Yesterday’s Heroes: Compensation and Creative Risk-Taking.

Haw Cheng (University of Michigan), Harrison Hong (Princeton University), Jose Scheinkman (Princeton University)

Abstract: We investigate the link between compensation and risk-taking among finance firms during the period of 1992-2008. First, there are substantial cross-firm differences in total executive compensation residualized for firm size. Second, residual pay is correlated with price-based risk-taking measures including firm beta, return volatility, tail cumulative return performance, and the sensitivity of firm stock price to the ABX subprime index. Third, residual compensation is also weakly correlated with balance-sheet based risk-taking measures such as holdings of non-GSE mortgage-backed securities and book leverage. Fourth, these risk-taking measures are correlated with residual pay even though executives are highly incentivized as measured by insider ownership. Finally, compensation and risk-taking are not related to governance variables but covary with ownership by institutional investors who tend to have short-termist preferences and the power to influence firm management policies. Our findings suggest that our residual pay measure is picking up other important high-powered incentives not captured by insider ownership. They also point to substantial heterogeneity in both firm culture and investor preferences for short-termism and risk-taking.

I. Introduction

Are Wall Street bonuses to blame for the most significant economic crisis since the Great Depression? The media and more importantly Treasury Secretary Timothy Geithner seem to think so. In his testimony (June 6, 2009) in front of Congress on the Treasury budget, Secretary Geithner argues, _I think that although many things caused this crisis, what happened to compensation and the incentives in creative risk taking did contribute in some institutions to the vulnerability that we saw in this financial crisis. We need to help encourage substantial reforms in compensation structures particularly in the financial industry_ (emphasis added).1 To address this issue, the Obama administration is implementing two key reforms. The first reform is the appointment of a _compensation czar_ who has been given the directive of not rewarding employees of finance companies receiving TARP funding for short-term or temporary increases in value. The second reform, the Corporate and Financial Institution Compensation Fairness Act currently before the US legislature, increases the say of shareholders in approving compensation and electing directors on compensation committees. Implicit in the second reform is that finance firms’ short-termist incentives reflect a misalignment with shareholder interest.

It is hoped that these reforms will curb what the Treasury Secretary refers to as _creative risk-taking_ by _some_ finance firms over the past five years. This risk-taking is perhaps best epitomized by the now infamous _musical chairs_ quote of Chuck Prince, then CEO of Citigroup, regarding their exposures to the subprime mortgage market. In his interview with the Financial Times back in July 2007, Chuck Prince, in referring to his company not backing away from risks at the beginning of the subprime crisis, remarked: _When the music stops, in terms of liquidity, things will be complicated. But as long as the music is playing, you've got to get up and dance. We're still dancing._ This quote is often attributed as competitive market pressure (presumably being fired by impatient shareholders) forcing Citi’s managers to take on such risks, whether or not they fully understood them.
The connection between compensation and risk-taking has basis in theory as there has been substantial theoretical work on the effects of short-termism in financial markets (see Stein 1989 and Stein 2003 for a review of this large literature). One set of sufficient ingredients are pay for short-term price performance (such as bonuses and equity option grants with short vesting periods) and performance measures which do not adequately capture tail risk (such as Jensen’s alpha or the Sharpe ratio). This would incentivize firms to follow strategies that effectively sell disaster insurance. Such strategies have a long history going back to Long-Term Capital Asset Management and more recently the AAA tranches of the collateralized default obligations associated with mortgages (see Coval, Jurek and Stafford 2009). However, theory does not necessarily imply that short-termism is a result of misalignment with shareholder interest. It may very well be that a certain clientele of short-termist investors want some sophisticated finance firms to take such risks and they therefore need to give these firms appropriate incentives to get them to do so (see, e.g., Froot, Perold and Stein (1992) on the critical role of institutional investors and Bolton, Scheinkman and Xiong 2006 for a broader theoretic take).
In this paper, we investigate the empirical link between compensation, short-termism and risk-taking. First, drawing inspiration from the Obama administration’s compensation reform proposal, we ask whether compensation practices could have predicted differences in outcomes among firms during the current financial crisis (i.e. separated Bear Stearns, Lehman, Citigroup, AIG from JP Morgan, Wells Fargo and Goldman Sachs). In other words, is there heterogeneity in firm compensation practices, and is it correlated with differences in risk-taking? Second, we ask, is any such correlation between compensation and risk-taking (if present) due to misalignment of interest between management and shareholders?
We use panel data on financial firm executive compensation and risk-taking from 1992-2008 to answer these two questions. Our measure of short-termism is the residual of total annual firm compensation (payouts) controlling for firm size and finance sub-industry classifications. We ask whether cross-sectional variation in our residual compensation measure is related to heterogeneity in subsequent risk-taking. Strong causality and normative statements will necessarily be limited given the nature of our exercise. Nonetheless, we view this analysis as a contribution for two reasons. First, as we discuss in the literature review below, we know very
little about how financial firms take risks, particularly for sophisticated trading firms such as Bear Stearns and Lehman Brothers. Second, given the topical nature of these questions and the importance of these reforms, we feel that there has been a lack of real empirical data in public dialogue and as such, any robust facts are valuable.
We compute our residual compensation measure as follows. We first average total compensation (including bonus, salary, equity and option grants, and other direct annual compensation) across the top five most highly paid executives at each firm. We aggregate across all forms of direct compensation because it is a less noisy measure of short-term pay practices than looking at particular components. Indeed, some authors such as Michael Jensen argue that option grants are just a cost-efficient way to pay bonuses and a large literature (Murphy 2000, Hall and Murphy 2003) convincingly shows that both bonus and option grants motivate short-termist behavior. Then we regress (cross-sectionally) total compensation on two control variables. The first is firm size since it is well known that the best personnel work for the biggest firms (Gabaix and Landier 2008, Murphy 1999). The second is heterogeneity in sub-industry classifications among financial firms (which we break into three categories: primary dealers, banks, lenders and bank-holding companies, and insurance companies) since primary dealers and banks may have different compensation practices than insurance companies.
Our measure differs from what is typically used in the literature. For instance, a more traditional measure of incentives is insider ownership. Indeed, recent work (notably Fahlenbrach and Stulz 2009) finds that insider ownership does not have much predictive power for risk-taking and that executives of finance firms tend to have high values of ownership stakes to begin with. One might therefore be tempted to conclude that there is no relationship between compensation and risk-taking.
However, this is premature since our residual pay measure can pick up other important incentives better than traditional measures. First, top executives, even if they have high ownership stakes, face other high-powered incentives related to market pressure from short-termist investors to out-perform rivals. The above quote from Chuck Prince and the recent firing of John Mack of Morgan Stanley after the collapse of Lehman (both of whom were well-incentivized and both facing pressure from impatient shareholders) are consistent with this perspective. In other words, implicit incentives related to firing also matter greatly. Second, many rank-and-file employees that matter for risk-taking (such as risk managers or proprietary traders) do not typically have high ownership stakes and hence our measure might better pick up the incentives of these employees. We would ideally like compensation data for a wide range of employees at each firm, but ExecuComp (our data source for compensation) typically only provides data for the top five executives. Nonetheless, higher annual payouts at the top level might pick up a firm culture for high powered incentives, whether they be bonuses or higher sensitivity of firing to short-term performance. As such, we view our residual pay measure as being a sensible proxy of both firm-wide explicit and implicit short-termist incentives.
One issue with our measure is that the level of total compensation could also reflect long-term incentives. To deal with this, we utilize insider ownership as a proxy for long-term compensation (similar to the literature) on the idea that stable insider holdings is a reflection that management is invested for the long-run and that insider holdings reflect long-term incentives. While we begin by focusing on pay residual firm size, we will look at the extent to which pay controlling for both firm size and insider ownership is correlated with risk-taking in an extended analysis. In other words, insider ownership, rather than being the object of interest as in other studies, is a control variable to help us sort out alternative explanations for our residual pay measure.
Our empirical design is as follows. We split our sample into two periods—an early, non-crisis period defined as 1992 (when we start having executive compensation data) up to 2000, which marks the end of the dot-com era, and a late, crisis period from 2001-2008 which marks the beginning and end of the housing boom. We then take the first three years 1992-1994 to create a ranking of executive compensation among firms at the beginning of the non-crisis period. Specifically, we take the log of average executive compensation from 1992-1994 and regress this on the log of a firm’s market capitalization in 1994, allowing for heterogeneity at the sub-industry level, to come up with a residual compensation ranking for each firm. We then take data from 1998-2000 to create a similar ranking for residual compensation before the crisis period.
Then, using data from 1995-2000 and 2001-2008, we calculate various risk-taking measures for the two sub-periods of non-crisis and crisis, respectively. The first set consists of price-based measures including firm beta, return volatility, and tail cumulative return performance. For the crisis period, we also compute the sensitivity of a firm’s stock price to the ABX subprime index. The second set consists of accounting-based measures including the average holdings of mortgage-backed securities not backed by one of the government-sponsored entities (GSEs) and book leverage. Our baseline analysis is to regress these risk-taking measures on our lagged residual CEO compensation (from 1992-1994) measure along with other firm characteristics. Similarly, we calculate risk-taking measures for the period of 2001-2008 and regress these on our residual compensation measures constructed from 1998-2000.
We work with this stark set-up rather than a full panel estimation for a few reasons. As we will show, residual pay in our two cross-sections is highly correlated, so we are essentially capturing permanent effects. This set-up makes it clear that residual pay in our cross-sections is very similar and allows for a simple and conservative framework to measure our effects. We will also work with a pooled panel set-up and cluster standard errors by firm in the robustness section. In addition, this set-up best captures cumulative returns over long horizons, which really gets at the idea behind the title of the paper. From 1995-2000, the market did very well and the risk-takers should have had good outlier performances, but during the period of 2001-2008, a poor time in market, the risk-takers should have had poor outlier performances.
We establish the following findings. First, there is substantial cross-sectional heterogeneity in the permanent component of residual executive compensation. More specifically, there is a strong and statistically significant linear relationship between firm size and compensation in both sub-samples. The residual compensation measures obtained from this regression are highly correlated across the two sub-samples. The correlation between these two residual pay lists is 0.61 with a p-value of zero. Firms with persistently high residual compensation include Bear Stearns, Lehman, Citicorp, Countrywide, and AIG. Low or moderate residual compensation firms include JP Morgan, Goldman Sachs, Wells Fargo, and Berkshire Hathaway. We then examine whether CEO turnover and stock price performance drive changes in the residual compensation measures across the two sub-periods. The idea is that if these variables do not drive changes in residual compensation then it is suggestive of something more fundamental about the culture or technology of the firm. We find that CEO turnover and past stock price performance have very little explanatory power. Although good past price performance leads a firm to have slightly higher residual compensation, the strongest predictor (both statistically and economically) of residual compensation in 1998-2000 is residual compensation in 1992-1994. As such, we interpret heterogeneity of our residual compensation measure as being due to permanent cross-firm differences.
Second, we find that our residual compensation measure is strongly correlated in both sub-samples with our price-based measures of subsequent risk-taking. The focus of our analysis is the use of price-based measures of risk-taking rather than traditional measures such as book leverage. After all, we know that many of the finance firms’ exposures during the recent crisis were off balance sheet. Accordingly, we find a very strong statistical and economic relationship between our price-based measures of risk and compensation. Firms with high executive compensation have a higher CAPM beta, higher return volatility, and are more likely to be in the tails of performance, with extremely good performance pre-crisis when the market did well and extremely poor performance during the crisis period when the market did poorly. For example, a one-standard deviation increase in residual compensation in 1998-2000 is associated with 21% lower returns over the market in the 2001-2008 period.3 Our results for exposure to the ABX subprime index are also interesting – a one-standard deviation increase in residual compensation is associated with a 0.37 standard deviation-increase in subsequent stock price exposure to movements in the subprime market, as measured by price movements in the ABX.
Third, for completeness, we also consider accounting-based measures of risk-taking. In addition to standard book leverage, we also use Federal Reserve data and compute average holdings of mortgage-backed securities not backed by one of the government-sponsored entities (GSEs),5 as a fraction of book value for banks and bank-holding companies. Since the MBS and book leverage measures are based on reported balance sheet values, we expect to find weaker results than our price-based risk measures. We find positive associations between residual compensation and holdings of non-GSE MBS holdings and book leverage. Compensation and risk-taking as measured by non-GSE holdings are more correlated in the crisis period than in the non-crisis period, reinforcing our ABX result that firms with high residual compensation are firms who took the largest bets on the subprime market. These findings are economically interesting and consistent with our expectation that the statistical significance in comparison to our price-based risk measures is more fragile and dependent on specification.
These three sets of findings suggest that there is substantial heterogeneity in financial firms in which high-compensation, high risk-taking and tail performance go hand in hand. As a result, the aggressive firms that were yesterday’s heroes when the stock market did well can easily be today’s outcasts when fortunes reverse, very much to the point of what we have experienced in the last twenty or so years. The important thing to note here is that our findings are qualitatively similar across different sub-industries. So our results are not due to just the primary dealers, though the results for primary dealers are stronger as one would expect since they have the most discretion to take risks. Moreover, the correlation between the risk-taking measures and residual compensation is primarily a compositional effect in that changes in the risk-taking measures are uncorrelated with changes in the residual compensation measure. Additionally, we examine components of pay and find that both bonuses and equity/option compensation are correlated with risk-taking (while salary is markedly less informative).
We also perform a series of robustness checks of these three sets of findings. We re-do our analysis by calculating residual compensation using book asset values rather than market values on the idea that asset values are more exogenous. We run our price-based risk-taking measures on residual compensation while controlling for both firm size and book leverage to show that our results are not just due to book leverage. This is essentially the same as working with delevered equity (or asset) betas as opposed to equity betas in our analysis. So our results are not simply mechanically driven by levered equity as measured by book leverage. However, we put this as a robustness analysis since one premise of our paper is that book leverage is also a measure of risk and might potentially be driven by residual compensation. Next, we exclude the CEO’s pay when computing our residual compensation measure and find nearly identical results. We do the same exercises for manufacturing industries as an out-of-sample check and find qualitatively similar though weaker effects. Finally, we run a pooled regression version of our set-up and find similar results.
Fourth, we go on to ask whether our compensation measure and results are picking up short-termism as opposed to unobserved heterogeneity. An alternative interpretation is that high residual compensation firms are like the growth firms of finance, where risk-taking requires talent and talent gets rewarded with higher compensation in general (both short-term and long-term) compared to low residual compensation and presumably low growth firms. We first repeat our above empirical exercises but now control for the market-to-book ratio on the premise that it is a proxy for whether a finance firm is a growth firm and find consistent results.
We then examine the hypothesis of short-term compensation directly by regressing risk-taking of firms on compensation while controling for insider ownership on the presumption that insider ownership is a proxy for long-term incentives. If indeed compensation is capturing long-term pay incentives (as opposed to short-term pay), then having insider ownership should mute our results and we should also expect insider ownership to predict risk-taking with the same sign as compensation. Instead, our baseline findings on compensation remain even after controlling for insider ownership, and we also find some evidence that insider ownership is negatively correlated with risk-taking. As such, we conclude that our risk-taking results are being driven by short-termism as opposed to a proxy for long-term incentives.
Finally, we ask whether our results are due to mis-governance or entrenchment as opposed to heterogeneity among investors who want to invest in high risk-taking firms and hence need to set compensation appropriately to induce such behavior. We find that standard governance measures such as the Gompers, Ishii and Metrick (2003) and Bebchuk, Cohen and Ferrell (2009) measures of entrenchment, as well as board independence, are not correlated with our results (if anything, the worst governance score firms are associated with less risk-taking). So it appears that there is no evidence using these standard metrics for mis-alignment of interest between shareholders and management, at least in the cross-section. But this may simply be that these measures are not very good measures of governance in finance.
In contrast, we find that residual compensation and risk-taking are positively correlated with institutional ownership. One interpretation, motivated by Froot, Perold and Stein (1992), is that there is heterogeneity in investor preferences with institutional investors (who tend to trade more and perhaps with shorter-horizons because of agency issues) wanting certain firms to take more risks and hence having to give them short-term incentives to do so. Indeed, both anecdotal and empirical evidence suggests that institutional investors are the ones with the power to pressure management (Graham, Harvey and Rajgopal 2005, Parrino, Sias and Starks 2003). In this interpretation, the high-powered incentives picked up by our residual pay measure is simply the carrot needed to get the firm to take risks desired by institutional investors. Of course, one has to be a bit careful in interpretations here since if institutional investors are too short-termist and say always flip the shares of the company, they will not have any influence over management. But in practice, there is plentiful evidence that institutional investors care greatly about companies making quarterly earnings targets, presumably because the accompanying growth in share prices helps the institutional investors’ portfolio performance.
Indeed, while we have focused on total direct compensation, which is easier to measure than firing pressure, it is likely that firing for failure to meet quarterly targets (while more difficult to measure) is a more powerful motivator.6 These two types of high-powered incentives are likely to be correlated across firms and may explain why short-term pay predicts risk-taking even though very rich executives had such large stakes in their companies. In point, the competitive pressure that Chuck Prince suggests in his musical chairs quote is likely due to firing

B. Variables

The construction of our variables is as follows. Our baseline measure of executive compensation is total direct compensation TDC1 from ExecuComp (Salary + Bonus + Value of Option Grants + Other Annual Compensation + Restricted Stock Grants + Long-term Incentive Payouts + All Other Compensation), averaged across the top five executives at the firm. Specifically, we measure top 5 executive compensation as the average compensation of the top 5 most highly paid executives (by TDC1), always including the CEO and CFO when available.8 For firm variables that overlap between CRSP and COMPUSTAT, we take the CRSP value. We compute Market Capitalization in a year as shares outstanding (SHROUT) times price (PRC) on December 31 of that year. The market-to-book ratio is Market Capitalization divided by book assets (ATQx1000).
We compute six measures of risk-taking and stock-price performance: 1) the beta of the firm’s stock, 2) the firm’s stock return volatility, 3) the cumulative return to the firm’s stock, 4) the correlation of a firm’s daily stock returns with returns to the ABX AAA index (ABX Exposure), 5) a firm’s balance sheet holdings of non-agency mortgage backed securities (MBS Exposure), and 6) book leverage. We follow Adrian and Shin (2009) who analyze the leverage characteristics of investment banks by computing leverage as the ratio of book assets (ATQ) to book equity (SEQQ).
We compute a firm’s Market Beta and Return Volatility for a given period (1995-2000 in the early period or 2001-2008 in the late period) using the CRSP Daily Returns File, and take our market return to be the CRSP Value-Weighted Index return (including dividends). Our data on
the risk-free return comes from Ken French’s website. In computing betas and volatility, we require at least one year’s worth of observations (252 trading days) in that period. We compute each firm’s cumulative compounded return in a given period and subtract it from the cumulative compounded return of the market to obtain each firm’s Cumulative Excess Return for that period. We follow Shumway (1997) in our treatment of delisting returns.
We use the on-the-run ABX daily price index obtained from Barclays Capital Live9 to compute a firm’s ABX Exposure. Following Longstaff (2008), we compute the ABX return as the log of the time-t price divided by the time t-1 price, where we ignore the coupon rates of each tranche (i.e. like Longstaff, we are assuming a coupon yield of zero). We compute a firm’s exposure to the AAA tranche by regressing returns obtained from the CRSP Daily Returns File on returns to the ABX AAA and returns to the market (defined as the CRSP Value-Weighted Index return, including dividends) for each firm from 2006 (when the ABX was created) through the end of 2008. We take the coefficient on ABX returns as the firm’s exposure to the ABX.
We obtain data on exposure to mortgage-backed securities (MBS) from the consolidated financial statements of bank holding companies (Form FR Y-9C), available electronically from the Federal Reserve Bank of Chicago. We define MBS exposure as total holdings of mortgage-backed securities not issued or guaranteed by government-sponsored entities (FNMA, GNMA and FHLMC), divided by total balance sheet size (BHCK2170). We include both pass-through securities (BHCK1710+BHCK1713) and non-pass-through securities (BHCK1734+BHCK1736) such as collateralized mortgage obligations (CMOs) and real-estate mortgage investment conduits (REMICs), and include holdings on the trading-side of the balance sheet (BHCK3536 on Schedule HC-D) as well as the securities balance sheet (aforementioned variables, on Schedule HC-B). We focus on non-GSE guaranteed mortgage-backed securities in order to focus attention on the riskiest securities such as subprime.
Our baseline computations relate total compensation to risk-taking. In extended results, we will also utilize insider ownership, which we measure as the number of shares plus the delta - weighted number of options owned by the top five executives divided by shares outstanding, as a noisy proxy for long-term compensation. We compute the delta-weights on the options using the Core and Guay (1999) methodology.
We also relate these measures of risk-taking and stock price performance to measures of governance. We obtain from RiskMetrics data on corporate governance including the G index (Gompers, Ishii and Metrick 2003), percentage of directors that are outsiders (classified as _Independent_ by RiskMetrics), and the board size. Since the RiskMetrics data on directors goes back to 1997, we have data on board size and independence in our _late_ period. To match the RiskMetrics data with CRSP, we use a historical ticker merge, as well as a _fuzzy_ merge based on SIC code, CUSIP, and company name utilizing the STATA routine _reclink._ We hand check the merged list and remove any discrepancies. We obtain data on the Entrenchment Index (Bebchuk, Cohen and Ferrell 2009) from Lucian Bebchuk’s website. For our measure of speculative activity, we use monthly stock turnover data from CRSP and compute the average 36-month stock turnover (VOL*100 / SHROUT*1000) for each period.
We obtain data on institutional ownership from the Thomson Reuters S13 database, which captures 13F filings by financial institutions electronically. We match 8-digit CUSIPs in Thomson to PERMNOs in CRSP , noting that the CUSIPs in Thomson are provided for the filing date (not the reporting date). For each PERMNO, we divide the shares held by each financial institution (SHARES) by the shares outstanding (as reported by Thomson in SHROUT1 before 1999 and SHROUT2 after 1999) and sum up over each stock. We take care to ensure that
holdings and shares outstanding both reflect stock splits when necessary. We censor the percentage of shares held by institutions at 1 for a few observations.
Lastly, we winsorize all variables except for our compensation variables and Market Capitalization at their 1% and 99% values. We do not winsorize the G Index, E Index, board size or the percentage of directors that are outsiders, since these are based on well-behaved count-data.

IV. Results

Our goal is to relate differences in risk-taking across finance firms to cross-sectional heterogeneity in their compensation. To this end, we split our sample into two periods—an early, non-crisis period defined as 1992 (when we start having reasonable executive compensation data) up to 2000, which marks the end of the dot-com era and a late, crisis period from 2001-2008 which marks the beginning and end of the housing boom. We then take 1992-1994 (1998-2000) to create a ranking of executive compensation among firms at the beginning of the non-crisis (crisis) period.13 As we mentioned earlier, in our comparison of firm compensation practices, it is important to control for two things. The first is firm size since it is well known that better personnel work for bigger firms (Gabaix and Landier 2008, Murphy 1999). The second is heterogeneity in sub-industry classifications among financial firms (described above). In other words, we work with a residual compensation measure in which we take the residual from a cross-sectional regression of compensation on firm size and sub-industry classifications. Ideally, we would like to control for heterogeneity by allowing both slopes and intercepts to vary across sub-industries. Unfortunately, the limited number of primary dealers per year does not allow us to form reliable estimates of the slope and intercept within that group. Instead, we take the log of average executive compensation in 1992-1994 (1998-2000 for the crisis-period) and regress it on the log of firms’ market capitalization in 1994 (2000 for the crisis-period), allowing intercepts to vary by sub-industry and allowing the insurers group to have a slope distinct from banks and primary dealers. This specification allows for heterogeneity in the levels of pay across sub-industries and for an insurer-specific slope (where we have enough observations to form a reliable estimate).
With these residual pay estimates in hand, we track the risk-taking of these firms from 1995-2000 and 2001-2008, respectively. Specifically, using data from 1995-2000, we calculate various risk-taking measures including firm beta, return volatility, tail cumulative return performance, average holdings of non-GSE backed mortgage-backed securities, and average book leverage. We then regress these risk-taking measures on our lagged residual CEO compensation (from 1992-1994) measure along with other firm characteristics. Similarly, we calculate risk-taking measures for the period of 2001-2008 and regress these on our residual compensation measures constructed from 1998-2000. During the crisis period, we can also compute the sensitivity of a firm’s stock price to the ABX subprime index.
Our final data set comprises two cross-sections: the first containing data on pay of 147 firms (14 primary dealers, 109 banks, 24 insurers) in 1992-1994 and their risk-taking activity in 1995-2000, and the second containing data on pay of 154 firms (11 primary dealers, 109 banks, 34 insurers) in 1998-2000 and their risk-taking in 2001-2008, with 75 firms reporting in both periods.
Table 1 and Table 1 (cont) report summary statistics for log compensation, risk-taking measures and various firm characteristics for our two periods. The figures are similar to those reported in other studies. A couple of comments are helpful here. Since compensation and
19
market capitalization do not scale linearly, we find it convenient to work with log compensation and log market capitalization. For convenience, we report here the raw compensation figures. The mean (median) executive compensation in 1992-1994 was $1.43M ($812K) with a standard deviation of $1.60M. In the 1998-2000 sample, the mean (median) executive compensation was $3.81M ($1.76M) with a standard deviation of $6.27M. Mean (median) firm market capitalization was $3.02B ($1.22B) with a standard deviation of $4.59B in 1994, and was $13.8B ($3.48B) with a standard deviation of $31.2B in 2000. Our sample encompases a broad-cross-section of finance. It includes the top investment banks, commercial banks, and insurers in both the early and late periods (Bear Stearns, Citigroup/Travelers, AIG, etc.), as well as smaller firms.

A. Heterogeneity in Compensation Practices

We first document that there is substantial cross-sectional heterogeneity in executive compensation controlling for firm size and finance sub-industry classifications. The formal regression results are presented in Panel A of Table 2. The first column shows the results for the early period and the second shows the results for the late period. Notice in the early period that the coefficient in front of Log Market Capitalization is positive (0.45) and very statistically significant. The coefficient in front of the insurer specific slope is -0.26 and also significant, indicating that insurer pay increases less quickly with firm size then for primary dealers and banks. The average level of pay also differs somewhat across these three groups, with primary dealers have the highest pay on average, followed by banks and then the insurers. The relationship is economically significant with an R-square above 0.6. The results for the late period in the second column are qualitatively similar.
Figure 1 plots the observations along with the fitted values from the regressions in Panel A of Table 2. Each panel plots the log of average total compensation among executives in each ranking period against log market capitalization, and highlights the relationship for our three groups. For example, Panel A plots, for the early period, the log of executive compensation during 1992-1994 against market capitalization at the end of 1994, with three lines representing
the linear fit of size to compensation for our three sub-industries. A quick eyeball of the figure suggests that there is indeed a strong linear relationship between log total compensation and log market capitalization, with primary dealers having a higher-than-average level of pay relative to banks and insurers and insurers having a lower pay-size slope compared to primary dealers and banks. Panel B of Figure 2 plots the results for the late period. Notice that the two figures are fairly similar. This is not a coincidence as the residual pays from these two periods are quite correlated, as we show below.
Panel B of Table 2 gives summary statistics for log compensation and log market capitalization by sub-industry and period. Together with the regression results from Panel A of Table 2, we can calculate the economic significance of the findings. For example, a one-standard deviation increase in log market capitalization is associated with a 0.73-standard deviation increase in total compensation in the early period among banks and bank holding companies. (A one-standard deviation increase in log market capitalization in the early period for banks is associated with a 1.1553 [1 SD] x 0.4559 [slope] = 0.5267 increase in log pay, which is 0.5267 / 0.7219 = 0.73-standard deviations of log pay for banks.) Given our small sample size and the fact that we have statistical significance, it is not surprising that the implied economic significance from our regression in Panel A of Table 2 is quite large. More interestingly, the residual compensation measures obtained from this regression are highly correlated across the two sub-samples, as shown in Panel C. The correlation between residual compensation in the two periods is 0.61 with a p-value of zero.
Table 3 lists quintile rankings of residual executive compensation (ranked within each sub-industry) for firms prominent in the financial crisis. High residual compensation firms include Bear Stearns, Citigroup, Countrywide, and AIG, and they tend to be high residual compensation firms even as far back as the 1992-1994 ranking period. We emphasize this point because we believe this suggests our residual compensation measure is a noisy proxy for firm-specific compensation practices.
To analyze this point further, we examine whether CEO turnover and stock price performance drive changes in the residual compensation measures. The idea is that if these variables do not drive changes in residual compensation then it is suggestive of something more fundamental about the culture or technology of the firm. Panel A of Table 4 presents the results
of an exercise where we regress quintile rankings of residual compensation in the late period on quintile rankings of residual compensation in the early period, cumulative returns in between the two periods (1995-1997), and whether there was any CEO turnover in between the two periods. The first column shows that the 1992-1994 quintile ranking is significant at the 1% level and explains 21.1% of the variance of 1998-2000 quintile rankings. The second column shows that introducing returns and CEO turnover between the two periods leads to an R-squared of 23.7%. Both coefficients are statistically insignificant. Good past price performance leads a firm to have slightly higher residual compensation in the late period and CEO turnover leads to lower residual compensation, but the bulk of explanatory power for what a firm’s residual compensation ranking is in the late period is provided by the ranking in the early period. Since the theoretical directional effect of CEO turnover on rankings is unclear, in the third column, we regress the absolute value of changes in rankings on an indicator for whether there was any CEO turnover in 1995-1997, and find a statistically insignificant coefficient of 0.10. We repeat this exercise to analyze whether movements in and out of the highest quintile and lowest quintile are driven by returns and turnover in Panel B and find no significant relationship.
Panel C repeats this exercise for raw residual compensation (not quintile rankings) and finds that the coefficient on early period compensation is 0.74; returns and CEO turnover are both statistically insignificant and provide almost no additional R-squared. We conclude that CEO turnover and stock price performance have weak explanatory power for changes in rankings and that the bulk of explanatory power is provided by past rankings. We note finally that a Breusch-Pagan-Godfrey test of serial correlation in the residual compensation between the two periods rejects the null hypothesis of no serial correlation with a p-value of zero.16 As such, we interpret our residual compensation measure as being largely a firm fixed-effect and that there is a substantial cross-sectional variation in this residual compensation measure.
Finally, because we are concerned that sample attrition between our early and late ranking periods may be driving our results, we examine whether there are systematic differences between the 72 firms who are not present in both samples and the 75 that survive. First, we examine whether persistence among firms that drop out of our 1998-2000 sample but are still present in the 1995-1997 sample (there are 33 such firms) is different than persistence for firms that survive. We regress 1995-1997 residual compensation as the dependent variable on 1992-1994 residual compensation and include an interaction with an indicator for whether a firm subsequently drops out. We find no statistical evidence that persistence for dropouts is different than persistence for survivors: in fact, the point estimate on 1992-1994 residual compensation is even higher for the 33 firms who subsequently drop out than for the 75 firms that survive, although the difference is not statistically significant. Second, we look at CRSP delisting codes for the 72 firms that do not survive and find that mergers account for nearly all the firms that drop out. Since targets are typically smaller firms, we examine whether there is a size bias in our results by dropping the bottom 25% of firms by market capitalization in both the 1992-1994 and 1998-2000 samples and repeating our analysis. We find that our estimates of persistence are if anything higher and our results on risk-taking below are virtually unchanged. We conclude that attrition between the two samples is not driving our persistence results.

B. Compensation and Risk-Taking

We now analyze the relationship between our residual compensation measure and risk-taking and find that residual compensation and subsequent risk-taking are strongly correlated in both sub-samples. We start with our price-based measures. Our first set of findings is that firms with high executive compensation have a higher CAPM beta, higher return volatility and are more likely to be in the tails of performance, with extreme good performance pre-crisis when the market did well and extreme poor performance during the crisis period when the market did poorly.
Figure 2 demonstrates the results of predictive regressions where we compute beta, volatility and returns in 1995-2000 (2001-2008) and regress this on residual compensation in 1992-1994 (1998-2000). The formal regressions are in Table 5. Residual compensation strongly predicts beta, volatility, and subsequent returns; the results are statistically significant for all price performance measures at the 5% level and are also economically significant. A one-standard deviation increase in residual pay in the late period is associated with a 0.1135 increase in beta (0.6611 [1 SD of residual pay] x 0.1717 [slope] = 0.1135), which is 0.34-standard deviations (0.1135/0.3364 [1 SD of beta] = 0.34). For the non-crisis period, this number is 0.33 standard deviations. In terms of volatility, a one-standard deviation increase in residual pay in the late period is associated with a 0.0689 increase in volatility (0.6611 [1 SD of residual pay] x 0.1037 [slope] = 0.0686), which is 0.26-standard deviations (0.0686 / 0.2595 [1 SD of volatility] = 0.26). In the early period, this association is 0.28-standard deviations.
The results on cumulative returns are particularly striking – residual compensation strongly predicts cross-sectional differences in subsequent cumulative performance. For example, for the crisis period, a one-standard deviation increase in residual compensation predicts a 21% decrease in returns (0.6611 [1 SD of residual pay] x -0.3205 [slope] = -0.21), which is a 0.26-standard deviation decrease in returns in the cross-section (0.21/0.8105 [1 SD of cumulative returns] = 0.26). In contrast, in the early period, a one-standard deviation increase in residual compensation predicts a 51% increase in returns (0.5236 [1 SD of residual pay] x 0.9773 [slope] = 0.51), or a 0.26-standard deviation increase (0.51/1.955 [1 SD of cumulative returns] = 0.26). A model which regresses returns on compensation in both early and late periods, allowing for different slopes and intercepts in the two periods, rejects the null hypothesis that the coefficient of residual compensation in the early period is equal to that of the late period at the 1% level.
Given the persistence of residual compensation, the results show that aggressive firms that were yesterday’s heroes when the stock market did well can easily be today’s outcasts when fortunes reverse. Bear Stearns (BSC), Citigroup/Travelers (C/TRV), and AIG (AIG) are prime examples. In other words, there is substantial heterogeneity in financial firms in which high-compensation, high risk-taking and tail performance go hand in hand. In particular, it is important to note that this link between compensation and risk-taking (as measured by beta, volatility, and returns) persists in both periods, even before the crisis. This suggests that the persistent effect picked up by our residual compensation measure is consistently linked to risk-taking over time.
We continue with our risk-taking analysis in Figure 2 (cont) where we consider the other risk-taking measures including the sensitivity of firm stock price to the ABX subprime index (ABX Exposure), holdings to non-GSE backed mortgages (MBS Exposure), and book leverage. Our exercise is to regress these measures of risk-taking in 1995-2000 on residual compensation in 1992-1994 and also to regress risk-taking in 2001-2008 on residual compensation in 1998-2000.
Figure 2 (cont) demonstrates the results of these predictive regressions. Table 5 provides formal regression results for ABX exposure for the late period, depicted graphically in Figure 2 (cont) Panel G. High residual compensation in 1998-2000 predicts high exposure to subprime in 2006-2008 at the 1% level. In economic terms, a one standard deviation increase in residual compensation leads to a 0.37-standard deviation increase in ABX exposure (0.6611 [1 SD of residual pay] x 0.1429 [slope] / 0.2577 [1 SD of ABX exposure] = 0.37). The figure immediately reveals that firms prominent in the crisis and most exposed to subprime, such as Bear Stearns (BSC), Lehman Brothers (LEH) and AIG (AIG), were high residual compensation firms in 1998-2000. Compensation also picks out a number of other firms who had high exposure to subprime – Hartford Financial (HIG), an insurer who received $3.4 billion in TARP money, is a high compensation firm, as is Fremont General (FMT). Fremont General was a relatively small California bank that nevertheless managed to originate a significant volume of subprime mortgages nationally and did not stop doing so until faced with a likely cease and desist order from the FDIC in 2007. Afterwards, Fremont General became embroiled in lawsuits alleging predatory lending. In sum, heterogeneity in compensation practices as early as 1998-2000 is able to pick out firms who were subsequently most exposed to risky subprime mortgages.
Since a portion of financial firms’ exposure to the subprime market operated through off-balance sheet vehicles, our ABX exposure measure, which is market-based, should more sharply capture the large risks that banks took than balance-sheet measures. Off-loading risky assets into structured investment vehicles (SIVs), which finance the purchase of these assets using short-term paper, did not off-load the risk from the sponsoring firms themselves. Sponsoring firms often retained risk by granting _liquidity backstops_ or credit lines to these vehicles, to be drawn in case these SIV’s could not continue to finance themselves in the market . This is exactly what happened, bringing enormous losses to the sponsoring firms (Brunnermeier 2009).
We also look at balance-sheet based measures by examining holdings of non-GSE-backed MBS (as a percentage of balance sheet size) and book leverage and their relationship with residual compensation. Figure 2 (cont) Panels H and I and Table 5 report that holdings of non-GSE backed MBS are associated with residual compensation. As mentioned before, these mortgage-backed securities included substantial pools of risky mortgages such as subprime and Alt-A. High residual compensation in 1998-2000 predicts high holdings of non-GSE MBS in 2001-2008 (though it is barely not statistically significant). The results imply that a one-standard deviation increase in residual compensation is associated with a 0.16-standard deviation increase in risky MBS holdings (0.6611 [1 SD of residual pay] x 0.0081 [slope] / 0.0343 [1 SD of MBS exposure] = 0.16). In the early period, we find virtually no association. Since the non-GSE backed MBS market did not become substantially risky until the early 2000’s, when the growth in subprime lending led to a boom in the non-GSE MBS market (Keys et al. 2009 and Chomsisengphet and Pennington-Cross 2006), we actually view this non-result in the early period as consistent with our hypothesis.
Table 5 and Panels J and K of Figure 2 (cont) and show the results for book leverage. In both periods, we find a positive but statistically insignificant relationship. As mentioned, we view book leverage as a very noisy measure of risk-taking since leverage encompasses many different aspects of firm activity and is based on book values.
Our results are not driven only by a particular finance sub-industry. Table 6 repeats our analysis relating risk-taking in 1995-2000 (2001-2008) to residual compensation in 1992-1994 (1998-2000), where we successively drop different groups of financial firms in our analysis. First, we exclude the primary dealers from our analysis and find consistent results across all our measures of risk-taking. Second, because we are also concerned that the results may be driven by the insurance companies, we repeat the analysis dropping insurers. While statistical significance is slightly weaker for volatility and cumulative excess returns, we broadly find consistent results. Finally, we run our results using only banks and bank holding companies, excluding both insurers and the primary dealers. Although statistical significance is more limited (not surprising given that we are losing 25-30% of our sample), our findings are again qualitatively similar. So our results are not just due to primary dealers, though the results are stronger when primary dealers are included. This is not surprising, since these firms have more discretion to take risks (e.g., Bear Stearns and Citigroup).
Moreover, in results not reported, the correlation between risk taking measures and residual compensation is primarily a compositional effect in that changes in the risk-taking measures are uncorrelated with changes in the residual compensation measure. This drives home again the point that we are dealing with permanent cross-firm differences.
Additionally, in Table 7, we look at the different components of pay and find that both bonuses and equity compensation are correlated with risk-taking, consistent with earlier empirical literature which finds that bonuses and equity compensation motivate short-term behavior.17 Although statistical significance in this exercise is more limited, and thus we use caution in interpreting our results, we find that, consistent with concerns about bonuses and risk-taking, a one-standard deviation increase in (residual) bonuses in 1998-2000 is associated with a 0.33-standard deviation drop in 2001-2008 returns (0.9733 [1 SD of residual bonus] x -0.2761 [slope] / 0.8105 [1 SD of cumulative returns] = -0.33). Salary is a much less informative predictor of risk-taking; in fact, firms with higher salaries did better in the late period than otherwise.

C. Robustness Checks

In Table 8, we perform a series of robustness checks of the above findings. First, we re-do our analysis by calculating residual compensation using book asset values rather than market value on the idea that asset values are more exogenous than firm size. This is reported in the first row. We report only the coefficient in front of residual compensation both for the early and late period for each of the risk-taking measures, which are given by the columns. The results are very similar to before.
One may worry that residual compensation is simply a proxy for book leverage. Our results indicating a low correlation between residual compensation and book leverage suggest this is not the case, but we present more formal analysis in the second row of Table 7. Controlling for size and book leverage does not significantly affect our results. To further examine this hypothesis, in results not reported, we also include leverage on the right-hand side when computing residual compensation in the first-stage and find that leverage only marginally improves the fit between compensation and size and does not affect our risk-taking results. In sum, our price-based risk-tking measures are not driven by differences in book leverage.
Third, we exclude the CEO’s pay when computing our residual compensation measure and find nearly identical results.18 Even after excluding the CEO, a one-standard deviation increase in residual compensation is associated with a 0.36-standard deviation increase in ABX exposure when using non-CEO compensation (0.6552 [1 SD of non-CEO residual pay] x 0.1400 [slope] / 0.2577 [1 SD of ABX Exposure] = 0.36). While ideally we would have data on compensation of other employees at financial firms (e.g., traders), the persistence in residual compensation and the positive association between non-CEO executive compensation and risk-taking suggest that residual compensation is more indicative of a firm effect such as culture.
Fourth, we do the same exercises for manufacturing industries as an out-of-sample check since the theory of short-termism and risk-taking should apply to non-financial industries as well. However, one might expect these effects to be stronger for finance firms where risk is a much bigger deal, except for book leverage. This is indeed what we find from the results reported in the fourth row. We get qualitatively similar results but markedly weaker statistical significance. Note in particular, that residual compensation has no explanatory power for ABX exposure among non-finance firms, which is also a good check that our ABX exposure results are not spurious. There is some effect for return volatility and beta among non-finance firms but much weaker than before for the late period and of little to no significance in the early period.
Fifth, we run a pooled regression version of our analysis. More specifically, rather than just running two cross-sectional regressions, an early period and a late period, we do the following exercise. For each year, we calculate the risk-taking measures (except for long-horizon returns) using only one year’s worth of daily data. We then run a pooled regression of each year’s risk-taking measure on lagged residual pay calculated using the previous three years worth of data. We include risk-taking measures for which we have data during the full 1995-2008 period (beta, volatility, MBS holdings, and leverage) and omit results for returns where we are looking for long-horizon effects and the ABX since it only applies to the last two years in our sample. We find similar results for beta, return volatility, and MBS holdings. The bulk of the analysis tells us that our results are not an artifact of how we cut the sub-periods in our analysis.

D. Is it Short-Termism?

Fourth, we go on to ask whether our compensation measure and results are picking short-termism as opposed to unobserved heterogeneity. An alternative interpretation is that high residual compensation firms are like the growth firms of finance, where risk-taking requires talent and talent gets rewarded with higher compensation in general (both short-term and long-term) compared to low residual compensation and presumably low growth firms. One way to examine this hypothesis is to see whether our compensation/risk-taking correlation is just due to growth options, proxied by a firm’s market/book ratio. Under this hypothesis, market/book should be positively correlated with subsequent risk-taking.
Table 9 presents results where we include in our regressions of risk-taking and compensation a firm’s market/book and market capitalization at the end of each ranking period. Importantly, our results for residual compensation still hold strongly across a broad array of risk measures such as ABX exposure, beta, volatility and subsequent returns during the crisis, even controlling for heterogeneity in firm market/book and firm size in risk-taking. In contrast, for many risk-taking measures, a firm’s market/book has limited explanatory power in comparison. For our ABX measure, higher market/book is associated with lower subprime exposure, which does not support the growth firm hypothesis.
In Table 10, we examine the hypothesis of short-term compensation explicitly by regressing risk-taking of firms on compensation while controling for insider ownership on the presumption that insider ownership is a proxy for long-term incentives. If indeed compensation is capturing long-term pay incentives (as opposed to short-term pay as we suspect), then having insider ownership should mute our results and we should also expect insider ownership to predict risk-taking with the same sign as compensation. We measure insider ownership by the average percentage of shares held by the top 5 five executives in 1992-1994 and 1998-2000 in the early and late periods, respectively.
Our baseline findings on compensation remain even after controlling for insider ownership. Our point estimates on the association between residual compensation and risk-taking are remarkably similar, and the statistical and economic significance are also of similar magnitudes. We find some evidence that insider ownership tends to mitigate risk-taking. In particular, high insider ownership was associated with lower beta and exposure to MBS in the late period. A one-standard deviation increase in insider-ownership in the late period is associated with a 0.18-standard deviation decrease in MBS exposure (0.0730 [1 SD of insider ownership] x -0.0868 [slope] / 0.0343 [1 SD of MBS Exposure] = -0.18). Importantly, higher insider ownership was qualitatively associated with higher returns in both periods, reinforcing the view that compensation provided short-term incentives while insider ownership provided long-term incentives. Although the effect is not statistically significant, in economic terms, a one-standard deviation increase in 1998-2000 insider ownership is associated with a 0.17-standard deviation increase in buy-and-hold returns in 2001-2008 ([0.0730 [1 SD of insider ownership] x 1.8507 [slope] / 0.8105 [1 SD of cumulative returns] = 0.17), or 13.5-percentage points. Because we are concerned about how insider ownership scales with firm size, we introduce an additional control for market capitalization and find effects of similar statistical significance and economic magnitude. This additional evidence indicates that our risk-taking results are being driven by short-termism.

E. Is it Mis-Governance? Or Investor Preference?

Finally, we ask whether our results are due to mis-governance or entrenchment as opposed to simply heterogeneity among investors who want to invest in high risk-taking firms and hence need to set compensation appropriately to induce such behavior. One particular hypothesis we explore is that institutional investors (e.g. mutual fund managers), who themselves have short-horizons due to agency issues, might want to pressure finance firms to take risks to meet quarterly targets.
To begin with, we repeat our exercise with various measures of governance on the right-hand side. The measures of governance that we examine are measures of entrenchment (Gompers, Ishii and Metrick 2003 and Bebchuk, Cohen and Ferrell 2009), board independence (the percentage of outside directors on the board), and board size (Yermack 1996). In Table 11, we consistently find that none of these standard measures of governance predict risk-taking, nor are they associated with our measure of residual compensation. Entrenchment measures do not predict risk-taking, nor, surprisingly, cumulative returns in either period.19 The exceptions are that the G index is negatively correlated with ABX exposure (at the 10% level), and that the E Index is negatively correlated with MBS exposure (at the 10% level). The negative correlation, however, is puzzling, as it suggests that managers who were more entrenched (i.e., managed firms with weaker shareholder rights) are associated with less exposure to subprime.
Also, weaker shareholder rights is not associated with high residual compensation or subsequent risk-taking in either the early or late periods. Our results on board composition and board size reinforce this non-correlation result. The economic significance of board composition is also not significant: a one-standard deviation (16-percent) increase in the percentage of outsiders on a board (the mean is 65% in the late period) is associated with only a 0.05-standard deviation drop in ABX exposure (0.1629 [1 SD of percentage outsiders] x -0.0771 [slope] / 0.2577 [1 SD of ABX Exposure] = -0.05). Overall, we consistently find that the percentage of outsiders on the board do not predict subsequent risk-taking such as ABX or MBS exposure, nor even returns, during the crisis period.20
Next, we explore an alternative hypothesis: that risk-taking and executive compensation may be related to heterogeneous shareholder preferences. One interpretation, motivated by Froot, Perold and Stein (1992), is that there is heterogeneity in investor preferences with short-termist investors (say institutional investors such as mutual funds who themselves have short horizons) who want certain firms to take more risks and hence give them short-term incentives to do so. To implement this hypothesis empirically, we use two proxies, institutional ownership as the direct measure and share turnover as an indirect measure of institutional ownership presence since institutional investors are well-known to trade more (Hong and Stein 2007).
There is significant variation in average institutional ownership. Institutional ownership (averaged over the 12 quarters of 1992-1994 within firms and then across across firms) averaged 48% with a standard deviation of 17%; the corresponding numbers in 1998-2000 are a mean of
46% and standard deviation of 18%. The interquartile ranges were 24% and 27% in the early and late periods, respectively. These numbers parallel Erkens, Hung, and Matos (2009) although we have a smaller standard deviation.
We compute the average monthly stock turnover in the 36 months of 1992-1994 (1998-2000) and then relate this to to residual compensation and subsequent risk-taking in 1995-2000 (2001-2008). From the summary statistics for our measure of speculative activity in Table 1 (cont), there is significant heterogeneity in share turnover – the mean monthly share turnover among stocks in the 1998-2000 period was 8% with a standard deviation of 4%, ranging from 1% (the minimum) to 23% (the maximum). Even in the early period, before the height of the dot-com bubble, the standard deviation of monthly turnover was 4%, with a maximum of 19%.
Table 12 presents the formal regression results relating residual compensation and all the risk-taking measures to institutional ownership and share turnover. Institutional ownership is linked to residual compensation at the 10% level in the late period; in economic terms, a one-standard deviation increase in institutional ownership is associated with a 0.17-standard deviation increase in residual compensation (0.1823 [1 SD of institutional ownership] x 0.6140 [slope] / 0.6611 [1 SD of residual pay] = 0.17). This relationship is even stronger in the early period, where the economic relationship is 0.35-standard deviations in residual compensation for one standard deviation in institutional ownership (0.1655 [1 SD of institutional ownership] x 1.1163 [slope] / 0.5236 [1 SD of residual pay] = 0.35). Institutional ownership and risk-taking also positively covary (the exception being for book leverage where it goes the wrong way). The effect is particularly strong in return volatility – in the late period, a one-standard deviation increase in 1998-2000 institutional ownership is associated with a 0.26-standard deviation increase in subsequent return volatility (0.1823 [1 SD of institutional ownership] x 0.3658 [slope] / 0.2595 [1 SD of volatility] = 0.26).
We next turn to share turnover, which we view as simply being another proxy for institutional ownership presence since institutional investors are known to trade more frequently than retail investors. However, there might be differences in terms of the short-termism even among institutional investors. So share turnover is one way to pick up such differences across firms. But one has to be a bit careful in interpreting this variable since institutional investors who say always quickly flip out of the shares of a company are not likely to have any influence over management.
In particular, a one-standard deviation increase in average monthly stock turnover is associated with a 0.27-standard deviation increase in residual compensation in the late period (0.0423 [1 SD of stock turnover] x 4.2065 [slope] / 0.6611 [1 SD of residual pay] = 0.27), an effect that is significant both in economic and statistical magnitude (at the 5% level). There is also positive link between share turnover and risk-taking. For instance, high share turnover in 1998-2000 predicts subsequent risk-taking in 2001-2008, in particular book leverage, exposure to ABX, beta, volatility, and tail performance. Because we are concerned that we may be introducing a scale effect in using share turnover (which is a ratio of volume over shares outstanding) as a proxy for speculative activity, we add a control for market capitalization and find remarkably similar results. For example, a one-standard deviation increase in share turnover in the late period is associated with with a 0.27-standard deviation increase in ABX exposure (statistically significant at the 10% level), a 0.32-standard deviation increase in beta (significant at 1%), and a 0.20-standard deviation decrease in returns (significant at the 5% level).
Broadly, the evidence supports a story where short-term investors incentivize management using short-term incentives to take large bets on risky propositions. This alternative does not necessarily imply that managers were fully aware of their risks. If shareholders in certain firms want their managers to take risks they will offer appropriate contracts. Managers will also select themselves into these firms. Ceteris paribus, these firms will end up with managers that have more tolerance for risks or that do not fully perceive risks. As an example, one might think that Joseph Casano (of AIG FP) or Stanley O’Neal were ideal managers for stockholders that wanted their firms to take a lot of risk.
We also stress that we do not view this hypothesis as incompatible with the hypothesis that entrenchment is a significant problem that led to the crisis, but in light of the non-correlation
between shareholder rights and both risk-taking and price performance, at a minimum our results suggests that further research should explore investor preferences as an alternative hypothesis to failures of governance.

V. Conclusion

Rather than restating our findings which suggest a link between compensation and risk-taking, we will draw some directions for future research and policy implications from our analysis. On the research direction front, it is worth restating again the lack of causal statements in this analysis. What this analysis points to is that there is important heterogeneity across firms in risk-taking (i.e. Bear Stearns, Lehman and Citigroup have always been skating on the edge and have come close to failing before the most recent events) and importantly, this is very correlated with compensation. While not causal, our analysis suggests a beginning at least in terms of being able to quantify these issues. It also suggests that deeper research into the nature of implicit incentives, peer effects, and organizational structure might bear fruit as far as understanding risk-taking by finance firms.
As far as policy implications, in principle, financial market taxes and regulations should be directed at curbing the activities that increase systemic risk. However, this form of regulation may not be enough because (i) it is often difficult to measure with a reasonable degree of precision the systemic impact of an individual firm’s actions and (ii) the regulated firms would respond with innovations that minimize the impact of regulations. Perhaps for these reasons, regulation that align the incentives of managers and shareholders with the externalities caused by financial firms are also being contemplated.
The Obama administration’s compensation reforms of longer vesting periods and _say on pay_ certainly are sensible and reasonable, but will they be effective? The answer depends crucially on our last finding regarding whether the correlation between short-termist pay and risk-taking is due to mis-governance or if simply reflects heterogeneous demand on the part of investors for risk-taking finance firms. If it is indeed due to entrenchment, then imposing better governance makes sense. If the issues emanate from preferences of shareholders, then more _say on pay_ may not be helpful. Limits to pay as a function of short-term performance may help, because they would restrict the incentives that stockholders can give to managers to induce risk- taking. But unless total pay is limited, stockholders may simply reward managers excessively and threathen to fire them if short term targets are not met. Our findings shed light on the importance of institutional investors and their role in the crisis, which thus far as been ignored.
If short-termism among stockholders and institutional investors in particular is the problem, one might consider a capital gains tax that more aggressively rewards long-term holdings. Indeed, some CEOs such as Louis Gerstner, the former CEO of IBM, have called for such a scheme in critiquing the effectiveness of the Obama compensation reforms. A dramatic scheme such as rewarding at the extremes long-term holdings with zero capital gains tax and punishing short-term holdings with say the ordinary income tax rate might be contemplated. The current US tax scheme for all intents and purposes makes no distinction between short-term and long-term as long-term is defined as greater than a couple of years. One disavantage of a tax scheme is that, unless it is restricted to financial stocks, it would also affect industries where more risk-taking may be socially advantageous, but at a minimum, it would seem that further research into the nature of how tax schemes influence stockholdings’ horizons, perhaps appealing to international variation in tax regimes, would be desirable.