Dividend policy and earnings management across countries

Wen He, Lilian Ng, Nataliya Zaiats, Bohui Zhang

PII: S0929-1199(16)30318-2
DOI: doi:10.1016/j.jcorpfin.2016.11.014
Reference: CORFIN 1127

To appear in: Journal of Corporate Finance

Received date: 7 November 2015

Revised date: 22 November 2016
Accepted date: 23 November 2016

Please cite this article as: He, Wen, Ng, Lilian, Zaiats, Nataliya, Zhang, Bohui, Dividend
policy and earnings management across countries, Journal of Corporate Finance (2016),
doi:10.1016/j.jcorpfin.2016.11.014

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 ACCEPTED MANUSCRIPT

P T
RI
Dividend Policy and Earnings Management Across

SC
Countries

NU
Wen He, Lilian Ng, Nataliya Zaiats ∗ and Bohui Zhang† ‡
MA
ED

Current Version: November 2016

PT
CE
AC

∗
Corresponding author.
†
He is from UQ Business School, University of Queensland, Australia; Ng is from Schulich School of Busi-
ness, York University, Canada; Zaiats is from Sawyer Business School, Suffolk University, USA; and Zhang is
from UNSW Business School, University of New South Wales, Australia. Authors’ contact information: He,
wen.he@uq.edu.au, +61 7 33668048; Ng, lng@schulich.yorku.ca, (416) 736-2100; Zaiats, nzaiats@suffolk.edu,
(617) 573-8340; and Zhang, bohui.zhang@unsw.edu.au, +61 2 9385-5834.
‡
Acknowledgments: We thank the editor Jeffry Netter, the anonymous referee, Michael Brennan,
Philip Dybvig, Ting Li, Stephen Taylor, Feifei Zhu, William O’Brien and participants in the 2012 Accounting
and Finance Association of Australia and New Zealand’s Conference, Institute of Financial Studies at the
Southwest University of Finance and Economics, 2012 China International Conference in Finance, 2013
Eastern Finance Association Conference, 2013 Financial Management Association Conference, and 2015
Financial Management Association’s European Conference, for many helpful comments and suggestions.
 ACCEPTED MANUSCRIPT

Dividend Policy and Earnings Management Across Countries

ABSTRACT

T
This paper examines whether dividend policy is associated with earnings management and

P
whether the relationship varies across countries with wide-ranging degrees of institutional strength

RI
and transparency. Based on a sample of 23,429 corporations from 29 countries, we show that

SC
dividend payers manage earnings less than dividend non-payers, and that this evidence is stronger

in countries with weak investor protection and high opacity. Further, we find that dividend payers

NU
manage earnings less when they issue equity following dividend payments, and that this result is

more pronounced in countries with weak institutions and low transparency. Overall, our evidence
MA
suggests that firms may employ dividend policies associated with less earnings manipulation to

mitigate agency concerns and to establish credible reputation, thereby facilitating access to external

funds.
ED
PT

Keywords: Dividends; Earnings management; Agency problems; Investor protection; Equity is-

suance
CE

JEL Classification Number: G15; G34; G35; G38; M41

AC
 ACCEPTED MANUSCRIPT

1. Introduction

Paying dividends limits private control benefits available to insiders as cash paid out provides fewer

opportunities to consume these benefits (Pinkowitz et al., 2006). Prior studies, however, show that

T
paying dividends does not prevent firms from committing accounting fraud, 1 thereby suggesting

P
that dividend paying firms may not necessarily act in line with shareholder interests. The goal

RI
of this study is to examine whether there is any association between dividend paying status and

SC
earnings manipulation across firms from a broad spectrum of markets with wide-ranging degrees

of investor protection and transparency. Such a pursuit is particularly relevant as extant studies

NU
report that non-U.S. firms may experience less pressure to maintain their dividend policies than

their U.S. counterparts.2

MA
Prior literature finds that earnings management is less prevalent when private control benefits

are limited (e.g., Leuz et al., 2003; Gopalan and Jayaraman, 2012). Thus, firm managers who

decide to pay dividends may have fewer private control benefits to consume and conceal and hence,
ED

could be less likely to fabricate accounting information. Furthermore, both theory and empirical

evidence indicate that the U.S. firms are unwilling to cut dividends, and that they engage in
PT

dividend smoothing to maintain a constant stream of dividends (e.g., Lintner, 1956; Brav et al.,
CE

2005; Skinner and Soltes, 2011). However, non-U.S. firms appear to be less concerned about

dividend cuts and thus alter their dividend policies more frequently (e.g., Dewenter and Warther,
AC

1998; Chemmanur et al., 2010; Hail, Tahoun, and Wang, 2013). Such evidence implies that the

relationship between dividend paying status and earnings management may not persist or may be

weaker in an international setting. Alternatively, it is plausible that the link between dividend

policy and earnings management is stronger in a cross-country examination. This may be due

to a large variation in country-level investor protection measures (e.g., Jensen, 1986; La Porta

1
For example, in their sample of 330 fraud-accused firms spanning between 1982 and 2005, Caskey and Hanlon
(2013) find that 72 alleged fraud firms paid out $20 billion in dividends while simultaneously committing accounting
fraud.
2
See, for example, Brav et al. (2005); Dewenter and Warther (1998); Chemmanur et al. (2010); Hail, Tahoun,
and Wang (2013); Lintner (1956); Skinner and Soltes (2011).

1
 ACCEPTED MANUSCRIPT

et al., 2000). Specifically, such differences may result in a potentially stronger need of foreign

firms to convey their governance quality via alternative credible signals, compared to their U.S.

counterparts.3 We therefore conjecture that dividend paying firms manipulate earnings less than

their non-paying counterparts and that the strength of the relationship may vary with country-level

T
investor protection and transparency.

P
RI
Our analyses focus on the following three closely related issues. First, we examine the relation-

ship between dividend policy and earnings management. We gauge a firm’s earnings management

SC
by the magnitude of its abnormal accruals with a larger magnitude of abnormal accruals indicating

more aggressive earnings management. We test the relationship on a large cross-section of 23,429

NU
firms, whose dividend policies vary substantially across 29 countries, over a 21-year period from

1990 to 2010. Second, our analyses exploit the richness of our data by examining whether and how
MA
the relationship between dividend policy and earnings management differs across countries with

varying degrees of investor protection and transparency. Third, our study attempts to provide fur-
ED

ther insights regarding the observed variation of the dividends-earnings management relationship

with investor protection. We do so by examining the established link in the context of subsequent
PT

access to external financing.

We find that dividend payers have smaller abnormal accruals than dividend non-payers, sug-
CE

gesting that the former are less likely to engage in aggressive accruals management to conceal

firm performance. These results are economically significant, even after controlling for firm-specific
AC

variables that have previously been shown to affect earnings management. For example, our model

estimates indicate that the size of abnormal accruals for dividend payers is 7% smaller than that

for non-payers. Our findings are also robust to subsamples of U.S. and non-U.S. firms; subsample

periods of 1990s and 2000s; subsamples partitioned by firm size, dividend volatility, and cash flow

volatility; subsamples focusing on dividend changes - initiations and omissions as well as increases

and decreases; subsamples of dividend payers and non-papers with at least 10 or 20 years of data;

3
See Fan and Wong (2005); Choi and Wong (2007); Pinkowitz, Stulz, and Williamson (2006); Francis et al. (2005),
among others.

2
 ACCEPTED MANUSCRIPT

and to the various alternative measures of payouts and of earnings management. We also conduct

several additional robustness tests that focus on the samples matched by selected firm character-

istics, propensity score matching, and fixed effects regressions. All results persist as we employ

these alternative estimation approaches. Overall, we offer strong evidence that paying dividends is

T
associated with and leads to lower earnings management.

P
RI
We next examine whether the relationship between dividend policy and earnings management

differs across countries with varying degrees of institutional strength. Based on a global sample

SC
of firms, La Porta et al. (2000) find evidence that dividend payouts are strongly associated with

minority shareholder rights. 4 Such evidence supports their hypothesis that dividend payments are

NU
“an outcome of an effective system of legal protection of shareholders.” Alternatively, firms could

pay dividends to mitigate adverse selection and moral hazard issues (e.g., Hail et al., 2014). If firms
MA
paying dividends are associated with less earnings manipulation to alleviate country-level agency

issues, we expect the negative relationship between dividend paying status and earnings manage-
ED

ment to be more pronounced in firms from countries that are susceptible to greater agency conflicts.

The results are consistent with the prediction that dividend paying status is associated with lower
PT

earnings management, especially in countries with weak institutions and low transparency.

Lastly, we attempt to further explain why the negative link between dividend policy and earnings
CE

management is more pronounced in countries with weak investor protection and high opacity. One

plausible explanation is that dividend policy associated with low earnings manipulation conveys the
AC

management’s intention to forgo private control benefits and to build reputation that will facilitate

future access to capital markets (La Porta et al., 2000). This is particularly important for firms

that want to maintain their ability to raise funds in capital markets, particularly in less financially

developed ones. Our evidence supports the above conjecture. The results indicate that the negative

link between dividend policy and earnings management is more pronounced in firms that implement

equity issuances in one to three years subsequent to the time of dividend payments, notably in those

4
Bartram et al. (2012) extend La Porta et al.’s sample and reach the same conclusion.

3
 ACCEPTED MANUSCRIPT

from countries with lower legal protection and higher opacity.

Our study contributes to the finance and accounting literature in several directions. First, to

the best of our knowledge, our research is the first to offer robust international evidence on the

link between dividend policy and earnings management. Extant literature establishes that firms

T
from countries with weak institutions have lower dividend payouts (La Porta et al., 2000) and that

P
RI
earnings management is more prevalent in developing than in developed economies (e.g., Leuz et

al., 2003; Gopalan and Jayaraman, 2012; Lang et al., 2006). We contribute to this literature by

SC
providing evidence that dividend payers exhibit lower levels of earnings management across both

developed and emerging markets, especially so in the latter economies.

NU
Second, our paper offers significant insights into the role of dividends-earnings management

link in the context of agency concerns. La Porta et al. (2000) suggest that effective legal systems
MA
empower minority shareholders to force insiders to disgorge cash, but provide no explanation for

why a large proportion of firms, even in countries with low investor protection, still pay dividends.
ED

Our results offer a plausible explanation for why firms from countries with weak investor protection

and low transparency pay dividends that are strongly associated with less earnings manipulation.
PT

These firms may commit to dividend payments thereby consuming and concealing fewer private

control benefits in line with less aggressive earnings management. They may do so to demonstrate
CE

commitment to shareholder interests and establish reputation in the marketplace.

Finally, our work contributes to the growing literature on the effect of institutional factors
AC

on firm behavior in international markets. Leuz et al. (2003) find that weak investor protection

allows insiders to consume more private control benefits and to manage earnings more aggressively.

La Porta et al. (2000) and Faccio et al. (2001) report that firms pay less dividends if country

governance does not effectively constrain insiders from expropriating minority shareholders. Our

study adds to this literature by showing that firms operating in countries with weak legal regimes

might optimally choose to employ dividends to underscore their intention to forgo private control

benefits, via the negative link to earnings manipulation, and do so to facilitate access to external

4
 ACCEPTED MANUSCRIPT

funds.

The rest of the paper is organized as follows. Section 2 discusses the motivation of our study

and develops a number of testable hypotheses. Section 3 describes the data and explains the

construction of various earnings management measures and control variables. Section 4 tests the

T
relationship between dividend policy and earnings management. Section 5 examines whether the

P
RI
dividends-earnings management relationship varies across countries with different degrees of in-

vestor protection and transparency. This section also explores the relationship between dividend

SC
paying status and earnings management in the context of subsequent equity issuance. The final

section concludes.

2. Motivation and hypotheses development NU

MA
2.1. Dividend policy and earnings management

Information asymmetries between corporate insiders and outside investors give rise to agency con-
ED

cerns (Jensen and Meckling, 1976). Existing studies recognize that managers have incentives to

keep excess cash, because this affords them the opportunity to misappropriate or waste corporate
PT

resources for personal benefits at the expense of outside investors. 5 If profits are not paid out
CE

to shareholders, they may be diverted by insiders for personal use or directed into unprofitable

investments (e.g., La Porta et al., 2000; DeAngelo, DeAngelo, and Stulz, 2006; Denis and Osobov,
AC

2008). Dividend non-paying firms may thus be prone to exploit corporate resources for personal

interests.

In contrast, theory posits that in imperfect markets, firm managers use dividends as a means

to resolve agency-based conflicts between insiders and the outside shareholders (Jensen, 1986).

Dividends reduce the agency costs of free cash flow and minimize suboptimal managerial behavior

(Easterbrook, 1984). Extant agency theories postulate that dividends may be important in an

agency context in the following two manners. First, dividend policy may be the outcome of an
5
See, for example, Jensen (1986), Stulz (1990), and La Porta et al. (2000).

5
 ACCEPTED MANUSCRIPT

efficient contracting under which minority shareholders force the company insiders to disgorge

cash, thereby limiting any potentially suboptimal managerial behavior (e.g., La Porta et al., 2000;

Shleifer and Wolfenzon, 2002; Pinkowitz, Stulz, and Williamson, 2006). Second, dividends may

be employed to convey a firm’s commitment to act in the best interests of outside investors and

T
thus attenuate the agency concerns (e.g., Rozeff, 1982; Easterbrook, 1984; Jensen, 1986; Lang and

P
Litzenberger, 1989; Allen, Bernardo, and Welch, 2000; La Porta et al., 2000; Myers, 2000). Thus,

RI
firms paying dividends limit private control benefits available to insiders, as cash paid out provides

SC
fewer opportunities for insiders to consume these benefits (Pinkowitz et al., 2006).

Notably, a strand of literature also documents that consumption of private control benefits by

NU
the company insiders is positively associated with earnings manipulation. Specifically, Leuz et al.

(2003) argue that misrepresentation of accounting information could arise from the incentives of
MA
insiders and controlling shareholders to camouflage their private control benefits and document

evidence that earnings management is less pervasive when these benefits are limited. 6 Gopalan and
ED

Jayaraman (2012) compare earnings management practices of insider and non-insider controlled

firms and conclude that earnings management is more prevalent in the former category of firms as
PT

consumption of private control benefits is higher in these firms. We therefore expect that dividend

paying firms may choose to limit their private control benefits consumption and may subsequently
CE

have fewer incentives to conceal these benefits. Dividend payers may thus be less inclined to distort

their true economic performance via earnings manipulation.

AC

The discussions above give rise to our first hypothesis:

H1: Dividend policy is negatively associated with earnings management.

6
Specifically, the authors show that insiders can use their discretion in financial reporting to overstate reported
earnings to cover up a loss that might prompt outsiders to question and intervene. Alternatively, if earnings are
unexpectedly high, insiders can divert more assets to themselves and report lower earnings by overstating expenses.

6
 ACCEPTED MANUSCRIPT

2.2. Dividends-earnings management link and institutional strength across countries

Prior literature suggests that non-U.S. firms alter their dividend policies more often than do their

U.S. counterparts. For example, Dewenter and Warther (1998) show that Japanese firms are more

T
inclined to omit or decrease dividends and that their dividends are more responsive to reported

P
earnings. Chemmanur et al. (2010) compare dividend policies of U.S. versus Hong Kong firms and

RI
find that the latter employ more flexible dividend policies and engage less in dividend smoothing.

SC
Hail, Tahoun, and Wang (2014) report that in a sample of 38 countries from 1993 to 2008, the U.S.

firms exhibit the lowest incidence of dividend decreases at 9.4%, relative to the sample average of

NU
16.5%. It is therefore plausible that the relationship between dividend paying status and earnings

management may not persist or may be weaker for non-U.S. firms.

MA
Alternatively, it is plausible that the link between dividend policy and earnings management

varies with country-level investor protection and transparency measures. Specifically, a growing

strand of literature emphasizes that firms operating in countries with weak legal institutions at-
ED

tempt to employ alternative mechanisms to convey their commitment to shareholder interests and

maintain credibility in the financial markets. For example, based on a sample of firms from eight
PT

East Asian economies, Fan and Wong (2005) demonstrate that firms with agency problems inherent

in their ownership structures are more likely to hire a BigN auditor, and that this choice bears
CE

positive value implications. The authors further conclude that a BigN auditor undertakes a cor-

porate governance role in emerging markets. Choi and Wong (2007) show that the appointment
AC

of a high quality auditor plays a more important governance role in firms from countries with

weak legal institutions than in those from countries with strong legal protection. Pinkowitz, Stulz,

and Williamson (2006) find that the relationship between dividends and firm value is stronger in

countries with weak investor protection.

We thus predict that foreign firms may employ dividend policies with an intent to consume and

conceal fewer private control benefits as demonstrated by a stronger negative link between dividend

policy and earnings manipulation. Notably, Lang, Raedy, and Wilson (2006) and Gopalan and

7
 ACCEPTED MANUSCRIPT

Jayaraman (2012) find that firms from countries with weak investor protection manage earnings

more aggressively. In light of such evidence, it is interesting to evaluate whether dividends-earnings

management relationship is amplified in environments that are more prone to manifest agency

problems.

T
Combined, the existing findings documented above give rise to our second hypothesis:

P
RI
H2: The dividends-earnings management relationship is more pronounced in firms from coun-
tries with poor legal protection or low transparency.

SC
To further understand why the link between dividend policy and earnings management is more

NU
pronounced in firms from countries with weak institutions and high opacity, we focus on a firm’s

access to capital markets for external funds. Specifically, we contend that if firms indeed employ
MA
dividend policies with a strong intent to consume fewer private control benefits as reflected in a

decreased inclination to manipulate earnings, these firms may do so to establish reputation as they

must be attempting to facilitate external financing (e.g., La Porta et al., 2000).

ED

The increasing body of literature suggests that firms may employ alternative governance mech-

anisms to mitigate their agency concerns in order to subsequently gain access to external funds.
PT

For instance, Chang, Dasgupta, and Hilary (2009) find that firms hiring a BigN auditor are more
CE

likely to issue equity and to have larger equity issues. The authors further show that the difference

between the firms employing BigN auditors versus those employing non-BigN auditors narrows
AC

when market conditions improve. In their 2006 study, the three authors show that firms with

greater analyst coverage are more likely to issue equity and also do so more frequently than their

counterparts with fewer analysts following. Fan and Wong (2005) demonstrate that firms with

agency problems pertinent to their ownership structures are more likely to employ a BigN auditor,

and that the relationship is more pronounced in firms that raise equity frequently. All these find-

ings suggest that the role of alternative mechanisms to convey commitment to shareholder interests

may strengthen in firms interested in accessing the external market, especially in those with fewer

country-level means to do so.

8
 ACCEPTED MANUSCRIPT

The evidence in the preceding discussion leads to our third hypothesis:

H3: The dividends-earnings management relationship is more pronounced in firms that subse-
quently issue equity, especially in those from countries with poor legal protection or low
transparency.

P T
3. Data and variable construction

RI
3.1. Sample construction

SC
This subsection briefly describes the construction of our sample and key variables. We obtain stock

NU
returns and financial data from Datastream and WorldScope. Our sample period spans from 1990

to 2010. To be included in the sample, we require that (i) a firm must have non-missing accounting
MA
data for calculating earnings management measures and control variables; and (ii) a country must

have at least 50 firms with non-missing data in a given year. These requirements result in 174,340

firm-year observations for 23,429 non-financial firms across 29 countries. Specifically, our sample
ED

comprises 18 developed (Australia, Belgium, Canada, Denmark, Finland, France, Germany, Hong

Kong, Italy, Japan, the Netherlands, Norway, Singapore, Spain, Sweden, Switzerland, the U.K.,
PT

and the U.S.) and 11 emerging markets (China, India, Indonesia, Malaysia, the Philippines, Poland,

South Africa, South Korea, Taiwan, Thailand, and Turkey).

CE

To capture dividend policy, our analysis employs an indicator variable that takes the value of

one if the firm pays dividends, and zero if otherwise. This indicator allows us to exploit the distinct
AC

role of dividends between dividend payers and non-payers. We, however, verify the robustness of

our findings to alternative continuous measures, such as dividend payout ratios and dividend-price

ratios, and find consistent results.

Previous studies have shown that insiders and managers use accounting accruals opportunisti-

cally to distort the true economic performance and mislead outside investors in various occasions,

including equity issuances (Teoh, Welch, and Wong, 1998a, 1998b), bond issuances (Caton et al.,

2011), and stock option exercises (Bartov and Mohanram, 2004). As there exists no uniformly supe-

9
 ACCEPTED MANUSCRIPT

rior measure of earnings management, our analysis employs three key metrics that have been widely

studied in the accounting and finance literature, the construction of which is described in detail in

Appendix A. These measures are - (i) the absolute value of residuals from Jones’ (1991) abnormal

accruals model, Accr(J); (ii) the standard deviation of residuals from Dechow and Dichev’s (2002)

T
accruals model, Accr(D); and (iii) the standard deviation of residuals from Francis et al.’s (2005)

P
accruals model, Accr(F ). While each measure has its own strengths and weaknesses (Dechow, Ge,

RI
and Schrand, 2011), using three different measures will provide more robustness to our findings.

SC
In addition, in selected tests, we supplement these measures by exploiting subsamples of positive

versus negative Jones’ (1991) accruals, as well as employing probability of small profits and prob-

NU
ability of small earnings per share increases, as alternative metrics of earnings management. We

interpret that larger Accr(J), as well as larger variation of Accr(D) and Accr(F ), indicate more
MA
aggressive earnings management.

3.2. Univariate statistics

ED

This subsection presents some basic statistics pertinent to our key variables. Table 1 shows the
PT

cross-country distribution for the starting year of the sample period, number of unique firms,

number of firm-year observations, average proportion of firms that pay dividends, 7 average dividend-
CE

price ratio, and average of each earnings management measure. The starting year reflects the data

availability for computing earnings management measures for each country. Many of the developed
AC

countries have data available starting from 1990, whereas among the emerging markets, Malaysia

has the longest data available from 1993. Correspondingly, the numbers of unique firms and firm-

year observations are larger for the former than for the latter. The number of unique firms ranges

from 91 in Turkey with 471 firm-year observations to 7,989 unique firms in the U.S. with 60,969

firm-year observations. Japan, the U.K., and the U.S. contribute about 63% of the total number

of observations, with the U.S. alone accounting for 35%. In our robustness tests, we exclude the

U.S. or all these three countries from our analysis, and the results remain unaltered.
7
The average proportion of firms that pay dividends is computed based on firm-year observations in each country.

10
 ACCEPTED MANUSCRIPT

The proportion of dividend payers is between 36.5% (U.S.) and 88.6% (Japan), with a cross-

country average of 59.6%. The small proportion of dividend payers in the U.S. may reflect the

findings of Grullon and Michaely (2002) and Skinner (2008), who report the increasing use of stock

repurchases by U.S. corporations as an alternative payout mechanism, especially in recent years. In

T
other words, their results imply that U.S. firms are distributing less dividends to their shareholders.

P
RI
To determine the time trend in dividend payers, we also report the proportions of dividend

payers for two almost equal subperiods: 1990-1999 and 2000-2010. We find evidence of a significant

SC
16.1% decline in the proportion of the U.S. firms that pay dividends, from 46.1% in the first

subperiod to 30.0% in the second. The overall drop in the fraction of dividend payers across the

NU
two subperiods is 12.1%; the drop is more pronounced among developed markets (14.4%), compared

with only 7.7% fall in emerging markets. Similarly, the average dividend-price ratio for developed
MA
markets is 1.5% while it is 2.4% for emerging economies, with a cross-country average of 1.6%,

which ranges from 0.8% (U.S.) to 4.1% (Thailand). Interestingly, such results are consistent with
ED

those of Fama and French (2001), who focus on U.S. markets and subsequent studies such as Fatemi

and Bildik (2012) who examine the world markets. Fama and French find that the percentage of
PT

dividend payers declines over time due to the increasing number of new listings of small firms

with low profitability and strong growth opportunities. Fatemi and Bildik provide corroborating
CE

evidence from a sample of more than 17,000 companies from 33 different countries; they show a

significant worldwide decline in the propensity to pay dividends due to the payout policies of smaller
AC

and less profitable firms with relatively more investment opportunities.

The cross-country averages of earnings management measures are 0.057, 0.045, and 0.042 for

Accr(J), Accr(D), and Accr(F ), respectively. While Japan displays the lowest level of earnings

management across all three metrics, Poland exhibits the highest based on Accr(J) and Australia

also shows the highest based on Accr(D) and Accr(F ). While Accr(J) average is lower for developed

countries at 0.055 compared to 0.064 for emerging markets, Accr(D) and Accr(F ) averages are

somewhat higher for developed markets at 0.046 and 0.042 than for emerging markets at 0.043 and

11
 ACCEPTED MANUSCRIPT

0.038, respectively.

Table 2 presents the mean, median, and standard deviation of each earnings management mea-

sure for dividend payers and non-payers. It also reports p-values associated with the t-test for mean

difference and the Wilcoxon test for median difference between the two groups of firms. Drawn

T
from the extant literature, our subsequent multivariate analyses also control for variables that have

P
RI
shown to affect earnings management: firm size (M V ); total assets (T A); free cash flow (F CF );

book-to-market equity ratio (BM ); leverage (DE); sales growth (Salesg ); firm age (Age); long-

SC
term tangible assets (P P E); profitability (ROA); an IAS dummy (IAS); a BigN dummy (BigN );

closely-held ownership (Chold); and an ADR dummy (ADR). Means and medians of these control

NU
variables, as well as their differences between dividend payers and non-payers, are depicted in Table

2, with the details of their construction in Appendix B.

MA
Based on the t− and W ilcoxon tests, all the variables are statistically and significantly different

across dividend payers and non-payers at the 1% level, except for the mean difference in leverage
ED

at the 5% level. The univariate statistics show that all earnings management metrics are higher for

dividend non-payers than for dividend payers, suggesting that the former engage in more aggressive
PT

earnings manipulation than the latter. The differences in the magnitude of these metrics provide

preliminary univariate evidence that dividend policy relates negatively to earnings management in
CE

an international setting.

Dividend payers also tend to be larger, older, more profitable, exhibit lower sales growth, larger
AC

free cash flow, have higher long-term tangible assets, book-to-market equity ratios, and larger

closely-held ownership, compared with their dividend non-paying counterparts. Furthermore, there

are larger proportions of IAS adopters and of ADR issuers, while a smaller proportion of BigN

adopters, among dividend payers than non-payers. These observations are broadly in line with

evidence in the existing literature pertinent to dividend determinants. 8

8
For instance, Fama and French (2001); Grullon and Michaely (2002); La Porta et al. (2000); Denis and Osobov
(2008), among others.

12
 ACCEPTED MANUSCRIPT

Table 3 reports Pearson cross-correlation coefficients of the variables employed in our study. All

earnings management measures - Accr(J), Accr(D), and Accr(F ), are significantly correlated with

Div, a dummy variable that takes a value of 1 if a firm pays dividends in year t, and 0 if otherwise.

The correlation coefficients of earnings management variables with Div are consistent with our

T
expectation that dividend paying firms exhibit lower levels of earnings manipulation. Dividend

P
paying status, Div, is significantly positively correlated with firm size, total assets, free cash flow,

RI
book-to-market equity ratio, leverage, firm age, long-term assets, profitability, IAS dummy, ADR

SC
issuance, and closely held ownership, while significantly negatively correlated with sales growth

and BigN adopters. In general, the correlation coefficients between the variables in our sample

NU
are moderately low, suggesting no evidence of multicollinearity. One exception is the correlation

coefficient of 0.87 between total assets (T A) and market value (M V ). To capture firm size, we thus
MA
only include M V in regressions. However, the unreported regression results, which employ both

T A and M V as control characteristics, remain unchanged.

ED

4. Dividend policy and earnings management

4.1. Assessing the relationship

PT

In this subsection, we examine the relationship between dividend policy and earnings management
CE

in a multivariate setting. Specifically, we conduct cross-sectional analyses by regressing a proxy

for earnings management, Accr(J), on a dividend indicator, Div, while controlling for firm-specific
AC

variables, Xi (i = 1, ..., N ), that have previously been found to affect earnings management, inclu-

sive of prior year’s Accr(J), as well as for country, industry, and year fixed effects,

N
Accr(J)t+1 = a + bDivt + cit Xit + Fixed Effects + ǫt+1 . (1)
i

Equation (1) forms our baseline regression throughout this study, where the key explanatory vari-

able Div is a dummy variable that takes the value of 1 if the firm pays dividends in year t, and 0 if

otherwise. We employ Jones’ (1991) absolute value of residuals (Accr(J)) as a proxy for earnings

management in (1), computed in year t + 1. Accr(J) gauges the extent to which a firm manages

13
 ACCEPTED MANUSCRIPT

its earnings with the larger accruals metric indicative of more aggressive earnings management.

Therefore, if our hypothesis H1 is true, then the b coefficient should be negative, implying that

dividend payers are less inclined to manipulate their earnings.

Firm-level regression results of Equation (1) and its variants are presented in Table 4. All

T
these regression models incorporate a multitude of firm-level control variables, including the lagged

P
RI
Accr(J); firm size (M V ); free cash flow (F CF ); book-to-market equity ratio (BM ); sales growth

(Salesg ); firm age (Age); leverage (DE); long-term tangible assets (P P E); profitability (ROA); a

SC
dummy variable that takes the value of 1 if the firm prepares its financial statements in accordance

with International Accounting Standards (IAS); a dummy indicator that takes the value of 1 if the

NU
firm employs a BigN auditor (BigN ); a dummy variable that equals 1 if the firm has issued American

Depository Receipts (ADR); and closely-held ownership (Chold).9 Throughout the study, all
MA
multivariate regressions also include country, industry, and year fixed effects (as applicable), and

all associated p-values are computed based on standard errors adjusted for heteroskedasticity and
ED

firm-level clustering. Also, all accruals measures have been multiplied by 100 and deflated by total

assets, and are thus reflected as a percentage of total assets.

PT

Models M1-M3 focus on the entire sample of firms from 29 countries. Model M1 shows the

relationship between earnings management and Div alone, while model M2 presents the relationship
CE

with the lagged Accr(J) alone. Model M3 expands model M1 by adding controls for the various

firm characteristics. We control for the lagged Accr(J), separately in model M2, and together with
AC

Div and the other controls in model M3, to assure that the link between Div and Accr(J) is not

driven by an omitted time-invariant determinant of earnings management. Models M4 and M5

examine the potential differential effects of the relationship between dividend policy and earnings

management for the U.S. versus non-U.S. firms. Models M6 and M7 separate out the possible

differences in the relationship during the 1990s versus the 2000s.

Consistent with our expectations, the results highlight a negative, statistically significant rela-
9
For example, see Becker et al. (1998); Lang, Raedy, and Wilson (2006); Francis and Wong (2008); Barth,
Landsman, and Lang (2008), among others.

14
 ACCEPTED MANUSCRIPT

tionship between dividend policy, as measured by dividend paying status, and earnings management

across all models. For instance, the coefficient of Div in model M4 is -0.487 and highly significant

at the 1% level, reflecting close to 0.5% decrease in abnormal accruals for firms paying dividends.

Based on the Accr(J) mean for dividend non-payers of 0.075, this coefficient is also economically

T
significant as it implies that accruals are about 7% smaller for dividend payers, compared to those

P
of non-payers. It is also noteworthy that the adjusted R-squared increases from 7.2 percent in

RI
model M1, where Div is the only explanatory variable, to 17.3 percent in model M3, which also ac-

SC
counts for all the firm characteristics. The adjusted R-squared of 13.1 percent in model M2, where

Accr(J)L is the only regressor, underscores that the lagged Accr(J) contributes substantially to

NU
such an increase.

While both negative and highly significant, the estimate of Div in model M3 is lower than its
MA
counterpart in M5, suggesting that paying dividends in non-U.S. firms decreases earnings manage-

ment to a larger degree than it does in the U.S. firms. In two unreported tests, we further find that
ED

the difference between the coefficients of Div in these two models, and that the interaction term

of dividend policy and non-U.S. firms when focusing on the full sample of firms, are both strongly
PT

significant. Such findings imply that earnings management is more sensitive to dividend policy in

non-U.S. firms than in the U.S. firms. This result also verifies that our key finding is not driven
CE

by the large proportion of the U.S. firms in the sample. The magnitudes of the coefficients of Div

in models M6 and M7 (-0.621 and -0.427, respectively) suggest that dividend payers exhibit lower
AC

earnings management during the 1990s than during 2000s. Both estimates are strongly significant

at conventional levels, confirming that our key finding is insensitive to the subsample periods.

The results on the control characteristics are broadly consistent with those of other studies. 10

Specifically, larger, older, more profitable firms, firms with higher book-to-market equity ratios,

leverage, and higher long-term tangible assets, and those hiring a BigN auditor are likely to exhibit

lower earnings management. In contrast, firms with higher sales growth, with larger proportion

10
See, for example, Becker et al. (1998); Lang, Raedy, and Wilson (2006); Francis and Wong (2008); Barth,
Landsman, and Lang (2008).

15
 ACCEPTED MANUSCRIPT

of closely-held shares, and those that issued ADRs are more prone to be associated with stronger

earnings manipulation. The signs of estimates of IAS and of F CF are mixed or insignificant.

In summary, we have established a strong negatively significant association between dividend

paying status and earnings management, while controlling for factors that have shown to be related

T
to earnings management as well as for country, year, and industry fixed effects. The negative

P
RI
association is also robust across the various subsamples. Such a relationship is consistent with our

conjecture that firms may employ dividend policies as they have a desire to consume and conceal

SC
fewer private control benefits, thereby exhibiting fewer incentives to manipulate their earnings. The

results therefore render support for the dividends-earnings management relationship across a wide

NU
range of firms from 29 developed and emerging markets that differ vastly in the degrees of strength

of legal institutions and investor protection.

MA

4.2. Robustness tests

ED

We recognize that dividend policy may be endogenously determined as firms self-select based on

the various characteristics such as earned/contributed capital mix, agency conflicts, profitability,
PT

or investment opportunities, among others (e.g., DeAngelo, DeAngelo, and Skinner, 2004; DeAn-

gelo, DeAngelo, and Stulz, 2006; Denis and Osobov, 2008; Chay and Suh, 2009). In our baseline
CE

regressions in Table 4, we attempt to mitigate the endogeneity concerns by regressing the earnings

management measures on the lagged dividend policy, while also controlling for the lagged earnings
AC

management and earnings variables, and obtain robust results. Importantly, however, it is also

plausible that dividend payments are an outcome of the past earnings which in turn may be related

to accruals metrics, and we do not aim to rule out such reverse causality possibilities.

To further address the endogeneity concerns, we employ six different tests and present the results

in Table 5. We employ Accr(D) as the dependent variable in model M1, Accr(J) in all other models,

except models M20-M24, where we use the remaining alternative earnings management proxies, as

elaborated below. In unreported models, we also employ all three earnings management measures

16
 ACCEPTED MANUSCRIPT

- Accr(J), Accr(D), and Accr(F ) for all tests, as applicable, and obtain consistent results.

First, we replicate the baseline regression (1), shown in model M3 of Table 4, while controlling

for firm fixed effects. Such an approach allows us to verify whether our findings are driven by

unobserved time-invariant attributes that cause firms to self-select into dividend payers. The result

T
in model M1 underscores a negative Div coefficient that is significant at the 1% level, suggesting

P
RI
that the relationship between dividend paying status and earnings management persists and is

insensitive to controlling for firm fixed effects.

SC
Second, we investigate whether dividend paying status indeed represents the link between div-

idend payments and earnings management rather than captures other effects on earnings manipu-

NU
lation. For example, one could argue that dividend payers are normally large firms, exhibit strong

dividend stability, and low cash flow volatility. A plausible expectation thus arises that firms with
MA
the above characteristics are associated with lower accruals-based earnings manipulation. It is thus

also possible that the role of dividend paying status in earnings manipulation is subsumed once
ED

the effects of these characteristics are examined in greater detail. Although we carefully control for

earnings management determinants in all regressions, certain non-linearities may affect the reported
PT

relationships.

To address this concern, we partition dividend payers as well as dividend non-payers into large
CE

and small by the median firm size (M V ), dividend payers into those with high and low dividend

volatility as well as payers and non-payers into those with high and low cash flow volatility by their
AC

respective medians for each country and year. We calculate dividend and cash flow volatilities as

standard deviations of dividends or of free cash flows, respectively, in the previous eight years and

require a minimum of five years of data. We replicate the baseline regression (1), while estimate

the tests separately for each partition. To conserve space, we report the results of the estimations

which include the differential effects of Div with Large Size, High Div V ol, or with High CF V ol

in models M2-M4, respectively. The coefficients of Div are negative and strongly significant, and

those of the differential effects indicate that small dividend payers or those with high cash flow

17
 ACCEPTED MANUSCRIPT

volatility exhibit less earnings manipulation than do their counterparts of larger size or with lower

cash flow volatility. The differential effect pertinent to dividend volatility is insignificant. Also,

the results of unreported models which separately focus on large and small firms, those with high

and low dividend and cash flow volatilities underscore negatively significant coefficients of Div at

T
the 1% level across all models. Combined, the above results offer a strong evidence that the effect

P
of dividend paying status on earnings management is distinct and persists when conditioned on

RI
measures which could be closely associated with dividend paying status.

SC
Third, we examine the effects of various changes in dividend policy on earnings management.

In models M5 and M6, we assess the effects of dividend initiation and omission, respectively, on

NU
a change in earnings manipulation. We designate Initiation in model M5, a dummy variable

that takes the value of 1 if the firm does not pay dividends in year t − 1, while initiates dividend
MA
payments in year t, and continues to pay dividends in year t + 1. In model M6, we incorporate

Omission - a dummy indicator that takes the value of 1 if the firm distributes dividends in year
ED

t − 1, but eliminates dividends in year t, and continues to not pay dividends in year t + 1. We

estimate a panel OLS regression that is similar to that of Table 4, except that the dependent
PT

variable now represents the change in earnings management metrics from year t to year t + 1,

and the key explanatory variables are Initiation (M5) or Omission (M6). The control variables
CE

incorporate changes in the firm characteristics, calculated in the same manner as is the change in

the dependent variable. Our sample contains 3,284 firm-year observations associated with dividend
AC

initiations and 2,770 firm-year observations pertinent to dividend omissions. To be included in the

sample, we require that firms have a total of eight years of non-missing information centered on

dividend initiation/omission event (i.e., from year t − 4 to year t + 4).

If dividend paying status is predictive of lower earnings management, we expect a negative

(positive) change in earnings management following a dividend initiation (omission). The result

in model M5 is consistent with our prediction. The estimate of Initiation coefficient is negative

and statistically significant. Model M6, however, underscores statistically insignificant coefficient

18
 ACCEPTED MANUSCRIPT

of Omission. It is thus evident that employing a dividend policy results in lower earnings man-

agement, whereas there is no evidence to suggest that dividend omissions lead to more aggressive

earnings management. Such finding is in line with prior literature that establishes that the role of

dividend cuts in firm performance is ambiguous. For instance, Grullon et al. (2005) document that

T
dividend cuts do not relate to future firm performance. Instead, Benartzi, Michaely, and Thaler

P
(1997), Grullon, Michaely, and Swaminathan (2002), and Healy and Palepu (1988) report increased

RI
earnings subsequent to a dividend cut, while DeAngelo, DeAngelo, and Skinner (1992) find that

SC
firms cut dividends after a period of underperformance.

We are further interested to examine whether the dividend decreases and increases relate to or

NU
exhibit varied impacts on earnings manipulation. In model M7, we construct an indicator variable,

Decrease, which takes the value of 1 if a firm decreases its dividend payments from year t − 1 to
MA
year t, and 0 if otherwise. In model M8, we incorporate Increase - a dummy variable that takes

the value of 1 if a firm increases its dividend payments from year t − 1 to year t, and 0 if otherwise.
ED

We estimate a panel OLS regression that is similar to that of Table 4, while also includes the

interaction terms of Decrease and Increase with Div in models M7 and M8, respectively. The
PT

results underscore negative significant coefficients of Div in both models M7 and M8, while those

of the differential effects are insignificant. Such findings indicate that the effect of divided policy
CE

on earnings manipulation persists and is insensitive to dividend decreases or increases. Such results

corroborate those of Skinner and Soltes (2011). The authors find that the relationship between
AC

dividends and earnings persistence is unrelated to the magnitude of dividends thereby underscoring

the effect of the dividend paying status rather than of the dividend payment amount.

We next focus on the subsamples of dividend payers and non-payers which persist in their

respective dividend paying status for at least 10 years in model M9, and for at least 20 years

in model M10. In all other tests throughout the study, we assure that the survivorship bias is

minimized and thus do not restrict our samples based on time periods or availability of measures.

We, however, recognize that firms may change from dividend payers to dividend non-payers and

19
 ACCEPTED MANUSCRIPT

vice versa. We are thus interested to assess whether the relationship between dividend paying

status and earnings management persists in subsamples that restrict firms to a dividend paying

status for a certain period of time. We estimate regressions similar to those of Table 4, while focus

on the subsamples with at least 10 years of data in a dividend paying status in model M9, and with

T
at least 20 years of data in a dividend paying status in model M10. We obtain 73,053 and 4,980

P
firm-year observations in models M9 and M10, respectively. Our results of the negative significant

RI
coefficients of Div remain unaltered in both models M9 and M10. Our key finding of the negative

SC
link between dividend paying status and earnings management is thus insensitive to firms changing

between dividend payers and non-payers.

NU
Fourth, we undertake the propensity score matching approach to mitigate the concern that

the observable firm characteristics associated with the dividend paying status cause differences
MA
in the relationship between dividend paying status and earnings management. In the first stage,

we model the likelihood of dividend paying status for each country-year, to obtain propensity
ED

scores; the logistic regression includes M V , F CF , Salesg , ROA, and industry fixed effects. We

next employ the propensity score-matched sample to estimate the OLS regression of the effect of
PT

dividend paying status on earnings manipulation, and report the results in model M11. The finding

is consistent with our expectation that divided payers exhibit less earnings management than do
CE

their dividend non-paying counterparts. To complement this finding, we also match dividend payers

with non-payers based on country, industry, and year fixed effects as well as on one of the following
AC

firm characteristics - M V , F CF , Salesg , Age, or ROA,11 and report the regression results in

models M12-M16, respectively. The coefficients of Div are negative and strongly significant at the

1% level across all models, thus reinforcing the negative link between dividend paying status and

earnings manipulation.

Fifth, we employ alternative measures of dividend policy to verify whether the relationship be-

tween dividend paying status and earnings management persists. Models M17 and M18 employ the

11
We restrict firms to those with positive return on assets.

20
 ACCEPTED MANUSCRIPT

dividend-price ratio (Div/P ) and dividend payout ratio (Div/E), respectively. With the increased

popularity of repurchases as a means to distribute cash flows to outside investors (e.g., Grullon,

Michaely, and Swaminathan, 2002; Skinner, 2008), we also exploit share repurchases (Repur) as an

alternative proxy for payout to shareholders and report the results in model M19. The results in

T
models M17-M19 further indicate that employing dividend-price ratio, dividend payout ratio, and

P
share repurchases as alternative dividend policy/payout constructs, renders no effect on our key

RI
finding of a negative relationship between the dividend paying status and earnings management. In

SC
an unreported test, we exclude firms with negative dividend payout ratios and obtain qualitatively

similar results.

NU
Finally, we re-examine the relationship between dividend policy and earnings management

while exploiting the alternative metrics of earnings manipulation. It is plausible that managers
MA
may prefer either earnings-increasing or earnings-decreasing manipulation. To disentangle the po-

tential differential effects of these two types of earnings management behavior, we partition the
ED

sample by the sign of Jones’ (1991) abnormal accruals. Model M20 focuses on the subsample of

income-increasing/positive accruals (P Accr), while model M21 examines the partition of income-
PT

decreasing/negative accruals (N Accr). Model M22 focuses on Accr(F ), while models M23 and

M24 exploit the probabilities of small profits - P rob(S), and of small earnings per share increases
CE

- P rob(SEP S), respectively. The latter two measures focus on management’s incentives to ma-

nipulate earnings to beat target earnings (e.g., Leuz, Nanda, and Wysocki, 2003; Dechow, Ge,
AC

and Schrand, 2011). Specifically, Burgstahler and Dichev (1997) argue that managers have strong

incentives to avoid reporting a loss or a decrease in earnings. Further, studies document that re-

porting a small profit to avoid losses is associated with earnings management using discretionary

loss reserves in insurance companies (e.g., Beaver et al., 2003) or deferred tax expenses (e.g., Philips

et al., 2003).

The results in models M20-M24 provide further support to our key findings. While the dividend-

earnings management relationship is stronger for negative than for positive accruals, models M20

21
 ACCEPTED MANUSCRIPT

and M21 indicate that dividend payers are less likely to manage earnings, either upward or down-

ward. Using three alternative measures of earnings management as dependent variables in models

M22-M24 also provides qualitatively similar results.

Overall, the multitude of tests documented above yield robust evidence that dividend payments

T
are negatively associated with and lead to an decrease in earnings management.

P
RI
5. Dividend policy, earnings management, and country-level agency

SC
concerns

5.1. Dividends-earnings management relationship, investor protection, and transparency

NU
Thus far, we have established corroborating evidence that dividends relate to earnings management
MA
across a broad spectrum of developed and emerging markets. We now proceed to ascertain whether

the above relationship persists when the sample is partitioned by the degree of country-level in-

vestor protection and transparency measures, and whether and how it varies in these measures. If
ED

firms pay dividends to consume and conceal fewer private control benefits, and thus exhibit fewer

incentives to manage earnings, we expect the negative relationship between dividend paying status
PT

and earnings management to be more pronounced in firms from countries with weak institutions

and low transparency. These firms may be attempting to mitigate country-level agency concerns to
CE

establish reputation or credibility in the market. Employing a large sample of firms from 29 devel-

oped and developing economies with a wide variation of legal protection measures and of differing
AC

degrees of transparency offers an opportune setting to determine such plausible effects.

Our study employs four legal protection metrics and three transparency measures. The former

are (i) a firm’s legal origin (Legal Origin), (ii) the rule of law (Rule of Law), (iii) law and order

(Law & Order), and (iv) the government effectiveness (Gov Effectiveness), and the latter include

(i) financial reporting extensiveness (Report Extensiveness), (ii) financial reporting quality (Interim

Report Quality), and (iii) the governance transparency (Gov Transparency). We follow La Porta

et al. (2000) to employ Legal Origin as a proxy for agency costs at the country level. We include

22
 ACCEPTED MANUSCRIPT

Rule of Law to measure the extent to which a country enforces its law and order protecting mi-

nority shareholders from La Porta et al. (1998). Law & Order comprises ”law” sub-component

which assesses the strength and impartiality of the legal system, and the ”order” sub-component

which assesses popular observance of the law from International Country Risk Guide (ICRG). Gov

T
Effectiveness reflects perceptions of the quality of public and of civil services, their independence

P
from political pressures, the quality of and government’s commitment to policies formation and

RI
implementation from Kaufmann, Kraay, and Mastruzzi (2009). Report Extensiveness captures the

SC
extensiveness of financial reporting by examining the inclusion or omission of 90 accounting items

in annual financial statements in 1995 from the Center for Financial Analysis and Research (CI-

NU
FAR). Interim Report Quality measures the frequency and breadth of interim financial disclosure

from CIFAR. Gov T ransparency represents the intensity of governance disclosures used by outside
MA
investors to hold officers and directors accountable from Bushman, Piotroski, and Smith (2004)

(BPS (2004)). The higher these seven metrics are, the stronger is a country’s investor protection

and transparency.
ED

To implement the tests, we replicate our baseline regression (1), while partitioning the sample
PT

into two groups by the median or by a dummy indicator to reflect the strength of each metric -

Strong and W eak. For example, in the case of Legal Origin, our sample of firms is divided into
CE

those from common versus civil law countries, with common law classified as Strong and civil law

as W eak investor protection. Table 6 reports the results, and to conserve space, it only highlights
AC

the coefficients of Div and the associated p-values for the two groups of firms, together with the

t−test for the difference in the Div coefficients between the two groups. This approach allows us to

evaluate whether the relationship between dividend paying status and earnings management varies

for firms operating in countries with strong versus those operating in weak investor protection

or low transparency environments. Panel A reports the results for all countries, while Panel B

highlights the results without the U.S. in the sample.

The results are strongly consistent with our prediction. The coefficients of Div are negative

23
 ACCEPTED MANUSCRIPT

and highly significant at the 1% level across all measures of investor protection and transparency

in both Panels A and B. Importantly, the absolute values of the Div coefficient are much higher

for the W eak than for the Strong category, and the differences in the estimates for these two

groups are highly significant at the 1% or 5% level across most proxies of investor protection and

T
transparency, with the exception of Gov Transparency 12 in both Panels A and B, and Legal Origin,

P
Rule of Law, and Report Extensiveness in Panel B. The results suggest that dividend payers from

RI
countries with the least investor protection or the lowest transparency display significantly lower

SC
earnings management than do their counterparts from countries with the most investor protection

or the highest transparency. It is plausible that firms from countries with weak investor protection

NU
or low transparency may employ dividend policies associated with lower earnings manipulation to

convey their commitment to shareholder interests.

MA
5.2. Dividends-earnings management relationship and equity issuance

To further verify whether the relationship between dividend policy and earnings management serves
ED

to mitigate country-level agency concerns, we investigate the above link in the context of subsequent

access to external financing. We postulate that if the dividend policy is employed as a means to
PT

build reputation in the capital markets, we should observe dividend payers to exhibit lower earnings
CE

management especially if they plan to subsequently issue equity. Also, we ought to find such

evidence to be more pronounced in firms from countries with weak institutions or low transparency.
AC

This is because firms in these countries may have fewer alternative country-level means to display

their commitment to act in the interests of their investors.

To perform the tests, we replicate the baseline regression (1), shown in model M3 of Table 4,

while controlling for Issue, a dummy indicator that takes the value of 1 if the firm issues equity in

years t + 1 to t + 3 subsequent to dividend payments, as well as its interactions with Div, and with

each of the control characteristics. We report the results in Table 7 across five different models as

we employ the alternative metrics of earnings management. Specifically, models M1-M5 focus on
12
Please note that the U.S. is classified as Weak for Gov Transparency measure.

24
 ACCEPTED MANUSCRIPT

Accr(J), Accr(D), Accr(F ), P rob(S), and P rob(SEP S), respectively. Panel A reports the results

for all countries, while Panel B presents the results without the U.S. in the sample. To conserve

space, we highlight the coefficients of Div, Issue dummy, and their interaction Div*Issue, along

with the associated p-values.

T
The results in Table 7 are fully supportive of our conjecture. The coefficients of dividend paying

P
RI
status are negative and significant across all models in both Panels A and B, with the exception

of model M4 in Panel B, reinforcing the earlier findings that dividend payers are less likely to

SC
display aggressive earnings management. The coefficients of interaction terms (i.e., Div*Issue)

are also negative and strongly significant in all models, with the exception of model M5 in Panel

NU
B. Therefore, dividend paying firms that issue equity following dividend payments, exhibit lower

earnings management than their non-issuing counterparts.

MA
We further assess how the dividends-earnings management relationship varies with subsequent

equity issuance and with differing degrees of investor protection and transparency. We employ the
ED

investor protection and transparency metrics as those in Table 6 - Legal Origin, Rule of Law, Law

& Order, Gov Effectiveness, Report Extensiveness, Interim Report Quality, and Gov Transparency.
PT

We perform the regression analyses as those in Table 7, but the sample is partitioned into two

groups by the median or by the dummy indicator to convey the strength of each investor protection
CE

or transparency metric - Strong and W eak.

Table 8 reports the results and emphasizes the coefficients of Div, Issue, and of their interaction
AC

Div*Issue, and the associated p-values. Results are fully consistent with our expectations. The

coefficients of Div are negative and strongly significant across all investor protection and trans-

parency metrics. The coefficients of differential effects, Div*Issue, are lower for firms in W eak

category than for those in Strong category across all country-level agency-related variables, except

for Legal Origin in both Panels A and B, and Gov Transparency in Panel B. Interestingly, the

coefficients of the differential effects are significant across all investor protection and transparency

measures, except Legal Origin in both Panels A and B, only in W eak, while are insignificant in

25
 ACCEPTED MANUSCRIPT

most Strong partitions. Such results underscore that dividend payers that subsequently issue eq-

uity, exhibit lower levels of earnings management than their non-issuing counterparts only in firms

from countries with weak investor protection and high opacity, while not in those from countries

with strong legal institutions and high transparency.

T
Taken together, the evidence in Tables 7 and 8 suggests that the dividends-earnings management

P
RI
link is stronger in firms that plan to issue equity following dividend payments, especially in those

with stronger country-level agency concerns. Such evidence renders a valuable support to our

SC
conjecture that the role of dividends associated with a strong commitment to shareholder interests

as manifested by lower levels of earnings manipulation is particularly underscored when firms wish

NU
to establish their credibility in the capital markets.
MA
6. Conclusion

This paper aims (i) to assess whether dividend policy affects earnings management; (ii) to evaluate
ED

whether and how the relationship between dividend paying status and earnings manipulation varies

with country-level investor protection and transparency; and (iii) to entertain a plausible explana-
PT

tion for the variation in the dividends-earnings management link with country-level agency concerns

by exploiting firms’ future access to external financing. Employing a large sample of 23,429 firms
CE

from 29 developed and emerging markets, with wide variation in country-level governance and the

strength of legal institutions, offers an opportune platform to ascertain whether and how dividend
AC

policies exert such effects.

We find a robust negative relationship between the dividend paying status and earnings man-

agement, even after conditioning on a number of variables that have been previously shown to

affect earnings management. The dividends-earnings management relationship is also robust to the

partitions into various subsamples, endogeneity tests, as well as to alternative measures of dividend

policy and of earnings management. Conducting a number of additional tests, we further verify

robustness of our key findings. Subsequent analyses show that the negative relationship between

26
 ACCEPTED MANUSCRIPT

dividend policy and earnings manipulation is more pronounced in firms operating in countries with

weak investor protection and high opacity. Finally, the dividends-earnings management link of

firms that issue equity following dividend payments is stronger than that of their counterparts that

do not seek subsequent access to external financing, and only in those with greater country-level

T
agency problems.

P
RI
Overall, our findings suggest that firms may employ dividend policies as they have a desire to

consume fewer private control benefits, thereby decreasing the necessity to conceal consumption of

SC
these benefits via aggressive earnings manipulation. Importantly, the negative relationship between

dividend paying status and earnings management is more pronounced in firms from countries with

NU
stronger agency problems. Such evidence implies that dividend policies associated with lower levels

of earnings manipulation may serve to mitigate agency concerns particularly of firms that may
MA
have fewer country-level alternative means to convey their commitment to shareholder interests.

This inference is primarily supported by the findings that the dividends-earnings management
ED

relationship is stronger in firms that subsequently access the equity markets, especially in countries

with weak legal protection and high opacity. It appears that if the benefits of acting in line with
PT

shareholder interests outweigh the costs of reduced expropriation by controlling shareholders, firms

may self-impose the constraint of paying dividends associated with lower earnings manipulation to
CE

capture reputation effects.

AC

27
 ACCEPTED MANUSCRIPT

References

Allen, F., A. Bernardo, and I. Welch. 2000. A theory of dividends based on tax clienteles. Journal of
Finance 55, 2499-2536.
Barth, M., W. Landsman, and M. Lang. International accounting standards and accounting quality. Jour-
nal of Accounting Research 46, 467-498.

T
Bartram, S.M., P. Brown, J.C. How, and P. Verhoeven. 2012. Agency conflicts and corporate payout

P
policies: A global study. Warwick Business School, University of New South Wales, and Queensland
University of Technology working paper.

RI
Beaver, W., M. McNichols, and K. Nelson. 2003. Management of the loss reserve accrual and the distri-
bution of earnings in the property-casualty insurance industry. Journal of Accounting and Economics

SC
35, 347-376.
Becker, C., M. DeFond, J. Jiambalvo, and K. Subramanyam. 1998. The effect of audit quality on earnings
management. Contemporary Accounting Research 15, 1-24.

NU
Benartzi, S., R. Michaely, and R. Thaler. 1997. Do changes in dividends signal the future or the past?
Journal of Finance 52, 1007-1034.
Brav, A., J. Graham, C. Harvey, and R. Michaely. 2005. Payout policy in the 21st century. Journal of
MA
Financial Economics 77, 483-525.
Burgstahler, D. and I. Dichev. 1997. Earnings management to avoid earnings decreases and losses. Journal
of Accounting and Economics 24, 99-126.
Bushman, R., J. Piotroski, and A. Smith. 2004. What determines corporate transparency? Journal of
ED

Accounting Research 42, 207-252.

Caskey, J. and M. Hanlon. 2013. Dividend policy at firms accused of accounting fraud. Contemporary
Accounting Research 30, 818-850.
PT

Caton, G., C. Chiyachantana, C. Chua, and J. Goh. 2011. Earnings management surrounding seasoned
bond offerings: Do managers mislead ratings agencies and the bond market? Journal of Financial and
Quantitative Analysis 46, 687-708.
CE

Chang, X., S. Dasgupta, and G. Hilary. 2006. Analyst following and financing decisions. Journal of Finance
61, 3009-3048.
Chang, X., S. Dasgupta, and G. Hilary. 2009. The effect of auditor quality on financing decisions. The
AC

Accounting Review 84, 1085-1117.

Chay, J.B. and J. Suh. 2009. Payout policy and cash-flow uncertainty. Journal of Financial Economics
93, 88-107.
Chemmanur, T., J. He, G. Hu, and H. Liu. 2010. Is dividend smoothing universal? New insights from a
comparative study of dividend policies in Hong Kong and the U.S. Journal of Corporate Finance 16,
413-430.
Choi, J. H. and T.J. Wong. 2007. Auditors’ governance functions and legal environments: An international
investigation. Contemporary Accounting Research 24, 13-44.
DeAngelo, H., L. DeAngelo, and D. Skinner. 1992. Dividends and losses. Journal of Finance 47, 1837-1863.
341-371.
DeAngelo, H., L. DeAngelo, and D. Skinner. 2004. Are dividends disappearing? Dividend concentration
and the consolidation of earnings. Journal of Financial Economics 72, 425-456.

28
 ACCEPTED MANUSCRIPT

DeAngelo, H., L. DeAngelo, and R. Stulz. 2006. Dividend policy and the earned/contributed capital mix:
a test of the life-cycle theory. Journal of Financial Economics 81, 227-254.
Dechow, P. and I. Dichev. 2002. The quality of accruals and earnings: The role of accrual estimation
errors. The Accounting Review 77, 35-59.
Dechow, P., W. Ge, and C. Schrand. 2011. Understanding earnings quality: A review of the proxies, their
determinants and their consequences. Journal of Accounting and Economics 50, 344-401.

T
Denis, D. and I. Osobov. 2008. Why do firms pay dividends? International evidence on the determinants

P
of dividend policy. Journal of Financial Economics 89, 62-82.

RI
Dewenter, K. and V. Warther. 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.
Easterbrook, F. 1984. Two agency-cost explanations of dividends. American Economic Review 74, 650-659.

SC
Faccio, M., L. Lang, and L. Young. 2001. Dividends and expropriation. American Economic Review 91,
54-78.

NU
Fama, E. and K. French. 2001. Disappearing dividends: Changing firm characteristics or lower propensity
to pay? Journal of Financial Economics 60, 3-43.
Fan, J. P. and T. J. Wong. 2005. Do external auditors perform a corporate governance role in emerging
markets? Evidence from East Asia. Journal of Accounting Research 43, 35-72.
MA
Fatemi, A. and R. Bildik. 2012. Yes, dividends are disappearing: Worldwide evidence. Journal of Banking
and Finance 36, 625-922.
Francis, J., R. LaFond, P. Olsson, and K. Schipper. 2005. The market pricing of accruals quality. Journal
of Accounting and Economics 39, 295-327.
ED

Francis, J. and D. Wong. 2008. The joint effect of investor protection and Big 4 audits on earnings quality
around the world. Contemporary Accounting Research 25, 157-191.
PT

Gopalan, R. and S. Jayaraman. 2012. Private control benefits and earnings management: Evidence from
insider controlled firms. Journal of Accounting Research 50, 117-157.
Grullon, G. and R. Michaely. 2002. Dividends, share repurchases, and the substitution hypothesis. Journal
CE

of Finance 57, 1649-1684.

Grullon, G., R. Michaely, and B. Swaminathan. 2002. Are dividend changes a sign of firm maturity?
Journal of Business 75, 387-424.
AC

Grullon, G., R. Michaely, S. Benartzi, and R.H. Thaler. 2005. Dividend changes do not signal changes in
future profitability. Journal of Business 78, 1659-1682.
Grullon, G., R. Michaely, and B. Swaminathan. 2002. Are dividend changes a sign of firm maturity?
Journal of Business 75, 387-424.
Hail, L., A. Tahoun, and C. Wang. 2014. Dividend payouts and information shocks. Journal of Accounting
Research 52, 403-456.
Healy, P., and K. Palepu. 1988. Earnings information conveyed by dividend initiations and omissions.
Journal of Financial Economics 21, 149-175.
Jensen, M., 1986. Agency costs of free cash flow, corporate finance, and takeovers. American Economic
Review 76, 323-329.
Jensen, M. and W. Meckling. 1976. Theory of the firm: Managerial behavior, agency costs, and ownership
structure. Journal of Financial Economics 3, 305-360.

29
 ACCEPTED MANUSCRIPT

Jones, J. 1991. Earnings management during import relief investigations. Journal of Accounting Research
29, 193-228.
Kaufmann, D., A. Kraay, and M. Mastruzzi. 2009. Governance matters VIII: Aggregate and individual
governance indicators, 1996-2008. The World Bank.
La Porta, R., F. Lopez-De-Silanez, A. Shleifer, and R. Vishny. 1998. Law and finance, Journal of Political
Economy 106, 1113-1155.

T
La Porta, R., F. Lopez-De-Silanez, A. Shleifer, and R. Vishny. 2000. Agency problems and dividend policies

P
around the world. Journal of Finance 55, 1-33.

RI
Lang, M., J. Raedy, and W. Wilson. 2006. Earnings management and cross listing: Are reconciled earnings
comparable to US earnings? Journal of Accounting and Economics 42, 255-283.
Lang, L. and R. Litzenberger. 1989. Dividend announcements: Cash flow signaling vs. free cash flow

SC
hypothesis? Journal of Financial Economics 24, 181-191.
Leuz, C., D. Nanda, and P. Wysocki. 2003. Earnings management and investor protection: An international
comparison. Journal of Financial Economics 69, 505-527.

NU
Lintner, J. 1956. Distribution of incomes of corporations among dividends, retained earnings, and taxes.
American Economic Review 46, 97-113.
Myers, S. 2000. Outside equity. Journal of Finance 55, 1005-1037.
MA
Philips, J., M. Pincus, and S. Rego. 2003. Earnings management: New evidence based on deferred tax
expenses. The Accounting Review 78, 491-521.
Pinkowitz, L., R. Stulz, and R. Williamson. 2006. Does the contribution of corporate cash holdings and
dividends to firm value depend on governance? A cross-country analysis. Journal of Finance 61,
ED

2725-2751.
Rozeff, M. 1982. Growth, beta and agency costs as determinants of dividend payout ratios. Journal of
Financial Research 5, 249-259.
PT

Shleifer, A. and D. Wolfenzon. 2002. Investor protection and equity markets. Journal of Financial
Economics 66, 3-27.
Skinner, D. 2008. The evolving relation between earnings, dividends, and stock repurchases. Journal of
CE

Financial Economics 87, 582-609.

Skinner, D. and E. Soltes. 2011. What do dividends tell us about earnings quality? Review of Accounting
Studies 16, 1-28.
AC

Stulz, R. 1990. Managerial discretion and optimal financing policies. Journal of Financial Economics 26,
3-26.
Teoh, S., I. Welch, and T. Wong. 1998a. Earnings management and the underperformance of seasoned
equity offerings. Journal of Financial Economics 50, 63-99.
Teoh, S., I. Welch, and T. Wong. 1998b. Earnings management and long-term underperformance of initial
public offerings. Journal of Finance 53, 1935-1974.

30
 ACCEPTED MANUSCRIPT

Table 1
Summary Statistics by Country

This table presents the starting year of each country’s data availability (SYear), the mean values of firm-level variables
for the number of unique firms (NFirms), the number of firm-year observations (NFirm-Yrs), the proportion of

T
dividend-paying firm-years for the full sample period (Full) and for two subperiods (1990-1999 and 2000-2010),
dividend-price ratio (Div/P ), and three different earnings management measures following Jones (1991) Accr(J),

P
Dechow and Dichev (2002) Accr(D), and Francis et al. (2005) Accr(F ). All variables are defined in Appendix B.

RI
The sample period is from 1990 to 2010.

SC
Proportion of Dividend Payers Earnings Management Measures
Country SYear NFirms NFirm-Yrs Full (Firm-Yrs) ’90-’99 ’00-’10 Div/P Accr(J) Accr(D) Accr(F )

Developed Markets

NU
Australia 1990 937 5,199 0.568 0.848 0.504 0.027 0.072 0.064 0.061
Belgium 2000 96 547 0.693 0.693 0.022 0.054 0.039 0.035
Canada 1990 782 3,366 0.504 0.701 0.445 0.021 0.057 0.052 0.048
Denmark 1996 131 915 0.737 0.871 0.680 0.020 0.053 0.040 0.036
Finland 1997 123 981 0.838 0.950 0.816 0.037 0.052 0.038 0.035
MA
France 1990 701 5,940 0.710 0.794 0.662 0.019 0.055 0.034 0.032
Germany 1990 620 5,094 0.625 0.771 0.535 0.020 0.068 0.050 0.047
Hong Kong 1995 861 6,013 0.554 0.736 0.533 0.026 0.073 0.057 0.053
Italy 1997 242 1,607 0.685 0.745 0.678 0.019 0.053 0.035 0.034
Japan 1990 2,582 29,353 0.886 0.885 0.886 0.013 0.035 0.019 0.018
Netherlands 2001 111 655 0.441 0.441 0.020 0.064 0.039 0.037
ED

Norway 1992 167 1,435 0.779 0.851 0.725 0.028 0.055 0.037 0.035
Singapore 1997 181 1,234 0.814 0.913 0.799 0.032 0.061 0.042 0.039
Spain 1994 144 1,116 0.739 0.735 0.741 0.022 0.052 0.032 0.030
Sweden 1996 236 1,560 0.692 0.850 0.645 0.023 0.051 0.043 0.040
Switzerland 1993 216 1,966 0.735 0.803 0.709 0.017 0.047 0.032 0.030
PT

United Kingdom 1990 2,107 15,968 0.723 0.899 0.602 0.026 0.062 0.048 0.044
United States 1990 7,989 60,969 0.365 0.461 0.300 0.008 0.060 0.058 0.050
CE

Emerging Markets
China 1997 337 1,994 0.483 0.559 0.475 0.016 0.074 0.040 0.037
India 2002 1224 4,202 0.777 0.777 0.019 0.071 0.057 0.044
Indonesia 1997 242 1,753 0.480 0.581 0.467 0.021 0.069 0.046 0.042
Malaysia 1993 708 5,836 0.684 0.868 0.647 0.021 0.063 0.041 0.036
AC

Philippines 2001 111 655 0.441 0.441 0.020 0.064 0.039 0.037
Poland 2004 124 480 0.490 0.490 0.017 0.076 0.044 0.042
South Africa 1997 181 1,234 0.814 0.913 0.799 0.032 0.061 0.042 0.039
South Korea 1995 810 4,417 0.696 0.730 0.691 0.021 0.059 0.038 0.035
Taiwan 2001 611 3,844 0.658 0.658 0.031 0.057 0.037 0.032
Thailand 1997 334 2,513 0.644 0.436 0.677 0.041 0.063 0.044 0.042
Turkey 2003 91 471 0.552 0.552 0.030 0.075 0.052 0.046

By Market Type
Developed 18,656 146,941 0.584 0.675 0.531 0.015 0.055 0.046 0.042
Emerging 4,773 27,399 0.660 0.730 0.653 0.024 0.064 0.043 0.038
Aggregate Measures 23,429 174,340 0.596 0.678 0.557 0.016 0.057 0.045 0.042

31
 ACCEPTED MANUSCRIPT

Table 2
Comparison between Dividend Payers and Dividend Non-Payers

T
This table presents univariate tests of the difference of each firm-level variable between dividend payers and dividend

P
non-payers across 29 countries. The variables are three different measurements of accruals - Accr(J), Accr(D) and
Accr(F ); market capitalization (M V ); free cash flow (F CF ); book-to-market equity ratio (BM ); sales growth

RI
(Salesg ); firm age (Age); debt-equity ratio (DE); property, plant, and equipment (P P E); return on assets (ROA);
international accounting standards (IAS); Big 4, 5, or 8 auditors (BigN ); closely-held ownership (Chold); and
American Depositary Receipts (ADR). All variables are defined in Appendix B. NObs is the number of firm-year

SC
observations. For each variable, the table shows the p-values associated with the t−test for the difference in mean
values of the dividend payers and dividend non-payers and with the Wilcoxon test for the difference in median
values. The sample period is from 1990 to 2010.

NU
Dividend Payers Dividend Non-Payers Tests for Differences
Variable NObs Mean Median Std NObs Mean Median Std Mean Median
MA
Accr(J) 103,931 0.049 0.034 0.050 70,409 0.075 0.051 0.073 < 0.01 < 0.01
Accr(D) 97,506 0.033 0.024 0.030 62,924 0.066 0.051 0.054 < 0.01 < 0.01
Accr(F ) 97,210 0.030 0.023 0.027 62,464 0.061 0.047 0.050 < 0.01 < 0.01
ED

MV 103,931 12.697 12.579 1.895 70,409 11.333 11.289 1.960 < 0.01 < 0.01
F CF 103,931 0.090 0.084 0.072 54,886 0.009 0.052 0.197 < 0.01 < 0.01
PT

BM 103,931 0.830 0.631 0.767 54,886 0.748 0.521 1.127 < 0.01 < 0.01
Salesg 103,931 0.103 0.076 0.293 70,409 0.231 0.084 0.849 < 0.01 < 0.01
CE

Age 103,931 16.183 15.000 9.989 70,409 11.385 9.000 8.178 < 0.01 < 0.01
DE 103,931 0.761 0.445 1.360 70,409 0.738 0.263 2.284 0.02 < 0.01
PPE 103,931 0.336 0.304 0.208 70,409 0.264 0.195 0.226 < 0.01 < 0.01
AC

ROA 103,931 0.055 0.053 0.081 70,409 -0.063 0.016 0.255 < 0.01 < 0.01
IAS 103,931 0.102 0.000 0.303 70,409 0.089 0.000 0.285 < 0.01 < 0.01
BigN 103,931 0.599 1.000 0.490 70,409 0.636 1.000 0.481 < 0.01 < 0.01
Chold 103,931 0.399 0.400 0.241 70,409 0.377 0.350 0.246 < 0.01 < 0.01
ADR 103,931 0.025 0.000 0.155 70,409 0.013 0.000 0.113 < 0.01 < 0.01

32
 Table 3
Correlation Matrix of Firm-Specific Variables
AC
The table reports the correlation matrix of Div, Div/P , Div/E, Accr(J), Accr(D), Accr(F ), market capitalization (M V ), free cash flow (F CF ),
book-to-market equity ratio (BM ), sales growth (Salesg ), firm age (Age), debt-equity ratio (DE), property, plant, and equipment (P P E), return
CE
on assets (ROA), international accounting standards (IAS), Big 4, 5, or 8 auditors (BigN ), closely-held ownership (Chold), and American
Depositary Receipts (ADR). The coefficients are all statistically significant at the 1% level. All variables are defined in Appendix B.
PT
Variables Div Div/P Div/E Accr(J) Accr(D) Accr(F )ED MV F CF BM DE Salesg Age PPE ROA IAS BigN Chold

Div/P 0.52
Div/E 0.31 0.29
Accr(J) -0.21 -0.10 -0.08
Accr(D) -0.37 -0.19 -0.13 0.47

33
MA
Accr(F ) -0.37 -0.19 -0.13 0.47 0.93
MV 0.33 0.08 0.10 -0.24 -0.32 -0.31
F CF 0.29 0.23 0.10 -0.20 -0.30 -0.31 0.31
NU
BM 0.05 0.14 0.02 -0.07 -0.10 -0.11 -0.26 0.03
DE 0.01 -0.03 0.00 -0.04 -0.08 -0.08 0.04 0.01 0.08
Salesg -0.11 -0.06 -0.05 0.08 0.09 0.09 0.02 -0.07 -0.08 -0.03
SC
Age 0.25 0.12 0.08 -0.15 -0.17 -0.17 0.26 0.10 0.04 0.06 -0.13
PPE 0.16 0.15 0.08 -0.16 -0.26 -0.27 0.08 0.17 0.09 0.11 -0.06 0.09
ROA 0.32 0.24 0.11 -0.22 -0.31 -0.33 0.29 0.85 0.06 0.04 -0.09 0.12 0.12
RI
IAS 0.02 0.07 0.03 0.03 0.03 0.04 0.05 0.01 -0.01 -0.01 0.03 0.00 -0.05 0.03
P
ACCEPTED MANUSCRIPT

BigN -0.04 0.07 0.00 -0.04 -0.02 -0.02 0.25 0.11 -0.09 -0.07 0.01 0.04 0.03 0.07 0.05
T
Chold 0.05 0.04 0.02 0.05 -0.01 -0.01 -0.25 0.02 0.09 0.02 -0.02 -0.21 0.02 0.05 0.07 -0.21
ADR 0.04 0.02 0.01 -0.03 -0.05 -0.05 0.19 0.03 -0.03 0.01 0.00 0.05 0.05 0.02 0.04 0.05 -0.06
 ACCEPTED MANUSCRIPT

Table 4
The Relationship between Dividend Policy and Earnings Management
This table presents firm-level OLS regressions, where the dependent variable is the earnings management
measure following Jones (1991), Accr(J) in year t + 1, and the key explanatory variable is the dummy
variable Div measured in year t, as are the remainder of explanatory variables. The control variables
include the lag of dependent variable (Accr(J)L ); firm size (M V ); free cash flow (F CF ); book-to-market
equity ratio (BM ); sales growth (Salesg ); firm age (Age); leverage (DE); long-term tangible assets (P P E);

T
profitability (ROA); the IAS dummy (IAS); a BigN dummy (BigN ); an ADR dummy (ADR); and closely-
held ownership (Chold). All variables are defined in Appendix B. Models M1-M3 employ the entire sample

P
of 29 countries; models M4-M5 focus on the U.S. and non-U.S. subsamples; and models M6-M7 examine
the subsamples of the 1990s versus the 2000s. All regressions include country, industry, and year fixed

RI
effects, and all associated p-values reported below parameter estimates are computed based on standard
errors adjusted for heteroskedasticity and firm-level clustering. NObs is the number of observations, and R̄2
is the adjusted R-squared. The sample period is between 1990 and 2010.

SC
Full Sample U.S. Non-U.S. 1990-1999 2000-2010
M1 M2 M3 M4 M5 M6 M7

NU
Div -1.928*** -0.487*** -0.437*** -0.572*** -0.621*** -0.427***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Accr(J)L 0.286*** 0.224*** 0.226*** 0.217*** 0.204*** 0.230***
MA
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
MV -0.375*** -0.413*** -0.293*** -0.302*** -0.388***
(0.00) (0.00) (0.00) (0.00) (0.00)
F CF -0.009 -0.082 0.266 2.221*** -0.743**
(0.98) (0.85) (0.57) (0.00) (0.05)
ED

BM -0.376*** -0.389*** -0.348*** -0.253*** -0.431***

(0.00) (0.00) (0.00) (0.00) (0.00)
Salesg 0.038 0.166** -0.062 0.057 -0.006
(0.31) (0.01) (0.18) (0.51) (0.89)
Age -0.009*** -0.007** -0.014*** -0.019*** -0.006**
PT

(0.00) (0.03) (0.00) (0.00) (0.01)

DE -0.017* -0.026 -0.003 0.002 -0.024*
(0.07) (0.12) (0.80) (0.89) (0.08)
CE

PPE -2.620*** -2.511*** -2.698*** -3.323*** -2.276***

(0.00) (0.00) (0.00) (0.00) (0.00)
ROA -3.044*** -2.861*** -3.430*** -4.681*** -2.705***
(0.00) (0.00) (0.00) (0.00) (0.00)
IAS -0.001 3.332*** -0.274*** -0.001 -0.169**
AC

(0.99) (0.00) (0.00) (1.00) (0.04)

BigN -0.298*** -0.909*** -0.176*** -0.292*** -0.401***
(0.00) (0.00) (0.00) (0.00) (0.00)
ADR 0.357*** 0.080 0.277 0.379***
(0.00) (0.51) (0.12) (0.01)
Chold 0.146* -0.101 0.277*** 0.094 0.264**
(0.09) (0.48) (0.01) (0.50) (0.01)

R̄2 7.20% 13.1% 17.3% 18.0% 17.1% 16.6% 17.7%

NObs 143,390 143,390 143,390 50,184 93,206 50,042 93,348
Country FE Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes

34
 Table 5
AC Robustness Tests
This table presents various robustness tests. Model M1 replicates the baseline model M3 of Table 4 while controlling for firm fixed effects, models M2-M4 incorporate interactions
of Div and Large Size, High Div V ol, and High CF V ol, respectively. We partition dividend payers and dividend non-payers into large and small by the median firm size ( M V ),
dividend payers into those with high and low dividend volatility, as well as payers and non-payers into those with high and low cash flow volatility by their respective medians for
each country and year. Models M5-M6 focus on dividend initiations and omissions, models M7-M8 test the effects of dividend decreases and increases, and models M9 and M10
CE
examine the subsamples of dividend payers and non-payers that persist in their respective dividend paying status for at least 10 years and 20 years. M11 replicates the baseline
model M3 of Table 4 using a propensity score matched (PSM) sample, and models M12-M16 employ samples that match dividend payers with non-payers on one of the following -
M V , F CF , Salesg , Age, or ROA. Models M17-M19 employ the alternative constructs of payouts - dividend-price ratio (Div/P ), dividend payout ratio (Div/E), and repurchases
PT
(Repur). Models M20-M24 employ the alternative measures of the dependent variable - positive Jones’ (1991) accruals ( P Accr), negative Jones’ (1991) accruals (N Accr), Francis
et al.’s (2005) accruals (Accr(F )), the probability of the small profit (P rob(S)), and the probability of the small increases in earnings per share (P rob(SEP S)). The control
characteristics are the same as those in Table 4. All variables are defined in Appendix B. All regressions include country, industry, and year fixed effects (as applicable), and all
ED
associated p-values reported below parameter estimates are computed based on standard errors adjusted for heteroskedasticity and firm-level clustering. NObs is the number of
observations, and R̄2 is the adjusted R-squared. The sample period is between 1990 and 2010.

F irm Large High High Dividend Matched Sample by Firm Characteristic

35
MA
Variable FE Size Div V ol CF V ol Initiation Omission Decrease Increase > 10 Yrs > 20 Yrs P SM MV F CF Salesg Age ROA

M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 M12 M13 M14 M15 M16

NU
Div -0.037 -0.639 -0.424 -0.243 -0.494 -0.455 -0.408 -0.469 -0.538 -0.480 -0.542 -0.510 -0.454 -0.516
(< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01)
V ar*Div 0.340 -0.007 -0.175 0.039 -0.039
(< .01) (0.88) (0.07) (0.26) (0.27)
SC
Initiation -0.165
(0.03)
Omission -0.103
(0.27)
RI
R̄2 86.2% 17.3% 16.5% 16.6% 85.5% 85.8% 16.9% 16.9% 15.3% 12.1% 15.0% 15.5% 15.5% 15.2% 14.1% 15.2%
P
ACCEPTED MANUSCRIPT

NObs 131,295 143,390 73,050 73,070 3,284 2,770 118,536 118,536 73,053 4,980 174,211 151,075 150,994 150,740 151,075 152,930
Country FE No Yes Yes Yes No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE No Yes Yes Yes No No Yes Yes Yes Yes Yes
T
Yes Yes Yes Yes Yes
Year FE No Yes Yes Yes No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Firm FE Yes No No No Yes Yes Yes Yes No No No No No No No No
 AC
Table 5
CE
Robustness Tests – Continued
PT
Alternative Measures of Dividend Payouts Alternative Measures of Earnings Management
ED
Variable Div/P Div/E Repur P Accr N Accr Accr(F ) P rob(S) P rob(SEP S)
M17 M18 M19 M20 M21 M22 M23 M24

36
Div -7.413*** -0.116*** -0.386*** -0.588*** -0.096*** -0.282*** -0.482***
MA
(< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01)
Repur -0.140***
(< .01)
NU
R̄2 17.2% 17.3% 17.2% 16.3% 18.2% 80.8% 0.1415 0.1242
NObs 142,631 132,566 142,942 74,682 68,260 130,823 143,390 143,390
SC
Country FE Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
RI
P
ACCEPTED MANUSCRIPT

T
 ACCEPTED MANUSCRIPT

Table 6
Dividends-Earnings Management Relation, Investor Protection, and Transparency
This table presents firm-level OLS regressions focusing on country-level investor protection and transparency
measures based on the following empirical model,


N
Accr(J)t+1 = a + bDivt + cit Xit + Fixed Effects + ǫt+1 .
i

T
The dependent variable is the earnings management measure following Jones (1991), Accr(J), and the key

P
explanatory variable is the dummy variable Div. The control characteristics are the same as those in Table
4, and all variables are defined in Appendix B. The table presents only the coefficients of Div, i.e., the

RI
estimates of b from the model above, for seven investor protection and transparency measures (i.e., legal
origin, rule of law, law & order, government effectiveness, report extensiveness, interim report quality, and
governance transparency), partitioned by the strength of each measure into Strong and W eak. For each

SC
variable, the table also reports the t−test of difference in mean values between Strong and W eak categories,
and the respective p-values. Panel A reports results based on all countries, whereas Panel B presents
those employing all countries except the U.S. All regressions include industry and year fixed effects, and all
associated p-values reported below parameter estimates are computed based on standard errors adjusted for

NU
heterosedasticity and firm-level clustering. The sample period is between 1990 and 2010.

Country Difference
Institution Strong Weak (Weak-Strong)
MA
Panel A: All Countries
Legal Origin -0.489 -0.563 -0.074
(< .01) (< .01) (0.03)
Rule of Law -0.422 -0.628 -0.206
(< .01) (< .01) (< .01)
ED

Law & Order -0.493 -0.627 -0.134

(< .01) (< .01) (< .01)
Gov Effectiveness -0.452 -0.679 -0.227
(< .01) (< .01) (< .01)
PT

Report Extensiveness -0.466 -0.596 -0.130

(< .01) (< .01) (< .01)
Interim Report Quality -0.428 -0.683 -0.256
(< .01) (< .01) (< .01)
CE

Gov Transparency -0.558 -0.469 0.090

(< .01) (< .01) (0.15)
AC

Panel B: All Countries excluding the U.S.

Legal Origin -0.598 -0.563 0.035
(< .01) (< .01) (0.63)
Rule of Law -0.565 -0.627 -0.063
(< .01) (< .01) (0.40)
Law & Order -0.467 -0.752 -0.285
(< .01) (< .01) (0.04)
Gov Effectiveness -0.485 -0.679 -0.194
(< .01) (< .01) (0.02)
Report Extensiveness -0.512 -0.618 -0.105
(< .01) (< .01) (0.21)
Interim Report Quality -0.474 -0.683 -0.209
(< .01) (< .01) (0.08)
Gov Transparency -0.555 -0.601 -0.046
(< .01) (< .01) (0.85)

37
 ACCEPTED MANUSCRIPT

Table 7
Dividends-Earnings Management Relation and Subsequent Share Issuance
This table tests the link between dividend policy and earnings management in
the context of subsequent equity issuance. The dependent variables are Accr(J),
Accr(D), Accr(F ), the probability of the small profit (P rob(S)), and probability
of the small increases in EPS (P rob(SEP S)), measured in year t + 1 in Models
M1-M5, respectively. The key independent variables are a dividend dummy, Div;

T
a subsequent equity issuance dummy, Issue; and their interaction, Div × Issue.

P
The unreported control characteristics are the same as those in Table 4, and all
variables are defined in Appendix B. Panel A reports results based on all countries,

RI
whereas Panel B presents those using all countries except the U.S. All regressions
also include country, industry, and year fixed effects, and all associated p-values
reported below parameter estimates are computed based on standard errors adjusted

SC
for heteroskedasticity and firm-level clustering. The sample period is between 1990
and 2010.

NU
Accr(J) Accr(D) Accr(F ) P rob(S) P rob(SEP S)
M1 M2 M3 M4 M5
Panel A: All Countries
MA
Div -0.373 -0.062 -0.063 -0.009 -0.348
(< .01) (< .01) (< .01) (0.09) (0.09)
Issue 2.057 0.562 0.508 -1.118 0.556
(< .01) (< .01) (< .01) (< .01) (0.67)
Div × Issue -0.184 -0.085 -0.070 -0.235 -0.338
(0.02) (< .01) (0.01) (< .01) (0.03)
ED

R̄2 17.0% 81.1% 81.2% 15.5% 14.9%

NObs 118,536 118,536 118,536 118,536 118,536
Controls Yes Yes Yes Yes Yes
PT

Controls × Issue Yes Yes Yes Yes Yes

Country FE Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes
CE

Year FE Yes Yes Yes Yes Yes

Panel B: All Countries excluding the U.S.

Div -0.401 -0.105 -0.094 0.086 -0.575
AC

(< .01) (< .01) (< .01) (0.18) (0.09)

Issue 2.372 0.565 0.524 -0.995 2.118
(< .01) (< .01) (< .01) (0.01) (0.28)
Div × Issue -0.251 -0.040 -0.064 -0.255 0.595
(0.05) (0.09) (0.06) (0.02) (0.30)

R̄2 16.9% 81.1% 80.8% 7.1% 5.9%

NObs 63,396 63,396 63,396 63,396 63,396
Controls Yes Yes Yes Yes Yes
Controls × Issue Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes

38
 Table 8
Dividends-Earnings Management Relation, Country-level Governance, and Subsequent Share Issuance

This table tests the link between dividend policy and earnings management in the context of subsequent equity issuance, separately for seven investor protection and
transparency measures (i.e., legal origin, rule of law, law & order, government effectiveness, report extensiveness, interim report quality, and governance transparency),
partitioned by the strength of each measure into Strong and W eak. The dependent variable is Accr(J), measured in year t + 1 across all models. The key independent
AC
variables are a dividend dummy, Div; a subsequent equity issuance dummy, Issue; and their interaction, Div × Issue. The unreported control characteristics are the same
as those in Table 4, and all variables are defined in Appendix B. Panel A reports results based on all countries, whereas Panel B presents those using all countries except the
U.S. All regressions also include country, industry, and year fixed effects, and all associated p-values reported below parameter estimates are computed based on standard
errors adjusted for heteroskedasticity and firm-level clustering. The sample period is between 1990 and 2010.
CE
Legal Origin Rule of Law Law& Order Gov Effectiveness Report Extensiveness Interim Report Quality Gov Transparency

Weak Strong Weak Strong Weak Strong Weak Strong Weak Strong Weak Strong Weak Strong
PT
Panel A: All Countries

Div -0.412 -0.292 -0.372 -0.334 -0.266 -0.311 -0.324 -0.357 -0.309 -0.314 -0.452 -0.272 -0.263 -0.393
ED
(< .01) (< .01) (0.01) (< .01) (0.06) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01)
Issue 1.542 1.685 2.923 1.899 2.760 1.920 2.478 1.394 1.858 1.618 1.643 2.067 1.574 2.613
(0.01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (0.04) (< .01) (< .01) (< .01)
Div × Issue -0.155 -0.334 -0.516 -0.210 -0.562 -0.176 -0.436 -0.249 -0.412 -0.254 -0.266 -0.227 -0.301 -0.284
(0.33) (< .01) (0.02) (0.02) (< .01) (0.07) (0.01) (0.02) (0.02) (0.02) (< .01) (0.32) (0.01) (0.07)

39
MA
R̄2 17.1% 14.7% 11.4% 17.5% 15.3% 17.6% 17.5% 16.5% 15.7% 14.7% 11.3% 17.6% 14.8% 18.4%
NObs 39,061 59,160 17,154 81,067 27,420 69,726 34,190 64,031 35,273 60,942 18,567 77,648 47,768 46,445
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Controls × Issue Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
NU
Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
SC
Panel B: All Countries excluding the U.S.

Div -0.412 -0.380 -0.372 -0.407 -0.303 -0.423 -0.324 -0.555 -0.318 -0.527 -0.452 -0.355 -0.331 -0.403
(< .01) (0.03) (0.01) (< .01) 0.05 (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (0.07) (< .01)
Issue 1.542 2.327 2.923 2.064 3.415 1.870 2.478 1.399 1.845 2.234 1.643 2.434 1.181 2.630
RI
(0.01) (< .01) (< .01) (< .01) (< .01) (< .01) (< .01) (0.02) (< .01) (< .01) (0.04) (< .01) (0.18) (< .01)
Div × Issue -0.155 -0.370 -0.516 -0.160 -0.689 -0.053 -0.436 -0.081 -0.423 -0.039
P
-0.266 -0.217 -0.340 -0.246
ACCEPTED MANUSCRIPT

(0.33) (0.10) (0.02) (0.29) (< .01) (0.73) (< .01) (0.69) (0.02) (0.86) (< .01) (0.16)
T (0.19) (0.12)

R̄2 17.08% 12.95% 11.4% 18.6% 12.0% 17.8% 17.5% 16.5% 18.9% 13.2% 11.3% 18.5% 12.22% 18.51%
NObs 39,061 24,335 17,154 46,264 15,679 46,642 34,190 29,206 34,440 26,950 18,567 42,823 13,561 45,827
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Ye
Controls × Issue Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
 ACCEPTED MANUSCRIPT

Appendix A
Measures of Earnings Management

1. Jones’ (1991) Abnormal Accruals Model (Accr(J))

We employ the absolute value of residuals from Jones’ abnormal accruals model as an earnings man-
agement measure. In this model, total accruals include changes in working capital and depreciation

T
expenses. Assuming that changes in working capital are a function of revenue growth, and that de-

P
preciation accruals are proportional to long-term tangible assets, one can estimate abnormal accruals
that cannot be explained by revenue growth and fixed assets. Empirically, accruals from Jones (1991)

RI
model are residuals from the following regression model,

TAccri,t = β0 + β1 ∆REVi,t + β2 P P E i,t + ǫi,t , (2)

SC
where TAccr is firm i’s total accruals in year t, ∆REV is its change in revenue, and P P E is its property,
plant and equipment. We estimate Jones’ model using each firm’s cross-sectional data in a given year
to extract its residual, ǫi,t . Generally, U.S. studies estimate Jones’ model by industry and year. We,

NU
however, run the regressions in a given year for each country, while controlling for industry fixed
effects. We do so because in an international setting, due to a limited number of observations across
numerous industries, industry-level regressions may produce unreliable estimates. Positive residuals
indicate income-increasing manipulations, while negative residuals suggest deflated reported earnings.
MA
We employ the absolute value of the residuals, |ǫ|, as the measure of earnings management.
2. Dechow and Dichev’s (2002) Accruals Model (Accr(D))
In Dechow and Dichev’s accruals model, the standard deviation of residuals from the model is employed
as a proxy for earnings management. Dechow and Dichev examine the quality of accruals, or the
extent to which accruals are related to cash flows. They argue that one of the functions of accruals
ED

is to shift or adjust the recognition of cash flows over time, because accruals anticipate future cash
collections/payments and reverse when cash previously recognized in accruals is received or paid.
Therefore, high quality accruals should be closely related to cash flows in years surrounding the
PT

recognition of accruals. Accordingly, Dechow and Dichev model working capital accruals as a function
of current, past, and future cash flows, and use the standard deviation of residuals from the model as
a measure of accruals management. Specifically, their time-series accruals model is given by
CE

WCAi,t = γ0 + γ1 F CF i,t−1 + γ2 F CF i,t + γ3 F CF i,t+1 + ηi,t , (3)

where WCA is firm i’s change in working capital, and F CF is the cash flow from operations. This
model, however, focuses on short-term working capital accruals and does not cover long-term accruals.
AC

3. Francis et al.’s (2005) Accruals Model (Accr(F ))

Francis et al. extend Dechow and Dichev’s (2002) model by adding sales growth to reflect performance,
and adding P P E to capture other accruals including depreciation. This extension results in the
following model,

TAccri,t = λ0 + λ1 F CF i,t−1 + λ2 CF i,t + λ3 F CF i,t+1 + λ4 ∆REVi,t + λ5 P P E i,t + ξi,t (4)

Francis et al. employ the standard deviation of residuals ξi,t to measure the quality of accruals.
Dechow, Ge, and Schrand (2011) argue that time-series regressions assume time-invariant parameter
estimates and induce sample survivorship biases. To circumvent these concerns, we employ the pro-
cedure of Francis et al. by estimating (3) and (4) using cross-sectional regressions for each country.
Following Ng (2011), we calculate the standard deviation of residuals using a five-year rolling window
for both Dechow and Dichev’s (2002) and Francis at al.’s (2005) earnings management metrics.

40
 Appendix B
Variable Definitions

Variable Definition Data Source

Accr(J) Absolute value of residual accruals obtained from Jones’ (1991) model Worldscope
P Accr Income increasing/positive residual accruals obtained from Jones’ (1991) model Worldscope
N Accr Income decreasing/negative residual accruals obtained from Jones’ (1991) model Worldscope
AC
Accr(D) Standard deviation of residual accruals obtained from Dechow and Dichev’s Worldscope
(2002) model
Accr(F ) Standard deviation of residual accruals obtained from Francis et al.’s (2005) model Worldscope
CE
P rob(S) Probability of small profit, measured as a dummy variable that takes the value Worldscope
of 1 if return on assets is between 0 and 1%
PT
P rob(SEP S) Probability of small earnings per share increase, measured as a dummy variable Worldscope
that takes the value of 1 if the ratio (earningst -earningst−1 )/totalassetst−1 is
between 0 and 1%
ED
Div A dummy variable that takes the value of 1 if a firm pays cash dividends in year Worldscope
t
Initiation A dummy variable that takes the value of 1 if a firm does not pay dividends Worldscope

41
in year t − 1, while initiates dividend payments in year t, and continues to pay
MA
dividends in year t + 1
Omission A dummy variable that takes the value of 1 if a firm distributes dividends in year Worldscope
NU
t − 1, but eliminates dividends in year t, and continues to not pay dividends in
year t + 1
Decrease A dummy variable that takes the value of 1 if a firm decreases its dividend pay- Worldscope
SC
ments from year t − 1 to year t, and 0 if otherwise
Icrease A dummy variable that takes the value of 1 if a firm increases its dividend pay- Worldscope
ments from year t − 1 to year t, and 0 if otherwise
Div/P Dividend-price ratio defined as dividends per share scaled by stock price Worldscope
RI
P
ACCEPTED MANUSCRIPT

Div/E Dividend payout ratio computed as dividends per share scaled by earnings Worldscope T
Repur A dummy variable that takes the value of 1 if a firm repurchases its shares in Worldscope
year t
High Div V ol A dummy variable that takes the value of 1 if a dividend payer is above the median Worldscope
dividend volatility for each country and year, computed as standard deviation of
dividends in the previous 8 years, with the minimum of 5 years of data required,
and 0 if the firm is below the median dividend volatility
MV Log of market capitalization in U.S. dollars Datastream
Large Size A dummy variable that takes the value of 1 if a firm is above the median firm Datastream
size (measured by M V ) for dividend payers and for non-payers, for each country
and year, and 0 if the firm is below the median firm size
DE Ratio of total debt to common equity Worldscope
Salesg Annual sales growth rate computed as log of sales in year t scaled by log of sales Worldscope
in year t − 1
 Appendix B - Continued
Variable Definitions
Variable Definition Data Source

Age Number of years since a firm is included in Datastream Datastream

PPE Ratio of fixed assets to total assets Worldscope
ROA Ratio of net income before extraordinary items plus interest expense to total Worldscope
AC
assets
TA Log of total assets Worldscope
IAS A dummy variable that takes the value of 1 if a firm has adopted International Worldscope
CE
Accounting Standards (IAS)
BigN A dummy variable that takes the value of 1 if a firm is audited by any of the Big Worldscope; Compustat
4, Big 5, or Big 8 auditors
PT
Chold Fraction of shares closely held by insiders and controlling shareholders Worldscope
ADR A dummy variable that takes the value of 1 if a firm has issued American De- Depository institutions; U.S. stock exchanges
ED
positary Receipts (ADRs)
BM Log of book-to-market equity ratio Worldscope
σRet Annualized standard deviation of monthly stock returns Datastream

42
CF Operating cash flows, measured as funds from operations scaled by total assets Worldscope
MA
High CF V ol A dummy variable that takes the value of 1 if a firm is above the median cash Worldscope
flow volatility for each country and year, computed as standard deviation of free
NU
cash flows in the previous 8 years, with the minimum of 5 years of data required,
and 0 if the firm is below the median cash flow volatility
Issue A dummy variable that takes the value of 1 if a firm has issued new shares in Worldscope
SC
years t + 1 to t + 3 subsequent to a dividend payment
Legal Origin A dummy variable that takes the value of 1 if a firm is of common law origin, La Porta et al. (1998)
and 0 if a firm is of civil law origin
RI
Rule of Law Strength of a country’s legal system and law enforcement protecting minority La Porta et al. (1998)
P
ACCEPTED MANUSCRIPT

shareholders T
Law & Order An index that comprises ”law” sub-component which assesses the strength and International Country Risk Guide (ICRG)
impartiality of the legal system, and the ”order” sub-component which assesses
popular observance of the law
Gov Ef f ectiveness An index that reflects perceptions of the quality of public and of civil services, the Kaufmann, Kraay, and Mastruzzi (2009)
degree of independence from political pressures, the quality of policy formulation
and implementation, as well as the credibility of the government’s commitment
to such policies
Report Extensiveness An index that captures the extensiveness of financial reporting by examining the Center for Financial Analysis and Research
inclusion or omission of 90 accounting items in annual financial statements in (CIFAR)
1995
Interim Report Quality An index that indicates the frequency and breadth of interim financial disclosure Center for Financial Analysis and Research
(CIFAR)
Gov T ransparency An index that represents the intensity of governance disclosures used by outside Bushman, Piotroski, and Smith (2004)
investors to hold officers and directors accountable
 ACCEPTED MANUSCRIPT

!"#$!"#%&'

T
!"#"$%&$ '()%*+ ,(&(-% %(*&"&-+ .%++ /0(& $"#"$%&$ &1&2'()%*+3

P
!"#"$%&$ '()%*+ ,(&(-% %(*&"&-+ .%++ 40%& /0%) "++5% %65"/) +57+%65%&/.) /1 '()15/+3
80% (71#% 9"&$"&-+ (*% +/*1&-%* "& :15&/*"%+ 4"/0 4%(; "&+/"/5/"1&+ (&$ 0"-0 1'(:"/)3

RI
!"#"$%&$+ ,() 0%.' ,"/"-(/% (-%&:) :1&:%*&+ (&$ 9(:"."/(/% (::%++ /1 %</%*&(. 95&$+3

SC
NU
MA
ED
PT
CE
AC