There are two prevailing schools of thought on what determines CEO pay. The principal-agent view (as articulated by Core et al. 2003, among many others) suggests that boards design CEO compensation contracts so as to provide incentives for CEOs to work hard and maximize shareholder value [3] . Alternatively, the managerial power view (Bebchuk and Fried, 2003) asserts that CEOs exert influence over their boards that effectively allows them to participate in setting their own pay [4] . The two approaches suggest different explanations for large pay packages; under the principal-agent approach, high average levels of pay may be necessary to compensate the CEO for the risk he bears by having his compensation tied to firm performance (via stock and options), whereas under the managerial power approach high pay might indicate weak corporate governance-specifically, a high degree of leverage for the CEO over the board and compensation committee. Crystal (1991) argues that boards of directors over-com- pensate senior executives because outside directors are hired by the CEO and can be removed by the CEO [5] . Likewise, Lambert et al. (1993) and Boyd (1994) find a positive relation between CEO compensation and the percentage of the board composed of outside directors [6] [7] . However, Finkelstein and Hambrick (1989) conclude that compensation is unrelated to the percentage of outside directors on the board [8] . Conyon (2006) provides a survey of the evidence supporting these alternative perspectives [9] .

Empirical evidence for the managerial power argument has focused on linking CEO pay to measures of board capture or weak corporate governance. Chhaochharia and Grinstein (2009) show that when new standards for board independence were introduced in the wake of corporate scandals in the early 2000s, CEO pay declined disproportionately at firms most affected by the new rules [10] . In a sample of IPO firms, Conyon and He (2004) find that CEOs with large stakeholders on the board (who are presumably stronger monitors) get pay packages tied more closely to firm performance (less cash and more equity) [11] . However, the evidence is not uniform; for example, Boyle and Roberts (2013) find that CEOs of New Zealand firms who sit on the board compensation committee are paid less than those who do not [12] . Sur et al. (2015) find cash compensation is mostly driven by firm specific factors, while equity driven compensation is largely determined by time-varying factors with firm and industry effects playing a minor role [13] . Frydman and Saks (2010) add that there has been a fundamental transformation in CEO compensation with stock options becoming a greater portion of total compensation.

Murphy (2002) rebuffs the managerial power hypothesis in part by finding that there is an insignificant difference in total compensation between CEOs hired from the inside versus the outside [14] . Murphy (2002) concludes that this finding is inconsistent with the managerial power view that insider CEOs use their relationships with the boards to extract rents. Instead, he attributes his findings to differences in preferences for riskier stock options between outsider and insider CEOs.

Meanwhile, there is a substantial body of empirical evidence that CEO pay is increasing in firm size (Kostiuk 1989, Murphy 1985) [15] [16] ; typical estimates put the elasticity of pay with respect to size at 0.25 (so a 10% increase in size is associated with a 2.5% increase in pay). Under the principal-agent perspective, this relationship can be seen as reflecting the need to give the CEOs of larger firms, incentives commensurate with the larger economic impact of their actions (Gayle and Miller 2009) [17] . In contrast, the more cynical view of the managerial power approach would be that larger firms present CEOs with more opportunities to carve out rents (Bebchuk and Fried 2003).

This empirical relationship adds nuance to a concern with older roots: namely that CEOs are interested in empire-building for its own sake. Writers, such as Marris (1964) and Aoki (1984) argued that, unencumbered by external constraints, executives are more interested in increasing firm size than maximizing shareholder value [18] [19] . Tosi et al. (2000) attributes this in part to the idea that CEOs can exert more complete control over firm size than performance [20] .

In light of this empirical evidence, a growing theoretical literature has attempted to elaborate on the reasons that firm size and CEO pay might be linked. Rosen (1992) suggests that the marginal productivity of a CEO may scale up with firm size because his decisions affect how a larger pool of employees and resources are deployed [21] . When this is so, competitive labor markets will tend to match more talented CEOs with larger firms where they are paid according to their productivity, establishing a positive association between size, CEO ability, and pay. Baker and Hall (2004) extend this theory and take it to the data, finding support for the conjecture that CEO productivity rises with firm size [22] . Building on these results, Gabaix and Landier (2008) develop a competitive equilibrium assignment model of the market for CEO talent and show that calibrated predictions of pay can help to explain not only the size-pay correlation but also time series and cross-country patterns in pay [23] . Tosi et al. (2000) concur that reasons for the size-pay correlation include greater complexity, more stratification, and greater human capital at larger firms.

Under the principal agent view, CEO compensation is linked to firm size, giving rise to the question whether CEO motives for consummating mergers and divestitures. Mergers and divestitures are highly visible actions that a CEO can take to grow or shrink his firm, respectively; if he expects these actions to be rewarded or punished in future compensation, there may be a temptation to make decisions that are not aligned with maximizing firm value. On the question of mergers, it has long been suspected that CEOs may be prone to empire-building at the expense of shareholder value. This view is bolstered by considerable empirical evidence that stock markets treat merger announcements as negative news for the acquiring firm and that acquirers tend to overpay for their acquisitions (Moeller et al. 2004) [24] .

A number of studies have found that CEOs tend to be rewarded for mergers with higher pay (Hartford and Li 2007, Grinstein and Hribar 2004, Yim 2013) [25] - [27] . Harford and Li (2007) find that growth via internal capital investments is not similarly rewarded and suggest that, for some reason, mergers give CEOs particular leverage in bargaining with boards for better pay. Yim (2013) shows that the pay increase in the year after an acquisition is retained in subsequent years (thus strengthening the incentives of all CEOs to complete mergers, but particularly younger ones who will enjoy these gains over a longer career horizon). If CEOs pursued only mergers that were in their shareholders’ interest, then rewarding them for mergers might be appropriate. Harford and Li (2007) address this by examining the relationship between CEO pay and his firm’s post-merger performance. They find that the overall pay gain from mergers conceals an asymmetry: CEOs are rewarded for mergers that turn out well but are not punished for mergers that turn out poorly. One conceivable explanation is that captured, or partially captured boards, tend to insulate CEOs from the downside consequences of their mistakes. In fact, there is a body of evidence indicating that unwarranted generosity after mergers is exacerbated by weak corporate governance (Harford and Li 2007, Grinstein and Hribar 2004). Grinstein and Hribar (2004) find that weaker boards tend to grant larger bonuses for mergers. Harford and Li (2007) show that CEOs are punished for poor post-merger performance when the board is strong. Furthermore, CEOs with weaker boards tend to make deals that are larger and less profitable (as judged by market reaction) (Harford and Li 2007), suggesting that they may not anticipate consequences for failure.

In contrast with the healthy literature on mergers, there is relatively little evidence on how divestitures and other decisions to shrink the firm affect CEO pay. In principle, divestitures should be rewarded or punished, just as with mergers, on the basis of their effect on shareholder value. Bebchuk and Grinstein (2005) find the asymmetric result that increases in firm size lead to higher pay, but decreases in size (measured either by sales or market capitalization) do not cause pay to fall [28] . In one of the few studies to focus explicitly on divestitures, Haynes et al. (2007) examine panel data from British firms [29] . They find that in general, CEOs are punished for divestitures via the robust link between firm size and pay, suggesting that CEOs will have little incentive to downsize even when it is the right thing to do. However, firms with strong boards are an exception to this rule: they tend to reward CEOs for divestitures.

This paper then examines the influence of mergers and divestitures on CEO pay. As part of the analysis, we also assess the influence of board composition and the way these factors have been changing over time, particularly pre and post financial crisis.

Finally, the level and mix of CEO compensation may have changed since the financial crisis of 2008. Gabaix et al. (2014) examine the size-CEO compensation relationship before and after the financial crisis and find the effect to be very similar between the two periods with compensation going up proportionally pre-crisis and going down at a similar rate post crisis. Vemala et al. (2014) also found that firm size along with firm performance and board composition had similar effects on CEO compensation, pre and post financial crisis [30] . They, however, also found that while total compensation increased during and after the 2008 financial crisis, cash compensation decreased. Faulkender et al. (2010) had similar findings that CEO compensation increased while stock market prices fell [31] . These findings conflict, in some respects, with the agency view of CEO compensation since stock prices and earnings declined significantly in 2009. This paper adds to the literature by assessing how the financial crisis altered the determinants of CEO compensation.

This paper tests three hypotheses:

Hypothesis 1: CEO compensation is influenced by the financial crisis. We hypothesize that the financial crisis had a disciplining effect altering CEO compensation contracts toward variable compensation (tied to stock market performance and earnings per share) as measured by TDC 1 or TDC 2 and away from cash compensation as measured by salary and/or total compensation. In addition, we expect pay for performance to have a greater effect on CEO compensation post versus pre-financial crisis. In other words we hypothesize that financial performance variables, such earnings per share of stock market performance, interacted with the financial crisis effect will have a positive influence on CEO compensation.

Hypothesis 2: CEO compensation is positively influenced by merger announcements and/or relative merger value, as measured by merger value percentage (transaction value/firm market capitalization), since mergers add to firm size. We, however, do not expect CEO compensation to be significantly impacted by divestitures because divestitures reduce firm size but also may trim poor performing businesses. As a corollary, we expect this effect to be altered in the post financial crisis period with mergers not having a significant impact on CEO compensation post financial crisis

Hypothesis 3: Following recent evidence on the disciplining effect of more independent boards, we expect CEO compensation to be decreasing in the fraction of independent board members (relative to affiliate and employee members). On the theory that independent boards will respond more forcefully to the shock of the financial crisis, we conjecture that this effect will grow stronger post-2008.

The paper analyzes CEO compensation for a set of 123 US companies (24.6% of the S&P 500) over a ten-year time period between Jan 1, 2003 and December 31, 2012, spanning equally before and after the financial crisis. In order to assess the effect of large mergers and divestitures, the sample was restricted to companies that engaged in a merger or a divestiture with a transaction value greater than $1 billion. Data on merger and divestiture values and announcement dates were found using the SDC Thomson database. The data set includes 97 companies that engaged in a merger or divestiture and 26 other companies² that did not engage in a large merger (>$1 million) during the ten-year time span. Each of the 26 companies was selected because it is a leading competitor for one of the 97 merging companies.

CEO compensation was measured in four ways: salary, total current compensation (salary plus bonus), TDC 1 (Total Direct Compensation 1), and TDC 2 (Total Direct Compensation 2). Compensation data was taken from the Execucomp section of the Compustat database, a Wharton Research Division Service (WRDS) database that provides company specific data to include various measures of CEO compensation, Company sales, profit, and company director information. CEO compensation can measured by salary, salary plus bonus, or salary plus bonus plus stock compensation (TDC 1 or TDC 2). TDC 1 estimates the value of total compensation realized by the executive in a given year, measured by salary, bonus, the value of restricted stock granted, the total value of stock options exercised, and the value of long-term incentive payouts. Thus TDC 1 comprises the value of total compensation awarded (but not necessarily realized) to the executive that year. TDC 2 estimates the value of total compensation using the same method except it replaces the value of restricted stock granted with the net value of stock options exercised. As Kaplan and Rauh (2010) explain: “…TDC 2 will be closer to an executive’s true adjusted gross income while TDC 1 will more closely approximate the compensation a company’s board expected to pay the executive” [32] .

Firm size was measured using the firm’s sales. Lagged firm sales³ were used because, when considering pay for performance, CEO compensation is awarded after the firm’s prior year financial performance is completed.

We also investigated the effect of profitability, as measured by earnings per share (EPS) from operations (lagged), on CEO pay to explore whether the link between pay and performance. We also used the firm’s annual stock market performance (lagged) as an independent variable to measure pay for performance.⁴ Profitability measures along with firm sales were taken from the Compustat database, while stock market and S&P data were gathered using the Center for Research in Security Prices (CRSP) database. We expect EPS and annual stock market returns to have a positive effect on incentive compensation.

In addition, we used dummy variables to control for whether the CEO was a new hire (1) and whether he was hired from outside (1) or inside the firm. We expect that incentive compensation will be lower if the CEO is a new hire, as incentives may be built in over time. Also, we expect that CEOs hired from the outside may receive higher salary or incentive compensation than inside hires, based on the theory that the firm faces more competition for outside hires (both from the CEOs’ current employers and from other firms trying to hire them). Finally, we interacted new hire and outside under the theory that new hires that were inside the firm may be in a better position to extract rents from the board than outside new hires.

Finally, we examined board composition using the Risk Metrics database. Each board member is categorized in this database as being an employee, linked (affiliated⁵), or independent. It has been assumed that the higher the percentage of affiliated board members the higher the CEO compensation, though statistical studies have not completely borne this out (see Hallock 1997; Core et al. 1999, and Fich and White 2003) [33] [34] . We also hypothesize that independent board members are chosen by the CEO suggesting they would award the CEO higher compensation than an employee board member.

Table 1 shows company sales, market capitalization, and compensation by industry. From the table we see that the sample is fairly well dispersed between industries. The

consumer products, technology, industrial, and healthcare sectors each accounted for 15% to 17% of the sample. The oil and gas and financial segments were 12% and 11% of the total, while the utility, retail, and transportation sectors were 6%, 5%, and 2% of the total. As a robustness check we will interact the industries with merger and divestiture announcement to test whether any one industry is biasing the results.

From Table 1 we see the highest average salaries among sampled firms are paid in the consumer industry, followed by the retail industry. One of the commonalities between these two industries is that they are relatively non-cyclical. The cyclical industries, such as oil and gas, industrial, and technology industries each pay lower salaries but higher stock and option compensation, as shown in TDC 1 and TDC 2. Companies in the financial sector are the largest as measured by market capitalization, while companies in the utility and transport sectors maintained the lowest market capitalization. The retail sector had the highest compensation using TDC 1 and TDC 2, as well as total current compensation.

Figure 1 shows annual pay by compensation category over the sample period. Here we see increasing compensation for all categories over the first three (2003-2005) to six years (2003-2008) of the sample. A sharp decline in the compensation level in each of the categories, except salaries, then occurs between 2006 and 2008. The compensation categories that include stock and option grants (TDC 1 and TDC 2) show a time trend similar to overall stock market performance over the period: rising through the first half of the decade and then declining in the years leading up to and including the recession of 2008/2009, before recovering around 2010. TDC 2 declined particularly sharply since this compensation category includes stock options exercised, and is therefore closely tied to contemporaneous market performance. TDC 1 declined more gradually, as TDC 1 includes stock options awarded. We also see in Figure 1 that average salary increased slightly (from $1.0 million to $1.3 million), while total current compensation (salary plus bonus) actually fell between 2003 and 2012. In contrast with the decline in TDC 2, which is sharpest in 2008, the declines in total current compensation and TDC 1 begin in 2006, preceding the recession.

Figure 2 shows the annual number of merger announcements and divestitures, while Figure 3 shows the average transaction value of merger announcements and divestitures. We see from Figure 2 that merger announcements increased for the first three years of the sample period and then declined over the last four years of the sample (2009-2012) to the 2003/2004 level. Divestiture announcements remained steady between two and five for the first six years of the sample and, like merger announcements, declined significantly in 2009. There were then a large number of divestitures that occurred in 2011.

Table 2 shows the total number of mergers and divestitures by industry as well as the average and relative merger value by industry. Here we see that the largest number of mergers occurred in the technology industry, while the highest merger values occurred in the financial industry. Also, we observe that the high relative merger value (merger value/market capitalization) occurred in the utility sector. Regarding divestitures, we see that the consumer sector accounted for the largest number and largest value of divestitures.

From Figure 3 we see that, in addition to the growing number of mergers between 2003 and 2005, the average merger value was also growing larger. Merger value then declined but rose sharply in 2008 and 2009, suggesting few large mergers were announced in the recession years.

Figure 4 shows CEO board composition by year. There is a gradual but sustained trend toward more independent board membership over the period, as the average fraction of independent members rises from 70% to 85%. This rise is made possible by a steady decline in both employee and affiliated board members over the ten years.

These trends in CEO compensation and board composition are confirmed in Table 3, which shows the averages pre and post financial crisis. From Table 3 we see that the average market capitalization decreased while the average firm sales increased. In addition, we observe that 11.6% of CEOs were newly hired in any given year, but this amount declined from 13% to 10% pre versus post financial crisis. In contrast, the number of outside versus inside hires increased post financial crisis from 13% to 20%. The change in each of these statistics will be explored further through our econometric modeling.

Finally, in Table 4 we segment each of the statistics between merging/divesting and non-merging firms. Here, we see the average non-merging/divesting firms in our sample were smaller than the merging/divesting firms. Accordingly, CEO compensation

among the non-merging/divesting firms was less in each pay category.

Our first set of results are based on variations on the estimation equation below, where y_jt represents logged compensation for the CEO at firm j in year t. The righthand side includes suite of regressors (described below), an indicator for the post-crisis period (2008 or later), and industry fixed effects. The estimation is based on a panel data

model with random effects. (A Hausman test does not reject random effects in favor of fixed effects.)

Separate regressions were run for each of the four executive compensation measures. is the firm’s (lagged) log sales.⁶ is a vector of lagged firm performance measures; it includes log earnings per share, stock market return, and (for benchmarking) the S&P return. The pair of variables includes the fraction of board members who are independent or affiliated; the fraction of employee board members is the omitted category.

The vector includes variables related to transactions (mergers and divestitures). Only large transactions (value of $1 billion or greater) are considered. For mergers, the variables include (an indicator for announcing a large merger) and (the transaction value of any large mergers as a fraction of the firm’s market capitalization at the time of the announcement). Similarly, we have and for divestitures. We also include an indicator variable for firms that carried out any large merger or divestiture over the entire time horizon of our sample.

Finally, x_jt is a vector of indicator variables specifying whether the current CEO was an outside hire, whether he or she is a new hire in year t, and an interaction of these two.

Specification (I) allows for a break in the level of compensation after the crisis, but otherwise it treats the coefficients on factors influencing pay as constant over time. This establishes a benchmark for comparison with other studies of the determinants of CEO pay. However, our conjecture is that the relationship between pay and the right hand side variables changed after the crisis. To test this, we estimate Specification (II) which includes all of the regressors in Specification (I) but adds interactions between the dummy and most of the regressors.

Finally, we conjecture that CEOs are rewarded differently for deal making across different industries. To test this, we estimate Specification (III) which includes all the regressors in Specification (I) and adds interactions between the variable and a firm’s industry.⁷