The Long-Term Consequences of Short-Term Incentives

This paper shows that short-term stock price concerns induce CEOs to take value-reducing actions. Vesting equity, our measure of short-term concerns, is positively associated with the probability of a firm repurchasing shares, the amount of shares repurchased, and the probability of the firm announcing a merger and acquisition (M&A). When vesting equity increases, stock returns are more positive in the two quarters surrounding both repurchases and M&A, but more negative in the two years following repurchases and four years following M&A. These results are inconsistent with CEOs buying underpriced stocks or companies to maximize long-run shareholder value, but consistent with these actions being used to boost the short-term stock price and improve the conditions for equity sales. Overall, by identifying actions that carry clear value implications, this paper documents the long-term negative consequences of short-term incentives.


Introduction
The short-termism of executive incentives is a major problem alleged by academics, practitioners, and policymakers. A central concern in Bebchuk and Fried's (2004) influential critique of executive pay is that CEOs are rewarded for short-term increases in the stock price; hence, their main proposal for pay reform is to escrow the CEO's equity until the long-term (Bebchuk and Fried (2010)). The UK Government's response to its Green Paper on Corporate Governance Reform proposes increasing the minimum vesting period of equity from three to five years.
The concern with short-term incentives is that they lead to the CEO taking myopic actions that boost the short-term stock price at the expense of long-run value. However, critics' allegations are rarely backed up by systematic evidence. Gathering such evidence is particularly challenging for two main reasons. First, it is very difficult to demonstrate a causal effect of short-run horizons since the CEO's contract is endogenous. Second, even if one could show that CEO incentives cause particular actions, it is difficult to show that such actions are myopic, i.e., erode long-term value.
Edmans, Fang, and Lewellen (2017, "EFL") address the first challenge by introducing a new measure of CEO incentives: the amount of stock and options scheduled to vest in a given quarter.
Vesting equity is highly correlated with same-quarter equity sales, so it leads to short-term stock price concerns, analogous to the relevance criterion for a valid instrument. 1 It also depends on the magnitude and vesting schedule of equity grants made several years ago, and so is unlikely driven by current economic conditions -analogous to the exclusion restriction for a valid instrument. EFL show that vesting equity is significantly correlated with reductions in investment growth. They study investment since it is arguably a firm's most important day-to-day decision. However, since we can only observe the level of investment and not its quality, it is difficult to assess the value implications of the investment cut and thus address the second challenge of showing that the cut is myopic. If the scrapped investments would have been wasteful, the implication of short-term stock price concerns is very different -far from inducing myopia, they encourage the CEO to rein in empire-building or excess expenditure. While EFL conduct cross-sectional tests that are suggestive of the myopia interpretation, they are unable to use long-run stock returns to study the long-term consequences of investment cuts, for three reasons. First, any association is unlikely to be causal, because long-term stock returns are likely affected by many firm decisions other than investment. Second, there is no announcement date for investment cuts, as firms are only required to report investment at a quarterly frequency. Third, their sample period is relatively short .
This paper studies two corporate actions whose long-run consequences can be more accurately measured, enabling us to assess the long-term consequences of short-term incentives. The first is stock repurchases. Like investment cuts, repurchases boost the short-term stock price (Ikenberry, Lakonishok, and Vermaelen (1995)) and so CEOs with short-term concerns might have incentives to undertake them. Also like investment cuts, repurchases can either be myopic (if financed by scrapping valuable projects) or efficient (if financed by free cash which would otherwise have been wasted). Critically, unlike investment cuts, the long-term stock return can be used to diagnose the value implications of the repurchase even if it were not caused by the repurchase. The long-term stock return measures the return that the firm obtains from the repurchased stock.
The second corporate action is M&A, which has different advantages to repurchases. First, M&A has an announcement date, enabling us to cleanly calculate short-and long-term returns. Moreover, the announcement date is relevant -the majority (72%) 2 of announced M&A is eventually completed.
In contrast, firms are only required to make an announcement when they first establish a repurchase program 3 , announced repurchases are often not completed (Stephens and Weisbach (1998)), and there are major problems with the standard data source used to approximate repurchase announcements, described later. Second, M&A is a much more significant event than an investment cut (or repurchase) -it is arguably the most transformative corporate decision that a firm can undertake -and so it is likely that at least a significant portion of long-run stock returns is attributable to the M&A. Indeed, prior literature (e.g. Agrawal, Jaffe, and Mandelker (1992)) uses long-run stock returns to assess the long-term value implications of M&A.
As an example of how a firm might use M&A to boost the short-term stock price with negative long-run consequences, Bazaarvoice acquired PowerReviews in June 2012, which led to its stock price soaring above $20. Bazaarvoice's officers and directors then sold $90 million of stock before the U.S. Department of Justice ("DoJ") commenced an antitrust lawsuit in January 2013, since PowerReviews was Bazaarvoice's closest competitor. The DoJ lawsuit forced Bazaarvoice to divest PowerReviews and led to its stock price falling below $7. In internal communications, Bazaarvoice executives stated that their motivation for the acquisition was "[e]limination of our primary competitor" to leave them with "literally, no other competitors." Thus, they likely knew that a DoJ lawsuit would be probable and that the long-term returns would be negative, but the acquisition announcement inflated the stock price in the short-term.
We study the relationship between vesting equity and both actual repurchases and M&A announcements over 2006-2015, a longer sample period than prior literature that allows us to study long-term returns. We find that a one standard deviation increase in vesting equity is associated with a 1.2% increase in a firm's likelihood of conducting a share repurchase in a given quarter (corresponding to an expected increase in shares repurchased of $1.5m for an average firm), controlling for other determinants of repurchase behavior and year-quarter fixed effects. This compares with the unconditional repurchase probability of 37.5%. When focusing on sizable repurchases, i.e. ones that exceed the sample mean, the increase is now 1.04% compared to an unconditional probability of 20%. These results are not driven by repurchases that result from investment cuts -instead, repurchases and investment cuts appear to be independent channels that a CEO may pursue to increase the stock price. We find similar results for M&A: a one standard deviation increase in vesting equity is associated with a 0.6% increase in a firm's likelihood announcing an M&A in a given quarter, compared with the unconditional probability of 15.8%.
Our main results are the short-and long-term returns to repurchases and M&A. Again, we find a consistent picture across both corporate events: vesting equity increases short-term returns but reduces long-term returns, consistent with it inducing the CEO to take myopic actions with negative long-term consequences. A one standard deviation increase in vesting equity is associated with an annualized 0.61% higher return over the two quarters surrounding a repurchase, but a 1.11% (0.75%) lower return during the first (second) year after the repurchase. The results are similar for M&A although the negative association with long-run returns persists for longer. A one standard deviation increase in vesting equity is associated with an annualized 1.47% higher return over the two quarters surrounding an M&A announcement, but a 0.81%, 0.35% (insignificant), 0.72%, and 0.62% lower return in the first, second, and third, and fourth subsequent years. This paper is related to three literatures. The first is on the effects of short-term equity incentives.
In addition to EFL, Edmans et al. (2017) show that CEOs reallocate news toward months in which their equity vests and away from adjacent months, and Gopalan, Huang, and Maharjan (2016) document that vesting equity leads to CEO departures. Ladika and Sautner (2016) find that the adoption of FAS 123R induced some firms to accelerate option vesting, which in turn led to a fall in investment; Jochem, Ladika, and Sautner (2016) show that accelerated vesting prompted CEO turnover. Our main contribution is to identify outcome variables (repurchases and M&A) whose long-run effects can be estimated, thus allowing us to demonstrate that short-term equity incentives may have negative long-term consequences.
While our main contribution is to study the long-term effects of short-term incentives, rather than to show the effect of vesting equity on different outcome variables, the outcome variables chosen are of independent interest as they relate the paper to two separate literatures. One is on the determinants and consequences of stock repurchases. Dittmar (2000) studies the effect of various characteristics on stock repurchases; we show that they are also determined by the CEO's contract horizon. Turning to the consequences, Ikenberry, Lakonishok, andVermaelen (1995, 2000) show that the long-term returns to stock repurchases are positive, suggesting the average repurchase creates value. In contrast, Almeida, Fos, and Kronlund (2016) show that repurchases that are motivated by the desire to beat earnings per share ("EPS") forecasts lead to reductions in employment and investment, and a fall in cash holdings (which could either be positive or negative for firm value). Whether a firm is motivated to beat EPS forecasts depends ex post on whether earnings would be below the forecast in the absence of the repurchase. We identify a determinant of repurchase activity that is predictable ex ante, and link it to long-run stock returns.
The third literature is on the determinants and consequences of M&A. Firms are more likely to engage in acquisitions if they have overconfident CEOs (Malmendier and Tate (2008)), young CEOs (Yim (2013)), less debt-based CEO compensation (Phan (2014)), and deviate from their target capital structure (Uysal (2013)). Turning to the consequences, the surveys of Jensen and Ruback (1983) and Andrade, Mitchell, and Stafford (2001) show that acquirers enjoy (modestly) positive short-term returns and negative long-term returns. Moreover, short-and/or long-term returns have been shown to be increasing in recent acquirer performance (Morck, Shleifer, and Vishny (1990)) and corporate governance (Masulis, Wang, and Xie (2007)), and decreasing in CEO overconfidence (Malmendier and Tate (2008)) and CEO debt-based pay (Phan (2014)). We show that CEO short-term incentives affect both the propensity to acquire and the short-and long-run returns to acquisitions.

Measuring short-term incentives
Our initial sample contains the entire 48,856 firm-CEO-years for which Equilar collects compensation data from 2006 to 2015. We closely follow the approach of EFL to calculate vesting equity, which is described in more detail in Appendix B. In short, this procedure involves three steps.
First, we use annual data from Equilar to infer the number of shares and options that vest, grant-bygrant, in a particular year. Second, we allocate this vesting equity to a particular quarter, since quarterly is the highest frequency available for actual repurchases. This requires the vesting date of equity, which we infer for options using their expiry date and estimate for stock using EFL's algorithm. Third, we calculate the effective value of quarterly vesting equity. Doing so requires the delta of each individual vesting option, which we are able to calculate since the first step yields grantby-grant vesting data. 4 The resulting measure reflects the dollar change in vesting equity for a 100% change in price, and we label it VESTING. We estimate VESTING for a sample of 150,914 firm-CEO-quarters, representing 6,122 unique firms and 9,623 unique CEOs.

Measuring stock returns to corporate actions
As discussed in the introduction, we link equity vesting to share repurchases and M&A, since we can assess their long-term value implications using long-run stock returns. 4 Prior to 2006, disclosure requirements do not allow us to infer vesting options on a grant-by-grant level.
We first measure a firm's actual repurchases in a given quarter. Historically, this task has been difficult for three reasons. First, firms were not required to disclose actual repurchases in their periodic filings prior to 2004. Second, while Thomson Reuters's Securities Data Company ("SDC") Platinum database collects repurchase announcements, these announcements are voluntary after a repurchase program is first established (which could be several years prior). Even when firms do announce their repurchases, they are not obligated to follow through with these announcements -they can subsequently choose to purchase more shares than announced, fewer shares than announced, or even none at all (Stephens and Weisbach (1998)). Third, although researchers have used several databases to approximate the actual amount of shares repurchased by a firm in a given year, such as SDC Platinum, the Compustat Annual files, and CRSP, each database has its unique challenges and the resulting proxies are often noisy. Banyi, Dyl, and Kahle (2008) find significant estimation errors in all of them.
We measure actual repurchases using Compustat Quarterly. This database takes advantage of the Securities and Exchange Commission's (SEC) enhanced disclosure requirements, which require public companies to report the number of common shares repurchased (CSHOPQ in Compustat Quarterly) as well as the average price paid for the shares repurchased (PRCRAQ) in their 10-Q filings for quarterly reporting for periods ending on or after March 15, 2004. We first define a binary variable REP to denote the existence of a share repurchase for a firm, which equals one if the firm reports either CSHOPQ or PRCRAQ in a given quarter, and zero otherwise. 5 We also calculate REP%, the value of the shares repurchased (CSHOPQ × PRCRAQ) as a percentage of market capitalization at the end of the prior quarter.
We collect data for all M&A announced between January 2006 and May 2016 from SDC Platinum. We define MA, a binary variable that equals one if a firm announced an M&A as an acquirer in a quarter, and zero otherwise.
To gauge the long-run value implications of share repurchases and M&A, we calculate the buyand-hold abnormal returns (BHAR) surrounding these events. We calculate BHAR also at the quarterly level, from quarter q-1 (the quarter prior to the event quarter q) to quarter q+16. We calculate a firm's quarterly BHAR by first geometrically compounding its three-month raw return and then subtracting the geometrically-compounded return on one of three benchmarks -the CRSP value-weighted index, Fama-French 49 industry portfolio, and Daniel, Grinblatt, Titman, and Wermers's (1997, DGTW) characteristic-based portfolio. The last two benchmarks control for industry-and firm-level factors that might account for a firm's stock returns.
For M&A, 72% of announced deals in our sample are eventually completed, and so an M&A announcement has real consequences. We thus also calculate CAR, the cumulative market-adjusted abnormal returns to the announcements of M&A, over [-1, +1], [-2, +2], and [-3, +3] (with day 0 being the announcement date), to capture the initial market reaction. We cannot similarly calculate the returns to share repurchase announcements for several reasons. First, firms are only required to make an announcement when they first establish a repurchase program; subsequent modifications to the program and actual repurchases do not need to be announced. Second, as Banyi, Dyl, and Kahle (2008) show, even for repurchases that are announced, data quality in SDC Platinum is poor. SDC's data coverage is far from comprehensive and systematically misses announced repurchases for low growth firms; on the other hand, it double counts other repurchases. Third, even for repurchase announcements that are accurately recorded, they are often not followed through and so are less relevant events than M&A announcements. It may take several years for the repurchase to be executed; Stephens and Weisbach (1998) study the three-year period after an announcement and find that the average repurchase is not completed. Nevertheless, we verify robustness to studying repurchase announcements in Section 5.

Controls
As standard controls, we include the CEO's incentives that might arise from his unvested equity holdings (UNVESTED), already-vested equity holdings (VESTED), salary (SALARY), and bonus (BONUS), to isolate the incentives provided by vesting equity rather than other components of a CEO's contract. We also include the CEO's age, tenure, and a new CEO indicator (AGE, TENURE, and NEWCEO). NEWCEO is measured for the year to which quarter q belongs, while UNVESTED, VESTED, SALARY, BONUS, AGE, and TENURE are measured for the year before.
We follow Huang and Thakor (2013) to construct additional controls used in the repurchase analysis. These controls include the natural logarithm of quarterly sales (SALES), market-to-book ratio (MB), the long-term debt-to-assets ratio (BKLEV), the operating and nonoperating return-onassets ratio (ROA and NROA) and recent stock returns (RET). These controls measure firm size, leverage, accounting performance (which affects excess capital) and stock performance (which affects undervaluation) -factors previously shown to affect repurchase behavior (e.g., Dittmar (2000), Jagannathan, Stephens, and Weisbach (2000), Guay and Harford (2000)). We measure these controls either over quarter q-1 or at the end of q-1.
The additional controls used in the M&A analysis are mainly taken from Uysal (2011). We first calculate the trailing one-year average market leverage ratio (MKLEV) prior to quarter q-1, which, as Uysal (2011) shows, is the primary driver of a firm's M&A decision. We also include SALES, MB, ROA, and RET to proxy for firm size and performance, MALIQ, the sum of M&A values in the firm's industry over a year to measure its industry M&A liquidity, as well as HFI, the Herfindahl index of the firm's industry to measure its product market concentration.

Sample and summary statistics
The sample that intersects vesting data with repurchase data and controls consists of 93,537 firm-CEO-quarters, and the sample that intersects vesting data with M&A and controls consists of 94,362 firm-CEO-quarters. Table 1 reports summary statistics. Comparable to EFL, vesting equity has a mean of $786,877 in our sample. In a given quarter, 37.5% of firms buy back stock and 15.8% announce at least one M&A. The average percentage of shares repurchased is 0.36% for all firms; this number becomes 0.95% if we limit the sample to firms that conducted repurchases.

Equity vesting and share repurchase
We assess whether vesting CEOs are more likely to engage in a share repurchase by running the following panel regression: REPq (REP%q)= α + βVESTINGq + γCONTROLS1q-1 + q. (1) The dependent variable is either the repurchase indicator REP or the repurchase amount REP%. The independent variables include VESTING measured for the CEO during quarter q, as well as the controls as discussed in Section 2.3. The sample is at the firm-CEO-quarter level, but we omit firm subscripts (and CEO subscripts if there are multiple CEOs for a firm in a quarter) for brevity. In all regressions henceforth, we cluster standard errors at the firm level. 6 Column (1) of Table 2 reports the regression results of estimating equation (1) with REP as the dependent variable using a probit model, which ensures that the predicted values of REP are bounded within [0, 1] and allows for heteroskedasticity. We include year-quarter fixed effects to control for time variation in share repurchases induced by common shocks, such as macroeconomic conditions. Vesting equity is positively associated with a firm's likelihood of conducting a share repurchase in a given quarter at the 1% level. Based on the reported marginal effect, a one standard deviation increase in VESTING is associated with 1.2% increase in the firm's likelihood of conducting repurchase in a quarter, which is modest compared to 37.5%, a firm's unconditional probability of repurchasing shares in a quarter in our sample.
The economic significance increases if we only consider sizable repurchases to begin with. If we redefine REP to equal one only when the percentage of shares repurchased exceeds the sample average of 0.36%, a one standard deviation increase in VESTING is associated with 1.04% increase in the firm's likelihood of repurchasing at least 0.36% of its shares in a quarter. This corresponds to 5.21% of the unconditional probability of engaging in such a repurchase in our sample (20%).

Column (2) reestimates equation (1) using a linear probability model (LPM): the coefficient on
REP is similar in magnitude to that reported in column (1) and remains significantly positive at the 1% level. Compared to a probit model, a LPM assumes a non-normal and homoscedastic error term and potentially gives unbounded predicted values of REP, but it allows us to estimate coefficients on certain binary covariants that are otherwise non-estimable under probit (Caudill (1987)). We exploit this feature of LPM and include firm fixed effects in column (3) to control for firm-level heterogeneity in the propensity to repurchase. The coefficient on REP remains significantly positive at the 1% level, albeit smaller in magnitude. Table 2 report the ordinary least squares (OLS) regression results of estimating equation (1) with REP% as the dependent variable. Again, we first include year-quarter fixed effects in column (4) and then add firm fixed effects in column (5). VESTING remains positive and significant at the 1% level. Based on the reported coefficient in column (4), a one standard deviation increase in VESTING is associated with a 0.03% increase in the amount of shares repurchased as a fraction of market capitalization (the average market capitalization in our sample is $5bn), or $1.54m. This compares to the sample mean of 0.36%.

Columns (4)-(5) of
Among the controls for the CEO's other equity incentives, UNVESTED is significantly positive in all five specifications and VESTED is significantly negative in three. The coefficients on these variables is difficult to interpret: the CEO's voluntary holdings of vested equity are endogenous. His holdings of unvested equity are also endogenous since they depend on recently-granted equity; moreover, unvested equity might mitigate or exacerbate myopia depending on when it vests in the short-term or long-term. Turning to other controls for CEO incentives and characteristics, repurchases are positively related to CEO salary and negatively related to CEO age. The coefficients on firm characteristics are generally consistent with prior literature -repurchases are more likely for firms that are large, low-value, less leveraged, more profitable, and recent stock market laggards.
The results in Table 2 do not control for investment, because EFL show that vesting equity leads to a reduction in investment. Thus, investment would be a "bad control", as it is affected by the independent variable of interest. However, it remains important to check whether our results are robust to controlling for investment. If repurchases are financed by investment cuts, the negative correlation between VESTING and repurchases could be due to repurchases simply proxying for investment cuts. Table 3 thus adds contemporaneous R&D and capital expenditure (both scaled by total assets) as additional controls. While repurchases are indeed strongly negatively correlated with capital expenditure and weakly negatively correlated with R&D, the coefficients on VESTING are almost unchanged compared to those reported in Table 2. 7 These results suggest that investment cuts and share repurchases are independent channels that a CEO may pursue to increase the stock price when his equity is vesting, rather than repurchases simply being financed by investment cuts.

Equity vesting and BHAR surrounding share repurchase
The prior section showed that vesting equity induces the CEO to increase stock repurchases. These repurchases could be myopic, if they are financed by cuts in value-enhancing investments not included in R&D or capital expenditure (e.g., organizational development, which is included in Selling, General, and Administrative expenses), or if they are of overvalued stock. Alternatively, they could be efficient, if they are financed by free cash that might otherwise be wasted, or if they are of undervalued stock. Under both hypotheses, short-term returns to repurchases should be positiveeven if a firm's stock is overvalued, repurchases may still boost the short-term stock price by (falsely) signaling undervaluation to the market. However, the two hypotheses have opposite predictions for long-run returns: if repurchases are myopic (efficient), long-run returns should be negative (positive).
As discussed in the introduction, this prediction does not require the assumption that future stock returns are caused by the repurchases.
We regress the BHAR surrounding repurchases on VESTING: This regression approach follows Chen, Harford, and Li (2007). The dependent variable, BHAR, is calculated at the quarterly level from quarter q-1, the quarter prior to the event quarter, to quarter q+16, 16 quarters after. We require a firm to be traded at least two years following quarter q to be included in the sample. We measure short-term returns by compounding BHAR over quarters q-1 and q for two reasons. First, stock returns in these two quarters will have the most direct effect on the CEO's payoff from equity sales induced by equity vesting in quarter q. Second, expanding the window into quarter q-1 helps capture market reaction if an announcement was made ahead of the actual repurchase. We measure long-run returns by annualizing BHAR over quarters q+1 to q+4, q+5 to q+8, q+9 to q+12, and q+13 to q+16, respectively. 8 We continue to include year-quarter fixed effects and firm fixed effects.
Columns (1)-(5) of Table 4 report the OLS regression results of estimating equation (2) with BHAR calculated over the short-term window and four long-run windows, respectively. In Panel A, BHAR is calculated relative to the returns on the CRSP value-weighted index. The coefficient on VESTING is positive and significant at the 5% level in column (1), which suggests that repurchases conducted by firms with larger amounts of equity vesting for their CEOs enjoy a higher return in the near term. A one standard deviation increase in VESTING is associated with 0.3% increase in BHAR over quarters q-1 to q, and 0.61% if annualized. However, the pattern quickly reverses as the coefficients on VESTING turn significantly negative in columns (2) and (3), both at the 1% level.
Firms with higher CEO vesting equity experience much lower returns over the two years following repurchases. A one standard deviation increase in VESTING is associated with a 1.11% (0.75%) decrease in BHAR during the first (second) year following the repurchase. 9 The coefficients become insignificant in the third and fourth year.
Panels B and C repeat the analyses in Panel A, but instead calculate BHAR relative to the returns on the Fama-French 49 industry portfolios and the returns on the DGTW characteristic-based portfolios, respectively. We observe a similar pattern: VESTING is positively related to BHAR over the two quarters surrounding repurchases but negatively related to BHAR over the next two years.
Overall, the results of Table 4 are more consistent with vesting equity inducing CEOs to undertake myopic repurchases that boost short-term returns at the expense of long-term value, rather than efficient repurchases that increase firm value in both the short-and long-run.

Equity vesting and M&A announcement
This section links vesting equity to another corporate action, M&A. Our hypothesis is that, similar to repurchases, vesting equity could induce a CEO to undertake M&A that boosts the short-term stock price at the expense of long-term returns. Unlike repurchases, the vast majority (72%) of announced M&A are completed, and so we test this hypothesis by linking vesting equity to M&A announcements.
We run the following panel regression: The dependent variable is the M&A indicator MA, and the independent variables include VESTING and the controls discussed in Section 2.3. As in the repurchase analyses, we build the sample at the firm-CEO-quarter level. Table 5 reports the regression results of estimating equation (3) using a probit model in column (1) and a LPM in columns (2)-(3). We include year-quarter fixed effects in all three columns, and firm fixed effects in the last column. Vesting equity is positively associated with a firm's likelihood of announcing an M&A in a given quarter at the 5% level or lower. Based on the reported marginal effect in column (1), a one standard deviation increase in VESTING is associated with a 0.6% increase in the firm's likelihood of announcing an M&A in a quarter, compared with the unconditional probability of 15.8%.
When firm fixed effects are included in column (3), the controls for other CEO incentives and CEO characteristics are all insignificant except for CEO's incentives from his unvested equity, which is significantly positive at the 10% level, and a new CEO indicator, which is significantly negative also at 10%. Turning to firm controls, market-to-book and the firm's accounting and stock performance are significantly positive. Market leverage is significantly negative, consistent with Uysal (2011).
Given the size of M&A, it is less likely that M&A (as opposed to repurchases) is financed by investment cuts. Nevertheless, we repeat the analysis in Table 5 controlling for contemporaneous R&D-to-assets and capital expenditures-to-assets. The results are reported in Table OA1 and remain robust.

Equity vesting and BHAR surrounding M&A announcement
We now evaluate the efficiency of vesting-induced M&A. As in the repurchase analyses, we similarly regress the BHAR surrounding M&A announcements on VESTING: Unlike repurchases, we do have the exact announcement dates for M&A so we calculate BHAR in the equation above, from quarter q-1, redefined as three months prior to the announcement date, to quarter q+16, 16 quarters after the announcement date. 10 Again, we require a firm to continue trading at least two years following quarter q and include year-quarter and firm fixed effects. Table 6 reports the regression results of estimating equation (4) with BHAR calculated relative to the returns on the CRSP value-weighted index, Fama-French 49 industry portfolios, and DGTW characteristic-based portfolios in Panels A, B, and C, respectively. All three panels demonstrate a similar pattern to Table 4: VESTING is positively related to short-term returns but negatively related to long-term returns. The one difference is that the negative relationship with long-term returns persists for up to four years, consistent with Agrawal, Jaffe, and Mandelker's (1992) finding of fiveyear negative long-term returns to M&A. Based on the coefficients reported in Panel A, a one standard deviation increase in VESTING is associated with an annualized 1.47% increase in BHAR over quarter q-1 to q. However, it is also associated with a 0.81%, 0.35% (insignificant), 0.72%, and 0.62% decrease in BHAR in the first, second, and third, and fourth years after the M&A, respectively.

Equity vesting and M&A announcement returns
While Table 6 studies stock returns in the two quarters around the M&A announcement, Table 7 hones in on the [-1, +1], [-2, +2] and [-3, +3] windows, to more precisely measure how M&A boosts the short-term stock price. We hypothesize a positive relationship between VESTING and CAR, i.e. vesting equity leads CEOs to announce deals that are perceived more positively by the market in the short-term. We run the following regression: As before, we control for the other components of CEO pay, age, tenure, and a new CEO dummy, as well as size and the market-to-book ratio due to size and value effects in stock returns. Consistent with our hypothesis, a CEO's vesting equity is positively related to his firm's M&A announcement returns. Based on the reported coefficients, a one standard deviation increase in VESTING is associated with 0.15% increase in three-day CAR, 0.18% increase in five-day CAR, and 0.20% increase in seven-day CAR. These results suggest that CEOs with high vesting equity undertake acquisitions that the market responds to positively in the short-term.

Robustness Checks
This section describes the results of additional robustness tests. The first set of tests verify robustness to alternative definitions of the dependent variables. Table OA2 studies the link between vesting equity and repurchase announcements. We do not use repurchase announcements in the core analyses primarily due to the data quality issues described in Section 2.2. In addition, a repurchase announcement is less meaningful (compared to an M&A announcement) since the repurchase may not eventually be completed. However, since repurchase announcements can increase the short-term stock price even if not eventually executed, a CEO with short-term concerns may have incentives to undertake them. The dependent variable is REPANN, an indicator for whether a firm announces a share repurchase program or actual share repurchase in a particular quarter. Under both probit and LPM specifications, VESTING is positive and significant at the 1% level. For example, a one standard deviation increase in VESTING is associated with 0.4% increase in a firm's likelihood of announcing a repurchase in a given quarter, compared with the unconditional probability of 4.3%. The economic significance is markedly higher than in Table 2, but we put less weight on these results given the data issues. Table OA3 studies robustness to alternative definitions of the M&A dependent variable. The first alternative is MANUM, the number of acquisitions announced in a particular quarter (while Table 5 used MA). Column (1) and (2), without and with firm fixed effects respectively, show that VESTING is positive and significant at the 5% level or better. The second alternative is MASUM, the aggregate value of all acquisitions made in a particular quarter, scaled by the acquirer's market capitalization at the end of the previous quarter. 11 Column (3) shows that VESTING is positive and significant at the 1% level without firm fixed effects. The significance level drops to 10% in column (4) with firm fixed effects, potentially due to the MASUM being underestimated because over half of the deals in our sample do not have their size recorded in SDC. Table OA4, Panels A and B, repeats the results of Tables 5 and 6 (respectively) only considering M&A announcements that are subsequently completed. Despite the smaller sample, the results are similar to including all M&A announcements. Table OA5 conducts the return analyses of Table 4 (for repurchases) and Table 6 (for M&A) studying CAR rather than BHAR. While BHAR geometrically compounds a stock's benchmarkunadjusted return and then subtracts the geometrically-compounded benchmark return, the CAR first calculates a stock's benchmark-adjusted monthly (or daily) returns and then arithmetically compounds them over several months. Conrad and Kaul (1993) argue that the BHAR method is more accurate for statistical reasons, hence using it in the main analyses, but here we verify robustness to CAR. The inferences are unchanged: both repurchases and M&A lead to significantly positive shortterm returns, but negative long-term returns over two years for repurchases and four years for M&A.
The final set of tables verifies robustness to alternative ways of calculating VESTING. One concern with VESTING is that an option's delta is increasing in the current stock price, which may be correlated with unobservable variables (such as growth opportunities) that also drive repurchase and M&A behavior. While this might seem to work against our repurchase results (since higher growth opportunities would encourage investment rather than repurchases), it may explain our M&A results (since a higher stock price may make it easier to stock-finance M&A, or obtain board approval for M&A). Table OA6 recalculates VESTING assuming that all options are at-the-money. This still allows option deltas to vary with their maturity date and the volatility of the underlying stock, but removes their dependence on the strike price.
A related concern is that the current stock price may affect VESTING through triggering vesting.
Our use of vesting equity is motivated by it being determined by equity grants made several years prior. While true for grants with time-based vesting, performance-based vesting is becoming more common, and Bettis et al. (2010) find that 46% of performance-based vesting provisions are contingent on stock price thresholds, twice as frequent as the next category. Table OA7 recalculates VESTING including only time-based vesting grants, and removes post-2006 grants labeled "performance-based," "contingent," or "accelerated;" post-2006 grants with unknown vesting schedule. Table OA8 addresses the concern that an option's delta depends on its time-to-maturity, but if CEOs exercise their options shortly after they vest, their effective horizons are shorter. We thus recalculate VESTING using options' intrinsic values: we assign a delta of one to all in-the-money options and zero to all out-of-the-money options, because only the former would be exercised immediately upon vesting. In Tables OA6-OA8, the inferences regarding both the frequency of and returns to repurchases and M&A are unchanged.

Conclusion
This paper shows that the impending vesting of equity may lead CEOs to take myopic actionsactions that boost the short-term stock price at the expense of long-term value. An increase in vesting equity is associated with a greater frequency of stock repurchases and M&A announcements, and higher short-term returns and lower long-term returns surrounding these events. These results provide suggestive evidence of the negative causal effects of short-term CEO incentives on long-term firm value, thus potentially justifying proposals to lengthen the vesting period of equity.
Note that, while we have provided evidence of the potential costs of short-term incentives, there may also be costs of lengthening vesting periods, as proposed by some academics and practitioners.
For example, longer vesting periods may subject the CEO to risk outside his control and lead to him demanding a risk premium, or avoiding value-creating risky projects as shown theoretically by Brisley (2006). Relatedly, the model of Laux (2012) shows that, if equity is forfeited upon dismissal, long vesting periods may encourage the CEO to take short-term actions that reduce the risk of being fired. Future research is needed to identify the costs of long-term incentives to provide further guidance on any reform.

Appendix A: Definition of variables
This appendix describes the calculation of variables used in the core analysis. Underlined variables refer to variable names within Compustat. t indexes the year to which quarter q belongs.

Outcome variables of interest REPq
An indicator variable that equals one if a firm reports either the number of shares repurchased (CSHOPQ) or average repurchase price (PRCRAQ) in quarter q, and zero otherwise.

REP% q
The value of shares repurchased in quarter q (CSHOPQ× PRCRAQ) as a percentage of market capitalization (CSHOQ× PRCCQ)) at the end of quarter q-1, and zero if no repurchase is conducted.

MA q
An indicator variable that equals one if a firm announced an M&A in quarter q, and zero otherwise.

BHARq-1 to q
A firm's buy-and-hold abnormal return (BHAR) over quarter q-1 and q, with quarter q being either the fiscal quarter in which a share repurchase occurred or the one quarter that follows an M&A announcement (with the first day of the quarter being the M&A announcement date). For repurchase events, BHAR is calculated as the firm's geometrically-compounded monthly raw returns minus a benchmark return geometrically compounded over the same period: either the CRSP value-weighted index, the Fama-French 49 industry portfolio (obtained from Kenneth French's data library), or the DGTW (1997) characteristic-based portfolio (obtained from Russell Wermers' website) BHAR and benchmark returns for M&A events are calculatedly similarly as those for repurchase events, but use daily returns rather than monthly returns. BHARq+1 to q+4, BHARq+5 to q+8 , BHARq+9 to q+12 , and BHARq+13 to q+16 are analogously calculated as a given firm's BHAR for quarter q+1 to q+4, q+5 to q+8, q+9 to q+12, and q+13 to q+16, respectively.

CARq
Cumulative market-adjusted abnormal announcement return surrounding an M&A announcement made by a firm during quarter q. It is calculated as the sum of the firm's daily abnormal returns over [-n, n] with the daily abnormal return being the firm's daily raw return minus the corresponding return on the CRSP value-weighted index, where day 0 is the announcement date and n = 1, 2, and 3 trading days. Natural logarithm of total sales of quarter q-1.

MB q-1
The ratio of market value of assets to book value of assets, calculated as [market capitalization plus book value of total debt (DLTTQ+DLCQ)] divided by total assets, both at the end of quarter q-1. BKLEV q-1 Long-term debt-to-asset ratio (DLTTQ)/AT of quarter q-1.

ROAq-1
Operating income (OIBDPQ) in quarter q-1 divided by the average of the total assets at the beginning and the end of quarter q-1. NROA q-1 Non-operating income (NIPIQ) in quarter q-1 by the average of the total assets at the beginning and the end of quarter q-1.

RETq-1
A firm's BHAR relative to the CRSP value-weighted index over quarter q-1. R&D q R&D (XRDQ) in quarter q divided by total assets at the end of quarter q-1, and set to zero if missing.

CAPXq
Capital expenditure (inferred from CAPXY) in quarter q divided by total assets at the end of quarter q-1, and set to zero if missing.

MKLEV q-1
Average quarterly market leverage over year t-1, calculated as book value of total debt divided by market value of total debt, where market value of total debt is the sum of book value of total debt, market capitalization, and preferred stock (PSTKQ) minus deferred taxes and investment tax credit (TXDITCQ).

MALIQ q-1
Industry M&A liquidity is the sum of acquisitions value for the year to which quarter q-1 belongs within three-digit SIC code divided by the total assets of all Compustat firms in the same three-digit SIC and year. HFI q-1 Herfindahl index, calculated as the sum of the squares of the market shares of the Compustat firms for the year to which quarter q-1 belongs in the same three-digit SIC industry. Market share is defined as the sales of a firm divided by the sum of sales in the firm's industry.

MV q-1
Natural logarithm of market capitalization at the end of quarter q-1.

Appendix B: Calculation of Vesting Equity
This Appendix describes our calculation of vesting equity, which also follows EFL. First, we retrieve a CEO's number of vesting shares in a given year using Equilar's variable "Shares Acquired on Vesting of Stock," which includes shares vested from restricted stock plans, restricted stock unit plans, and long-term incentive plans. We then infer a CEO's number of vesting options in the year, grant by grant, from his unvested options at the beginning and the end of the year as well as his newly awarded options during the year. Option grants are sorted using their strike price and expiry date.
Second, we convert vesting equity from an annual to quarterly basis by estimating the vesting date of equity. For options, this is simple. Options vest and expire on the anniversary of a grant (as assumed in the literature and as we verify in a random sample). For shares, there is no expiry date, and grant dates are only available for shares awarded after 2006 in Equilar, so we follow EFL's algorithm to assign them to a particular quarter. In the first step, a CEO's vesting shares in a given year are attributed to stock awards post 2006 for which we know the grant dates from Equilar. These include cliff-vesting grants, which vest at the end of the vesting period, and graded-vesting grants, which we assume to vest annually on a straight-line basis following Gopalan et al. (2014). In the second step, the remaining vesting shares are attributed to pre-2006 grants evenly across all the grant dates that we observe from post-2006 awards in Equilar.
For robustness, EFL propose two alternative algorithms to assign vesting shares. The first uses post-2006 cliff and graded 12 stock awards without performance provisions (as opposed to all post-2006 cliff and graded stock awards) in the first step. This addresses the concern that, for performance-vesting equity, the grant date anniversaries may not be a good guide to the vesting date. The second algorithm similarly uses post-2006 nonperformance-vesting cliff and graded stock awards in the first step, but the second step uses only grant dates for performance-vesting stock -since non-performance-vesting stock was used in the first step, so the remaining unmatched shares are unlikely from this pool. Our results are unchanged under either alternative algorithm.   (1), the marginal effect for VESTING is displayed below the standard errors. *** ( ** ) ( * ) indicates significance at the 1% (5%) (10%) two-tailed level, respectively.  This table presents the OLS regression results on the relationship between buy-and-hold abnormal return (BHAR) over the period from one quarter prior to the quarter in which a share repurchase occurred to four years after the repurchase quarter and the CEO's vesting equity. BHAR is calculated over the value-weighted market index in Panel A, the Fama-French industry portfolio in Panel B, and the DGTW benchmark portfolio in Panel C. Variable definitions are in Appendix A. VESTING is in billions. Standard errors are in parentheses, clustered by firm. *** ( ** ) ( * ) indicates significance at the 1% (5%) (10%) two-tailed level, respectively.  (1), the marginal effect for VESTING is displayed below the standard errors. *** ( ** ) ( * ) indicates significance at the 1% (5%) (10%) two-tailed level, respectively. VESTING is in billions. Standard errors are in parentheses, clustered by firm. *** ( ** ) ( * ) indicates significance at the 1% (5%) (10%) two-tailed level, respectively.   (1), the marginal effect for VESTING is displayed below the standard errors. *** ( ** ) ( * ) indicates significance at the 1% (5%) (10%) two-tailed level, respectively.  (1), the marginal effect for VESTING is displayed below the standard errors. *** ( ** ) ( * ) indicates significance at the 1% (5%) (10%) two-tailed level, respectively.   Panel A presents the OLS regression results on the relationship between long-term cumulative market-adjusted abnormal return (CAR) over the period from one quarter prior to the quarter in which a share repurchase occurred to four years after the repurchase quarter and the CEO's vesting equity. Panel B presents the OLS regression results on the relationship between long-term CAR over the period from one quarter prior to the M&A announcement date to four years after the announcement date and the CEO's vesting equity. CAR is calculated over the value-weighted market index in both panels. Variable definitions are in Appendix A and Table OA9. VESTING is in billions. Standard errors are in parentheses, clustered by firm. *** ( ** ) ( * ) indicates significance at the 1% (5%) (10%) two-tailed level, respectively. Panel A (C) presents the regression results on the relationship between share repurchases (M&A announcements) and the CEO's vesting equity. Panel B (D) presents the regression results on the relationship between BHAR over the period from one quarter prior to the quarter in which a share repurchase occurred to four years after the repurchase quarter (from one quarter prior to the M&A announcement date to four years after the announcement date) and the CEO's vesting equity. Variable definitions are in Appendix A and Table  OA9. All are estimated using an OLS model unless otherwise specified. VESTING is in billions. BHAR is calculated over the value-weighted market index. Standard errors are in parentheses, clustered by firm. *** ( ** ) ( * ) indicates significance at the 1% (5%) (10%) two-tailed level, respectively. Panel A (C) presents the regression results on the relationship between share repurchases (M&A announcements) and the CEO's vesting equity. Panel B (D) presents the regression results on the relationship between BHAR over the period from one quarter prior to the quarter in which a share repurchase occurred to four years after the repurchase quarter (from one quarter prior to the M&A announcement date to four years after the announcement date) and the CEO's vesting equity. Variable definitions are in Appendix A and Table  OA9. All are estimated using an OLS model unless otherwise specified. VESTING is in billions. BHAR is calculated over the value-weighted market index. Standard errors are in parentheses, clustered by firm. *** ( ** ) ( * ) indicates significance at the 1% (5%) (10%) two-tailed level, respectively. Panel A (C) presents the regression results on the relationship between share repurchases (M&A announcements) and the CEO's vesting equity. Panel B (D) presents the regression results on the relationship between BHAR over the period from one quarter prior to the quarter in which a share repurchase occurred to four years after the repurchase quarter (from one quarter prior to the M&A announcement date to four years after the announcement date) and the CEO's vesting equity. Variable definitions are in Appendix A and Table  OA9. All are estimated using an OLS model unless otherwise specified. VESTING is in billions. BHAR is calculated over the value-weighted market index. Standard errors are in parentheses, clustered by firm. *** ( ** ) ( * ) indicates significance at the 1% (5%) (10%) two-tailed level, respectively.

Table OA9: Definition of variables used in the Online Appendix
This table describes the calculation of variables used only in this online appendix. The variables used also in the core analysis are described in Appendix A of the paper.

REPANN q
An indicator variable that equals one if a firm announced either the establishment of a new share repurchase program or actual repurchase(s) under an existing repurchase program in quarter q as captured by the SDC Platinum, and zero otherwise.

MANUM q
The number of M&A that a firm announced in quarter q, and zero if none was announced.

MASUMq
The sum of deal size for all M&A that a firm announced in quarter q, as a percentage of market capitalization at the end of quarter q-1, and zero if none was announced. We delete a firm-quarter if a firm announces at least one M&A in a quarter but none of the M&A has transaction size recorded in the SDC Platinum. CARq-1 to q A firm's cumulative market-adjusted abnormal return over quarter q-1 and q, with quarter q being either the fiscal quarter in which a share repurchase occurred or onequarter time that follows an M&A announcement (with the first day of the quarter being the M&A announcement date). For repurchase events, it is calculated as the sum of the firm's monthly abnormal returns over the two quarters with the monthly abnormal return being the firm's monthly raw return minus the corresponding return on the CRSP value-weighted index. For M&A events, it is calculated as the sum of the firm's daily abnormal returns over the two quarters with the daily abnormal return being the firm's daily raw return minus the corresponding return on the CRSP valueweighted index. CARq+1 to q+4, CARq+5 to q+8 , CARq+9 to q+12 , and CARq+13 to q+16 are analogously calculated as a given firm's CAR for quarter q+1 to q+4, q+5 to q+8, q+9 to q+12, and q+13 to q+16, respectively. VESTING_ATM q Similar to VESTING q , except that all options are assumed to be at the money. VESTING_TBq Similar to VESTINGq, except that it includes only post-2006 time-based vesting grants without performance provisions (i.e., we remove post-2006 grants labeled "retirement," "performance-based," "contingent," or "accelerated," and post-2006 grants with unknown vesting schedule). VESTING_INTq Similar to VESTINGq, except that options' deltas are replaced with their intrinsic values, i.e., delta is set to one for all in-the-money options and is set to zero for all outof-the-money options.