The effect of exogenous firm mispricing on forced CEO turnover
Are CEOs likely to be fired when facing bad luck?
MSc Thesis, K.V. Maassen (student number 5776546) Supervisor: Assistant Professor of Finance F.S. Peters, PhD MSc Business Economics, Track Finance
1
Statement of Originality
This document is written by Student Kees Volkan Maassen who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
2 Abstract
Classic contracting models suggest that performance evaluation of CEOs should not be linked to performance shocks beyond the CEOs control. Jenter & Kanaan (2010) find evidence that CEOs are fired after bad performance caused by exogenous industry and market shocks. This thesis tests whether corporate boards incorporate exogenous shocks to firm value in their decision to retain a CEO. A unique data set of 1515 turnovers consisting of 331 forced turnovers is used. No evidence is found that a CEO turnover is more likely when stock price performance is bad due to pure mispricing. It seems that corporate boards do filter out these idiosyncratic shocks. As an exogenous shock to firm performance we use price pressure to the stock price caused by mutual fund flows. This measure was initially used by Coval and Stafford (2007) and Edmans et al. (2010).
Keywords: CEO Turnover, Performance Evaluation, Corporate Governance, CEO performance.
3
1.
Introduction
The topic of firm performance and the evaluation of CEOs remains a topic of continued interest in the field of corporate governance. The decision to retain or fire a CEO after bad firm performance is a key task of the corporate board and a strong signal to the share- and stakeholders of the firm. Although classic contracting theory predicts that CEO evaluation by the corporate board should not be based on factors which are beyond the CEO’s control (Holmström, 1979), current research finds that CEO evaluation is partially based on performance shocks that are not related to CEO ability or characteristics. Bertrand and Mullainathan (2001) found that CEO remuneration is based on factors that cannot be assigned to CEO ability or firm-specific factors (the authors use changes in the global oil price as a shock). Unlike what standard economic theory would predict CEOs are often rewarded for luck, for factors beyond the firm’s and thus CEO’s control. Jenter & Kanaan (2010) for example find that CEOs are significantly more likely to be fired after negative performance shocks to their industry or the market in general.
In this thesis we use an exogenous shock to firm performance which is also outside the CEO’s and firm’s control, namely pure mispricing due to mutual fund outflow. This measure holds less economic relevancy to firm fundamentals than peer group performance. There is no literature based on standard economic theory about whether the decision of a corporate board to take these idiosyncratic performance shocks into account when evaluating the position of a CEO is economically efficient. A temporary mispricing shock that causes stocks prices to deviate from their fundamental value is introduced, which is initially constructed and used by Edmans, Goldstein and Jiang (2010) to research the impact of firm prices on takeover probability. They use price pressure caused by mutual fund flows as a shock to the firm’s stock price. The price pressure does not hold any information, but is by caused mutual funds that experience capital flows when investors invest in or retain capital from a mutual fund. (pp. 4). Mutual funds have to sell or buy shares it holds proportionally to increase or decrease the size of the portfolio. This shock is exogenous to firm performance, as Edmans, Goldstein and Jiang (2010) argue it is unlikely that the fund decisions to invest or divest mutual fund shares directly correlate with the takeover prospects of individual firms. This is supported by a low correlation between the measure and the business cycle1.
The question whether boards do incorporate these shocks in their decision framework on CEO retention may have implications for our view of the effectiveness of corporate governance. Specifically,
4 the purpose of this thesis is to examine the following question: “Do corporate boards misattribute temporary and exogenous shocks to firm performance to CEOs, and do boards incorporate these shocks in their decision to fire or retain a CEO?” Does a positive relationship between forced CEO turnover and shocks to firm performance exist or, in other words, does a CEO’s position depend on a shock that can just be seen as “bad or good luck”?
Recent research is performed by Jenter and Kanaan (2010) and Cai and Wei (2010). Both papers use an exogenous shock to test the effect on respectively CEO turnover or CEO remuneration. Jenter and Kanaan find evidence to support the hypothesis that shocks to peer group performance have a positive effect on CEO turnover, because it seems that boards do not behave economically efficient; they misattribute changes in firm value because of exogenous performance to the CEO. Cai and Wei (2010) test an exogenous shock caused by mutual fund flow on the stock performance and find evidence that boards do change their sensitivity to accounting fundamentals when the stock price deviates from its fundamental value due to mispricing. Boards increase the weight of accounting fundamentals when stock price is affected by an exogenous shock.
In this thesis we test the hypothesis that corporate boards systemically misattribute bad stock performance due to pure mispricing to CEOs. Mainly, because boards do not observe bad performance due to mispricing as easily as bad peer performance and consequently the probability of misattribution is higher.
We add to the body of research because almost no literature exists which looks into corporate board decisions and exogenous shocks beyond the firm’s control to firm performance. It differs from Cai and Wei (2010) and others in the following ways; first we use a different specification model which completely filters price pressure of mutual fund flow, by performing a two-stage least square regression; Second, forced CEO turnover is a significantly more observable and discrete event than variations in CEO pay and bonus. Oppositely, a CEO can be retained or forced to resign, it is a strong and costly signal made by the board, and will be scrutinized by shareholders. Third, a unique data set is composed and collected on forced CEO turnovers. Different data sets of turnovers (Jenter & Kanaan, Peters & Wagner, Eisfeldt and Kuhnen, Nini and Taylor) are combined and at least double collected. Disagreements between datasets are sorted out using a classification scheme. Unlike other existing data sets we account for survivorship bias and we do not automatically classify turnovers as voluntarily when the CEO stays as a chairman of the board.
The empirical methodology is based on the work of Jenter and Kanaan (2011). A two-stage least square (2sls) regression is used to assign between the random exogenous shocks to firm performance and the
5 fundamental firm performance, CEO ability and other shocks. The specification is modified to include price pressure instead of peer performance.
Performing estimation over 331 forced turnovers and 1500 turnovers we find no conclusive evidence that price pressure caused by mutual fund flow is misattributed as an economically efficient response of the corporate board to decide whether to fire a CEO from his or her position at the firm. The findings are in agreement with Cai and Wei (2010), who found that boards rely less on stock performance in setting CEO pay when the mutual fund flow driven price pressure occurs.
This thesis also relates on corporate governance at a broader sense. The result that CEO systematically misattributes bad firm performance due to luck supports the statement that corporate governance is still in development. Misattribution is highly undetectable and a potential source of erroneous decisions made by boards. Retaining or firing a CEO is a strong and potentially costly signal to share and stakeholders of the firm and so it must be conducted in an effective framework.
The rest of the thesis is organized as follows. The next section briefly explores the hypothesis and related literature. Section 3 describes the model and the regression specification to test the hypothesis. Section 4 provides a description of the data and summary statistics. The empirical results and some robustness checks are provided in section 5, and section 6 concludes and provides some discussion for further research.
2. Hypothesis development
The relation between CEO evaluation and firm performance has been researched over time. The pioneering study of Coughlan and Schmidt (1985), which uses stock price performance and later studies such as Warnets et al. (1988) and Weisbach (1988), all find a positive relationship between a firm stock price performance and boards decisions to fire or retain the top management of a firm. Later literature such as Morck, Sheifner and Vishy (1989), Barro and Barro (1990) and Gibbons and Murphy (1990) find evidence that there is no relationship between exogenous shocks, such as peer group performance and CEO turnover, and that, according to classic contracting theory, boards filter out these performance shocks which are beyond CEO control.
Recent research by Jenter and Kanaan (2010) find that in contrast to the prior research boards do not filter out industry shocks from firm performance in evaluating the current CEO’s positions and performance. They suggest that there are three main reasons why CEOs are fired when their peer group is not doing well. First, CEOs may be evaluated on peer group performance if a CEO’s actions affect peer
6 performance; this especially happens in oligopolistic industries. To test this hypothesis, they restrict the sample to small firms, because smaller firms are likely to be in an oligopolistic industry. Subsequently, these firms are unlikely to affect market equilibrium in their respective market. CEO’s are not able to exercise power on exogenous mispricing and Jenter and Kanaan find no evidence that they affect industry performance. No economically efficient reason for the corporate board exists to evaluate a CEO based on this. Second, boards receive more information about their CEOs in times of bad peer performance.Jenter and Kanaan find evidence that peer performance increases the probability of a CEO dismissal of an outperforming firm but has significantly more effect on CEO of underperforming firms. Boards do make an economically efficient decision if underperformance of a firm (and thus a CEO) is less revealing about CEO ability. They note that this means that boards do not monitor CEO ability and performance well when peer performance and firm performance is high. In contrast no evidence is found that peer performance has a smaller effect on CEOs with longer tenure. CEOs who sit on their position for a long period would already have proven their ability in different market performances. Finally, peer group performance may affect CEO turnover because boards do misattribute exogenous performance to the CEO. Jenter and Kanaan find evidence that forced CEO turnover of underperforming firms are occurring more frequently in (recessions than in booms), while no relationship exist between a CEO position of an outperforming firm and peer performance. Underperforming CEOs are less able to defend against attribution in bad times, but they will be able to hide behind good industry and market performance in booms. Boards are better able to evaluate a CEO’s performance against an easily observed benchmark such as the S&P 500 and the largest firms in their industry than more hidden exogenous performance components, such as in this paper performance shocks caused by mutual fund flow.
Jenter and Kanaan dismiss the first hypothesis quickly, but they find results which are broadly consistent with the latter two hypotheses. They find it surprising that no significant effects of peer performance on CEOs with different tenures and between CEOs with more or less power exists. They notice that these findings have consequences on the correct design of CEO turnover studies. Given that CEO turnover are determined jointly by firm specific industry and market performance these regressions suffer from omitted variable basis, unlike others they use a 2sls regression model.
Exogenous systemic shocks which reflect changes in industry conditions or in shocks to firm fundamentals can signal an economically efficient message to the board in specific situations. Two papers introduce a frame work to explain that CEO evaluation can be economically efficient linked to systemic shocks and not only to firm-specific performance. Eisfeldt and Kuhnen (2011) develop a competitive assignment model which is based on two ideas; first, if industry conditions affect the outside of options of matched firms and managers; industry shocks will also drive turnover. Second, that any fixed firing cost
7 leads to a threshold rule for termination at the firm level. Also Oyer (2004) suggests that due the cost to adjust the terms of an employee’s compensation scheme are relatively fixed, this implies variability in an employee’s reservation utility. Ultimately, the firm wants to transfer risk to the worker as a means of insuring participation during various states of the economy.
Most related literature to our paper is performed by Cai and Wei (2010). They use the same exogenous shock to peer performance to show that corporate boards rely more on accounting performance than stock performance to decide on CEO bonus pay when the stock price is temporarily less informative due to fund flow pressure. They find that boards pay attention to when a stock price deviates from its fundamental value due to exogenous price shocks and that they change CEO bonus pay according to. Consequently, boards reduce the relation between CEO bonus pay to stock performance and base their decision more on accounting fundamentals. Conclusively, firms that face price pressure do pay their CEOs more cash and give them less equity based compensation. Also boards are likely to lower the sensitivity of CEO pay in degree to overvaluation of the stock and increase the pay to performance when stock is underpriced. Boards do understand firm fundamentals and the monitoring is sufficiently active to find a balance between price pressure and the equity-based component of CEO compensation
Bushman et al. (2006), Jayaraman and Milbourn (2011) show that stock prices reflecting firm fundamentals better are also better reflecting managerial efforts, and consequently the board can depend more on the stock value when setting equity based performance remuneration and the question to retain a CEO.
Taking into account the findings of Jenter and Kanaan, and Cai and Wei we test the hypothesis that boards do incorporate price pressure caused by mutual fund flow to their evaluation of CEO performance. Only when mutual fund flow coincides with specific or industry specific firm performance can it be economically efficient to incorporate it into a decision framework. The oligopolistic hypothesis does not hold, because we assume that mutual fund flow is not related to industry or firm concentrations, It shows that boards can response economically efficient on exogenous shocks outside firm and CEO control. Stock performance shocks due to pure mispricing is more difficult to observe than specific industry performance (for example many analysts publish research on specific industries), so this subsequently lowers the probability that performance due to pure mispricing has any effect. Oppositely, finding evidence to support the hypothesis that corporate boards systemically misattributes bad performance due to pure mispricing to CEOs is more likely. Jenter and Kanaan specifically show situations how this misattribution can be shown. In our case, to show that mutual fund flow has an effect on turnover is already sufficient. The reason is that boards do not observe bad performance due to mispricing as easily as bad peer performance, and therefore the probability of misattribution is higher.
8 The hypothesis that boards receive stronger informative signals about CEO ability when firms independently perform badly due to pure mispricing is more unlikely than in the case of bad peer performance but cannot be dismissed yet.
3. Empirical Methodology
To test our prediction that transitory mispricing of the firm does not have an effect on the probability of CEO turnover, a model by Jenter and Kanaan (2010) is used. They use an empirical model derived from Holmström (1982) and Gibbons and Murphy (1990) to test that peer performance has no predictive power on the probability of forced CEO turnover. In this thesis a similar model is introduced to test whether a exogenous shock to firm performance has predictive power on forced CEO turnover; whether non-relevant shocks affect managerial decisions of the corporate board. They use a two-stage least square (2sls) regression to decompose firm performance (a firm’s stock return is used as a measure for firm performance) into a systemic component, which is peer group performance and a firm-specific part which reflects components such as CEO ability, accounting fundamentals and other firm-specific characteristics. In the second stage the probability of forced CEO turnover is predicted by regressing the estimated peer group performance and the estimated residual, which reflects firm-specific components. Jenter and Kanaan note that this 2sls procedure accomplishes the same as an instrumental regression in which peer group performance serves as an instrument to firm performance, in this case an IV regression is not used because the residuals in the first-stage are also used in the second stage, which also captures the possible reverse causality.
This procedure tests a pure-form hypothesis that corporate boards completely filter peer performance, aside from firm stock return performance. This is in contrast to Cai & Wei (2000.), who tests a weak form hypothesis but can find that an estimated effect can be relatively small, because boards do not distinguish between peer and firm-specific performance (pg. 9)
We replace peer group performance for price pressure driven by mutual fund flow. A 2sls procedure is used, so the effect of stock return on forced CEO probability was effectively assigned to the mutual fund flow and to a part of CEO and firm-specific characteristics and other non-relevant shocks. Mutual fund flow acts as an instrument because it is exogenous to firm performance and it does not have an economic relationship with the probability of forced CEO turnover. Avoiding a simultaneity bias is not relevant, because the effect of forced CEO turnovers on firm performance ends up in the residuals and not in the estimated mutual fund flow.
9 (i: First stage)
(ii: Second stage)
Where
Note that before. Now it is
is the estimated exogenous component of mispricing of the stock and thus not attributable to managerial actions or the firm itself. And is the estimated firm-specific component and components which are not related to our mutual fund flow price pressure but do not hold any relevancy to the firm’s performance.
The first stage is estimated by using a simple regression, the second stage is estimated by using a probit regression. Following Jenter & Kanaan (2010) a common price pressure beta is estimated for all firms which avoid an estimation error in the second stage. Standard errors are adjusted by using Stata’s built in linear probability model.
Additionally, also a weak form of the hypothesis based on Cai and Wei is estimated. The specification2 is an OLS model in which they do not decompose perform into firm-specific performance, it does not take into account what exactly the relation holds between the price pressure and the firm specific performance and whether corporate boards do distinguish between mutual fund flow price pressure and firm-specific performance when deciding to retain a CEO or not.
4. Data
4.1 Voluntarily and forced CEO turnover
CEO turnover is obtained from the S&P ExecueComp database from 1992 through 2009. The ExecueComp database contains CEO characteristics and compensation for top executives of all public firms in the U.S. A CEO turnover is recognized as such when a firm’s CEO identified in ExecueComp changes. News reports and SEC announcements are obtained from Lexis-Nexis which are used to examine the reason why a CEO turnover occurred.
2 The weak form is specified like this
10 The classification of a turnover as voluntary or as forced is based on Parrino (1997) and Jenter and Kanaan (2011). Turnovers are classified as forced if it is explicitly stated in news reports or company press announcements that the CEO was ‘’forced out’’, ‘’fired’’, ‘’ousted’’, or retires or resigns due to ‘’policy differences’’, ‘’pressure’’ or similar descriptions. In case a CEO is younger than 60 years old, the reason to resign is further reviewed. A turnover is classified as forced if the news articles or press releases do not report that the reason of departure is death, poor health or the acceptance of another position. Additionally, a turnover is classified when the departure appears to be unexpected, in case when the date has not been announced at least 6 months before the turnover. But turnovers are re-classified as voluntarily when the CEO takes a comparable position in another company soon after his departure (the same or the following fiscal year) as well as when press reports convincingly explain that the turnover was due to previously undisclosed personal or business reasons that are unrelated to the firm’s activities. Table 1 provides an overview with these strict criteria whether to determine a turnover as forced or voluntary.
A distinction between this turnover dataset and other datasets is we correct for survivorship bias, firms which are dropped from ExecueComp are included in the analysis. Turnovers in the case of a merger, leverage buy-out or spin-offs are always classified as voluntarily. In other cases, for example when firms go into bankruptcy protection (Chapter 11 or Chapter 12) or index exclusion it is determined whether a turnover occurred (in a reasonable time window of approximately 2 or 3 months) and how to classify them.
Classification of CEO turnover is difficult since CEOs are almost never openly fired from their positions, even with an extensive classification scheme the distinction between a voluntarily or forced departure can be unclear and ambiguous. A CEO departure of a company with poor performance, even though the CEO is applauded loudly and has a long-track record of ‘success’ can still be classified as forced. A departure of an interim CEO which does not fulfil his contract agreement and is succeeded by another short-term non-permanent CEO can also be classified as forced. In cases of ambiguity when still strictly following the classification scheme common sense can thus be helpful.
Unlike Jenter and Kanaan (2011) forced turnovers are not mechanically re-classified as voluntary if the CEO remains on the board as for example as a chairman. Beside that the CEO turnover dataset is unique because it captures a longer timespan (1993-2009, additionally all turnovers in existing datasets (Jenter & Kanaan, Peters & Wagner, Eisfeldt & Kuhnen, Nini and Taylor) at least have been double collected. Table 2 provides an overview of the different data sources the turnover dataset consists of. Any disagreements in classification whether a CEO turnover is forced or voluntary are reclassified by
11 researching news articles, press releases and SEC announcements published in a three year window around the particular CEO turnover, obtained via Lexis-Nexis. In case double collected turnover data is not yet available, e.g. for the time period 2007-2009, the classification of a CEO turnover is performed from scratch and later cross-checked by the data provided by Taylor (2010) or Nini (2012). As well, disagreements between Jenter and Kanaan (2012) and Eisfeldt and Kuhnen (2011) are clarified following the classification criteria.
4.2 Price pressure driven by mutual fund flow
To test the hypothesis that the corporate board misattributes an exogenous temporary shock to firm performance as an economically efficient signal and thus has an effect on the probability of forced CEO turnover a variable must be found that de facto acts as an instrument. It must hold a relationship with the changes in stock price of a firm, which is easily observable and which shareholders take in to account. But it should not have an economic relationship with firm-specific and CEO specific characteristics such as fundamentals, industry performance and CEO ability. Actually, reverse causality is not a problem, because the both residuals are incorporated in the second-stage. The model uses a shock exogenous to firm-specific characteristics (firm fundamentals) and CEO ability that was first introduced by Coval and Stafford (2007). Edmans, Goldstein and Jiang (2010) use a modified version. Price pressure is caused by investors’ capital flows to mutual funds. Consequently, mutual funds are obliged to sell or buy in proportion to current holdings. In- and outflows of at least 5% are taken into account, because only extreme outflows will induce enough price pressure on the firm stock. Smaller flows can be absorbed by internal cash and liquidity providers. Unlike Coval and Stafford (2007), Edmans, Goldstein and Jiang (2010, pg. 24) are using a mutual fund flow that not actually consists of actual trades, but which is based on hypothetical orders projected from their previous disclosed portfolio. It takes into account the selling and buying of stock which is based on an informational driven part, but purely the mechanical selling and buying caused by investors’ flows. In this paper we use the same mutual fund flow variable as used by Edmans, Goldstein and Jiang (2010)3, but it is slightly modified to incorporate not only outflows but also inflows. In addition, this measure is transitory; it has no long term effect on the stock price.
Mutual fund flows are calculated by obtaining total net assets and the return from each mutual fund from the CRSP Mutual Fund, Monthly Returns dataset. The result is monthly fund flows of each fund j for month t, as the change in total net assets during the month adjusted for returns, thereby assuming that the flows occur at the end of the month:
12
(1)
Where are the total net assets of fund j at the end of month t, and is the monthly return of fund j during month t.
Fund returns and total net assets are obtained from CRSP Mutual Funds, monthly returns and Asset
Values Database. Funds can have different share classes but they represent the same underlying portfolio
and therefore do not differ in their underlying trades, all share classes are combined, by using a matching table, which connect the different class shared (CRSP handle_fund_no) to the underlying portfolios, (CRSP port handle port_no which is a subset from fund_no). As Edmans, Goldstein and Jiang (2010) only flows of at least 5% of total assets are considered, because only extreme flows are likely to have an impact on pricing, whereas moderate flow shocks could be absorbed by internal cash of the mutual fund or external liquidity providers. Equation 2 shows the mutual fund flow and equation 3 shows price pressure. The summation of only flows >= 5% leads to pressure.
(2)
(3)
Where I is the stock, j is the fund, t is the month, F is the dollar inflow into fund j, s is security I’s percentage of total net assets of fund j, and VOL is the trading volume of stock i. So is the the dollar outflow and scaled by the stock’s dollar trading volume.
Sias, Starks and Titman (2006) and Coval and Stafford (2007) note that order imbalances affect stock prices because the impact of a given outflow on prices will depend on the stock’s liquidity, the dollar outflow is scaled by stocks volume, unlike Coval and Stafford (2007) which focuses on fire sales and use trades that are driven by information. The stock’s trading volume consists of the change of fund’s existing positions that is mechanically induced by investor flows. Such flows are in turn unlikely to be driven by information of an individual firm held by the fund, since these views would be expressed through direct trading of the stock and not by changing the total portfolio. Following Coval and Stafford (2007) an event is a firm-year pair where the mutual fund flow falls below the 10th/90th percentile value of the full sample. A price event is defined as a firm-month pair in which flows exceed or are equal to |5%|.
13 Where the θth percentile is respectively the 10th and 90th of the full sample.
Following Coval and Stafford (2007), an event of extreme price pressure is defined as a firm-year where mutual fund flow falls above the 90th percentile or under the 10th percentile of the sample4.
4.3 Summary Statistics
Table 1 presents the summary statistics of the mutual fund flow induced price pressure for yearly and monthly periods. For the yearly interval the average mutual fund flow induced price pressure is -4.539 with a standard deviation of 65.32 and for the monthly statistic the average is -0.30 with a standard deviation of 22.99. Table 2, Panel A presents the frequency of voluntarily and forced CEO turnovers. The sample has 2361 distinctive U.S. Firms with 15.675 firm-year observations from 2001 to 2009. The data set holds 1515 (100%) turnovers of which 331 (22%) are classified as forced and 1184 (78%) as voluntarily. Turnovers of CEO older than 60 years which are voluntary amounts to 288 (19%) and 72(22%) of them are Panel B presents the firm performance as the yearly stock return, the average return is 14% of the firms in the sample and the standard deviation is 0.61. Panel C shows that the data set consists of 3726 distinctive CEOs. A CEOs tenure that was forced to resign is on average 4.0 years. As expected a CEO that was not forced to resign holds his position significantly longer on average 13.8 years. Figure 2 shows the total CEO turnovers split in voluntarily and forced over time. The ratio forced/voluntarily stays relatively constant around 20% and the number of total turnover does not deviate substantially over the years.
5. Results
5.1 Mutual fund flow price pressure and firm performance
Figure 2 presents the price pressure on stock performance that is caused by mutual fund flow and the business cycle. It shows that the fund flow is reasonably independent from the business cycle. Mutual fund flow pressure and the S&P500 index hold a Pierson’s correlation of 0.019. Therefore, it can be assumed that mutual fund flow pressure is quite exogenous and independent from the business cycle.
4 We aggregate mutual fund flow over the monthly periods in the following way: if in the majority of the months of
a certain year significant (of |5%|) mutual fund flow exists, then subsequently the whole year is categorized as experiencing pressure.
14 Table 6 presents an OLS specification with extreme mutual fund flow price pressure as the regressor and stock returns as a dependant variable; the coefficient is significantly at zero.On a firm-specific level mutual fund flow affects the price discovery process, and on average mutual fund flows increases the stock price with 0.7%, when adjusted for time fixed effects. Similar to Cai and Wei (2012), Edmans, Goldstein and Jiang (2010) also find that stock prices do react on mutual fund flow. The effect of mutual fund flow on firm value and that it affects the informativeness of the stock price is important, boards must be able to observe these shocks in the stock return. Although boards can observe these shocks to the stock price, it is almost impossible to link them to price pressure caused by mutual fund flow and subsequently they can be easily misattributed.
We further explore the effect of mutual fund flows on the firm stock price. Cumulative and abnormal returns around an event of extreme price pressure are estimated and plotted in figure 2 (total flow) and figure 3 (in- and outflow separately). An extreme flow is defined as a 1/10th percentile of sample. Total flows before the event are relatively volatile compared to after the event of extreme price pressure. In both cases the stock price decreases, around the event, but then recovers right after the event. Unlike Cai and Wei (2012) no clear trends exist. It seems that extreme in- and outflow does not have an opposite effect on stock performance. Stocks experiencing extreme outflow have lower average abnormal returns than stocks experiencing extreme inflows, but importantly around the event (t=0) the average abnormal returns are almost equal and zero for both groups. The abnormal returns are calculated as Edmans, Goldstein and Jiang (2010) and Cai and Wei (2012) by subtracting firm stock returns from equally weighted market index. We can conclude that extreme mutual fund flows temporarily affects the stock price from the fundamental value, although the effect is low, it is statistically significant.
5.2 The effect of price pressure caused by mutual fund flow on forced turnover
The main purpose of this paper is to test the hypothesis that boards incorporate temporary mispricing of stock prices due to an exogenous shock in their evaluation of a CEO a two-stage least square regression model is used (see Table 8). Because turnovers in the data set are only dated yearly and not monthly, mutual fund flow with a one-year lag is used. The time difference between the actual announcement date and the fiscal year is bounded to 6 months, because the actual announcement of the turnover would be somehow in the middle of the years. As a robustness check, a mutual fund flow price pressure without lag is added to the table as well. Thereby, besides for controlling for firm effects it is important to control for time fixed effects. All our estimated coefficients have the same sign as expected, but the magnitudes and statistically significance differ. Column 1 is our main regression and it shows that the predicted residual, namely the part of the return that is decomposed to the temporarily mispricing shock not due to CEO ability and firm fundamentals does hold a negative relationship with the probability of forced turnover. A
15 1% increase of mutual fund flow mispricing on the firm stock has on average a decrease of 0.2% on the probability of a forced CEO turnover.
Additionally, we use different dependent variables to check the base line regression for robustness (See Table 7, column 1). Theoretically, it would be expected that the magnitude of a transitory exogenous shock to firm performance has a smaller effect on voluntarily turnover, because a CEO would not decide to resign voluntarily in case of a temporary mispricing of the stock from its fundamental value. It is thus expected that regressing on total turnover should result in a lower coefficient on the predicted residual. We do not find results to support this; there is no consistent increase in the value of the coefficients of the predicted residual. See column 2, 3 and 4 for the estimated coefficient predicted on respectively total turnover, turnover when the CEO’s age is higher or equal than 60 and higher or equal than 65. The estimated coefficients indicate that the effect caused by mutual fund flows is small and not significantly different from zero.
Table 11 splits the specification by three groups of different CEO tenure. Column 1, 2 and 3 presents the estimated coefficients on forced turnover for CEO tenure lower or equal to 4 years, between 5 years and 12 years, and higher than 12 years. It shows that the component of the price pressure caused by mutual fund flow has a positive effect on forced turnover when the CEO has a short tenure and negative when the CEO has held his position longer than 4 years. If we assume that CEO ability is better observed by the board for CEOs who already holding their position for a longer time, this regression shows that misattribution happens for CEO of shorter tenure, but because it is not broadly supported by the other regressions, we will not conclude it, but it shows significant differences between age groups of CEOs, providing an additional robustness check.
Table 10 presents the results of a simple OLS regression, a similar model is also estimated by Cai and Wei (2010). Column 1 presents results for the lagged variables and Column 2 for the non-lagged variables, for both specifications the estimated coefficients are close to zero and not statistically significant. This model is prone to statistical errors because it only controls for fixed and time effects but any other control factors. Although it is statistically not conclusive, it does not falsify our baseline regression by presenting totally different or sign reversed coefficients. As expected we find the same kind of signs but with smaller magnitude in comparison with the base line regression. This result differs from Cai and Wei (2012), they find a positive relationship but estimated on the dependent variable CEO pay. CEO pay is a less important and less costly signal than forced CEO turnover, because the board can only decide whether to fire or retain a CEO.
16
6. Conclusion
No evidence is found that corporate board misattribute non-relevant exogenous shocks to firm performance as relevant. Although we do not test strictly for misattribution, we can conclude that boards do not incorporate mutual fund flow pressure in their decision to retain a CEO. There is no statistically significant relationship between forced CEO turnover and price pressure caused by mutual fund flows. Different robustness checks have shown that the effect is probably close to zero. But differences among regression show that the results are robust. We cannot conclude that corporate boards do take exogenous shocks to firm performance in to account to decide on CEO turnover. This research has somehow shown that no further research is really required to dive further into price pressure caused by mutual fund flow, probably because the effect is just too small for the corporate board to take into account (even when the board is not attributing it correctly to factors beyond firm control) and the decision of forced turnover is probably too rigid, due to corporate governance decision processes, to let it be influenced by these flows.
Adding the announcement date of the turnover could improve the empirical methodology because then the price pressure effect can be exactly pinpointed to the event. By definition the effect is averaged on about 6 months. But if we assume that corporate boards do not observe the stock price of their firm on a daily to basis to decide whether or not to fire a CEO also this research is not that relevant. The mispricing is just too transitory for the board to practically base a decision on it.
17 Table 1. Turnover Data
Turnover data have at least been doubly collected for the period of 1993-2006. Turnover data covering the period of 2007-2009 is collected from scratch. Disagreements between datasets are further reviewed into and solved using the classification rules. E.g. the data collected from scratch from 2007-2009 is crosschecked with Nini (2012) and Taylor (2010).
Jenter/Kanaan (2011) 1993 - 2000
Peters/Wagner (2009) 2001 - 2005
Nini (2012) 2001 - 2007
Taylor (2010) 2008 – 2009
Kuhnen (2011) 2005 -2010
Peters/Maassen/RA (collected from scratch) 2007 - 2009
Table 2. Classification Criteria to determine the nature of a Turnover.
The classification criteria listed in the table are used to determine whether a turnover should be perceived as forced or voluntarily. In addition to this, common sense can be helpful when these rules are not applicable.
Forced A. Press reports state that CEO was “forced out”, “fired”, “ousted”, or retires or resigns due to “policy differences, “”pressure” or similar descriptions.
B. CEO is under the age of 60 and (i) the articles do not report the reason to be death, poor health or acceptance of another position or (ii) the departure appears to be unexpected, i.e. the company does not announce the retirement date at least 6 months before the succession.
C. Identify a firm that are dropped from the Execucomp database and identify whether there was a CEO turnover and whether it was forced. (in case of a merger, leveraged buy-out or spin-off, Forced
18 is always coded as 0)
Voluntarily D. CEO is under the age of 60 and classified as forced in B. are reclassified as voluntarily if (i) the CEO takes a comparable position in another company (CEO) soon after his departure (the following or the same fiscal year) or (ii) the press reports convincingly explain that the turnover was due to previously undisclosed personal or business reasons that are unrelated to the firm’s activities or (iii) do not mechanically re-classify as voluntarily if the CEO remains on the board (as do Jenter and Kanaan, 2011)
E. Turnovers in case of a merger, leverage buy-out or spin-offs are always classified as voluntarily. (See Criterion C)
19 Table 3: Definition of Variables
Turn: equals one if a CEO changes within a firm.
Turn60: equals one if a CEO changes within a firm & CEO is under age 60.
Turn65: equals one if a CEO changes within a firm & CEO is under age 65.
Forced: equals one when a CEO is forced to resign its post.
Return (monthly): monthly stock returns collected from CRSP. (PRC/PRC[-1]-1)
Return (yearly): yearly stock returns calculated by multiplying return
Abnormal return: abnormal returns obtained from subtracting the equal weighted S%P 500 normalized returns (Cai and Wei (2012) Pg. 9)
MFflow Price Pressure (monthly): Price pressure to the firm stock price caused by monthly mutual fund flow, the measure is obtained from Edmans, Goldstein and Jiang (2010) and Coval and Stafford (2007) It differs from the one used by Edmans et al. because also outflows are taking into account. Et al. (Negative or Positive).Only stocks experiencing fund flows of higher or lower than 5% during the month are taken into account
MFflow Price Pressure (yearly): Identical to MFflow Price Pressure (monthly), but for yearly periods. Only stocks experiencing flows higher or lower than 5% in the majority of the months the year in which
20 Table 4. Summary Statistics: Mutual Fund Flow driven Price Pressure
Summary Statistics Yearly Mutual Fund flow (%)
Percentiles 1% -8.15903 5% -2.13639 10% -1.04707 25% -0.24816 N 414446 50% -0.00709 mean -0.1385771 75% 0.121603 sd 11.14048 90% 0.740779 Var 124.1102 95% 1.743192 skewness -79.56327 99% 7.475172 kurtosis 10287.11
Summary Statistics Monthly Mutual Fund flow (%)
Percentiles 1% -1.06588 5% -0.19715 10% -0.07258 25% -0.00933 N 414446 50% -0.00014 mean -0.0106871 75% 0.001188 sd 11.31475 90% 0.042092 Var 3.36374 95% 0.14826 skewness -233.9862 99% 1.027124 kurtosis 100502.7
21 Table 5. Summary Statistics: Turnover and Firm Stock Performance
Panel A Frequency of voluntary and forced CEO turnovers
freq. % of Total Turnovers Number of firm-years 15675 Total turnovers 1515 Forced turnovers 331 22% Voluntary turnovers 1184 78%
Voluntarily turnovers when CEO age >=60 288 19% Voluntarily turnovers when CEO age >=65 95 6% Forced turnovers when CEO age>=60 72 22% Forced turnovers when CEO age >=65 35 3%
Panel B Firm performance (yearly stock return)
Percentiles 1% -0.93 5% -0.74655 10% -0.55676 US Firms 7527 25% -0.16733 N 40657 50% 0.183242 mean 0.571175 75% 0.752485 sd 2.220498 90% 1.77648 Var 4.930611 95% 2.862778 skewness 28.18584 99% 6.71151 kurtosis 1713.691
Panel C CEO characteristics
freq. years
distinct CEOs 3726
CEO tenure (avg.) 10.3
CEOs tenure when turnover = forced (avg.) 4.0 CEOs tenure when turnover = voluntarily
22 Figure 1. Forced and voluntary CEO Turnovers
This figure plots the amount of CEO turnovers for U.S. Firms (the right axis). A distinction is made between forced CEO turnovers and voluntary CEO turnovers.
23 Table 6. Time series Mutual Fund Flow price pressure and S&P500 monthly returns
This figure plots the time series of the average monthly mutual fund flow when price pressure exists (the right axis), and the return of the S&P 500 (the left axis). The Pierson correlation coefficient is 0.019. Almost no correlation exists between the two variables.
24 Table 7. Regression of mutual fund flow induced price pressure on monthly stock return
Parameters are estimated by an OLS model. Dependant variable is the firm monthly stock return. Standard errors are clustered by firm. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
(1) (1)
VARIABLES Monthly Return Monthly Return
MFflow price pressure (monthly) .0317741*** .0074474*** (0.003) (0.003) Constant -0.0119185*** -.0117434 *** (0.0004) (0.0004) R^2 0.003 0.07 F-Statistic 132.6 7.8
Time Fixed Effects No Yes
25 Figure 2. (Cumulative) Abnormal average return (CAAR & AAR) of firms’ stock experiencing extreme price pressure caused by mutual fund flows.
This figure plots the time series of the (cumulative) average abnormal return (the left axis) before and after a price event. Only periods of at least 15 days are taking into consideration, consequently more observations at [t=-7, t=7] are used to plot this figure than beyond or after this interval. An extreme flow is defined as an event in which mutual fund flow falls in the 1st/10th percentile of the sample.
26 Figure 3. (Cumulative) Abnormal average return (CAAR & AAR) of firms’ stock experiencing extreme price pressure caused by mutual fund in- and outflows
This figure plots the time series of the (cumulative) average abnormal return (the left axis) before and after a price event. Only periods of at least 15 days are taking into consideration, consequently more observations at [t=-7, t=7] are used to plot this figure than beyond or after this interval. An extreme flow is defined as an event in which mutual fund flow falls in the 1st/10th percentile of the sample.
27 Table 8. Baseline regression results
Parameters are estimated of a 2sls probit model. Dependant variable is the probability that a CEO is forced by the executive board to resign. Standard errors are clustered by firm. The first-stage is estimated by a linear regression. Standard Errors are adjusted. Time fixed effects are used to control for aggregate shocks in (fund flows) and forced CEO turnover. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
(1) (2) (3) (4)
VARIABLES Forced Forced Forced Forced
Predicted yearly return (L-1) -.0330993* -.2928659
(.0229999) (.3643373)
Marginal Effects -.0021918* -.0188981
(.0015257) (.0432964)
Predicted yearly return -.0905244* -.9330133
(.0300595) (.6146448) Marginal Effects -.0058233* -.0626106 (.0019473) (.0413399) Predicted Residual (L-1) -.0318462* -.0227432* (.0199161) (.023011) Marginal effects -.0021088* -.0014676* (.0013217) (.0014867) Predicted Residual -.0816663 -.1041016 (.0228033) (.0293142) Marginal effects -.0052535 -.0069858 (.0014816) (.0019928) Constant -1.873224*** -1.860367*** -1.747916*** -1.381768*** (.0281143) (.0275291) (.3643373) ( .3325736) LR Chi2 2.96 16.27 1.27 18.07 Pseudo R2 0.01 0.005 0.006 0.07
FIRST STAGE L.yearly return yearly return L.yearly return yearly return
L.MFflow_year -.1547249*** -.1131144 *** ( .0379621) (.0382899) Mfflow_year -.1547249*** -.1131144 *** (.0379621) (.0382899) Constant .5368464*** .5368464*** .5378428 *** .5378428 *** (.0117923) (.0117923) (.0116998) (.0116998) adj. R^2 0.005 0.005 0.02 0.02 F-Statistic 16.61 16.61 8.73 8.73
Time fixed effects No No yes yes
28 Table 9. Different Dependant Variables
Parameters are estimated of a 2sls probit regression model. The dependant variable is respectively the probability of a forced turnover, a voluntarily turnover, a turnover when CEO's age >60 years or CEO’s age>65 years old. Standard errors are clustered by firm. The first-stage is estimated by a linear regression. Standard Errors are adjusted. The first stage regression is omitted (See baseline). Time fixed effects are controlled for. Appendix 1 presents an age distribution of the CEO’s in the sample. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
(1) (2) (3) (4)
VARIABLES Forced Turn
Turn60 (CEO age>60y) Turn65 (CEO age>65y)
Predicted yearly return (L-1) -.0330993* .0101312* .0059541* -.0062218
(.0229999) (.0076909) (.0101161) (.01572) Marginal Effects -.0021918* .0008725* .0003406* -.0002252 (.0015257) (.0006625) (.0005787) (.0005689) Predicted Residual (L-1) -.0318462* .0049928* .0040965* -.0064145* (.0199161) (.0057129) (.0067303) (.0113143) Marginal effects -.0021088* .00043 .0002343 -.0002321 .(0013217) (.0004921) (.000385) (.0004095) Constant -1.873224*** -1.756887*** -1.974385*** -2.18749 (.0281143) (.0130489) (.0156465) (.019595) LR Chi2 2.96 1.52 1.40 18.07 Pseudo R2 0.01 0.01 0.01 0.01 Observations 34,718 34,718 34,718 34,718
29 Table 10. OLS Regression
Parameters are estimated of a simple regression model. Dependant variable is probability of a forced CEO turnover. Standard errors are clustered by firm. The first-stage is estimated by a linear regression. Standard errors are adjusted. First stage is omitted. Controlling for time fixed effects. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
(1) (2)
VARIABLES Forced Forced
MFflow_year (L-1) 0.0001916 (0.0017589) Marginal Effects .0000126 (.000116 ) Mfflow_year 0.0013706 (0.0030184) Marginal Effects .0000882 (.0001942 ) Yearly return (L-1) -0.0304839 (.0197951) Marginal effects -.0020102 ( .001308 ) Yearly return -.0816663 (.0228033) Marginal effects -.0052535 (.0014816) Constant -1.876926 *** -1.865123 *** (.0273033) (.0251624 ) LR Chi2 2.74 16.27 Pseudo R2 0.01 0.03 Observations 10,824 12,528
30 Table 11. Different CEO tenure and Forced Turnover
Parameters are estimated of a 2sls probit regression model. The dependant variable is the probability of a forced turnover. The data set is split in three groups, respectively CEO with tenure shorter or equal to 4 years, between 5 and 12 years and longer than 12 years. Standard errors are clustered by firm. The first-stage is omitted. Standard Errors are adjusted. Time fixed effects are controlled for. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
1 2 3
VARIABLES Forced Forced Forced
CEO tenure <= 4 yrs. >4 & <=12 yrs. >12 yrs.
Predicted yearly return (L-1) -0.1134525 -.1279053 ** .2970749**
(0.1151701) (.0472945) (.1509861) Marginal Effects -.0081188 -.008546** .0089269* (.0082533) (.0032077) (.0047486) Predicted Residual (L-1) 0.065373** -.1182475** -.0027255 (0.030694) (.0423974) (.0638781) Marginal effects .0046782** -.0079007** -.0000819 (.0022116) (.0028787) (.0019195) Constant -1.78642*** -1.814552*** -2.431751*** (.08111) (.0460683) (.1169004) LR Chi2 5.82 10.71 4.49 Pseudo R2 0.02 0.02 0.02 Observations 5500 4322 2145
31 Appendix CEO age and Turnovers
Voluntarily Turnovers Forced Turnovers
CEO
Age Freq. Percent Cum. Freq. Percent Cum.
31 1 0.09 0.09 1 0.32 0.32 33 1 0.09 0.18 0 0 0.32 35 2 0.18 0.26 0 0 0.32 36 1 0.09 0.44 0 0 0.32 38 1 0.09 0.53 1 0.64 0.64 39 8 0.70 0.62 2 0.64 1.28 40 9 0.79 1.32 4 1.28 2.56 41 10 0.88 2.11 3 0.96 3.51 42 13 1.14 2.99 7 2.24 5.75 43 17 1.50 4.13 6 1.92 7.67 44 27 2.37 5.63 6 1.92 9.58 45 33 2.90 8.00 12 3.83 13.42 46 28 2.46 10.91 8 2.56 15.97 47 43 3.78 13.37 14 4.47 20.45 48 52 4.57 17.15 13 4.15 24.6 49 58 5.10 21.72 16 5.11 29.71 50 48 4.22 26.82 19 6.07 35.78 51 77 6.77 31.05 17 5.43 41.21 52 68 5.98 37.82 17 5.43 46.65 53 61 5.36 43.80 17 5.43 52.08 54 59 5.19 49.16 21 6.71 58.79 55 67 5.89 54.35 17 5.43 64.22 56 71 6.24 60.25 16 5.11 69.33 57 50 4.40 66.49 14 4.47 73.8 58 50 4.40 70.89 17 5.43 79.23 59 41 3.61 75.29 11 3.51 82.75 60 56 4.93 78.89 12 3.83 86.58 61 40 3.52 83.82 12 3.83 90.42 62 46 4.05 87.34 5 1.6 92.01 63 33 2.90 91.38 6 1.92 93.93 64 18 1.58 94.28 2 0.64 94.57 65 12 1.06 95.87 1 0.32 94.89 66 6 0.53 96.92 3 0.96 95.85 67 9 0.79 97.45 2 0.64 96.49 68 3 0.26 98.24 4 1.28 97.76 69 4 0.35 98.50 1 0.32 98.08 70 2 0.18 98.86 3 0.96 99.04 71 4 0.35 99.03 2 0.64 99.68 73 3 0.26 99.38 0 0 99.68 74 4 0.35 99.65 0 0 99.68 75 0 0.00 100 1 0.32 100 81 1 0.09 100 0 0 100 Total 1137 100 313 100
32 References
Bertrand, M. and Mullainathan, S., 2001, Are CEOS rewarded for luck? The ones without principals are,
Quarterly Journal of Economics 116, 903-932.
Barro, Jason R., and Robert J. Barro, 1990, Pay, performance, and turnover of bank CEOs, Journal of
Labor Economics, 8, p. 48-481.
Cai, Jie, Martijn Cremers, and Kelsey D. Wei. "CEO Compensation and Stock Mispricing: How do boards react to mutual fund flow-driven price pressure?." (2010), working paper.
Coval, Joshua D., and Stafford, Erik, 2007, Asset fire sales (and purchases) in equity markets, Journal of
Financial Economics 86 (2), 479–512
Coughlan, Anne, and Ronald Schmidt, 1985, Executive compensation, management turnover, and firm performance: An empirical investigation, Journal of Accounting and Economics, 7(1-3), p. 43-66.
Edmans, Alex, Goldstein, Itay, and Jiang, Wei, 2010, The real effects of finance markets: The impact of prices on takeovers, Journal of Finance, forthcoming.
Eisfeldt, Andrea L., and Camelia M. Kuhnen. "CEO turnover in a competitive assignment framework."
Journal of Financial Economics 109.2 (2013): 351-372.
Gibbons, Robert, and Kevin J. Murphy, 1990, Relative performance evaluation for chief executive officers, Industrial and Labor Relations Review, 43(3), p. 30-51.
Holmström, Bengt, 1979, Moral hazard and observability, Bell Journal of Economics, 10, 74-91. Holmström, Bengt, 1982, Moral hazard in teams, Bell Journal of Economics, 13, 324-340.
Jenter, Dirk, and Fadi Kanaan, 2011, CEO turnover and relative performance evaluation, Journal of
Finance, forthcoming.
Jenter, Dirk, and Katharina Lewellen, 2010, Performance-induced CEO turnover, Working paper.
Morck, Randall, Andrei Shleifer, and Robert W. Vishny. "Management ownership and market
valuation: An empirical analysis." Journal of financial Economics 20 (1988): 293-315.
Nini, Greg, David C. Smith, and Amir Sufi. "Creditor control rights, corporate governance, and
firm value." Review of Financial Studies 25.6 (2012): 1713-1761.
33 Oyer, Paul, 2004, Why do firms use incentives that have no incentive effects?, Journal of Finance, 59(4), 1619-1649.
Parrino, Robert, 1997, CEO turnover and outside succession: A cross-sectional analysis, Journal of
Financial Economics, 46(2), p. 165-197.
Peters, Florian S., and Alexander F. Wagner. "The executive turnover risk premium." The Journal of
Finance (2014)
Talor, Lucian. “Why Are CEOs Rarely Fired? Evidence from Structural Estimation.” The
Journal of Finance 65, no. 6 (2010): 2051–87.
Warner, Jerold B., Ross L. Watts, and Karen H. Wruck, 1988, Stock prices and top management changes,
