Does Ex Ante Severance Pay Affect the Timeliness of Bad News Disclosure? The Role of Managerial Exit Costs

Abstract

Extant literature documents a positive association between ex ante severance pay and timeliness of bad news disclosure, suggesting that the provision of severance pay is consistent with efficient contracting. Relying on an empirically unexplored theory, we investigate whether and how managerial exit costs (i.e., financial and nonfinancial losses triggered by employment termination) affect the effectiveness of severance pay in curbing bad news withholding. We find that managerial exit costs attenuate the positive association between severance pay and timely disclosure of bad news. Moreover, we document that severance pay does not prompt managers to reveal bad news when their exit costs are sufficiently high (i.e., in the top quartile). This result suggests that exit costs erode the efficacy of ex ante severance pay in curtailing bad news withholding. Overall, our findings support the notion that a “one-size-fits-all” approach to structuring severance agreements undermines the potential of severance pay to benefit investors.

Keywords

ex ante severance pay disclosure timeliness managerial exit costs bad news disclosure efficient contracting

Introduction

Ex ante severance pay provides top executives with financial compensation in the event of employment termination and is determined when the initial job offer between the executive and the firm is signed.¹ Managers are entitled to receive severance pay to the extent that they leave the firm for good reasons or are fired without cause. Recent years have witnessed outrage at severance agreements from a variety of parties, including investors, activists, regulators, governments, and the press (Dash, 2007; Deutsch, 2005; Lee, 2014; Stewart, 2011).² They point out that severance pay weakens the link between executive compensation and firm performance, thereby undermining managerial incentives to expend effort. Facing this criticism, advocates justify that severance pay facilitates executive recruiting by providing financial compensation for an uncertain future event—employment termination (Kumar, 2016; Priyan, 2016; Tuttle, 2015). The continuing controversy surrounding this compensation practice suggests that a more complete understanding of the efficacy of severance pay as an incentive alignment mechanism is imperative.

Despite the public furor over severance contracts, extant research provides empirical evidence primarily supporting the notion that severance pay curtails agency problems and enhances firm value (Baginski et al., 2018; Cadman et al., 2016; Chen et al., 2015). Nevertheless, a stream of theoretical research posits that whether severance pay constitutes part of an efficient incentive contract is contingent upon multiple factors (He, 2012; Inderst & Mueller, 2010). In particular, Laux (2015) theorizes that the amount of severance pay in an optimal compensation contract depends on the manager’s desire to keep the current job, represented by the implicit and explicit costs (forgone benefits) incurred by the manager at dismissal. He predicts that when the manager’s disutility of being dismissed is large (i.e., the costs incurred at dismissal are high), the board should include more severance pay to better align managerial actions with investor interests. We refer to these financial and nonfinancial consequences triggered by employment termination as exit costs, an empirically underexplored component of managerial career concerns. Specifically, we examine whether and how exit costs affect severance pay’s effectiveness in curbing managers’ withholding of bad news.

Prior research shows that executives withhold bad news when they are concerned about the negative assessment of their ability and competence by the board and outside investors (Hermalin & Weisbach, 2007; Kothari et al., 2009; Verrecchia, 2001). Ex ante severance pay alleviates this concern in that it establishes a predetermined financial compensation in the case where revealing negative information results in employment termination (Baginski et al., 2018; Ling, 2012). Although severance pay makes disclosing bad news less threatening to the manager than it would be in the absence of severance pay, managers face varying amounts of exit costs if dismissed. We expect that managers evaluate the costs and benefits associated with disclosing bad news that could result in dismissal. This tradeoff suggests that the financial benefit created by severance pay becomes less pronounced for managers who will incur (forgo) greater financial and nonfinancial losses (benefits) when leaving the current employer. As a result, despite promised severance pay, managers with higher exit costs would be more reluctant to provide timely disclosure of adverse outcomes to the extent that bad news elicits the board’s disappointment in managerial competence and increases dismissal likelihood. Hence, we predict that managerial exit costs weaken the effectiveness of ex ante severance pay in deterring bad news withholding.

While prior research provides mixed evidence on the determinants of severance pay (Rau & Xu, 2013), it is possible that boards consider firm- and manager-level characteristics related to exit costs in the design of severance agreements. We suggest that in the context of bad news hoarding, it is difficult for the board to determine the amount of severance pay to appropriately offset exit costs for at least two reasons. First, certain compensation schemes complement severance pay to enhance alignment between managerial actions and investor interests even though they can generate high financial exit costs. For example, recent research shows that managers engage in value-enhancing risk taking when offered higher amounts of both stock options and severance pay (Almazan & Suarez, 2003; Inderst & Mueller, 2010). Thus, while severance pay curbs bad news withholding, option compensation unvested at dismissal makes it costlier for managers to leave the current firm and can deter them from revealing negative information. Second, the nonfinancial aspects of exit costs cannot be precisely determined. For instance, exit costs associated with managers’ social and geographic ties are typically developed over time, making it difficult for the board to project in the hiring process. Moreover, losses in reputation, prestige, and professional records will likely persist and thus cannot be fully offset by a predetermined amount of financial compensation. As a result, the effect of severance pay on the timeliness of bad news disclosure is unlikely to be uniform as exit costs vary across managers.

We test our prediction using the I/B/E/S and Standard & Poor’s Capital IQ (CIQ) databases with data from 2006 to 2014. We focus on chief executive officers (CEOs), and collect severance pay data from CIQ. We validate the credibility of this information source by cross-checking with a hand-collected sample from proxy statements. We follow Donelson et al. (2012) and measure disclosure timeliness based on how quickly earnings news is revealed via the evolution of analyst earnings forecasts. This measure better captures the total earnings news managers reveal to the capital markets in that analyst forecasts incorporate information from a variety of managerial disclosure channels (e.g., earnings guidance, conference calls/presentations, analyst/investor days, and corporate websites). Relying on the labor market literature on managers’ career prospects and retention incentives, we adopt multiple measures to capture both financial and nonfinancial aspects of managerial exit costs. Specifically, we use the extent of business affiliations, the amount of forfeitable equity compensation, and the number of local board positions held by the CEO.

We first confirm the findings in prior research by documenting that severance pay is associated with timelier disclosure of bad news relative to good news. We then find that the positive association between ex ante severance pay and timeliness of bad news disclosure becomes less pronounced as managerial exit costs increase, consistent with our expectation. This result suggests that although severance pay has potential value for investors, managerial exit costs undermine the efficacy of severance pay in curbing bad news withholding. Furthermore, we document that when managerial exit costs reach a sufficiently high level (i.e., the top quartile), severance pay no longer positively relates to timelier disclosure of bad news relative to good news. This finding is consistent with the notion that managers are reluctant to reveal negative information early when the financial and nonfinancial costs of dismissal from the current position outweigh the financial benefit created by ex ante severance pay.

Our study makes several important contributions to extant literature. First, we expand the scope of the research on the incentive effects of ex ante severance pay. Despite negative public perception on severance pay, prior research provides empirical evidence that severance pay serves as a governance mechanism to alleviate agency conflicts rather than a means of rent extraction for executives (Cadman et al., 2016). This literature supports the notion that severance pay protects managers from risky actions that enhance firm value (i.e., an insurance role) as opposed to shielding managers from decisions to the detriment of investors. However, analytical research predicts that whether severance pay represents a form of efficient contracting hinges critically on managerial desire to keep the job (Laux, 2015). In light of little empirical research investigating this issue, our study fills the gap by showing that the efficacy of severance pay as insurance against poor firm performance outside managerial control is significantly weakened when managers face sufficiently high exit costs (i.e., stronger desire to stay with the current employer).

Second, we add to the growing research on the role of ex ante severance pay in corporate financial reporting practices. This literature in general supports a positive association between severance pay and financial reporting quality, such as timelier disclosure of bad news (Baginski et al., 2018), increased accounting conservatism (Ling, 2012), less real earnings management (Chen et al., 2015), and reduced likelihood of accounting fraud and meeting/beating analyst forecasts via discretionary accruals (Brown, 2015). In contrast, our findings suggest that managers trade off the predetermined financial benefit created by severance pay and the explicit and implicit costs triggered by employment termination when they determine corporate disclosure transparency.

In addition, we provide timely and practical implications about the design of executive compensation contracts. Cowen et al. (2016) theorize that the potential of severance pay to serve the interests of both investors and executives is weakened by the “one-size-fits-all” approach boards typically take to structuring severance agreements nowadays. They therefore call for empirical research examining contingencies that affect the efficacy of severance pay in reducing agency costs. Answering this call, our results suggest that the ultimate value of an ex ante severance agreement to investors is likely eroded to the extent that the boards do not account for the constantly changing dynamics between managerial exit costs and severance pay over a CEO’s career. Rather than adopting institutionalized severance contracts that simply conform to conventional structures, firms should tailor severance terms to better reflect the evolving exit costs of individual executives to increase the effectiveness of severance pay as a governance tool.

Related Literature and Hypothesis

Ex Ante Severance Pay and Efficient Contracting Theory

Despite the public’s rebuke at the excessive amount of severance pay, academic research in general supports the notion that severance pay functions as a governance tool that serves the interests of both investors and executives. For example, Cadman et al. (2016) document that severance pay is positively associated with the extent to which a CEO invests in risky projects with a positive net present value. In the context of corporate financial reporting, empirical evidence also supports severance pay representing a form of efficient contracting to restrain managerial myopia. Chen et al. (2015) find that managers with higher severance pay are less likely to avoid earnings declines through engaging in real earnings management. Brown (2015) documents that severance pay is associated with a reduced likelihood of committing accounting fraud and earnings management via discretionary accruals.

Recent research investigates the effect of severance pay on managers’ bad news disclosure. Managers have incentives to withhold bad news when they are concerned about the effect of revealing negative information on the assessment of their ability, hoping that subsequent events will turn in their favor (Hermalin & Weisbach, 2007; Kothari et al., 2009; Verrecchia, 2001). To the extent that disclosing bad news jeopardizes the manager’s career prospects and personal wealth, severance pay should reduce their incentive to withhold bad news by providing financial compensation in the event of dismissal. In line with this argument, Baginski et al. (2018) document that with sufficiently large severance pay, managers no longer delay the release of bad news relative to good news in management earnings forecasts. In a similar vein, Ling (2012) documents a positive association between the existence of severance agreements and accounting conservatism (i.e., timelier recognition of bad news in reported earnings).

Ex Ante Severance Pay and Managerial Exit Costs

Managers’ career concerns arise from the assessment of their ability and competence by the board and outside investors and how this assessment affects their current and future financial compensation (Gibbons & Murphy, 1992). While extant empirical research generally supports severance pay alleviating managerial career concerns and reducing agency costs, a stream of theoretical research posits that the ability of severance pay to benefit investors is contingent upon different factors and thus this compensation practice may not always be a form of efficient contracting. For example, Inderst and Mueller (2010) predict that to induce incompetent managers to quit, it is more efficient to make the rewards of continued employment less attractive rather than reward quitting by offering severance pay. He (2012) indicates that the board should consider managers’ payoffs from outside alternative options when structuring severance agreements.

Laux (2015) theorizes that the amount of severance pay in an optimal compensation contract hinges on the manager’s desire to keep the job, represented by the costs of losing their current position (including benefits forgone upon employment termination). His model predicts that when the CEO’s disutility of being dismissed is large, the optimal compensation contract should include higher severance pay to align managerial actions with investor interests. Relying on this underexplored literature, we refer to such disutility at dismissal as managerial exit costs. We suggest that managers’ exit costs represent a subset of career concerns that encompass financial and nonfinancial costs they would incur from employment termination.³ Building on the labor market literature, we adopt multiple measures to capture both the financial and nonfinancial (e.g., social and geographic) aspects of exit costs. Specifically, we expect managers to incur greater exit costs when they have less developed business affiliations, hold more forfeitable equity compensation, and hold more local board positions. We provide a more detailed discussion of these measures in the research design section.

Hypothesis

In this study, we examine whether and how managerial exit costs affect the efficacy of severance pay in curbing bad news withholding. Prior research provides inconclusive evidence on what firm- and executive-level characteristics the board considers in the design of severance agreements (Cadman et al., 2016; Rau & Xu, 2013). In particular, results from prior research on the determinants of the contracted amount of severance pay are not robust to alternative specifications.⁴ We suggest that even though certain firm and managerial characteristics are considered in the negotiation of severance agreements, severance pay may not appropriately compensate for exit costs in the context of bad news withholding for at least two reasons.

First, combined with ex ante severance pay, certain executive compensation components provide complementary incentives that benefit investors even though they tend to generate high managerial exit costs. For example, stock options reward managers for exceptional stock performance, while severance pay provides downside protection. Recent research shows that severance pay and stock options are offered together in executive pay to encourage managerial effort to undertake risky projects on behalf of investors (Almazan & Suarez, 2003; Inderst & Mueller, 2010). However, these two compensation schemes have opposing effects on bad news withholding. While severance pay encourages timelier disclosure of bad news, managers with larger unvested option holdings would incur greater financial losses at dismissal and would thus withhold bad news to avoid employment termination. As a result, when managers hold both high severance pay and high option compensation, it is ex ante unclear which effect would dominate.

Second, nonfinancial exit costs cannot be precisely determined when the manager is hired. For example, in addition to lost pay from the current firm, dismissed managers with less developed social networks suffer further financial losses if they have difficulty finding a new comparable position. Similarly, managers who have developed strong geographic ties may not want to leave the current firm if, for example, they lose benefits linked to the local labor market (e.g., income streams from local directorships, country club memberships, or family ties). Furthermore, owing to employment termination, losses in reputation, professional accomplishment, and personal fulfillment will likely persist and are difficult for the board to project when hiring a new executive.⁵

Ex ante severance pay guarantees financial compensation if managers are fired without cause or leave the firm for good reasons, thereby reducing the sensitivity of managerial personal wealth to dismissal caused by poor firm performance. As a result, managers promised severance pay are more likely to reveal bad news in a timely manner even though disclosure of negative information increases their likelihood of being dismissed by the board. However, we suggest that not all managers are willing to forgo the rewards of continued employment for a predetermined amount of financial compensation. That is, although severance pay makes disclosing bad news less threatening, managers face varying exit costs upon losing their current job, which in turn affects the efficacy of severance pay in prompting bad news disclosure.

We expect that managers weigh the financial benefit created by severance pay against personal (financial and nonfinancial) losses triggered by employment termination when they assess whether to provide early warnings about bad firm performance that could elicit the board’s disappointment and result in dismissal. This cost–benefit tradeoff suggests that managers are less likely to communicate bad news with outside stakeholders in a timely manner as the net costs of leaving the current firm increase. As a result, the effectiveness of severance pay in prompting managers to deliver timely bad news is undermined by exit costs they incur upon employment termination. Hence, we hypothesize in H1 that for a given amount of severance pay, managers are more likely to withhold bad news as their exit costs increase.

Hypothesis 1 (H1): Managerial exit costs weaken the ability of ex ante severance pay to deter bad news withholding.

Sample and Research Design

Sample Selection

Our sample begins with all firms in the intersection of I/B/E/S and CRSP with available analyst forecast and stock price data for fiscal years 2006 to 2014.⁶ We retain firms with positive or negative earnings news in each fiscal quarter.⁷ This results in 146,640 firm-quarter observations with valid measures of the timeliness of earnings news. We collect annual severance data for CEOs from the Standard and Poor’s Capital IQ (CIQ) database, keeping only those observations where the CEO has nonzero total compensation. We obtain control variables from CIQ and exclude firms in the financials and utilities sectors (based on Global Industry Classification Standard [GICS]), resulting in a final sample of 46,750 firm-quarter observations (2,762 unique firms). Some tests have fewer observations due to missing values. Appendix A provides our sample selection procedure.

Measure of Disclosure Timeliness

Our measure of disclosure timeliness (TIMELY) is based on the movement of the consensus analyst forecast over an entire quarter, adopted from Donelson et al. (2012). Managers can release earnings news through a variety of disclosure channels, such as earnings guidance, conference calls and presentations, analyst/investor days, and corporate websites. As a widely used proxy for market expectations of firm prospects, analyst forecasts incorporate earnings information released in various managerial disclosures, including those unavailable to academic researchers.⁸ Using the change in the consensus forecast as a quarter progresses allows us to capture how quickly all earnings news is revealed to the market through any disclosure channel that is incorporated by analysts into their forecasts (Donelson et al. 2012). We examine the timeliness of all disclosures (as they are incorporated into the consensus analyst forecast) rather than focusing on one specific disclosure type because the board is unlikely to take actions solely based on market reactions to a single form of bad news revelation such as management’s earnings guidance (Pae et al., 2016).

Knowing ex post the total earnings news (earnings per share [EPS] at the end of the quarter relative to the consensus forecast at the beginning of the quarter), we calculate the proportion of the total earnings news that has been revealed and incorporated by analysts by each day of the quarter.⁹ For example, if the consensus forecast on the first day of a quarter is $1.00 per share and the actual earnings at the end of the quarter are $1.50 per share, then the implied good news for the quarter is $0.50 per share. If on a given day, the consensus forecast is $1.20 per share, then 40% of the quarter’s news has been revealed by that day [($1.20 − $1.00)/($1.50 − $1.00)]. TIMELY is the average of these daily proportions of news revealed over the entire quarter.¹⁰ Therefore, higher values of TIMELY represent quarters in which more news is revealed earlier in the quarter.¹¹ We provide a detailed example calculation of TIMELY in the Online Appendix A.

Measures of Managerial Exit Costs

To measure managerial exit costs, we rely on the labor market literature on managerial career prospects (Laux, 2012; Liu, 2014) and retention incentives (Balsam & Miharjo, 2007; Finkelstein, 1992). Given that exit costs are incurred only when changing jobs, managers face lower exit costs when their outside career prospects are more promising and higher exit costs when their retention incentives in the current firm are stronger. We measure these exit costs using (a) the lack of business connections (UNCONNECTED), (b) the amount of forfeitable equity compensation (FORFEIT), and (c) the number of local board positions (LOCALLINK). As these three proxies capture different aspects of managerial exit costs, we create an overall measure of managerial exit costs (EXITCOST), calculated as the sum of UNCONNECTED, FORFEIT, and LOCALLINK, scaled to fall between zero and one.

The first exit cost measure captures the extent of a CEO’s business connections (UNCONNECTED). Prior research shows that executives with richer business connections are more likely to find alternative employment opportunities with high compensation (Liu, 2014; Renneboog & Zhao, 2011). As a result, to the extent that CEOs with weaker business connections face greater difficulty securing new employment with high pay, they likely bear higher costs from leaving the current firm. We measure UNCONNECTED as the average decile ranking of the following, multiplied by negative one and scaled to range between zero and one: the number of firms the CEO is affiliated with as an employee or director, the number of words in the CEO’s biography, and the number of affiliated companies visible enough to have available revenue information in the CIQ database. We use the CEO’s biography because it explains connections that are not reflected in the number of business affiliations (e.g., club and fraternity membership). In addition, we include business affiliations with sales information to consider the ability of potential employers to provide high pay and stable employment.¹² CEOs with higher values of UNCONNECTED have higher exit costs.

The second exit cost measure is the amount of the CEO’s forfeitable compensation (FORFEIT). Equity compensation is used to align the interests of managers and shareholders by exposing mangers’ personal wealth to their firm’s stock prices. Employees often forfeit a significant amount of unvested stock options and restricted stock upon leaving their current employer. Theoretical and empirical studies suggest that managers have stronger incentives to stay in their current employment when they would forfeit equity compensation upon employment termination (Balsam & Miharjo, 2007; Edmans et al., 2012; Gopalan et al., 2014; Jochem et al., 2018; Laux, 2012; Oyer & Schaefer, 2005). This suggests that CEOs face higher exit costs when they hold a greater amount of forfeitable stock options and restricted stock shares. We measure FORFEIT as the scaled decile ranking of the CEO’s unvested stock options and restricted stock. CEOs with high values of FORFEIT have higher exit costs.

The third exit cost measure is number of local board positions held by the CEO (LOCALLINK). Strong geographic ties elicit retention incentives because they represent “soft” benefits that managers forgo by leaving the firm. For example, CEOs with local geographic connections earn a significant compensation premium (Engelberg et al., 2013). CEOs with connections to the location of the firm have strong incentives to keep their job because they would otherwise lose the reputational and financial benefits associated with local ties. LOCALLINK equals the number of board positions held by the CEO for companies headquartered in the same state as the CEO’s firm, scaled (out-of-sample) to fall between zero and one. CEOs with higher values of LOCALLINK have higher exit costs.

Models

We first confirm that high severance pay is associated with timelier disclosure of bad news by estimating the following ordinary least squares (OLS) regression model:

\begin{matrix} TIMEL Y_{i, t} = α_{0} + α_{1} SEVERANC E_{i, t} + α_{2} BADNEW S_{i, t} \\ + α_{3} BADNEW S_{i, t} \times SEVERANC E_{i, t} + β_{n} Control s_{i, t} + ε_{i, t} \end{matrix}

(1)

We use two measures of SEVERANCE to capture the magnitude of severance pay. The first measure captures the raw level of severance pay, calculated as the scaled decile rank of the CEO’s contracted severance pay (RSEVERANCE). The second measure captures the relative magnitude of the severance pay for that specific manager, calculated as the scaled decile rank of the CEO’s contracted severance pay divided by total compensation (%SEVERANCE). BADNEWS is an indicator variable equal to one if the actual EPS for the quarter is lower than the consensus forecast on the first day of the quarter and zero otherwise. Consistent with larger contracted severance pay being associated with timelier disclosure of bad news, we expect α₃ to be positive.

To test our hypothesis that the effect of severance pay on the disclosure of bad news is weaker when managers’ exit costs are higher, we estimate the following regression:

\begin{matrix} TIMEL Y_{i, t} = α_{0} + α_{1} SEVERANC E_{i, t} + α_{2} BADNEW S_{i, t} + α_{3} BADNEW S_{i, t} \times SEVERANC E_{i, t} \\ + α_{4} EXITCOS T_{i, t} + α_{5} EXITCOS T_{i, t} \times SEVERANC E_{i, t} \\ + α_{6} EXITCOS T_{i, t} \times BADNEW S_{i, t} \\ + α_{7} EXITCOS T_{i, t} \times BADNEW S_{i, t} \times SEVERANC E_{i, t} + β_{n} Control s_{i, t} + ε_{i, t} \end{matrix}

(2)

In Equation 2, we expect the coefficient α₇ to be negative. This result is consistent with less timely disclosure of bad news for CEOs with both high severance pay and high exit costs (relative to CEOs with only high severance pay or only high exit costs). We control for career concerns using the CEO’s founder status (FOUNDER), block ownership (BLOCKHOLDER), board influence (CHAIRMAN), age (LAGE), and tenure (TENURE). Controls for the firm’s information environment include size (LOGMV), whether the firm is audited by one of the Big 4 accounting firms (BIG4), the volatility of the firm’s stock return during the current year (RETURNVOL), and the number of analysts following the firm (LOGANALYSTS). To control for the effect of other compensation on the timeliness of disclosure, we include the natural log of the CEO’s total compensation (LOGCOMP). We control for the amount of news revealed during the quarter (NEWS) and the number of days in the quarter (LOGNUMDAYS) because both may influence how quickly the news is revealed. As our timeliness measure is based on analyst forecasts, we also include several variables to control for differences in analyst characteristics across firm-quarters: the total number of forecasts (LOGFORECASTS), the average frequency of forecasts issued for that firm quarter (FORECASTFREQ), the average firm specific experience (FIRMEXPERIENCE), the average overall experience (TOTEXPERIENCE), and the average number of firms followed (FIRMSFOLLOW) by analysts issuing forecasts for the firm. We also include earnings surprise (SURPRISE) to capture analyst forecast accuracy and the interplay of analyst and management incentives.

We gauge the timeliness of managers’ disclosure of good and bad news based on analyst forecasts. This choice is appropriate because of the previously mentioned advantages and also because analysts rely on managerial disclosures to update their earnings forecasts (Altschuler et al., 2015; Beyer, 2008). However, it is possible that analysts use other information sources (e.g., industry-specific and macro-economic news) when creating and updating their earnings forecasts. While we do not expect these other information sources to influence the relative timeliness of bad news disclosures compared with good news disclosures, we include industry and year fixed effects to control for this possibility. Finally, we are concerned with endogeneity and one source is time invariant unobservable firm characteristics that determine both compensation packages and firms’ disclosure policies. We address this concern by including firm fixed effects.

Empirical Results

Descriptive Statistics

Table 1 reports descriptive statistics for the sample of 46,750 firm-quarter observations representing 2,762 unique firms.¹³ The average timeliness of earnings news (TIMELY) is 0.09, suggesting that for the average firm, the average daily percentage of news that has been revealed in a quarter is about 9%. Only one other study, Donelson et al. (2012), uses this measure of earnings timeliness, and the distribution of TIMELY is different from the reported distribution in their study. However, their study consists of a select sample of sued firms and their matched counterparts (total of 423 observations), which have much higher magnitudes of negative news relative to our larger sample containing both good and bad news quarters. Hence, the distribution of variables in their sample is not necessarily generalizable to a broad population. The table reports only ranked values of severance pay but in unreported analysis, the average CEO has an ex ante contracted severance amount of $2.6 million. This figure is lower than the $7.3 million reported in Baginski et al. (2018), but is expected because their sample consists of S&P 1,500 firms that are significantly larger than many firms in our sample. The correlation matrix of variables used in the regression models is presented in the Online Appendix B.

Table 1.

Descriptive Statistics.

Variables	N	M	SD	Minimum	25th	Medium	75th	Maximum
Dependent variables
TIMELY	46,750	0.09	0.36	−1.00	0.00	0.00	0.19	1.00
Independent variables
RSEVERANCE	46,750	0.53	0.31	0.00	0.22	0.56	0.78	1.00
%SEVERANCE	46,750	0.83	0.12	0.67	0.78	0.78	0.89	1.00
BADNEWS	46,750	0.44	0.50	0.00	0.00	0.00	1.00	1.00
EXITCOST	46,750	0.35	0.19	0.00	0.20	0.33	0.49	0.90
UNCONNECTED	46,750	0.49	0.28	0.00	0.26	0.52	0.70	1.00
FORFEIT	46,750	0.83	0.13	0.67	0.67	0.78	0.89	1.00
LOCALLINK	46,750	0.01	0.02	0.00	0.00	0.00	0.00	0.32
Control variables
FOUNDER	46,750	0.08	0.27	0.00	0.00	0.00	0.00	1.00
BLOCKHOLDER	46,750	0.07	0.25	0.00	0.00	0.00	0.00	1.00
CHAIRMAN	46,750	0.46	0.50	0.00	0.00	0.00	1.00	1.00
LAGE	46,750	4.03	0.34	0.00	3.97	4.06	4.13	4.51
TENURE	46,750	3.06	0.56	0.69	2.83	3.22	3.50	3.61
LOGMV	46,750	6.91	1.89	0.00	5.72	6.89	8.10	12.75
BIG4	46,750	0.85	0.36	0.00	1.00	1.00	1.00	1.00
RETURNVOL	46,750	3.60	0.82	0.00	3.34	3.68	4.03	6.08
LOGCOMP	46,750	14.84	1.14	0.00	14.13	14.86	15.57	18.87
NEWS	46,750	0.00	0.03	−0.22	0.00	0.00	0.00	0.06
LOGANALYSTS	46,750	2.22	0.69	0.69	1.79	2.20	2.71	4.01
LOGFORECASTS	46,750	3.20	0.93	0.69	2.56	3.26	3.87	5.81
LOGNUMDAYS	46,750	4.51	0.18	3.43	4.47	4.52	4.60	5.01
SURPRISE	46,750	0.00	0.02	−0.20	0.00	0.00	0.00	0.06
FORECASTFREQ	46,750	3.11	1.09	1.00	2.33	3.00	3.72	6.33
FIRMEXPERIENCE	46,750	2.65	2.07	0.00	0.00	2.84	4.13	7.57
TOTEXPERIENCE	46,750	4.38	3.26	0.00	0.00	5.00	6.73	11.80
FIRMSFOLLOW	46,750	12.56	8.63	0.00	0.00	15.59	18.50	31.86

Note. Table 1 presents descriptive statistics. Please refer to Appendix A for sample selection criteria and Appendix B for variable definitions.

Figure 1 provides a graphical depiction of our findings. For each day within the quarter, we calculate the proportion of the news that has been revealed up to that day by taking the difference between that day’s consensus and the beginning consensus, scaled by total news for the quarter. In Figure 1, we divide each quarter into 20 equal intervals and present the mean proportion of news that has been revealed by each interval. We separate the firm-quarter observations into four groups based on whether the CEO severance was above or below the median and whether the overall news for the quarter was good or bad. Panel A presents firm quarters with above median EXITCOST while Panel B presents firms quarters with below median EXITCOST.

Figure 1.

Provides a graphical depiction of our results.

Consistent with Donelson et al. (2012), both Panel A and Panel B of Figure 1 show that bad news is disclosed in a timelier manner than good news on average. This result supports the notion that managers tend not to withhold bad news when total earnings news sources are considered.¹⁴ Figure 1 also indicates higher severance pay is associated with timelier disclosure of bad news. A comparison of Panel A and Panel B suggests that when managerial exit costs are high (Panel A), the difference between the low and high severance groups in disclosing bad news is less pronounced. In other words, Figure 1 shows that high exit costs appear to reduce the effect of severance on timelier disclosure of bad news, thereby providing preliminary support for H1.

Severance Pay, Managerial Exit Costs, and Timeliness of Bad News Disclosure

Prior literature documents a positive association between severance pay and the timeliness of bad news disclosure (Baginski et al., 2018). As a starting point, we confirm this result with our measure of disclosure timeliness. Table 2 reports the results of estimating Equation 1 using OLS. Column 1 uses the level of severance (RSEVERANCE) and column 2 uses its scaled counterpart (%SEVERANCE). Consistent with theory and prior empirical literature (Baginski et al., 2018; Levitt & Snyder, 1997), the interaction between bad news and severance pay (BADNEWS × SEVERANCE) positively relates to earnings news timeliness (TIMELY). This implies that more bad news is revealed earlier in the quarter for firms with high CEO severance pay. Estimated coefficients for control variables are intuitive and generally consistent with Donelson et al. (2012).

Table 2.

OLS Regression Results for Severance and Timeliness of Bad News Disclosures.TIMELY_i,t = α₀+α₁SEVERANCE_i,t+α₂BADNEWS_i,t+α₃BADNEWS_i,t×SEVERANCE_i,t+β_nControls_n,i,t+ϵ _i,t (1)

SEVERANCE = Variables	RSEVERANCE Coef. (t-stat)	%SEVERANCE Coef. (t-stat)
BADNEWS	0.168*** (23.751)	0.109*** (4.560)
SEVERANCE	−0.027** (−2.041)	−0.033 (−1.273)
BADNEWS × SEVERANCE	0.106*** (9.463)	0.139*** (4.923)
FOUNDER	−0.010 (−0.210)	−0.015 (−0.312)
BLOCKHOLDER	0.011 (0.572)	0.013 (0.674)
CHAIRMAN	−0.009 (−0.743)	−0.009 (−0.704)
LAGE	0.000 (0.008)	−0.000 (−0.005)
TENURE	−0.009 (−1.138)	−0.009 (−1.084)
LOGMV	0.003 (1.105)	0.003 (1.044)
BIG4	0.022 (1.591)	0.021 (1.541)
RETURNVOL	−0.000 (−0.038)	−0.000 (−0.082)
LOGCOMP	0.002 (0.709)	0.005 (1.478)
NEWS	−2.163*** (−11.037)	−2.121*** (−10.814)
LOGANALYSTS	−0.024 (−0.793)	−0.022 (−0.721)
LOGFORECASTS	0.017 (0.672)	0.015 (0.588)
LOGNUMDAYS	−0.013 (−1.302)	−0.012 (−1.244)
SURPRISE	3.665*** (17.028)	3.692*** (17.144)
Constant	0.021 (0.226)	−0.011 (−0.106)
Analyst controls	Yes	Yes
Firm, year & industry fixed effects	Yes	Yes
N	46,750	46,750
R ²	.091	.090

Note. Table 2 presents regression results of estimating Equation 1. For each regression, the estimated coefficients are presented with t-statistics in the brackets to the right. Please refer to Appendix A for sample selection criteria and Appendix B for variable definitions. ***, **, and * indicate significance (two-tailed) at the 1%, 5%, and 10% levels, respectively.The lines are highlighted in bold to emphasize they are the coefficients of interest.

Our hypothesis predicts that the positive association between severance pay and timeliness of bad news disclosure is less pronounced when managerial exit costs are high. Table 3 presents the results of this test using multiple specifications. Panel A reports results from estimating Equation 2 using two different measures of severance pay (RSEVERANCE and %SEVERANCE) and a composite measure of exit costs (EXITCOST). Consistent with our prediction, the coefficient on EXITCOST × BADNEWS × SEVERANCE is significantly negative using both measures of SEVERANCE. The coefficient of −0.411 on this interaction term in column 1 indicates that for firms with high severance pay, the average proportion of bad news known during the quarter is 41% lower for CEOs with high exit costs relative to CEOs with low exit costs, suggesting less timely disclosure of bad news by managers with higher exit costs.¹⁵ Although not directly related to our research question, we note a positive and significant coefficient on EXITCOST × BADNEWS, suggesting that for CEOs with low severance pay, high exit costs are associated with timelier disclosure of bad news.¹⁶ In unreported analysis, we dichotomize severance pay using the top decile, above median, above mean, and above zero severance pay, respectively. We continue to find that for these varied definitions of high and low severance pay, high exit costs are associated with timelier disclosure of bad news.

Table 3 Panel A.

OLS Regression Results for Severance and Timeliness of Bad News Disclosures.TIMELY_i,t = α₀+α₁SEVERANCE_i,t+α₂BADNEWS_i,t+α₃BADNEWS_i,t×SEVERANCE_i,t+α₄EXITCOST_i,t+α₅EXITCOST_i,t×SEVERANCE_i,t+α₆EXITCOST_i,t×BADNEWS_i,t+α₇EXITCOST_i,t×BADNEWS_i,t×SEVERANCE_i,t+β_nControls_i,t+ϵ_i,t (2)

SEVERANCE = Variables	RSEVERANCE Coef. (t-stat)	%SEVERANCE Coef. (t-stat)
BADNEWS	0.016 (0.477)	−0.256** (−2.267)
SEVERANCE	−0.124** (−2.405)	−0.181 (−1.606)
BADNEWS×SEVERANCE	0.377*** (7.012)	0.561*** (4.173)
EXITCOST	−0.064 (−1.190)	−0.167 (−1.196)
EXITCOST×BADNEWS	0.232*** (4.797)	0.552*** (3.315)
EXITCOST×SEVERANCE	0.143* (1.898)	0.217 (1.319)
EXITCOST × BADNEWS × SEVERANCE	−0.411*** (−5.162)	−0.638*** (−3.222)
FOUNDER	−0.009 (−0.179)	−0.012 (−0.245)
BLOCKHOLDER	0.013 (0.674)	0.014 (0.730)
CHAIRMAN	−0.008 (−0.649)	−0.009 (−0.669)
LAGE	−0.000 (−0.002)	0.000 (0.011)
TENURE	−0.010 (−1.170)	−0.009 (−1.127)
LOGMV	0.003 (1.071)	0.003 (0.953)
BIG4	0.021 (1.553)	0.021 (1.495)
RETURNVOL	0.000 (0.011)	−0.000 (−0.066)
LOGCOMP	0.002 (0.606)	0.005 (1.373)
NEWS	−2.170*** (−11.068)	−2.125*** (−10.825)
LOGANALYSTS	−0.024 (−0.782)	−0.022 (−0.729)
LOGFORECASTS	0.017 (0.659)	0.015 (0.591)
LOGNUMDAYS	−0.013 (−1.320)	−0.012 (−1.259)
SURPRISE	3.654*** (16.984)	3.684*** (17.105)
FORECASTFREQ	0.011 (1.391)	0.012 (1.479)
FIRMEXPERIENCE	−0.009*** (−2.911)	−0.009*** (−2.957)
TOTEXPERIENCE	0.002 (1.066)	0.002 (1.137)
FIRMSFOLLOW	−0.000 (−0.201)	−0.000 (−0.274)
Constant	0.072 (0.727)	0.111 (0.812)
Firm, year, and industry fixed effects	Yes	Yes
N	46,750	46,750
R ²	.092	.090

Note. Panel A of Table 3 presents regression results of estimating Equation 2 based on the composite measure of exit costs. For each regression, the estimated coefficients are presented with t-statistics in the brackets to the right. Please refer to Appendix A for sample selection criteria and Appendix B for variable definitions. ***, **, and * indicate significance (two-tailed) at the 1%, 5%, and 10% levels, respectively.The lines are highlighted in bold to emphasize they are the coefficients of interest.

To further assess the magnitude of our results, we decile rank EXITCOST and create TOPEXITCOST which equals one (zero otherwise) for the top decile values of EXITCOST. Panel B of Table 3 reports results from estimating Equation 2 using the two measures of severance pay and replacing EXITCOST with TOPEXITCOST. The coefficient of −0.070 on TOPEXITCOST × BADNEWS × SEVERANCE in column 1 indicates that for firms with high severance pay, the average proportion of bad news known during the quarter is 7% lower for CEOs in the top decile of exit costs relative to all other high severance pay CEOs. In untabulated results, we also create TOPSEVERANCE which equals one (zero otherwise) for the top decile values of total severance pay. The coefficient on TOPEXITCOST × BADNEWS × TOPSEVERANCE is −0.072, implying the average proportion of bad news known during the quarter is 7% lower for CEOs in the top decile of exit costs relative to other CEOs in the top decile of severance pay.

Table 3 Panel B.

SEVERANCE = Variables	RSEVERANCE Coef. (t-stat)	%SEVERANCE Coef. (t-stat)
BADNEWS	0.149*** (14.733)	0.040 (1.166)
SEVERANCE	−0.035** (−2.092)	−0.076** (−2.143)
BADNEWS×SEVERANCE	0.141*** (8.868)	0.223*** (5.449)
TOPEXITCOST	−0.001 (−0.111)	−0.057 (−1.502)
TOPEXITCOST×BADNEWS	0.037*** (2.707)	0.130*** (2.741)
TOPEXITCOST×SEVERANCE	0.013 (0.659)	0.077* (1.709)
TOPEXITCOST × BADNEWS × SEVERANCE	−0.070*** (−3.145)	−0.159*** (−2.822)
CONSTANT	0.047 (0.500)	0.026 (0.260)
Controls	Yes	Yes
Firm, year, and industry fixed effects	Yes	Yes
N	46,750	46,750
R ²	.091	.090

Note. Panel B of Table 3 presents regression results of estimating Equation 2 where TOPEXITCOST equals one, zero otherwise if EXITCOST is in the top decile. For each regression, the estimated coefficients are presented with t-statistics in the brackets to the right. Please refer to Appendix A for sample selection criteria and Appendix B for variable definitions. ***, **, and * indicate significance (two-tailed) at the 1%, 5%, and 10% levels, respectively.The lines are highlighted in bold to emphasize they are the coefficients of interest.

Panel C of Table 3 reports results from estimating Equation 2 using RSEVERANCE and the individual measures that constitute our main measure of exit costs (UNCONNECTED, FORFEIT, and LOCALLINK).¹⁷ The coefficient on EXITCOST × BADNEWS × SEVERANCE is significantly negative across all measures of exit costs except LOCALLINK, suggesting that the results presented for the composite measure of exit costs are not driven by any one individual measure.¹⁸ Moreover, the magnitude of the coefficient using the composite measure of exit costs (−0.411) is larger than the magnitude of either of the individual measures (largest is −0.288). This result suggests the three components are jointly more potent, thereby reinforcing managerial exit costs as a combined measure relative to their individual levels.

Table 3 Panel C.

SEVERANCE = EXITCOST = Variables	RSEVERANCE UNCONNECTED Coef. (t-stat)	RSEVERANCE FORFEIT Coef. (t-stat)	RSEVERANCE LOCALLINK Coef. (t-stat)
BADNEWS	0.128*** (8.525)	−0.014 (−0.309)	0.167*** (23.002)
SEVERANCE	−0.060** (−2.560)	−0.066 (−0.994)	−0.027** (−2.005)
BADNEWS×SEVERANCE	0.180*** (8.088)	0.327*** (4.156)	0.105*** (9.026)
EXITCOST	−0.021 (−0.673)	−0.054 (−1.018)	−0.167 (−0.598)
EXITCOST×BADNEWS	0.079*** (3.201)	0.239*** (3.988)	0.161 (0.524)
EXITCOST×SEVERANCE	0.079*** (3.201)	0.055 (0.711)	0.078 (0.204)
EXITCOST × BADNEWS × SEVERANCE	−0.156*** (−3.911)	−0.288*** (−3.103)	−0.030 (−0.068)
Constant	0.028 (0.292)	0.088 (0.885)	0.038 (0.405)
Controls	Yes	Yes	Yes
Firm, year, and industry fixed effects	Yes	Yes	Yes
N	46,750	46,750	46,750
R ²	.091	.091	.091

Note. Panel C of Table 3 presents regression results of estimating Equation 2 based on individual measures of exit costs. For each regression, the estimated coefficients are presented with t-statistics in the brackets to the right. Please refer to Appendix A for sample selection criteria and Appendix B for variable definitions. ***, **, and * indicate significance (two-tailed) at the 1%, 5%, and 10% levels, respectively.The lines are highlighted in bold to emphasize they are the coefficients of interest.

Additional Analyses and Robustness Tests

Effect of Managerial Exit Costs on Ex Ante Severance Pay

A potential concern with our main tests is that managerial exit costs might determine the level of ex ante severance pay. Hence, the interaction between EXITCOST and SEVERANCE might simply reflect different levels of SEVERANCE rather than an incremental effect of SEVERANCE attributable to variations in EXITCOST. Controlling for the level of severance in our regression models does not necessarily address this concern if the association between exit costs and severance is nonlinear. Thus, we split our sample into quartiles of EXITCOST and estimate Equation 1 within each quartile. This process effectively groups firms into subsamples within which SEVERANCE is independent of EXITCOST.

In untabulated analysis using RSEVERANCE and %SEVERANCE, the coefficient on SEVERANCE × BADNEWS is significantly positive only in the bottom three quartiles of EXITCOST. The insignificant coefficient in the highest quartile of EXITCOST suggests that with sufficiently high exit costs, severance pay does not result in timelier disclosure of bad news. F-tests confirm that the coefficient on SEVERANCE × BADNEWS in the highest EXITCOST quartile is significantly lower than all other EXITCOST quartiles while there is no significant difference in the coefficient across the three lowest EXITCOST quartiles. This provides additional evidence that as managerial exit costs increase, the effect of ex ante severance pay on timely disclosure of bad news becomes less pronounced.

Identification

To improve identification, we consider a scenario in which we expect our predicted effect to be likely stronger (or weaker) and test whether this is indeed the case (DeFond & Zhang, 2014; Rajan & Zingales, 1998). Specifically, we expect the effect of managerial exit costs on the association between severance pay and bad news disclosure timeliness to be weaker when outside career opportunities are bleak. The most recent financial crisis decreased outside career opportunities for managers as the financial health of potential employers suffered (Jenter & Kanaan, 2015). This decline in the labor market reduces managers’ confidence about securing outside employment upon employment termination. We expect that managers have stronger incentives to keep their current jobs during the financial crisis, as exit costs for all managers would be higher during this period of reduced employment options in the case of termination.

We separate our firm-quarters into three distinct periods—before, during, and after the financial crisis—and estimate Equation 2 in each period with the results presented in Table 4. The crisis period represents heightened exit costs for all managers due to fewer external employment options, and the effect of the crisis is not captured in our EXITCOST measure. Consistent with all CEOs facing high managerial exit costs, we find that the coefficient on EXITCOST × BADNEWS × SEVERANCE is insignificant during the crisis period. That is, we find no difference in bad news withholding across differences in our measured exit costs when all CEOs face high exit costs from reduced employment options. However, consistent with H1, this coefficient is significantly negative both before and after the crisis period, although the magnitude of the coefficient is smaller in the postcrisis period. These results suggest that the effect of managerial exit costs on the association between severance pay and bad news disclosure timeliness is weakened when managers lack outside career opportunities or when retention incentives at the current employer are greater. Overall, we find consistent evidence that the effect of managerial exit costs varies predictably, thereby improving the identification of our main results.

Table 4.

Identification Test Results for Severance and Timeliness of Bad News Disclosures.TIMELY_i,t = α₀+α₁SEVERANCE_i,t+α₂BADNEWS_i,t+α₃BADNEWS_i,t×SEVERANCE_i,t+α₄EXITCOST_i,t+α₅EXITCOST_i,t×SEVERANCE_i,t+α₆EXITCOST_i,t×BADNEWS_i,t+α₇EXITCOST_i,t×BADNEWS_i,t×SEVERANCE_i,t+β_nControls_i,t+ϵ_i,t (2)

SEVERANCE = Period = Variables	Pre-crisis (2006–2007)Coef. (t-stat)	During crisis (2008–2010)Coef. (t-stat)	Post-crisis (2011–2014)Coef. (t-stat)
BADNEWS	−0.065 (−0.858)	0.148*** (2.635)	−0.007 (−0.137)
SEVERANCE	−0.132 (−0.586)	−0.302*** (−2.602)	−0.137 (−1.497)
BADNEWS×SEVERANCE	0.442*** (3.263)	0.254*** (2.700)	0.365*** (4.348)
EXITCOST	−0.322 (−1.203)	−0.281** (−2.041)	−0.026 (−0.281)
EXITCOST×BADNEWS	0.386*** (3.252)	0.014 (0.165)	0.280*** (3.726)
EXITCOST×SEVERANCE	0.307 (0.867)	0.513*** (2.863)	0.090 (0.697)
EXITCOST × BADNEWS × SEVERANCE	−0.561*** (−2.639)	−0.221 (−1.524)	−0.416*** (−3.487)
Constant	0.005 (0.017)	0.162 (0.601)	−0.127 (−0.691)
Controls	Yes	Yes	Yes
Firm, year, and industry fixed effects	Yes	Yes	Yes
N	8,253	19,084	22,725
R ²	.092	.081	.092

Note. Table 4 presents regression results of estimating Equation 2 before, during, and after the financial crisis with RSEVERANCE as the dependent variable. For each regression, the estimated coefficients are presented with t-statistics in the brackets to the right. Please refer to Appendix A for sample selection criteria and Appendix B for variable definitions. ***, **, and * indicate significance (two-tailed) at the 1%, 5%, and 10% levels, respectively.The lines are highlighted in bold to emphasize they are the coefficients of interest.

Alternative Measure of Bad News Withholding

To lend more credence to our main results, we follow the approach in Bao et al. (2018) and use the level of short interest as another comprehensive measure of bad news withholding. This measure captures the existence and extent of managers’ private bad news (Khan & Lu, 2013; Massa et al., 2015). Prior research shows high short interest reliably predicts negative information that is held by the manager but yet to be reflected in stock prices. High short interest is associated with negative future stock returns, future stock price crash risk, negative earnings surprises, and analyst downgrades (Callen & Fang, 2015; Christophe et al., 2004, 2010; Desai et al., 2002). These results support the notion that managers of firms with higher short interest hide and accumulate bad news before releasing it all at once.

Bao et al. (2018) document a negative relation between residual short interest at the end of a quarter and the frequency of bad news disclosure in the subsequent quarter. This evidence is consistent with managers of firms with higher residual short interest, on average, possessing and withholding more negative private information than do those of firms with lower residual short interest. Following their approach, we measure bad news withholding using the residual short interest (SHORT) obtained from regressing short interest on institutional ownership and convertible debt. We reexamine H1 using this alternative measure by estimating the following equation:

\begin{matrix} SHOR T_{i, t} = & α_{0} + α_{1} SEVERANC E_{i, t} + α_{2} BADNEW S_{i, t} + α_{3} BADNEW S_{i, t} \\ \times SEVERANC E_{i, t} + α_{4} EXITCOS T_{i, t} + α_{5} EXITCOS T_{i, t} \\ \times SEVERANC E_{i, t} + α_{6} EXITCOS T_{i, t} \times BADNEW S_{i, t} \\ + α_{7} EXITCOS T_{i, t} \times BADNEW S_{i, t} \times SEVERANC E_{i, t} \\ + β_{n} Control s_{i, t} + ε_{i, t} \end{matrix}

(3)

Our untabulated results show that the coefficient on EXITCOST × BADNEWS × SEVERANCE is significant and positive using both measures of SEVERANCE.¹⁹ Hence, our main results presented in Table 3 are robust to this alternative measure of bad news timeliness.

Sensitivity Tests

Our timeliness measure may contain noise because analysts’ forecasts do not always monotonically increase over the forecast period in the direction of the news for the period. Moreover, for some firms, more than 100% of the total news of the quarter was revealed on some days during the quarter (due to oscillations of the consensus). We employ several exclusion criteria to address these concerns. First, we delete observations where the proportion of news revealed was greater than 1 for more than 10% of the quarter (3,677 observations where the consensus during the quarter moved further than the ultimate news revealed during the quarter). Second, we delete observations where relative to the beginning consensus, the daily consensus was in the opposite direction of the total news for the quarter for more than 10% of the days in the quarter (11,314 observations where the consensus went up or down early in the quarter before ultimately going the opposite direction by the end of the quarter). Third, we exclude any observation where the consensus on more than 10% of the days in the quarter moves further from, rather than closer to, the actual earnings for the quarter (16,110 observations). We apply these exclusion criteria separately and estimate Equation 2 in unreported analyses. Using RSEVERANCE, we find that consistent with our prediction, the coefficient on EXITCOST×BADNEWS×SEVERANCE in these analyses is −0.325, −0.191, and −0.203, respectively. We find similar results in sign and statistical significance when we use %SEVERANCE. Hence, our main results are robust to addressing several nuances in the way news is revealed in the forecast period.

Intuitively, analyst-specific biases have a stronger effect on our timeliness measure when only a few analysts are issuing forecasts for a given firm. To mitigate the effect of such biases, we require a minimum of three analysts per firm (42,105 observations), and re-estimate Equation 2. In untabulated analysis, we find that consistent with our prediction, the coefficient on EXITCOST × BADNEWS×SEVERANCE is significant and negative. We find similar results in sign and statistical significance when we use %SEVERANCE, consistent with our main results.

Finally, our main measure of bad news disclosure timeliness is advantageous because it reflects various channels managers use to disclose private information. This measure is potentially confounded by analysts’ incentives and the interplay between management incentives and analysts’ incentives. In our main tests, we control for these potential confounds using a battery of analysts’ characteristics. Nevertheless, managers may “walk down” analyst forecasts over the quarter to meet or beat the consensus as the end of the quarter (Richardson et al., 2004). To address the potential effect of managers walking down analyst forecasts on our results, we exclude all quarters where the consensus at the end of the quarter is less than the consensus at the beginning of the quarter and re-estimate Equation 2. In untabulated analyses, we find the coefficient on EXITCOST×BADNEWS×SEVERANCE in this reduced sample is significantly negative, consistent with our main results.

Conclusion

Despite the public criticism about severance contracts, extant research suggests that severance pay curtails agency problems and enhances firm value (Baginski et al., 2018; Cadman et al., 2016; Chen et al., 2015). Nevertheless, recent theory posits that whether severance pay constitutes part of an efficient incentive contract is contingent on the implicit and explicit costs (forgone benefits) managers incur at dismissal (Laux, 2015). Referring to these financial and nonfinancial consequences triggered by employment termination as managerial exit costs, we examine whether and how exit costs affect the effectiveness of severance pay in curbing managers’ bad news withholding. Consistent with our expectation, we find that the positive association between ex ante severance pay and timeliness of bad news disclosure becomes less pronounced as managerial exit costs increase. Furthermore, we document that when managerial exit costs reach a sufficiently high level (i.e., the top quartile), severance pay no longer positively relates to timelier disclosure of bad news relative to good news. These results suggest that although severance pay has potential value for investors, managerial exit costs undermine the efficacy of severance pay in curbing the withholding of bad news.

Our study expands the scope of the research on the incentive effects of ex ante severance pay by identifying a scenario where the efficacy of severance pay as a governance mechanism to alleviate agency conflicts is significantly weakened—when managers face greater financial and nonfinancial losses upon leaving their current firm. Our study also adds to the growing research on the role of ex ante severance pay in corporate financial reporting practices. Whereas this literature in general supports a positive association between severance pay and financial reporting quality, our findings suggest that managers trade off the predetermined financial benefit created by severance pay and the explicit and implicit costs triggered by employment termination when they determine corporate disclosure transparency. In addition, our findings have timely and practical implications about the design of executive compensation contracts. Our result suggests that boards should tailor severance agreements by carefully and regularly considering the exit costs of individual executives rather than taking the “one-size-fits-all” approach that undermines the potential of severance pay to serve the interests of investors.

Supplemental Material

Online_Appendix – Supplemental material for Does Ex Ante Severance Pay Affect the Timeliness of Bad News Disclosure? The Role of Managerial Exit Costs

Supplemental material, Online_Appendix for Does Ex Ante Severance Pay Affect the Timeliness of Bad News Disclosure? The Role of Managerial Exit Costs by Herita Akamah, Bryan Brockbank and Sydney Qing Shu in Journal of Accounting, Auditing & Finance

Footnotes

Appendix

Appendix B.

Variable Definitions.

Dependent variable
TIMELY	= Average proportion of a quarter’s total news that has been revealed by each day during the quarter, from Donelson et al. (2012). For each day in the quarter, we calculate the proportion of news that has been revealed up to that day. For example, if the consensus forecast on the first day of a quarter is $1.00 per share and the actual earnings are $1.50 per share, then the implied good news for the quarter is $0.50 per share. If on a given day, the consensus forecast is $1.20 per share, then 40% of the quarter’s news has been revealed by that day [($1.20-$1.00)/($1.50-$1.00)]. TIMELY is the average of these daily proportions of revealed news over the entire quarter. See the Online Appendix A for detailed example of this calculation.
SHORT	= The residual from regressing short interest on institutional ownership and convertible debt for firms with stock price greater than $1 (Bao et al 2018).
Independent variables
RSEVERANCE	= Decile ranking of the CEO’s severance pay, scaled so values range from 0 to 1.
%SEVERANCE	= Decile ranking of the ratio of the CEO’s severance pay to his/her total compensation, scaled so values range from 0 to 1.
BADNEWS	= 1 (0 otherwise) if actual EPS is lower than the consensus forecast on the first day of the quarter (one day after the previous quarter earnings announcement).
UNCONNECTED	= Average decile ranking of the following, scaled to range from 0 to 1: (1) The number of companies a CEO has been affiliated with as an employee or member of the board of directors (multiplied by minus one). (2) The number of words in the CEO’s biography on CIQ (multiplied by minus one). (3) The number of affiliated companies visible enough to have available revenue information on CIQ (multiplied by minus one).
FORFEIT	= The scaled decile ranking of CEO compensation that would be forfeited upon ending employment (e.g., restricted stock, unvested stock and stock options).
LOCALLINK	= The number of board positions held by the CEO for companies headquartered in the same state as the CEO’s firm, scaled (out of sample) so values range from 0 to 1.
EXITCOST	= Sum of UNCONNECTED, FORFEIT, and LOCALLINK, scaled to range between 0 and 1.
Control variables
FOUNDER	= 1 (0 otherwise) if the CEO is the founder of the firm.
BLOCKHOLDER	= 1 (0 otherwise) if the CEO’s equity ownership is at least 5%.
CHAIRMAN	= 1 (0 otherwise) if the CEO is the chair of the board.
LAGE	= Natural logarithm of the CEO’s age (missing values replaced with zero).
TENURE	= Natural logarithm of the number of years the CEO has served as the firm’s CEO during our sample period (2006 to 2014).

LOGMV	= Natural log of the market value of equity as of the end of year t.
BIG4	= 1 (0 otherwise) if the firm is audited by one of the Big 4 accounting firms.
RETURNVOL	= Natural logarithm of the standard deviation of the firm’s daily stock return over year t.

LOGCOMP	= Natural logarithm of the CEO’s total annual compensation.
NEWS	= The amount of news that is revealed during quarter t. Calculated as actual EPS minus the consensus EPS forecast on the first day of the quarter, scaled by beginning of quarter stock price.


LOGANALYSTS	= Natural log of the number of analysts providing EPS forecasts for the firmin quarter t.

LOGFORECASTS	= Natural log of the number of EPS forecasts for the firm in quarter t.
LOGNUMDAYS	= Natural log of the number of days in quarter t. The quarter begins on the first day after the previous quarter’s earnings announcement for which analyst forecasts are issued to calculate the consensus.


SURPRISE	= Earnings surprise for the firm in quarter t, calculated as the actual EPS minus the consensus on the date of the earnings announcement.

FORECASTFREQ	= The average frequency of forecasts issued by analysts following the firm in quarter t.

FIRMEXPERIENCE	= The average number of years that the analysts issuing forecasts for the firm in quarter t have issued forecasts for that firm.

TOTEXPERIENCE	= The average number of years that the analysts issuing forecasts for the firm in quarter t have issued forecasts for any firm.

FIRMSFOLLOW	= The average number of firms followed by the analysts issuing forecasts for the firm in quarter t.

Note. Appendix B provides the definitions of the key variables used in the empirical analysis. EPS = earnings per share; CEO = chief executive officers; CIQ = Capital IQ.

Acknowledgements

We gratefully acknowledge the contributions of the Editor-in-Chief Bharat Sarath, the associate editor Carolyn Levine, and an anonymous reviewer in helping to improve our article. We appreciate helpful comments from workshop participants at the University of Oklahoma.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Bryan Brockbank

Sydney Qing Shu

Data Availability

Data are available from the sources cited in the text.

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Almazan

Suarez

(2003). Entrenchment and severance pay in optimal governance structures. The Journal of Finance, 58(2), 519–547.

Altschuler

Chen

Zhou

(2015). Anticipation of management forecasts and analysts’ private information search. Review of Accounting Studies, 20(2), 803–838.

Baginski

S. P.

Campbell

J. L.

Hinson

L. A.

Koo

D. S.

(2018). Do career concerns affect the delay of bad news disclosure? The Accounting Review, 93(2), 61–95.

Balsam

Miharjo

(2007). The effect of equity compensation on voluntary executive turnover. Journal of Accounting and Economics, 43(1), 95–119.

Bao

Kim

Mian

G. M.

(2018). Do managers disclose or withhold bad news? Evidence from short interest. The Accounting Review.

Beyer

(2008). Financial analysts’ forecast revisions and managers’ reporting behavior. Journal of Accounting and Economics, 46(2–3), 334–348.

Brown

K. E.

(2015). Ex ante severance agreements and earnings management. Contemporary Accounting Research, 32(3), 897–940.

Bushee

B. J.

Gow

I. D.

Taylor

D. J.

(2018). Linguistic complexity in firm disclosures: Obfuscation or information?. Journal of Accounting Research, 56(1), 85–121.

Cadman

B. D.

Campbell

J. L.

Klasa

(2016). Are ex-ante CEO severance pay contracts consistent with efficient contracting? Journal of Financial and Quantitative Analysis, 51(3), 737–769.

10.

Callen

J. L.

Fang

(2015). Short interest and stock price crash risk. Journal of Banking and Finance, 60, 181–194.

11.

Chen

Cheng

A. K.

Wang

(2015). CEO contractual protection and managerial short-termism. The Accounting Review, 90(5), 1871–1906.

12.

Christophe

S. E.

Ferri

M. M.

Angel

J. J.

(2004). Short-selling prior to earnings announcments. Journal of Finance, 59(4), 1845–1876.

13.

Christophe

S. E.

Ferri

M. M.

Hsieh

(2010). Informed trading before analyst dowgrades: Evidence from short seller. Journal of Financial Economices, 95(1), 85–106.

14.

Cowen

A. P.

King

A. W.

Marcel

J. J.

(2016). CEO severance agreements: A theoretical examination and research agenda. Academy of Management Review, 41(1), 151–169.

15.

Dash

(2007, April 8). Has the exit sign ever looked so good? The New York Times. https://www.nytimes.com/2007/04/08/business/yourmoney/08axe.html.

16.

DeFond

Zhang

(2014). A review of archival auditing research. Journal of Accounting and Economics, 58(2–3), 275–326.

17.

Desai

Ramesh

Thiagarajan

Balachandran

B. V. B.

(2002). An investigation of the informaitoanl role of short interest in the NASDAD market. Journal of Finance, 57(5), 2263–2287.

18.

Deutsch

C. H.

(2005, April 3). My big fat C.E.O. paycheck. The New York Times.

19.

Donelson

D. C.

McInnis

J. M.

Mergenthaler

R. D.

(2012). The timeliness of bad earnings news and litigation risk. The Accounting Review, 87(6), 1967–1991.

20.

Edmans

Gabaix

Sadzik

Sannikov

(2012). Dynamic CEO compensation. The Journal of Finance, 67(5), 1603–1647.

21.

Engelberg

Gao

Parsons

C. A.

(2013). The price of a CEO’s rolodex. Review of Financial Studies, 26(1), 79–114.

22.

Finkelstein

(1992). Power in top management teams: Dimensions, measurement, and validation. Academy of Management Journal, 35(3), 505–538.

23.

Gibbons

Murphy

K. J.

(1992). Optimal incentive contracts in the presence of career concerns: Theory and evidence. Journal of Political Economy, 100(3), 468–505.

24.

Goldman

Huang

(2011). Contractual versus actual severance pay following CEO turnover [Working paper]. Indiana University.

25.

Gopalan

Milbourn

Song

Thakor

A. V.

(2014). Duration of executive compensation. The Journal of Finance, 69(6), 2777–2817.

26.

(2012). Dynamic compensation contracts with private savings. Review of Financial Studies, 25(5), 1494–1549.

27.

Hermalin

B. E.

Weisbach

M. S.

(2007). Transparency and corporate governance (NBER Working Paper No. 12875). 195–233.

28.

Inderst

Mueller

H. M.

(2010). CEO replacement under private information. Review of Financial Studies, 23(8), 2935–2969.

29.

Jenter

Kanaan

(2015). CEO turnover and relative performance evaluation. The Journal of Finance, 70(5), 2155–2184.

30.

Jochem

Ladika

Sautner

(2018). The retention effects of unvested equity: Evidence from Accelerated Option Vesting. The Review of Financial Studies, 31(11), 4142–4186.

31.

Kalfen

(2012). Key findings in the 2012/2013 study of executive severance arrangements. Meridian Compensation Partners, LLC.

32.

Kasznik

Lev

(1995). To warn or not to warn: Management disclosures in the face of an earnings surprise. The Accounting Review, 70(1), 113–134.

33.

Khan

(2013). Do short sellers front-run insider sales? The Accounting Review, 88(5), 1743–1768.

34.

Kothari

S. P.

Shu

Wysocki

P. D.

(2009). Do managers withhold bad news. Journal of Accounting Research, 47(1), 241–276.

35.

Kumar

(2016). Are massive severance packages to fired CEOs justified ? Boardroom Memo. https://scroll.in/article/824414/are-massive-severance-packages-to-fired-ceos-justified.

36.

Laux

(2012). Stock option vesting conditions, CEO turnover, and myopic investment. Journal of Financial Economics, 106(3), 513–526.

37.

Laux

(2015). Executive pay, innovation, and risk-taking. Journal of Economics and Management Strategy, 24(2), 275–305.

38.

Lee

(2014, April 17). Short stint, big payout: View CEO severance packages. USA Today Network. https://www.usatoday.com/story/news/nation-now/2014/04/17/ceos-severance-packages-yahoo/7826743/.

39.

Levitt

Snyder

(1997). Is no news bad news? Information transmission and the role of early warning in the principal-agent model. The Journal of Economics, 28(4), 641–661.

40.

Ling

(2012). Ex ante severance agreements and timely disclosure of bad news. Journal of Accounting, Auditing & Finance, 27(2), 177–207.

41.

Liu

(2014). Outside options and CEO turnover: The network effect. Journal of Corporate Finance, 28, 201–217.

42.

Massa

Qian

Zhang

(2015). Competition of the informed: Does the presence of short sellers affect insider selling? Journal of Financial Economices, 118(2), 268–288.

43.

Oyer

Schaefer

(2005). Why do some firms give stock options to all employees: An examination of alternative theories. Journal of Financial Economics, 76(1), 99–133.

44.

Pae

Song

C. J.

A. C.

(2016). Career concerns and management earnings guidance. Contemporary Accounting Research, 33(3), 1172–1198.

45.

Priyan

D. C.

(2016, May 20). Is the proposed $55 million sevearnce package for Yahoo chief Marissa Myer Justified? Linkedin.com.

46.

Rajan

R. G.

Zingales

(1998). Financial dependence and growth. The American Economic Review, 88(3), 559–586.

47.

Rau

P. R.

(2013). How do ex ante severance pay contracts fit into optimal executive incentive schemes? Journal of Accounting Research, 51(3), 631–671.

48.

Renneboog

Zhao

(2011). Us knows us in the UK: On director networks and CEO compensation. Journal of Corporate Finance, 17(4), 1132–1157.

49.

Richardson

Teoh

S. H.

Wysocki

P. D.

(2004). The walk-down to beatable analyst forecasts: The role of equity issuance and insider trading incentives. Contemporary Accounting Research, 21(4), 885–924.

50.

Skinner

(1994). Why firms voluntarily disclose bad news. Journal of Accounting Research, 32(1), 38–61.

51.

Skinner

(1997). Earnings disclosures and stockholder lawsuits. Journal of Accounting and Economics, 23(3), 249–283.

52.

Stewart

J. B.

(2011, September 30). Rewarding C.E.O.’s who fail. The New York Times. https://www.nytimes.com/2011/10/01/business/lets-stop-rewarding-failed-ceos-common-sense.html.

53.

Tuttle

(2015, September 11). Why disgraced CEOs get insanely generous payout packages when they are filed. Mondy.com.

54.

Verrecchia

R. E.

(1991). Essays on disclosure. Journal of Accounting and Economics, 32(1-3), 97–180.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.26 MB