In our paper, CEO Job Security and Risk-Taking, which was recently made publicly available on SSRN, we use the length of employment contracts to estimate CEO turnover probability and its effects on risk-taking. Protection against dismissal should encourage CEOs to pursue riskier projects. Indeed, we show that firms with lower CEO turnover probability exhibit higher return volatility, especially idiosyncratic risk. An increase in turnover probability of one standard deviation is associated with a volatility decline of 17 basis points. This reduction in risk is driven largely by a decrease in investment and is not associated with changes in compensation incentives or leverage.
Dismissal is a serious threat to executives. It leads to the loss of current employment, to reduced future career options (Brickley et al., 1999) and, sometimes, to loss of unvested equity-based compensation (Dahiya and Yermack, 2008). As public demand for increased managerial responsibility has grown over the last three decades, the risk of CEO turnover has risen substantially: the incidence of CEO turnover increased from 13% during 1992–1997 to 17% during 1998–2005 (Kaplan and Minton, 2012). It is therefore important to document the effects of turnover risk on managerial incentives. Yet theoretical predictions concerning this topic are ambiguous: job security should make executives more comfortable to take risk. However, reappointment may be so valuable to the manager that he is willing to pursue negative NPV projects as long as their upside offers improved odds of that manager being retained (“gambling for resurrection”).
Empirical measurement of the effect of job security on risk-taking is complicated by the endogeneity of turnover with respect to performance. In order to circumvent this endogeneity problem, we use the variation in turnover probability that results from ex-ante contract length. Many CEOs of US firms operate under fixed-term employment agreements. Dismissal before the contractual termination date is costly and can lead to litigation. More importantly, firms are likely to set contract lengths according to their horizons for planning and personnel review. Because the remaining contract length decreases with time (and changes upon renewal), we can track the behavior of a given CEO under the same contract, as the horizon changes. Based on 3,954 of these contracts, we find that contractual protection matter for turnover risk. Using a hazard model, we estimate the likelihood of turnover as a function of the CEO’s contract horizon and tenure. Being one year closer to the contract’s expiration date translates into a 21% higher probability of termination, controlling for tenure.
Following the literature (e.g., Guay, 1999; Cohen et al., 2000; Hayes et al., 2012; Gormley et al., 2013; Shue and Townsend, 2013), we use realized stock return volatility as our primary measure of risk-taking. We find a significant negative relationship between turnover probability and risk-taking so defined. A one-standard-deviation increase in the likelihood of CEO turnover is associated with a 17-basis-point decrease in volatility. Since this relation might not be linear, we split firms into quantiles of high and low turnover risk. We discover no evidence of gambling for resurrection in either the high-risk or the medium-risk quantiles.
Throughout the analysis, we control for executive—firm fixed effects. In other words, we hold the CEO—firm pair constant and exploit the variation in turnover risk throughout the CEO’s tenure. One may be concerned that the remaining time under the contract is correlated with the CEO’s tenure at the firm. However, contracts get renewed and their length reset and changed during the renewal. Our results hold when we only use the subsample of CEOs under renewed contracts. Not all CEOs sign fixed-term employment contracts. We control for selection into such contracts using a Heckman (1979) selection model based on variation in employment law across states. Our findings are also robust to controlling separately for firm fixed and executive fixed effects and hold also when we control explicitly for age or tenure groups, as well as time-varying compensation elements.
Our analysis makes several contributions to the existing literature. First, the paper contributes to the literature on CEO turnover. The debate in this area has focused on how CEO turnover is related to firm performance and corporate governance. We introduce a predictor of CEO turnover that not only improves the precision the turnover probability estimation, but also has distinct advantages. Contractual terms establish a turnover timing structure ex-ante, independent of performance. This enables empirical researchers to isolate the causal effects of turnover risk.
Second, our work adds to the literature on risk-taking incentives by establishing an empirical link between career concerns and stock volatility. Such a link has been assumed by the theoretical literature as far back as Fama (1980) and Holmstrom (1982), but despite the theoretical interest, empirical evidence linking career concerns and risk-taking is sparse. A notable exception is Gormley and Matsa (2011), which uses risk arising from large, left-tail events as shocks to job security. They find that mangers respond to these risks by acquiring unrelated businesses with high cash flows in an attempt to limit their downside. A related branch of empirical literature, starting with Chevalier and Ellison (1997 and 1999), establishes a link between the career concerns of mutual fund managers and the risks that they take. However, incentives of fund managers differ from the ones of CEOs as funds are in competition to be ranked as top performers, while firms compete for investors by generating higher absolute returns (irrespective of between-firm rankings).
Third, there are only few empirical studies of CEO employment contracts. Schwab and Thomas (2005) describe a sample of 375 contracts from a legal perspective. Gillan et al. (2009) report that many CEOs operate without an explicit contract. These authors study, the choice between explicit and implicit contracts. We build on their work by describing the effect of contract horizon on career outcomes, risk-taking and performance.
The full paper is available for download here.