compensation, acquisition propensities, and other corporate decisions. Interestingly, the peer effects are more than twice as strong in the year immediately following staggered alumni reunions. To investigate the effect of prominent alumni, we count how many people who went to the same college as the CEO around the same time (with overlapping years) are officers or directors of private companies, public firms, or nonprofit organizations. The more prominent classmates a CEO has the stronger the social pressures the CEO faces because classmates are typically considered to be one s equals. Results reported in Table 8 confirm this hypothesis. We find that the alumni network and local CEO network are two separate social circles: each affects CEO pay, but they do not interact with each other, as indicated by the insignificant loading on the cross term in Column 5. The conclusion is the same using the maximum or average wealth level of the CEO s prominent classmates, but the sample size is dramatically reduced due to the lack of wealth information. CEOs that choose to serve on multiple non-profit organizations might be more concerned about their social status. Non-profit organizations also provide those CEOs venues to display wealth or compare wealth with a different set of wealthy people via offering their own homes or being invited to other homes for charity functions, and competing for largest charitable 26 The average pay for S&P 1500 CEOs in a social circle with three nearby Forbes 400 individuals (the 25th percentile) is $4.284 million. The predicted average pay for S&P 1500 CEOs in a social circle with 25 nearby Forbes 400 individuals (the 75th percentile) is calculated as follows: Ln(pay(25)) Ln(pay(3)) = *(Ln(25) Ln(3)), where the value of is obtained from Column 2 of Table 7. Thus, pay(25) is predicted to be 4.284*Exp(0.0734*Ln(25/3)) = $5.005 million. This is higher than the average compensation for CEOs in circles with three nearby Forbes 400 individuals by = $0.721 million, all else equal. 15

18 contributions or highest bids on auctions. Thus, CEOs serving for more non-profit organizations have more exposures to and likely face stronger social pressures. To investigate, we count the number of non-profit organizations at which a CEO serves as an officer or director, and analyze its effect on CEO pay in a way similar to that of prominent alumni. The results reported in Columns 2 and 4 of Table 9 show that a CEO who serves more non-profit organizations indeed receives higher pay. In addition, the pressure from local CEOs elevates CEO pay beyond that from serving on non-profit boards; see Column 3. When we keep both the number of local CEOs and the indicator for more non-profit board seats (above sample median) in the same regression, only the number of local CEOs retains its significance; see Column 5. The coefficient on the cross term is insignificant, indicating that a CEO with more prominent alumni is not more sensitive to social pressures from local CEO peers. Table 10 reports the effects of four additional social channels through which social pressures may affect CEO compensation. We show that the social premium associated with the number of local CEOs remains after controlling for those alternative sources of social pressures. First, we examine the effect of the number of local social elites 27 and the number of top wealth holders in the state on social premiums in CEO pay. According to the Conspicuous Consumption Theory of Veblen (1934) and Frank (2000), wealthy people often use luxury homes to display wealth and signal social status to their peers. Next, we examine the effect of the value of luxury homes (the 99 th percentile value of homes sold in the MSA) and the value of CEO homes (for a subsample of 523 executives who were CEOs in 2004) on social premiums. We orthogonalize the number of local CEOs to each of those four alternative sources of social pressures and find that each source has an influence on CEO pay, while the pressures from local CEOs elevate CEO pay beyond these alternative sources. 4. CEO personal traits, corporate governance, and social pressures Given that social premiums are the amount of CEO compensation in excess of that which can be explained by firm performance, risk, governance, and local economic conditions, it is useful to examine the demand and supply: which CEOs desire to receive and which boards are 27 Very few social elites are current CEOs. The average and median numbers of social elites in the local area are 1,356 and 272, respectively. 16

19 willing to grant such social premiums. More specifically, in this section we examine how CEO personal traits and corporate governance affect the magnitude of social premiums. First, we examine the effect of CEO personal traits on their sensitivity and response to social pressures. In particular, we look at whether a CEO is old (older than 55, the sample median), whether the CEO graduated from a prestigious university, and whether the CEO is a narcissist (Chatterjee and Hambrick 2007). As shown in Table 11, young CEOs and CEOs who did not graduate from prestigious universities receive lower compensation in general but higher social premiums in their compensation as indicated by the negative coefficient on the interaction term. These CEOs seem eager to establish their social status with high income, especially those without pedigree. Interestingly, narcissistic CEOs (with a narcissism score greater than 2.5, the sample median), driven by a superior self-image, receive much higher total pay. However, because narcissistic CEOs shun social interactions with people deemed inferior, which include most other CEOs, their sensitivity to social pressures is less than a half of that of otherwise similar but nonnarcissistic CEOs. This evidence helps differentiate the social pressure theory from an alternative one: more connected CEOs receive higher pay for their greater values to the firm (beyond what is reflected in the past and current performance) because, the alternative theory does not predict lower values for narcissistic CEOs. Further, we examine which boards are more sympathetic to social pressures. We first look at the effect of eight conventional corporate governance measures on social premiums, first jointly, then individually. The results are reported in Table 12. Social premiums are higher at firms with a higher block ownership, a smaller board, and a lower GIM index; each of which indicates strong corporate governance. On the other hand, social premiums are also higher at firms with a higher fraction of inside directors, which can be interpreted as either as an outcome of weak corporate governance or a conformation to social norms by boards with deeper local roots. In addition to these conventional corporate governance measures, we test the effect of prior connections between the CEO and directors via work, non-profit, and education experiences. This measure captures the social dependence of the board (Hwang and Kim 2009; Engelberg, Gao, and Parsons 2012; Nguyen 2011). The results reported in Table 13 suggest that CEOs who are connected to directors are paid more generously, consistent with the findings in 17

20 the prior literature. However, their social premiums are indistinguishable from those of unconnected CEOs. In general, the results reported in Tables 12 and 13 do not support the conjecture that social premiums are only granted at firms with weak corporate governance. We believe that in a labor market with scarce managerial talent, even a strong board has to grant social premiums if competing firms are doing so (Acharya and Volpin, 2010). 4. Conclusion Let me tell you about the very rich. They are different from you and me. 28 It is often attributed to greed for highly paid CEOs to demand even higher pay. This study helps us understand the phenomenon from a perspective of social interactions and pressures from social peers. Our approach explicitly recognizes that one s well-being, by definition, must be measured in the context of one s social setting. CEOs socialize with other CEOs and social elites in their community and compare wealth through published sources as well as visible displays of wealth. CEOs also include their industry peers and school ties in their reference groups. They are aware of their social rankings, which depend, at least partially, on their consumption. Thus, they are propelled to demand greater pay to secure or improve their rankings, anticipating other CEOs will do the same. The evidence presented in this paper suggests that the compensation for otherwise identical CEOs varies from one location to another, not simply due to difference in living expenses, but due to the wealth level needed to maintain the CEO s status in the respective social circle. If there are compelling reasons for a company s headquarters to be located in a certain area, the board of directors would be compelled to follow the social norm and grant the social premium to its highly valuable CEO. However, the collective actions of the boards will inevitably raise CEO pay year after year and accelerate the pace of the hedonic treadmill. 28 By F. Scott Fitzgerald in the short story Rich Boy in All the Sad Young Men. 18