Article history:Received 17 September 2012Received in revised form 8 July 2013Accepted 9 July 2013Available online 24 July 2013
Using seventeen observable demographic characteristics, we investigate the impact of six CEOprofiles on dividend policy. Firms headed by married, Republican, Christian CEOs with childrenmaintain higher dividend yields and are more likely to considerably increase their dividendpayout. Following substantial dividend hikes, firms led by CEOs with these more traditionalpersonal lives exhibit deteriorating performance. Potential explanations include managerialoptimism coupled with dividend signaling and the possibility that CEO profiles proxy for anunobserved firm effect such as firm maturity. However, the associations above continue topersist in both mature firm and turnover sub-samples. Overall, this suggests that theserelationships are related to characteristics of the CEOs themselves.
In 2012, Cisco and Exxon Mobil both made headline-worthy dividend hikes, with Cisco increasing its dividend per share by75%1 and Exxon's 21% increase rendering it the largest dividend payer in the world.2 While evidence has shown that a firm'sdividend policy is affected by a multitude of factors, including changing market conditions3 and characteristics of the firm itself,4
we assert that a company's dividend policy may additionally be influenced by its leader. Specifically, the bold increases in Ciscoand Exxon Mobil's payout policies may be linked to common traits of the executives themselves (i.e., similarities include bothleaders being married, Christian, Republican parents). Consequently, we contend that demographic factors may play an importantrole in dividend policy decisions.
A subset of recent literature has considered the effect of executive qualities and backgrounds on firm decisions. Kaplan et al.(2012) employ factor analysis to analyze the impact of personality traits (e.g., respect, teamwork, and interpersonal abilities) onthe hiring success of CEO candidates. Turning to demographic attributes, Bertrand and Schoar (2003) find that older managers
55G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
or those lacking MBA educations implement more conservative investment and financial policies. Similarly, while Malmendieret al. (2011) also find that leverage decreases with CEO age, they further discover that leverage increases with CEO tenure andprior military service, particularly amongWorldWar II veterans, suggesting lower risk aversion. Gender is another demographiccharacteristic associated with corporate decision-making. For example, the female CFOs in Huang and Kisgen's (2013) sampleare less likely than male CFOs to make acquisitions or issue debt. Additionally, Hutton et al. (forthcoming) discover that firmswith Republican managers adhere to more conservative policies such as lower levels of corporate debt, as well as capital andR&D expenditures. Further, Nicolosi and Yore (2013) reveal that a CEO's marital status is also related to corporate activities suchas mergers, joint ventures, security issuances, and divestitures. Finally, the religiosity of a firm's operating area (thus suggestiveof the executives' personal preferences) has also been linked to corporate actions. Specifically, firms headquartered in morereligious counties tend have lower capital and R&D investments (Hilary and Hui, 2009) and are less likely to engage inundesirable behavior such as options backdating, earnings manipulation, or excess executive compensation (Grullon et al.,2009).
The relationship between CEOs' demographic attributes and firm dividend policy is, however, unclear. One possibility is thatmanagerial optimism is more prevalent among subjects with certain traits, and their corresponding overinflated forecasts offuture performance encourage erroneous increases in dividend payout. An alternative explanation is that older firms concurrentlyface limited investments and possess hiring preferences for specific types of CEOs. In this case, any supposed correlation wouldactually be a byproduct of the effect of firm maturity on investment opportunities.
To investigate this matter, we study the relationship between six CEO profiles – formed by factor analysis from seventeendemographic characteristics – and firms' dividend yield, proclivity to substantially increase dividends, and subsequent firmperformance. We discover that companies led by married, Republican, Christian CEOs with children maintain higher dividendyields, are more likely to significantly hike dividends, and experience deteriorating performance following large dividendincreases. These tendencies exist in both the overall sample and a subset of mature firms, suggesting that firmmaturity and hiringpreferences are not the only explanation. Further, an event study centered on turnover between executives fitting differingprofiles also supports the previous results regarding CEOs with more traditional personal lives.
This study contributes to the literature on two fronts. First, we discover that another corporate decision – dividend policy – isrelated to leadership attributes. Unlike some studies in which executive qualities are time-limited or onerous for the averageinvestor to utilize as a sorting mechanism (e.g., Kaplan et al., 2012, or Malmendier and Tate, 2008), the profiles in this paper areconstructed from readily observable characteristics and thus have practical applicability. Second, if the tendency of married,Republican, Christian parent CEOs to raise (seemingly groundlessly) dividends is a result of overoptimistic firm forecasts, theseresults help resolve a conflict in the literature as to whether managerial overconfidence leads to an increase or decrease individend payout. While some authors (i.e., Ben-David et al., 2013; Cordeiro, 2009; Deshmukh et al., 2013) argue that managerialoptimismmay amplify Myers and Majluf's (1984) pecking order of capital structure, others (i.e., Bouwman, 2010; DeAngelo et al.,1996) counter that managerial optimism may entice biased managers to overuse dividends as a means of signaling superiorfuture firm performance (Bhattacharya, 1979; Miller and Rock, 1985). Our finding that CEOs with more traditional personal livesdisplay a propensity to increase dividends coupled with existing results that these same CEOs engage in other types ofoverconfident behavior (i.e., Nicolosi, 2012a,b) lends support to the latter argument.
2. Data
Using daily CRSP data for quarterly ordinary cash dividends of the common stock of non-financial and non-utilityfirms, a qualifying dividend increase is defined as a hike in adjacent dividends (i.e., those with 20 to 90 trading daysbetween announcement dates) of at least 12.5% (after adjusting for the effects of stock splits/dividends) but no more than500% (to exclude outliers). This methodology follows that of Grullon et al. (2002). Additionally, qualifying dividendincreases must also have non-missing non-problematic CRSP returns in the 252 trading days prior to and subsequent tothe event, or [−252,252].
We next pair qualifying dividend changes with the six demographic profiles constructed by Nicolosi (2012b) via principalcomponents analysis. The characteristics used to build the profiles include the CEO's age in years when first appointed to office;the number of years the executive was employed by the firm before the appointment; the CEO's gender; the CEO's education leveland prestige (i.e., whether the CEO attended Ivy League institutions); the CEO's educational field (i.e., business, technical, law, orother); the CEO's geographic background (i.e., birth place and higher education location); the CEO's political and religiousaffiliations; the CEO's military or entrepreneurial history; and whether the CEO is married, divorced, and/or has children. Detaileddefinitions for each characteristic are provided in the description of Table 1. Profile #1 describes a married, Christian, RepublicanCEO with children. Profile #2 describes a highly-educated CEO with business training. Profile #3 describes a highly-educated CEOwith legal training. Profile #4 describes an older, long-term employee-turned-CEO. Profile #5 describes an Asian CEO withtechnical training. Finally, Profile #6 describes a female Democrat CEO.
Altogether, the final sample spans 1980 to 2008 and covers 2386 CEOs at 1528 firms for a total of 16,185 firm-years duringwhich 2325 qualifying dividend increases occur. Sample descriptions are found in Table 1. Summary statistics for the firms and forthe qualifying dividend changes are provided in Panels A and B, respectively. For comparison, information regarding dividenddecreases as well as increases is supplied. Panel C contains the distribution of companies, executives, and dividend changes acrossthe years. Finally, Panel D is a duplication of the profiles' factor loadings from Nicolosi (2012b).
Table 1Summary statistics.
Panel A: Firm descriptors
Following Fama and French (2001), the size of firm f in fiscal year t, SIZEf,t, is proxied by the percent of NYSE-listed firms with smaller market capitalizations atthe end of the firm's previous fiscal year. A firm's age, FIRMAGEf,t, equals the number of years since the firm's initial public offering (if available) or the numberof years the firm has been listed in the COMPUSTAT database. A firm's market-to-book ratio, MVBVf,t, equals its equity market value plus its total asset bookvalue less its equity book value, where the sum is then scaled by its total asset book value. A firm's equity market value equals its outstanding number ofshares multiplied by the share price at the end of its fiscal year. The book value of a firm's equity equals the book value of its stockholders' equity less theliquidating value of the firm's preferred stock plus the firm's deferred taxes and tax credits (if available) less post-retirement assets (if available). A firm'sleverage ratio, LEVRATf,t, equals the book value of its short-term and long-term debt scaled by the market value of the firm's assets (the latter defined as thefirm's equity market value plus its total asset book value less its common equity book value). A firm's financial slack, SLACKf,t, equals its cash and short-termsecurities, scaled by its total asset book value. A firm's return on assets, ROAf,t, equals its operating income before depreciation, scaled by its beginning totalasset book value. A firm's dividend yield, DIVYLDf,t, equals the dividends per share paid during the fiscal year, scaled by the fiscal year's closing share price. Afirm's return volatility, VOLf,t, equals the standard deviation of the firm's daily equity CRSP holding period returns over the fiscal year. The annual risk-freerate, RFt, equals the Federal Reserve documented yield on a ten-year Treasury bond at the end of the firm's fiscal year. MVBV, SLACK, ROA, and DIVYLD havebeen Winsorized at the 1% and 99% levels. Excludes financial firms and utilities.
All distributions d are quarterly taxable cash dividends paid in U.S. dollars for ordinary common shares and have nonproblematic CRSP daily return data forthe 252 trading days prior to the distribution's announcement. The distributions must also have Compustat financial data available to construct the variablesin the previous panel. Following Grullon et al. (2002), we define a qualifying dividend change as one in which the absolute percent change in dividends(compared to the dividend announced 20 to 90 trading days prior) is between 12.5% and 500%. A firm's dividend yield, DIVYLDf,t, equals the dividends pershare paid during the fiscal year, scaled by the fiscal year's closing share price. The dividend per share of distribution d in dollars is denoted by DIVAMTd. Thedollar change of distribution d compared to the dividend announced 20 to 90 trading days prior is denoted by DIVCHGd. The percent change of distribution dcompared to the dividend announced 20 to 90 trading days prior is denoted by DIVCHG%d. Excludes dividends of financial firms or utilities.
This panel tracks the distribution of observations across time. The “Early” time period spans 1980 to 1994, the sample period studies by Malmendier and Tate(2005, 2008). The “Late” time period spans 1995 to 2008. We define a dividend initiation as the first cash dividend payment reported by CRSP for a firm.Following Michaely et al. (1995), a dividend omission occurs when a firm which has paid 6 consecutive quarterly cash dividends (no more than 90 tradingdays apart) fails to pay a cash dividend in the next calendar quarter (90 trading days). In order to match the omitted dividend to the executive in office, weassume an omission date equal to three months after the last declared dividend. We adhere to Boehme and Sorescu's (2002) methodology by defining adividend resumption as the first cash dividend announced by a firm after a payment hiatus between 33 to 180 months. Like Grullon et al. (2002), we identifyqualifying dividend changes as those in which the absolute percent change in dividends (compared to the dividends announced 20 to 90 trading days prior)are between 12.5% and 500%.
56 G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
Table 1 (continued)
Panel C: Distribution of data across time
This panel tracks the distribution of observations across time. The “Early” time period spans 1980 to 1994, the sample period studies by Malmendier and Tate(2005, 2008). The “Late” time period spans 1995 to 2008. We define a dividend initiation as the first cash dividend payment reported by CRSP for a firm.Following Michaely et al. (1995), a dividend omission occurs when a firm which has paid 6 consecutive quarterly cash dividends (no more than 90 tradingdays apart) fails to pay a cash dividend in the next calendar quarter (90 trading days). In order to match the omitted dividend to the executive in office, weassume an omission date equal to three months after the last declared dividend. We adhere to Boehme and Sorescu's (2002) methodology by defining adividend resumption as the first cash dividend announced by a firm after a payment hiatus between 33 to 180 months. Like Grullon et al. (2002), we identifyqualifying dividend changes as those in which the absolute percent change in dividends (compared to the dividends announced 20 to 90 trading days prior)are between 12.5% and 500%.
The table below is taken from Nicolosi (2012b) and displays the factor loadings of the size executive demographic profiles created via principal components.AGEi,f is CEO i's age in years when first appointed chief executive of firm f. TENUREi,f equals the number of years the executive was employed by firm f beforebeing appointed CEO. GENDERi is an indicator variable that equals one if the CEO is male. EDUCi is an indicator variable that equals four if the CEO earned agraduate degree; three if the CEO attended but did not graduate from a graduate school; two if the CEO earned an undergraduate degree; one if the CEOattended but did not graduate from an undergraduate school; zero otherwise. EDUC_IVYi is an indicator variable that equals one if the CEO was educated at anIvy League institution at either the undergraduate or graduate level. EDUC_BUSi is an indicator variable that equals one if the CEO was educated in a businessfield at either the undergraduate or graduate level. EDUC_TECHi is an indicator variable that equals one if the CEO was educated in a technical field (e.g.,engineering) at either the undergraduate or graduate level. EDUC_LAWi is an indicator variable that equals one if the CEO was educated in a legal field ateither the undergraduate or graduate level. CULTUREi is an indicator variable that equals one if the CEO was born or educated in an Asian country. FOUNDERi isan indicator variable that equals one if the CEO started his/her own company.MILITARYi is an indicator variable that equals one if the CEO served in the armedforces or attended military school. DEMOCRATi is an indicator variable that equals one if the CEO has been identified as a member or supporter of theDemocratic political party. REPUBLICANi is an indicator variable that equals one if the CEO has been identified as a member or supporter of the Republicanpolitical party. CHRISTIANi is an indicator variable that equals one if the CEO has been identified as a follower of a Christian religion. MARRIEDi is an indicatorvariable that equals one if the CEO has been married at least once. DIVORCEDi is an indicator variable that equals one if the CEO has been divorced at leastonce. CHILDRENi is an indicator variable that equals one if the CEO has at least one son or daughter.
57G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
58 G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
3. Analysis
3.1. Profiles and dividend yield
We begin our analysis by testing whether the CEO profiles are linked to firms' dividend yields. In Table 2, we run the followingTobit regression using annual pooled firm data:
Table 2Do CEOdependscaled band depless itsequipmexcludin
pendent variable, dividend yield (DivYld), is censored at zero. FirmControls is a set of control variables that includes lagged
The deannual firm sales, free cash flow, market-to-book ratio, and tangible assets. The construction of these variables is located in thetable description. However, the coefficients of interest (λ) are associated with the Profiles set, which contains the six principlecomponents or demographic profiles. Additionally, the regression model includes fixed effects for fiscal years and/or industries. Inall analyses, variables have been Winsorized at the 1% and 99% levels.
Looking at the results in Table 2, dividend yield is consistently higher under firms headed by CEOs with traditional personallives (i.e., Profile #1) and those who were older employees when they were promoted (i.e., Profile #4). For example, a one-unitincrease in Profile #1 significantly increases firm dividend yield in Model 2 by an average of 0.014%. To put this in context, a shiftfrom the first to third Profile #1 quartile (i.e.,−0.97 to 0.80) would result in an increase in firm dividend yield of 0.025%. Likewise,a shift from the first to third Profile #4 quartile (i.e., −0.64 to 0.52) would result in an increase in firm dividend yield of 0.034%.
However, while the profiles have distinct benefits (e.g., practicability, lack of multicollinearity, etc.), the drawback of factoranalysis is that a given company's CEO might not neatly align with any of the constructed profiles. Therefore, as a robustnesscheck, we next rerun the above regressions, substituting in for the six profiles the underlying individual characteristics originallyused to form the components. The Tobit regression model is adjusted as follows:
pendent variable DivYld is again censored at zero and the set FirmControls is unchanged. The coefficients of interest remain
The deλ, but they are now associated with the Characteristics set. This set contains only one regressor, rotating between each of theseventeen individual CEO characteristics. These individual characteristics are the CEO's age and firm tenure at positionassumption; gender; higher education level; Ivy League ties; previous training in business, engineering, or law; Asian heritage or
profiles affect firms' dividend yields? In this table, tobit regression is used to determine whether CEO profiles are related to dividend payout levels. Theent variable equals the dividend yield of firm f led by CEO i in fiscal year t. Dividend yield is defined as the dividends per share paid during the fiscal year,y the fiscal year's closing share price. Salesf,t equals the natural log of a firm's annual sales. A firm's cash flow equals its income before extraordinary itemsreciation, scaled by its lagged total asset book value. A firm's market-to-book ratio, MVBVf,t, equals its equity market value plus its total asset book valueequity book value, where the sum is then scaled by its total asset book value. A firm's tangible assets, TangAssetf,t, equal its net property, plant, andent scaled by total asset book value. The regression model includes fixed effects for fiscal year and industry (i.e., the twelve Fama-French industries,g financials and utilities). t-statistics are presented in brackets. One (two) asterisk(s) indicate significance at the 5% (1%) level.
16,185 16,185 16,185 16,185 16,185o R2 0.7619 0.5608 0.6907 0.7000 0.8129Es Yes No Yes No Yestry FEs Yes No No Yes Yes
Table 3Do CEO characteristics affect firms' dividend yields? In this table, tobit regression is used to determine whether individual CEO characteristics are related to dividend payout levels. The dependent variable equals thedividend yield of firm f led by CEO i in fiscal year t. Dividend yield is defined as the dividends per share paid during the fiscal year, scaled by the fiscal year's closing share price. Salesf,t equals the natural log of a firm's annualsales. A firm's cash flow equals its income before extraordinary items and depreciation, scaled by its lagged total asset book value. A firm's market-to-book ratio, MVBVf,t, equals its equity market value plus its total assetbook value less its equity book value, where the sum is then scaled by its total asset book value. A firm's tangible assets, TangAssetf,t, equal its net property, plant, and equipment scaled by total asset book value. The 17characteristics are defined in Table 1's Panel D. The regression model includes fixed effects for fiscal year and industry (i.e., the twelve Fama-French industries, excluding financials and utilities). t-statistics are presented inbrackets. One (two) asterisk(s) indicate significance at the 5% (1%) level.
60 G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
experience; entrepreneurial or military background; Democrat, Republican, or Christian affiliations; and whether the CEO hasever been married, divorced, or had children. The characteristics are included one at a time due to their correlation. As before, themodels also include fixed effects for industries and years.
Looking at the results in Table 3, half of the individual CEO attributes are related to firm dividend yield. Specifically, firmsheaded by CEOs who are older, highly-educated, previous long-term employees, married, or have children (male, founders, ordivorced) display significantly higher (lower) dividend yields. For example, firms led by individuals who marry exhibit dividendyields that are, on average, 0.01% higher. Overall, we find solid (partial) support for Table 2's positive association betweendividend yield and Profile #4 (#1) leaders in that both (half) of the underlying individual characteristics that positively load onthe profile generate the same significantly positive relationship with dividend yield.
3.2. Profiles and dividend increases
Next, we question whether, controlling for initial dividend yield, CEO profiles are linked to a firm's propensity to increase itsdividend payout. To that end, we use a conditional logistic regression model with pooled annual observations in which thedependent variable equals one if firm f led by CEO i significantly increases a quarterly cash dividend in fiscal year t.
ntrol variables in FirmControls are now firms' lagged size, age, market-to-book ratio, leverage ratio, financial slack, return on
The coassets, dividend yield, and stock return volatility, as well as the yield on a 10-year Treasury bond. The construction of thesevariables is found in the table description. The coefficients of interest (λ) are again associated with the Profiles set containing thesix demographic profile components. Fixed effects for both industries and/or years are included. The logistic coefficients arepresented as marginal effects.
As with dividend levels, firms' hike likelihoods in Table 4 are also linked to the demographic executive profiles. Supporting thefindings in Table 2, CEOs with traditional personal lives (i.e., Profile #1) or older employee-turned-CEOs (i.e., Profile #4) are morelikely to substantially increase dividends in a given fiscal year. Additionally, highly educated lawyers (i.e., Profile #3) andtechnically-trained Asians (i.e., Profile #5) are less likely to substantially increase dividends. For example, a one-unit increase inProfile #1 in Model 5 is associated with a 10.4% higher likelihood of the firm significantly increasing its dividend in a given year(i.e., the odds that dividends will be substantially increased in any such year shifts from an unadjusted odds ratio of 0.157 to0.173). In terms of probabilities, a one-unit increase in Profile #1 would be associated with a 1.16% higher probability of the firmsignificantly increasing its dividend in a given year.
As before, for robustness, in Table 5 we investigate the relationship between the same annual hike likelihood and theunderlying CEO characteristics. The conditional logistic regression model in Eq. (3) is adjusted by replacing the dropped Profilesset with the previously described Characteristics set. The remainder of the model is unchanged.
f the characteristics used to construct the component profiles are separately significantly linked to annual hike likelihoods.
Half oSpecifically, firms led by CEOs who are older, have prior employee experience, are Republican or Christian, or marry (have higherlevels of education or possess Ivy League or technical backgrounds) are significantly more (less) likely to substantially increasetheir dividend payout in a given fiscal year. For example, a CEO whomarries is 20.7% more likely than a CEO who never marries tosignificantly raise their firm's dividend per share (i.e., the odds that dividends will be substantially increased in any such yearshifts from an unadjusted odds ratio of 0.157 to 0.189). In terms of probabilities, a CEOwhomarries has a 2.20% higher probabilityof the firm significantly increasing its dividend in a given year than a CEO who never marries. Overall, we find solid (partial)support for Table 4's positive relation between hike likelihood and Profile #4 (#1) leaders in that both (three-quarters) of theunderlying individual characteristics that positively load on the profile generate the same significantly positive relationship.
3.3. Profiles, dividend increases, and subsequent performance
Thus far, it appears that two types of CEOs (i.e., Profiles #1 and #4) lead firms with higher dividend payout levels and hikelikelihoods. We next consider the performance of these CEOs' companies after significant dividend increases with the followingordinary least squares regression:
Y f ;t ¼ α þ βDivIncf ;t−1 þ FirmControlsϕþ Profilesγ þ λ DivInc f ;t−1�Profiles
� �þ ε ð5Þ
1 investigates firms' subsequent operating performance, and consequently the dependent variable Yf,t equals firm f's return
Modelon assets in year t. Considering subsequent firm performance over a longer time horizon, in Models 2 through 4 the dependentvariable equals firm f's compound return over a 36-, 48-, and 60-month holding period starting at the beginning of year t. Thecontrol variable DivIncf,t is an indicator variable that equals one if firm f increases its ordinary cash dividend payment by between12.5% to 500% in fiscal year t; zero otherwise. The set of control variables FirmControls remains the same as in the last two tables—i.e., it includes firms' lagged size, age, market-to-book ratio, leverage ratio, financial slack, return on assets, dividend yield, and
61G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
stock return volatility, as well as the yield on a 10-year Treasury bond. Similarly, the set of six demographic profile components,Profiles, also remains the same. The coefficients of interest in these models, however, are associated with the last six interactionterms. Each CEO profile factor is interacted with the dummy variable DivInc, resulting in the attached coefficient λ revealing therelation between the CEO profile values and firm performance in the fiscal year(s) following a significant dividend increase. Fixedeffects for both industries and years are included.
Looking at the results in Table 6, firm performance significantly suffers in the year(s) following a dividend increase incompanies headed by CEOs with traditional personal lives (i.e., Profile #1). Specifically, looking at the interaction coefficients inthe first model focusing on subsequent operating performance, a one-unit increase in Profile #1 in a dividend-hiking firm isassociated with a return on assets in the following fiscal year that is 0.50% lower. Turning to longer-term stock performance, aone-unit increase in Profile #1 in a dividend-hiking firm is associated with a 36-month return that is 5.3% lower, a 48-monthreturn that is 9.4% lower, and a 60-month return that is 13.7% lower. Profile #1 is the only executive profile that is consistentlyrelated to variations in subsequent firm performance following substantial dividend increases; dividend increases initiated byProfile #4 CEOs are not associated with any significant variations in subsequent firm stock or operating performance.
4. Discussion
4.1. Profiles and the dividend signaling hypothesis
At this point, it appears that firms led by CEOs with traditional personal lives select higher dividend yields, are more prone tosubstantially increase their dividend payout, and experience deteriorating firm performance in the year(s) following dividendhikes. We next consider why these relationships occur. One possible explanation is that Profile #1 CEOs may share a common biassuch as managerial optimism that leads them to overuse dividends as a means of signaling superior future firm performance. Thatis, if managers utilize dividend policy to communicate their private information regarding firm/project quality (e.g., Bhattacharya,1979; Miller and Rock, 1985), then overly-optimistic managers erroneously expecting improved firm performance would therebycommit to higher payout levels which would not subsequently be followed by substantial performance improvement. Indeed,surveys reveal that managerial forecasts regarding future firm performance do enter the dividend decision process (Brav et al.,2005; Lintner, 1956). This managerial optimism argument is offered by DeAngelo et al. (1996) as an explanation for the lack ofinformation embedded in dividend changes pursued by formerly successful firms experiencing earnings slumps. Similarly, themodel utilized in Bouwman (2010) predicts that optimistic CEOs will overestimate future cash flows and thus pursueunexpectedly high dividend payouts, all else equal. Supporting her premise, she discovers that dividend increases announced byoverconfident CEOs exhibit significantly higher cumulative abnormal announcement returns.
The psychology literature provides clues as to why Profile #1 CEOs may be susceptible to optimistic outlooks. Puri andRobinson (2007) reveal a positive relationship between marriage and over-optimism. Specifically, individuals who aremiscalibrated in terms of life expectancy forecasts are more likely to remarry. Other support for a link between marriage andoptimism comes from Grinblatt and Keloharju (2009). Studying Finnish males, the authors discover higher self-confidence amongthose who subsequently marry. Additionally, optimistic biases may increase with religiosity. After interviewing 600 adults withthe Attributional Style Questionnaire, Sethi and Seligman (1993) observe more optimistic outlooks among subjects belonging tofundamentalist faiths compared to members of more liberal faiths. Further, some research suggests that Republicans may berelatively more optimistic. With regard to concerns about technology, the environment, war, social deviance, and economictroubles, Wildavsky and Dake (1990) find conservatives to be less risk averse. With regard to 25 health hazards, Flynn et al.(1994) also detect lower perceptions of risk among conservative subjects. Finally, Costantini and Craik (1980) discover higheroptimism and more positive attitudes toward life among Republican versus Democrat presidential delegation members.
Empirical research also suggests that these CEOs with traditional personal lives exhibit optimistic behaviors at the corporatelevel. Specifically, these same CEOs have been shown to condition investment on free cash flow (Nicolosi, 2012b), anoverconfident activity investigated extensively by Malmendier and Tate (2005). In addition, these CEOs with traditional personallives also exhibit an increased propensity to issue callable debt which is unlikely to be subsequently called (Nicolosi, 2012a).
4.2. Profiles and the dividend free cash flow hypothesis
While the aforementioned results could be explained by Profile #1 CEOs attempting to erroneously signal their overoptimisticexpectations regarding future firm performance via increased dividend payout, an alternative explanation also exists. That is, it ispossible that any relationship between dividend payout and CEO profiles might simply be a reflection of firm characteristics.According to the free cash flow hypothesis (i.e., Jensen (1986)'s residual theory of dividends), CEOs who anticipate poor futurefirm prospects may return profits to shareholders in order to avoid the temptation to overinvest. In such a scenario, CEOs at moremature firms would have a proclivity to increase dividends. The negative MVBV coefficients in Tables 2 through 5 support thenotion that higher dividend payouts may be inversely associated with investment opportunities. Further, if mature firms facingdwindling prospects both increase their dividends as a means of disgorging themselves of tempting cash and prefer to hire Profile#1 CEOs, this would explain the suggested relationship between Profile #1 CEOs and both higher dividend payouts and lowersubsequent returns. In other words, if firms at various stages of their life cycle employ different types of CEOs, then therelationship observed thus far between dividend payment and CEO type might actually be proxying for the relationship betweendividend payment and firm maturity.
Table 4Do CEO profiles affect the likelihood that a firm will increase its dividends? In this table, conditional logistic regression is used to determine whether CEO profilesare related to dividend increase proclivity. The dependent variable equals one if firm f led by CEO i significantly increased a quarterly cash dividend in fiscal year t,indicated by an absolute percent change in the current announced dividend relative to the dividend announced 20 to 90 trading days prior of between 12.5% and500%. The control variables are defined in Table 1's Panel A. The regression model includes a constant as well as fixed effects for fiscal year and industry (i.e., thetwelve Fama-French industries, excluding financials and utilities). Coefficients are presented as marginal effects with Chi-square statistics in brackets. One (two)asterisk(s) indicate significance at the 5% (1%) level.
NObs 16,185 16,185 16,185 16,185 16,185NEvents 2194 2194 2194 2194 2194Pseudo R2 0.1395 0.1308 0.1575 0.1151 0.1427Year FEs Yes No Yes No YesIndustry FEs Yes No No Yes Yes
62 G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
To investigate this possibility, we next explore the distribution of CEO types across firms. In order to focus on CEOs who stronglyfit a particular profile, the sub-sample analysis in Table 7 considers only observations in which the CEO belongs to the top quintileof a particular factor profile. Panel A provides the mean values of various annual firm characteristics (e.g., total assets,MVBV, sales,profitability, free cash flow, book leverage, and firm age) associated with each top factor quintile. Looking at the results withoutconsidering significance, CEOs with traditional personal lives tend to work at larger, more profitable, and more heavily-leveredfirms.
Taking a closer look, Panel B tracks the proportion of these top quintile CEOs across various groupings of Compustat firms.Specifically, a sub-sample of the universe of Compustat companies5 is used to determine annual quintile breakpoints for the abovefirm characteristics. High factor CEO firms are then annually assigned to these quintiles in order to observe the distribution of CEOprofiles across different types of firms. For example, the first row in Panel B reveals that 38.18% of CEOs who strongly fit the Profile#1 description lead firms whose total assets place them in the largest Compustat size quintile; 22.02% of strong Profile #1 CEOslead firms belonging to the second largest Compustat size quintile when total assets are considered; 19.20% of the same CEOs leadfirms belonging to the middle size quintile; and 12.98% and 7.62% of the same Profile #1 CEOs lead firms belonging to the secondsmallest and smallest size quintiles, respectively. Looking at the results, CEOs who strongly fit the first profile do indeed tend to beconcentrated in mature firms (i.e., the distribution is monotonically increasing as firms become older, larger, and moreprofitable). However, the higher incidence of employment among larger, older firms is evident for most CEO types in Panel B,suggesting that these relationships may be an artifact of the data structure. That is, the biographical data for company leaders (ofany type) used to construct the profiles may only be readily available for larger, more established firms.
5 In order to better match the large capitalization firms featured in Execucomp, the entire CRSP/Compustat database is first annually reduced to only the firmswhich are current or previous constituents of either the S&P 500 or 1500 index. If the entire universe is instead examined, the Execucomp firms would largely befound in the upper size quintiles, thus obscuring the discrimination of finer size differences. The overall patterns remain, however, if the entire universe isconsidered when calculating breakpoints.
63G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
Therefore, we next try to determine whether CEO type is a proxy for firm maturity with a sub-sample regression analysis.Specifically, if older, established firms both rely on higher dividend payouts due to their limited prospects (i.e., the free cash flowhypothesis) and are more likely to hire a particular type of CEO, the relationship between CEO type and dividend payment amongmature firms should be weak or nonexistent. Thus, in Table 8 we rerun Table 4's logistic regression using only mature firms— i.e.,observations in which firms are in both the upper medians for their age (i.e., 22 years or greater) and size (i.e., marketcapitalizations exceed at least 69.9% of NYSE-listed firms that year).
Looking at the results in Table 8, mature firms led by CEOs with traditional personal lives are still more likely to increasedividends. For example, a one-unit increase in Profile #1 in Model 5 is associated with a 12.0% higher likelihood of the firmsignificantly increasing its dividend in a given year. In terms of probabilities, a one-unit increase in Profile #1 would be associatedwith a 1.84% higher probability of the firm substantially increasing its dividend in a given year. Thus, the observed relationshipbetween traditional CEOs and dividend increases does not simply reflect a proclivity of mature firms to simultaneously disgorgethemselves of cash due to inferior investment opportunities (as the free cash flow hypothesis suggests) and hire CEOs withtraditional personal lives. There is something about the leaders themselves which additionally impacts firms' payout likelihoods.
Our final examination into the possibility that a firm effect may be influencing the observed relationship between CEO typeand dividend payment utilizes the natural experiment provided by CEO turnover. Specifically, if dividend policy is unrelated toCEO attributes and is instead driven by firm characteristics (e.g., limited investment opportunities) then a switch in leadershipshould leave its dividend policy unaffected. Consequently, we approach this sub-sample analysis by first considering only theannual observations surrounding CEO turnover. Specifically, we retain the seven fiscal years centered on each regime change (i.e.,CEO switch)— three years before, three years after, and the transition year (i.e., the incoming CEO's first year). Each windowmustcontain non-missing control variable values and the incoming CEOmust stay at least three years following the transition year. Theindicator variable Postt equals one in the fiscal years under the new CEO's leadership (i.e., in event years 0 to +3). Next, qualifyingCEO switches are then classified based on a comparison between the incoming and the outgoing CEOs' profile values —
specifically, whether there is a substantial increase of 20% or more. The indicator variables IncPr1s–IncPr6s equal one if switch sresults in a 20% or greater increase in Profile 1–Profile 6 values, respectively; zero otherwise.
With the sub-sample formed and every CEO switch characterized along six dimensions according to relative changes in each ofthe six profiles, regression models similar to Eqs. (1) and (3) are utilized. First, Table 2's Tobit regression is modified asfollows.
The Controls set again includes the same lagged firm controls as in Table 2 but now additionally contains Postt. Further, insteadof the Profiles set, the set IncPr contains the six dummy variables (one at a time) indicating whether there has been a substantialincrease in each of the six demographic profiles. The coefficients of interest are those attached to the interaction of Post and theswitch dummy variables. For example, Post*IncPr1 equals one only in the years after a firm has hired a CEO who substantiallybetter fits the traditional personal life profile. Therefore, the λ associated with this interaction reveals the effect on dividend yieldthat occurs in the years following a switch to a demographically different type of CEO. The regression models again include fixedeffects for years and industries.
Looking at the results in Table 9, CEO turnover that considerably alters the leadership demographics has a significant impacton firm dividend payout. Specifically, in the years immediately following a switch to a CEO who substantially better fits thetraditional personal life profile (the highly-educated lawyer or older employee profiles), firms experience significantly higher(lower) dividend yields, on average.
Finally, we investigate the effect of CEO switches on hike likelihood. To that end, Table 4's logistic regression is modified asfollows.
ntrols set again includes the same lagged firm controls as in Table 4 but now additionally contains Postt. Further, instead of
The Cothe Profiles set, the set IncPr contains the six dummy variables (one at a time) indicating whether there has been a meaningfulincrease in each of the six demographic profiles. The coefficients of interest remain those attached to the interaction of Post andthe switch dummy variables.
Looking at the results in Table 10, CEO turnover that substantially alters the leadership demographics also has a significantimpact on hike likelihood. Specifically, in the years immediately following a switch to a CEO who considerably better fits thetraditional personal life or older employee profile (the highly-educated lawyer or technically-trained Asian profiles), firms aresignificantly more (less) likely to materially raise their cash dividend per share. For example, in a year soon after switching to aCEO who substantially better fits Profile #1 (i.e., Model 2 in Table 10), a firm has a 40.4% higher likelihood (or a 3.50% higherprobability) of significantly increasing its dividend. Altogether, the results suggest that CEO type is not just a proxy for firmmaturity and thus limited investment opportunities, but instead firm dividend policies are related to leadership demographics.
5. Conclusions
A growing subset of the corporate finance literature reveals that managerial traits impact workplace decisions. In this paper,we consider the relationship, if any, between dividend policy and six different CEO profiles constructed from seventeen
Table 5Do CEO characteristics affect the likelihood that a firm will increase its dividends?In this table, conditional logistic regression is used to determine whether individual CEO characteristics are related to dividend increase proclivity. The dependent variable equals one if firm f led by CEO i significantlyincreased a quarterly cash dividend in fiscal year t, indicated by an absolute percent change in the current announced dividend relative to the dividend announced 20 to 90 trading days prior of between 12.5% and 500%. Thecontrol variables and individual characteristics are defined in Table 1's Panel A and D, respectively. The regression model includes a constant as well as fixed effects for fiscal year and industry (i.e., the twelve Fama-Frenchindustries, excluding financials and utilities). Coefficients are presented as marginal effects with Chi-square statistics in brackets. One (two) asterisk(s) indicate significance at the 5% (1%) level.
Table 6Subsequent firm performance and CEO profiles. In this table, ordinary least squares regression is used to determine whether CEO profiles are related to firmperformance following significant dividend increases. In the first model, the dependent variable equals the firm's return on assets in the year following thedividend increase. The dependent variable the final three models equals the compound holding return over a 36, 48, and 60 month period, respectivelybeginning the fiscal year after the dividend increase. The first nine control variables are defined in Table 1's Panel A. DivInc is an indicator variable that equals oneif the firm increased its ordinary cash dividend payment by 12.5–500% in a given fiscal year. The regression model includes a constant as well as fixed effects fofiscal years and industries (i.e., the twelve Fama-French industries, excluding financials and utilities). One, two, and three asterisks indicate significance at the10%, 5%, and 1% levels, respectively.
66 G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
,
r
observable demographic characteristics. Specifically, we investigate the association between these CEO types and dividend yield,dividend hike likelihood, and firm performance subsequent to substantial dividend increases. Pooled regression analysis revealsthat firms led by married, Republican, Christian CEOs with children select higher dividend yields and are more likely tosignificantly increase their dividend payout. Following substantial dividend hikes, firms led by these same CEOs exhibitdeteriorating operating and stock performance.
One possible explanation for these findings is that CEOs with these backgrounds are more susceptible to optimistic tendencies.Their overinflated forecasts regarding future firm performance, which subsequently fail to materialize, lead them to increasedividend payout. Alternatively, if mature firms facing dwindling investment opportunities both increase their dividends as ameans of disgorging themselves of tempting cash and prefer to hire this type of CEO, then CEO type would be proxying for a firm'sstage in its life cycle. While CEOs with these more traditional backgrounds do indeed tend to be employed at larger, older, morestable firms, when we restrict the sample to only mature firms (i.e., older companies with larger market capitalizations), theseCEOs still display a higher propensity to substantially increase dividends. Further, an event study centered on CEO turnoverreveals higher dividend yields and hike likelihoods in the fiscal years immediately following transitions to CEOs that more closely
Table 7Distribution of CEO profiles across firms. This table considers only CEOs who strongly fit a particular profile (i.e., after sorting each profile by factor loading, onlyCEOs in the top quintile for each factor are included). Panel A provides the mean values of the various annual firm variables (defined below) associated with eachof the top quintile CEOs. Panel B tracks the proportion of the top quintile CEOs across Compustat firms. Specifically, in order to better match the largecapitalization firms featured in Execucomp, the entire CRSP/Compustat database is reduced to only the firms which are current or previous constituents of eitherthe S&P 500 or 1500 index. After computing the firm-level variables defined below (and winsorizing at the 1% and 99% level), quintile breakpoints are calculatedannually for each variable using this reduced universe of large capitalization Compustat firms. High factor CEO firms are then assigned to annual quintiles in orderto observe the distribution of CEO profiles across different types of firms. TotalAssets equals the natural log of the firm's total assets. MVBV equals the ratio of themarket value relative to book value of the firm's assets. The market value of its assets is defined as the sum of the market value of its equity, its total debt incurrent liabilities, its long term debt, the liquidating value of its preferred stock, less its deferred taxes and investment tax credit. The market value of a firm'sequity equals its share price at the close of its fiscal year multiplied by its common shares outstanding. Sales equals the natural log of the firm's total net sales.Profitability equals the firm's operating income before depreciation scaled by lagged total assets. FCF is the sum of the firm's cash flow income beforeextraordinary items plus its depreciation and amortization, scaled by lagged total net property, plant and equipment. A firm's BookLeverage equals the sum of itstotal debt in current liabilities plus its long term debt, scaled by total assets. FirmAge equals the number of years elapsed since the firm's first appearance in CRSP.
Panel A: Mean firm characteristics of top quintile CEOs
67G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
Table 8Do CEO profiles affect the likelihood that a mature firm will increase its dividends? In this table, conditional logistic regression is used to determine whether CEOprofiles are related to dividend increase proclivity. This sub-sample includes only those observations in which firm age exceeds the sample median of 22 yearsand firm size exceeds the sample median of 69.9% (i.e., the firm's market capitalization exceeds at least 69.9% of NYSE listed firms in a given year). The dependentvariable equals one if firm f led by CEO i significantly increased a quarterly cash dividend in fiscal year t, indicated by an absolute percent change in the currentannounced dividend relative to the dividend announced 20 to 90 trading days prior of between 12.5% and 500%. The control variables are defined in Table 1'sPanel A. The regression model includes a constant as well as fixed effects for fiscal year and/or industry (i.e., the twelve Fama-French industries, excludingfinancials and utilities). Coefficients are presented as marginal effects with Chi-square statistics in brackets. One (two) asterisk(s) indicate significance at the 5%(1%) level.
NObs 5442 5442 5442 5442 5442NEvents 1116 1116 1116 1116 1116Pseudo R2 0.1573 0.1060 0.1636 0.1024 0.1606Year FEs Yes No Yes No YesIndustry FEs Yes No No Yes Yes
68 G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
fit this traditional personal life profile. Overall, the results suggest that these relationships are not simply a firm effect but areinstead related to CEO demographic characteristics.
In addition to illuminating another corporate decision affected by leadership attributes, this study contributes to the CEOoptimism literature on several fronts. First, we help resolve the uncertainty surrounding the effect of CEO overconfidence oncorporate dividend policy. Specifically, given that CEOs with traditional personal lives have been linked to overconfident actionselsewhere (i.e., Nicolosi, 2012a,b, reveals that these same CEOs display positive investment-cash flow sensitivity and rely oncallable bond issuance, respectively), our findings that they additionally display a propensity to unnecessarily increase dividendslend further support to the conjecture that optimism and payout are positively related (e.g., Bouwman, 2010; DeAngelo et al.,1996). Second, while the existing CEO overconfidence literature relies on behavioral proxies not easily utilized by the commoninvestor, this study's use of observable executive characteristics allows investors to effortlessly act on the relationships revealed.As such, our results are more accessible and applicable to the average layman. Third, while the corporate overconfidenceempirical studies have mostly been confined to the period spanning only to 1994, our study extends to the more recent past (i.e.,to 2008).
References
Baker, M., Wurgler, J., 2004a. Appearing and disappearing dividends: the link to catering incentives. J. Financ. Econ. 73, 271–288.Baker, M., Wurgler, J., 2004b. A catering theory of dividends. J. Finance 59, 1125–1165.Ben-David, I., Graham, J.R., Harvey, C.R., 2013. Managerial miscalibration. Q. J. Econ. 128 (4).Bertrand, M., Schoar, A., 2003. Managing with style: the effect of managers on firm policies. Q. J. Econ. 118, 1169–1208.Bhattacharya, S., 1979. Imperfect information, dividend policy, and “The bird in the hand” fallacy. Bell J. Econ. 10, 259–270.
69G. Nicolosi / Journal of Corporate Finance 23 (2013) 54–70
Boehme, R.D., Sorescu, S.M., 2002. The long-run performance following dividend initiations and resumptions: underreaction or product of chance? J. Finance 57,871–900.
Bouwman, C.H.S., 2010. Managerial optimism and the market's reaction to dividend changes. Working Paper. Case Western Reserve University.Brav, A., Graham, J.R., Harvey, C.R., Michaely, R., 2005. Payout policy in the 21st century. J. Financ. Econ. 77, 483–527.Cordeiro, L., 2009. Managerial overconfidence and dividend policy. SSRN Working Paper. London Business School.Costantini, E., Craik, K.H., 1980. Personality and politicians: California party leaders, 1960–1976. J. Personal. Soc. Psychol. 38, 641–661.DeAngelo, H., DeAngelo, L., Skinner, D.J., 1996. Reversal of fortune-dividend signaling and the disappearance of sustained earnings growth. J. Financ. Econ. 40,
341–371.Denning, L., Peaple, A., 2012. Clear gap between Shell and Exxon. Wall Street Journal. Dow Jones & Company, Inc, New York, NY.Deshmukh, S., Goel, A.M., Howe, K.M., 2013. CEO overconfidence and dividend policy. J. Financ. Intermed. 22, 440–463.Fama, E.F., French, K.R., 2001. Disappearing dividends: changing firm characteristics or lower propensity to pay? J. Financ. Econ. 60, 3–43.Flynn, J., Slovic, P.S., Mertz, C.K., 1994. Gender, race, and perception of environmental health risks. Risk Anal. 14, 1101–1108.Grinblatt, M., Keloharju, M., 2009. Sensation seeking, overconfidence, and trading activity. J. Finance 64, 549–578.Grullon, G., Michaely, R., Swaminathan, B., 2002. Are dividend changes a sign of firm maturity? J. Bus. 75, 387–424.Grullon, G., Kanatas, G., Weston, J.P., 2009. Religion and corporate (mis)behavior. Working Paper. Rice University.Hilary, G., Hui, K.W., 2009. Does religion matter in corporate decision making in America? J. Financ. Econ. 93, 455–473.Huang, J., Kisgen, D.J., 2013. Gender and corporate finance: are male executives overconfident relative to female executives? J. Financ. Econ. 108, 822–839.Hutton, I., Jiang, D., Kumar, A., 2013. Corporate policies of republican managers. J. Financ. Quant. Anal. (Forthcoming).Jensen, M.C., 1986. Agency costs of free cash flow, corporate finance, and takeovers. Am. Econ. Rev. 76, 323–329.Kaplan, S.N., Klevanov, M.M., Sorenson, M., 2012. Which CEO characteristics and abilities matter? J. Finance 67, 937–1007.Lintner, J., 1956. Distribution of incomes of corporations among dividends, retained earnings, and taxes. Am. Econ. Rev. 46, 97–118.Lohr, S., 2012. Cisco Revenue and Profit Exceed Forecasts as Orders in Asia Stay Strong, New York Times. New York Times Company, New York, NY.Malmendier, U., Tate, G., 2005. CEO overconfidence and corporate investment. J. Finance 60, 2661–2700.Malmendier, U., Tate, G., 2008. Who makes acquisitions? CEO overconfidence and the market's reaction. J. Financ. Econ. 89, 20–43.Malmendier, U., Tate, G., Yan, J., 2011. Overconfidence and early-life experiences: the effect of managerial traits on corporate financial policies. J. Finance 66,
1687–1733.Michaely, R., Thaler, R.H., Womack, K.L., 1995. Price reactions to dividend initiations and omissions — overreaction or drift? J. Finance 50, 573–608.Miller, M.H., Rock, K., 1985. Dividend policy under asymmetric information. J. Finance 40, 1031–1051.Myers, S.C., Majluf, N.S., 1984. Corporate financing and investment decisions when firms have information that investors do not have. J. Finance 13, 187–221.Nicolosi, G., 2012a. Callable debt demographics. Working Paper. Northern Illinois University.Nicolosi, G., 2012b. Profiling CEOs: an investment analysis. Working Paper. Northern Illinois University.Nicolosi, G., Yore, A., 2013. “I do”: does marital status affect how much CEOs “do”? Working Paper. Northern Illinois University.Puri, M., Robinson, D.T., 2007. Optimism and economic choice. J. Financ. Econ. 86, 71–99.Sethi, S., Seligman, M.E.P., 1993. Optimism and fundamentalism. Psychol. Sci. 4, 256–259.Smith, C.W., Watts, R.L., 1992. The investment opportunity set and corporate financing, dividend, and compensation policies. J. Financ. Econ. 32, 263–292.Wildavsky, A., Dake, K., 1990. Theories of risk perception: who fears what and why? Daedalus 119, 41–60.
Table 9Dividend yield following changes in CEO profiles. The subsample in this table contains annual observations for the seven year window (centered on the incomingexecutive's first year) surrounding CEO turnover. Incoming CEOs must hold their position for four years, during which the indicator variable Post equals one underthe new leadership. The dependent variable, DivYld, in the tobit regressions is defined as the dividends per share paid during the fiscal year, scaled by the fiscalyear's closing share price. IncPr1–IncPr6 are indicator variables that equal one if the new CEOs' Profiles 1–6, respectively, are at least 20% greater than theirsuccessors'. Models include fixed effects for years and industries. t-statistics are listed in brackets. One (two) asterisk(s) indicate significance at the 5% (1%) level.
Variable Model 1 Model 2 Model 2 Model 4 Model 5 Model 6 Model 7
Table 10Dividend increase likelihood following changes in CEO profiles. The subsample in this table contains annual observations for the seven year window (centered onthe incoming executive's first year) surrounding CEO turnover. Incoming CEOs must hold their position for four years, during which the indicator variable Postequals one under the new leadership. The dependent variable, DivInc, in the conditional logistic regressions equals one if firm f led by CEO i significantly increaseda quarterly cash dividend in fiscal year t, indicated by an absolute percent change in the current announced dividend relative to the dividend announced 20 to 90trading days prior of between 12.5% and 500%. IncPr1–IncPr6 are indicator variables that equal one if the new CEOs' Profiles 1–6, respectively, are at least 20%greater than their successors'. Models include fixed effects for years and industries. Coefficients are presented as marginal effects with Chi-square statistics inbrackets. One, two, and three asterisks indicate significance at the 10%, 5%, and 1% levels, respectively.
Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7
