Ex-military CEOs and income smoothing

Income smoothing, managers’ practice of dampening earnings fluctuations, remains debated in accounting and finance. Critics portray it as opportunistic, masking true performance (Leuz et al., 2003; Lang et al., 2012); proponents argue it conveys forward-looking insight (Subramanyam, 1996; Tucker and Zarowin, 2006). Recent research ties smoothing’s usefulness to managerial traits, including ability and intent (Demerjian et al., 2020; Baik et al., 2020), yet little research asks whether imprinting life experiences also shape it.

We examine whether income smoothing differs for CEOs with prior military service, given evidence that military training has lasting effects on decision-making (Franke, 2001; Benmelech and Frydman, 2015). This evidence yields contrasting predictions: ex-military CEOs are less likely to commit fraud (Benmelech and Frydman, 2015), produce readable disclosures (Jha et al., 2024) and pursue conservative tax strategies (Law and Mills, 2017), suggesting military ethics discourage accrual discretion; yet military training also instills discipline and forward-looking judgment (Bamber et al., 2010; Benmelech and Frydman, 2015), traits enabling smoothing as a credible signal (Spence, 1973). Our first research question asks whether ex-military CEOs are more or less likely to smooth income than civilian CEOs.

Our second research question asks whether smoothing by ex-military CEOs enhances or diminishes information environment quality: under the signaling view, it may embed forward-looking information into earnings, improving earnings’ and prices’ ability to predict future performance, whereas military-linked overconfidence and risk-taking (Elder et al., 1991) may instead produce smoothing that conceals rather than clarifies performance. Addressing both questions clarifies how formative experiences shape not just whether managers manage earnings, but the quality of that choice.

We identify military CEOs using multiple sources, including BoardEx, Capital IQ, Marquis Who’s Who and proxy-statement verification, following prior research (Benmelech and Frydman, 2015; Law and Mills, 2017; Jha et al., 2024), and measure income smoothing as a composite factor score from three established proxies (Baik et al., 2020).

Our findings address both questions. First, firms led by ex-military CEOs engage in significantly greater income smoothing than civilian-led firms, resolving the theoretical ambiguity in favor of signaling. Second, this smoothing is informative rather than opportunistic, positively associated with earnings’ and stock prices’ ability to predict future performance.

Empirical governance research faces well-known endogeneity challenges, including unobserved firm characteristics, reverse causality and selection effects, which we address through complementary identification strategies: firm fixed effects, propensity score matching (PSM) and an instrumental variables (IV) approach exploiting variation in military service likelihood across birth cohorts (Benmelech and Frydman, 2015). Results remain robust across all approaches.

Our study makes two key contributions. First, we extend the smoothing literature by identifying military experience as a novel, psychologically grounded determinant of smoothing, distinct from the formal business education (Nurakhmet and Park, 2024) or compensation incentives (Grant et al., 2009) emphasized in prior work: formative experiences outside the corporate domain leave lasting imprints on executives’ reporting practices. Our work relates closely to Baik et al. (2020), who document a similar, ability-based smoothing channel. Rather than managerial ability, which aggregates innate talent, education and career experience, we focus on stable, early-life imprinting; our data reveal no meaningful correlation between military CEO status and Demerjian’s ability score, confirming the two constructs are empirically distinct.

Second, this paper enriches the literature on ex-military CEOs by documenting a distinct, strategic channel through which military imprinting shapes corporate reporting. Prior work emphasizes avoidance: military training deters CEOs from fraud (Benmelech and Frydman, 2015; Koch-Bayram and Wernicke, 2018), aggressive tax strategies (Law and Mills, 2017) and earnings management generally (Lai et al., 2023; Li et al., 2024). We instead document a strategic channel: ex-military CEOs use discretionary accruals to improve earnings’ and stock prices’ informativeness about future performance, evidence that intent, not just magnitude, shapes earnings management’s consequences.

Our findings also carry practical relevance for stakeholders inside and outside the firm. For boards, evidence that military-imprinted CEOs smooth income informatively rather than opportunistically suggests that observable, pre-career characteristics can complement internal monitoring and ability-based signals such as those in Baik et al. (2020). For standard setters, our results challenge the view that all smoothing misrepresents performance, because informative smoothing arises through independent channels, and blanket restrictions risk destroying value-relevant information. Auditors, similarly, may condition accrual-based risk indicators on CEO characteristics rather than apply them uniformly, because the same pattern can reflect signaling or a precursor to misstatement. Finally, analysts and investors can interpret smoothed earnings in light of CEO background and information asymmetry.

The remainder proceeds as follows: Section 2 reviews the literature and develops hypotheses; Section 3 describes the methodology; Section 4 presents the data; Section 5 reports results; Section 6 addresses identification and robustness; and Section 7 concludes.

Income smoothing entails managers’ discretionary actions to attenuate period-to-period fluctuations in reported earnings (Beidleman, 1973; Healy and Wahlen, 1999). The implications of intentional smoothing remain theoretically and empirically contested.

Anchored in agency theory (Jensen and Meckling, 1976), one stream of research conceptualizes smoothing as an opportunistic practice through which managers manipulate earnings to obscure performance and protect private interests such as job security or bonus compensation. Consistent with this view, Graham et al. (2005) documented that 78% of surveyed executives would sacrifice economic value to achieve smoother earnings, and cross-country evidence shows smoothing is more prevalent where investor protection is weak, linking smoothing to opportunism (Leuz et al., 2003). Critics further argue that smoothing diminishes earnings’ ability to capture true volatility, obscuring signals about underlying risk (Lang et al., 2012).

Conversely, drawing on signaling theory, income smoothing may serve an informational role by enabling managers to leverage private knowledge of future prospects to convey a clearer view of long-term earnings potential (Spence, 1973; Tucker and Zarowin, 2006). Analytical models suggest that smoothing can align reported earnings more closely with permanent earnings (Sankar and Subramanyam, 2001), and empirical evidence shows that smoother earnings enhance analysts’ forecast accuracy and lower firms’ cost of capital (Francis et al., 2004). More recently, smoothing has been found to improve the informativeness of banks’ loan loss provisions, underscoring its potential as an informational enhancement (Narayanamoorthy and Wheeler, 2025).

Taken together, these competing views suggest that income smoothing can either mislead investors or clarify economic fundamentals, depending on managerial context and intent. Examining managers’ personal characteristics and professional backgrounds therefore offers important insight into the motivations underlying smoothing practices.

Upper echelons theory (Hambrick and Mason, 1984) posits that organizational outcomes reflect top executives’ experiences, values and cognitive styles, contrasting with neoclassical views that treat managers as interchangeable: decision-making is filtered through individual biases that affect strategic choices. Consistent with this, Bamber et al. (2010) found manager fixed effects explain significant variation in disclosure attributes, including forecast frequency, specificity and bias, even after controlling for firm-level determinants.

One notable extension of this framework to income smoothing is Baik et al. (2020), who find that CEOs with higher managerial ability are more likely to engage in smoothing that enhances earnings informativeness, interpreting this as evidence that able CEOs deploy smoothing strategically to signal firm fundamentals. However, the ability measure used by Baik et al. (2020), based on Demerjian et al. (2012), aggregates diverse sources including innate talent, formal education and professional experience, and does not isolate the specific channels through which managerial characteristics influence reporting behavior.

Importantly, managerial ability differs conceptually from imprinting experiences, defined as formative life events that leave enduring marks on an individual’s preferences and decision-making style ^[1]. While ability may evolve with context and experience, imprinting emphasizes stable traits shaped early in life. To our knowledge, no study has examined whether noncorporate imprinting experiences affect income smoothing ^[2]. We address this gap by investigating whether prior military service shapes CEOs’ propensity to engage in accrual-based income smoothing.

We focus on military experience as a salient imprinting trait, motivated by extensive evidence of its enduring influence on executive behavior and corporate outcomes (Marquis and Tilcsik, 2013). Benmelech and Frydman (2015) argued that military service instills discipline, integrity and risk aversion, fostering a conservative leadership style that persists long after active service, while also equipping executives to make sound judgments under stress and uncertainty, a capacity reflected in ex-military CEOs leading firms that outperform during industry downturns.

This capacity for sound judgment motivates our first hypothesis. Income smoothing can serve as a credible signal of firm quality when used by capable managers (Spence, 1973), as the ability-based smoothing discussed above (Baik et al., 2020) illustrates. If ex-military CEOs possess comparable decision-making acumen, they may use smoothing strategically to signal the firm’s underlying resilience rather than to obscure it. These considerations lead to our first hypothesis:

This hypothesis, however, may confront tension from the ethical, rule-following dimension of the same military imprint. As noted above, ex-military CEOs adopt more conservative investment and tax policies, produce more readable disclosures and exhibit lower earnings management overall (Benmelech and Frydman, 2015; Law and Mills, 2017; Jha et al., 2024; Li et al., 2024), evidence rooted in the honesty and accountability norms of military culture (Franke, 2001; Soeters, 1997). Military-trained CEOs may therefore be reluctant to exercise accrual discretion at all, rendering the relationship between military background and smoothing an open empirical question.

While we expect a positive association between military experience and income smoothing, the underlying motivation remains an empirical question. If ex-military CEOs possess superior judgment in forecasting future performance (Bamber et al., 2010) and managing firms through uncertainty (Benmelech and Frydman, 2015), their smoothing may embed private information about future cash flows into current earnings, serving a signaling function (Sankar and Subramanyam, 2001). This reasoning motivates our second hypothesis:

Smoothing may also amplify stock price informativeness by conveying managerial foresight (Tucker and Zarowin, 2006). Military experience is associated with discipline, accountability and the ability to make sound decisions under pressure (Franke, 2001; Benmelech and Frydman, 2015), qualities that may improve a CEO’s ability to assess future performance and convey it credibly through accruals. This leads to our third hypothesis:

Both hypotheses remain open empirical questions, however, given the competing possibility, raised earlier, that military-linked overconfidence and risk-taking (Elder et al., 1991) could instead produce smoothing that masks rather than clarifies performance.

We obtain firm identifiers and executive names from Execucomp to identify CEOs with military backgrounds, drawing primarily on BoardEx, which contains educational, career and network information for more than 1.4 million global directors and executives, supplemented by S&P Capital IQ and Marquis Who’s Who. Where military service cannot be validated through at least two of these three databases, we conduct supplementary searches using resources including Gale in Context and company annual reports. We exclude CEOs whose involvement is limited to reserve duty without active service. Our ex-military CEO variable equals one if the CEO has prior military experience and zero otherwise, consistent with established literature (Benmelech and Frydman, 2015; Law and Mills, 2017).

We construct a composite measure of income smoothing (SMTH) by integrating three widely adopted proxies to reduce measurement error, following Baik et al. (2020); SMTH is the common latent factor extracted via factor analysis from three individual measures, SMTH1, SMTH2 and SMTH3. SMTH1 is the ratio of the standard deviation of operating earnings to that of operating cash flows, both scaled by lagged total assets. SMTH2 is the Spearman correlation between changes in total accruals and changes in operating cash flows, scaled by lagged assets, capturing the extent to which managers use accruals to offset cash flow fluctuations. SMTH3 is the Spearman correlation between changes in discretionary accruals and changes in premanaged income (net income less discretionary accruals), assessing whether managers use discretionary accruals to stabilize earnings. Higher values, after multiplying by negative one, indicate greater smoothing; calculations require data for at least three of the prior five years.

To examine our H1 concerning the association between CEOs with military backgrounds and income smoothing, we estimate the following regression model adapted from prior literature (Lang et al., 2012; Baik et al., 2020) ^[3]:

$SMTHt, t+4$ represents the aggregate measure of income smoothing over years t to t + 4. $MILITARY_CEOt$ is an indicator variable equal to one if the CEO has military experience, and zero otherwise.

The coefficient β₁ captures the effect of military service on income smoothing, controlling for firm-level characteristics known to influence smoothing behaviors. These controls include firm size, leverage, book-to-market ratio, sales volatility, the frequency of losses, operating cycle, sales growth, operating leverage and average cash flows, following Lang et al. (2012)^[4]. In addition, we examine the relation between military experience at time t and subsequent income smoothing to mitigate concerns about reverse causality. We estimate this OLS model with firm and year fixed effects specifically to isolate the imprinting effect of military service from time-invariant unobserved firm heterogeneity and macroeconomic shocks. The use of fixed effects also helps alleviate the omitted variable concern, improving the efficiency of our estimations. Standard errors are clustered at the firm level throughout to account for serial correlation within firms across our panel. Consistent with our hypothesis, we predict the coefficient of MILITARY_CEO (⁠$β1$⁠) to be significantly positive, indicating that CEOs with military backgrounds are more likely to engage in income smoothing.

We obtained accounting variables from COMPUSTAT and stock return data from CRSP. To identify ex-military CEOs, we relied on multiple sources, including BoardEx, ExecuComp, S&P Capital IQ and Marquis Who’s Who, as detailed in Section 3. Managerial ability measures were drawn from Professor Demerjian’s website. Our sample period spans from 1993 to 2022.

Table 1 summarizes our sample selection process. We began with 298,666 firm-year observations from COMPUSTAT and excluded 194,104 observations not covered by ExecuComp. We further removed 58,119 observations with missing financial data. We exclude firms in the financial and utilities sectors (12,086 observations), due to their unique regulatory requirements. Our final full sample consists of 34,357 firm-year observations from 3,440 unique US firms. To mitigate the effect of outliers, we winsorize all the continuous variables at the 1st and 99th percentiles.

Table 2 presents the summary statistics for the dependent variables in Panel A and the independent variables in Panel B. The mean of MILITARY_CEO is 0.099, indicating that 9.9% of firm-year observations in our sample are led by CEOs with military backgrounds. Overall, the distribution of variables in our sample is consistent with prior literature on military CEOs (Law and Mills, 2017; Jha et al., 2024) and income smoothing (Baik et al., 2020).

The correlation table in Panel C of Table 2 shows that there is no significant correlation between managerial ability score and CEO’s military experience. In other words, Demerjian’s managerial ability does not capture the effect of military experience that shapes CEOs’ styles over corporate policies.

Table 3 presents baseline OLS results testing H1, that firms led by ex-military CEOs exhibit higher income smoothing. Columns (1) and (2) exclude and include Demerjian’s MA_SCORE, respectively, to assess whether the military effect is distinct from managerial ability. In both specifications, the coefficient on MILITARY_CEO is positive and significant at the 1% level: firms led by ex-military CEOs engage in significantly greater income smoothing than civilian-led firms. Benchmarked against SMTH’s standard deviation of 0.602 (Table 2), these coefficients imply income smoothing is approximately 1.21 and 0.97 standard deviations higher for military-CEO firms in columns (1) and (2), respectively ^[5]. Military experience thus captures a dimension of managerial influence on income smoothing distinct from conventional, efficiency-based measures of ability.

To test H2, which predicts that income smoothing by military-experienced CEOs strategically enhances the informativeness of current earnings about future performance, we estimate the following OLS regression to capture the interaction between smoothing, military leadership and current earnings:

where CFO_{t + 1} is cash flow from operations in year t + 1 scaled by lagged total assets, E_t is earnings before extraordinary items in year t scaled by lagged total assets, CFO_{t − 1} controls for expectations at the start of year t, and SMTH_{t − 4,t} represents income smoothing over the preceding five-year period. A significantly positive coefficient on γ₁₀ would indicate that income smoothing by military CEOs further enhances the ability of current earnings to predict future cash flows.

Table 4 reports the results of our tests for H2. In Columns (1) and (2), we estimate a baseline model excluding CEOs’ military background. In Column (1) the coefficients on current earnings (E_t) and lagged cash flows (CFO_{t − 1}) are both positive and significant, confirming that both contribute to predicting future cash flows. Column (2) shows that income smoothing generally enhances the informativeness of current earnings, as evidenced by the significantly positive coefficient on SMTH_{t − 4,t} × E_t. More importantly, Column (3) introduces the three-way interaction MILITARY_CEO × SMTH_{t − 4,t} × E_t, whose significantly positive coefficient supports H2, suggesting that income smoothing by ex-military CEOs further enhances the informativeness of current earnings about future cash flows. Notably, the three-way interaction MILITARY_CEO × SMTH_{t − 4,t} × CFO_{t × 1} is not significant, indicating the military effect operates through earnings informativeness specifically, not through cash-flow persistence.

To test H3, which posits that income smoothing by military-trained executives improves the market’s ability to incorporate future performance into current valuations, we extend the CKSS FERC framework (Collins et al., 1994) as adapted by Tucker and Zarowin (2006):

The dependent variable R_t is the annual ex-dividend stock return for year t, and R_t3 is the annually compounded stock return from year t + 1 to t + 3. X_{t − 1} and X_t represent EPS for years t − 1 and t, while X_t3 is the cumulative EPS from years t + 1 through t + 3, all excluding extraordinary items, adjusted for stock splits and dividends following Christie (1987), and deflated by beginning-of-year stock price. We predict β₂ (ERC) and β₃ (FERC) to be positive. If smoothing by military CEOs strengthens the relation between current stock prices and future earnings, β₁₈, the coefficient on MILITARY_CEO_t × SMTH_{t − 4,t} × X_t3, should be significantly positive.

Table 5 reports the results. Consistent with Collins et al. (1994), β₃, the coefficient on future earnings (X_t3), is positive across all specifications. Consistent with Tucker and Zarowin (2006), the coefficient of SMTH_{t − 4,t} x X_t3 is positive and significant in Columns (2) and (3), suggesting that income smoothing generally improves the market’s ability to incorporate future earnings into current valuations. More importantly, the coefficient on MILITARY_CEO_t × SMTH_{t − 4,t} × X_t3 is significantly positive, consistent with H3, implying that income smoothing by ex-military CEOs further enhances stock price informativeness through strategic signaling ^[6].

Taken together, our results support a signaling interpretation of income smoothing by ex-military CEOs: consistent with Spence’s (1973) signaling theory, the incremental informativeness of both current earnings and stock prices for future performance indicates that these CEOs use accrual discretion to convey credible information rather than obscure it. Because smoothing requires acting on private knowledge of future cash flows, it is a costly signal that less-informed managers cannot easily replicate, making it credible specifically when exercised by CEOs with military-imprinted judgment and foresight. These findings extend the ability-based signaling channel documented by Baik et al. (2020) to a distinct, noncorporate source of managerial credibility.

While the empirical analyses in the previous section establish a robust statistical association between military-experienced CEOs and informative income smoothing, this association must extend beyond statistical artifacts to consider what it means in practice for boards, market participants and regulators.

For boards and nominating committees, our findings identify a further channel through which CEO imprinting experience, specifically prior military service, can enhance firm value: ex-military CEOs strategically use accrual discretion to communicate credible information about the firm’s future rather than avoid it altogether.

For market participants, CEO characteristics offer a practical basis for interpreting earnings signals. Because smoothed earnings can reflect either informative signaling or opportunistic concealment, investors and analysts may weigh them more heavily as an indicator of future performance when CEOs are ex-military or score highly on managerial ability (Baik et al., 2020), and with more skepticism otherwise. Auditors may similarly condition accrual-based risk assessments on these characteristics rather than apply uniform thresholds, because the same pattern can signal credible communication or a precursor to misstatement depending on who exercises the discretion.

For standard setters, our findings speak to the debate over the appropriate level of accounting flexibility (Healy and Wahlen, 1999). Evidence that an observable manager characteristic predicts informative rather than opportunistic discretion challenges blanket restrictions: treating all smoothing alike risks discarding a valuable signal, alongside the ability-based channel documented by Baik et al. (2020).

Firms with certain traits, such as a preference for structured governance, may favor both smoother earnings and military-trained executives, raising endogeneity and selection concerns. To isolate the imprinting effect of military service from these confounds, we use three complementary strategies, each targeting a distinct source: PSM for reverse causality and functional form misspecification, IV analysis for unobserved heterogeneity and reverse causality and entropy balancing (EB) for covariate imbalance.

Firms predisposed to greater income smoothing may also be more inclined to appoint military-experienced CEOs. PSM addresses this by constructing a control sample of observationally similar firms without military CEOs, without imposing a specific functional form on the MILITARY_CEO-smoothing relationship (Rosenbaum and Rubin, 1983). Following Shipman et al. (2017), we use the matched sample for subsequent multivariate regression.

We estimate propensity scores via a logit model using the controls from equation (1) with year and industry fixed effects, then apply one-to-one nearest neighbor matching without replacement, yielding a balanced sample of 6,416 firm-year observations (3,208 treated, 3,208 control). Panel A of Table 6 reports the first-stage logit; Panel B presents balance diagnostics, confirming treatment and control groups are well-balanced across covariates.

Panel C presents OLS results on the matched sample. The coefficient on MILITARY_CEO remains significantly positive, confirming our findings are not an artifact of endogenous sample selection or omitted variables.

To address concerns that the association between military CEOs and income smoothing reflects spurious correlation or unobserved firm heterogeneity, we use an IV strategy. Consistent with established research on military leadership and corporate performance (Benmelech and Frydman, 2015; Koch-Bayram and Wernicke, 2018), we instrument CEO military background using birth year and a dummy variable indicating birth during the conflict era (1930–1943).

Table 7 presents the results. First-stage estimates confirm instrument strength: birth year is negatively, and the conflict-era dummy positively, associated with military service, both significant and consistent with prior literature.

The Cragg–Donald Wald F-statistic of 382.47 and the Kleibergen–Paap Wald F-statistic of 61.28 both substantially exceed the Stock and Yogo’s (2005) critical threshold of 19.93, rejecting weak instrument concerns; the Hansen p-value of 0.481 further supports instrument validity.

The second-stage coefficient on the instrumented military CEO variable remains positive and statistically significant, confirming our findings withstand this endogeneity correction^[7].

EB (Hainmueller, 2012) addresses covariate imbalance without PSM’s model dependence. The results from EB confirm the ex-military CEO and income smoothing association is robust; full details appear in  Appendix 2 B1.

This study investigates whether and how CEOs’ military background shapes the practice and consequences of accrual-based income smoothing. Using a comprehensive sample of US firms, we find that firms led by ex-military CEOs engage in greater income smoothing, and that such smoothing enhances the informativeness of current earnings and stock prices with respect to future performance, indicating that military CEOs use smoothing strategically rather than opportunistically. These findings are robust across multiple identification strategies, including firm fixed effects, PSM, EB and IV analyses.

Our study makes two key contributions. First, we identify military experience as a novel, psychologically grounded determinant of accrual-based income smoothing, distinct from managerial ability, demonstrating that formative experiences outside the corporate domain leave enduring marks on executives’ reporting practices. Second, we broaden the literature on ex-military CEOs by documenting a distinct, strategic channel through which military imprinting shapes corporate reporting: rather than simply avoiding earnings management, as prior work on military executives emphasizes, ex-military CEOs use accrual-based smoothing to enhance the informativeness of both earnings and stock prices, underscoring that the intent behind earnings management, not just its magnitude, shapes its consequences.

This study is not without limitations. Our sample is drawn exclusively from US public firms, and the effect of CEO characteristics on financial reporting behavior may not generalize to other regulatory and institutional settings. Corporate governance mechanisms and enforcement intensity vary considerably across countries, and prior research shows that earnings management and the influence of executive characteristics on firm outcomes both vary across investor-protection regimes (Leuz et al., 2003; Al Sa’Ed and Saleh, 2024). Future research could examine whether the role of military-imprinted CEOs in smoothing extends to non-US settings with weaker investor protections.

While PSM helps address endogeneity concerns, it has limitations – most notably, its reliance on a model-specified propensity score and the resulting loss of sample size. To mitigate these issues and reinforce the robustness of our findings, we conduct an additional analysis using entropy balancing (Hainmueller, 2012). In contrast to PSM, entropy balancing does not rely on a specific model to estimate matching probabilities. Instead, it uses an iterative reweighting algorithm to ensure that the treatment and control groups are statistically equivalent across the selected covariates in terms of their moments (e.g. means and variances). This approach retains the full sample while improving covariate balance between military and nonmilitary CEO firms.

Table A2 presents the results of this entropy-balanced analysis. The coefficient on MILITARY_CEO remains positive and statistically and economically significant, reinforcing our main finding that firms led by ex-military CEOs engage in significantly greater income smoothing. The stability and significance of the results across specifications suggest that our conclusions are not driven by model specification or sample selection biases associated with PSM.

Our main instrumental variable analysis (Table 7) instruments CEO military background using birth year and an indicator for birth during the conflict era (1930–1943), following Benmelech and Frydman (2015). Because CEOs born in this era are concentrated in the earlier portion of our 1993–2022 panel, a natural concern is that the instrument’s identifying power, and consequently our second-stage estimates, may be driven disproportionately by observations from the early years of the sample. We address this concern using two complementary approaches: restricting the sample to the period in which the instrument retains meaningful support, and using an alternative instrument that does not rely on birth cohort timing at all.

1) Restricted sample period: We restrict the sample to 1993–2011, consistent with the sample period used in Law and Mills (2017), who study military CEOs in a closely related setting over 1992–2011.

Table A3 reports OLS estimates replicating our main specification in Table 3 over this restricted period. The coefficient on MILITARY_CEO remains positive and statistically significant, confirming that the direct association between military experience and income smoothing is not confined to, or dependent on, the later years of our panel.

Table A4 reports the corresponding two-stage least squares estimates using the same birth cohort instruments as Table 7. Rather than weakening, the first stage instrument strengthens over this restricted window: the conflict era coefficient rises to 0.276 from 0.168, and the Kleibergen–Paap F-statistic rises to 66.23 from 61.28 in the full sample. The second-stage MILITARY_CEO coefficient remains similar. This indicates that the identifying variation is, if anything, concentrated where the instrument is strongest, and that our full sample estimates are not an artifact of weak, diluted identification in later years.

2) Industry year average instrument: While the birth cohort approach is the dominant identification strategy used in the military CEO literature (Benmelech and Frydman, 2015; Koch-Bayram and Wernicke, 2018), its identifying power is necessarily concentrated in the earlier years of any long panel. To provide a test free of this limitation, we construct an alternative instrument based on the industry year average proportion of military CEOs, excluding the focal firm, lagged two and three years, an approach widely used elsewhere in the accounting and finance literature (Dass et al., 2014; Dhaliwal et al., 2016) and whose identification does not depend on the timing of any historical cohort. Peer firms’ propensity to hire military experienced CEOs plausibly reflects industry norms and executive labor market conditions that also shape the focal firm’s own hiring, while remaining unrelated to firm-specific determinants of income smoothing once firm and year fixed effects are included.

Table A5 reports first stage and second stage two-stage least squares results using these alternative instruments. The first stage coefficients on both lagged industry averages are positive and highly significant, and the Kleibergen–Paap and Cragg–Donald F-statistics both indicate a strong instrument. The second-stage MILITARY_CEO coefficient remains positive and statistically significant, consistent with our main finding.

Dass, N., Kini, O., Nanda, V., Onal, B. and Wang, J. (2014), “Board expertise: do directors from related industries help bridge the information gap?”, Review of Financial Studies, Vol. 27 No. 5, pp. 1533-1592.

Dhaliwal, D., Judd, J.S., Serfling, M. and Shaikh, S. (2016), “Customer concentration risk and the cost of equity capital”, Journal of Accounting and Economics, Vol. 61 No. 1, pp. 23-48.