Professional Documents
Culture Documents
Working Paper
Dividend Smoothing : An Agency Explanation and
New Evidence
Reference: Knyazeva, Anzhela (2014). Dividend Smoothing : An Agency Explanation and New
Evidence. [S.l.] : SSRN.
https://ssrn.com/abstract=2504715.
https://doi.org/10.2139/ssrn.2504715.
doi:10.2139/ssrn.2504715.
Kontakt/Contact
ZBW – Leibniz-Informationszentrum Wirtschaft/Leibniz Information Centre for Economics
Düsternbrooker Weg 120
24105 Kiel (Germany)
E-Mail: rights[at]zbw.eu
https://www.zbw.eu/econis-archiv/
Anzhela Knyazeva†
U.S. Securities and Exchange Commission
Diana Knyazeva
U.S. Securities and Exchange Commission
Abstract
In spite of considerable research into firm dividend behavior, dividend smoothing has eluded a definitive
explanation. This paper provides an agency interpretation of dividend smoothing and offers evidence that
variation in corporate governance and managerial incentive conflicts explains differences in intertemporal
properties of dividends. We argue that smooth dividends are an alternative to traditional corporate
governance mechanisms. Empirically, we document a greater degree of dividend smoothing, fewer
dividend cuts, and a trend towards regular incremental dividend increases at firms with weak traditional
monitoring mechanisms. The effect of governance on dividend changes is largest for firms with high free
cash flow. We document consistent patterns for total shareholder payout and overall commitment to
external claimholders. However, dividends and repurchases are not perfect substitutes and adjustments to
repurchases are secondary to the weakly governed managers’ need to sustain dividends.
*
An earlier version of the paper was circulated as “Delivering on the Dividend Promise: Corporate Governance, Managerial
Incentives, and Dynamic Dividend Behavior.” The authors acknowledge financial support from the Stern School of Business at
New York University and the Simon School of Business at the University of Rochester. The authors thank Kose John, Joseph
Stiglitz, Daniel Wolfenzon, David Yermack, and Bernard Yeung for valuable discussions and Bin Chang, Vieira Elisabete,
William Greene, Gustavo Grullon, Shane Heitzman, Michael Lemmon, John Long, Sattar Mansi, Andrew Metrick, Roni
Michaely, Eli Ofek, Micah Officer, Carrie Pan, Anthony Saunders, Bill Schwert, Cliff Smith, Raghu Sundaram, Jerry Warner,
Toni Whited, Jeffrey Wurgler, Jerry Zimmerman, and seminar and conference participants at New York University, University of
Rochester, University of Michigan, Ohio State University, University of Maryland, University of North Carolina – Chapel Hill,
Emory University, University of Wisconsin, Arizona State University, University of Georgia, Baruch College, Georgia State
University, Georgia Tech, Financial Management Association, and European Financial Management Association meetings for
helpful comments. We thank Institutional Shareholder Services for providing Corporate Governance Quotient data. All errors and
omissions are our own.
The Securities and Exchange Commission, as a matter of policy, disclaims responsibility for any private publication or statement
by any of its employees. The views expressed herein are those of the authors and do not necessarily reflect the views of the
Commission or of the authors’ colleagues on the staff of the Commission.
†
Corresponding author: Anzhela Knyazeva, Securities and Exchange Commission, 100 F Street NE, Washington, DC 20549. E-
mail: anzhela.knyazeva@gmail.com.
Electroniccopy
Electronic copy available
available at:
at:https://ssrn.com/abstract=2504715
http://ssrn.com/abstract=2504715
1. Introduction
The issue of intertemporal patterns in dividends has remained unresolved in the finance literature
for over half a century. Since Lintner (1956) and Fama and Babiak (1968) documented the
propensity of firms to smooth dividends and reluctance to make cuts, there has been a need for a
better understanding of dynamic dividend behavior. We approach this question from a corporate
governance perspective. In the presence of agency costs, stable dividends can help disgorge free
cash flow and constrain inefficient managerial behavior, however, their effectiveness is limited if
This paper contributes to the existing literature in several ways. First, the analysis
dividend smoothing, changes to dividend levels, and the role of dividend changes in the broader
context of payout policy and financial structure adjustments. Second, our findings have
implications for the credibility of dividend commitments at firms with weak traditional
governance mechanisms. Although such commitments are implicit, our evidence strongly
supports the adherence of managers to dividend commitments. Third, our paper sheds light on
the differences between firms with weak versus strong traditional governance mechanisms with
Holding the firm’s investment opportunities constant, weak governance mechanisms can
result in misalignment of managers and shareholders and lead to inefficient investment. Stable
dividends limit managerial discretion and free cash flow at the manager’s disposal. In firms that
find it too costly to use traditional oversight devices, an implicit commitment to stable dividends
The decision to adhere to a dividend policy over time can be reconciled with managerial
self-interest. While shareholders cannot observe the quality of investment choices, they can
Electroniccopy
Electronic copy available
available at:
at:https://ssrn.com/abstract=2504715
http://ssrn.com/abstract=2504715
observe dividends, cash flow realizations, and monitoring mechanisms. All else equal, when
traditional monitoring mechanisms are weak, dividend deviations are more likely to contribute to
agency costs, resulting in firm value loss. This is rationally expected by investors. The manager
weighs gains from a dividend cut against adverse shareholder response to dividend deviations.
Myers (2000) models a dynamic tradeoff facing a self-interested manager that enters into an
implicit dividend contract with investors: the manager can forgo dividends but investors can
choose not to fund the firm’s projects (that the manager seeks to pursue). Further, as Zwiebel
(1996) and Fluck (1998, 1999) show, shareholders can attempt to intervene and increase scrutiny
of the manager if the loss of firm value from the deviation is expected to be high.
The alternative hypothesis is that weak traditional governance mechanisms lead to more
deviations from the dividend commitment as misaligned managers have more discretion to cut
dividends.
The main empirical findings are as follows. Weak traditional governance mechanisms are
associated with more stable dividends and fewer dividend cuts or omissions. Weak governance
firms pass through more cash flow increases as dividend increases, however, weakly governed
managers are cautious about the magnitude of increases (since they have to be sustained over
time) and prefer small increases. Changes in dividend levels are decreasing in the strength of
traditional governance mechanisms, particularly in the presence of large cash flow increases.
Similarly, smoothing and changes in total payout are decreasing in the strength of corporate
governance. However, repurchases and dividends are not perfect substitutes: compared to
strongly governed firms, weakly governed firms are more likely to cut repurchases when they
need to decrease payout and to increase dividends when they need to make additional cash
distributions. Our results hold across a number of governance measures, including our new
endogeneity.
The rest of the paper is organized as follows. The second section develops the hypothesis
and testable predictions. The third section discusses data and methodology. The fourth section
2. Hypotheses
Related work
Prior work shows that dividends are persistent in the data (e.g., Lintner, 1956; Fama and Babiak,
1968; Dewenter and Warther, 1998; Gugler, 2003). Michaely and Roberts (2012) find that
private firms have less persistent dividends. Several papers theoretically model dividend
smoothing (e.g., Kumar, 1988; Warther, 1994; Fudenberg and Tirole, 1995; Karpavicius, 2014).
across firms.
Other studies document a positive market reaction to dividend increases and a negative
market reaction to dividend cuts (e.g., Aharony and Swary, 1980; Asquith and Mullins, 1983).
Officer (2011) finds a more favorable reaction to dividend initiations for weakly governed firms.
Evidence of the value relevance of dividend changes in a setting with frictions helps motivate
our argument about the equity market response to observable dividend decisions.
structure (e.g., Rozeff, 1982; Amihud and Li, 2006), governance (e.g., La Porta et al., 2000;
John, Knyazeva, and Knyazeva, 2011b; Hu and Kumar, 2004; Pan, 2007; Harford, Mansi, and
Maxwell, 2008), or information asymmetries (e.g., Booth and Xu, 2008). An important question
This paper contributes to existing literature in the following ways. Our findings provide
new insights into intertemporal patterns in corporate cash distributions. We examine weakly
governed managers’ choice to smooth, uphold or deviate from dividends over time; dividend
changes in response to cash flow shocks; and dividend decisions in the context of overall payout
policy adjustment. The analysis has implications for the credibility of the implicit dividend
contract when traditional governance mechanisms are weak, hence dividends matter most from
the agency perspective yet managers can deviate from nonbinding commitments more easily.
The findings also suggest that any analysis of aggregate dividend trends should be conditional on
governance quality. Our empirical setting refines governance and alignment measures and
to observe investment decisions gives rise to agency conflicts and inefficient managerial
investments motivated by empire building, diversification, or private benefits (e.g., Jensen and
Meckling, 1976; Rozeff, 1982). Dividends can alleviate such agency conflicts (at a cost) by
limiting free cash flow available to the manager and exposing the manager to additional
monitoring (Jensen, 1986; Easterbrook, 1984). Dividends can be viewed as an implicit contract
between the manager and shareholders (Myers, 2000). Following the initial dividend adoption
decision, managers can choose whether to uphold this implicit dividend promise or to deviate
from it by cutting or eliminating dividends. Unlike the quality of investment projects chosen by
the manager, dividend changes and governance mechanisms are observable and verifiable. When
efficient investment and a larger firm value loss. In contrast, in the presence of strong
governance mechanisms, additional constraints on managerial decisions are not necessary and
manager in Myers (2000). A dividend cut can improve the manager’s contemporaneous utility by
offering more discretion in investment decisions. However, the manager’s expected utility from
continuing in the firm is decreased by adverse shareholder reaction to the observable dividend
decision: In the Myers (2000) model, investors can limit the amount of capital provided to the
firm in response to a deviation from the implicit dividend contract. The resulting decrease in the
availability of capital constrains the manager from undertaking privately beneficial investment
projects in the future, should internal cash flow become insufficient. If expected firm value loss
from the manager’s decision is substantial, the manager can also face the threat of intervention or
additional scrutiny from shareholders. (In a related vein, Zwiebel (1996) and Fluck (1998, 1999)
model the manager’s dynamic tradeoff between current gains and future control challenge threats
in the context of capital structure choice.) Therefore, the manager’s dividend change decision
The main takeaway is that weak corporate governance can be consistent with a credible
implicit dividend contract. Importantly, a stable dividend policy can offset weaknesses in
traditional corporate governance mechanisms. This yields several implications for intertemporal
dividend behavior. The main testable prediction is that weakly governed managers uphold the
1
Existing empirical literature is divided on the question of signaling benefits of dividends in relation to unexpected increases in
future earnings (e.g., Benartzi, Michaely, and Thaler, 1997).
We now turn to dividend increases at weakly governed firms. If cash flows do not
change, managers are expected to maintain the dividend level. If cash flows temporarily
decrease, managers raise capital from external financing markets (Easterbrook, 1984). In case of
investments. Thus, when traditional monitoring mechanisms are weak, the probability of
dividend increases is weakly higher overall and strictly higher if cash flows increase. However,
given the anticipated pressure to uphold dividends in the future, weakly governed managers will
exercise caution and avoid large increases that may become unsustainable in the future. If cash
flows continue to increase, weakly governed managers will raise dividends incrementally.
The described dynamic dividend behavior patterns have implications for overall payout
decisions of weakly governed managers. First, total cash distributions are expected to be more
persistent and payout changes are expected to be higher when governance is weak. Second,
payout policy adjustments will minimize deviations from the implicit dividend contract, leading
to more cuts in repurchases rather than cuts in dividends when managers decrease payouts.
Similarly, shareholders expect misaligned managers to distribute extra cash in the form of higher
dividends, which sustains the constraint on managerial investment behavior. Higher share
repurchases have a weak commitment effect since the timeline and amounts of payouts are less
regular payments and imposes the threat of bankruptcy in the case of deviations (Jensen, 1986;
John, Knyazeva, and Knyazeva, 2011b). Both debt and implicit dividend commitments reduce
the firm’s cash flow through a cash distribution to external claimholders. Both forms of
as specific costs such as tax disadvantage of dividends, bankruptcy costs, and agency costs of
debt). Like with dividends, deviations from debt commitments lead to significant negative
consequences, so changes to debt and dividend commitments could act as partial substitutes.
Our testable predictions are summarized below. All else equal, we expect weak
(4) more overall payout smoothing (with a preference towards the use of repurchase cuts for
lowering payouts and dividend increases for raising payouts and partial substitution between debt
discretion with respect to dividend deviations, resulting in less smoothing and more dividend
cuts. We believe this is less likely for two reasons. First, this paper focuses on US firms and
examines the role of firm-level monitoring mechanisms in an environment with strong overall
investor protections, which is different from countries with few investor protection laws where
potential for adverse shareholder action is limited (La Porta et al., 2000). Second, in a setting
with a developed stock market and more dependence on equity financing, meeting shareholder
concerns is likely to be more important than in economies that are more dependent on bank
financing.
2
Given that dividend cuts tend to be significantly larger in magnitude, changes in dividend levels are most likely decreasing in
the strength of traditional corporate governance mechanisms.
Sample
The main tests use the sample of U.S. firms from Compustat Annual for 1993–2004. Sensitivity
tests extend the sample period through 2009 (certain governance metrics are unavailable, so we
modify the governance index definition accordingly). Tests of the propensity to pay dividends
use the full sample of firms. Tests of intertemporal dividend behavior use the sample of past
dividend payers. We exclude financial firms (SIC 6000–6999), regulated utilities (SIC codes
4900–4949), observations with assets below 20 mln, firms incorporated outside the US, and
LBOs. CRSP monthly file is used to obtain dividend data for ordinary common shares, unless
specified otherwise. Certificates, ADRs, shares of beneficial interest, units, Americus trust
components, closed-end funds, and REITs are excluded. Dividend is defined as the annual sum
of ordinary quarterly cash dividends (CRSP distribution code 1232), adjusted for splits and stock
The Appendix contains detailed variable definitions. We use proxies for monitoring by boards,
institutional blockholders, corporate control markets, lower separation of ownership and control
in single class firms, and CEO-level alignment characteristics (CEO Characteristics), as well as
the aggregate of the above measures, Alignment and Governance Index (AGI). We convert
continuous measures into annual rankings, defined such that higher values correspond to stronger
The proportion of outside directors on the board (see, e.g., Rosenstein and Wyatt, 1990;
Agrawal and Knoeber, 1996; Knyazeva, Knyazeva, and Masulis, 2013), small board size
(Yermack, 1996), separation between the CEO and key committees (Shivdasani and Yermack,
1999), and the frequency of board meetings (Vafeas, 1999) have been tied to the intensity of
as more active monitors (Cremers and Nair, 2005). Further, concentration of institutional
Antitakeover defenses can shield managers from the discipline of the corporate control market
and shift more bargaining power from to shareholders to managers (see, e.g., Gompers, Ishii, and
Metrick, 2003; Bebchuk and Cohen, 2005; Bebchuk, Cohen, and Ferrell, 2005; Bertrand and
Mullainathan, 2003). The G index comes from the Gompers et al. (2003) dataset and the
Bebchuk et al. (2005) E index is constructed from RiskMetrics data. Certain CEO characteristics
can affect the degree of alignment of manager and shareholder interests. Longer CEO tenure in
the firm may increase managerial power over the board (Shivdasani and Yermack, 1999). CEO
age could be associated with a shorter horizon of the manager in the firm and potential
misalignment with shareholder interests. Higher CEO ownership can facilitate alignment (used in
some of the tests). Dual classes of shares exacerbate the separation of voting and cash flow
rights. Dual class firms are identified in Gompers, Ishii, and Metrick (2005) for 1994–2002 and
in RiskMetrics for the remaining years. The five measures described above are equally weighted
to construct AGI. For robustness, we define AGI using rescaled continuous variables instead of
rankings; simplify AGI to the core index dimensions available for a longer sample period; and
perform factor analysis of individual governance measures from the main sample and
construct governance factors. Our results are highly robust to alternative governance metrics.
Prior work has raised concerns about endogeneity in the relation between managerial
ownership, governance, and firm performance (see, e.g., Himmelberg, Hubbard, and Palia, 1999;
3
Amihud and Li (2002) argue that a lower degree of information asymmetry between management and institutional owners is the
underlying reason for the negative relation between institutional ownership and dividend payout.
and Rusticus, 2006; Hermalin and Weisbach, 1998). To the extent that dividend commitments
and strong corporate governance are alternative ways of curbing inefficient investment, firms
may jointly choose optimal dividend policy and governance quality depending on the relative
costs of monitoring and persistent dividends. Although no empirical test or correction can fully
eliminate endogeneity, we recognize this concern and attempt to address it in several ways.
First, we use the instrumental variables approach.4 Instruments should predict corporate
governance but they should not affect dividend changes, except through the governance channel.
We propose the following determinants of firm-level governance quality: (i) industry governance
practices; (ii) state laws; (iii) initial corporate control market conditions; and (iv) market demand
individual firms, but it offers a standard against which shareholders can benchmark firm-level
governance. Strong governance practices in the firm’s industry increase the likelihood of strong
governance at the sample firm and reduce the manager’s ability to resist monitoring. State
antitakeover laws are likely to influence the adoption of antitakeover defenses by individual
firms and may also affect internal governance mechanisms (for instance, firms may implement
stronger internal governance mechanisms to offset reduced corporate control market discipline).
Initial corporate control market conditions are likely to affect the initial governance structure but
not future dividend changes. We measure initial market-wide takeover activity with the fraction
of firms delisted due to M&As in the first year that the firm was listed in CRSP. Investor demand
for governance is also likely to affect firm-level governance, without a direct effect on dividend
4
Instrumental variables estimates are consistent if governance is endogenous but instruments satisfy excludability and relevance
conditions. Ordinary least squares estimates are more efficient if governance is exogenous. We use the auxiliary regression
version of the Durbin-Wu-Hausman test (see Davidson and MacKinnon, 1993; Hausman, 1978) and the difference of Hansen-
Sargan statistics (C statistic) to test the exogeneity of governance measures (probit model tests follow Rivers and Vuong (1988)).
We recognize potential concerns regarding the power of these tests and consider the results suggestive.
10
market-to-book for the strong governance subsample (AGI above median) and the weak
governance subsample (AGI below median) of industry firms.5 In addition, portfolio structure of
the firm’s institutional owners affects investor attention span and focus on monitoring
mechanisms. While investors with concentrated portfolios are more likely to demand good
governance, distracted investors are more likely to exit. We use the average Herfindahl index of
portfolio holdings (excluding the stake in the sample firm) across all institutional investors with
Second, we consider external shocks to governance structure associated with the adoption
adopted business combination, control share acquisition, and fair price laws in the 1980s and
early 1990s (Bertrand and Mullainathan, 2003). In the more hostile takeover market of the
eighties, the passage of antitakeover laws was equivalent to a decrease in governance. The event
represented a shock to the governance of firms incorporated in affected states. If our hypothesis
holds, the addition of antitakeover laws would, all else equal, weaken monitoring and necessitate
listing standards following Enron and related scandals represents another governance shock
(Grinstein and Chhaochharia, 2007). If our hypothesis holds, stronger governance resulting from
these reforms should lead to lower dividends and more significant dividend cuts.
Finally, we use firm fixed effects to control for unobservable firm-level heterogeneity.
We also consider dividend changes as a function of lagged dividend and governance metrics.
5
The construction of the measure is similar in spirit to the Baker and Wurgler (2004) dividend premium.
11
An important part of the manager’s objective function is the strength of ownership incentives.
Fenn and Liang (2001) find that managerial ownership is positively associated with payout in
firms with high agency costs, while stock option compensation is negatively associated with
dividends. We obtain data on CEO ownership and stock options from Execucomp. Dividend
choice is also conditional on the availability of cash flow and investment opportunities, which
reflects the extent of the free cash flow problem. Firm size is expected to have a positive effect
on dividends (Fama and French, 2001). Risky firms are expected to pay lower dividends (Hoberg
and Prabhala, 2009; Jagannathan, Stephens, and Weisbach, 2000). Other controls include
information asymmetry, liquidity, and industry conditions (change in median industry dividend).
4. Results
Univariate evidence
The first set of results presents univariate tabulations of governance and dividend trends in our
sample. Since univariate tests do not account for omitted controls potentially correlated with
dividends, we use these results to illustrate the governance effects and motivate our further
multivariate analysis.
strongest governance index quartiles. Dividend persistence is highest for firms in the bottom
quartile of governance.
governance subsample. Consistent with our predictions, firms with the strongest traditional
governance mechanisms make twice as many dividend cuts as firms with the weakest
12
dividends.
Table 1 presents univariate tests of dividend characteristics for high and low governance
subsamples using quartile and median definitions. Weakly governed firms are two times more
likely to be dividend payers. Conditional on paying dividends, weakly governed firms exhibit
significantly more dividend smoothing, avoid dividend cuts, and are more likely to increase
dividends. (Since dividend cuts are considerably larger in magnitude than dividend increases,
avoidance of dividend cuts drives greater dividend smoothing among weakly governed firms.)
Univariate differences are highly statistically significant and in line with our conceptual
arguments.
[Table 1]
Fama and French (2001) document the aggregate trend of disappearance of dividend
payers, which they attribute to changing sample composition and propensity to pay among past
payers. Although governance data does not extend to the start of their thirty-year sample period,
we can draw some relevant inference from our sample. Figure 3 shows the evolution of
dividends at firms in the bottom versus top governance quartiles. Since we require the
availability of governance and compensation data, our sample firms are larger, have more assets
in place, and are more likely to pay dividends compared to firms in the Fama and French (2001)
sample. The weakest governance subsample saw less dividend disappearance and more
experienced a larger relative decline in dividend incidence between the start and the end of our
sample period (Figure 3). Overall, during our sample period, dividend payers disappeared at
13
firms, it is important to address the drivers of being a dividend payer in the first place. The
governance analysis of the propensity to pay dividends follows John, Knyazeva, and Knyazeva
(2011b). The adoption of a dividend commitment is less likely when firms can fall back on less
[Table 2]
The signs and significance of the controls are generally consistent with the expectations
based on earlier literature (e.g., Fama and French, 2001). For instance, firms with more assets in
place pay higher dividends. Dividends are more likely to be present in firms with a high return
on assets. Similarly, larger firms are more likely to adopt dividends. Risky cash flows reduce the
likelihood of dividends since it is more costly for firms to commit to cash outflows. Presence of
significant managerial equity stakes and CEO stock options decreases the probability that the
firm will be a dividend payer, consistent with agency theory and empirical findings in Fenn and
Liang (2001). Liquidity is negatively associated with dividends, consistent with Banerjee,
Gatchev, and Spindt (2007). Analyst following, included as a proxy for low information
asymmetry between the firm and investors, is negatively related to the incidence of dividends,
partial adjustment dividend model. We regress the difference between current and lagged
dividend level on lagged dividend and controls. The degree of dividend smoothing is computed
as one plus the coefficient on lagged dividend. A higher coefficient on lagged dividend reflects a
higher level of dividend smoothing. To capture the effect of governance on dividend smoothing,
14
with a significant negative coefficient. Consistent with our testable predictions, firms with
[Table 3]
The effects are strongest for board and blockholder monitoring measures. Takeover
defenses are not significant. The corporate control market has become less hostile over time, and
our sample for this set of analyses starts in 1993 due to the availability of firm-level governance
data. Tests of intertemporal dividend behavior are performed in the sample of past dividend
payers, so we use a selection model to evaluate potential selection bias.7 Although the selection
term is significant, interaction terms of interest continue to enter with negative signs (Columns
III and IV). Overall, the findings support prediction (1): weakly governed managers engage in
more dividend smoothing, consistent with upholding the implicit dividend contract.
Table 4 reports the effects of governance on the direction of dividend changes. Consistent
with prediction (2), weakly governed managers are less likely to announce dividend cuts and
omissions. Board and blockholder monitoring and CEO alignment have the most significant
effects. The results in the table also support prediction (3): weakly governed managers are
overall more likely to raise dividends, all else equal. This suggests that a typical dividend
commitment entails not only dividend smoothing, but also a smooth upward-sloping dividend
path. These results provide a more accurate picture of evolution of dividends over time
6
Tests in Columns I and II do not reject the null hypothesis of exogeneity of AGI, CEO ownership, and interaction terms, which
suggests that ordinary least squares estimates are consistent.
7
The selection equation uses the lagged proportion of dividend payers in the firm’s industry, lagged industry median dividend
level, dummy for having a positive dividend in the firm’s initial year in the sample (since 1980), and the industry-level version of
the market dividend premium (based on Baker and Wurgler, 2004), as well as controls from the main equation (with log of
lagged net sales as a proxy for size), sales growth, and index of state laws. The selection model is also partly identified through
the nonlinearity of the functional form.
15
[Table 4]
into account the magnitude of dividend adjustments. Corporate governance has a negative effect
on dividend changes, consistent with more dividend cuts and fewer dividend increases at well
governed firms. In turn, managers at weakly governed firms are less likely to deviate from the
implicit dividend commitment. The most significant effects are due to board structure,
blockholder monitoring, CEO alignment characteristics, and the presence of a single class of
shares. Governance effects continue to hold after correcting for selection (Columns III and IV)
and endogeneity (Columns V and VI).8 An increase in the governance index by one standard
[Table 5]
Panel B (Columns V and VI) examines the magnitude of dividend increases. Since
weakly governed managers face more shareholder pressure to adhere to the level of dividends,
they are more conservative with respect to the magnitude of increases. We find that while weakly
governed managers are more likely to raise dividends overall, they prefer moderate dividend
increases to substantial increases. Small incremental adjustments make the new dividend level
Panel B also reports the effects of governance on dividend changes around cash flow
shocks. As expected, the governance effect is concentrated in subsamples of large and persistent
cash flow increases. Weakly governed managers choose dividend increases both to meet current
8
Tests in Columns I and II reject the exogeneity of suspected governance variables, which suggests that ordinary least squares
estimates are inconsistent and a correction for endogeneity is needed. First-stage statistics - Anderson likelihood ratio test and
Cragg-Donald test – support the relevance and strength of instruments. With the caveat about the power of the Hansen-Sargan
excludability test, instruments also satisfy the excludability condition.
16
sustained in the future without costly cuts. As a result, weakly governed managers raise
dividends in response to large or persistent cash flow increases, which are more likely to draw
Governance shocks
state antitakeover laws and post-Enron governance reforms. Corporate governance changed
around those events for reasons that were arguably exogenous to individual firms’ decisions and
largely unrelated to dividend behavior. To the extent that changes in corporate governance
around the shock had an exogenous component, we are able to sharpen the identification of
governance effects on dividends and mitigate some of the potential concerns about the
The first shock involves the passage of second-generation state antitakeover laws. By
deterring hostile takeover attempts, state antitakeover laws can weaken corporate control market
discipline (see, e.g., Bertrand and Mullainathan, 2003; John, Knyazeva, and Knyazeva, 2011b).
We treat the passage of state antitakeover laws as a governance-decreasing shock. Bertrand and
Mullainathan (2003) provide years of passage of business combination, control share acquisition,
and fair price laws by state.9 Consistent with our hypothesis, the results in Table 6, Panel A show
that the passage of antitakeover laws had a positive effect on dividend changes.
[Table 6]
reforms (the passage of Sarbanes-Oxley and the adoption of board and committee independence
9
The main sample began in 1993 due to the availability of firm-level governance data. In turn, the sample used in this test mainly
spans the eighties. Greater prevalence of hostile takeover attempts during that period likely made antitakeover provisions a more
influential mechanism of external monitoring.
17
governance proxies around the shock helps us identify the governance effect. In line with our
argument and earlier results, the levels and changes of dividends are decreasing in changes in
governance around the shock (Table 6, Panel B). Further, the decrease in dividends after the
reforms was larger for firms that did not have a majority of independent directors on the board
prior to the shock, thus had to raise board governance quality, and for firms that have attained a
While it is hard to rule out endogeneity, taken together these results reduce the likelihood
that reverse causality or omitted variables affecting governance levels are driving our findings.
The analysis presented so far has focused on intertemporal dividend behavior. However,
dividends are only one element of the larger set of financial policies that firms establish with
consideration of agency conflicts (John, Knyazeva, and Knyazeva, 2011b). Our predictions
extend to persistence and changes in total payout. In the first two columns of Table 7, persistence
and changes in total payout to shareholders are decreasing in the strength of governance,
[Table 7]
The next question is whether repurchase changes serve as substitutes for dividend
changes when weakly governed managers implement payout adjustments. Although repurchases
are also a way of returning cash to shareholders, they lack the commitment features of dividends.
If changes in dividends and repurchases are perfect substitutes from the standpoint of remedying
governance problems, governance quality should not affect the choice between dividend changes
and repurchase changes, holding the direction of the payout change constant. We do not find
perfect substitutability. Weakly governed managers prefer to cut payout through repurchases.
18
Debt can constrain managers through regular interest payments, in which case our
predictions for dividend changes should also extend to changes in total commitments to external
claimholders (dividends paid to shareholders and interest paid to debtholders). Consistent with
prediction (4), Table 8 shows that changes in total payments to external claimholders are
decreasing in the strength of governance. Thus, the effect of governance on dynamic dividend
behavior extends to adjustment in overall commitments. Unlike repurchases, debt payments are
associated with a significant cost of deviation, which increases the likelihood that debt
commitments and dividend commitments are partly substitutable in the intertemporal context.
We examine this issue in the first two columns of the table by regressing dividend changes on
debt changes and controls (without assumption of causality). The coefficient on debt is negative
and significant, consistent with partial substitution between dividend and debt changes.
[Table 8]
We perform a number of sensitivity tests to evaluate the robustness of our findings. Table 9
explores interactions of the governance effect on intertemporal dividend behavior with various
[Table 9]
influence the relation between dividend changes and cash flow changes. The relation between
cash flow changes and dividend changes is strongest when corporate governance is weak
(Column I). In a related result, the effect of corporate governance on dividend changes is more
pronounced for firms with limited growth or investment opportunities, which are likely to have
19
changes respond to shareholder demand. There is more pressure to increase (maintain) dividends
after periods of poor performance (Column IV-V) and in industries where investors value
dividends more (Column VI). Similarly, governance has a larger effect on dividend changes in
industries dependent on external financing (Column III), consistent with the Myers (2000)
capital market rationale for sustaining dividends over time. Managers in such industries are more
likely to have to raise external financing in the future. Therefore, they face a greater need to meet
The first three columns of Panel A of Table 10 include firm effects to mitigate
endogeneity concerns by controlling for unobserved firm-level heterogeneity. The results are
[Table 10]
Additional causality checks predict dividend changes using lags of governance and
changes in governance instead of governance levels. Past governance and changes in governance
have a negative effect on dividend changes (even after controlling for past dividend changes).
Simultaneous equations regressions examine the effect of dividends on governance as well as the
effect of governance on dividends. Governance has a negative effect on dividend changes, but
possible nonlinearities by looking at the effect of AGI above and below the median on dividend
changes. Both effects are negative, significant, and similar in magnitude, which does not appear
to indicate nonlinearity.
version of the index that uses rescaled values of continuous characteristics instead of rankings
20
factor analysis of underlying governance variables is used to check the reliability of AGI. The
governance characteristics can be identified. The requirement that eigenvalues exceed one yields
seven factors that cumulatively explain 86.6% of variation in the underlying variables. Based on
factor loadings, governance factors that affect dividend changes are qualitatively similar to
ownership and control in dual class firms, board independence, and CEO characteristics.
Several other factors can affect costs and benefits of dividends, so additional controls are
introduced in Panel C of Table 10. DeAngelo, DeAngelo, and Stulz (2006) support the life-cycle
theory of dividends and find a higher propensity to pay dividends among firms with a higher
fraction of earned equity. The governance result continues to hold after controls for the share of
retained earnings in assets and year of entry dummies are added. Miscellaneous controls include
the share of tangible assets in total assets, stock performance, and bond rating (to proxy for the
cost of debt as an alternative commitment device). For robustness we also use alternative
dependent variable and sample definitions. Tests of changes in dividends per share scaled by
lagged price, changes in cash overall cash dividends scaled by lagged market value, and tests in
the full sample of firms, including firms that did not pay dividends in the previous period, yield
similar results.
21
pay dividends in the previous period, is analyzed instead of the sample of past dividend payers.
It is possible that an omitted variable associated with the proximity of the firm to
financial distress is driving some of the managerial behavior associated with a dividend
commitment. For instance, firms that are at low risk of financial distress can feasibly sustain a
stable dividend policy whereas distressed firms may undertake dividend cuts out of necessity. To
check whether the governance findings are driven by distress, Panel D of Table 10 uses Altman’s
z-score in robustness tests of dividend changes and dividend cuts. First, z-score itself is used as a
control in the regression. Second, the analysis is repeated using an indicator variable for firms
that are less likely to be financially distressed (with z-scores of 3 and higher). As expected,
dividend changes are decreasing in the likelihood of financial distress and dividend cuts are more
frequent among distressed firms. However, the governance results continue to hold and do not
In Panels E and F of Table 10, we extend the sample period to 2009. We make some
modifications to the AGI index due to data availability, however, the index continues to capture
the main dimensions of internal and external corporate governance and CEO characteristics. It
averages firm rankings based on board governance (board independence and inverse of board
size), CEO characteristics and alignment (CEO ownership and inverse of CEO tenure), and
corporate control market oversight (inverse of the Bebchuk et al. (2005) index of takeover
defenses). In Panel E, we control for additional industry variation by using three-digit SIC
industry fixed effects instead of Fama-French industry definitions. In Panel F, we filter out
geographic variation using state-county fixed effects. As John, Knyazeva, and Knyazeva (2011a)
22
location dimension as well as a broad range of differences across remotely located firms that can
affect agency conflicts. (The time-invariant central location indicator is absorbed by county fixed
effects, so it is not reported.) The main results are highly robust to these changes. We note there
is somewhat less dividend persistence in the extended sample. Governance and alignment quality
is negatively related to dividend smoothing and dividend changes. As before, firms with better
governance and managerial alignment have a significantly higher likelihood of dividend cuts and
omissions and a lower likelihood of dividend increases or initiations. The effect of single versus
dual classes of shares is not statistically significant after we account for the main governance and
alignment index.
Table 11 uses an event study to examine the market reaction to dividend cuts and
increases by dividend payers. Event study results reveal that shareholders react differently to
dividend decisions of weakly governed and strongly governed managers.10 The market reaction
to dividend changes is significantly larger for weakly governed managers. Since the dividend
constraint is more important for preventing suboptimal managerial behavior when governance is
weak, dividend changes have more significant effects on shareholder value, all else equal.
[Table 11]
an extension of the Bertrand and Mullainathan (2003) “quiet life” hypothesis. Dividend changes
potentially release new information that the market can use to update beliefs about the manager’s
quality. Dividend cuts attract costly market scrutiny of managerial actions while dividend
increases expand the constraint imposed on the manager, which can precipitate future cuts.
10
Our event study tests of dividend changes by dividend payers complement the evidence in Officer (2011) on dividend
initiations by non-payers.
23
in either direction (cuts as well as increases), and make additional cash distributions through
more flexible repurchases that can be cut easily in the future. Contrary to the quiet life
explanation, our results suggest that weakly governed managers also raise and initiate dividends
more often and distribute additional cash through repurchases less often.
5. Conclusion
considerably across firms, and systematic empirical evidence on the underlying determinants of
dynamic dividend behavior is scarce. This paper analyzed the issue from the perspective of
sustainability of the implicit dividend contract conditional on agency conflicts and provided new
Weakly governed managers engage in more dividend smoothing and make fewer
dividend cuts. Changes in the dividend level are decreasing in the strength of governance.
Besides making fewer cuts, weakly governed managers undertake more dividend increases. As
weakly governed managers balance the need to meet shareholder expectations and set a
sustainable dividend level, they adjust dividends only in response to large cash flow shocks and
Weak governance has a positive effect on the persistence and changes in total payout.
However, dividend and repurchase changes are not perfect substitutes: weakly governed
managers cut payouts through decreases in repurchases and distribute additional cash through
higher dividends. Payout adjustments appear to preserve (and when necessary, expand) the
pre-specified timeline and levels of cash payouts. Changes in total commitments to shareholders
24
act as partial substitutes in the dynamic context. The paper has also examined changes in
Some issues are open for future research. While this paper has focused on dividend
smoothing, an extensive accounting literature has examined earnings smoothing and Leuz,
Nanda, and Wysocki (2003) have linked earnings management to weak legal environments. It
would be of interest to examine the joint determination of dividend and earnings smoothing in
settings with varying corporate governance quality. We have considered the intertemporal
relation of dividends, repurchases and debt changes. Future work could analyze the relation
between dividend changes and other value-relevant corporate decisions, such as investment.
25
Agrawal, A., Knoeber, C., 1996. Firm performance and mechanisms to control agency problems
between managers and shareholders. Journal of Financial and Quantitative Analysis
31(3), 377–397.
Aharony, J., Swary, I., 1980. Quarterly dividend and earnings announcements and stockholders'
returns: An empirical analysis. Journal of Finance 35(1), 1–12.
Amihud, Y., Li, K., 2006. The declining information content of dividend announcements and the
effect of institutional holdings. Journal of Financial and Quantitative Analysis 41(3),
637–660.
Asquith, P., Mullins, D. W., 1983. The impact of initiating dividend payments on shareholders’
wealth. Journal of Business 56(1), 77–96.
Baker, M., Wurgler, J., 2004. Appearing and disappearing dividends: The link to catering
incentives. Journal of Financial Economics 73(2), 271–288.
Banerjee, S., Gatchev, V., Spindt, P., 2007. Stock market liquidity and firm dividend policy.
Journal of Financial and Quantitative Analysis 42, 369–398.
Bebchuk, L., Cohen, A., 2005. The costs of entrenched boards. Journal of Financial Economics
78, 409–433.
Bebchuk, L., Cohen, A., Ferrell, A., 2005. What matters in corporate governance? Review of
Financial Studies 22(2), 783–827.
Benartzi, S., Michaely, R., Thaler, R., 1997. Do changes in dividends signal the future or the
past? Journal of Finance 52, 1007–1034.
Bertrand, M., Mullainathan, S., 2003. Enjoying the quiet life? Corporate governance and
managerial preferences. Journal of Political Economy 111(5), 1043–1075.
Booth, L., Xu, Z., 2008. Who smoothes dividends? Working paper.
Brick, I., Palia, D., Wang, C.-J., 2005. Simultaneous estimation of CEO compensation, leverage,
and board characteristics on firm value.
Chidambaran, N., Palia, D., Zheng, Y., 2006. Does better corporate governance “cause” better
firm performance? Working paper.
Core, J. E., Guay, W. R., Rusticus, T. O., 2006. Does weak governance cause weak stock
returns? An examination of firm operating performance and investors’ expectations.
Journal of Finance 61(2), 655–687.
Cragg, J., Donald, S., 1993. Testing identifiability and specification in instrumental variable
models. Econometric Theory 9, 222–240.
Cremers, M., Nair, C., 2005. Governance mechanisms and equity prices. Journal of
Finance 60 (6), 2859–2894.
Davidson, R., MacKinnon, J. G., 1993. Estimation and inference in econometrics. Oxford
University Press.
DeAngelo, H., DeAngelo, L., Stulz, R., 2006. Dividend policy and the earned/contributed capital
mix: a test of the life-cycle theory. Journal of Financial Economics 81, 227–254.
Dewenter, K., Warther, V., 1998. Dividends, asymmetric information, and agency conflicts:
Evidence from a comparison of the dividend policies of Japanese and U.S. firms. Journal
of Finance 53, 879–904.
26
27
28
Sample
The main sample is based on Compustat Annual (1993–2004), excluding foreign firms, firms with total assets less than 20 mln,
financials (SIC 6000–6999), regulated utilities (SIC 4949–4999), and firms with missing Compustat, governance or
compensation data. The extended sample in Panels E and F of Table 10 uses the 1996–2009 sample period.
The extended sample uses Alignment and Governance Index 2 (AGI2) and Single Class of Shares for 1996-2009. Due to data
changes in RiskMetrics that affect the availability of the G index and the constraints on our availability of blockholder data, we
use a modified AGI definition, AGI2. AGI2 is constructed as the average of firm rankings, rescaled to [0,1], based on Board
Independence (fraction of independent directors on the board), Board Size (inverse), the Bebchuk et al. (2005) index of takeover
defenses (inverse), CEO tenure (inverse), and CEO ownership. Data is available from RiskMetrics. Gap years for takeover
defense data are filled in using adjacent years.
Instruments
Instruments for AGI and CEO Ownership:
State Laws – the index that assigns 1 for the presence of each of the following state antitakeover laws: business combination law;
control share acquisition law; cash out law; fair price law; director’s duties law; antigreenmail (recapture of profits) law; firms
are ranked based on the index in a given year; the variable is rescaled to [0,1] such that higher values reflect the presence of
fewer antitakeover laws.
11
Where the name of a director’s primary employer was available, the accuracy of the CEO flag was checked manually.
29
30
31
32
33
Decrease Decrease
14% 7%
Increase
44%
Increase
No
68%
Change
25%
No
change
42%
34
35
***
D_Div (all firms) 0.77 0.36
***
ρ(Div,LagDiv) 0.98 0.91
***
D_DivDecr 0.07 0.14
***
D_DivIncr 0.67 0.43
***
∆Div 0.09 0.01
***
D_Div (all firms) 0.74 0.45
***
ρ(Div,LagDiv) 0.97 0.93
***
D_DivDecr 0.07 0.12
***
D_DivIncr 0.64 0.48
***
∆Div 0.08 0.02
36
Probit estimation of the probability of being a dividend payer. Sample and variable definitions are presented in Appendix A.
Regressions include Fama-French industry dummies, year dummies, and the proportion of dividend payers in the firm’s industry
in the previous period. Robust z-statistics with clustering by firm are in parentheses. Significance at 1%, 5%, and 10% levels is
denoted with ***, **, and *, respectively.
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
[-1,+1] (incl.
[-1,+1] [-1,+1] [-2,+2]
omissions)
I II III IV
** *** *** ***
∆Div 0.514 0.758 0.851 0.922
2.55 2.98 3.17 3.26
*** *** *** *
D_DivDecr -0.013 -0.012 -0.020 -0.011
-2.86 -2.79 -2.72 -1.83
∆Div x D_DivDecr -0.354 -0.420 -0.568 -0.537
-0.92 -1.09 -1.36 -1.20
** ** ** **
∆Div x AGI -0.882 -0.891 -0.843 -0.835
-2.38 -2.43 -2.18 -2.09
∆Div x D_DivDecr x AGI 0.803 0.829 1.096 1.039
1.15 1.19 1.47 1.27
∆Div x Cash Flow -0.557 -0.405 -0.762
-1.24 -0.88 -1.36
∆Div x Market-to-Book 0.019 -0.012 -0.007
0.52 -0.36 -0.18
∆Div x CEO Ownership -0.004 -0.002 -0.005
-0.66 -0.36 -0.99
*
∆Div x CEO Stock Options -0.032 -0.055 -0.026
-1.75 -1.63 -1.05
* *
∆Div x Analyst Following -0.054 -0.083 -0.094
-1.15 -1.76 -1.75
Obs. 3385 3334 3413 3334
2
R 0.12 0.12 0.14 0.11
Adj. R2 0.10 0.10 0.13 0.10
54