How Does Intentional Earnings Smoothing Vary With Managerial Ability?

Abstract

We investigate if high-ability managers are more likely to intentionally smooth earnings, a form of earnings management, and when they are more likely to do so. Although prior studies provide evidence that high-ability managers report higher quality earnings, the literature does not indicate whether this behavior is common because of (or happens in spite of) high-ability managers’ intentional smoothing activities. We find that (a) high-ability managers are significantly more likely to engage in intentional smoothing, (b) their intentional smoothing is associated with improved future operating performance, and (c) their intentional smoothing is more prevalent when the smoothing either benefits shareholders, the manager, or both. We do not, however, find evidence that high-ability managers who smooth are more likely to have engaged in informed trading or are more likely to consume perquisites. High-ability managers’ intentional smoothing is also associated with increased voluntary (but not forced) executive turnover, consistent with high-ability managers being motivated, at least in part, by how the capital market consequences of smoothing are expected to benefit shareholders, thereby bolstering their reputation.

Keywords

managerial ability earnings smoothing accruals earnings management real earnings management

Introduction

Prior research provides substantial evidence that high-ability managers generate more accurate future earnings forecasts and more effectively implement their chosen strategies than lower ability managers (Baik, Farber, & Lee, 2011; Bertrand & Schoar, 2003; Demerjian, Lev, & McVay, 2012; Holcomb, Holmes, & Connelly, 2009). These skills are the building blocks that underpin the superior earnings quality reported by high-ability managers (Aier, Comprix, Gunlock, & Lee, 2005; Demerjian, Lev, Lewis, & McVay, 2013). Yet, these same skills can also facilitate earnings management, including earnings smoothing. Earnings smoothing requires managers to accurately forecast future earnings, and then increase or decrease current income to both reduce earnings volatility and generate future reporting slack (DeFond & Park, 1997). Thus, we investigate if high-ability managers are more likely than other managers to intentionally smooth earnings.¹ We also explore the future performance consequences and incentives associated with high-ability managers’ intentional smoothing.

Ex ante, the relation between managerial ability and intentional smoothing is unclear. We expect that high-ability managers will have a better understanding of the trend line around which to smooth because they are able to generate better earnings forecasts (Baik et al., 2011; Beidleman, 1973; DeFond & Park, 1997; Moses, 1987). As a result, they are likely more capable of reporting smooth earnings than low-ability managers. We also expect that high-ability managers are better able to identify adjustments that smooth earnings at a lower cost than other managers, thereby maximizing the net benefits of intentionally smoothed earnings.²

Although we expect that high-ability managers have the ability to smooth earnings more effectively than other managers, this innate ability does not mean that they will necessarily choose to intentionally smooth. First, it is possible that high-ability managers more effectively manage their companies, obviating the need for discretionary smoothing. Second, even with the need to intentionally smooth, higher ability managers may opt not to do so. For example, if high-ability managers have valuable reputations, which garner them greater lifelong compensation (e.g., Demerjian et al., 2012; Fee & Hadlock, 2003), and intentional smoothing could harm their reputations, then they may avoid intentional smoothing.³

If high-ability managers intentionally smooth earnings, why they choose to do so is an empirical question. Smoothing could benefit the firm by improving the quality of information for outside users, for example, by bringing reported earnings closer to permanent earnings. Alternatively, high-ability managers could use smoothing for extractive purposes, such as claiming perquisites from the firm. For managers to choose smoothing for extractive purposes, however, the benefits would need to outweigh any reputational or litigation costs.

Understanding the relation between managerial ability and intentional smoothing is important for at least two reasons: First, boards of directors design compensation contracts to elicit desired actions from executives (Holmstrom, 1979), so it is important for them to know whether intentional smoothing by high-ability managers should be encouraged or discouraged. Second, as prior research provides evidence that earnings reported by high-ability managers are of higher quality (Aier et al., 2005; Demerjian et al., 2013), it is vital for capital providers to understand whether this relation is a result of (or happens in spite of) intentional smoothing.

At a conceptual level, intentional smoothing is managers’ deliberate use of accrual and real earnings management to reduce earnings variability over time. Although prior archival research has focused on accrual-based smoothing activities, analytical smoothing models and survey evidence indicate that managers use both accrual and real earnings management to smooth earnings (Acharya & Lambrecht, 2015; Graham, Harvey, & Rajgopal, 2005; Lambert, 1984). Thus, our measure of intentional smoothing considers both types of earnings management. We find that more able managers are more likely to intentionally smooth earnings, consistent with the notion that these managers have the confidence and technical expertise to undertake this complex reporting strategy.⁴ Our evidence suggests that our measure captures both upward and downward adjustments, leading to smoother earnings rather than systematically overstated or understated earnings. We use a firm fixed effects research design, which allows us to assess the effect of different managers in the same firm over time, as well as minimize the effect of time-invariant firm features. To further reduce concerns about endogeneity, we estimate a two-stage least-squares regression, which yields consistent results.⁵ These analyses indicate that our results are attributable to variation in managerial ability, and that intentional smoothing is significantly greater among high-ability managers than among lower ability managers.

Because prior research provides evidence that high-ability managers are more effective at implementing financing and investing strategies than low-ability managers (Bertrand & Schoar, 2003; Demerjian et al., 2012; Holcomb et al., 2009), we also expect that when high-ability managers implement intentional smoothing strategies, their techniques are more effective than those of lower ability managers. To assess effectiveness, we examine the future operating benefits/costs of intentional smoothing, and the results are consistent with high-ability managers’ smoothing being less costly (or more beneficial) than smoothing by other managers.

Finally, we examine if intentional smoothing varies with incentives to smooth. To start, we document an increase in intentional smoothing in response to incentives. This cross-sectional variation provides some assurance that our analyses capture intentional intervention into the financial reporting system, rather than natural smoothing due to the neutral application of generally accepted accounting principles (e.g., Dechow & Skinner, 2000). We then examine the types of incentives that influence high-ability managers’ intentional smoothing. This analysis contributes to our understanding of the underlying motives for high-ability managers’ intentional smoothing. We find evidence that high-ability managers intentionally smooth earnings when the firm is near debt covenant thresholds and when managers are younger (our measure of reputation incentives).⁶ We do not, however, find evidence of high-ability managers smoothing for personal insider trading gains, or when the potential for consumption of perquisites is greater. Finally, we examine subsequent executive turnover, and find a positive association between high-ability managers’ intentional smoothing activities and the incidence of voluntary executive turnover, but no association with forced executive turnover. In summary, these results indicate that high-ability managers’ intentional smoothing is motivated, at least in part, by how they expect the capital market benefits of smoothing to benefit shareholders and bolster their own reputation.

Our results offer insights into the two opposing viewpoints of earnings smoothing: whether (on average) it (a) benefits shareholders (e.g., Badertscher, Collins, & Lys, 2012; Subramanyam, 1996; Tucker & Zarowin, 2006) or (b) obfuscates performance to facilitate perquisite consumption to the detriment of shareholders (e.g., Lang, Lins, & Maffett, 2012; Leuz, Nanda, & Wysocki, 2003; Levitt, 1998). Our results suggest that the ability to effectively smooth varies across managers, and that intentional smoothing by high-ability managers is a low-cost mechanism to help the firm avoid debt covenant violations, but it does not appear to facilitate insider trading or perquisite consumption. This evidence should be informative to boards as they assess the value of management and consider the desirability of intentional smoothing.

In addition, our analyses inform the academic debate on whether intentionally smoothed earnings represent high-quality earnings. Our results suggest that, when undertaken by managers with sufficient skill, smoothing leads to earnings that more consistently reach important earnings thresholds and that are associated more positively with future operating performance. To the extent that these earnings characteristics are important to shareholders, our results suggest that intentional smoothing improves earnings quality. Finally, our model of intentional smoothing broadens the scope of smoothing to consider real earnings management and should be useful to researchers examining intentional earnings smoothing.

Motivation and Hypothesis Development

Intentional Earnings Smoothing

At the construct level, intentional smoothing is management’s purposeful intervention into the firm’s operating and reporting processes to reduce the volatility of reported earnings (Acharya & Lambrecht, 2015; Beidleman, 1973; Graham et al., 2005; Lambert, 1984; Moses, 1987). Beidleman (1973) describes it as the intentional dampening of earnings fluctuations over time, and Graham et al. (2005) note that 78% of executives say that they would “give up economic value in exchange for smooth earnings” (p. 5). Thus, intentional smoothing reflects the ongoing and overtime use of income-increasing and income-decreasing accrual and real activities earnings management to reduce the volatility of reported earnings.

Some research provides evidence that intentionally smoothed earnings represent a deliberate distortion of reported performance (e.g., DeFond & Park, 1997; Lang et al., 2012; Leuz et al., 2003). This literature suggests that earnings smoothing distorts firm performance measures and, in general, benefits the manager at the expense of other stakeholders. An alternative motive for intentional smoothing is to improve the usefulness of earnings. Many analytical studies support this view, showing that smooth earnings are more informative and useful for contracting (Chaney & Lewis, 1995; Demski, 1998; Kirschenheiter & Melumad, 2002; Sankar & Subramanyam, 2001). A number of archival studies complement these models and provide evidence that smoothing improves the information content of earnings (e.g., Subramanyam, 1996; Tucker & Zarowin, 2006), and is associated with higher equity valuations when firms consistently meet the market’s earnings expectations (Barth, Elliott, & Finn, 1999; Kasznik & McNichols, 2002; Myers, Myers, & Skinner, 2007).⁷ This literature provides evidence that earnings smoothing enhances firm performance measures, improves contracting, and, in general, benefits stakeholders.

Intentional Earnings Smoothing and Managerial Ability

Our discussion has so far highlighted numerous motives for intentional smoothing, some of which benefit stakeholders and some of which benefit the manager at the expense of other stakeholders. We next discuss how we expect managerial ability to influence managers’ propensity to smooth in each situation.

Smoothing income toward recurring or permanent earnings is a complex task. Managers must first be able to forecast the firm’s future earnings, and then determine how to adjust reported earnings toward the forecast. As noted in DeFond and Park (1997), this process requires managers to increase or decrease current income both to achieve current period smoothness and to build up reporting slack to continue reporting smoothly in the future. An income smoothing strategy requires considerable foresight by managers, both to project the future conditions that the firm may face and to anticipate the reporting implications of those conditions.

We expect that high-ability managers are more able to make these projections and plan their reporting actions accordingly. The logic is similar to that of Demerjian et al. (2013), who contend that high-ability managers make superior estimates and judgments, and that these superior abilities are reflected in high-quality earnings.⁸ Similarly, Trueman (1986) and Baik et al. (2011) posit that more able managers are better able to forecast earnings and should thus be able to identify the appropriate trend around which to smooth.⁹ In addition, prior research suggests that high-ability managers are more effective at implementing chosen strategies than lower ability managers. For example, high-ability managers make better financing and investing decisions, and are less likely to restate earnings than lower ability managers (Aier et al., 2005; Bertrand & Schoar, 2003; Demerjian et al., 2013; Holcomb et al., 2009).

Reporting earnings that correspond more closely to economic performance is one of numerous possible benefits of intentional smoothing. Prior research also provides evidence of a market premium for firms that consistently beat earnings expectations (Barth et al., 1999; Kasznik & McNichols, 2002), where intentional smoothing is one path to this outcome (Myers et al., 2007), and reduced contracting costs for firms that intentionally smooth earnings (Demerjian, Donovan, & Lewis-Western, 2017). Chief financial officers (CFOs) indicate that they believe that meeting benchmarks is “very important,” and that “hitting earnings benchmarks builds credibility with the market” and increases the firm’s stock price (Graham et al., 2005, p. 5). If intentional smoothing reflects managers’ aims to report the smoothed earnings desired by stakeholders, then we expect that high-ability managers are better able to use smoothing to achieve the desired characteristics than lower ability managers whose forecasting skills and implementation acumen are weaker.

If intentional smoothing reflects opportunistic behavior, then the relation between managerial ability and intentional smoothing is unclear ex ante. If high-ability managers have valuable reputations, which, for example, garner them greater lifelong compensation (e.g., Demerjian et al., 2012; Fee & Hadlock, 2003), then they have an incentive to avoid reputation-harming behavior. High-ability managers’ skills may, however, facilitate opportunistic smoothing because they are able to implement opportunistic smoothing strategies with a lower likelihood of detection, thereby garnering private benefits through insider trading or perquisite consumption (e.g., Wang, 2016). Thus, even if intentional smoothing is opportunistic, high-ability managers may still be more likely to implement the strategy.

Overall, this discussion suggests numerous situations that motivate intentional smoothing, as well as that high-ability managers are better able to smooth earnings than lower ability managers. Thus, we expect that, on average, high-ability managers are more likely to engage in intentional smoothing. This logic leads to our first hypothesis, which we state in alternative form:

Hypothesis 1 (H1): Intentional smoothing by high-ability managers is significantly greater than that by other managers.

Regardless of the aim of high-ability managers’ intentional smoothing, we expect that high-ability managers are better able to assess the amount of expected slack (to rein in) or shortfall (to bump up) earlier in the period and also to implement strategies more effectively than lower ability managers, both of which will increase the net benefits of high-ability managers’ smoothing activities.¹⁰ Similar to the prior discussion, we expect high-ability managers to have a better understanding of the actual trajectory of earnings around which to smooth, whereas lower quality managers might project an unreasonable trajectory, creating a costly snowballing effect (e.g., Schrand & Zechman, 2012). Thus, we expect high-ability managers to implement a more effective intentional smoothing strategy than lower ability managers, leading to greater net benefits. This leads to our second hypothesis, which is stated in alternative form:

Hypothesis 2 (H2): The net future operating benefits of intentional smoothing are significantly greater for high-ability managers than for other managers.

We next consider if high-ability managers’ intentional smoothing increases in response to specific incentives. We expect that if, on average, managers deliberately smooth earnings, we should observe increases in smoothing in response to incentives. Prior research has offered numerous motives for earnings smoothing; we consider six of these incentives, described in Section “Incentives.” Some incentives (e.g., consumption of perquisites) are clearly not in shareholders’ best interests, while others may be (e.g., avoiding technical default). If high-ability managers use their ability to benefit themselves to the detriment of shareholders, we expect intentional smoothing to increase when it primarily benefits the manager (e.g., in the presence of informed insider trades). It is also possible, however, that management benefits from smoothing through reputational enhancements, which occur as the manager reports earnings that exceed important earnings benchmarks and increases her credibility with the market (Graham et al., 2005), or via increases in share price that benefit managers whose compensation is linked to firm value. If these incentives are the underlying motives for high-ability managers’ intentional smoothing, then we should observe increases in smoothing for high-ability managers at times when the smoothing is more likely to benefit shareholders, such as when it allows the firm to avoid a debt covenant violation, meet earnings expectations, or when executives’ compensation is more closely linked to firm value. Although the smoothing may still benefit the manager in these situations (via enhanced reputation or an increase in the value of equity compensation), the smoothing also benefits shareholders and would thus be desirable (or more desirable) than if opportunistic incentives motivate intentional smoothing. In summary, consideration of incentives provides additional evidence of intentional earnings management and offers insight into the motivation behind high-ability managers’ intentional smoothing. We state our third hypothesis in alternative form:

Hypothesis 3 (H3): In response to smoothing incentives, high-ability managers increase their intentional smoothing more than other managers.

Data, Variable Definitions, and Descriptive Statistics

We obtain our data from the Annual Compustat file for our intentional smoothing and control variables, Center for Research in Security Prices (CRSP) to form returns variables, Institutional Brokers’ Estimate System (I/B/E/S) for the consensus analyst forecast, ExecuComp for executive compensation data, Thomson Reuters for insider trades, and RiskMetrics for dual-class voting shares. We also obtain two publicly available datasets from researchers, managerial ability (Demerjian et al., 2012), and a list of U.S. firms with dual-class shares (Gompers, Ishii, & Metrick, 2010).¹¹

We begin with all Compustat firms with nonmissing assets. Following McNichols (2002), we exclude firm-years experiencing accounting changes, merger or acquisition activity, or discontinued operations.¹² To remain in our sample, we require firms to have information available to calculate managerial ability and Compustat data necessary for the calculation of our control variables, including the innate earnings quality variables, which require four prior years of data for calculation. To examine how managerial ability maps into intentional smoothing, we require that the same management team be in place during the period over which we measure managerial ability (t − 2 and t− 1) and smoothing (t − 2 to t). Thus, we exclude firms with executive turnover in the 3-year period from t − 2 to t.¹³ The period begins in 1995 because we require Securities and Exchange Commission (SEC) filings to be available electronically on Edgar to identify executive turnover, and ends in 2013 to allow for realizations of future operating performance. Our final sample consists of 13,153 firm-year observations and 3,523 firms. We summarize the sample selection procedure in Table 1.

Table 1.

Sample Selection.

Data requirement	Firm-years
Initial sample without M&A activity or discontinued operations and nonmissing assets, post 1994	103,690
Less firms with cash flows, sales, and returns information	(20,146)
Less firms with sufficient data to calculate other control variables, including innate earnings quality variables	(53,957)
Less firms with sufficient data to calculate managerial ability and intentional smoothing	(2,835)
Less firms with sufficient data to calculate future operating performance	(4,384)
Less firms without CEO or CFO turnover from years t− 2 to t	(9,215)
Final sample	13,153

Note. M&A = merger and acquisition.

Variable Definitions

Managerial ability

We base our assessment of managerial ability on the MA-Score, developed by Demerjian et al. (2012). They estimate the score in two stages: The first stage uses a frontier analysis method Data envelopment analysis (DEA) to provide an estimate of how efficiently managers use their firms’ resources (including capital, labor, and innovative assets) to generate revenues relative to their industry peers. The second stage uses regression analysis to purge firm-level drivers of efficiency. Demerjian et al. (2012) attribute the unexplained efficiency to the management team (see Demerjian et al., 2012, for details). In essence, high-quality managers generate more sales for a given level of inputs than lower quality managers. Demerjian et al. (2012) conduct numerous validity tests, concluding that their measure outperforms existing measures such as historical returns and media citations.

To identify high-ability managers, we first form quartiles (by industry and year) of the MA-Score.¹⁴ We define High-Ability Managers as those in the top quartile of MA-Score in both years t − 2 and t − 1. This approach reduces the likelihood that idiosyncratic performance in a single year affects our identification of high-ability managers. Note that we do not expect managerial ability to change in the short run. Rather, we consider the scores across 2 years to reduce possible measurement error. As an untabulated robustness check, we also define High-Ability Managers based on only 1 year. The results are similar, but weaker, as we would expect.

Intentional smoothing

As intentional smoothing is multidimensional and can be implemented using many different strategies (e.g., Acharya & Lambrecht, 2015; Dhole, Manchiraju, & Suk, 2016; B. Francis, Hasan, & Li, 2016; Graham et al., 2005; Lambert, 1984), our measure is based on four empirical proxies for reporting discretion and real activities management.¹⁵ We begin with abnormal accruals (AbnAcc). We define abnormal accruals using the modified Jones model following Dechow, Hutton, Kim, and Sloan (2012).¹⁶ Specifically, we estimate Equation 1 by industry (Fama & French, 1997):

{\frac{WCAcc}{A_{avg}}}_{t} = α + α_{1} \frac{1}{A_{avg}} + β_{2} \frac{{(Δ Sales - Δ AR)}_{t}}{A_{avg}} + β_{3} \frac{PP E_{t}}{A_{avg}} + \sum_{i = 1}^{n} δ_{i} Fir m_{i} + \sum_{i = 1}^{t} \emptyset_{i} Yea r_{i} + ε_{t} .

Following Kothari, Mizik, and Roychowdhury (2016), we include firm and year fixed effects in this and all subsequent models of real earnings management.¹⁷ Including the firm fixed effects in the first stage lessens the “bad model” problem (Kothari et al., 2016).¹⁸ We define working capital accruals (WCAcc) as the change in current assets plus the change in short-term debt less the change in both current liabilities and cash. $Δ Sales$ is the annual change in sales, $Δ AR$ is the annual change in accounts receivable, and PPE reflects the firm’s net property, plant, and equipment. The residual from Equation 1 serves as our measure of abnormal accruals, AbnAcc.

To capture real activities manipulation, we follow Roychowdhury (2006) and measure three activities that could be used to affect reported financial results: increasing sales by offering aggressive sales discounts or extending lenient credit terms (resulting in lower than expected cash flows given the level of sales, AbnCFO), overproducing inventory (to lower the per-unit fixed cost component of cost of goods sold, AbnProd), and cutting discretionary expenses to increase earnings (AbnExp).¹⁹ For each activity, we use the empirical model from Roychowdhury (2006) supplemented with firm and year fixed effects following Kothari et al. (2016) to measure the normal level of the activity where the residual captures the “abnormal” activity level.

The first real activities management metric is abnormal operating cash flow, which we measure with the following model (estimated by industry Fama & French, 1997):

\frac{CF O_{t}}{A_{avg}} = α_{0} + α_{t} \frac{1}{A_{avg}} + β_{1} \frac{Sale s_{t}}{A_{avg}} + β_{2} \frac{Δ Sale s_{t}}{A_{avg}} + \sum_{i = 1}^{n} δ_{i} FirmFixe d_{i} + \sum_{i = 1}^{t} \emptyset_{i} Yea r_{i} + ε_{t} .

CFO is cash flow from operations.²⁰Sales and ΔSales measure the level and change in sales, respectively. The residual reflects abnormal cash flows. Thus, we multiply the residual by −1, so that it is increasing in the extent of real activities management.

The second measure of real activities management is overproduction. The model of normal production (which we estimate by Fama & French, 1997) industry is as follows:

\frac{PRO D_{t}}{A_{avg}} = α_{0} + α_{t} \frac{1}{A_{avg}} + β_{1} \frac{Sale s_{t}}{A_{avg}} + β_{2} \frac{Δ Sale s_{t}}{A_{avg}} + β_{3} \frac{Δ Sale s_{t - 1}}{A_{avg}} + \sum_{i = 1}^{n} δ_{i} FirmFixe d_{i} + \sum_{i = 1}^{t} \emptyset_{i} Yea r_{i} + ε_{t} .

PROD is costs of goods sold plus the change in inventory. The residual from this model is our measure of real activities management from overproduction (AbnProd). The third measure of real activities management is abnormal discretionary expenses. We estimate the model of normal discretionary expenses by industry (Fama & French, 1997) as follows:

\frac{Expense s_{t}}{A_{avg}} = α_{0} + α_{1} \frac{1}{A_{avg}} + β_{1} \frac{Sale s_{t}}{A_{avg}} + β_{2} \frac{Δ Sale s_{t}}{A_{avg}} + \sum_{i = 1}^{n} δ_{i} FirmFixe d_{i} + \sum_{i = 1}^{t} \emptyset_{i} Yea r_{i} + ε_{t} .

We measure expenses subject to discretion (Expenses) as the sum of R&D and SG&A over the year. The residual reflects abnormal expenses. Thus, we multiply the residual by −1, so that it is increasing in the extent of real activities management (AbnExp).

We are interested in intentional smoothing, which may be implemented using both accrual and real earnings management. Because managers likely use the individual mechanisms concurrently (and thus the metrics may be highly correlated) and because a simple summation of the metrics may result in double counting or offsetting (particularly for the real earnings management metrics, for example, Roychowdhury, 2006), we use a principal components analysis to combine the individual metrics into one variable reflecting overtime income-increasing and income-decreasing earnings management. First, we sum the absolute value of each metric over years t − 2 to t. We then perform a principal components analysis with a Varimax rotation. The analysis results in 1 factor with an eigenvalue exceeding 1, which we retain as our variable of interest. The rotated factor pattern, as presented in Figure 1, indicates that all of the individual metrics load positively on the factor, with the real activities management metrics having the highest coefficients. Thus, this measure increases at times when management has made greater use of both income-increasing and income-decreasing abnormal operating and reporting decisions relative to the firm’s own average level of abnormal operating and reporting activities. Thus, this factor is our measure of intentional smoothing (IntentionalSmoothing_t_-2,t).

Figure 1.

Rotated factor pattern for the intentional smoothing factor.

Incentives

We consider six incentives for managers to intentionally smooth earnings. First, we examine the firm’s recent (current and prior two years) tendency to report performance metrics that just exceed its debt covenant thresholds (i.e., the firm’s tendency to avoid covenant violations over the same time frame that we examine intentional smoothing). Demerjian et al. (2017) find that intentional smoothing helps the firm to avoid spurious technical default but does not aid the firm in delaying performance technical default.²¹ As a result, intentional smoothing can improve the usefulness of earnings for contracting. Following Demerjian and Owens (2016), we define Tight_t years as those where the firm’s tightest debt covenant falls into the lowest decile of slack. These covenants are close to exceeding the contract threshold but do not actually violate the covenant. We set this variable to 0 for firms without private debt in the Dealscan database. To capture the proportion of years the manager is exposed to this incentive, we cumulate this variable over years t − 2 to t prior to ranking into deciles by industry and year (Tight_t_-2,t).²²

Similarly, we consider the firm’s recent propensity to report earnings that just beat the market’s earnings expectations because prior research provides evidence that firms are rewarded with higher stock valuations when they more frequently meet expectations (Barth et al., 1999; Kasznik & McNichols, 2002) where intentional smoothing is one path to this outcome (Myers et al., 2007). We set JustBeat_t equal to 1 in years where the firm meets or beats analyst earnings per share (EPS) expectations by 1 cent or less. Again, to capture the proportion of years, the manager is exposed to this incentive, we cumulate this variable over years t − 2 to t prior to ranking into deciles by industry and year (JustBeat_t_-2,t).

Third, we calculate a measure of perquisite consumption. The existence of two classes of shares with different voting rights limits the ability of noncontrolling shareholders to control the firm (e.g., Gompers et al., 2010). Reducing shareholders’ rights has been found to negatively affect firm value (e.g., Gompers, Ishii, & Metrick, 2003). Shares with high voting rights and low cash flow rights are an extreme example of reducing the noncontrolling shareholders’ rights, and have been associated with lower firm value and greater consumption of perquisites by managers (e.g., Gompers et al., 2010). In each year, we flag firms with dual-class shares where one class of shares has preferential voting rights. As with our other incentive variables, we sum the annual metric over years t − 2 to t, and rank by industry and year to obtain the proportion of years with shares traded that have unequal voting rights (Perquisite Consumption_t_-2,t).

Fourth, we measure the sensitivity of executives’ wealth to a 1% change in the firm’s stock price from the average delta (over years t − 2, t) of the executive with the greatest sensitivity.²³ Specifically, High_Delta is an indicator set equal to 1 if the average delta of the highest delta executive falls among the top quartile for the sample year, 0 otherwise.²⁴ The literature is mixed with respect to the extent that equity-based incentives encourage opportunistic earnings management (e.g., Armstrong, Guay, & Weber, 2010; Cheng & Warfield, 2005), and we are not aware of research examining the influence of equity incentives on income smoothing. We conjecture that if intentional smoothing benefits shareholders, then managers with wealth more closely linked to the firm’s stock price will engage in more intentional smoothing.

Fifth, we set Informed Trade to one in years where the executive team engaged in informed trade in year t. We measure informed trade following L. Cohen, Malloy, and Pomorski (2012), who classify traders as routine or informed based on their historical pattern of trades over the preceding years. Routine traders are those who consistently trade at regular intervals, whereas we classify traders with no discernible pattern of trades as informed. We cumulate this variable over the years t − 2 to t, and rank by industry and year to capture the proportion of years the manager is exposed to this incentive (Informed Trade_t_-2,t). We posit that if intentional smoothing increases with informed trade, it is less likely to benefit shareholders.

Finally, we examine if reputation concerns motivate intentional smoothing. We calculate two variables based on the CEO’s age that reflect either greater reputation-building incentives or short-employment horizons (i.e., reduced reputation-building incentives). We posit that if intentional smoothing increases when the firm is led by a younger (older) CEO, the smoothing is more likely to reflect reputation building (short horizons) (Ali & Zhang, 2015; Dechow & Sloan, 1991). We expect that if intentional smoothing aids the executive in reputation development, then it is more likely to be beneficial to shareholders; otherwise, it would not improve reputation. We measure reputation incentives with an indicator variable for CEOs whose age is less than or equal to 45 (Young CEO) and short-horizon incentives with an indicator variable for CEOs whose age is greater than or equal to 65 (Mature CEOs).²⁵ We provide details on variable definitions and measurement in Panel B of Table 2.

Table 2.

Descriptive Statistics.

Panel A: Descriptive Statistics.
Variable	n	M	Median	SD	25%	75%
High-Ability Managers	13,153	0.190	0.000	0.391	0.000	0.000
FirmSize_t	13,153	6.012	5.884	1.813	4.750	7.141
MB Ratio_t	13,153	2.983	2.163	6.859	1.310	3.681
SalesVolatility_t _-4,t	13,153	0.184	0.135	0.167	0.078	0.235
CFOVolatility_t _-4,t	13,153	0.076	0.053	0.085	0.031	0.089
OperCycle_t _-4,t	13,153	4.667	4.737	0.764	4.304	5.108
Losses_t _-4,t	13,153	0.252	0.200	0.329	0.000	0.400
BigNAuditor_t	13,153	0.864	1.000	0.342	1.000	1.000
SalesGrowth_t ^♦	13,153	0.154	0.083	0.612	−0.007	0.201
AbnRet_t ^♦	13,153	0.121	−0.032	0.908	−0.278	0.275
LitigationInd_t	13,153	0.376	0.000	0.484	0.000	1.000
NumAnalysts_t ^♦	13,153	6.715	4.000	6.544	2.000	9.000
IndRev%_t	13,153	0.013	0.000	0.043	0.000	0.010
PostSOX	13,153	0.560	1.000	0.496	0.000	1.000
ROA_t+ ₁	13,153	0.069	0.095	0.234	0.023	0.167
ReportedEarnVolatility_t_-2, _t ^♦	13,153	0.067	0.031	0.135	0.013	0.073
IntentionalSmoothing_t _-2,t	13,153	0.017	−0.219	1.023	−0.650	0.416
Tight_t_-2, _t ^♦	13,153	0.293	0.000	0.722	0.000	0.000
JustBeat_t_-2, _t ^♦	13,153	0.380	0.000	0.633	0.000	1.000
Perquisite Consumption_t_-2, _t ^♦	13,153	0.123	0.000	0.565	0.000	0.000
Delta_t_-2, _t ^♦	6,880	1,224.94	268.713	12,183.7	108.462	679.659
Informed Trade_t_-2, _t ^♦	13,153	1.614	2.000	1.070	1.000	3.000
Young CEO_t	6,475	0.080	0.000	0.271	0.000	0.000
Mature CEO_t	6,475	0.104	0.000	0.305	0.000	0.000
Executive Turnover	12,303	0.050	0.000	0.219	0.000	0.000
Voluntary Executive Turnover	12,303	0.027	0.000	0.163	0.000	0.000
Forced Executive Turnover	13,303	0.023	0.000	0.150	0.000	0.000

Note. All continuous variables are winsorized at the extreme 1%. The definition of and timing for each of the variables are provided in Panel B. High-Ability Managers are managers in the top quartile of MA-Score in both years t − 2 and t − 1. The “^♦” denotes a variable that is transformed in regression analyses reported in subsequent tables, but the untransformed variable is reported in Table 1 for ease of interpretation. The sample consists of 13,153 firm-year observations from 1995 to 2013.

Panel B: Variable Definitions.
Variable	Description	Definition
Ability and smoothing metrics
High-Ability Managers	Highest managerial ability	Managers in the top quartile (by industry-year) of MA-Score in both years t− 2 to t− 1. The MA-Score is from Demerjian, Lev, and McVay (2012).
IntentionalSmoothing	The extent of intentional smoothing	The factor based on the absolute value of discretionary revenue, discretionary CFO, discretionary production, and discretionary expenses over years t− 2 to t.
Control variables
ReportedEarnVolatility	Reported earnings volatility	The standard deviation of [earnings (IBC) / average assets (AT)] over the last 3 years (t− 2, t). In the regression analysis, we use the decile rank (by industry and year) of this variable.
FirmSize	Firm size	The natural log of the firm’s assets (AT) reported at the end of year t. Assets is the untransformed variable and reported in millions.
SalesVolatility	Sales volatility	The standard deviation of [sales (SALE) / average assets (AT)] over at least 3 of the last 5 years (t− 4, t).
CFOVolatility	Cash flow volatility	The standard deviation of [cash from operations (OANCF) / average assets (AT)] over at least 3 of the last 5 years (t− 4, t).
OperCycle	Operating cycle	The natural log of the length of the firm’s operating cycle, defined as sales turnover plus days in inventory [(SALE / 360) / (average RECT) + (COGS / 360) / (average INVT)], averaged over at least 3 of the last 5 years (t− 4, t).
Losses	Loss history	The percentage of years reporting losses in net income (IBC) over at least 3 of the last 5 years (t− 4, t).
Big N Auditor	Big N auditor Indicator	An indicator variable set equal to 1 for firms audited by Big N audit firms in year t; 0 otherwise. Big N audit firms are defined from Compustat as firms with “AU” codes between 1 and 8, inclusive. Thus, Arthur Andersen, Ernst & Young, Deloitte & Touche, KPMG, and PricewaterhouseCoopers are classified as Big N audit firms.
SalesGrowth	One year sales growth	The decile rank of industry-adjusted (by industry, year) sales growth defined as current year’s sales (SALE_t) less prior year’s sales (SALE_t_-1) less the increase in receivables, all scaled by prior year’s sales.
MB Ratio	Market-to-book ratio	The market-to-book ratio defined as the firm’s market capitalization (PRCC×CSHO) divided by book value (SEQ) for year t.
AbnRet	Abnormal return	The decile rank (by industry and year) of the firm’s market-adjusted buy-and-hold return for year t, where market returns are value weighted. Industry is classified per Fama and French (1997).
LitigationInd	Litigation industry	An indicator variable set equal to 1 for firms in litigious industries. Following prior research (J. Francis, Philbrick, & Schipper, 1994), we define high-litigation-risk industries as SIC codes: 2833-2836 (biotechnology), 3570-3577 and 7370-7374 (computers), 3600-3674 (electronics), and 5200-5961 (retailing).
NumAnalysts	Analyst coverage	The log of the number of analysts covering the firm in year t.
IndRev%	Industry revenue percentage	The firm’s sales in year t− 1 divided by the total sales for the firm’s industry in year t− 1.
PostSOX	Indicator for post-SOX years	An indicator variable set equal to 1 for fiscal years 2003 forward; 0 otherwise.
Other variables
ROA_t₊₁	Future earnings	ROA (operating income after depreciation scaled by average total assets) in year t+ 1.
ROA_t_+1,t+3	Mean future earnings	Mean ROA (the mean of operating income after depreciation for years t+ 1, t+ 2, and t+ 3 scaled by average total assets over the same period).
AbnRet_t₊₁	Future market-adjusted return	The firm’s buy-and-hold return over year t+ 1 less the market’s buy-and-hold return over year t+ 1. This variable is used in untabulated analysis.
Tight_t	Indicator for years close to violation of a debt covenant	Following Demerjian and Owen (2016), years where the firm’s tightest debt covenant falls into the lowest decile of slackness, that is, it is close to exceeding the contract threshold, but does not actually violate the covenant. We set this variable to 0 for firms without private debt in the Dealscan database.
Tight_t_-2,t	Number of years of Tight	The sum of Tight over years t− 2 to t. In the regression analysis, we use the decile rank (by industry and year) of this variable.
JustBeat_t	EPS just beat market expectations	An indicator variable set equal to 1 in years where the firm meets or beats analysts’ EPS expectations by 1 cent or less.
JustBeat_t_-2,t	Number of years of JustBeat	The sum of JustBeat over years t− 2 to t. In the regression analysis, we use the decile rank (by industry and year) of this variable.
Perquisite Consumption	Proxy for perquisite consumption	An indicator variable set equal to 1 in years where the firm has dual classes of shares with unequal voting rights. We identity firms with unequal voting rights using data from both RiskMetrics and the dataset provided by Gompers, Ishii, and Metrick (2010). Firms with greater voting than cash flow rights are more likely to consume perquisites.
Perquisite Consumption_t_-2,t	Number of years with unequal voting	The sum of Perquisite Consumption over years t− 2 to t. In the regression analysis, we use the decile rank (by industry and year) of this variable.
Delta	Sensitivity of wealth to 1% change in stock price	Delta is calculated per the method of Core and Guay (2002). To calculate Delta for more recent years, we use the code provided by Kai Chen (https://csclub.uwaterloo.ca/~k55chen/?p=211).
Delta_t_-2,t	Average delta over years t− 2 to t	Mean Delta of the executive with the greatest sensitivity of wealth to the firm’s stock price over years t− 2, t. High Delta is an indicator set equal to 1 if Delta_t_-2,t falls among the top quartile of the sample year; 0 otherwise.
Informed Trade_t	Indicator for years with informed trade	An indicator variable set to 1 in years where the executive team engaged in informed trade in year t, with informed trade defined as in L. Cohen, Malloy, and Pomorski (2012).
Informed Trade_t_-2,t	Number of years of informed trade	The sum of Informed Trade over years t− 2 to t. In the regression analysis, we use the decile rank (by industry and year) of this variable.
Young CEO	Reputation incentives of the CEO	An indicator set equal to 1 when the CEO’s age is 45 years or less; 0 otherwise.
Mature CEO	Short-horizon incentives of the CEO	An indicator set equal to 1 when the CEO’s age is 65 years or older; 0 otherwise. We chose 65 as the cutoff because Dechow and Sloan (1991) find that 65 is the age upon which most CEOs retire.
Executive Turnover	Executive turnover	An indicator variable set equal to 1 in year t if CEO or CFO turnover occurs in the following year. This variable is used in untabulated analysis.
Voluntary Executive Turnover	Voluntary executive turnover	An indicator variable set equal to 1 in year t if CEO or CFO turnover occurs in the following year, and the turnover is not classified as forced turnover. This variable is used in untabulated analysis.
Forced Executive Turnover	Forced executive turnover	An indicator variable set equal to 1 in year t if CEO or CFO turnover occurs in the following year and the turnover is classified as forced per Hazarika, Karpoff, and Nahata (2012). Specifically, the turnover is classified as forced if (a) the CEO was fired, forced out from the position, or departed due to policy differences; or (b) the departing CEO’s age is less than 60, and the announcement does not report that the CEO died, left because of poor health, or accepted another position elsewhere or within the firm; or the CEO “retires” but leaves the job within 6 months of the “retirement” announcement. The circumstances surrounding CEO turnovers in the third group are further investigated by reading press articles to reduce the likelihood that a turnover is incorrectly classified as forced. These departures are reclassified as voluntary if the incumbent takes a comparable position elsewhere, or departs for previously undisclosed personal or business reasons that are unrelated to the firm’s activities. This variable is used in untabulated analysis.

Note. SOX = Sarbanes–Oxley Act of 2002.

Control variables

Our main control variables are based on the determinants of earnings quality noted by Dechow and Dichev (2002) and Hribar and Nichols (2007), including firm size, proportion of losses, sales volatility, cash flow volatility, and operating cycle. We also control for the use of a Big N audit firm, which is associated with earnings quality (Becker, DeFond, Jiambalvo, & Subramanyam, 1998). We control for sales growth, the firm’s market-to-book ratio, and market-adjusted returns to control for growth and economic shocks to performance, both of which could potentially affect our measures of managerial ability and intentional smoothing (Demerjian et al., 2013). We include an indicator variable for high-litigation industries to control for the increased incentive to avoid negative earnings surprises in highly litigious environments (J. Francis, Philbrick, & Schipper, 1994). Other controls include the number of analysts following the firm and the firm’s share of industry revenue. We include these variables to control for investor recognition and SEC scrutiny, both of which increase the likelihood that abnormal reporting is detected (e.g., Beneish, 1997). We include an indicator variable for years following the passage of the Sarbanes–Oxley Act in 2002 because prior research suggests that the regulation changed managers’ earnings management strategies (e.g., D. Cohen, Dey, & Lys 2008).²⁶ We provide variable definitions and measurement periods in Panel B of Table 2.

Descriptive Statistics

We provide descriptive statistics in Table 2. For the transformed variables (SalesGrowth, AbnRet, NumAnalysts, ReportedEarnVolatility, Tight, JustBeat, Perquisite Consumption, Delta, Informed Trade), we present the untransformed variable for ease of interpretation in Table 2. We classify about 19% of firm-years as having a high-ability manager. Mean reported earnings volatility (ReportedEarnVolatility) is 0.07. Mean (median) IntentionalSmoothing is 0.02 (–0.22). The large difference in the value of IntentionalSmoothing at the lower quartile (–0.65) relative to the upper quartile (0.42) indicates wide variation in intentional smoothing.

In Table 3, we present both Pearson’s and Spearman’s correlations. High-Ability Managers are associated with more profitable firm-years (ROA) and greater sales volatility (SalesVolatility) but with lower cash flow volatility (CFOVolatility). High-ability managers also appear to use intentional smoothing to a greater extent than other managers as we observe a significantly positive correlation between IntentionalSmoothing and High-Ability Managers. High-ability managers’ smoothing efforts also appear successful as evidenced by the negative correlation between High-Ability Managers and ReportedEarnVolatility. In untabulated analyses, we also consider a factor comprised of the sum of signed reporting discretion over the same time frame, and find a negative association between High-Ability Managers and income-increasing discretionary reporting and real activities management. Thus, high-ability managers are not associated with greater income-increasing earnings management but rather appear to engage in greater smoothing activities.

Table 3.

Pearson’s (Spearman’s) Correlation Coefficients Below (Above) the Diagonal.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
(1) High-Ability Managers_t_-1		.05^a	−.03^a	.21^a	−.08^a	.17^a	−.02^a	.04^a	.02^a	.11^a	.04^a	.03^a	.01
(2) SalesVolatility	.08^a		.44^a	−.05^a	.31^a	.08^a	.04^a	−.04^a	−.04^a	−.06^a	−.01	.09^a	−.04^a
(3) CFOVolatility	−.04^a	.37^a		−.23^a	.40^a	.24^a	.02^a	−.08^a	−.07^a	−.11^a	−.02^a	.12^a	−.03^a
(4) ROA_t+₁	.17^a	−.07^a	−.39^a		−.25^a	.07^a	−.02^a	.10^a	.02^a	.17^a	.10^a	.02	.00
(5) ReportedEarnVolatility_t_-2,t	−.08^a	.27^a	.33^a	−.23^a		.00	.09^a	−.11^a	−.06^a	−.14^a	−.04^a	.08^a	−.05^a
(6) IntentionalSmoothing_t_-2,t	.15^a	.17^a	.39^a	−.19^a	.06^a		−.04^a	.07^a	−.06^a	.12^a	.08^a	.09^a	.01
(7) Tight_t_-2,t	−.02^a	.03^a	−.03^a	.03^a	.08^a	−.03^a		−.03^a	.03^a	−.04^a	.04^a	.01	.00
(8) JustBeat_t_-2,t	.04^a	−.03^a	−.05^a	.08^a	−0.11^a	.05^a	−.03^a		.00	.09^a	.03^a	.02	−.01
(9) PerquisiteConsumption_t_{-2, t}	.03^a	−.04^a	−.06^a	.03^a	−.05^a	−.05^a	.03^a	.00		−.03^a	−.05^a	.00	.08^a
(10) High Delta_t_{-2, t}	.11^a	−.05^a	−.10^a	.15^a	−.14^a	.11^a	−.03^a	.09^a	−.03^a		.07^a	−.02^a	.11^a
(11) Informed Trade_t_-2,t	.04^a	−.00	−.02^a	.09^a	−.04^a	.04^a	.04^a	.04^a	−.05^a	.06^a		.02^a	−.06^a
(12) Young CEO	.03^a	.09^a	−.09^a	.02	.08^a	.07^a	.00	.01	.00	−.02^a	.02^a		−.10^a
(13) Mature CEO	.01	−.05^a	−.02^a	.01	−.05^a	.02	.01	−.01	.07^a	.11^a	−.06^a	−.11^a

Note. Variable definitions are provided in Panel B of Table 2.

“^a” indicates a significant correlation at the 5% alpha level or higher.

Test Design and Results

Managerial Ability and Intentional Earnings Smoothing (H1)

In Table 4, we present the estimation of the following model:

\begin{array}{l} I n t e n t i o n a l S m o o t h i n g_{t - 2, t} = α + α_{1} High - Ability Manager s_{t - 1} + α_{2} F i r m S i z e_{t} + α_{3} S a l e s V o l a t i l i t y_{t - 4, t} \\ α_{4} C F O V o l a t i l i t y_{t - 4, t} + α_{5} O p e r C y c l e_{t - 4, t} + α_{6} L o s s e s_{t - 4, t} + α_{7} B i g N A u d i t o r_{t} \\ + α_{8} S a l e s G r o w t h_{t} + α_{9} M B R a t i o_{t} + α_{10} A b n R e t_{t} + α_{11} L i t i g a t i o n I n d_{t} \\ + α_{12} N u m A n a l y s t_{t} + α_{13} I n d R e v % + α_{14} P o s t S O X + \sum_{i = 1}^{n} δ_{i} F i r m F i x e d_{i} + ε_{t - 2, t} . \end{array}

Table 4.

Managerial Ability and Earnings Smoothing.

\begin{array}{l} I n t e n t i o n a l S m o o t h i n g_{t - 2, t} = α + α_{1} H i g h - A b i l i t y M a n a g e r s_{t} + α_{2} F i r m S i z e_{t} + α_{3} S a l e s V o l a t i l i t y_{t - 4, t} \\ + α_{4} C F O V o l a t i l i t y_{t - 4, t} + α_{5} O p e r C y c l e_{t - 4, t} + α_{6} L o s s e s_{t - 4, t} + α_{7} B i g N A u d i t o r_{t} \\ + α_{8} S a l e s G r o w t h_{t} + α_{9} M B R a t i o_{t} + α_{10} A b n R e t + α_{11} L i t i g a t i o n I n d_{t} \\ + α_{12} N u m A n a l y s t_{t} + α_{13} I n d R e v %_{t} + α_{14} P o s t S O X_{t} ε_{t - 2, t} . \end{array}

Panel A: OLS Regressions.
	Dependent variable =
	IntentionalSmoothing_t_-2,t		ReportedEarnVolatility_t_-2,t
High-Ability Managers	0.125***	0.426***	−0.012	−0.010
	5.63	9.63	−1.32	−0.87
IntentionalSmoothing_t_-2,t			0.022***	−0.012***
			3.35	−2.91
High-Ability Managers_t_-1× IntentionalSmoothing_t_-2,t			−0.025***	−0.001
High-Ability Managers_t_-1× IntentionalSmoothing_t_-2,t			−3.01	−0.01
FirmSize	−0.105***	−0.094***	−0.016**	−0.026***
	−3.55	−6.09	−2.34	−6.80
SalesVolatility	0.018	0.160	0.188***	0.299***
	0.15	1.39	6.39	12.74
CFOVolatility	1.572***	3.358***	0.431***	0.417***
	2.57	7.36	5.90	5.01
OperCycle	0.047	0.038	0.014	−0.002
	0.48	1.36	1.10	−0.28
Losses	−0.326***	−0.240***	0.301***	0.324***
	−5.76	−3.66	13.57	21.92
BigNAuditor	−0.022	−0.053	−0.01	0.020**
	−0.54	−0.95	−0.34	2.21
SalesGrowth	−0.011	−0.052	−0.041***	−0.048***
	−0.55	−1.20	−5.06	−2.62
MB Ratio	−0.002	0.007**	0.001**	0.001*
	−1.22	2.14	2.46	1.67
AbnRet	0.000	0.067	−0.025***	−0.053***
	0.04	1.26	−3.55	−3.19
LitigationInd	0.005	0.335***	−0.058	−0.071***
	0.05	7.64	−1.51	−6.92
NumAnalysts	0.060***	0.147***	−0.028***	0.002
	2.80	6.07	−3.37	0.36
IndRev%	0.923***	0.710**	0.362***	0.193***
	2.84	1.93	2.88	2.59
PostSOX	0.069***	0.138***	0.030***	0.025***
	3.42	4.52	3.89	3.02
Intercept	0.235	0.413**	0.458***	0.393***
	0.54	2.36	5.93	8.05
Firm fixed effects	Yes	No	Yes	No
Industry fixed effects	No	Yes	No	Yes
N	13,153	13,153	13,153	13,153
R² (%)	11.93	26.18	23.85	28.84

Panel B: Two-Stage Least-Squares Regression Analysis.
	First stage	Second stage
	Dependent variable =
	Managerial Ability	IntentionalSmoothing_t_-2,t
Pred. Managerial Ability		0.892***
		4.27
FirmSize	0.002**	−0.095***
	2.04	−9.15
SalesVolatility	0.128***	0.138
	18.55	1.35
CFOVolatility	0.063***	3.34***
	3.81	7.14
OperCycle	−0.007***	0.037**
	−4.32	2.35
Losses	−0.114***	−0.230***
	−29.85	−5.09
BigNAuditor	−0.010***	−0.053*
	−3.08	−1.87
SalesGrowth	−0.021***	−0.036
	−6.46	−1.24
MB Ratio	0.001	0.001***
	2.22	2.78
AbnRet	0.001	0.068***
	0.31	2.62
LitigationInd	0.002	0.328***
	0.59	13.91
NumAnalysts	0.020***	0.149***
	10.90	9.15
IndRev%	−0.240***	0.689***
	−5.86	3.86
PostSOX	−0.001	0.138***
	−0.46	8.78
MSA Average Ability	0.729***
	43.23
Firm fixed effects	No	No
Industry fixed effects	Yes	Yes
N		13,153
R² (%)		25.62
	Statistic	p value
Underidentification Test : χ² statistic	680.88	<.01
Weak Instrument Test: F statistic	1,558.83	<.01
Managerial Ability is Exogenous : χ² statistic	3.591	.06

Note. Variable definitions are provided in Panel B of Table 2. We present t statistics below the coefficients. For models that include (exclude) firm fixed effects, standard errors are clustered by firm (firm and year). Panel B reports the results of a two-stage least-squares regression. In the first stage, MSA Average Ability is the average ability of all managers in the same MSA as the firm’s headquarters and measured in year t − 1. The dependent variable for the first stage is managerial ability (the average of managerial ability in years t − 2 and t − 1). Critical values for the underidentification test are based on Stock-Yago (2005). The weak instrument test is based on the Cragg–Donald Wald F statistic. The Hausman test examines the null hypothesis that managerial ability is exogenous.

*, **, and *** denote a two-tailed p value of less than .10, .05, and .01, respectively, for all control variables. For hypotheses tests, *, **, and *** denote a one-tailed p value of less than .10, .05, and .01, respectively.

Our primary models include firm fixed effects to mitigate concerns of time-invariant correlated omitted variables. We also report results excluding firm fixed effects to illustrate differences in smoothing in the cross-section (rather than across time for each firm). Our dependent variable includes data from multiple years, so the errors are not independent across years within a firm. When the error terms are correlated within a firm, clustering standard errors by firm produces unbiased standard errors (Petersen, 2009).²⁷ Thus, we use robust standard errors clustered by firm in all models. For the models that exclude firm fixed effects, we also cluster the standard errors by year, and we supplement Equation 5 with industry fixed effects.

We present results in Panel A of Table 4. We find strong evidence that both across firms and across years (within a firm), High-Ability Managers are associated with significantly greater intentional smoothing.²⁸ To better understand the economic magnitude, we estimate model (a) using the decile rank of intentional smoothing as the dependent variable (results not tabulated). The significant coefficient for High-Ability Managers of 0.04 (p < .01) indicates that high-ability managers increase the rank of their firm’s intentional smoothing by about half a decile. As a reference point, the coefficient for NumAnalyst in the same regression is approximately 0.02 (p < .01). Thus, the influence of analysts’ demand for intentional smoothing is half the influence of high-ability managers. On this basis, we conclude that the ability of management has an economically meaningful impact on the magnitude of firms’ intentional smoothing.

Our main variable of interest, IntentionalSmoothing, is increasing when a manager has made greater use of income-increasing and income-decreasing earnings management over time. We refer to this behavior as intentional smoothing. To provide further evidence that the measure does, in fact, reflect attempts to smooth, we consider incentives in Section “Managerial Ability and the Incentives to Intentionally Smooth.” To provide additional evidence, we also examine the relation between high-ability managers’ intentional smoothing and earnings volatility in the last two columns of Table 4. The results reported in the third column indicate that when a high-ability manager leads the firm and engages in greater intentional smoothing activities, the firm reports lower earnings volatility than when the same firm is led by a lower ability manager. The results in Column 4 do not yield similar inferences, but this lack of significance is likely due to differences across firms that are more difficult to control for in the models that exclude firm fixed effects. For example, it is possible that intentional smoothing in the most volatile firms leads to earnings that are less volatile than they would have been otherwise, but that are still more volatile than the average firm. Nonetheless, the within-firm analysis suggests that high-ability managers’ intentional smoothing is associated with lower reported earnings volatility, and provides some reassurance that our measure reflects managers’ actual attempts to smooth. Overall, the results in Panel A of Table 4 provide evidence that high-ability managers engage in greater intentional smoothing consistent with H1.

Next, we conduct a two-stage least-squares analysis. As we lack a natural experiment where a firm’s managerial ability is exogenously shocked, we use an instrumental variable to better assess causality. To conduct the analysis, we must identify an instrument that is related to managerial ability but unrelated to the firm’s intentional smoothing strategies. We consider the availability of high-ability managers in the firm’s local labor market, and expect that a greater supply of high-ability managers increases the likelihood that the firm’s directors include in their network more high-ability managers and are thus, ceteris paribus, more likely to employ a high-ability manager.²⁹ At the same time, we expect that the average ability of other executives in the same geographic region as the firm is unrelated to the firm’s intentional smoothing, and thus meets the exclusion criteria required for a valid instrument. The two-stage least-squares analysis requires the use of a continuous variable in the first stage, and thus we examine managerial ability rather than the indicator for high-ability managers (although we find similar results when we use High-Ability Managers in the first stage).

We present the results in Panel B of Table 4. We observe a significantly positive coefficient for the instrument (MSA Average Ability). Also, we present two diagnostic tests as suggested by Larcker and Rusticus (2010). The first, a test of underidentification, rejects the null that our instrument is irrelevant (based on critical values from Stock & Yogo, 2005). The second, a weak instrument test, rejects the null that the instrument is weak (based on the Cragg–Donald Wald F Statistic). We also conduct a Hausman test which rejects the null that there is no endogeneity in this setting. The second stage results provide evidence of a positive relation between the instrumented managerial ability measure and intentional smoothing. These results corroborate our assertion that differences in managerial ability rather than omitted firm characteristics influence differences in firms’ intentional smoothing and provide evidence in support of H1.

Intentional Smoothing and Future Performance (H2)

H2 investigates the future performance consequences associated with high-ability managers’ intentional smoothing. We explore these relations with Equation 6:

\begin{array}{l} F u t u r e R O A_{t + n} = α + α_{1} High - Ability Manager s_{t - 1} + α_{2} IntentionalSmoothin g_{t - 2, t} \\ + α_{3} (High - Ability Manager s_{t - 1} \times IntentionalSmoothin g_{t - 2, t}) + α_{4} F i r m S i z e_{t} \\ + α_{5} S a l e s V o l a t i l i t y_{t - 4, t} + α_{6} C F O V o l a t i l i t y_{t - 4, t} + α_{7} O p e r C y c l e_{t - 4, t} + α_{8} L o s s e s_{t - 4, t} \\ + α_{9} B i g N A u d i t o r_{t} + α_{10} S a l e s G r o w t h_{t} + α_{11} M B R a t i o_{t} + α_{12} A b n R e t_{t} + α_{13} L i t i g a t i o n I n d_{t} \\ + α_{14} N u m A n a l y s t_{t} + α_{15} I n d R e v %_{t} + α_{16} P o s t S O X_{t} + \sum_{i = 1}^{n} δ_{i} F i r m F i x e d_{i} + ε_{t + n} . \end{array}

We consider two measures of future performance: 1-year forward ROA and the 3-year average ROA beginning in year t + 1.³⁰ We report the results from the estimation of Equation 2 in Table 5. First, we find a positive association between high-ability managers and future performance. Second, we examine if the positive influence of high-ability managers on future performance remains when they intentionally smooth. We find that the effect on 1- and 3-year-ahead future earnings of high-ability managers’ intentional smoothing is positive. That is, we examine the coefficient on IntentionalSmoothing for high-ability managers (i.e., the sum of α₁+α₃); the F tests indicate that intentional smoothing by high-ability managers is significantly positively associated with future performance (i.e., α₁+α₃ are significantly greater than 0). We find similar results when we exclude the firm fixed effects from the model (see the last two columns of Table 5). Finally, the coefficient for IntentionalSmoothing is not significantly different from 0 in all but one specification where it is significantly negative. Overall, these analyses allow us to reject H2 that the future operating consequences of intentional smoothing are not different for high-ability managers, and conclude instead that the intentional smoothing of high-ability managers is associated with increases in future performance.

Table 5.

Future Operating Performance, Managerial Ability, and Intentional Smoothing.

\begin{array}{l} F u t u r e R O A_{t + x} = α + α_{1} High - Ability Manager s_{t - 1} + α_{2} I n t e n t i o n a l S m o o t h i n g_{t - 2, t} \\ + α_{3} (IntentionalSmoothin g_{t - 2, t} \times High - Ability Manager s_{t - 1}) \\ + α_{4} F i r m S i z e_{t} + α_{5} S a l e s V o l a t i l i t y_{t - 4, t} + α_{6} C F O V o l a t i l i t y_{t - 4, t} \\ + α_{7} O p e r C y c l e_{t - 4, t} + α_{8} L o s s e s_{t - 4, t} + α_{9} B i g N A u d i t o r_{t} \\ + α_{10} S a l e s G r o w t h_{t} + α_{11} M B R a t i o_{t} + α_{12} A b n R e t_{t} + α_{13} L i t i g a t i o n I n d_{t} \\ + α_{14} N u m A n a l y s t_{t} + α_{15} I n d R e v %_{t} + α_{16} P o s t S O X_{t} + ε_{t - 2, t} . \end{array}

	Dependent variable =
	FutureROA_t+ ₁ 1 year		FutureROA_t+ _1,t+3 3 year		FutureROA_t+ ₁ 1 year	FutureROA_t+ _1,t+3 3 year
High-Ability Managers_t _-1	0.017***	0.016***	0.012**	0.011**	0.028***	0.021***
	3.33	3.05	2.43	2.23	5.24	4.93
IntentionalSmoothing_t _-2,t		−0.008		0.000	−0.032***	−0.016
		−0.68		0.02	−2.52	−1.42
High-Ability Managers_t_-1×IntentionalSmoothing_t_-2,t		0.018**		0.010*	0.050***	0.033***
		2.10		1.41	4.55	2.88
Firm fixed effects	Yes	Yes	Yes	Yes	No	No
Industry fixed effects	No	No	No	No	Yes	Yes
Controls variables	Yes	Yes	Yes	Yes	Yes	Yes
N	13,153	13,153	8,494	8,494	13,153	8,494
R² (%)	3.99	5.05	4.10	3.88	40.18	37.20
Coefficient tests	(F statistic/p value)
α₁+α₃ = 0		8.24***<.01		4.62**.03	31.75***<.01	17.40***<.01

Note. This table reports the results from the regression of future earnings on IntentionalSmoothing_t_-2,t, managerial ability and controls. Variable definitions are provided in Panel B of Table 2. We present t statistics below the coefficients. Statistical significance is assessed with robust standard errors. For models that include firm fixed effects, standard errors are clustered by firm. For models that exclude firm fixed effects, standard errors are clustered by firm and year.

Managerial Ability and the Incentives to Intentionally Smooth (H3)

We continue by investigating if incentives influence the relation between high-ability managers and intentional smoothing. We investigate these relations with the following model:

\begin{array}{l} I n t e n t i o n a l S m o o t h i n g_{t - 2, t} = α + α_{1} High - Ability Manager s_{t - 1} + α_{2} I n c e n t i v e_{t - 2, t} \\ + α_{3} (Incentiv e_{t - 2, t} \times High - Ability Manager s_{t - 1}) + α_{4} F i r m S i z e_{t} + α_{5} S a l e s V o l a t i l i t y_{t - 4, t} \\ + α_{6} C F O V o l a t i l i t y_{t - 4, t} + α_{7} O p e r C y c l e_{t - 4, t} + α_{8} L o s s e s_{t - 4, t} + α_{9} B i g N A u d i t o r_{t} \\ + α_{10} S a l e s G r o w t h_{t} + α_{11} M B R a t i o_{t} + α_{12} A b n R e t_{t} + α_{13} L i t i g a t i o n I n d_{t} + α_{14} N u m A n a l y s t_{t} \\ + α_{15} I n d R e v %_{t} + α_{16} P o s t S O X_{t} + \sum_{i = 1}^{n} δ_{i} F i r m F i x e d_{i} + ε_{t - 2, t} . \end{array}

In Equation 7, the firm fixed effects allow us to investigate how high-ability managers’ intentional smoothing is associated with reporting incentives relative to both their own reporting discretion in other periods (i.e., α₁ vs. α₃) as well as the reporting discretion of other managers facing the same incentives in the same firm (α₂).

We report the results in Table 6. The first specification excludes the interaction term Incentive×High-Ability Managers. We note that none of the incentive variables have coefficients that differ significantly from 0. In the second specification, which includes the interaction terms, we observe a significantly positive coefficient for the interaction between High-Ability Managers and Tight (coefficient is 0.11, p < .10). In contrast, the interaction of High-Ability Managers and Perquisite Consumption, our proxy for greater agency costs, and the interaction of High-Ability Managers and Informed Trade do not differ significantly from 0. Thus, informed trading and agency conflicts do not appear to motivate high-ability managers to intentionally smooth earnings.

Table 6.

Managerial Ability, Intentional Earnings Smoothing, and Incentives.

\begin{array}{l} I n t e n t i o n a l S m o o t h i n g_{t - 2, t} = α + α_{1} High - Ability Manager s_{t - 1} + α_{2} I n c e n t i v e_{t - 2, t} \\ + α_{3} (Incentiv e_{t - 2, t} \times High - Ability Manager s_{t - 1}) + α_{4} F i r m S i z e_{t} + α_{5} S a l e s V o l a t i l i t y_{t - 4, t} \\ + α_{6} C F O V o l a t i l i t y_{t - 4, t} + α_{7} O p e r C y c l e_{t - 4, t} + α_{8} L o s s e s_{t - 4, t} + α_{9} B i g N A u d i t o r_{t} \\ + α_{10} S a l e s G r o w t h_{t} + α_{11} M B R a t i o_{t} + α_{12} A b n R e t_{t} + α_{13} L i t i g a t i o n I n d_{t} + α_{14} N u m A n a l y s t_{t} \\ + α_{15} I n d R e v %_{t} + α_{16} P o s t S O X_{t} + \sum_{i = 1}^{n} δ_{i} F i r m F i x e d_{i} + ε_{t - 2, t} . \end{array}

	Dependent variable = IntentionalSmoothing_t_-2,t
	All firms	All firms	Low-to-mid leverage	All firms	All firms
High-Ability Managers_t _-1	0.125***	0.091	0.048	0.069	0.333*
	5.65	1.04	0.52	0.89	1.51
Tight _t_-2,_t×High-Ability Managers_t_-1		0.114*	0.138**	0.198***	0.333***
		1.56	1.81	2.53	2.28
JustBeat_t_-2,_t×High-Ability Managers_t_-1		0.039	0.086*	0.054	0.005
		0.73	1.48	1.05	0.04
Perquisite Consumption _t_-2, _t×High-Ability Managers_t_-1		−0.114	−0.123	−0.098	−0.133
		−0.97	−0.91	−0.87	−0.48
Informed Trade_t_-2, _t×High-Ability Managers _t_-1		0.027	0.024	−0.054	−0.151
		0.57	0.48	−1.19	−1.21
High Delta _t_-2,_t×High-Ability Managers_t_-1				0.051	−0.006
				1.35	−0.08
Young CEO_t×High-Ability Managers_t_-1				0.155***	0.248**
				2.40	1.66
Mature CEO_t×High-Ability Managers_t_-1				−0.015	−0.049
				−0.29	−0.39
Tight_t _-2,t	0.026	0.007	0.019	−0.011	0.102
	0.77	0.20	0.47	−0.26	1.36
JustBeat_t _-2,t	−0.006	−0.014	−0.017	−0.001	0.023
	−0.28	−0.55	−0.64	−0.04	0.48
Perquisite Consumption_t _{-2, t}	−0.079	−0.056	−0.09	0.035	−0.073
	−1.32	−0.88	−1.22	0.53	−0.63
Informed Trade_t _{-2, t}	0.031	0.026	0.037	0.05**	0.089**
	1.49	1.16	1.50	2.04	1.80
High Delta_t _-2,t				−0.000	0.089**
				−0.09	2.17
Young CEO_t				−0.134	−0.081
				−0.36	1.55
Mature CEO_t				−0.014	0.063
				−0.49	1.06
Firm fixed effects	Yes	Yes	Yes	Yes	No
Industry fixed effects	No	No	No	No	Yes
Control variables	Yes	Yes	Yes	Yes	Yes
N	13,153	13,153	10,051	6,398	6,398
R² (%)	12.00	12.04	3.22	9.43	28.12

Note. This table reports the results from the regression of IntentionalSmoothing, on managerial ability, reporting incentives and controls. Variable definitions are provided in Panel B of Table 2. We present t statistics below the coefficients. Statistical significance is assessed with robust standard errors. For models that include firm fixed effects, standard errors are clustered by firm. For models that exclude firm fixed effects, standard errors are clustered by firm and year. Low-to-Mid Leverage Firm Years are years when the firm’s average leverage over years t − 2 to t is among the bottom three quartiles of the sample. Leverage is calculated as long-term debt divided by average total assets.

Demerjian et al. (2017) provide evidence that intentional smoothing reduces the likelihood of spurious technical default, but is not useful in avoiding defaults resulting from increases in credit risk. Motivated by their results, we partition our sample based on credit risk (as measured by the extent of leverage). We only find a significantly positive relation between Tight×High-Ability Managers and IntentionalSmoothing for firms with low credit risk (i.e., firms with low-to-mid leverage).³¹ This result is consistent with Demerjian et al. (2017) who find that intentional smoothing is used to avoid spurious technical default, but is not helpful in delaying performance-driven default. Demerjian et al. (2017) conclude that using intentional smoothing to reduce the likelihood of spurious technical default reduces contracting costs. Thus, our results suggest that high-ability managers’ intentional smoothing benefits shareholders via reduced contracting costs.

Next, we examine the influence of compensation structure (High Delta) and CEO age (Young CEO, Mature CEO) on the relation between High-Ability Managers and IntentionalSmoothing. We consider them last because the sample is smaller for these tests due to missing data. We observe an insignificant coefficient for the interaction between High Delta and IntentionalSmoothing. We find a significantly positive coefficient for Young CEO×High-Ability Managers. This result indicates that young CEOs are more likely to intentionally smooth earnings, and is consistent with young executives having greater incentives to build their reputation via smoothing. We do not observe a significant coefficient for the interaction of Mature CEO×High-Ability Managers. The final specification excludes firm fixed effects but yields similar results.

Overall, our results suggest that high-ability managers engage in greater intentional smoothing over years when their firm reports earnings that are more frequently in close proximity to private debt covenant thresholds. Intentional smoothing among high-ability managers is concentrated among younger executives with greater reputation-building incentives, but does not increase when executives have a short-term focus because they are nearing the end of their careers. Finally, we do not find evidence that the high-ability managers’ intentional smoothing increases with perquisite consumption or informed trading.

Examination of Additional Consequences

To provide additional evidence on the consequences of intentional smoothing, we consider both future returns and future executive turnover in tests not tabulated. We measure future returns as the firm’s 1-year forward buy-and-hold return adjusted for the market return over the same period. Similarly, we measure executive turnover in year t + 1, and classify turnover of the CEO or CFO as forced or voluntary following the method of Hazarika, Karpoff, and Nahata (2012), where voluntary turnover is presumed to occur following superior executive performance and forced turnover indicates poor executive performance.³²

The results suggest that the market appropriately prices the future performance implications of high-ability managers and their intentional smoothing as we find no association between High-Ability Manager or High-Ability Manager×IntentionalSmoothing_t_-2,t and future abnormal returns. Next, we examine total, voluntary, and forced executive turnover. The results suggest that high-ability managers are more likely to experience turnover, and that the increase in turnover is attributable to voluntary decisions to leave the firm, presumably for improved employment opportunities. Moreover, we do not find evidence of an association between high-ability managers’ intentional smoothing and forced executive turnover. Overall, these analyses suggest that high-ability managers are motivated, at least in part, by how the capital market benefits of smoothing benefit shareholders, thereby bolstering their reputation.

Conclusion

We investigate whether high-ability managers are more likely to intentionally smooth earnings, a form of earnings management, and when they are more likely to do so. Our evidence indicates that high-ability managers are, on average, more likely to intentionally smooth earnings. The results also suggest that high-ability managers more effectively implement intentional smoothing strategies: We find that firms with high-ability managers experience incrementally superior earnings performance in the periods following intentional smoothing. Finally, we examine specific incentives related to smoothing, including those that benefit all shareholders (e.g., avoiding debt covenant violations, meeting or beating earnings benchmarks) and those that benefit the manager alone (e.g., perquisite consumption or informed trading). Our results reveal that high-ability managers smooth earnings when it benefits all shareholders but do not smooth earnings solely for their own personal benefit.

Our results offer insights into the two opposing viewpoints of earnings smoothing, whether it is beneficial or detrimental to shareholders. Our evidence is consistent with high-ability managers deploying their superior skill to report an earnings stream that avoids various reporting pitfalls to benefit all shareholders. We interpret these results as evidence that, when executed by a manager, intentional earnings smoothing can be viewed as a beneficial activity by managers. There are, however, limits to the inferences that can be drawn from this study. First, to appropriately test our hypotheses, we utilize a sample without executive turnover over the period that we measure intentional smoothing and with sufficient data to calculate our control variables. This process results in a sample of large, well-governed firms. The extent to which our results extend to smaller firms with less sophisticated governance systems is unclear but might be fruitful ground for future research. Second, we cannot observe managers’ intentions in making reporting choices; we can only infer intention from observed behavior. It is thus possible that measurement error or omitted factors could allow different conclusions to be drawn from our evidence (e.g., compensation structure might vary with managerial ability and influence smoothing; Dhole et al., 2016). This being said, we believe the results of our tests (and particularly the results related to incentives) point to intentional smoothing by high-ability managers as providing benefits to shareholders.

Footnotes

Acknowledgements

We thank Marcus Caylor, Ted Christensen (the discussant at the JAAF conference), Peter Easton (the editor), Weili Ge, Radha Gopalan, Won Kim, Panos Patatoukas, Divesh Sharma, Jeff Tsay, the anonymous reviewers, and workshop/conference participants at the 2017 Journal of Accounting, Auditing and Finance Conference, the Nick Dopuch Conference at Washington University, the Fraud and Misconduct Conference at University of California Berkeley, Brigham Young University, the University of Alberta, the University of California at Davis, the University of Houston, and Kennesaw State University for their comments. We thank Andrew Metrick for dual-class share information and Kai Chen for code used to calculate the sensitivity of executives’ wealth to changes in the firm’s stock price.

Authors’ Note

An earlier version of this article was circulated under the title “Earnings Smoothing: For Good or Evil?”

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 iD

Melissa Lewis-Western

Notes

References

Acharya

Lambrecht

(2015). A theory of income smoothing when insiders know more than outsiders. Review of Financial Studies, 28, 2534-2574.

Aier

Comprix

Gunlock

Lee

(2005). The financial expertise of CFOs and accounting restatements. Accounting Horizons, 19, 123-135.

Ali

Zhang

(2015). CEO tenure and earnings management. Journal of Accounting & Economics, 59, 60-79.

Allen

Dechow

Pope

(2017). Reference-dependent preferences: Evidence from marathon runners. Management Science, 63, 1657-1672.

Allison

(2005). Fixed effects regression methods for longitudinal data. Cary, NC: SAS Institute.

Allison

(2009). Fixed effects regression models (Quantitative applications in social sciences). Thousand Oaks, CA: SAGE.

Armstrong

Guay

Weber

(2010). The role of information and financial reporting in corporate governance and debt contracting. Journal of Accounting & Economics, 50, 179-234.

Badertscher

Collins

Lys

(2012). Discretionary accounting choices and the predictive ability of accruals with respect to future cash flows. Journal of Accounting & Economics, 53, 330-352.

Baik

Farber

Lee

(2011). CEO ability and management earnings forecasts. Contemporary Accounting Research, 28, 1645-1668.

10.

Barth

Elliott

Finn

(1999). Market rewards associated with patterns of increasing earnings. Journal of Accounting Research, 37, 387-413.

11.

Becker

DeFond

Jiambalvo

Subramanyam

K. R.

(1998). The effect of audit quality on earnings management. Contemporary Accounting Research, 15, 1-24.

12.

Beidleman

(1973). Income smoothing: The role of management. The Accounting Review, 48, 653-668.

13.

Beneish

(1997). Detecting GAAP violation: Implications for assessing earnings management among firms with extreme financial performance. Journal of Accounting and Public Policy, 16, 271-309.

14.

Bertrand

Schoar

(2003). Managing with style: The effect of managers on firm policies. The Quarterly Journal of Economics, 118, 1169-1208.

15.

Cameron

Miller

(2015). A practitioner’s guide to cluster-robust inferences. The Journal of Human Resources, 50, 317-372.

16.

Chaney

Lewis

(1995). Earnings management and firm valuation under asymmetric information. Journal of Corporate Finance, 1, 319-345.

17.

Cheng

Warfield

(2005). Equity incentives and earnings management. The Accounting Review, 80, 441-476.

18.

Cohen

Dey

Lys

(2008). Real and accrual-based earnings management in the pre- and post- Sarbanes-Oxley period. The Accounting Review, 83, 757-787.

19.

Cohen

Malloy

Pomorski

(2012). Decoding inside information. The Journal of Finance, 67, 1009-1043.

20.

Core

Guay

(2002). Estimating the value of employee stock option portfolios and their sensitivities to price and volatility. Journal of Accounting Research, 40, 613-630.

21.

Dechow

Dichev

(2002). The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review, 77, 35-59.

22.

Dechow

Hutton

Kim

J. H.

Sloan

(2012). Detecting earnings management: A new approach. Journal of Accounting Research, 50, 275-334.

23.

Dechow

Skinner

(2000). Earnings management: Reconciling the views of accounting academics, practitioners, and regulators. Accounting Horizons, 14, 235-250.

24.

Dechow

Sloan

(1991). Executive incentives and the horizon problem: An empirical investigation. Journal of Accounting & Economics, 14, 51-89.

25.

DeFond

Park

(1997). Smoothing income in anticipation of future earnings. Journal of Accounting & Economics, 23, 115-139.

26.

Demerjian

Donovan

Lewis-Western

(2017). Income smoothing and debt covenants: Evidence from technical default. Retrieved from https://ssrn.com/abstract=2745631

27.

Demerjian

Lev

Lewis

McVay

(2013). Managerial ability and earnings quality. The Accounting Review, 88, 463-498.

28.

Demerjian

Lev

McVay

(2012). Quantifying managerial ability: A new measure and validity tests. Management Science, 58, 1229-1248.

29.

Demerjian

Owens

(2016). Measuring the probability of financial covenant violation in private debt contracts. Journal of Accounting & Economics, 61, 433-447.

30.

Demski

(1998). Performance measure manipulation. Contemporary Accounting Research, 15, 261-285.

31.

Dhole

Manchiraju

Suk

(2016). CEO inside debt and earnings management. Journal of Accounting, Auditing & Finance, 31, 515-550.

32.

Fama

French

(1997). Industry costs of equity. Journal of Financial Economics, 43, 153-193.

33.

Fee

Hadlock

(2003). Raids, rewards, and reputations in the market for managerial talent. Review of Financial Studies, 16, 1315-1357.

34.

Francis

Hasan

(2016). Evidence for the existence of downward real-activity earnings management. Journal of Accounting, Auditing & Finance, 31, 212-248.

35.

Francis

Philbrick

Schipper

(1994). Shareholder litigation and corporate disclosures. Journal of Accounting Research, 32, 137-164.

36.

Gompers

Ishii

Metrick

(2003). Corporate governance and equity prices. Quarterly Journal of Economics, 118, 107-155.

37.

Gompers

Ishii

Metrick

(2010). An analysis of dual-class companies in the United States. Review of Financial Studies, 23, 1051-1088.

38.

Graham

Harvey

Rajgopal

(2005). The economic implications of corporate financial reporting. Journal of Accounting & Economics, 40, 3-73.

39.

Hazarika

Karpoff

Nahata

(2012). Internal corporate governance, CEO turnover, and earnings management. Journal of Financial Economics, 104, 44-69.

40.

Holcomb

Holmes

Connelly

(2009). Making the most of what you have: Managerial ability as a source of resource creation. Strategic Management Journal, 30, 457-485.

41.

Holmstrom

(1979). Moral hazard and observability. The Bell Journal of Economics, 10, 74-91.

42.

Hribar

Nichols

(2007). The use of unsigned earnings quality measures in tests of earnings management. Journal of Accounting Research, 45, 1017-1053.

43.

Kasznik

McNichols

(2002). Does meeting earnings expectations matter? Evidence from analyst forecast revisions and share prices. Journal of Accounting Research, 40, 727-759.

44.

Kirschenheiter

Melumad

(2002). Can “big bath” and earnings smoothing co-exist as equilibrium financing reporting strategies?Journal of Accounting Research, 40, 761-796.

45.

Kothari

S. P.

Mizik

Roychowdhury

(2016). Managing for the moment: The role of earnings management via real activities versus accruals in SEO valuation. The Accounting Review, 91, 559-586.

46.

Lambert

(1984). Income smoothing as rational equilibrium behavior. The Accounting Review, 59, 604-618.

47.

Lang

Lins

Maffett

(2012). Transparency, liquidity, and valuation: International evidence on when transparency matters most. Journal of Accounting Research, 50, 729-774.

48.

Larcker

Rusticus

(2010). On the use of instrumental variables in accounting research. Journal of Accounting & Economics, 49, 186-205.

49.

Leuz

Nanda

Wysocki

(2003). Earnings management and investor protection: An international comparison. Journal of Financial Economics, 69, 505-527.

50.

Levitt

(1998). The “numbers game.” Retrieved from www.sec.gov/news/speech/speecharchive/1998/spch220.txt

51.

Liu

(2008). Employment networks and the CEO labor market (Working paper). Retrieved from https://ssrn.com/abstract=1361776

52.

McNichols

(2002). Discussion of the quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review, 77, 61-69.

53.

Moses

(1987). Income smoothing and incentives: Empirical tests using accounting changes. The Accounting Review, 62, 358-377.

54.

Myers

Skinner

(2007). Earnings momentum and earnings management. Journal of Accounting, Auditing, & Finance, 22, 249-284.

55.

Palmrose

Z. V.

Scholz

(2004). The circumstances and legal consequences of non-GAAP reporting: Evidence from restatements. Contemporary Accounting Research, 21, 139-180.

56.

Petersen

(2009). Estimating standard errors in financial panel data sets: Comparing approaches. Review of Financial Studies, 22, 435-480.

57.

Roychowdhury

(2006). Earnings management through real activities manipulation. Journal of Accounting & Economics, 42, 335-370.

58.

Sankar

Subramanyam

K. R.

(2001). Reporting discretion and private information communication through earnings. Journal of Accounting Research, 39, 365-386.

59.

Schrand

Zechman

(2012). Executive overconfidence and the slippery slope to financial misreporting. Journal of Accounting & Economics, 53, 311-329.

60.

Stock

Yogo

(2005). Testing for weak instruments in linear IV regression, In Andrews

Stock

(Eds.), Identification and inference for econometric models: Essays in honor of Thomas Rothenberg (1st ed., pp. 80-108). Cambridge, UK: Cambridge University Press.

61.

Stubben

(2010). Discretionary revenues as a measure of earnings management. The Accounting Review, 85, 695-717.

62.

Subramanyam

K. R.

(1996). The pricing of discretionary accruals. Journal of Accounting & Economics, 22, 249-281.

63.

Trueman

(1986). Why do managers voluntarily release earnings forecasts?Journal of Accounting & Economics, 8, 53-71.

64.

Tucker

Zarowin

(2006). Does income smoothing improve earnings informativeness?The Accounting Review, 81, 251-270.

65.

Wang

(2016). Smarter money: High-ability CEOs’ insider trading profitability (Working paper).

66.

Zang

(2012). Evidence on the trade-off between real activities manipulation and accrual-based earnings management. The Accounting Review, 87, 675-703.