Professional Documents
Culture Documents
Taxation and Dividend Policy: The Muting Effect of Agency Issues and Shareholder Conflicts
Taxation and Dividend Policy: The Muting Effect of Agency Issues and Shareholder Conflicts
January 2017
ABSTRACT
Using proprietary data on the entire spectrum of ownership-structure and exact
tax-status of investors and firms we examine how dividend taxation affects
payout. Utilizing an exogenous shock to dividend taxation, we show that absent
any frictions, dividend taxation has a large impact on payout. As agency issues
and shareholder conflicts increase, owners’ tax-preferences have significantly
smaller impact on payout. We identify three mechanisms that reduce the dividend-
tax sensitivity: Coordination among owners, heterogeneity in tax preferences, and
diverging objectives between managers and owners. Altogether, taxation has a
first-order impact on payout, but agency issues and shareholder conflicts mute its
impact substantially.
* Jacob is at WHU – Otto Beisheim School of Management (martin.jacob@whu.edu) and Michaely is at the Samuel Curtis
Johnson School of Management of Cornell University (rm34@cornell.edu). We thank participants at the NBER corporate
finance meeting, the Utah Winter Finance Conference, the Washington University Corporate Finance conference, the 2015
FIRS Conference, the Multinational Finance Society, the IDC Summer Finance Conference, the arqus conference, and the
MaTax Conference for many helpful comments. We would also like to thank Francesca Cornelli (the editor), two
anonymous referees, Annette Alstadsæter, Jennifer Blouin, Philip Bond, Tony Cookson, Kathy Kahle, Ambrus Kecskes,
Kristine Hankins, Jochen Hundsdoerfer, Shimon Kogan, Sebastian Michenaud, Maximilian Müller, Francisco Pérez-
González, Alessandro Previtero, Jim Poterba, Ed Rice, Michael Roberts, and seminar participants at the Drexel University,
Rice University, Temple University, University of Arizona, University of Chicago, University of Georgia, University of
Oxford, University of Pittsburgh, University of South Florida, University of Texas Dallas, University of Washington,
University of Western Ontario, and York University. Martin Jacob acknowledges financial support from the Research
Council of Norway.
Electroniccopy
Electronic copyavailable
available at:
at: https://ssrn.com/abstract=2516927
https://ssrn.com/abstract=2516927
1. Introduction
Policymakers frequently use changes in capital taxation to affect dividend payout. For example, when
introducing the 2003 tax cut in the United States, President George W. Bush argued that the double taxation of
corporate dividends “may […] discourage the payment of dividends,” even though dividends “provide a number
of important benefits to investors.” 1 Consistent with the argument in the political debate, theory suggests a large
effect of dividend taxation on payout (e.g., Miller and Modigliani 1961; Chetty and Saez 2010). Recent empirical
evidence on the effect of taxes on dividend payment using the 2003 dividend tax cut suggest that firm payout is
responsive to changes in dividend taxation (e.g., Chetty and Saez 2005; Yagan 2015). Several studies find
evidence of the increased dividend payouts of listed firms, for example, those with high executive stock
ownership (e.g., Blouin, Raedy, and Shackelford 2011; Hanlon and Hoopes 2014). However, the economic
magnitude of the estimated affects does not appear to be very large. For example, for his sample, Yagan (2015,
p. 3560) concludes that the payout “increase was small in dollar terms and may have been irrelevant for real
outcomes.” Other researchers are not fully convinced that the 2003 dividend tax cut had a lasting effect. 2 This
is surprising, since we would expect dividend taxes to play a larger role in firms’ payout decisions. 3
Using a large dividend tax cut, we examine the dividend tax responsiveness and how it is affected by other
frictions, such as agency issues and conflicting objectives among shareholders. Unlike prior studies, we use
comprehensive linked firm–owner data with information on the actual investors’ tax rates for practically every
owner, and information on the exact ownership structure of a large panel of all closely held corporations (CHCs)
in Sweden. Our setting with these small, entrepreneurial firms has three important advantages over prior studies
that enable us to shed new light on the effect of taxes on dividend policy and how this tax effect interacts with
1
Pages 202–203, Chapter 5 of the February 2003 Economic Report of the President.
2
There are some studies casting doubt on a causal effect around the 2003 U.S. tax cut (e.g., Julio and Ikenberry 2004;
Edgerton 2013; Floyd, Li, and Skinner 2015). One argument is that while profits and dividends increased, the dividend-
to-earnings ratio remained effectively constant. Further, share repurchases grew more than dividends during 2002–
2007 even though one would expect a substitution of share repurchases with dividends (Floyd, Li, and Skinner 2015).
Real estate investment trusts increased dividends even though they were not affected by the tax cut (Edgerton 2013).
Survey evidence casts further as managers do not seem to view shareholders’ taxation as an important determinant of
their dividend policy, even if asked specifically about the 2003 tax cut (Brav et al. 2005, 2008).
3
Another example is Chetty and Saez (2005). Their implied elasticity estimate is well below prior estimates (e.g.,
Poterba 1987, 2004).
Electroniccopy
Electronic copyavailable
available at:
at: https://ssrn.com/abstract=2516927
https://ssrn.com/abstract=2516927
other frictions. First, while, by necessity, all prior studies had to rely on proxies for investors’ tax rates, we were
able to obtain, for the first time, complete and exact data on the tax status of each owner and firm in our sample.
This alone allows us to better identify firms that are likely to change their dividends in response to a tax change.
Second, we isolate the effect of taxes on dividend policy in an environment that is almost void of frictions that
might reduce tax sensitivity. Using detailed information on ownership structure and on the marginal tax rate of
each owner, we are then able to control the extent of agency issues, that is, conflicts between owners and
managers (e.g., Jensen 1986, Chetty and Saez 2010), as well as shareholder conflicts, that is, conflicting
objectives among owners. Third, our data enable us to shed light on the dividend tax sensitivity of small, private
firms, an important part of the economy. In fact, small and medium-sized enterprises represent about 95% of all
enterprises, over half of the employment, and over half of the value added in terms of GDP in OECD countries. 4
Our sample firms represent a substantial fraction of the Swedish economy, for example, employing about 30%
of all private sector employees. At the same time, however, using data on small, entrepreneurial firms has the
limitation that we cannot directly draw conclusions for larger, listed firms.
Our first finding is that, absent any frictions, taxes have a large impact on dividend policy. Second, we
show that conflicts of interest between owners and management (e.g., Chetty and Saez 2005) result in the weaker
response of dividends to tax changes. Just as importantly, we find that this is not only friction that interacts with
the effect of taxes on dividends. In addition, shareholder conflicts, such as owner heterogeneity in tax preferences
or coordination issues among owners, explain why dividend tax responsiveness declines as ownership becomes
more dispersed. Our third main finding is that the owners of small, entrepreneurial firms trade off dividend and
labor income taxes in payout decisions. Hence, the tradeoff between labor income taxes and dividend taxes is
relevant not only for executive compensation (Goolsbee 2000), but also for the payout decisions of small,
entrepreneurial firms. Firms pay a smaller (larger) portion of the payout through wages if the owners face lower
(higher) taxes on dividends than on wages. However, the ability to substitute between these forms of income
4
Source: http://siteresources.worldbank.org/CGCSRLP/Resources/SME_statistics.pdf.
The reform contained two changes concerning dividend taxation. First, the combined dividend tax rate was
reduced from 49.6% prior to 2006 to 42.4% after the reform. Second, the dividend allowance—the amount of
dividends that can be paid out at this reduced dividend tax rate, rather than the higher income tax rate—more
than doubled. Taken together, the tax reform increased the incentive for highly taxed shareholders to increase
dividends vis-à-vis wages, since dividends were taxed at a much lower tax rate than wages (51% or higher) in
Using the exact investors’ tax rates, we first investigate how variation in investor taxation affects dividend
payout. We use a difference-in-difference (DD) approach where the first difference compares the dividend
payouts of CHCs before and after the tax reform. The second difference uses differences in tax preferences for
dividends across owners. We categorize each owner as either high tax or low tax. We empirically show that,
relative to owners with lower tax rates on wages than on dividends, the owners of firms with a tax preference
for dividends increased dividend payouts by about 44% following the tax reform. In other words, the relative
taxation of dividend and labor income had a large impact on the payout policies of firms.
However, owners’ tax preferences may not be the only factor that affects the payout response to changes in
dividend taxation. Our setting and data allow us to identify and isolate two such potential frictions. The first
friction is conflicting objectives among shareholders that could arise because of different tax preferences across
owners or because of coordination issues. Such shareholder conflicts could mute the effect of dividend policy to
tax changes and consequently result in lower after-tax cash flow to shareholders. The second friction is agency
conflicts between managers and shareholders (e.g., Jensen 1986; Chetty and Saez 2005), which in turn could
result in lower dividend tax elasticity, distorted investment policy, and lower company value. To test the
potential impact of shareholder conflicts as well as agency issues on the dividend tax sensitivity of payouts, we
employ a research design that additionally exploits the heterogeneity in ownership structure in a difference-in-
difference-in-difference (DDD) approach. To be more precise, we allow the response to the tax reform (first
difference) depending on the tax preference (second difference) to vary across firms that differ in the extent of
agency issues and shareholder conflicts proxied by the number of owners (third difference). We argue that, as
increases, tax preferences are more likely to be heterogeneous among owners and the optimization process that
maximizes the after-tax payout for all owners becomes substantially more complex, introducing coordination
costs. Similarly, agency issues become more severe as controlling and monitoring the manager becomes much
harder. In an environment with more shareholder conflicts or agency issues, we would expect the effect of taxes
to be muted.
Our results from the DDD analysis show that the owners of wholly owned firms, which by construction do
not suffer from shareholder conflicts, increased dividend payouts following the tax reform if they were subject
to higher taxes on labor income than on dividends. Around the reform, these owner–managers increased the ratio
of dividends to total cash to shareholders (dividend compensation ratio) by about 7.3 percentage points, or 73%,
relative to the pre-reform average dividend compensation ratio. This tax sensitivity gradually decreases as the
number of owners and associated frictions increase. For a firm with three owners, owner–managers with the
same tax preference for dividends increased their dividend compensation ratio by only 4.4 percentage points,
40% less than the effect for firms with one owner. This effect is economically and statistically different from the
response of wholly owned firms. Once the firm has five owners and conflicting objectives among owners or
between owners and managers are so acute, the individual owners’ tax preferences have very little impact, if
any, on the firm’s payout policy. We subject this DDD analysis to an extensive set of robustness tests that all
confirm our main findings. First, to ensure that differences in firm size and cash holdings do not drive the results,
we repeat the DDD analysis on a more restricted subsample that is matched on these variables or on a sample of
the largest firms. Second, we run the DDD analysis at the firm level rather than the owner level because,
ultimately, the payout decision is made at the firm level. Third, we verify that our results are not driven by
In the second part of the empirical analysis, we investigate the mechanisms—agency issues versus
shareholder conflicts—that could affect dividend tax sensitivity. The first possible explanation we explore is the
conflicting interests between owners about the form of payout, dividends versus wages. Owner–managers in
corporations with one owner can maximize after-tax profits by easily substituting dividends and wages. The
of owner involvement in the daily operations of the firm make it more difficult to substitute dividends with
wages because different owners receive different wages and could also own different stakes in the firm. Hence,
coordination across owners and, consequently, payout optimization are more difficult in firms with dispersed
ownership than in wholly owned firms. Therefore, one possible reason why diverse-ownership firms do not fully
respond to changes in dividend taxation is the complexity in optimizing the tradeoff between dividends and
substitutability and, thus, the tax responsiveness of dividends and wages. Consistent with this explanation, we
find that firms with only one owner (and obviously with no coordination issues) show economically significant
substitution between dividends and wages. Firms with higher coordination costs experience smaller changes in
their wage compensation ratio, even when tax preferences suggest they should increase dividends and reduce
wages. For example, holding owners’ tax preferences constant, the propensity to substitute wages with dividends
decreases by 39% for firms with three owners relative to wholly owned firms.
A second type of shareholder conflict that could affect dividend tax elasticity relates to heterogeneity in tax
preferences across owners: It is harder to reach a consensus about dividend policy when investors do not agree
about the value of after-tax cash flows. We empirically show that, in firms with greater tax heterogeneity across
owners, the response to the reform is significantly weaker than for similar firms with more homogeneous tax
A different friction that we examine is the effect of agency issues, that is, conflicts between managers and
owners on the tax sensitivity of dividends (e.g., Chetty and Saez 2005; Hanlon and Hoopes 2014). 5 We find that
dividend compensation is less responsive to taxation when there is separation of ownership and control.
5
Similar to our findings concerning the effect of conflicts between managers and owners, a concurrent working paper
by Berzins et al. (2014) uses private firm data from Norway and argues that agency issues moderate dividend tax effects
(see also Chetty and Saez 2005). Our approach and data differs from theirs in several aspects. First, we exploit the
exact ownership structure of each sample firm as well as information on the exact tax status of each owner and variation
in tax preferences for dividends across firms. This unique data enable us to examine how tax preferences shape payout
policy and how heterogeneity in tax preferences affects the payout response to dividend tax changes. Second, we are
also able to explore the role of conflicting objectives among owner, i.e., shareholder conflicts that might reduce the tax
sensitivity of payout. Lastly, our experiment and data enable us to examine substitution across alternative payout
channels and the tradeoff between dividend taxation and labor income taxation.
payouts (e.g., Easterbrook 1984; Jensen 1986; La Porta et al. 2000; Chetty and Saez 2010).
In one of our final tests, we are able to construct a sample where we can simultaneously examine whether
both agency issues and tax heterogeneity (shareholder conflicts) have an effect on dividend tax response. We
find that both tax heterogeneity and agency conflicts decrease dividend tax responsiveness. In the final tests, we
show that other coordination issues among shareholders also seem to play a significant role: While controlling
for the separation of ownership and control, we show that firms with only a few owners responded more strongly
to the reform than firms with more dispersed ownership, even when tax preferences for dividends are the same
across all owners. 6 Altogether, these findings suggest that shareholder conflicts as well as agency issues
contributes to the declining dividend tax elasticity decreases as the number of owners increases.
This paper’s findings are important to policymakers who frequently use dividend taxation to affect
corporate payout policies. The mechanisms we uncover suggest when and why tax policy could be more
effective. Dividend taxation appears to be an effective tool to affect payout policy when agency conflicts and
shareholder conflicts are small. Our findings further suggest that theoretical and empirical work examining the
impact of taxation on payout policy might be incomplete without considering the interaction of other frictions
Finally, this paper also contributes to a growing literature on a very important part of the world economy,
namely, private firms. 7 One important implication of our study is that not only agency issues but also shareholder
conflicts can affect the shareholder value of these small, entrepreneurial firms. Shareholder conflicts among
owners could mute firms’ payout response to a tax reform. Owners then forgo potential increases in after-tax
income if firms do not change their payout policy according to the individual tax preferences of each owner.
These forgone increases in disposable income—which can be over 1% of pre-tax income per year in our
6
We find the similar support for the role of coordination issues when using the tax preference of the dominating
shareholder instead of using homogeneous tax preferences for dividends among all owners.
7
See, for example, Djankov et al. (2002), Haltiwanger, Jarmin, and Miranda (2013), Adelino, Schoar, and Severino
(2015), or Yagan (2015). Prior research uses data from the U.S. (Petersen and Rajan 1994, 2002; Schoar 2002;
Maksimovic and Phillips 2008; Chemmanur, Krishnan, and Nandy 2011) or from countries such as France (Bertrand,
Schoar, and Thesmar 2007), Denmark (Bennedsen et al. 2007), and the United Kingdom (Michaely and Roberts 2012).
results from a sample of private firms that might not be comparable to large, publicly listed firms. Still, the
economic mechanisms we document—shareholder conflicts and agency issues as well as income shifting
We focus on Sweden because it is a country with access to detailed tax data (see Section 3) and with well-
developed institutions. Sweden’s institutional structure in terms of corporate governance and the general tax
structure is comparable to that of many developed countries. For example, Sweden is comparable with respect
to private enforcement and shareholder rights to Australia, Canada, Finland, France, Germany, Israel, New
Zealand, and the United States (see, e.g., the anti-self-dealing index of Djankov et al. 2008). Sweden’s creditor
rights are similar to those of the United States, Brazil, Canada, Finland, Ireland, and Switzerland (Djankov,
McLiesh, and Shleifer 2007). Furthermore, Sweden has a rule of law comparable to countries such as Canada,
Germany, Switzerland, Singapore, the United Kingdom, and the United States. Relevant to our study on the
importance of dividend taxation, Sweden has a classical corporate tax system with a double taxation of
dividends, with corporate tax and dividend taxation, as many other countries (e.g., Jacob and Jacob 2013).
About 315,000 corporations are registered in Sweden that are subject to corporate taxes. A non-listed
corporation is closely held for tax purposes if four or fewer shareholders possess at least 50% of the voting rights
and if at least one shareholder is active in the operation of the firm. 8 According to the tax law, a shareholder is
active if contributing to the firm’s profit generation. CHCs play an important role in the Swedish economy. As
of 2009, 64% of corporations are closely held. During our sample period, CHCs generated over 24% of the
operating profits and employed 29% of all employees in the entire private sector—corporate and non-corporate
businesses—in Sweden (see Table 1). Our sample firms thus cover an important part of the Swedish economy.
8
CHCs can have five (or more) active owners in case family members are active owners. Specifically, if five family
members hold more than 50% of the voting rights and if at least one of the family members is active in the profit
generation, then the corporation is a CHC according to the tax law.
2006 tax reform, dividends were taxed at a proportional tax of 30% at the individual level, resulting in a
combined dividend tax burden of 49.6% (= 28% + (1 – 28%) × 30%). Since the dividend tax burden of 49.6%
can be below the marginal tax rate on wages (which can be as high as 67%, see Figure 1), owners could exploit
this tax wedge and choose the least taxed payout channel. To limit this type of tax minimization in CHCs, the
tax law defines an upper cap of dividends—the imputed dividend allowance—that can be taxed as dividends.
The dividend allowance is calculated per firm, based on wages to employees and on equity. 9 Each owner receives
a percentage of this allowance according to the ownership share. Any dividends in excess of the dividend
allowance are taxed at the progressive income tax rate on wages and not at the dividend tax rate of 30%. If
dividend payout is below the dividend allowance, the unused allowance can be carried forward indefinitely with
interest. Prior to the reform, only 1.6% of the firms had no dividend allowance left. For wholly owned firms,
this number was 1.1%, increasing to 1.7% (2.8%) for firms with two (three) owners and reaching 3.8% (5.2%)
In January 2005, an expert committee presented a report recommending changes to dividend taxation. The
proposals were, however, not immediately implemented by policymakers and it was uncertain if the proposed
changes would be implemented at all. Only in late 2005, a last-minute tax relief of SEK 1 billion was announced
to be implemented by January 1, 2006. Following the recommendations by the expert committee, the Swedish
government decided to cut the dividend tax rate with the intention of generally spurring entrepreneurship and
investment. Since this dividend tax reduction was announced late in 2005, it was not anticipated by firms and
owners during 2005. The reform was implemented permanently and, due to Sweden’s history of a stable tax
environment, it is reasonable to assume that corporations and their owners perceived the reform as permanent. 11
9
For example, in 2004, the dividend allowance per firm was equal to 11.7% of nominal equity plus 10% of wages paid
to employees (excluding the owners’ wages). The imputed interest on nominal equity results from the interest rate on
government bonds plus a risk premium of seven percentage points (five percentage points before 2004).
10
We also test that this difference does not explain our main findings. The main results replicated without individuals
(and firms) who fully utilized the dividend allowance are very similar to our main results.
11
There were also other policy changes around 2006 (“The Swedish Reform Programme for Growth and Employment,
2005–2008,” http://www.regeringen.se/contentassets/b80397b337f544988a55e9d3714447cb/the-swedish-reform-
programme-for-growth-and-employment-2005-2008). These changes relate to environmental expenditures,
investments in sustainable transport, climate programs, changes to competition legislation, or adjustments to youth
dividend tax rate for CHCs was cut from 30% to 20%. This reduced the combined tax burden on dividends
within the dividend allowance to 42.4%. Dividends in excess of the dividend allowance are still taxed at the
progressive income tax rate on wages and are thus subject to a combined tax burden of up to 68.8%. Table 2
summarizes the dividend tax rates before and after the 2006 tax reform. Second, for the vast majority of firms,
the dividend allowance increased substantially from before to after the reform and, depending on the asset
structure and wages paid to employees, the increase could even be 10 or more times higher. 12 Figure 2 presents
the average accumulated dividend allowances per owner before and after the reform. We find that the dividend
allowance per owner substantially increases for all groups. We also split the sample into CHCs with one to five
owners. Most importantly, the average increase in dividend allowance and the average dividend allowance per
owner are similar across groups. Taken together, as the reform substantially increased the amount of dividends
that can be paid out at the reduced rate of 42.4%, highly taxed owner–managers have been able to pay out
significantly more dividends since the 2006 reform at the reduced tax rate.
In addition to dividends, owner–managers can pay out cash in the form of wages. Wages are tax deductible
at the corporate level but subject to income taxation at the shareholder level. Figure 1 presents the progressive
labor income tax schedule for 2002 and 2008. The combined marginal tax burden on labor income ranges from
31.4% to 67.1% (2008 values). 13 There is one large and important kink in the tax schedule at which the marginal
income tax rate increases by 20 percentage points, from about 31% to 51%. 14 The comparison of the tax rates
on dividends and wages shows that if an owner’s marginal income tax rate is to the right of the kink, that is, 51%
unemployment support. Another tax-related change was the repeal of the inheritance and gift tax in 2005. This reform
was already announced in 2004 and was not part of the tax bill that we analyze. In addition, the parallel trends over
2002–2005 in Figures 3 and 4 indicate no effect of the 2005 inheritance tax abolishment on payout.
12
The 2006 reform increased both the wage-based allowance by more than 100% and capital-based allowance by two
percentage points, or about 17% (using the 2004 value). The 2006 reform also introduced an optional fixed dividend
allowance that amounted to SEK 64,950 in 2006 and that increased to SEK 120,000 in 2009. See Alstadsæter and Jacob
(2012) for more detailed calculations.
13
In our calculations, we account for contributions for health insurance, unemployment insurance, and a defined
contribution pension plan at the corporate level. We neglect the standard deduction and the earned income tax credit,
which was introduced in 2006, since these did not alter the relative taxation of wages and dividends.
14
The kink was at SEK 290,100 (about USD 39,416) in 2002 and increased to SEK 380,200 (about USD 51,658) in 2009.
The kink increases every year (Alstadsæter and Jacob 2012, Table AI.2).
tax rate is to the left of the kink (31% or lower), the owner prefers wages over dividends and should not respond
3.1 Data
We base our study on the Firm Register and Individual Database (FRIDA), provided by Statistics Sweden.
The data contain the corporate tax returns of all Swedish corporations over the period 2000–2009. Corporate tax
returns comprise the tax balance sheet and the profit and loss statement. The individual data cover income and
tax variables, along with demographic characteristics such as age, gender, and education. Through unique
identifiers, we link information on firms and their shareholders to obtain the ownership structure.
Administrative tax data, such as FRIDA, have the advantage that all firms are required to file a corporate
tax statement. Studies based on accounting data face the problem that the data coverage of private firms is very
poor due to the lack of filing requirements for small enterprises (e.g., Michaely and Roberts 2012). While our
data have comprehensive tax coverage, we need to exclude public firms and private firms that are not classified
as closely held, since we do not have data on ownership or payout of these firms. Since we analyze the effect of
taxation on payout decisions, we require information on dividends to owners, along with their tax status.
Therefore, we focus on unlisted CHCs. Our data contain the number of active owners and their ownership share.
Some firms could have additional passive owners. However, on average, approximately 90% of shares in CHCs
are held by active owners, indicating that ownership by passive owners plays a less important role. This share is
also similar across firms that differ in the number of active owners. For example, in firms with one active owner,
88% of shares are held by the active owner (2007 values). In firms with five active owners, a total of 86% of the
shares are held by the active owners. We therefore use the terms owner and active owners synonymously
throughout the paper, since the vast majority (over 85%) of firms are fully owned by active owners. Further,
there are some additional limitations of our data: We have no information on the firm’s age or foreign operations,
and, finally, we have no information on the family relations of the owners in our sample.
10
demographic characteristics, income composition, and, most importantly for our study, payout information. Each
active shareholder of a CHC is required to file a tax return that declares the dividend and wage income from
(each of) his or her CHC(s), as well as income from all other sources. We winsorize shareholder-level and firm-
level variables at the 1% and 99% levels to control for outliers and require firms to exist for at least three
consecutive years. 15 A potential concern about the analysis is that some of the CHCs are used only for tax
avoidance purposes and hence it would be difficult to extrapolate from their behavior to other firms. Such tax
shelter firms typically have one owner (e.g., Alstadsæter and Jacob 2012). We thus exclude such passive firms
because their inclusion could bias our results. 16 Importantly, our results are not sensitive to this sample
restriction. In Part III of the Online Appendix, we replicate all results, including passive firms, and show that all
results hold when we include these firms. Our final sample contains 1,172,617 observations for the owner–
Table 2 presents summary statistics for payout variables as well as firm-level and shareholder-level
characteristics. We use the following payout variables, which are measured at the shareholder level: Dividend
CHCi,t is the total amount of dividends (in Swedish kronor, or SEK) received by owner i from the CHC(s) of the
owner in year t. Similarly, Wages CHCi,t is the amount of wages received (in SEK) by owner i from the CHC(s).
The variable %Div Compensationi,t (%Wage Compensationi,t) is the percentage of total compensation from the
CHC(s), Dividend plus Wages paid out as dividends (wages). If total compensation is zero, we set %Div
Compensation and %Wage Compensation to zero. Panel A of Table 3 summarizes our payout variables. On
15
This restriction allows firms to exit and enter the sample around the reform. While this addresses potential survivorship
bias, the reform could induce firms to change their status to a CHC. We test the robustness of our results and require
firms to exist during at least two years before the reform and during two years after the reform. Our main results remain
unchanged when we use this alternative sample restriction. Table A.I of the Online Appendix (available at
http://www.whu.edu/fileadmin/data-facct/JM_Online_Appendix.pdf) reports the coefficient estimates.
16
Specifically, we exclude income-shifting firms, such as holding companies (firms with 90% or more of their turnover
arising from interest, dividends, and capital gains), low-turnover companies (firms with less than SEK 200,000 in sales
and turnover), and corporations that never pay dividends or wages in any sample year. See Alstadsæter and Jacob
(2012) for a more detailed discussion on passive firms as well as tax shelter firms.
11
70% of total compensation from the CHC is in the form of wages, while 16% is in the form of dividends. 18
We measure ownership structure, growth opportunities, internal funds, profitability, capital structure, and
size for each firm j in year t. Panel B of Table 3 summarizes our main firm-level variables. As discussed above,
we use the number of owners (Ownersj,t) in the firm as our measure of agency issues and shareholder conflicts.
Since Tobin’s q is not observable due to lack of market prices, we proxy for growth opportunities through the
percentage change in fixed assets from t - 1 to t (Investmentj,t). As measures of internal funds, we include
Cash_Assetsj,t, defined as cash holdings over total assets, and RE_Assetsj,t, defined as the ratio of internally
generated equity (retained earnings) to the prior year’s total assets (DeAngelo, DeAngelo, and Stulz 2006). We
use taxable corporate income over total assets as the measure for profitability (Profit_Assetsj,t). We control for
capital structure by including total book debt over total assets (Leveragej,t). We use the natural logarithm of total
assets (Ln(Total Assets)j,t) and the number of employees (Num Employeesj,t) as measures of size. On average, a
CHC has two active owners, holds 26% of its assets as cash, has a ratio of retained earnings (debt) to assets of
about 24% (63%), and has about six employees. The investment ratio varies considerably, from 0% (25th
percentile) to 64% (75th percentile). Furthermore, we present aggregate statistics on the importance of each
number of owners bin in Table 1. Firms with one (two) owner(s) account for about 47% (33%) of employment,
wages, and operating profits among the CHCs in our sample. Firms with three (four and five) owners account
for about 13% (10%) of employment, wages, and operating profits. Hence, a substantial part of economic activity
average, 29% of their total income from sources outside the CHC(s) (Relevance Non-CHCi,t). We use a dummy
variable for the tax status of owner–managers (High Taxi,t-1) that we set equal to one if the owner’s marginal tax
17
In 2006, SEK 7.36 equaled USD 1.
1 𝐷𝐷𝐷𝐷𝐷𝐷
18
The average portion of dividends is calculated as ∑ , unless the owner did not receive any compensation in
𝑛𝑛 (𝐷𝐷𝐷𝐷𝐷𝐷+𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊)
𝐷𝐷𝐷𝐷𝐷𝐷
year t. In that case, is set to zero. Since these statistics include owners who did not receive any form of
(𝐷𝐷𝐷𝐷𝐷𝐷+𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊)
compensation from their CHC(s) in one year, the share of wages and dividends do not add to one.
12
on dividends. Our summary statistics for High Tax i,t-1 in Table 3 indicate that 47% of owner–managers are
subject to a marginal income tax rate of over 51%. There is little variation across number of owners bins. In
wholly owned firms (firms with two owners), about 48% (45%) of owner–managers are subject to a marginal
income tax rate of over 51% (see Table A.X of the Online Appendix). This fraction is 49.6% (49.2%) for owner–
managers in firms with three (five) owners. The variable Number of Firmsit is the number of CHCs in which an
owner actively participates. The majority of owners participate in exactly one firm. On average, owner–
managers hold shares in 1.12 CHCs. In Part II of our Online Appendix, we present descriptive statistics on firm-
level and shareholder-level characteristics for each number of owners bin that ranges from one to five owners.
To test how taxation affects dividend payout, we exploit differences in shareholder tax preference for
dividend income. An owner in a high tax bracket (51% or higher) has an incentive to shift more of the payout to
dividends that are taxed at 42.4% after the reform. An owner in a low tax bracket who is subject to a 31% tax
rate does not have an incentive to shift. We thus split the sample into highly taxed owners (High Tax = 1) and
owners subject to a low income tax on wages (High Tax = 0) and track the difference in the dividend
compensation ratio between these two groups over time. If taxation affects dividend payout, we would observe
an increase in the difference between high- and low-tax owners around the tax reform. Figure 3 plots this
difference for the full sample of owners and the DD estimate without control variables. We use the tax status
from the prior year to avoid the current payout (wages and dividends) affecting the tax rate.
Figure 3 shows that, prior to 2006, the difference in dividend compensation between owners with a marginal
income tax rate above 51% and lower-taxed owners was about 5.5 percentage points and constant over time.
After the reform, the difference doubled to about 11.5 percentage points and remains at this high level. The
observations in Figure 3 have two main implications. First, there is a common trend before and after the reform
in dividend compensation between the two groups. Second, the dividend tax cut increases the difference in
dividend compensation in accordance with the tax preferences for dividends of highly taxed owners. The
13
response to the tax reform using a DD approach with control variables and estimate
where the dependent variable is %Div Compensation for shareholder i in year t. We use the 2006 tax reform as
an exogenous event (Post). Since the reform decreased dividend taxes and increased the amount of dividends
for which the reduced dividend tax rate is available, the reform provides incentives to increase dividend payout
if dividends are taxed at a lower rate than wages. This is the case when individuals are subject to a marginal
income tax rate of 51% or higher (High Tax = 1). We thus expect 𝛼𝛼1 > 0.
Π𝑗𝑗,𝑡𝑡 is a vector of firm-level variables that includes six control variables. First, we control for the availability
of internal resources for distribution to shareholders. We include cash holdings over total assets (Cash_Assets)
and operating income relative to total assets (Profit_Assets) as controls. Second, we include the ratio of retained
earnings to assets (RE_Assets) to control for the life-cycle model of dividend payouts. Third, we control for
growth opportunities through change in fixed assets (Investment). Fourth, we include leverage to control for the
effect of creditors on payout policies. Creditors could have an influence on payout policies because, for example,
debt covenants could restrict dividend payouts to protect creditor rights. Fifth, we control for firm size (Ln(Total
Assets) and Num Employees), since larger firms tend to have higher dividend payouts.
We include a vector of shareholder-level variables, χ𝑖𝑖,𝑡𝑡 . Specifically, we include the percentage of total
income derived from income sources other than the CHC(s) (Relevance Non-CHC). If owners generate
substantial income from sources other than the CHC(s), compensation from the CHC(s) becomes less important
and owners could be less tax sensitive. Previous literature further shows that the elasticity of taxable income
differs, for example, across age cohorts, married and unmarried individuals, or with respect to the level of
education (e.g., Chetty et al. 2011). We thus include the demographic characteristics Age, marital status
(Married), and High Education as control variables. We include shareholder fixed effects (𝛼𝛼𝑖𝑖 ) and year fixed
effects (𝛼𝛼𝑡𝑡 ). We measure all variables at the shareholder level. That is, firm-level variables are linked to the
14
firms for that owner. 19 We base our statistical inference on robust standard errors clustered at the owner level.
Clustering at the owner level addresses concerns that our dependent variable, dividend payout, is correlated over
Table 4, Column (1) presents the coefficient estimates for 𝛼𝛼1 , or the interactive effect on dividend payout
from the post-reform time period and a high tax bracket owner (Post×High Tax). Our results show that investor-
level taxation has a strong influence on dividend payout. The owners of CHCs with tax preferences for dividends
increase the percentage of compensation paid as dividends by about 5.9 percentage points. The 5.9 percentage
point increase in the dividend compensation ratio represents a 44% increase relative to the pre-reform sample
average of a dividend compensation ratio of 13.4%. Further, the DD coefficient estimate with controls is very
close to the DD estimate without controls from Figure 3. In Column (2), we estimate the payout response at the
firm level rather than the owner level, using the ratio of dividends to total assets for each firm as the dependent
variable. We now use the dummy variable Avg High Tax to operationalize tax preferences at the firm level. We
set Avg High Tax equal to one if the average tax rate on the wages of all owners (weighted by the ownership
share) is above 50%. In other words, Avg High Tax equals one if, on average, owners prefer dividends to wages.
The positive coefficient of 0.011 indicates that firms whose owners, on average, prefer dividends increase the
dividend-to-asset ratio by 1.1 percentage points, or 42% of the sample average dividend-to-assets ratio of 2.6%.
To summarize, the regression results, using either the firm-level or owner-level approach, paint a picture similar
to that in Figure 3: Changes in investors’ level of taxation have a first-order effect on dividend payouts and the
4.2 Tax Sensitivity and Agency & Shareholder Conflicts: Graphical Evidence
In firms with more dispersed ownership, owners may be less likely to adjust dividend payouts according to
their own tax preferences because of conflicts among owners or conflicting objectives between managers and
19
As a robustness test, we re-estimate our baseline results from Table 5 using the 90% of owners who own exactly one
CHC. We do not have to use aggregated firm characteristics across firms for these owners. The results are reported in
the Online Appendix, Table A.II, and are very similar to our baseline results.
15
we take advantage of the detailed ownership information in our data. We replicate Figure 3 for different number
of owners bins, our proxy for differences in conflicts among owners or conflicting objectives between managers
and owners. We split the sample into groups according to the number of owners, ranging from one to five, to
test how agency issues and shareholder conflicts affect dividend tax elasticity. Panel A of Figure 4 uses owners
in wholly owned firms. This setting is almost void of any other frictions. As for the full sample, we observe a
constant difference between high- and low-tax owners prior to the reform and a sharp increase in the difference
after the reform. The corresponding DD estimate amounts to 0.071. Panels B and C repeat this analysis for
owners in firms with two and three owners, respectively. We again observe a pattern similar to that of firms with
one owner, but the DD estimates are smaller (0.059 and 0.045, respectively) once there are more potential
conflicts agency issues and conflicts among shareholders. The decrease in the dividend payout response to the
2006 tax cut is even more apparent when moving to owners of firms with four or five owners. While the DD
estimate for wholly owned firms is 0.071, the DD estimate for owners in firms with four (five) owners decreases
to 0.018 (0.034), less than a third (half) of the effect relative to wholly owned firms.
The graphical evidence in Figure 4 seems to suggest that the dividend tax responsiveness is declining as
conflicting objectives among owners or between owners and managers become more acute. In Figure 5, we thus
plot the unconditional triple difference: The difference between high- and low-tax owners (first difference) and
between owners of wholly owned firms to owners of firms with multiple owners (second difference) over time.
The comparison before the reform to after it results in the third difference. The triple difference (DDD) estimate
quantifies the decline in dividend tax responsiveness as ownership becomes more dispersed. Panel A of Figure
5 compares owners of wholly owned firms to owners in closely held firms with two owners. We observe a
parallel trend prior to 2006 and an increase in the DD after the reform. That is, the reform response by owners
of wholly owned firms is more pronounced than the response of owners in CHCs with two owners. The
corresponding DDD estimate is 0.012 and is statistically significant. As the number of owners and thus agency
issues or shareholder conflicts increase, the DDD estimate becomes larger. For example, the DDD estimate
increases to 0.025 (0.036) when we contrast wholly owned firms to firms with three (five) owners. In each panel,
16
Overall, the positive and significant triple difference estimates indicate that an increase in agency issues and
shareholder conflicts mutes tax responsiveness. The owners of very CHCs appear to respond more strongly to
the 2006 tax reform than owners of CHCs with more dispersed ownership, even if they have a preference for
dividends.
4.3 Tax Sensitivity, Agency Issues, and Shareholder Conflicts: Tax Reform Analysis
We next formally analyze how the tax responsiveness of dividend compensation varies with the number of
owners while controlling for owner-level as well as firm-level characteristics. Our empirical strategy uses a
DDD approach, as illustrated in Figures 4 and 5. The first difference compares the dividend payouts of CHCs
before and after the tax reform. The second difference compares differences in tax preferences for dividends
across owners, which we operationalize by the dummy variable High Tax. The third difference exploits
heterogeneity in ownership structure, our proxy for agency issues and shareholder conflicts. We thus estimate:
%𝐷𝐷𝐷𝐷𝐷𝐷 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖,𝑡𝑡 = 𝛼𝛼0 + 𝛼𝛼𝑖𝑖 + 𝛼𝛼𝑡𝑡 + 𝛼𝛼𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑟𝑟𝑟𝑟𝑗𝑗,𝑡𝑡 + 𝛽𝛽Π𝑗𝑗,𝑡𝑡 + 𝛾𝛾χ𝑖𝑖,𝑡𝑡 (2)
+𝛼𝛼1 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐻𝐻𝐻𝐻𝐻𝐻ℎ 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑟𝑟𝑟𝑟 = 1𝑗𝑗,𝑡𝑡 + 𝛼𝛼2 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐻𝐻𝐻𝐻𝐻𝐻ℎ 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑟𝑟𝑟𝑟 = 2𝑗𝑗,𝑡𝑡
+𝛼𝛼3 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐻𝐻𝐻𝐻𝐻𝐻ℎ 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑟𝑟𝑟𝑟 = 3𝑗𝑗,𝑡𝑡 + 𝛼𝛼4 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐻𝐻𝐻𝐻𝐻𝐻ℎ 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑟𝑟𝑟𝑟 = 4𝑗𝑗,𝑡𝑡
+𝛼𝛼5 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐻𝐻𝐻𝐻𝐻𝐻ℎ 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑟𝑟𝑟𝑟 = 5𝑗𝑗,𝑡𝑡 + 𝛿𝛿𝐻𝐻𝐻𝐻𝐻𝐻ℎ 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑟𝑟𝑟𝑟𝑗𝑗,𝑡𝑡 + 𝜀𝜀𝑖𝑖,𝑡𝑡
where the dependent variable is %Div Compensation for shareholder i in year t. We again use the 2006 tax
reform as an exogenous event. To test the hypothesis that conflicting objectives among shareholders and between
owners and managers affects tax sensitivity, we estimate the tax response coefficient (Post×High Tax) separately
for each number of owners bin that ranges from one owner to five owners (Owners). 20 We thus allow the
response to the tax reform (Post) for owners with tax preferences for dividends (High Tax) to vary across the
Owners bins. If agency issues and shareholder conflicts affect tax responsiveness, we would observe declining
20
As indicated by the subscript t, the number of owners varies over time. One potential concern is that the reform could
affect the number of active owners in a firm. We therefore run two additional robustness tests. First, we include only
firms for which the number of owners did not change over the 2004–2007 period. Second, we fix the number of owners
at the 2005 level. Table A.III of the Online Appendix presents the regression results for α1 to α5. The results are similar
to our baseline results from estimating equation (2).
17
Table 5 presents the coefficient estimates for 𝛼𝛼1 to 𝛼𝛼5 . The owners of very closely held firms with tax
preferences for dividends increase the percentage of compensation paid as dividends by about 7.3 percentage
points (one-owner CHCs). These are economically large effects. The 7.3 percentage point increase in the
dividend compensation ratio for owners of wholly owned firms with a tax preference for dividends represents a
69% increase relative to the pre-reform sample average dividend compensation ratio of 10.6% (see Panel B of
Table A.IV of the Online Appendix). This tax sensitivity gradually decreases as agency issues and shareholder
conflicts are more likely to be present. Owners in firms with three owners increase the dividend compensation
ratio by only 4.4 percentage points. That is, tax sensitivity decreases by about 40% (from a 7.3 percentage point
increase to a 4.4 percentage point increase) if the number of owners increases from one to three. The difference
in effects is therefore not only statistically significant (t-stat = 4.78) but also economically significant. Once a
firm has five owners, the individual owners’ tax preferences do not have a statistically significant influence on
the dividend compensation ratio. In other words, in firms with more dispersed ownership where agency issues
and shareholder conflicts are likely to be more acute, the individual owner is not able to adjust the dividend
compensation ratio around the 2006 tax reform according to his or her own tax preferences.
Our DDD approach has two potential concerns. The first issue is that some of our controls variables and
their interaction with dividend payout may be affected by the 2006 tax reform such as the ratio of retained
earnings to assets. To this end, we alternatively use first differences instead of levels of our control variables to
measure for their growth. Results from re-estimating equation (2) using first differences instead of levels in
controls are reported in Column (2) of Table 5. The magnitude of the estimates is very close to our baseline
estimates from Column (1) and supports our main findings. In addition, we run a fully-interacted model, where
we allow interact each control variable with the Post dummy, the number of owners indicator variables as well
as the interactions of High Tax and number of owners. This addresses concerns that the effect of observable
firm- and owner-level variables can potentially vary from before to after the reform and across treatment groups.
Again, the results in Column (3) of Table 5 are consistent with our baseline estimate.
18
shareholder preferences. As long as such differences are constant over time, the inclusion of shareholder fixed
effects in equation (2) is equivalent to matching on observable and unobservable characteristics. Moreover, our
graphical analysis in Figure 3 supports the common trend assumption. To address remaining concerns that
differences in size and cash holdings across firms could explain the observed differences in dividend tax
elasticity, we construct two additional sub-samples. In the first subsample (see Column (4) of Table 5), we
restrict the sample to firms with at least 10 employees (the top decile). 21 We use propensity score matching to
construct the second subsample. Specifically, we run a matching DDD approach based on the pre-reform
characteristics size (Ln(Total Assets) and Num Employees) and cash holdings (Cash_Assets). Since the owners
of firms with five owners are the smallest group in terms of number of observations, we individually match them
to owners of firms with one, two, three, and four owners, respectively. 22 After these four one-to-one matching
procedures, there are no differences in size and cash holdings across firms with different ownership structures
(see Part II of the Online Appendix). We then use this sample and rerun our fully interacted DDD model from
equation (2). Column (5), Table 5 presents the tax response coefficients (Post×High Tax) for the matched
sample. Consistent with our main results, we obtain a significant response for the owners of wholly owned firms
in both sub-samples. The response to the 2006 tax reform again declines in the number of owners for the sample
of firms with at least 10 employees and for the matched sample. These results further suggest that our results are
not driven by differences in size and cash holdings across firms. In addition, in Table A.VI of the Online
Appendix, we document that the effect of agency issues and shareholder conflicts on tax responsiveness is
present across the firm size distribution. We find similar results in all quartiles of the total assets as well as the
number of employees distribution. Taken together, the consistency in results and economic magnitudes across
21
Part II of our Online Appendix presents descriptive statistics for firm-level and individual-level variables. Using the
sample of firms with at least 10 employees substantially reduces differences in size across number of owners bins.
22
We run the matching in 2005 and also match on two lags of Ln(Total Assets), Num Employees, and Cash_Assets,
Alternatively, we use pairwise matching and compare the owners of wholly owned firms to the owners of firms with
multiple owners. Table A.V of the Online Appendix presents the coefficient estimates for four DDD regressions using
the matched samples. The results are similar when using this alternative matching approach.
19
Third, to address concerns that measuring payout responses at the owner level explains our findings, we
run the analysis at the firm level instead of using owner-level data. Similar to the approach based on the owner
level, we estimate
𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 − 𝑡𝑡𝑡𝑡 − 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑗𝑗,𝑡𝑡 = 𝛼𝛼0 + 𝛼𝛼𝑗𝑗 + 𝛼𝛼𝑡𝑡 + 𝛼𝛼𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑟𝑟𝑟𝑟𝑗𝑗,𝑡𝑡 + 𝛽𝛽Π𝑗𝑗,𝑡𝑡 (3)
where Dividend-to-Assetsj,t is the ratio of dividends to total assets of firm j in year t. 23 We use the dummy variable
Avg High Tax to operationalize tax preferences at the firm level. Again, we allow the DD coefficient Avg High
Tax × Post as well as the main coefficient Avg High Tax to vary across the number of owners bins (𝛼𝛼1 to 𝛼𝛼5 and
the vector 𝛾𝛾). We further include firm-level control variables (Π𝑗𝑗,𝑡𝑡 ), firm fixed effects (𝛼𝛼𝑗𝑗 ), and year fixed effects
Figure 6 illustrates the resulting 𝛼𝛼1 to 𝛼𝛼5 coefficients with the 95% upper and lower confidence intervals.
We normalize the estimated coefficients by the unconditional pre-reform average to obtain the relative change
in the dividend-to-asset ratio. That is, we obtain the percentage change in the dividend-to-asset ratio from before
to after the reform. The firm-level analysis confirms our results from the owner-level analysis. The results in
Figure 6 show that the effect of the dividend tax cut is strongest in wholly owned firms and then gradually
declines in the number of owners, i.e., when agency issues and shareholder conflicts are more acute. For
23
To address concerns that the scaling variable, total assets, is affected by the tax reform, we scale dividend payout by
the total assets of year 2005. The results are very similar and are reported in Figure A.I of the Online Appendix.
20
lower than in wholly owned firms. The results in Tables 5 and 6 and Figure 6 show that the dividend tax elasticity
We implement four alternative procedures in our robustness tests. First, we use alternative dividend
measures to address concerns that our dependent variable %Div Compensation only relates to the choice of the
payout channel (dividend versus wages). We use a dummy variable equal to one if the owner receives a dividend
and zero otherwise (Dividend Payer). This variable captures the fraction of owners receiving dividends. Table
6 presents the regression results from estimating equation (2) using this alternative dependent variable. We again
obtain a significant response for owners of wholly owned firms as well as for owners of CHCs with two owners.
The response is significantly smaller as ownership becomes more dispersed. In other words, the fraction of
highly taxed owners seeing an increase in dividends is significantly higher in very closely held firms. Table 6
also uses a different scaling variable for dividend payout to additionally address potential concerns that scaling
by total cash to shareholders only relates to the choice of the payout channel. Instead of using total cash to
shareholders, we use the owner’s total income to normalize dividends. Our results are again in line with our
previous findings. Following the dividend tax cut, highly taxed owners increase dividends relative to lower-
taxed individuals but this response decreases as the number of owners increases.
Second, another possible concern is that the magnitude of the tax rate difference between dividends and
wage payouts could differ across number of owners bins. This could explain the observed differences in tax
responsiveness to the reform, since our main tax measure High Tax does not account for the magnitude of the
tax wedge. To this end, we test whether our main result holds if we separate effects for the top tax bracket
(income tax rate of about 67%) as well as for the middle tax bracket (income tax rate of 51–67%). In our DDD
model from equation (2), we include an additional dummy variable, Top Tax Bracket, and its interactions with
Post and each bin of 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑟𝑟𝑟𝑟𝑖𝑖,𝑡𝑡 . We thus separate the tax response of individuals in the top tax bracket (Top Tax
Bracket = 1) from highly taxed owners who are not in the top tax bracket (High Tax = 1). In both tax brackets
(see Table 7), we find that dividend tax elasticity is declining as there are more owners. The response is
21
shareholder conflicts become more acute: For example, the tax reform effect of High Tax (Top Tax Bracket)
decreases by over 54% (25%) as the number of owners increases to three. Finally, Column (3) presents the t-
statistics of the differences in the High Tax and Top Tax Bracket coefficients. As expected, the effect is stronger
for owners in the top tax bracket for each number of owners bin.
The results in Table 7 also enable us to indirectly calculate an implied elasticity estimate. Since our
identification of the tax effect is based on a DDD approach, we cannot use the dividend tax rate to calculate
elasticity directly. We need to translate the coefficient estimating the resulting change in dividends into an
implied elasticity (see also Chetty and Saez 2005). To do this, we relate the payout change to the tax rate change.
In wholly owned firms, owners subject to an income tax rate of 51% (High Tax = 1) increase the dividend payout
ratio by 6.3 percentage points or by 74% of their pre-reform average. Relative to the dividend tax cut of 7.2
percentage points—14% of the income tax rate of 51% (14% = 7.2% / 51%)—this results in an implied elasticity
of -5.3 (=74% relative dividend increase / 14% relative tax cut). For owners in firms with three owners and an
income tax of 51%, the implied elasticity decreases to -2.4. In firms with five owners, the elasticity is statistically
not different from 0. The estimates for wholly owned are above prior estimates using aggregated data (e.g.,
Poterba 2004). Even for firms with three owners, the estimate is above Poterba’s (1987) estimate of -1.57, again
using aggregate data. One reason for the seemingly large elasticity is that owners substitute wages with dividends
(see Section 5.1 below). Shifting income across tax bases—in our case from wages to dividends—typically has
larger tax elasticities than real responses, i.e., a higher level of total payout (see, e.g., le Maire and Schjerning
Third, we demonstrate that the payout responses of CHCs are not the result of macroeconomic
developments in Sweden. Since our data comprise only CHCs, we use data on Swedish publicly traded firms as
counterfactuals and examine whether their dividend policy changed around 2006. We compare the aggregated
statistics of CHCs and public firms, since dividend taxation did not change for public firms. 24 Figure 7 presents
24
Our data contract does not allow us to append public firm data to our CHC data. Therefore, we can only compare
aggregated statistics.
22
firms. The comparison of the treated group (CHCs) and the counterfactuals (public firms) shows a parallel trend
prior to 2006 and a sharp increase in the dividend payout and fraction of dividend-paying CHCs after 2006. The
unconditional DD estimates are statistically significant despite the few observations in this aggregated analysis.
Finally, we do not exploit the policy experiment and examine the relation between dividends and ownership
using a simple cross-sectional analysis. While we cannot draw causal inference from this analysis, this test can
provide external validation to the results around the exogenous 2006 tax change. To be more precise, higher
taxes on wages than on dividends should be positively associated with dividend payout. This association is
expected to weaken as the number of owners increases. To test this, we use equation (2) and simply include the
dummy variable High Tax as well as its interaction with the number of owners (Owners) instead of running the
DDD model. Table 8 reports the regression results from this test. 25 The variable High Tax is positively associated
with the dividend compensation ratio. The economic magnitude of the coefficient implies that, if an owner is
subject to a marginal income tax rate of 51% or higher, the dividend compensation ratio increases by 2.2
percentage points, or 14% of the sample average, from before to after the reform. The negative coefficient for
the interaction High Tax×Owners suggests that the positive association between High Tax and the dividend
compensation ratio decreases as the number of owners increases. For example, going from one owner to three
owners decreases the effective coefficient of High Tax from 0.0218 to 0.0120 (= 0.0218 – (3–1) × 0.049), or by
about 45%. The effect of taxes becomes economically very small if a firm has five owners. Going from a wholly
owned CHC to a firm with five owners decreases the effective coefficient of High Tax by 0.0196 (= 0.0049 × (5
– 1)) from 0.0218 to 0.0022. Overall, these results are reassuring that our findings are not limited to one particular
point in time—the 2006 tax cut. Dividend tax response declines in the number of owners also in the overall
sample.
25
We find associations of firm-level variables with the dividend compensation ratio that are in line with prior literature.
For example, cash, profits, and retained earnings are positively associated with dividend compensation, while leverage
is negatively associated with dividend compensation. Consistent with the monitoring role of dividends, an increase in
the number of owners is positively associated with the level of dividend compensation.
23
What type of conflicts can explain the reduction in dividend tax sensitivity in firms with more owners? One
reason could be conflicting interests among owners (Shleifer and Vishny 1986). Diverse-ownership firms could
have a lower propensity to optimize the tradeoff between payout channels. One aspect of shareholder conflicts
could thus be coordination issues among owners, for example, on the optimal payout channel (Section 5.1). A
second aspect of shareholder conflicts could be differences in tax preferences among owners (Section 5.2). It is
important to note that conflicting interests among owners are distinct from agency issues between owners and
managers, which we analyze in Section 5.3. While in Sections 5.1 to 5.3 we use samples that are able to isolate
one friction (e.g., heterogeneous tax preferences), in Sections 5.4 and 5.5 we construct different subsamples that
enable us to examine the interactions of these frictions and how they jointly affect dividend tax responsiveness.
Private firms can pay out cash to owners as dividends or wages. Wages have two objectives. First, they
serve as compensation for labor supply. Second, they can serve as a tax-efficient channel to distribute cash to
shareholders. Indeed, about 78% of wholly owned firm owners receive wages in our sample. In firms with four
(five) owners, 65% (61%) of owners still receive wages. Owner–managers in very CHCs can easily substitute
dividends and wages to maximize the after-tax cash flows they receive from the firm. The more owners in a
firm, the more complex the maximization process that optimizes the payout for all the owners, particularly in
firms with different levels of owner involvement in daily operations. Even if all the owners receive wages, the
optimization process is more complex and requires more coordination among owners in a firm with many owners
than in a wholly owned firm. Coordination costs therefore increase with the number of owners. This friction can
then decrease the substitutability of payout channels and, consequently, result in muted tax responsiveness.
To analyze how the interaction of shareholder conflicts and taxation affects the substitutability of dividends
and wages, we examine how wages have been affected by the change in taxes. We rerun our DDD approach
from equation (2) using %Wage Compensation as dependent variable. Figure 8 plots the resulting 𝛼𝛼1 to 𝛼𝛼5
coefficients. For firms with no or little frictions (one or two active owners), the response to the 2006 tax reform
is strongest. The percentage of wage compensation decreases by about 5.4 percentage points (one-owner CHC)
24
owners decrease the wage compensation ratio by only 3.3 percentage points. The difference in coefficient
estimates is statistically (t-stat = 4.65) and economically significant: Tax responsiveness decreases by 39% if a
firm has three owners instead of one owner. In firms with five owners, the individual owners’ tax preferences
have little influence on wage compensation. 26 We obtain similar results when using the sample of firms with at
least 10 employees or the matched sample (see Table A.VII of the Online Appendix).
The documented differences in tax reform responses across firms with different ownership structures
suggest that the methods of payout (dividends vs. wages) are tax responsive if there are no shareholder conflicts
or agency issues. If firms have many owners and thus more potential conflicts among owners or between owners
and managers, this substitutability and tax responsiveness substantially decreases. To illustrate this
substitutability, Figure 8 additionally plots the Post×High Tax coefficient estimates using dividends as
dependent variable for numbers of owners from Tables 5. The results suggest that pre-tax wages plus dividends
do not change. We find the same zero-payout effect when we examine firm-level data and use the ratio of total
payout (wages plus dividends) to total assets as the dependent variable (Figure A.III of the Online Appendix).
Although pre-tax income does not change, after-tax income changes and, consequently, shareholder value.
The owners of very CHCs are able to substitute wages with dividends. Both effects are statistically and
economically significant and suggest a one-to-one substitution. The increase in dividends (decrease in wages)
of 7.3 (5.4) percentage points translates into annual tax savings of about 6,881 SEK (or about 936 USD) for each
highly taxed owner of a wholly owned CHC. This result is equivalent to an income of 1.1% relative to total pre-
tax income . The tax responsiveness of both dividend and wages decreases with the number of owners. In firms
with three owners, each individual owner with tax preferences for dividends still experiences an increase in
dividends and a decrease in wages. However, the magnitude of the one-to-one substitution is smaller. The
increase in dividends (decrease in wages) by 4.4 (3.3) percentage points is equivalent to tax savings of 4,400
26
One potential econometric concern is that, since dividends and wages are related, their residuals are also related. We
therefore re-estimate equation (2) as a seemingly unrelated regression model. Figure A.II of the Online Appendix
presents Post×High Tax estimates and the 95% confidence intervals for each bin of owners from the seemingly
unrelated regression model estimation. The findings are consistent with our main results.
25
tax responses to a large tax reform is significantly smaller and wages and dividends are no longer responsive to
the individual owners’ tax preferences. Highly taxed owners of CHCs with five owners therefore forgo tax
savings of about 1.1% of pre-tax income relative to owners of wholly owned CHCs. 27
One potentially important part of the conflicting interests among owners is the heterogeneity in tax
preferences (e.g., Michaely and Vila 1996). In firms with dispersed ownership, it may be harder to reach a
consensus about the level of dividends when shareholders have heterogeneous tax preferences. Therefore,
changes in dividend taxation could have a smaller impact on dividend policy when tax preferences are not
aligned. To test this, we examine how firms that are similar in owners’ average tax rates but that differ in the
variation of tax rates among shareholders responded to the 2006 dividend tax cut. We expect that firms with
homogeneous tax preferences for dividends increase dividend payouts after the reform. In contrast, we would
expect tax heterogeneity, that is, high variation of income taxes among shareholders, to mute the effect of
taxation on dividend payouts. Since wholly owned firms cannot have variations in tax rates among owners, we
restrict our sample to firms with at least two owners. Using firm-level data, we estimate the following model:
27
One possible concern about this result is that the substitutability of wages and dividends is limited to owners with
wages from CHCs and hence the reduction in wages is smaller for firms with many owners. It is possible that a larger
fraction of owners in dispersed firms do not receive wages than in very closely held firms. To this end, we rerun our
main DDD estimations and restrict the sample to owners that received a wage prior to the 2006 tax reform. This
approach ensures that the reform does not affect our sample selection. Table A.VIII of the Online Appendix presents
the regression results for our DDD approach from equation (2), with %Div Compensation and %Wage Compensation
as the dependent variables. The results support our main findings from above and show that the effects are significant
for owners in wholly owned firms and firms with two owners. In firms with four or more owners, the individual owner’s
tax preference does not affect dividend or wage compensation.
26
Avg High Tax and HighTaxSD as our tax measures, where Avg High Tax is defined as in equation (3) and
HighTaxSD is our proxy for tax heterogeneity. It is a dummy variable equal to one if the standard deviation of
the marginal income tax rate on wages among shareholders is in the top quintile of the standard deviation
distribution. In other words, if HighTaxSD equals one, the firm is characterized by high variation in the tax rate
among owners, that is, substantial tax heterogeneity. We use dummy variables instead of the actual average tax
rate and standard deviation. Using dummy variables allows us to estimate the DDD model and to interpret the
coefficients accordingly. The coefficient 𝛼𝛼2 is the DD coefficient and indicates how firms with a high average
tax rate on wages change dividends in response to the 2006 tax reform. The effect of tax heterogeneity on the
reform response of firms with tax incentives to pay dividends is captured by the DDD coefficient, 𝛼𝛼6 . If tax
preferences are heterogeneous among shareholders (HighTaxSD = 1), we would expect firms to respond less to
the reform, even if their owners have, on average, tax preferences for dividends (𝛼𝛼6 < 0). Our results are
qualitatively similar if we use continuous tax measures (see Table A.IX of the Online Appendix).
Table 9 presents the 𝛼𝛼2 and 𝛼𝛼6 coefficients from estimating equation (4) including firm-level controls, firm
fixed effects, and year fixed effects. We use firms with at least two owners (Column (1)) and firms with at least
four owners (Column (2)). Our results show that firms in which the owners’ average income tax on wages is
above the dividend tax rate increase dividend payouts from before to after the reform. The increase is
economically and statistically significant. If owners prefer dividends, firms increase the dividend-to-asset ratio
by about 1.49 (1.58) percentage points in the sample of firms with at least two (four) owners. The response to
the dividend tax cut is drastically reduced by tax heterogeneity. If the variation in tax rates among owners is high
(HighTaxSD = 1), the reform response decreases by 1.52 (1.03) percentage points in the sample of firms with
two (four) or more owners. The overall response (𝛼𝛼2 + 𝛼𝛼6 ) for firms with heterogeneous tax preferences
(HighTaxSD = 1) is insignificant in both cases. Hence, tax heterogeneity can partly explain why the estimated
dividend tax elasticity in firms with four or five owners in Table 5 is lower than in wholly owned firms.
27
In addition to shareholder conflicts, agency issues that result from separation of ownership and control and
conflicting objectives of managers and owners could explain the decrease in tax sensitivity (Chetty and Saez
2010). Using different settings and less precise data on ownership and owners’ tax preferences, prior literature
indeed finds that managers with larger stakes in the company respond more to changes in dividend taxation (e.g.,
Blouin, Raedy, and Shackelford 2011, Hanlon and Hoopes 2014). In our setting, ownership and control can also
be separated. Consider a firm with three owners. Either the firm hires an external manager to run the firm or one
of the owners acts as the manager. In both cases, ownership and control are separated and agency conflicts are
more acute than in a wholly owned firm, which can also be run by an external manager. To proxy for the
separation of ownership and control that goes beyond differences in ownership structure, we test how firms with
a manager without an equity stake responded to the 2006 tax reform. For this purpose, we test how the dividend-
to-asset ratio of CHCs changed around the 2006 tax reform. We use a DDD design and test the following
𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑒𝑒𝑒𝑒𝑒𝑒 − 𝑡𝑡𝑡𝑡 − 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑗𝑗,𝑡𝑡 = 𝛽𝛽0 + 𝛽𝛽1 𝐴𝐴𝐴𝐴𝐴𝐴 𝐻𝐻𝐻𝐻𝐻𝐻ℎ 𝑇𝑇𝑇𝑇𝑇𝑇𝑗𝑗,𝑡𝑡−1 + 𝛽𝛽2 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑑𝑑𝑗𝑗,𝑡𝑡 (5)
+𝛽𝛽6 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐴𝐴𝐴𝐴𝑔𝑔 𝐻𝐻𝐻𝐻𝐻𝐻ℎ 𝑇𝑇𝑇𝑇𝑇𝑇𝑗𝑗,𝑡𝑡−1 × 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑑𝑑𝑗𝑗,𝑡𝑡 + 𝛾𝛾Π𝑗𝑗,𝑡𝑡 + 𝛽𝛽𝑗𝑗 + 𝛽𝛽𝑡𝑡 + 𝜀𝜀𝑖𝑖,𝑗𝑗,𝑡𝑡
where Dividend-to-Assetsj,t is the dividend-to-asset ratio measured at the firm level, Avg High Tax is defined as
in equation (3), Post is a dummy equal to one for post-reform years, and 𝛽𝛽3 is the DD coefficient indicating how
payout responds to owners’ tax preferences, on average. We expect 𝛽𝛽3 to be positive, indicating that if owners
have tax preferences for dividends, firms should pay higher dividends after the reform. Note that the coefficients
on Avg High Tax cannot be compared across equations (4) and (5). To be more precise, the coefficient from Avg
High Tax in equation (5) (equation (4)) captures the effect for firms with no separation of ownership and control
(low variation in tax rates) but potentially with shareholder conflicts (agency issues).
28
to one if the highest wage in the firm to an employee exceeds the highest payout (dividend and wages) to one of
the owners. About half of our sample firms meet this criterion. If the separation of ownership and control affects
the tax sensitivity of payout, we should observe a smaller tax response when the firm is run by a manager without
an ownership share. In our model, this effect is captured by the DDD coefficient 𝛽𝛽6 that indicates how the
response to the tax reform changes if the firm has a manager without an ownership share. We expect 𝛽𝛽6 to be
negative. This would imply that, in firms with separation of ownership and control and with tax preferences for
dividends relative to wages, the increase in dividends is lower than in firms without an external manager but
with tax preferences for dividends. We also include the interaction between Post and Separated as well as the
interaction between Separated and High Tax. Equation (5) is estimated at the firm level, including firm-level
Table 10 presents the coefficient estimates using the full sample (Column (1)) and for firms with one owner
(Column (2)). We find positive 𝛽𝛽3 coefficients in both cases. That is, on average, firms increase the dividend-
to-asset ratio from before to after the reform when the owners have tax preferences for dividends. The DDD
coefficient 𝛽𝛽6 is negative. This finding implies that when ownership and control are separate, the response to the
tax reform is reduced. The economic magnitudes are large: The tax reform response in firms with tax preferences
for dividends decreases by about 72%, from 0.0085 to 0.0024, when an external manager runs the firm.
One concern about this approach is that the number of owners and having a separation of ownership and
management are related. In other words, Separated might be just an alternative proxy for the number of owners.
Even though the correlation between Separated and Number of Owners is rather small, we address this concern
by allowing the DD coefficient 𝛽𝛽3 to vary across number of owner bins. We thus control for potential differences
in the tax sensitivity across firms that differ in the number of owners when estimating 𝛽𝛽6 , the DDD coefficient
on the effect of agency issues on the (average) tax sensitivity. In Column (2) of Table 10, we present the DDD
coefficient. Relative to firms without separation of ownership and control, the tax reform response decreases by
about 0.6% of total assets if there is a manager in the firm without an ownership share. Importantly, the DDD
29
not simply capture the number of owners. Taken together, these results suggest that, in addition to conflicts
among shareholders, conflicting objectives between owners and managers can explain why dispersed ownership
firms have a lower dividend tax responsiveness than wholly owned firms.
In the three previous tests, we were able to construct samples that allow us to examine each of the potential
mechanisms in isolation and show that each has a negative effect on firms’ dividend policy responses to dividend
tax changes. A relevant question that remains is the interaction between these frictions. To this end, we construct
a sub-sample that allows us to examine tax heterogeneity and agency issues together. We focus on the subset of
firms with at least two owners because shareholder conflicts in the form of tax preference heterogeneity cannot
be a friction in wholly owned firms. We start with a simple DD approach where we include the interaction of
Avg High Tax and Post. We then add the two triple interactions to obtain a DDD setting. First, we include the
triple interaction Avg High Tax × Post × HighTaxSD to proxy for tax heterogeneity. Second, we include the
triple interaction Avg High Tax × Post × Separated to proxy for agency issues. 28 In this approach, the Avg High
Tax × Post coefficient captures the reform response of firms in which (1) owners on average prefer dividends,
(2) there is low tax heterogeneity, and (3) there is no separation of ownership and control. The triple interaction
Avg High Tax × Post × Separated captures the effect on the dividend tax elasticity when there is tax
heterogeneity but no separation of ownership and control. Likewise, the Avg High Tax × Post × Separated
coefficient captures the effect on dividend tax sensitivity when there is separation of ownership and control but
low tax heterogeneity. If only one of these mechanisms is the force behind the declining tax responsiveness, the
proxy for the other mechanism will be insignificant in this model. Columns (3) and (4) of Table 10 present the
regression results for the DD coefficient as well as the two DDD coefficients. In Column (4), we additionally
allow the DD coefficient to vary across the number of owners for the reasons discussed above. Therefore, the
main coefficient of Avg High Tax × Post is not identified and not reported.
28
We also include the other double interactions and main coefficients so that the resulting DDD approach is identified.
30
that agency issues as well as shareholder conflicts decrease dividend tax elasticity. In addition, we find that the
coefficient of Post × Avg High Tax× HighTaxSD significantly differs from the coefficient of Post × Avg High
Tax× Separated (t-stat = 2.13). This result implies that, in our setting, shareholder conflicts have a slightly larger
economic impact on the dividend tax sensitivity of payouts. These results are very similar when the DD
coefficient is additionally allowed to vary across number of owners bins (see Column (4)). Taken together, our
results suggest that both agency issues and shareholder conflicts can explain the observation in Section 4 that
In the final step, we attempt to isolate whether both types of shareholder conflicts—tax heterogeneity and
coordination costs—affect the dividend tax sensitivity. We use two distinct experiments to address the relative
roles of coordination costs and tax heterogeneity. We first examine how firms in which all owners have
homogeneous tax preferences for dividends but that differ in the number of owners responded to the tax reform.
In this experiment, we control for agency issues to isolate coordination issues from tax heterogeneity. If tax
heterogeneity solely explains the decline in dividend tax responsiveness as the number of owners increases, the
response to the dividend tax change would be similar across firms with homogeneous tax preferences for
dividends. In contrast, if coordination across owners also explains the declining tax responsiveness, we would
observe a decline in the response of dividend payouts to the tax reform as the number of owners increases, even
To this end, we plot the difference in dividend payout of firms where all owners have tax preferences for
dividend payout relative to other firms for each number of owners bin (Figure 9) Based on these five plots, we
make three important observations. First, for each number of owners bin, there is a parallel trend in the difference
in payout across groups before the reform. Second, there is a sharp decrease in the dividend payout for firms
with aligned tax preferences for dividends. Third, this increase in dividend payout decreases as the number of
owners decreases. This suggests that coordination might be part of shareholder conflicts. To address concerns
31
𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 − 𝑡𝑡𝑡𝑡 − 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑗𝑗,𝑡𝑡 = 𝛼𝛼0 + 𝛼𝛼1 ⋅ 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑗𝑗,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 = 1𝑗𝑗,𝑡𝑡 (6)
+𝛼𝛼2 ⋅ 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑗𝑗,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 = 2𝑗𝑗,𝑡𝑡 + 𝛼𝛼3 ⋅ 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑗𝑗,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 = 3𝑗𝑗,𝑡𝑡
+𝛼𝛼4 ⋅ 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑗𝑗,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 = 4𝑗𝑗,𝑡𝑡 +𝛼𝛼5 ⋅ 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑗𝑗,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 = 5𝑗𝑗,𝑡𝑡
where Dividend-to-Assetsj,t is the dividend-to-asset ratio measured at the firm level. We use a dummy variable,
Aligned, as our measure of homogeneous tax preferences for dividends. We set Aligned equal to one if all owners
have a higher tax rate on wages than the combined tax rate on dividends, measured in t – 1. We then allow the
DD coefficient Aligned × Post to vary firms with more owners to examine whether coordination issues can also
explain part of the muted dividend tax responsiveness. We also allow the main coefficient on Aligned to vary
across number of owners bins and the vector β in equation (6) thus comprises five coefficients, 𝛽𝛽1 to 𝛽𝛽5 . We
again include firm-level control variables, firm fixed effects, and year fixed effects. We also control for the
agency issues by including Separated and its interactions with Aligned, Post, and Post × Aligned.
Figure 10 presents estimated changes in the dividend-to-asset ratio around the 2006 tax reform for firms
with aligned tax preferences for dividends across owners relative to the unconditional pre-reform average. When
tax preferences for dividends are homogeneous across owners, even firms with many owners increase dividend
payouts in response to the tax reform. The estimated response for firms with five owners is significant and
positive. Importantly, our results show that wholly owned firms respond more strongly to the reform than firms
with more dispersed ownership, even when the tax preferences are the same among all owners. Going from a
wholly owned firm to a firm with four owners, we find that the reform response decreases by 46%. The estimated
responses are significantly different from each other, since the confidence intervals of the estimated reform
32
there are also coordination issues among owner muting the response to dividend tax changes.
Second, we relax the assumption that the tax preferences of all owners in a firm need to be aligned for
dividends. Instead, we use the tax preference of the dominating shareholder. If the dominating shareholder
holding more than 50% of equity has tax preferences for dividends, tax heterogeneity among shareholders
becomes less important, since the majority shareholder can shape the CHC’s payout policy (Shleifer and Vishny
1986). Firms’ payout policies could thus respond to dividend tax changes if the dominating shareholder has tax
preferences for dividends. In other words, if tax heterogeneity and not coordination across shareholders explains
the decline in tax elasticity as the number of owners increases, we would expect the effect of the dominating
shareholder’s tax preference to be similar across number of owners bins. To test this, we again first plot the
difference in dividend payout of firms where the dominating owners prefer dividend payouts to other firms for
each number of owner bin. Figure 11 plots the resulting five charts. Again, we observe that there is (1) a parallel
trend before the reform in each number of owners bin, (2) that there is a sharp decrease in the dividend payout
for firms where the dominating prefers dividends, and (3) that this increase in dividend payout decreases as the
number of owners decreases. To test more formally whether in addition to tax heterogeneity and separation of
ownership and control, coordination issues can explain the declining tax elasticity, we run the following model:
𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 − 𝑡𝑡𝑡𝑡 − 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑗𝑗,𝑡𝑡 = 𝛼𝛼0 + 𝛼𝛼1 ⋅ 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐷𝐷𝐷𝐷𝐷𝐷_𝑇𝑇𝑇𝑇𝑇𝑇𝑗𝑗,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 = 1𝑗𝑗,𝑡𝑡 (7)
+𝛼𝛼2 ⋅ 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐷𝐷𝐷𝐷𝐷𝐷_𝑇𝑇𝑇𝑇𝑇𝑇𝑗𝑗,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 = 2𝑗𝑗,𝑡𝑡 + 𝛼𝛼3 ⋅ 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐷𝐷𝐷𝐷𝐷𝐷_𝑇𝑇𝑇𝑇𝑇𝑇𝑗𝑗,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 = 3𝑗𝑗,𝑡𝑡
+𝛼𝛼4 ⋅ 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐷𝐷𝐷𝐷𝐷𝐷_𝑇𝑇𝑇𝑇𝑇𝑇𝑗𝑗,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 = 4𝑗𝑗,𝑡𝑡 + 𝛼𝛼5 ⋅ 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑡𝑡 × 𝐷𝐷𝐷𝐷𝐷𝐷_𝑇𝑇𝑇𝑇𝑇𝑇𝑗𝑗,𝑡𝑡−1 × 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 = 5𝑗𝑗,𝑡𝑡
where Dividend-to-Assetsj,t is the dividend-to-asset ratio measured at the firm level and Dom_Taxj,t–1 is a dummy
variable equal to one if a dominating shareholder (holding more than 50% of equity) has a tax preference for
dividends. We again include firm-level control variables, firm fixed effects, and year fixed effects. The vector β
33
bins. We again control for the agency issues by including Separated and its interactions with Dom_Tax, Post,
Figure 12 presents the 𝛼𝛼1 to 𝛼𝛼5 coefficients along with the upper and lower 95% confidence intervals. Once
again, the reform response is significant for all firms at the 1% level. However, the effect of the dominating
shareholder’s tax preference for dividends on the firm’s dividend payout gradually decreases with the number
of owners. For example, the response to the tax reform decreases by 49% when the number of owners increases
from one to five, even if the dominating shareholder has tax preferences for dividends in both firms. Taken
together, our results suggest that agency issues as well as shareholder conflicts in the form of coordination
problems and tax heterogeneity can explain why dividend tax elasticity declines when there are more owners in
the firm. One main implication of our findings is that—at least for our sample—shareholder conflicts can have
an impact on shareholders’ wealth. As shareholders substitute wages with dividends, firm owners forgo increases
in after-tax income if they cannot optimize the mix of wages and dividends according to their own tax
6. Conclusion
This paper shows that, in an environment where, to a large extent, the only friction is taxation, the effect of
payout taxes on dividends is large: When there are no conflicting objectives between owners and managers or
among shareholders, the effect of dividend taxation is economically and statistically significant. As other
frictions become more acute, dividend taxation becomes less important for payout policy. We show that agency
issues (e.g., Chetty and Saez 2005, 2010) as well as shareholder conflicts contribute to this declining tax
sensitivity. Shareholder conflicts mute the dividend tax effect in two ways. The first mechanism involves
different tax preferences among owners. Heterogeneity in owners’ tax preferences reduces dividend tax
elasticity. For any given number of owners, greater tax-related disagreement about optimal dividend policy
results in more muted responses to tax changes. Second, coordination problems among owners in firms with
many owners make the substitution between dividends and other forms of payments more difficult and complex,
even if the tax preferences are homogeneous. We find strong empirical evidence that, with a limited number of
34
in private firms). This substitutability is a function of relative taxation on dividends and wages. With one owner–
manager, there is a high rate of substitution. Holding owners’ levels of dividend and wage taxation constant and
controlling for agency issues, we find this substitutability declines with more owners.
The findings of this paper are clearly relevant to private, entrepreneurial firms with close ties between
owners and firms. Taking our results at face value, their implications could possibly be relevant to large, listed
firms. Both shareholders conflicts and agency issues are likely to be present in larger, traded companies as well
and, thus, one could expect firms’ payout responses to dividend tax changes to be a function of agency issues
(e.g., Chetty and Saez 2005; Blouin et al. 2011; Hanlon and Hoopes 2014) and conflicting objectives among
shareholders (Shleifer and Vishny 1986). However, the extent to which shareholder conflicts as well as agency
issues matter in other settings and for other types of firms is an empirical question that needs to be answered by
future research.
Our results also have policy implications, since dividend tax reforms are frequently aimed at affecting
payout policy and, ultimately, disposable income. When introducing the 2003 tax cut in the United States,
President George W. Bush said “abolishing double taxation of dividends will leave nearly 35 million Americans
with more of their own money to spend and invest.” 29 His economic advisor at that time, R. Glenn Hubbard,
argued that a dividend tax reduction would boost stock prices and increase spending (Brav et al. 2008). At least
in our sample of small, entrepreneurial firms, the effect of capital taxation on after-tax income is large for firms
with low levels of shareholder conflicts or agency issues. The responsiveness to dividend taxation and thus the
effect on disposable income becomes much weaker as these frictions become stronger. It is thus challenging for
policymakers to effectively change the payout policy of firms in the presence of agency issues and shareholder
conflicts. Interestingly, the 2006 Swedish tax reform is consistent with this mechanism: Tax rates on dividends
paid by public firms did not change while the dividend taxes for privately held firms were cut and the tax
decrease was twice as large for closely held firms as for widely held private corporations.
29
President Bush‘s speech on economic policy in Chicago, IL, on January 7, 2003.
35
as well as agency issues affects corporate investments. Recent empirical studies using data on private firms do
not find any aggregate investment effect and only very small payout responses (e.g., Yagan 2015). While
dividend taxation does not appear to affect aggregate investment, it has an effect on the allocation of corporate
investment across firms (Becker, Jacob, and Jacob 2013; Alstadsæter, Jacob, and Michaely 2015). However,
none of these papers considers the interaction of shareholder conflicts, agency issues, and taxation. Our results
suggest that these frictions could be one reason why prior literature fails to identify an effect of dividend taxation
36
rules. Report to the Expert Group for Public Economics, Ministry of Finance, Stockholm.
Alstadsæter, Annette, Martin Jacob, and Roni Michaely (2015): Do dividend taxes affect corporate investment?
Bennedsen, Morten, Kasper Meisner Nielsen, Francisco Perez-Gonzalez, and Daniel Wolfenzon (2007): Inside
the family firm: The role of families in succession decisions and performance. Quarterly Journal of
structure: Evidence from the French banking reforms of 1985. Journal of Finance, 62, 597–628.
Berzins, Janis, Øyvind Bøhren, and Bogdan Stacescu (2014): Dividends, taxes, and agency costs. Working
Paper.
Blouin, Jennifer L., Jana S. Raedy, and Douglas A. Shackelford (2011): Dividends, share repurchases, and tax
clienteles: Evidence from the 2003 reductions in shareholder taxes. The Accounting Review, 86, 887–914.
Brav, Alon, John R. Graham, Campbell R. Harvey, and Roni Michaely (2005): Payout policy in the 21st century.
dividend tax cut on corporate dividend policy: Empirical and survey evidence. National Tax Journal, 61,
381–396.
Chemmanur, Thomas J., Karthik Krishnan, and Debarshi K. Nandy (2011): How does venture capital financing
improve efficiency in private firms? A look beneath the surface. Review of Financial Studies, 24, 4037–4090.
Chetty, Raj, John N. Friedman, Tore Olsen, and Luigi Pistaferri (2011): Adjustment costs, firm responses, and
labor supply elasticities: Evidence from Danish tax records. Quarterly Journal of Economics, 126, 749–804.
Chetty, Raj and Emmanuel Saez (2005): Dividend taxes and corporate behavior: Evidence from the 2003
37
DeAngelo, Harry, Linda DeAngelo, and Rene M. Stulz (2006): Dividend policy and the earned/contributed
capital mix: A test of the life-cycle theory. Journal of Financial Economics, 81, 227–254.
Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer (2002): The regulation of
entry. Quarterly Journal of Economics, 117, 1–37.
Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer (2008): The law and
Djankov Simeon, Caralee McLiesh, and Andrei Shleifer (2007): Private credit in 129 countries. Journal of
Edgerton, Jesse (2013): Four facts about dividend payouts and the 2003 tax cut. International Tax and Public
Finance, 20, 769–784.
Easterbrook, Frank H. (1984): Two Agency-cost explanations of dividends. American Economic Review, 74,
650–659.
Floyd, Eric, Nan Li, and Douglas J. Skinner (2015): Payout policy through the financial crisis: The growth of
repurchases and the resilience of dividends. Journal of Financial Economics, 118, 299–316.
Goolsbee, Austan (2000): What Happens When You Tax the Rich? Evidence from Executive Compensation.
Haltiwanger, John, Ron S. Jarmin, and Javier Miranda (2013): Who creates jobs? Small versus large versus
Julio, Brandon, and David L. Ikenberry (2004): Reappearing dividends. Journal of Applied Corporate Finance,
16, 89–100.
38
Le Maire, Daniel and Bertel Schjerning (2013): Tax bunching, income shifting and self-employment. Journal
Maksimovic, Vojislav and Gordon Phillips (2008): The Industry Life Cycle, Acquisitions and Investment: Does
Firm Organization Matter? Journal of Finance, 63, 673–708.
Michaely, Roni and Michael R. Roberts (2012): Corporate dividend policies: Lessons from private firms. Review
Michaely, Roni and Jean-Luc Vila (1996): Trading volume with private valuation: Evidence from the ex-
Miller, Merton H. and Franco Modigliani (1961): Dividend policy, growth, and the valuation of shares. Journal
of Business, 34, 411–433.
Petersen, Mitchell A. and Raghuram G. Rajan (1994): The benefits of lending relationships: Evidence from
small business data. Journal of Finance, 49, 3–37.
Petersen, Mitchell A. and Raghuram G. Rajan (2002): The information revolution and small business lending:
Does distance still matter? Journal of Finance, 57, 2533–2570.
Poterba, James M. (1987): Tax policy and corporate saving. Brookings Papers on Economic Activity, 2, 454–
504.
Poterba, James M. (2004): Taxation and corporate payout policy. American Economic Review, 94, 171–175.
Schoar, Antoinette (2002): Effects of corporate diversification on productivity. Journal of Finance, 57, 2379–
2403.
Saez, Emmanuel (2010): Do taxpayers bunch at kink points? American Economic Journal: Economic Policy, 2,
180–212.
Shleifer, Andrei, and Robert W. Vishny (1986): Large shareholders and corporate control. Journal of Political
Economy, 94, 461–488.
Yagan, Danny (2015): Capital tax reform and the real economy: The Effects of the 2003 dividend tax cut.
39
600,000
500,000
400,000
300,000
200,000
100,000
0
All Owners 1 Owner 2 Owners 3 Owners 4 Owners 5 Owners
Pre-reform Post-Reform
40
41
42
43
44
45
Figure 10: Ownership and Dividend Payout: Homogeneous Tax Preferences for Dividends
This figure presents changes in the dividend-to-asset ratio around the 2006 tax reform estimated at the firm level.
For each ownership bin, we estimate the differences between the pre- and post-reform dividend-to-asset ratios for
firms where the marginal tax rate on wages is above the dividend tax rate for all owners (Aligned = 1). The
estimated model is defined as in equation (6). The figure plots the increase in the dividend-to-asset ratio relative
to the unconditional pre-reform average. The gray lines indicate the upper and lower 95% confidence intervals of
the point estimates.
46
Figure 12: Ownership and Dividend Payout: Tax Preferences of the Dominating Shareholder
This figure presents changes in the dividend-to-asset ratio around the 2006 tax reform estimated at the firm level.
For each ownership bin, we estimate the difference between the post-reform and pre-reform dividend-to-asset
ratios for firms where the marginal tax rate on the wages of the dominating shareholder (with over 50% ownership)
is above the dividend tax rate (Dom_Tax = 1). The estimated model is defined as in equation (7). The figure plots
the increase in the dividend-to-asset ratio relative to the unconditional pre-reform average. The gray lines indicate
the upper and lower 95% confidence intervals of the point estimates.
47
*
In 2009, the corporate tax rate was reduced to 26.3%.
+
Due to the corporate tax rate reduction, the combined tax rate amounted to 41.0% in 2009.
48
49
50
51
52
53
54
Table 10: Dividend Tax Sensitivity and Separation of Ownership and Control
This table presents the regression results on the payout policies of CHCs over the period 2001–2009, estimated at the
firm level. We use the ratio of dividend payout to the prior year’s total assets (Dividend-to-Asset Ratio) scaled by the
prior year’s total assets as the dependent variable. Independent variables comprise Avg High Tax, which is a dummy
variable equal to one if the weighted average income tax rate on the wages of all owners exceeds 50%; Post, a dummy
variable equal to one for the years 2006 and later; and Separation, a dummy variable equal to one if ownership and
management are separate. We set the dummy to one if an employee in the firm who is not an owner has a higher wage
than any owner receives in dividends and wages. Other firm controls include the log of total assets, the ratio of sales
to prior year assets, cash holdings relative to the prior year’s assets, retained earnings relative to the prior year’s assets,
and the debt-to-assets ratio. In Columns (2) and (4), we additionally include interactions of High Tax and the
respective number of owners bin ranging from one to five owners. In Columns (3) and (4), we restrict the analysis to
firms with at least two owners. In Column (4), we also include interactions of Post and High Tax with HighTaxSD,
our proxy of tax heterogeneity. The variable HighTaxSD is a dummy variable equal to one if the firm is in the top
quintile of the distribution of the standard deviation of income tax rates among shareholders. We include control
variables and firm and year fixed effects. We report robust standard errors clustered at the firm level in parentheses.
*** denotes significance at the 1% level.
Full Sample Firms with at least 2 Owners
(1) (2) (3) (4)
Post × Avg High Tax 0.0085*** 0.0113***
(0.0005) (0.0009)
Post × Avg High Tax × Separation -0.0061*** -0.0061*** -0.0078*** -0.0079***
(0.0006) (0.0006) (0.0009) (0.0009)
Post × Avg High Tax × HighTaxSD -0.0111*** -0.0109***
(0.0012) (0.0013)
Interactions of Tax × Number of Owners No Yes No Yes
Controls Yes Yes Yes Yes
Year Fixed Effects Yes Yes Yes Yes
Firm Fixed Effects Yes Yes Yes Yes
Observations 608,519 608,519 276,858 276,858
R-Squared 0.5961 0.5968 0.6257 0.6264
55