The Effect of Moving to a Territorial Tax System on

Profit Repatriation: Evidence from Japan∗

Makoto Hasegawa† Kozo Kiyota‡

August 2017

Abstract
In an increasingly globalized world, the design of international tax systems in terms
of the taxation of foreign corporate income has attracted much attention from policy
makers and economists alike. In the past, Japan’s worldwide tax system taxed foreign
source income upon repatriation. However, to stimulate dividend repatriations from
Japanese-owned foreign affiliates, the Japanese government introduced a foreign divi-
dend exemption system in 2009 that exempted dividends remitted by Japanese-owned
foreign affiliates to their parent firms from home-country taxation. This paper exam-
ines the effect of this dividend exemption system on profit repatriation by Japanese
multinationals. We find that the response of Japanese-owned affiliates to the dividend
exemption was heterogeneous. More particularly, foreign affiliates with a large stock
of retained earnings were generally more responsive to the reform and significantly
increased dividend payments to their parent firms in response to the enactment of the
dividend exemption system. Dividend payments by these affiliates also became more
sensitive to withholding tax rates on dividends levied by host countries under the new
exemption system.

Keywords: International taxation; Worldwide tax system; Territorial tax system; Profit
repatriation; Dividend exemption
JEL classification: H25; F23

∗
This paper is part of research project “Determinants of the Productivity Gap among Firms in Japan,”
undertaken at the Research Institute of Economy, Trade and Industry (RIETI) and an extension of a chapter
in Hasegawa’s doctoral dissertation. The study utilizes the micro data underlying the Basic Survey of
Japanese Business Structure and Activities and the Basic Survey of Overseas Business Activities conducted
by the Ministry of Economy, Trade and Industry (METI), and uses the Kikatsu–Kaiji converter provided by
the RIETI. Hasegawa is especially grateful to the members of his dissertation committee— Joel Slemrod,
James Hines, Stephen Salant, and Jagadeesh Sivadasan—for their invaluable advice, encouragement, and
mentoring. The authors also appreciate the helpful comments and suggestions of David Albouy, Christian
Gillitzer, Masaki Hotei, Arjan Lejour, Kevin Markle, Kiyoshi Matsubara, Hisahiro Naito, Tom Neubig,
Kazuki Onji, Masanori Orihara, Eiji Tajika, and Ryuhei Wakasugi, along with seminar participants at the
University of Michigan, the 2013 National Tax Association Annual Meeting and the 2014 International
Institute of Public Finance Annual Congress. Hasegawa gratefully acknowledges the financial support of the
Nomura Foundation, the University of Michigan, the National Graduate Institute for Policy Studies, and
the Japan Society for the Promotion of Science (JSPS) through KAKENHI Grant Nos. JP26780172 and
JP16H03610. Kiyota also acknowledges the financial support of the JSPS through KAKENHI Grant Nos.
JP26220503 and JP26285058. The usual disclaimers apply.
†
Graduate School of Economics, Kyoto University. E-mail: mhasegawa@econ.kyoto-u.ac.jp.
‡
Keio Economic Observatory, Keio University and RIETI. E-mail: kiyota@sanken.keio.ac.jp
1 Introduction
In an increasingly globalized world, the design of international tax systems, in terms of the
taxation of foreign corporate income has attracted much attention from policy makers and
economists alike. While taxing foreign source income may increase tax revenue, interna-
tional tax rules significantly influence the business activities of multinational corporations,
including the location of foreign direct investment, income reallocation (shifting) through
transfer pricing, and profit repatriation. For instance, the US taxes foreign income upon
repatriation, allowing foreign tax credits for corporate income and other related taxes paid
to foreign governments under the so-called worldwide income tax system. In contrast, a ter-
ritorial tax system exempts foreign income from home-country taxation, and such systems
are employed by many developed countries, including Australia, Belgium, Canada, France,
Germany, Italy, the Netherlands, and the UK.1 Accordingly, there is a longstanding debate
in the US on the merits of moving from its present worldwide tax system to a territorial tax
system (Joint Committee on Taxation, 2011).
Japan, the focus of this study, applied a worldwide income tax system until the end of
March 2009.2 At that time, the Japanese government was concerned that under the existing
system, Japanese multinational corporations kept a large portion of the profits earned by
their foreign affiliates abroad, and did not repatriate them back to Japan. Japanese firms
arguably had an incentive to do so because these foreign incomes were taxed at high rates (as
much as 40%) upon repatriation.3 As a means of stimulating dividend repatriations, Japan
introduced a foreign dividend exemption system in April 2009 that exempted dividends
remitted by foreign affiliates to their Japanese parent firms from home-country taxation.
Thus, with the introduction of the dividend exemption system, the Japanese corporate tax
system effectively moved to a territorial tax system. Because this reform drastically changed
the way in which the Japanese government taxed the foreign profits of Japanese corporations,
the reform would significantly affect the business activities of multinational corporations,
including decisions on foreign direct investment, income shifting, and profit repatriation.
This paper examines the effect of moving to a territorial tax system (with dividend
exemption) on profit repatriation by Japanese multinationals. A compelling motivation is
that few studies have empirically examined the outcomes of a change in the regime from
a worldwide tax system to a territorial tax system.4 In particular, studies concerning the
1
As of 2012, 28 of the 34 OECD countries employed territorial tax systems, and only the remaining six
countries (US, Chile, Ireland, Israel, South Korea, and Mexico) used worldwide tax systems (PwC, 2013).
2
In Japan, the fiscal year runs from April 1 to March 31.
3
In 2009, the corporate income tax rate in Japan was the highest in the OECD (OECD, 2010).
4
For the most part, the extant literature utilizes cross-country differences in international tax systems to
examine the effect of corporate taxes under both tax regimes on foreign direct investment (Slemrod, 1990;

1
impact of a territorial tax reform on dividend repatriations remain scarce. Tajika et al.
(2014), which is published in Japanese, investigate the impact of Japan’s dividend exemption
on dividends received by Japanese parent firms from their foreign subsidiaries using parent
firm-level data from 2008 to 2009.5 They find that more parent firms, especially those facing
greater demands for cash, increased dividends received from their foreign affiliates in response
to the enactment of the Japanese dividend exemption in 2009. However, their analysis does
not consider either the location of foreign affiliates owned by Japanese parent firms or the
tax rates of the host countries.
Elsewhere, Egger et al. (2015) examine the foreign dividend exemption enacted under
a UK tax reform in 2009 and find that foreign affiliates owned by UK multinational firms
responded to the tax reform by increasing dividend payments to their owners.6 Nevertheless,
two limitations remain regarding their data. First, they are unable to observe dividend
payments directly, and thus, they rely on an approximate measure of dividend payments.7
Second, because they calculate dividend payments using the balance sheet, they are unable
to identify the exact recipients.
Using unique confidential survey data for Japanese multinational corporations, we provide
the first evidence concerning the behavioral response of Japanese-owned foreign affiliates
regarding dividend repatriations to the adoption of a territorial tax system. As detailed
later, each foreign affiliate faces different tax costs to remit dividends to its parent firm
in Japan before and after the tax reform, depending on the corporate tax rate and the
withholding tax rate on dividends imposed by the host country. In other words, the tax
benefit (or cost) from the switch to the territorial tax system depends on the location of
the foreign affiliates. The contribution of our study is to identify the differential impact
of the tax reform on dividend repatriations using detailed information on foreign affiliates
and the tax rates of the host countries. In addition, the use of affiliate-level panel data
Hines, 1996; Altshuler and Grubert, 2001; Bénassy-Quéré, Fontagné, and Lahrèche-Révil, 2005). Desai and
Hines (2004) also estimate the tax burden on foreign income as being some $50 billion per year under the
US worldwide income tax system, while Grubert (2005) argues that considering the allocation of parent
overhead expenses (such as interest and R&D) to foreign income is important for obtaining more precise
estimates.
5
Some studies also examine the impact of the Japanese dividend exemption on firm value. Sakurada
and Nakanishi (2011) and Bradley et al. (forthcoming) examine investor reactions to news of the dividend
exemption during fiscal year 2008 using an event study methodology. However, they do not investigate the
actual consequences of the transition to the dividend exemption system after the 2009 tax reform.
6
Feld et al. (2016) show that the territorial tax reforms implemented in Japan and the UK increased
cross-border merger and acquisitions by Japanese and UK multinationals. Liu (2015) finds that the territorial
tax reform in the UK increased the capital investment of UK-owned foreign affiliates in low-tax countries.
7
Dividend payments in 2009 are calculated as the difference between the shareholder funds available
for distribution after current profits in 2008 and shareholder funds available for distribution before current
profits in 2009.

2
containing information on foreign subsidiaries’ dividend payments from 2007 to 2013 enables
us to evaluate the effect of the tax reform over a longer period than in the extant studies.
For our analyisis, we use the micro database of the annual survey conducted by the
Ministry of Economy, Trade and Industry of Japan (METI), the Survey of Overseas Business
Activities. This survey provides information on the financial and operating characteristics of
Japanese-owned foreign affiliates, including the dividends paid to Japanese investors. We also
use another METI survey, the Basic Survey of Japanese Business Structure and Activities,
containing financial information on parent firms. We merge these two surveys to construct
affiliate-level panel data that include parent firm information, and we analyze the period
from 2007 to 2013. Unlike Egger et al. (2015), the key advantage of our data is that we can
directly observe the dividends paid by foreign affiliates to their Japanese parent firms.
We find that the response of Japanese multinationals to the dividend exemption was
heterogeneous. Foreign affiliates with a large stock of retained earnings in the previous year
or before the tax reform were more responsive to the tax reform and significantly increased
dividend payments to their parent firms. Dividend payments by these affiliates also became
more responsive to withholding tax rates on dividends after the tax reform. This is because
multinationals can no longer claim foreign tax credits for withholding taxes on dividends and
thus must bear higher tax costs when repatriating dividends from foreign affiliates located
in countries with higher withholding tax rates. We also expected that foreign affiliates in
low-tax jurisdictions would exhibit a stronger response to the tax reform because the tax
costs for dividend repatriations from these affiliates were higher under the worldwide tax
system. However, we find no evidence that the change in dividend payments was positively
associated with the grossed-up tax rate differential between Japan and the host countries.
The remainder of the paper is organized as follows. Section 2 describes the background
and the provisions for dividend exemption enacted in Japan. Section 3 calculates the tax
costs of remitting profits from foreign subsidiaries to their parent firm in Japan through
dividends, royalties, or interest, and shows how the dividend exemption system changed
the tax costs of profit repatriation. Section 4 describes the data, while Section 5 presents
the empirical results from a preliminary analysis of the response of Japanese multinationals
to the dividend exemption system. Section 6 extends the empirical model in Section 5 to
analyze the situation where the heterogeneous responses to dividend exemption depend on
the magnitude of the stock of retained earnings in the foreign affiliates. Section 7 analyzes
alternative specifications of the dividend equation and checks the robustness of the main
results presented in Sections 5 and 6. Section 8 concludes.

3
2 The Dividend Exemption System Enacted in Japan
in 2009
In May 2008, a subcommittee on international taxation at the METI began to discuss the
introduction of a dividend exemption system in the corporate tax reforms slated for 2009.8
In August 2008, the subcommittee released an interim report and proposed introducing a
dividend exemption system (METI, 2008a). In the report, the METI estimated that the
stock of retained earnings of Japanese-owned foreign affiliates in 2006 was some 17 trillion
Japanese yen.9 Their main concern was that an excessive amount of profit was retained in
foreign countries to avoid home-country taxation in Japan, which then distorted the decisions
of Japanese corporations on the timing of profit repatriation and reduced the amount of
domestic R&D investment financed from foreign source income.10
As for the expected effects of dividend exemption, METI (2008a) argued that the dividend
exemption system would help: (1) to stimulate dividend repatriations with the removal of
the tax burdens on repatriated dividends, (2) to increase domestic capital investment and
R&D investment financed by repatriated foreign profits, and (3) to reduce tax compliance
costs by simplifying the international tax system.11
Following the release of the METI’s interim report, tax reform plans containing the
adoption of a territorial tax regime were sequentially released by the Government Tax Com-
mission, the Liberal Democratic Party (the ruling party in the Japanese House of Represen-
tatives), the Ministry of Finance, and the Cabinet from November 2008 to January 2009.
Finally, this regime change was included in the legislation for the 2009 tax reform. The
legislation passed into law on March 27, 2009 and came into effect on April 1, 2009.12
The Japanese dividend exemption system as enacted permits Japanese resident corpo-
8
This discussion of the foreign dividend exemption largely draws on Aoyama (2009) and Masui (2010).
9
This equates to some 146 billion US dollars at the 2006 exchange rate of 1 USD = 116.299 JPY (UNC-
TAD, 2016).
10
The subcommittee also examined the possibility of introducing a one-time dividend exemption similar to
the American Jobs Creation Act of 2004, limiting the use of dividends exempted from home-country taxation.
However, the subcommittee concluded that a one-time dividend exemption would only stimulate dividend
repatriations during the period under the exemption rule, and its aftereffects would counteract the initial
beneficial effect of dividend exemption. They were also concerned that limiting the use of exempted dividends
would distort the managerial decisions and undermine the managerial efficiency of Japanese corporations
(METI, 2008a).
11
To claim foreign tax credits under the worldwide tax system, Japanese multinationals were required to
submit numerous evidential documents to prove tax payments by their foreign affiliates and to calculate the
amount of foreign tax credits. Therefore, the foreign tax credit system imposed large tax compliance costs
on Japanese multinationals. The METI thus expected the dividend exemption system to help to reduce tax
compliance costs by abolishing the foreign tax credits for repatriated dividends (METI, 2008a).
12
Bradley et al. (forthcoming) provide additional details on the development of the dividend exemption
legislation.

4
rations to exempt 95% of their dividends received from foreign affiliates from home-country
taxation in accounting years commencing on or after April 1, 2009. The remaining (5%) of
dividends are regarded as expenses incurred by parent firms in earning the dividends and are
added to the calculation of their taxable incomes in Japan.13 In order to qualify for dividend
exemption, a parent firm must have held at least 25% of the shares in its affiliate for at least
six months prior to the dividend declaration date.14 While dividend exemption would reduce
corporate tax liabilities on repatriated dividends in Japan, foreign tax credits no longer apply
to withholding taxes on repatriated dividends imposed by the host countries. We also note
that the new system is still quite distant from purely source-based taxation. As the term
“dividend” exemption suggests, it only exempts foreign income in the form of dividends and
does not apply to other types of foreign source income, including royalties, interest payments,
income earned by foreign branches, and capital gains. Japanese multinationals can continue
to claim foreign tax credits for foreign taxes paid on these incomes.
Finally, because this tax reform concerns a foreign dividend exemption, the differences
between Japan’s dividend exemption enacted under the 2009 tax reform and the dividend
tax deduction enacted in the US under the American Jobs Creation Act of 2004 (AJCA)
are also noteworthy.15 First, while the AJCA provides US multinational corporations with a
special one-time deduction of 85% of dividends received from their foreign affiliates, Japan’s
dividend exemption is permanent. Second, under the AJCA, the 85% exemption applies only
to “extraordinary dividends,” defined as dividend payments exceeding average repatriations
over a five-year period ending before July 1, 2003, and after excluding the years with the
highest and lowest repatriations.16 Therefore, the exemption is limited to only part of the
dividends paid (extraordinary dividends), and US multinationals could claim dividend tax
reductions only if they received foreign dividends that exceeded the average. On the other
hand, Japan’s dividend exemption applies to 95% of all dividends as long as the conditions
described are satisfied.17 Thus, we note that the exemption permitted under the new tax
13
Expenses corresponding to 5% of the repatriated dividends are assumed to have been deducted from the
taxable incomes of parent firms when they invested in their subsidiaries and thus are not exempted upon
repatriation under the new exemption system.
14
The 25% minimum shareholding requirement is reduced under bilateral tax treaties with several coun-
tries. For example, the minimum shareholding is set at 10% for foreign affiliates in the US, Brazil, Australia,
and Kazakhstan, and at 15% for those in France (Aoyama, 2009).
15
Several studies investigate the effects of the one-time dividend deductions permitted by the American
Jobs Creation Act of 2004 on profit repatriation, domestic investment and employment, market values,
and income-shifting behavior of US multinational corporations (Oler et al., 2007; Blouin and Krull, 2009;
Redmiles, 2008; Bradley, 2011; Dharmapala et al., 2011).
16
In addition, to be eligible for the tax reduction, dividends must be paid in cash and invested in approved
activities in the US, although this requirement might not actually bind the decisions of US multinationals
(Blouin and Krull, 2009; Dharmapala et al., 2011).
17
METI (2008b) estimates that given its requirements, more than 95% of foreign affiliates would be eligible

5
system in Japan is quite different from, and arguably much more generous than, the dividend
tax reduction implemented under the AJCA in the US.

3 How Dividend Exemption Affects the Profit Repa-

triation of Japanese Multinationals
Hartman (1985) demonstrated that under certain conditions, repatriation taxes do not affect
the decisions on marginal investment and dividend payments made by “mature” subsidiaries
financing marginal investment out of their own retained earnings. However, this result
depends on the assumption that repatriation tax rates are constant over time. This may not
hold because, as we show below, the repatriation tax rates on dividends may change not only
depending on the foreign tax credit positions of the parent firms under a worldwide income
tax system and the definition of taxable income (tax bases) in the host countries, but also
when the corporate tax rate of the home or host country changes. Repatriation taxes could
then affect dividend repatriations of multinational corporations.18 In addition, repatriation
tax rates also vary because of changes in the international tax regime, as was the case for
Japan in 2009.
In what follows, we calculate the tax costs of remitting profits from foreign subsidiaries
to their parent firms in Japan through dividends, royalties, or interest given their decisions
on foreign direct investment, and we show how the dividend exemption system has changed
the tax costs of profit repatriation. We then make predictions for our empirical analysis
based on these changes in repatriation tax costs.
To consider the tax liabilities on foreign dividends under Japan’s worldwide tax system
(i.e., that before April 2009) and the new exemption system (i.e., after April 2009), we
calculate the tax costs of remitting an additional dollar of foreign income to Japan through
dividends, royalties, or interest. Let Yijct denote the pretax profit of affiliate i operating in
country c owned by parent j in year t, and let Tijct denote the foreign corporate income
tax paid by affiliate i. We define the average affiliate tax rate as τ̂ijct = Tijct /Yijct and
denote the statutory corporate income tax rates of Japan and country c in year t as τHt and
τct , respectively. We denote the withholding tax rates on dividends, royalties, and interest
D R I
payments in country c and year t as wct , wct , and wct , respectively.
for dividend exemption.
18
There is evidence that repatriation taxes discourage dividend payouts by US corporations (Hines and
Hubbard, 1990; Altshuler et al., 1995; Grubert, 1998; Desai et al., 2001). In contrast, using Japanese
affiliate-level data, Tajika and Nakatani (2008) find no evidence of the significant effect of corporate taxes
on dividend repatriations by Japanese multinationals.

6
 Under the worldwide tax system operating in Japan prior to April 2009, the tax liability
of parent j when it receives one dollar of additional dividends from its affiliate i in country
c depends on the foreign tax credit position of parent j; that is, whether the parent is in the
situation of an excess limit or an excess credit. A parent firm whose foreign tax payments
are less than the foreign tax credit limit, where the foreign tax credit limit is calculated
as the total foreign taxable income multiplied by the Japanese corporate income tax rate,
is referred to as being in excess limit. In contrast, if the foreign tax payments are greater
than the foreign tax credit limit, the parent is referred to as being in excess credit, and can
then use excess foreign tax credits (the difference between the foreign tax payments and the
foreign tax credit limit) to reduce its Japanese tax obligations on foreign source income over
the following three years.
Suppose that parent firm j is in excess limit. It could then claim foreign tax credits for
the taxes paid to host country c when affiliate i remits an additional dollar of dividends.
The dollar of dividends would be deemed as 1/(1 − τ̂ijct ) dollars of taxable income in Japan
(the gross-up formula), which yields a corporate tax liability of τH /(1 − τ̂ijct ). Parent j also
D
has to pay withholding taxes on the dividend wct to country c. Thus, the total tax payment
 D

needed to receive one additional dollar of dividends is τHt /(1 − τ̂ijct ) + wct . Parent j can
also claim foreign tax credits for the taxes paid to country c, which include the corporate
D
income tax payment τ̂ijct /(1 − τ̂ijct ) and the withholding tax on the dollar of dividends wct .
Therefore, the net tax payment of parent j to receive one dollar of dividends from its affiliate
i in country c in year t can be written as Pijct such that
   
τHt D τ̂ijct D τHt − τ̂ijct
Pijct ≡ + wct − + wct = ,
1 − τ̂ijct 1 − τ̂ijct 1 − τ̂ijct

which is the difference between the Japanese statutory tax rate and the subsidiary average
tax rate grossed up by the subsidiary average tax rate.
If parent j is in a position of excess credit, the parent can use the foreign tax credits to
remove its Japanese corporate tax liability.19 Then, the net tax payment is wct D
. In sum, the
tax costs of remitting a dollar of dividends can be written as
(
Pijct = (τHt − τ̂ijct )/(1 − τ̂ijct ) if parent j is in excess limit;
D
(1)
wct if parent j is in excess credit.

Following the introduction of the dividend exemption system (in April 2009), parent j
can exclude 95% of the dividends from its taxable income and thus only has to include 5%
19
Even when parent j is in an excess credit position, the foreign tax credit that parent j can claim is
limited up to the Japanese tax liability on the dollar of dividends (τHt /(1 − τ̂ijct )).

7
of the dividends in its taxable income. Conversely, the parent firm must incur withholding
D
tax on the dollar of dividends, wct . Thus, the net tax payment needed to receive the dollar
of dividends from affiliate i, or the repatriation tax rate under the new exemption system, is

D
0.05τHt + wct . (2)

Therefore, if parent j is in an excess limit position under the worldwide tax system,
the dividend exemption system eliminates almost all of its corporate tax liability in Japan
(except for the tax liabilities on 5% of the dividends).20 The tax cost of repatriating dividends
decreases from (τHt − τ̂ijct )/(1 − τ̂ijct ) to 0.05τHt when controlling for the withholding tax
D 21
rate on dividends wct . Alternatively, because withholding taxes on dividends are no longer
D
creditable under the dividend exemption system, parent i has to pay wct , which would have
been creditable under the worldwide tax system before 2009.
D
When the repatriation tax costs decrease to 0.05τHt (controlling for wct ), which is the
same for all firms, foreign affiliates will increase dividend payments under the new exemption
system as long as the repatriation taxes serve as a binding constraint on their dividend
payout decisions. Alternatively, the dividend exemption system reduces tax compliance
costs by abolishing the foreign tax credit system for dividends, and as a result, this could
possibly stimulate dividend repatriations. In addition, Japanese multinationals face different
repatriation tax costs depending on their foreign tax credit positions and the corporate tax
policies of the respective host countries. Because dividend exemption eliminates Japanese
corporate tax liability on repatriated dividends (Pijct ), dividend payments should become
less sensitive to (i.e., less affected by) the grossed-ed up differential between the Japanese
statutory tax rate and the average tax rate of the subsidiary (Pijct ) from 2009 on. In
other words, under the exemption system, foreign affiliates in low-tax countries (higher
Pijct ) should pay more dividends than before the tax reform. Finally, the withholding tax
rates on dividends would influence dividend repatriations under the new exemption system
more than before because, as (2) shows, multinationals must bear the burden of withholding
tax payments.
As we detail later, to control for firm size and the appreciation of the Japanese yen, we
20
Because of the relatively high corporate tax rate in Japan, it was highly likely under the previous
worldwide tax system that the foreign
 tax liability on the dollar of dividends was lower than the tax liability
τ̂ijct D τHt
in Japan 1−τ̂ijct + wct < 1−τ̂ijct . In fact, we find that this holds for 85.6% of all foreign affiliates from 2007
to 2008 in our data. Thus, it is reasonable to assume that most parent firms were in excess limit situations,
or that even if they were in excess credit, they did not have substantial excess foreign tax credits under the
worldwide tax system before 2009.
21
In this section, we assume Pijct = (τHt − τ̂ijct )/(1 − τ̂ijct ) > 0.05τHt . In our data from 2007 to 2013,
91.3% of foreign affiliates satisfy this condition.

8
measure dividend payments as the share of affiliate sales in our regression analysis. From
the above discussion, we hypothesize the expected effects of dividend exemption on dividend
payment scaled by affiliate sales to Japanese investors as follows.

H1: Dividend repatriations from foreign affiliates increase on average when controlling for
withholding tax rates on dividends.

H2: Dividend payments become less sensitive to the tax rate differentials between Japan and
foreign jurisdictions (Pijct ). In other words, foreign affiliates in low-tax jurisdictions
(higher Pijct ) should increase dividends (as scaled by sales) more than other affiliates.

H3: Dividend payments become more sensitive to withholding tax rates on dividends levied
by host countries.

While the dividend exemption system substantially changes the tax costs of repatriating
foreign dividends, it does not change the tax treatment of repatriated royalties and interest
payments at all. Consider the tax costs of remitting one additional dollar of a royalty or
interest from affiliate i to its parent j. Because these are deductible payments, remitting a
dollar as a royalty or interest will reduce the corporate tax payment in country c by τct . The
corporate tax liability on the dollar of deductible payments in Japan is τHt . Parent j also has
R
to remit to the government of country c the withholding tax on the dollar of royalty wct or
I R
on a dollar of interest wct . The total tax payment for the dollar of royalty is (τHt − τct + wct ),
I
while that for the dollar of interest is (τHt − τct + wct ).
If parent j is in excess limit, it would claim a foreign tax credit for the withholding tax
R I
on the dollar of royalty or interest, (wct or wct ). Thus, the net tax payment of remitting one
dollar of deductible payments is (τHt − τct ). If parent j is in an excess credit position, foreign
tax credits would eliminate the tax liability in Japan (τHt ), and the net tax costs would then
R

I


be wct − τct for the royalty payment and wct − τct for the interest payment.
In summary, regardless of the introduction of the dividend exemption system, the net
tax costs of remitting one additional dollar of a royalty can be written as
(
τHt − τct if parent j is in excess limit;
R
wct − τct if parent j is in excess credit.

The net tax costs of remitting one additional dollar of interest payments can be written
as (
τHt − τct if parent j is in excess limit;
I
wct − τct if parent j is in excess credit.

9
 As Grubert (1998) shows, those tax costs could affect dividend repatriations to the extent
that royalties and interest payments substitute for or complement dividends as an alterna-
tive means of profit repatriation. In the following sections, we empirically examine how
the response of dividend payments by Japanese-owned foreign affiliates to the repatriation
tax costs changed because of the introduction of the dividend exemption regime and test
hypotheses H1–H3.

4 Data
We use the micro database of the annual survey conducted by the METI, the Survey of
Overseas Business Activities. The main purpose of this survey is to obtain basic information
on the business activities of the foreign subsidiaries of Japanese firms. The survey covers all
Japanese firms (except those in the finance, insurance, and real estate industries) that owned
affiliates abroad at the end of the fiscal year (March 31). A foreign affiliate of a Japanese firm
is defined as a firm that is located in a foreign country and of which the Japanese firm owns
directly or indirectly at least a 10% equity share, covering up to second-tier subsidiaries (i.e.,
subsidiaries and sub-subsidiaries). The survey provides data on the financial and operating
characteristics of Japanese firms operating abroad, including dividends and royalties paid to
Japanese investors.
To control for parent firm characteristics, we use another METI survey, the Basic Survey
of Japanese Business Structure and Activities. This survey covers all firms with 50 or more
employees and capital or investment funds of at least 30 million yen, for both manufacturing
and non-manufacturing industries. The survey provides data on the financial and operating
characteristics of Japanese parent firms.
We merge these two annual cross-sectional surveys to develop a longitudinal (panel) data
set for foreign subsidiaries from 2006 to 2013. Each subsidiary is traced throughout the
period using the parent and affiliate IDs as a key.22 The merged panel data from the METI
surveys contain 75,250 observations for foreign affiliates in the sample from 2007 to 2013
with information on dividend payments available.23 Another 37,568 observations between
2007 and 2013 have missing values for dividend payments, and thus are not available for our
22
To merge the parent information from the latter survey into the subsidiary information from the former
survey, we use the converter provided by RIETI, which provides the correspondence between the parent IDs
from the two surveys.
23
Before 2007, the METI survey collected information in dividend payments to Japanese investors every
four years, and thus the information on dividends is unavailable for 2006. We collect the survey information
for 2006 to use the lagged values of sales and retained earnings of foreign affiliates and pretax profit, total
assets, and total debt of the parent firms in our regression analysis.

10
analysis.24
Table 1 provides summary statistics of the dividend payments by foreign affiliates (in
millions of yen) for each year from 2007 to 2013. As shown, the mean dividend payment
in 2009, the first year of the dividend exemption system, was larger than in either 2007 or
2008. However, even though the mean dividend was larger in 2009 than in 2007 or 2008,
the dividend payments for the 75th to 99th percentiles in 2009 were smaller than in 2007
or 2008. This implies that dividend payments above the 99th percentile in 2009 were much
larger than those in 2007 and 2008. As a result, the mean dividends paid by affiliates are
larger in 2009 than the previous two years.25 The mean dividend payments in each year
from 2011 to 2013 are also higher than under the worldwide tax system in 2007 and 2008,
while that in 2010 is the lowest among the seven years in the sample. We also note that the
median dividend is zero in all years, signifying that most foreign affiliates paid no dividends
(as detailed in Table 4).

=== Table 1 ===

There is a caveat when we look at dividend payments measured in Japanese yen, as the
yen sharply appreciated from 2007 to 2012. Thus, without normalization, the increase in
dividend repatriations could be undervalued, at least in terms of the domestic currency.26
Table 2 provides summary statistics of the dividend payments by foreign affiliates scaled by
sales to control for the size of the affiliates and allow for the change in the foreign exchange
rate. Table 3 also provides summary statistics of dividends scaled by affiliate sales after
winsorizing the top and bottom 1% to mitigate the influence of outliers. Comparison of
Tables 2 and 3 reveals that the mean values in each year, particularly 2012 and 2013, are
inflated because of the existence of outlier firms that paid huge amounts of dividends relative
to their sales. Nevertheless, both tables show that the mean dividends scaled by sales are
larger from 2010 to 2013 than in 2007 and 2008. Tables 2 and 3 show that the dividend
payments scaled by sales at the 75th percentile in 2009 and 2010 are smaller, but still larger
at the 90th percentile and above, than those in 2007.
24
There are many missing values for the dividend information because the respondents (parent firms)
left their response to the question on dividends blank when no dividends were paid. We also found that
the number of observations with missing values for dividends increased from 2010, while the number of
observations with zero dividends decreased by almost the same extent. We speculate that zero dividends
were recorded as missing values for these affiliates from 2010 to 2013. Thus, we replaced missing values for
dividends from 2010 to 2013 by zero if an affiliate reported dividend payments in all of the years in and
before 2009 when the affiliate appeared in the sample.
25
To ensure confidentiality, we are obliged not to provide the maximum and minimum values.
26
The average exchange rates between US dollars and Japanese yen during the data collection period are
as follows: 1 USD = 117.75 JPY in 2007, 103.36 JPY in 2008, 93.57 JPY in 2009, 87.78 JPY in 2010, 79.81
JPY in 2011, 79.79 JPY in 2012, and 97.60 JPY in 2013. (UNCTAD, 2016)

11
 === Tables 2 and 3 ===

Table 4 details the number of foreign affiliates that paid no dividends, the number of
affiliates paying dividends to Japanese investors, and the proportion of affiliates paying
dividends in each year from 2007 to 2013. As shown, 71.3% of all affiliates in the sample pay
no dividends. Strikingly, while the proportion of foreign affiliates paying dividends is lowest
in 2009 (25.5%), it is relatively stable over time, at between 25.5% and 31.3%, and does not
clearly increase after the tax reform in 2009.

=== Table 4 ===

In summary, while dividend payments scaled by sales increased after the tax reform, the
proportion of foreign affiliates paying dividends did not. In each year after the reform, the
size of dividend payments was generally larger than before the reform at the 90th percentile
and above. These are suggestive of the heterogeneous response of Japanese multinationals
to dividend exemption. That is, although the dividend exemption system may not have
stimulated dividend repatriations from most foreign affiliates that had not hitherto paid
dividends under the worldwide tax system, a small portion of firms that had paid large
amounts of dividends under the worldwide tax system may have further increased dividends
because of dividend exemption. These observations motivate our regression analysis in the
following sections to consider the possibility that the response of foreign affiliates to dividend
exemption varies depending on the payout capacity for dividends, as measured by the stock
of retained earnings.
Finally, Tables A1, A2, and A3 provide the number of foreign affiliates, the total amount
of dividends paid by these affiliates to Japanese investors, and the mean affiliate average tax
rates for each country in each year, respectively.27 We note that the locations of Japanese-
owned foreign affiliates favor the US, China, and other Asian countries, including Thailand,
Hong Kong, Singapore, Taiwan, Malaysia, South Korea, and Indonesia.

5 Preliminary Analysis
To test hypotheses H1–H3, we examine how the dividend exemption system affected the
repatriation behavior of Japanese multinational corporations and changed the responsiveness
of repatriated dividends to repatriation taxes (corporate taxes and withholding taxes) from
2009 onwards. As a preliminary analysis of dividend repatriation patterns before and after
27
In these tables and Table A-5, we focus on observations where information on dividends is available. We
also select 41 countries where there are at least 70 observations in total from 2007 to 2013.

12
the tax reform, our identification strategy in this section employs a version of a before-and-
after comparison using a post-reform dummy variable (set equal to one if t ≥ 2009 and zero
otherwise) in our regression analysis.28
We attempt to capture the causal effects of dividend exemption with the post-reform
dummy variable and its interaction terms with repatriation taxes while controlling for con-
founding factors that potentially affect dividend payments (measured in Japanese yen), such
as macroeconomic conditions, foreign exchange rates, the tax policies of the host countries,
and firm characteristics, as follows. First, we scale dividend payments by affiliate sales to
control for affiliate size and to make them robust to the appreciation of the Japanese yen.
Second, we include foreign exchange rates between Japanese yen and the local currencies and
real GDP growth rates of the host countries as country-level control variables to take into
account foreign exchange gains upon repatriation and reinvestment opportunities, respec-
tively. Third, we also control for firm-level financial characteristics. We use the annual sales
growth rate to capture investment opportunities in the host country at the affiliate level,
and we control for affiliate profitability to reflect payout capacity. To mirror the demand for
internal cash by parent firms, we control for the profitability and total debt of the parent
firms. Finally, we include country and industry dummies (or affiliate fixed effects) to con-
trol for systematic differences in dividend payments across different industries and countries,
possibly caused by different industry-specific and country-specific economic conditions that
are stable over the entire data period. We also include a set of year dummies whenever
possible.
We estimate the following equation in the spirit of Grubert (1998):

D R I
(Dividend/Sales)ijct = α0 + α1 Pijct + α2 wct + α3 wct + α4 wct + α5 τct
D R



+β0 DEt + β1 (DEt ∗ Pijct ) + β2 DEt ∗ wct + β3 DEt ∗ wct
I


+β4 DEt ∗ wct + β5 (DEt ∗ τct ) + γXijct + µc + νij + uijct , (3)

where (Dividend/Sales)ijct is the dividend payments of affiliate i located in country c to its

Japanese parent j divided by affiliate sales, in year t. The dummy variable DEt is equal
to one if t ≥ 2009, and zero otherwise. As defined earlier, Pijct is the grossed-up tax rate
differential between Japan and country c.29 The withholding tax rates of country c in year
28
Blouin et al. (2004) and Chetty and Saez (2005) employ a before-and-after comparison approach to
examine the effects of the 2003 dividend tax cut in the US on dividend payments to individual investors.
29
To bind the average affiliate tax rates (τ̂ijct = Tijct /Yijct ) between zero and one and to apply the gross-up
calculation to Pijct = (τHt − τ̂ijct )/(1 − τ̂ijct ) appropriately, we dropped observations with negative corporate
tax payments (Tijct < 0) and those with nonpositive pretax profits but positive tax payments (Yijct ≤ 0
and Tijct > 0). The average affiliate tax rate is set to zero if Tijct = 0 and Yijct ≤ 0 and one if τ̂ijct > 1,
which binds the average tax rate between zero and one. Foreign tax credits applied to up to 50% (35% from

13
 D R I
t on dividends, royalties, and interest payments are wct , wct , and wct , respectively.30 The
statutory corporate income tax rate of country c in year t is τct .31 The post-reform year
dummy DEt and the interaction terms of DEt with these tax variables are intended to
capture the change in dividends paid and the changes in the responsiveness of dividends to
the tax variables, respectively.
The vector of other control variables, denoted Xijct , includes the nominal exchange rate
of the Japanese yen per local currency in country c normalized to one at the 2005 level, the
annual real GDP growth rates, affiliate net profitability (after-tax profit scaled by sales),
affiliate annual sales growth, lagged parent net profit scaled by lagged total assets, lagged
parent total debt scaled by lagged total assets, the year dummy variable for 2008, and
industry dummies for the two-digit industry codes.32 Country dummies are included and
denoted as µc . Affiliate fixed effects are denoted as νij and included in some specifications.
To mitigate the influence of outliers, we winsorize all the firm-level variables used in the
analysis at the top and bottom 1%. Table 5 provides definitions of these variables, and
Table 6 details their summary statistics after winsorization of the firm-level variables.

=== Tables 5 and 6 ===

From the hypotheses proposed in the previous section, we expect the signs of the key
parameters to be as follows. If the dividend exemption system stimulated dividend repatria-
tions by foreign affiliates of Japanese multinational firms on average, the coefficients on DEt
should be positive (β0 > 0), as hypothesized in H1. The coefficient on Pijct is expected to
be negative (α1 < 0) because higher repatriation tax costs would discourage dividend pay-
ments under the worldwide tax system. If dividend payments became less sensitive to the
tax rate differential between Japan and foreign countries under the new exemption system
2011) of foreign taxable income under Japanese tax law. To consider this provision, and also to mitigate the
influence of extreme average tax rates, we replace τ̂ijct with 0.5 if τ̂ijct > 0.5 when calculating Pijct .
30
We collect information on withholding tax rates on dividends, royalties, and interest from Ernst & Young
Worldwide Corporate Tax Guide (2006–2013), the documents released by Japan’s National Tax Agency that
summarize the revisions of tax treaties between Japan and its partner countries (“Summary of the Revision
of Withholding Tax Rates”), PricewaterhouseCoopers Worldwide Tax Summaries, and also the database
of the Japan External Trade Organization (JETRO), J-FILE (http://www.jetro.go.jp/world/search/cost/).
If bilateral tax treaties between Japan and the host countries reduce the withholding tax rates for large
shareholders, we use the reduced tax rates in our analysis.
31
We obtain data on statutory corporate income tax rates from the KPMG Corporate and Indirect Tax
Survey 2011 and KPMG’s Tax Rates Online. The statutory tax rates include subcentral statutory corporate
income tax rates.
32
As a measure of affiliate profitability, we use after-tax profit (net income) following Desai et al. (2001).
We tried several other profitability measures, including the lagged after-tax profit scaled by lagged sales,
a loss dummy variable that took a value one if net income was negative otherwise zero, and pretax profit
scaled by sales. We found that the main results presented in this and the next section remained qualitatively
identical.

14
as hypothesized in H2, the estimated coefficient on (DEt ∗ Pijct ) would be positive (β1 > 0).
Another interpretation of H2 is that if dividend repatriations from low-tax countries (high
Pijct ) were discouraged under the worldwide tax system, foreign affiliates in these countries
should increase dividends (as scaled by sales) more than other affiliates, given that dividend
exemption substantially eliminates the Japanese tax liabilities upon repatriation.
D
We expect the coefficient on wct to be insignificant or possibly negative (α2 < 0) because
D
the tax price of dividends does not depend on (wct ) if a parent firm is in excess limit,
which would be the case for most Japanese multinationals, while the tax price equals the
D
withholding tax rate on dividends (wct ) if a parent firm is in excess credit. If Japanese
corporations care more about the tax costs of withholding taxes on dividends under the
new exemption system and thus their dividend repatriations become more sensitive to the
withholding tax rates on dividends, as hypothesized in H3, the estimated coefficient on
D
(DEt ∗ wct ) should be negative (β2 < 0). The signs of the coefficients on the withholding
tax rates on royalties and interest, and the statutory tax rates will depend on how strongly
dividends substitute for royalties or interest as alternative means of profit repatriation.
As for the estimation methods, we follow the previous literature (Grubert, 1998; Desai
et al., 2001; Tajika and Nakatani, 2008) and estimate the equations using either a Tobit
procedure or ordinary least squares (OLS). We employ Tobit specifications because many
affiliates (71.3% of all affiliates in the sample) pay zero dividends, and thus the dependent
variable in equation (3) can be considered as a left-censored variable. In the Tobit estima-
tion, we include country and industry dummy variables to control for systematic differences
in dividend payments across industries and countries, and hence, we use across-affiliate vari-
ations to identify the parameters.33 We also estimate the equation using fixed-effects OLS.
The fixed-effects OLS estimation uses within-affiliate variations to identify the parameters
after controlling for the time-invariant unobservable characteristics of each affiliate.
Table 7 presents the estimation results. Columns (1)–(3) report the coefficients estimated
by a Tobit procedure. The point estimates are the marginal effects on the latent dependent
variable, which can be interpreted as the “desired” amount of dividend payments.34 Columns
33
We do not include affiliate fixed effects in the Tobit models because of the incidental parameters problem,
which renders estimators in nonlinear panel data models with fixed effects inconsistent and biased. This
would be an especially serious problem in a short panel like ours (Greene, 2007).
34
In our analysis, the key parameters of interest are the coefficients on the interaction terms for DEt and
the other tax variables. As Ai and Norton (2003) show, the interaction effect on a dependent variable in
nonlinear models differs from the marginal effect of the interaction term. Therefore, in the estimation of
our empirical models using the Tobit procedure, the marginal effect of the interaction terms on the observed
dividend payments (conditional on positive dividend payments) cannot be calculated in the usual manner.
Thus, when we use the Tobit estimation, we focus on the marginal effects on the latent variable for dividend
payments, which is a linear function of the independent variables.

15
(4)–(6) report the results from the fixed-effects OLS specifications.35 The estimated coef-
ficient on DEt is positive in all specifications and significant at either the 1% or 5% level.
The OLS estimate in column (4) indicates that that, holding affiliate sales fixed, dividend
payments are higher than the level in 2007 by 0.76% of affiliate sales in or after 2009, which
amounts some 68.3 million JPY (0.78 million USD).36 This result is consistent with hypoth-
esis H1.37

=== Table 7 ===

The estimated coefficient on the tax differential between Japan and the host countries
(Pijct ) is negative and statistically different from zero at the 1% level in the Tobit models
but not estimated precisely in the fixed-effects OLS model, possibly because of the lack of
variation in Pijct within affiliates over time. This suggests that the worldwide tax system
significantly discouraged dividend repatriations from foreign affiliates in low-tax countries
because it would trigger an additional tax liability proportional to the difference between
the Japanese and foreign tax rates. However, the estimated coefficient on (DEt ∗ Pijct ) is
negative in all specifications and even significant in some specifications. This is inconsistent
with hypothesis H2, and suggests that dividend payments did not become less sensitive to
the tax rate differential between Japan and foreign countries following the enactment of
the dividend exemption system. In other words, foreign affiliates in low-tax countries did
not increase dividend payments to their parents significantly more than other affiliates.38
D
The coefficient on (DEt ∗ wct ) is estimated to be negative and is statistically significant
in all specifications. This is consistent with hypothesis H3. The OLS estimate in column
(4) implies that if the withholding tax rate on dividends is one percentage point higher
35
In the fixed-effects specifications, we do not include either the industry or country dummies because
they are absorbed within the affiliate fixed effects.
36
Hereafter we use the mean of affiliate sales from 2007 to 2013 (8,981 million JPY) and convert Japanese
yen to US dollars using the average exchange rate between 2009 and 2013: 1 USD=87.71 to evaluate the
economic magnitude of the estimated coefficients on the key parameters.
37
As Altshuler et al. (1995) point out, the transitory and permanent responses to changes in repatriation
tax costs may differ. In particular, we suspect that there may have been a spike in the level of repatriations in
2009. To explore this issue, we specified the post-reform year dummies (2009–2013) separately instead of as
DEt . We found that the coefficients on all of the post-reform year dummies from 2009 to 2013 were positive
in all specifications and statistically significant in the fixed-effects OLS specifications, while the estimated
coefficients on the year dummies in 2009 and 2010 lost significance in the Tobit specifications. This suggests
that the increase in dividend repatriations did not occur only in the initial year of the dividend exemption
regime.
38
We suspect that the repatriation response from low-tax countries may be captured by the negative
coefficient on the interaction term of DEt with statutory corporate income tax rates (DEt ∗ τct ), which is
marginally significant in the fixed-effects specifications in Table 7. However, the coefficient on the interaction
term is no longer significant in any specification when we estimate the equation after dropping Pijct and
(DEt ∗ Pijct ). Therefore, we find no evidence of a response from low-tax countries, even when we measure
the corporate income tax burden using the statutory corporate income tax rates of the host countries.

16
under the dividend exemption system, a foreign affiliate reduces its dividend payout by
0.025% of sales, which amounts to some 2.2 million JPY (25.6 thousand USD). In other
words, dividend payments become more sensitive to the withholding tax rates on dividends
following the enactment of the dividend exemption system.
D
We note that, similar to the coefficient on Pijct , the estimated coefficient on wct is also
negative in all specifications and significant at the 1% level in the Tobit specifications while
it is not statistically significant in the fixed effects specifications. This is not surprising be-
cause withholding tax rates change only infrequently over time. As Grubert (1998) showed
using US tax return data, this result suggests that withholding tax rates negatively affected
dividend repatriations under the worldwide tax system, possibly because Japanese multina-
tionals cared about being in excess credit and bearing some of the withholding tax liabilities
that they could not offset with foreign tax credits.39 The coefficient on annual sales growth,
which is intended to capture investment opportunities for foreign affiliates, is negative and
significant at the 1% level, suggesting that foreign affiliates pay fewer dividends when they
have bigger investment opportunities in the host country. The positive and statistically sig-
nificant coefficient on net profitability implies that affiliates that are more profitable are able
to pay more dividends.
In summary, from the preliminary analysis we obtain some evidence that foreign affili-
ates, particularly those located in a country with a lower withholding tax rate on dividends,
increased dividend payments following the adoption of the territorial tax system. However,
we find no evidence of a strong repatriation response from foreign affiliates in low-tax coun-
tries. There are some caveats when interpreting these estimation results. First, relying on
the post-reform year dummy variables to measure the average change in the level of divi-
dend payments may incorrectly capture the effects of cyclical and sectoral macroeconomic
factors in or after 2009 on profit repatriation (in spite of our attempt to control for those
confounding factors with the various control variables). Second, as Tables 1–4 suggest, the
response of foreign affiliates to dividend exemption may be heterogeneous. Foreign affiliates
that have a larger dividend payout capacity than other affiliates—for example, those with
a larger stock of retained earnings—may have responded more flexibly to the dividend ex-
emption system by increasing dividend payments to their parent firms. Alternatively, they
could have changed their dividend payouts depending on the foreign corporate taxes and the
dividend withholding tax rates of the host countries.
39
As shown in equation (1), if the parent is in excess credit, the repatriation tax cost under the worldwide
tax system is equal to the withholding tax rate.

17
6 Heterogeneous Response to Dividend Exemption: By
Stock of Retained Earnings
As discussed in Section 2, one of the main goals of dividend exemption is to stimulate dividend
repatriations from foreign affiliates that had retained and accumulated large foreign profits to
avoid taxation in Japan. Therefore, it is worthwhile to consider how these affiliates responded
to dividend exemption. Moreover, foreign affiliates that had retained a large amount of
profits would be able to change (increase) dividend payments flexibly in response to the tax
reform because dividends are from after-tax profits and the stock of retained earnings. We
also observe from Tables 2 and 3 that dividend payments as a proportion of sales generally
increase at the 90th percentile and above in each year from 2009 onwards, while more than
half of all affiliates paid no dividends at all through the entire data period. This suggests
that a portion of foreign affiliates with large dividend payout capacities exhibited a stronger
response to dividend exemption. To address this issue, we investigate the possibility that
the heterogeneous response to dividend exemption depends on the magnitude of the retained
earnings of the foreign affiliates and examine whether foreign affiliates with a large stock of
retained earnings changed dividend payments in a manner consistent with hypotheses H1–
H3.
We use the information on the stock of retained earnings at the end of each fiscal year
and construct a dummy variable equal to one if the stock of retained earnings scaled by sales
is greater than the value at the 90th percentile in the previous year’s sample.40 Denote the
dummy variable by Rijct , where i is the index for the affiliate owned by parent firm j. We
first select the threshold at the 90th percentile because we observe from Tables 2 and 3 that
the distribution of dividends scaled by sales generally increases at the 90th percentile and
above in each year from 2009 onwards.
To consider the heterogeneity of the response to dividend exemption in the regression
equation, we estimate equation (3) including this dummy variable Rijct and the interaction
D D
terms of the dummy variable with each of DEt , Pijct , wct , (DEt ∗ Pijct ), and (DEt ∗ wct )
as independent variables. We also include year dummies for 2009 to 2013 instead of the
post-reform dummy DEt itself to capture the common yearly changes in dividends for all
affiliates. We specify a set of observations where Rijct = 1 is a treatment group and conduct
a version of difference-in-differences estimation.
Columns (1) and (2) in Table 8 present the estimation results. Column (1) reports
the coefficients estimated by the Tobit procedure, and column (2) reports the results from
40
In the survey, the stock of retained earnings is defined as “Proprietary capital − Capital − Capital
reserve.”

18
the fixed-effects OLS estimation. Both specifications include as control variables a full set
of year dummies from 2008–2013, Exchange Ratect , GDP Growthct , and affiliate and par-
ent firm financial characteristics including Profitabilityijct , Sales Growthijct , (Parent Net
Profit/Assets)ijc(t−1) , and (Parent Debt/Assets)ijc(t−1) .41 The coefficient on Rijct is positive
in both specifications and statistically significant in the Tobit models.42 In addition, the co-
efficient on (Rijct ∗ DEt ) is significantly positive in both specifications. These results suggest
that a foreign affiliate with a larger stock of retained earnings in the previous year paid a
larger amount of dividends under the worldwide tax system and increased dividend payments
further following enactment of the dividend exemption system in 2009. This is consistent
with hypothesis H1. The OLS estimate in column (2) indicates that after the tax reform, a
foreign affiliate within the top 10% of retained earnings relative to sales in the previous year
increased the amount of dividends by about 0.86% of sales (some 77.2 million JPY, or 0.88
million USD) compared to other affiliates, holding affiliate sales fixed.

=== Table 8 ===

D
The coefficients on (Rijct ∗ DEt ∗ wct ) are negative and statistically significant at the
1% level in both specifications. This suggests that a foreign affiliate with a larger stock of
retained earnings in the previous year cares more about the tax costs of withholding taxes on
dividends under the dividend exemption system. This finding is consistent with hypothesis
H3. The fixed-effects OLS estimates imply that if the withholding tax rate on dividends
is one percentage point higher, a foreign affiliate with a large stock of retained earnings
reduces its dividend payout by 0.066% of sales (some 5.9 million JPY, or 67.6 thousand USD)
relative to other affiliates from 2009 onwards, holding affiliate sales fixed. The coefficients
on (DEt ∗ Pijct ) and (Rijct ∗ DEt ∗ Pijct ) are not estimated to be statistically significantly
positive in any of the two specifications, which is still inconsistent with hypothesis H2.
These findings are somewhat obscured, but still hold with smaller magnitudes of the
estimated coefficients, when we define the dummy variable Rijct at a lower threshold for
the stock of retained earnings. We estimate the same equation after redefining Rijct as a
dummy variable equal to one if the stock of retained earnings scaled by sales is greater than
the median value in the previous year’s sample. Columns (3) and (4) in Table 8 provide the
estimation results. The estimated coefficient on (Rijct ∗DEt ) loses its significance but remains
41
We tried the specifications excluding these affiliate and parent firm financial variables or including only
affiliate control variables as we did in columns (2)–(3) and (5)–(6) in Table 7 and have confirmed that the
inclusion or exclusion of these control variables does not alter the results presented in this section.
42
We cannot precisely estimate the coefficient on Rijct when including affiliate fixed effects, possibly
because the composition of foreign affiliates within the top 10% of retained earnings scaled by sales is stable
over time, and thus there is little within-affiliate variation in Rijct .

19
positive, in the Tobit specification. In the fixed-effects OLS specification, the estimated
coefficient is significantly positive (0.004) but smaller than that in column (2) of Table 8
D
(0.0086). We observe a similar tendency for the estimated coefficient on (Rijct ∗ DEt ∗ wct ).
As an alternative specification, we also try defining Rijct as the ratio of the stock of retained
earnings in year (t − 1) to sales in year (t − 1) instead of as a dummy variable. Table A-4
provides the estimation results. As shown, they are qualitatively similar to those in Table 8,
although the coefficient on (Rijct ∗ DEt ) loses statistical significance in some specifications.
This finding confirms that foreign affiliates with a larger stock of retained earnings in the
previous year were more responsive to the tax reform.
In summary, the response of Japanese-owned affiliates to dividend exemption was hetero-
geneous depending on the magnitude of the stock of retained earnings. A portion of foreign
affiliates that had sufficiently large payout capacity responded to dividend exemption in a
manner consistent with our hypotheses H1 and H3. The top 10% of foreign affiliates in
terms of retained earnings exhibited a stronger response to the tax reform because they had
sufficient capacity to change flexibly dividends paid from their retained earnings. These af-
filiates increased their dividend payouts and their dividend payments became more sensitive
to withholding tax rates on dividends with the enactment of the dividend exemption system.
However, we find no evidence that the change in dividend payments is positively associated
with the grossed-up tax rate differentials between Japan and foreign countries. In other
words, we do not find evidence for a sharp increase in dividend repatriations from low-taxed
foreign affiliates after the reform.
There are some possible explanations for this insignificant result. First, the tax rate
differential may not be a good measure of the repatriation tax costs for Japanese multina-
tionals. As pointed out by Altshuler et al. (2015), Japan allows corporations to repatriate
foreign profits without incurring domestic taxes through loans from foreign subsidiaries to
their parents. Unlike Japan, the US treats such loans as the distribution of profits and taxes
them under Section 956 of the Internal Revenue Code. This feature of the Japanese tax
system may attenuate the response from low-taxed foreign affiliates.43
Second, the insignificant repatriation response from low-taxed affiliates may imply a lack
of tax aggressiveness on the part of Japanese corporations. Altshuler et al. (2015) note that
43
Another feature of the Japanese tax system is that, unlike the US, Japan has tax sparing agreements with
several countries under bilateral tax treaties. Tax sparing is a treatment to allow multinational corporations
to benefit from a tax holiday provided in developing countries and to reduce repatriation tax costs under the
foreign tax credit system. We suspected that tax sparing agreements might be the reason for the insignificant
result. To examine this issue further, we estimated the same specifications as in Tables 7 and 8 after dropping
observations in the countries with which Japan had tax sparing agreements as of 2008, including Bangladesh,
Brazil, China, Philippines, Sri Lanka, Thailand, Vietnam, and Zambia. However, the results did not change
qualitatively.

20
“A notable feature of the Japanese tax environment is a compliant international tax-planning
culture. ... Although changes in attitudes are occurring, many Japanese companies consider
paying taxes a matter of loyalty, and the amount of taxes paid are considered a measure of
the company’s success” (pp. 24–25). The Japanese government was concerned that adopting
a territorial tax system would facilitate tax avoidance by multinational corporations shifting
foreign income to low-tax countries. Our results possibly suggest that Japanese parent firms
did not respond to dividend exemption by reallocating foreign profits to their foreign affiliates
in low-tax countries, thereby increasing dividend repatriations from these affiliates after the
tax reform, and this may alleviate at least this concern of the Japanese government.

7 Alternative Specifications and Robustness Checks

We have so far found the response of Japanese-owned foreign affiliates to the tax reform
depended on the stock of retained earnings in the manner predicted by hypotheses H1 and H3
but have found no evidence of a strong response from foreign affiliates in low-tax jurisdictions
as hypothesized by H2. In this section, we examine the results from the perspective of some
alternative specifications to test their robustness.

7.1 Sample Splitting based on the Stock of Retained Earnings

before the Tax Reform
In the previous section, we defined the dummy variable of Rijct based on the information
on the stock of retained earnings in year (t − 1) to test the heterogeneous response. In this
subsection, we split the sample based on the stock of retained earnings in a pre-reform year
(2008 or 2007), estimate equation (3) separately for the subsamples, and compare the results.
We first split the sample into two groups based on the stock of retained earnings in 2008
as follows. If foreign affiliates had a stock of retained earnings scaled by sales larger than
the sample median for 2008, these affiliates are classified in the “Large Retained Earnings”
group, while other affiliates are classified in the “Small Retained Earnings” group as long
as the information on the stock of retained earnings is available in 2008. If the information
on the stock of retained earnings of a foreign affiliate in 2008 is not available, the affiliate is
not classified in either group. All other things being equal, we expect to identify a stronger
response to the reform for the Large Retained Earnings group than for the Small Retained
Earnings group.
This estimation method has both advantages and disadvantages compared with the speci-
fications used in the previous section. The advantage is that, as Grubert and Altshuler (2013)

21
suggest, the stock of retained earnings in 2008 would reflect not only the payout capacity
of dividends but also the repatriation tax costs faced by foreign affiliates under the world-
wide tax system.44 This is because higher repatriation tax costs would have discouraged
foreign subsidiaries from paying dividends to their parents before the tax reform, which
in turn would have created a stronger incentive for them to accumulate retained earnings.
Therefore, the stock of foreign retained earnings before the tax reform should reflect the
repatriation costs faced by foreign subsidiaries under the worldwide tax system. Thus, we
expect that foreign affiliates with a large stock of retained earnings before the tax reform
should be more responsive to the introduction of the dividend exemption system.
Conversely, the disadvantage of this sample-splitting method is that when we split the
sample based on the stock of retained earnings scaled by sales in 2008, foreign affiliates that
did not appear or did not report the stock of retained earnings in 2008 are excluded from
the two subsamples over the entire data period. This substantially reduces the sample size
for estimation.45
We estimate equation (3) separately for the Large and Small Retained Earnings groups by
either Tobit or fixed-effects OLS. Columns (1)–(4) in Table 9 present the estimation results.
Columns (1)–(2) and (3)–(4) provide the results for the Large and Small Retained Earnings
groups, respectively. The estimated coefficient on DEt is statistically significant in both
specifications in columns (1)–(2) for the Large Retained Earnings group but significant only
in column (4) and smaller in magnitude for the Small Retained Earnings group. This suggests
that foreign affiliates that had accumulated large amounts of retained earnings before the
tax reform increased dividend payments to their parent firms after the tax reform.

=== Table 9 ===

To explore this finding visually, Figure 1 plots the mean of affiliate dividend payments
scaled by sales for each of the two groups from 2007 to 2013. The figure shows that the mean
dividend payment (scaled by sales) is larger for the Large Retained Earnings group than the
44
Grubert and Altshuler (2013) studied the dividend repatriation behavior of US multinationals under the
temporary tax holiday provided by the AJCA. They found that dividend repatriations increased as the stock
of retained earnings in foreign subsidiaries was larger, suggesting that the burden of repatriation tax costs
before the tax holiday increased with the stock of foreign retained earnings.
45
We also tried classifying foreign affiliates in the Large Retained Earnings group if their stock of retained
earnings was greater than the 90th percentile (instead of the median) in the sample for 2008. However,
we found that the sample size of the Large Retained Earnings group became too small (only about 4,000
observations) for us to estimate the parameters precisely and to obtain reliable results. In addition, the
sample size of the Small Retained Earnings group was much larger (more than 35,000 observations) than
that of the Large Retained Earnings group, and the estimates of the parameters were more statistically
significant because of the larger sample size. Therefore, it is not appropriate to compare directly the results
between the two subsamples.

22
Small Retained Earnings group and generally increases from 2009 onward for both groups.
In particular, the mean dividend payment for the Large Retained Earnings group spiked in
2009 and increased sharply in 2011 relative to that for the Small Retained Earnings group.
This figure appears to imply a stronger response of foreign affiliates in the Large Retained
Earnings group to the tax reform (relative to the Small Retained Earnings group), which is
consistent with the result for the coefficient on DEt shown in columns (1)–(4) of Table 9.

=== Figure 1 ===

D
We find a similar (but less significant) pattern for the coefficient on (DEt ∗ wct ) to that
for the coefficient on DEt in Table 9. Comparing columns (1) to (2) with (3) to (4), we find
D
that the coefficient on (DEt ∗wct ) is negative, larger in absolute terms, and more statistically
significant for the Large Retained Earnings group than for the other group. This result may
imply that foreign affiliates in the Large Retained Earnings group paid higher dividends after
the tax reform if the withholding tax rates on dividends were lower in the host countries.
Next, we check the results when we split the sample based on the stock of retained earnings
in 2007 instead of 2008. Columns (5)–(8) in Table 9 present the results. In columns (5) and
(6), the sample consists of foreign affiliates whose stock of retained earnings is larger than
the sample median in 2007. In columns (7) and (8), the sample consists of foreign affiliates
with the stock of retained earnings smaller than or equal to the median in the sample for
2007.
Unlike the results in columns (1)–(4) in Table 9, the coefficients on DEt do not clearly
differ between the two groups and are larger for the Small Retained Earnings group.46 How-
D
ever, as expected, the coefficient on (DEt ∗ wct ) is statistically significant only for the Large
Retained Earnings group in columns (5) and (6) but not for the Small Retained Earnings
group in columns (7) and (8). The coefficient is also much (more than 5–10 times) larger
in magnitude for the Large Retained Earnings group than the other group. This finding
suggests that foreign affiliates with a large stock of retained earnings as of 2007 paid more
dividends after the tax reform when they faced lower withholding tax rates on dividends.
To explore this finding, we split each of the two groups further into two subgroups
(four subgroups in total) based on the dividend withholding tax rate imposed by the host
country: foreign affiliates located in countries with withholding tax rates on dividends higher
than or equal to 10% (high withholding tax countries) and those located in countries with
withholding tax rates lower than 10% (low withholding tax countries). Note that the median
46
When we drop the affiliate and parent financial variables, the coefficient on DEt is larger for the Large
Retained Earnings group (0.0076) in the OLS specification than for the other group (0.0067), and both
coefficients are statistically significant at the 10% level.

23
(and the 75th percentile) of dividend withholding tax rates is 10%. In Figure 2, we plot the
mean of affiliate dividend payments scaled by sales for the four subgroups separately from
2007 to 2013. In the Small Retained Earnings group, foreign affiliates in high and low
withholding tax countries show a similar trend in the mean dividend payment over time.
In contrast, in the Large Retained Earnings group, foreign affiliates in low withholding tax
countries increased dividend payments scaled by sales relative to those in high withholding
tax countries from 2009 onward, which is consistent with the result for the coefficient on
D
(DEt ∗ wct ) shown in columns (5)–(8) of Table 9.

=== Figure 2 ===

In Table 9, the coefficient on (DEt ∗ Pijct ) is not statistically positive for the Large
Retained Earnings group. In sum, the results from these specifications are consistent with
our findings in the previous section. We have confirmed that the heterogeneous repatriation
response to the tax reform depends on the stock of retained earnings before the tax reform.
Foreign affiliates with a large stock of retained earnings before the tax reform increased
dividends and their dividend payments became more responsive to the dividend withholding
tax rates under the dividend exemption system.

7.2 Response to the Repatriation Tax Rates

We have thus far tested the hypotheses by focusing on the change in dividend payments after
the implementation of the 2009 tax reform and have failed to identify an increase in dividend
repatriations positively associated with Pijct , which is the grossed-up tax differential between
Japan and the foreign subsidiaries. The analysis in the previous sections examined the change
in dividend repatriations depending on the tax differential (Pijct ) and the withholding tax rate
D
(wct ), separately. In this subsection, as an alternative specification for examining the impact
of the reform depending on repatriation tax costs, we attempt to capture the repatriation tax
costs using a single variable and examine the response of divided payments to the repatriation
tax cost variable.
As shown in equations (1) and (2), for foreign affiliates owned by Japanese parents in the
excess limit status, the tax costs of remitting one dollar of dividends changed from Pijct to
D
0.05τHt + wct . Thus, the repatriation tax costs from 2007 to 2013 can be written as
(
Pijct if t ≤ 2008
Repatriation Taxijct = D
(4)
0.05τHt + wct if t ≥ 2009.
Table 10 provides the summary statistics of Repatriation Taxijct in each year, with its
mean plotted from 2007 to 2013 in Figure 3. The graph shows that the mean repatriation

24
tax cost is almost constant over time, except for 2009, and falls sharply from 26.6% in 2008
to 8.8% in 2009 because of the tax reform.47

=== Table 10 and Figure 3 ===

Given that repatriation tax costs within affiliates over time changed because of the tax
reform in 2009, we examine how the tax reform influenced dividend repatriations through
the changes in repatriation tax costs by estimating the following equation:

R I
(Dividend/Sales)ijct = α0 + α1 Repatriation Taxijct + α2 wct + α3 wct + α4 τct
+γXijct + πt + µc + νij + uijct ,

where the notations of the variables are the same as in equation (3) except for Repatriation Taxijct ,
defined by equation (4) and πt , which represents a full set of year dummies from 2008 to
2013. The key parameter of interest is the coefficient on Repatriation Taxijct (α1 ). If foreign
affiliates increased dividends in response to the decrease in repatriation costs caused by the
tax reform, we expect the estimated coefficient on Repatriation Taxijct to be negative.
Table 11 presents the estimation results. Columns (1)–(3) report the coefficients esti-
mated using a Tobit procedure. Columns (4)–(6) report the results from the fixed-effects
OLS specifications. The estimated coefficient on Repatriation Taxijct is negative in all spec-
ifications but statistically significant only in the Tobit specifications in columns (1)–(3).

=== Table 11 ===

The within-affiliate variations in Repatriation Taxijct caused by the tax reform consist of
two parts: (1) the removal of the residual burden in Japan (Pijct ), and (2) the additional tax
D
burden of the withholding tax (wct ). Our results in Sections 5 and 6 demonstrate that the
dividend payments became more sensitive to the withholding tax rates after the tax reform.
This finding implies that dividend repatriations respond to the additional tax burden of the
withholding tax given the tax reform. Therefore, we could attribute the insignificant result
in the fixed-effects specifications to the lack of the repatriation response to the removal of
the residual burden in Japan, which is consistent with our results regarding hypothesis H2
in the previous sections. In sum, even when we measure the repatriation tax costs using a
single variable incorporating the withholding tax rates on dividends, we cannot find clear
evidence that foreign affiliates that faced high repatriation tax costs before the tax reforms
(those located in low-tax jurisdictions) significantly increased dividends with the enactment
of the dividend exemption system.
47
The mean for each country from 2007 to 2013 is found in Table A5.

25
7.3 Censored Fixed-Effects Model
There may be some concern that our fixed-effects OLS specifications do not appropriately
address the censoring of the dependent variable (dividend payments). For example, Honoré
(1992) develops a censored fixed-effects estimation method for panel data. The estimator,
known as the trimmed least squares estimator, is consistent and asymptotically normal
without the need for parametric assumptions (such as normality) on the error term under
certain conditions.48 We estimate the Tobit specifications estimated in Sections 5 and 6
using the trimmed least squares estimator.49
Table 12 presents the estimation results for equation (3) using the trimmed least squares
estimator. The results are qualitatively similar to those presented in Table 7. As we found in
D
Table 7, the coefficient on DEt is statistically significantly positive and that on (DEt ∗ wct )
is statistically significantly negative in all specifications, while the coefficient on (DEt ∗ Pijct )
is not significantly different from zero. When we reestimate the specifications presented in
columns (1) and (2) of Table 8 using the trimmed least squares estimator, we obtain qualita-
tively similar results, as shown in Table 13. In sum, our results are robust to incorporating
affiliate fixed effects into the censored regression model.

=== Tables 12 and 13 ===

8 Discussion and Conclusion

In April 2009, Japan introduced a permanent dividend exemption system and moved to a
territorial tax system. Using confidential survey data for Japanese multinational corpora-
tions from 2006 to 2013, we provide the first evidence concerning the dividend repatriation
responses of Japanese-owned foreign affiliates to the transition from a worldwide income tax
system to a territorial tax system. We find that the response of Japanese-owned affiliates to
dividend exemption was heterogeneous, whereby foreign affiliates that retained a large stock
of retained earnings in the previous year or before the tax reform significantly increased
dividend payments to their parent firms in response to Japan’s adoption of a territorial tax
regime. This implies that the dividend exemption system helped to fulfill its primary goal
of stimulating dividend repatriations from foreign affiliates that had amassed large amounts
of foreign profits.
48
The estimator requires the exchangeability assumption that the error terms are symmetrically and
identically distributed conditional on the explanatory variables and the fixed effects.
49
For the reason described in footnote 34, we focus on the marginal effect on the latent dependent variable
(the desired amount of dividend payments).

26
 We also examined the changes in the responsiveness of dividends to foreign tax rates
(withholding tax rates on dividends and corporate tax rates in the host countries). Dividend
payments by affiliates with a large reserve of retained earnings became more sensitive to
withholding tax rates on dividends levied by host countries because Japanese multinationals
could no longer claim foreign tax credits for withholding tax payments on repatriated divi-
dends under the new exemption system. This suggests that under the new exemption system,
lowering withholding tax rates through the revision of existing tax treaties, or concluding
new tax treaties to lower withholding tax rates on dividends, would be a more effective way
of stimulating dividend repatriations than before. In contrast, we found no evidence that
the responsiveness of dividend repatriations to foreign corporate income tax rates changed
with the enactment of dividend exemption. More precisely, there is no clear evidence of a
strong repatriation response from low-taxed foreign affiliates after the tax reform.
Our results should be informative for the debate on international corporate tax policy
in the US as well as in other countries that have recently enacted, or are contemplating,
similar reforms. Our findings may be especially relevant to the US because of the similarity
between the US and Japanese worldwide tax systems, and given that the two countries
together have the highest corporate tax rates in the OECD. As discussed in Section 6, Japan
does not have a tax provision similar to Section 956 of the US Internal Revenue Code.50
Due to this feature of the international tax system, US multinationals may incur larger
tax burdens on profit repatriation from low tax jurisdictions than Japanese multinationals
did under the worldwide tax system. Indeed, there is evidence that dividend repatriations
from low tax jurisdictions surged in response to the one-time dividend tax reduction under
the AJCA (Dharmapala et al., 2011). Considering these points, the impact of dividend
exemption (if enacted in the US) on the dividend repatriation of US multinationals might be
larger than that of Japan’s dividend exemption. However, attempts to extend our findings
to other countries need to consider carefully the differences in international tax systems in
other respects including the Controlled Foreign Company (CFC) rules as well as possible
differences in cultural attitudes toward tax compliance and general tax morale. Nonetheless,
our finding that the heterogeneous response depends on the stock of retained earnings is
definitely worth noting.
50
Another notable difference between the US and Japan (before 2009) in the worldwide tax system is that
the US tax system requires multinational firms to calculate their foreign tax credits for foreign taxes on
passive and active income separately, while the Japanese worldwide tax system did not.

27
References
[1] Ai, Chunrong and Edward C. Norton. 2003. “Interaction Terms in Logit and Probit
models.” Economics Letters 80(1), pp. 123–129.
[2] Altshuler, Rosanne and Harry Grubert. 2001. “Where Will They Go if We Go Ter-
ritorial? Dividend Exemption and the Location Decisions of U.S. Multinational
Corporations.” National Tax Journal 54(4), pp. 787–809.
[3] Altshuler, Rosanne, T. Scott Newlon, William C. Randolph. 1995. “Do Repatriation
Taxes Matter? Evidence from the Tax Returns of U.S. Multinationals.” In Martin
Feldstein, James R. Hines, Jr. and R. Glenn Hubbard (eds.), The Effects of Taxation
on Multinational Corporations, Chicago: University of Chicago Press, pp. 253–272.
[4] Altshuler, Rosanne, Stephen E. Shay and Eric J. Toder. 2015. “Lessons the United States
Can Learn from Other Countries’ Territorial Systems for Taxing Income of Multi-
national Corporations.” Available at SSRN:http://ssrn.com/abstract=2557190.
[5] Aoyama, Keiji. 2009. “A Note on the New Japanese Exemption Policy of Foreign Sub-
sidiary’s Dividends (Gaikoku Kogaisha Haito Ekikin Fusannyu Seido no Kosatsu).”
Tsukuba Law Journal 6, pp. 99–117 (In Japanese).
[6] Bénassy-Quéré, Agnès, Lionel Fontagné, and Amina Lahrèche-Révil. 2005. “How Does
FDI React to Corporate Taxation?” International Tax and Public Finance 12(5),
pp. 583–603.
[7] Blouin, Jennifer and Linda Krull. 2009. “Bringing It Home: A Study of the Incen-
tives Surrounding the Repatriation of Foreign Earnings Under the American Jobs
Creation Act of 2004.” Journal of Accounting Research 47(4), pp. 1027–1059.
[8] Blouin, Jennifer L., Jana Smith Raedy, and Douglas A. Shackelford. 2004. “The Initial
Impact of the 2003 Reduction in the Dividend Tax Rate.” Working Paper, University
of Pennsylvania.
[9] Bradley, Sebastien. 2011. “Round-tripping of Domestic Profits under the American Jobs
Creation Act of 2004.” Working Paper.
[10] Bradley, Sebastien, Estelle Dauchy, and Makoto Hasegawa. forthcoming. “Investor Val-
uations of Japan’s Adoption of a Territorial Tax Regime: Quantifying the Direct
and Competitive Effects of International Tax Reform.” International Tax and Pub-
lic Finance.
[11] Chetty, Raj and Emmanuel Saez. 2005. “Dividend Taxes and Corporate Behavior: Ev-
idence from the 2003 Dividend Tax Cut.” Quarterly Journal of Economics 120(3),

28
 pp. 791–833.
[12] Desai, Mihir A., C. Fritz Foley, and James R. Hines Jr. 2001. “Repatriation Taxes and
Dividend Distortions.” National Tax Journal 54(4), pp. 829–851.
[13] Desai, Mihir A. and James R. Hines Jr. 2004. “Old Rules and New Realities: Corporate
Tax Policy in a Global Setting.” National Tax Journal 57(4), pp. 937–960.
[14] Dharmapala, Dhammika, C. Fritz Foley, and Kristin J. Forbes. 2011. “Watch What I
Do, Not What I Say: The Unintended Consequences of the Homeland Investment
Act.” Journal of Finance 66(3), pp. 753–787.
[15] Egger, Peter, Valeria Merlo, Martin Ruf, and Georg Wamser. 2015. “Consequences of
the New UK Tax Exemption System: Evidence from Micro-Level Data.” Economic
Journal 125, pp. 1764-1789.
[16] Feld, Lars P., Martin Ruf, Uwe Scheuering, Ulrich Schreiber, and Johannes Voget. 2016.
“Repatriation Taxes and Outbound M&As.” Journal of Public Economics 139, pp.
13-27.
[17] Greene, William. 2007. Econometric Analysis, 6th Edition. Upper Saddle River, New
Jersey: Prentice Hall.
[18] Grubert, Harry. 1998. “Taxes and the Division of Foreign Operating Income among
Royalties, Interest, Dividends and Retained Earnings,” Journal of Public Economics
68(2), pp. 269–290.
[19] Grubert, Harry. 2005. “Comment on Desai and Hines, “Old Rules and New Realities:
Corporate Tax Policy in a Global Setting”, ” National Tax Journal 58(2), pp. 263–
274.
[20] Grubert, Harry and Rosanne Altshuler. 2013. “Fixing the System: An Analysis of
Alternative Proposals for the Reform of International Tax,” National Tax Journal
66(3), pp. 671–712.
[21] Hartman, David G. 1985. “Tax Policy and Foreign Direct Investment.” Journal of Public
Economics 26(1), pp. 107–121.
[22] Hines, James R. Jr. 1996. “Altered States: Taxes and the Location of Foreign Direct
Investment in America.” American Economic Review 86(5), pp. 1076–1094.
[23] Hines, James R., Jr. and R. Glenn Hubbard. 1990. “Coming Home to America: Dividend
Repatriations by U.S. Multinationals.” In Assaf Razin and Joel Slemrod (eds.),
Taxation in the Global Economy, Chicago: University of Chicago Press, pp. 161–
200.

29
[24] Honoré, Bo E. 1992. “Trimmed Lad and Least Squares Estimation of Truncated and
Censored Regression Models with Fixed Effects.” Econometrica 60(3), pp. 533–565.
[25] Joint Committee on Taxation. 2011. Background and Selected Issues Related to the
U.S. International Tax System and Systems that Exempt Foreign Business Income.
JCX-33-11. May 20.
[26] Liu, Li. 2015. “International Taxation and MNE Investment: Evidence from the UK
Change to Territoriality.” Oxford University Centre for Business Taxation WP15/25.
[27] Masui, Yoshihiro. 2010. “Taxation of Foreign Subsidiaries: Japan’s Tax Reform
2009/10.” Bulletin for International Taxation 64(4), pp. 242–248.
[28] Ministry of Economy, Trade and Industry of Japan (METI). 2008a. Repatriations of
Foreign Profits by Japanese Enterprises: Toward the introduction of a dividend ex-
emption regime (Wagakuni Kigyo no Kaigairieki no Kanryu nitsuite). August 22,
(In Japanese).
[29] Ministry of Economy, Trade and Industry of Japan (METI). 2008b. On the 2009 Tax
Reform (Heisei 21 Nendo Zeiseikaisei nitsuite). December, (In Japanese).
[30] Oler, Mitchell, Terry Shevlin, and Ryan Wilson. 2007. “Examining Investor Expecta-
tions concerning Tax Savings on the Repatriations of Foreign Earnings under the
American Jobs Creation Act of 2004.” Journal of the American Taxation Association
29(2), pp. 25–55.
[31] Organisation for Economic Co-operation and Development (OECD). 2010. OECD Tax
Database, Paris: OECD.
[32] PricewaterhouseCoopers (PwC). 2013. “Evolution of Territorial Tax Systems in the
OECD.” Prepared for The Technology CEO Council.
[33] Redmiles, Melissa. 2008. “The One-Time Received Dividend Deduction.” Statistics of
Income Bulletin, pp. 102–114.
[34] Sakurada, Joe and Yoshiyuki Nakanishi. 2011. “The Stock Market Reaction to the
Announcement Regarding Japan’s Foreign Dividend Exemption (Gaikoku Kogaisha
Rieki no Kokunai Kanryu ni kansuru Zeiseikaisei to Shijo no Hanno).” Working
Paper, Hokkaido University. (In Japanese).
[35] Slemrod, Joel. 1990. “Tax Effects on Foreign Direct Investment in the United States:
Evidence From a Cross-Country Comparison,” in Assaf Razin and Joel Slemrod
(eds.), Taxation in the Global Economy, Chicago, IL: University of Chicago Press,
pp. 79–117.

30
[36] Tajika, Eiji, Masaki Hotei, and Keiko Shibata. 2014. “The Effects of Dividend-
Exemption Method on Repatriation of Income from Abroad: The Case of 2009
Japanese Tax Reform (Zeisei to Kaigai Kogaisha no Rieki Sokin: Honsha Shikin
Juyo Karamita 2009 Nendo Kaisei no Bunseki).” Economic Analysis (Keizai Bun-
seki). (In Japanese).
[37] Tajika, Eiji and Ryota Nakatani. 2008. “Welcome Home to Japan: Repatriation of
Foreign Profits by Japanese Multinationals.” Discussion Paper No. 2008-04, Hitot-
subashi University.
[38] United Nations Conference on Trade and Development (UNCTAD). 2016. UNCTAD-
Stat Database, http://unctadstat.unctad.org/ReportFolders/reportFolders.
aspx.

31
 Table 1: Dividend Payments by Foreign Affiliates (in millions of yen)

Year Mean SD Median P75 P90 P95 P99 N

2007 128 1473 0 12 136 350 2121 9492
2008 99 872 0 6 104 300 1615 10420
2009 132 2146 0 1 89 245 1535 11057
2010 96 873 0 4 98 285 1660 10951
2011 134 2849 0 8 121 333 1838 11102
2012 137 1685 0 9 126 320 1906 11193
2013 138 1139 0 16 171 471 2373 11035
Total 124 1725 0 7 119 327 1878 75250

Table 2: Dividend Payments by Foreign Affiliates as a Proportion of Sales

Year Mean SD Median P75 P90 P95 P99 N

2007 .0461 1.23 0 .00531 .0318 .0623 .21 9078
2008 .0262 .743 0 .00408 .031 .0632 .216 9912
2009 .0376 1.26 0 .00214 .0334 .0739 .286 10455
2010 .0414 1.25 0 .00458 .0325 .0649 .242 9979
2011 .0474 1.45 0 .00667 .0385 .074 .276 10175
2012 .119 5.06 0 .00742 .0403 .0795 .38 9935
2013 .275 12.5 0 .00942 .0409 .0793 .298 9772
Total .0843 5.17 0 .00566 .0356 .071 .272 69306

32
Table 3: Dividend Payments by Foreign Affiliates as a Proportion of Sales (after
Winsorizing at the Top and Bottom 1%)

Year Mean SD Median P75 P90 P95 P99 N

2007 .0113 .034 0 .00531 .0318 .0623 .21 9078
2008 .0114 .0348 0 .00408 .031 .0632 .216 9912
2009 .0127 .0397 0 .00214 .0334 .0739 .272 10455
2010 .012 .0364 0 .00458 .0325 .0649 .242 9979
2011 .0138 .0398 0 .00667 .0385 .074 .272 10175
2012 .0147 .0423 0 .00742 .0403 .0795 .272 9935
2013 .0149 .0411 0 .00942 .0409 .0793 .272 9772
Total .013 .0385 0 .00566 .0356 .071 .272 69306

Table 4: Proportion of Foreign Affiliates Paying Dividends

Year Dividend > 0 Dividend = 0 Total Number of Affiliates Proportion
2007 2848 6644 9492 30%
2008 2941 7479 10420 28.2%
2009 2823 8234 11057 25.5%
2010 3003 7948 10951 27.4%
2011 3238 7864 11102 29.2%
2012 3284 7909 11193 29.3%
2013 3456 7579 11035 31.3%
Total 21593 53657 75250 28.7%

33
 Table 5: Variable Definitions
Variable Definition
Salesijct Subsidiary operating revenues without in-
cluding non-operating income (in million
yen)
(Dividend/Sales)ijct Subsidiary dividend payments scaled by sales
Pijct Grossed-up difference between Japanese
statutory tax rate and the subsidiary aver-
age tax rate: Pijct = (τHt − τ̂ijct )/(1 − τ̂ijct ),
where τHt is Japan’s statutory corporate tax
rate
D
wct Withholding tax rate on dividends
R
wct Withholding tax rate on royalties
I
wct Withholding tax rate on interest
τ̂ijct Average subsidiary tax rate, which is defined
as the corporate tax payment divided by the
pretax profit of subsidiary i
τct Statutory corporate income tax rate
Exchange Ratect Exchange rate of the Japanese yen per local
currency, which is normalized to one in 2005
GDP Growthct Real GDP growth rate (%)
Profitabilityijct Net profit scaled by sales of subsidiary i
Sales Growthijct Annual sales growth rate of subsidiary i:
(Salesijct − Salesijc(t−1) )/Salesijc(t−1)
(Parent Net Profit/Assets)ijct Parent net profit scaled by total assets
(Parent Debt/Assets)ijct Parent total debt (total current and fixed li-
abilities) scaled by total assets
Retained Earningsijct Fiscal year-end balance of retained earnings,
which is defined as “Proprietary capital −
Capital − Capital reserve”
(Retained Earnings/Sales)ijct Subsidiary retained earnings at the end of the
account year scaled by sales
Repatriation Taxijct Tax costs of remitting one dollar of divi-
D
dends: Pijct if t ≤ 2008 and 0.05τHt + wct
if t ≥ 2009, where τHt is Japan’s statutory
corporate tax rate
Notes: The subscripts i, j, c, and t indicate the subsidiary, its parent firm,
the country where the subsidiary is located, and the year, respectively.

34
 Table 6: Descriptive Statistics

Variable Mean SD P25 Median P75 N

Salesijct 9171 27260 293 1293 5250 108792
(Dividend/Sales)ijct .013 .0385 0 0 .00566 69306
Pijct .245 .166 .158 .279 .402 77256
D
wct .0669 .0654 0 .1 .1 125822
R
wct .0883 .0583 .05 .1 .1 125822
I
wct .0998 .05 .1 .1 .1 125822
τ̂ijct .183 .194 0 .167 .286 77256
τct .279 .0717 .25 .25 .33 125866
Exchange Ratect .953 .158 .848 .955 1.08 125469
GDP Growthct 4.91 4.18 2.01 4.76 7.75 125535
Profitabilityijct -.0105 .335 .000851 .0269 .0718 99570
Sales Growthijct .248 1.03 -.119 .0642 .294 75817
(Parent Net Profit/Assets)ijct .0184 .0463 .0058 .0203 .0402 125626
(Parent Debt/Assets)ijct .564 .215 .4 .583 .739 125751
(Retained Earnings/Sales)ijct .0405 1.03 -.00965 .0881 .279 88576
Repatriation Taxijct .12 .118 .0203 .119 .12 105669
Notes: The subscripts i, j, c, and t indicate the subsidiary, its parent firm, the
country where the subsidiary is located, and the year, respectively. See Table 5
for the definition of each variable. Except for Repatriation Taxijct , all firm-level
variables with subscripts ijct are winsorized at the top and bottom 1% levels. The
winsorized values of Pijct are used when calculating Repatriation Taxijct .

35
 Table 7: Dividend Equation Regressions
(Dividend/Sales)ijct
(1) (2) (3) (4) (5) (6)

DEt 0.0187*** 0.0157** 0.0171** 0.0076*** 0.0078*** 0.0082***

(0.0068) (0.0068) (0.0068) (0.0025) (0.0029) (0.0029)
Pijct -0.1068*** -0.0783*** -0.0782*** 0.0016 0.0028 0.0028
(0.0051) (0.0052) (0.0053) (0.0017) (0.0019) (0.0020)
DEt ∗ Pijct -0.0026 -0.0107* -0.0114* -0.0034* -0.0036* -0.0033
(0.0059) (0.0061) (0.0061) (0.0018) (0.0021) (0.0022)
D
wct -0.1618*** -0.1274*** -0.1305*** -0.0320 -0.0271 -0.0306
(0.0411) (0.0399) (0.0399) (0.0210) (0.0228) (0.0240)
D
DEt ∗ wct -0.0417* -0.0497** -0.0487** -0.0245*** -0.0272*** -0.0292***
(0.0231) (0.0228) (0.0228) (0.0082) (0.0092) (0.0094)
R
wct 0.0054 0.0305 0.0321 -0.0036 -0.0085 -0.0108
(0.0451) (0.0443) (0.0447) (0.0166) (0.0193) (0.0213)
R
DEt ∗ wct -0.0514 -0.0481 -0.0439 -0.0010 -0.0016 0.0000
(0.0362) (0.0355) (0.0355) (0.0124) (0.0141) (0.0143)
I
wct -0.0044 -0.0356 -0.0360 -0.0244 -0.0296 -0.0238
(0.0618) (0.0585) (0.0584) (0.0194) (0.0212) (0.0265)
I
DEt ∗ wct 0.0215 0.0378 0.0378 0.0110 0.0177 0.0163
(0.0290) (0.0281) (0.0282) (0.0099) (0.0109) (0.0112)
τct -0.0524* -0.0252 -0.0153 -0.0019 0.0002 0.0002
(0.0270) (0.0270) (0.0270) (0.0088) (0.0101) (0.0104)
DEt ∗ τct -0.0246 -0.0065 -0.0066 -0.0140** -0.0144* -0.0154*
(0.0194) (0.0192) (0.0192) (0.0071) (0.0081) (0.0082)
Exchange Ratect 0.0040 0.0150** 0.0138** 0.0011 0.0029 0.0028
(0.0060) (0.0058) (0.0058) (0.0020) (0.0022) (0.0022)
GDP Growthct 0.0002 -0.0000 -0.0001 -0.0002*** -0.0002** -0.0002***
(0.0002) (0.0002) (0.0002) (0.0001) (0.0001) (0.0001)
Year 2008 0.0006 0.0015 0.0025 -0.0002 0.0001 0.0000
(0.0021) (0.0021) (0.0021) (0.0006) (0.0007) (0.0007)
Profitabilityijct 0.2492*** 0.2500*** 0.0112*** 0.0111***
(0.0038) (0.0038) (0.0017) (0.0018)
Sales Growthijct -0.0120*** -0.0120*** -0.0019*** -0.0019***
(0.0007) (0.0007) (0.0002) (0.0002)
(Parent Net Profit/Assets)ijc(t−1) 0.0371*** 0.0029
(0.0097) (0.0057)
(Parent Debt/Assets)ijc(t−1) -0.0108*** 0.0051
(0.0021) (0.0033)
Observations 60,568 51,186 49,261 60,568 51,186 49,261
R-squared 0.0031 0.0090 0.0092
Country & Industry Dummies Yes Yes Yes No No No
Affiliate Fixed Effects No No No Yes Yes Yes
Tobit or OLS Tobit Tobit Tobit OLS OLS OLS
Notes: All specifications contain a constant term. DEt is a dummy variable that is equal to one if t ≥ 2009
and zero otherwise. Year 2008 is the year dummy variable for 2008. See Table 5 for the definitions of all other
variables. Standard errors are in parentheses. Standard errors are clustered by affiliate in the OLS specifications.
***, **, and * denote significance at the 1%, 5%, and 10% level, respectively.

36
 Table 8: Dividend Equation including the Stock of Retained Earnings
(Dividend/Sales)ijct
90th Percentile Threshold for Rijct Median Threshold for Rijct
(1) (2) (3) (4)

Rijct 0.0113** 0.0038 0.0283*** 0.0019

(0.0047) (0.0043) (0.0034) (0.0015)
Rijct ∗ DEt 0.0150*** 0.0086* 0.0051 0.0039**
(0.0055) (0.0045) (0.0039) (0.0015)
Pijct -0.0812*** 0.0016 -0.0975*** -0.0013
(0.0057) (0.0018) (0.0079) (0.0018)
DEt ∗ Pijct -0.0064 -0.0010 -0.0085 -0.0014
(0.0066) (0.0020) (0.0093) (0.0018)
Rijct ∗ Pijct 0.0397** 0.0054 0.0729*** 0.0050
(0.0159) (0.0127) (0.0101) (0.0040)
Rijct ∗ DEt ∗ Pijct -0.0134 0.0029 -0.0017 0.0050
(0.0188) (0.0146) (0.0119) (0.0045)
D
wct -0.1310*** -0.0253 -0.1033** -0.0435*
(0.0409) (0.0238) (0.0437) (0.0246)
D
DEt ∗ wct -0.0182 -0.0192** -0.0252 -0.0133
(0.0241) (0.0097) (0.0313) (0.0099)
D
Rijct ∗ wct 0.0643** 0.0227 0.0017 0.0299**
(0.0315) (0.0285) (0.0243) (0.0137)
D
Rijct ∗ DEt ∗ wct -0.1259*** -0.0655** -0.0292 -0.0230**
(0.0376) (0.0273) (0.0294) (0.0114)
R
wct 0.0180 -0.0124 0.0200 -0.0138
(0.0452) (0.0218) (0.0445) (0.0219)
R
DEt ∗ wct -0.0320 0.0027 -0.0232 0.0002
(0.0362) (0.0148) (0.0359) (0.0150)
I
wct -0.0294 -0.0231 -0.0326 -0.0177
(0.0606) (0.0277) (0.0598) (0.0275)
I
DEt ∗ wct 0.0319 0.0143 0.0228 0.0134
(0.0288) (0.0115) (0.0285) (0.0114)
τct 0.0232 0.0146 0.0290 0.0144
(0.0283) (0.0113) (0.0281) (0.0113)
DEt ∗ τct -0.0061 -0.0136 -0.0035 -0.0111
(0.0195) (0.0084) (0.0194) (0.0084)
Observations 47,051 47,051 47,051 47,051
R-squared 0.0151 0.0166
Year Dummies Yes Yes Yes Yes
Country & Industry Dummies Yes No Yes No
Affiliate Fixed Effects No Yes No Yes
Tobit or OLS Tobit OLS Tobit OLS
Notes: All specifications contain a constant term, Exchange Ratect , GDP Growthct , Profitabilityijct , Sales
Growthijct , (Parent Net Profit/Assets)ijc(t−1) and (Parent Debt/Assets)ijc(t−1) as control variables. DEt is
a dummy variable that is equal to one if t ≥ 2009 and zero otherwise. Rijct is a dummy variable that is equal
to one if the stock of retained earnings scaled by sales is greater than the 90th percentile value (median) in
the sample for year (t − 1) in columns (1)–(2) (in columns (3)–(4)). See Table 5 for the definitions of all
other variables. Standard errors are in parentheses. Standard errors are clustered by affiliate in the OLS
specifications. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively.

37
 Table 9: Regressions of the Dividend Equation Using the Subsamples Split Based on the Stock of Retained
Earnings
(Dividend/Sales)ijct
Base Year: 2008 Base Year: 2007
Large Retained Earnings Small Retained Earnings Large Retained Earnings Small Retained Earnings
(1) (2) (3) (4) (5) (6) (7) (8)

DEt 0.0194** 0.0088* 0.0023 0.0077** 0.0136 0.0068 0.0236** 0.0086**

(0.0094) (0.0045) (0.0101) (0.0038) (0.0094) (0.0048) (0.0105) (0.0038)
Pijct -0.0428*** -0.0012 -0.0834*** 0.0010 -0.0315*** 0.0039 -0.0888*** -0.0015
(0.0079) (0.0039) (0.0069) (0.0018) (0.0075) (0.0039) (0.0070) (0.0017)
DEt ∗ Pijct 0.0009 0.0035 -0.0040 -0.0040** -0.0038 -0.0004 -0.0019 -0.0025
(0.0094) (0.0045) (0.0080) (0.0018) (0.0092) (0.0046) (0.0084) (0.0017)
D
wct -0.0991* -0.0222 -0.0908 -0.0649** -0.0915* -0.0118 -0.1009 -0.0721**
(0.0513) (0.0338) (0.0671) (0.0330) (0.0519) (0.0364) (0.0665) (0.0309)
DEt ∗ wct D -0.0442 -0.0305** -0.0238 -0.0115 -0.0649** -0.0452*** 0.0118 -0.0042
(0.0293) (0.0135) (0.0375) (0.0110) (0.0297) (0.0141) (0.0385) (0.0118)
R
wct -0.0277 -0.0369 0.1166 0.0488* -0.0296 -0.0389 0.1593* 0.0547*
(0.0562) (0.0301) (0.0827) (0.0286) (0.0558) (0.0287) (0.0824) (0.0292)
R

38
DEt ∗ wct -0.0492 -0.0107 0.0065 -0.0054 -0.0314 -0.0003 -0.0694 -0.0069
(0.0461) (0.0209) (0.0592) (0.0190) (0.0464) (0.0221) (0.0613) (0.0186)
I
wct -0.0448 -0.0268 -0.0254 -0.0256 -0.0689 -0.0125 -0.1941 -0.0484
(0.0730) (0.0407) (0.1194) (0.0198) (0.0774) (0.0412) (0.1229) (0.0311)
DEt ∗ wct I 0.0300 0.0176 -0.0066 0.0204* 0.0321 0.0226 0.0443 0.0150
(0.0382) (0.0171) (0.0428) (0.0114) (0.0376) (0.0180) (0.0445) (0.0120)
τct -0.0100 0.0053 -0.0961** -0.0062 -0.0261 -0.0020 0.0358 0.0084
(0.0369) (0.0172) (0.0409) (0.0094) (0.0374) (0.0178) (0.0423) (0.0108)
DEt ∗ τct -0.0163 -0.0147 0.0274 -0.0149 -0.0006 -0.0112 -0.0264 -0.0199*
(0.0258) (0.0123) (0.0294) (0.0112) (0.0259) (0.0130) (0.0305) (0.0107)
Observations 21,360 21,360 17,994 17,994 19,178 19,178 16,927 16,927
R-squared 0.0149 0.0141 0.0143 0.0176
Country & Industry Dummies Yes No Yes No Yes No Yes No
Affiliate Fixed Effects No Yes No Yes No Yes No Yes
Tobit or OLS Tobit OLS Tobit OLS Tobit OLS Tobit OLS
Notes: The large (small) retained earnings group consists of foreign affiliates whose stock of retained earnings is larger than (smaller than
or equal to) the median in the sample for 2008 in columns (1)–(4) and for 2007 in columns (5)–(8). All specifications contain a constant
term, Exchange Ratect , GDP Growthct , Year 2008 (the year dummy variable for 2008), Profitabilityijct , Sales Growthijct , (Parent Net
Profit/Assets)ijc(t−1) and (Parent Debt/Assets)ijc(t−1) as control variables. DEt is a dummy variable that is equal to one if t ≥ 2009 and
zero otherwise. See Table 5 for the definitions of all other variables. Standard errors are in parentheses. Standard errors are clustered by
affiliate in the OLS specifications. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively.
Table 10: Repatriation Tax Costs (Repatriation Taxijct ) Each Year

Year Mean SD P25 Median P75 N

2007 .251 .172 .146 .301 .407 10313
2008 .266 .169 .166 .321 .407 10941
2009 .0877 .0677 .0203 .12 .12 15302
2010 .0859 .0627 .0203 .12 .12 15845
2011 .0863 .0626 .0203 .12 .12 16590
2012 .0843 .063 .019 .119 .119 17905
2013 .0847 .0624 .019 .119 .119 18773
Total .12 .118 .0203 .119 .12 105669
Note: Repatriation Taxijct is the tax cost of remitting one
D
dollar of dividends: Pijct if t ≤ 2008 and 0.05τHt + wct if
t ≥ 2009, where Pijct = (τHt −τ̂ijct )/(1−τ̂ijct ), τHt is Japan’s
statutory corporate tax rate, τ̂ijct is the affiliate average tax
D
rate, and wct is the withholding tax rate on dividends. In
2007 and 2008, all values at the 75th percentile or above
take the value of 0.407 because more than 25% of foreign
affiliates pay no tax (τ̂ijct = 0).

39
 Table 11: Dividend Response to Repatriation Tax Rates
(Dividend/Sales)ijct
(1) (2) (3) (4) (5) (6)

Repatriation Taxijct -0.0923*** -0.0652*** -0.0643*** -0.0012 -0.0002 -0.0007

(0.0048) (0.0049) (0.0050) (0.0015) (0.0017) (0.0018)
R
wct -0.0667 -0.0339 -0.0306 -0.0100 -0.0161 -0.0182
(0.0445) (0.0439) (0.0442) (0.0154) (0.0183) (0.0202)
I
wct 0.0651 0.0346 0.0326 -0.0102 -0.0109 -0.0059
(0.0578) (0.0549) (0.0548) (0.0154) (0.0169) (0.0217)
τct 0.0188 0.0624*** 0.0650*** 0.0069 0.0113 0.0106
(0.0235) (0.0235) (0.0235) (0.0079) (0.0089) (0.0092)
Exchange Ratect -0.0023 0.0098 0.0111 0.0018 0.0069* 0.0075**
(0.0096) (0.0096) (0.0096) (0.0032) (0.0036) (0.0038)
GDP Growthct 0.0000 0.0000 0.0000 -0.0000 -0.0000 -0.0001
(0.0002) (0.0002) (0.0002) (0.0001) (0.0001) (0.0001)
Profitabilityijct 0.2381*** 0.2381*** 0.0083*** 0.0082***
(0.0036) (0.0036) (0.0015) (0.0015)
Sales Growthijct -0.0135*** -0.0134*** -0.0017*** -0.0017***
(0.0007) (0.0007) (0.0002) (0.0002)
(Parent Net Profit/Assets)ijc(t−1) 0.0330*** -0.0014
(0.0098) (0.0053)
(Parent Debt/Assets)ijc(t−1) -0.0125*** 0.0047
(0.0021) (0.0031)
Observations 67,744 56,721 54,452 67,744 56,721 54,452
R-squared 0.0023 0.0070 0.0073
Year Dummies Yes Yes Yes Yes Yes Yes
Country & Industry Dummies Yes Yes Yes No No No
Affiliate Fixed Effects No No No Yes Yes Yes
Tobit or OLS Tobit Tobit Tobit OLS OLS OLS
Notes: All specifications contain a constant term. Repatriation Taxijct is the tax cost of remitting one dollar
of dividends from foreign affiliate i to its parent firm j. See Table 5 for the definitions of all other variables.
Standard errors are in parentheses. Standard errors are clustered by affiliate in the OLS specifications. ***, **,
and * denote significance at the 1%, 5%, and 10% level, respectively.

40
 Table 12: Censored Fixed-Effects Estimation
(Dividend/Sales)ijct
(1) (2) (3)

DEt 0.0235** 0.0237** 0.0242**

(0.0103) (0.0109) (0.0110)
Pijct -0.0041 -0.0069 -0.0092
(0.0085) (0.0104) (0.0105)
DEt ∗ Pijct 0.0025 0.0011 0.0033
(0.0099) (0.0113) (0.0116)
D
wct -0.0990 -0.0432 -0.0445
(0.0738) (0.0669) (0.0695)
D
DEt ∗ wct -0.0792*** -0.0854*** -0.0919***
(0.0294) (0.0311) (0.0317)
R
wct -0.0196 -0.0412 -0.0412
(0.0734) (0.0818) (0.0837)
R
DEt ∗ wct -0.0431 -0.0489 -0.0404
(0.0488) (0.0539) (0.0531)
I
wct -0.1265 -0.1230 -0.0953
(0.0917) (0.1010) (0.1168)
I
DEt ∗ wct 0.0437 0.0792** 0.0748**
(0.0373) (0.0383) (0.0380)
τct 0.0037 -0.0035 -0.0045
(0.0311) (0.0338) (0.0346)
DEt ∗ τct -0.0406 -0.0310 -0.0340
(0.0282) (0.0296) (0.0296)
Exchange Ratect 0.0014 0.0142** 0.0132*
(0.0063) (0.0071) (0.0071)
GDP Growthct -0.0007*** -0.0002 -0.0004
(0.0002) (0.0002) (0.0002)
Year 2008 -0.0007 0.0000 0.0000
(0.0022) (0.0025) (0.0025)
Profitabilityijct 0.1678*** 0.1768***
(0.0264) (0.0250)
Sales Growthijct -0.0201*** -0.0198***
(0.0025) (0.0026)
(Parent Net Profit/Assets)ijc(t−1) 0.0076
(0.0216)
(Parent Debt/Assets)ijc(t−1) 0.0144
(0.0113)
Observations 60,568 51,186 49,261
Affiliate Fixed Effects Yes Yes Yes
Notes: DEt is a dummy variable that is equal to one if t ≥ 2009 and
zero otherwise. Year 2008 is the year dummy variable for 2008. See
Table 5 for the definitions of all other variables. Standard errors are in
parentheses. ***, **, and * denote significance at the 1%, 5%, and 10%
level, respectively.

41
Table 13: Censored Fixed-Effects Specifications including the Stock of Retained
Earnings (the 90th Percentile Threshold)
(Dividend/Sales)ijct
(1) (2) (3)

Rijct 0.0101 0.0166 0.0170

(0.0143) (0.0139) (0.0140)
Rijct ∗ DEt 0.0212 0.0252* 0.0250*
(0.0144) (0.0138) (0.0139)
Pijct -0.0114 -0.0085 -0.0083
(0.0092) (0.0096) (0.0097)
DEt ∗ Pijct 0.0190* 0.0118 0.0121
(0.0108) (0.0109) (0.0110)
Rijct ∗ Pijct 0.0380 0.0032 0.0046
(0.0441) (0.0423) (0.0426)
Rijct ∗ DEt ∗ Pijct -0.0008 0.0080 0.0079
(0.0501) (0.0468) (0.0473)
D
wct -0.0604 -0.0231 -0.0175
(0.0741) (0.0656) (0.0662)
D
DEt ∗ wct -0.0462 -0.0429 -0.0495
(0.0328) (0.0319) (0.0321)
D
Rijct ∗ wct 0.0274 0.0364 0.0267
(0.0720) (0.0718) (0.0717)
D
Rijct ∗ DEt ∗ wct -0.1668** -0.1676** -0.1685**
(0.0657) (0.0657) (0.0662)
R
wct -0.0432 -0.0503 -0.0517
(0.0786) (0.0783) (0.0795)
R
DEt ∗ wct -0.0274 -0.0295 -0.0268
(0.0509) (0.0503) (0.0502)
I
wct -0.1020 -0.0776 -0.0767
(0.1145) (0.1190) (0.1187)
I
DEt ∗ wct 0.0563 0.0688* 0.0656*
(0.0398) (0.0364) (0.0363)
τct 0.0342 0.0499 0.0468
(0.0375) (0.0368) (0.0371)
DEt ∗ τct -0.0234 -0.0224 -0.0236
(0.0300) (0.0287) (0.0288)
Observations 47,390 47,369 47,051
Affiliate Control Variables No Yes Yes
Parent Control Variables No No Yes
Year Dummies Yes Yes Yes
Affiliate Fixed Effects Yes Yes Yes
Notes: All specifications contain Exchange Ratect and GDP
Growthct as control variables. DEt is a dummy variable that
is equal to one if t ≥ 2009 and zero otherwise. Rijct is a
dummy variable that is equal to one if the stock of retained
earnings scaled by sales is greater than the 90th percentile
value in the sample for year (t − 1). See Table 5 for the def-
initions of all other variables. Affiliate control variables in-
clude Profitabilityijct and Sales Growthijct . Parent control
variables include (Parent Net Profit/Assets)ijc(t−1) and (Par-
ent Debt/Assets)ijc(t−1) . Standard errors are in parentheses.
***, **, and * denote significance at the 1%, 5%, and 10%
level, respectively.

42
Figure 1: Mean Dividend Payments Scaled by Sales for the Large and Small Retained
Earnings Groups

.03
Mean Dividend Payment Scaled by Sales
.005 .01 .0150 .02 .025

2007 2008 2009 2010 2011 2012 2013

Year

Large Retained Earnings Group Small Retained Earnings Group

43
Figure 2: Mean Dividend Payments Scaled by Sales for the Large and Small Retained
Earnings Groups in High or Low Withholding Tax Countries

.005 .01 .015 .02 .025 .03

Mean Dividend Payment Scaled by Sales
0

2007 2008 2009 2010 2011 2012 2013

Year
Large Retained Earnings-High Withholding Tax Large Retained Earnings-Low Withholding Tax

Small Retained Earnings-High Withholding Tax Small Retained Earnings-Low Withholding Tax

Figure 3: Graph of Mean Repatriation Tax Costs

.3 .25
Mean Repatriation Tax Cost
.1 .15 .2
.05
0

2007 2008 2009 2010 2011 2012 2013

Year

44
 Table A1: Number of Foreign Affiliates in Each Country
Country 2007 2008 2009 2010 2011 2012 2013 Total
Argentina 11 10 14 13 15 13 11 87
Australia 155 147 180 183 171 179 161 1176
Austria 17 17 18 19 20 22 17 130
Belgium 55 61 60 61 53 53 47 390
Brazil 98 112 124 126 122 131 123 836
Canada 97 108 119 103 111 107 110 755
Cayman Islands 15 17 12 9 7 11 11 82
Chile 15 19 22 19 20 23 21 139
China 2,504 2,868 3,112 3,070 3,130 3,126 3,109 20919
Czech Republic 36 34 35 36 34 35 36 246
Denmark 9 10 9 14 12 14 12 80
France 109 124 128 126 123 131 122 863
Germany 276 281 294 279 285 283 280 1978
Hong Kong 545 587 635 613 617 637 617 4251
Hungary 17 19 21 17 24 23 22 143
India 88 114 129 145 156 165 174 971
Indonesia 307 324 363 341 385 382 378 2480
Ireland 19 17 16 17 10 18 17 114
Italy 61 52 57 59 60 52 56 397
Korea, Rep. 323 331 360 381 386 389 401 2571
Malaysia 347 362 374 377 372 389 361 2582
Mexico 79 73 91 84 91 98 117 633
Netherlands 145 137 139 150 160 163 149 1043
New Zealand 29 36 39 42 40 30 28 244
Panama 69 111 102 100 95 98 86 661
Philippines 178 198 220 231 221 219 203 1470
Poland 27 30 33 29 30 29 24 202
Portugal 12 12 9 12 10 11 10 76
Russia 18 24 22 28 24 30 29 175
Singapore 488 507 493 509 519 509 486 3511
South Africa 14 15 19 19 25 20 22 134
Spain 54 55 55 56 52 51 48 371
Sweden 24 23 24 25 24 23 17 160
Switzerland 22 20 22 15 21 22 19 141
Taiwan 456 486 490 501 507 493 521 3454
Thailand 801 875 942 915 929 959 979 6400
Turkey 13 8 10 11 11 10 15 78
United Arab Emirates 18 23 26 26 27 30 29 179
United Kingdom 281 290 300 293 288 285 267 2004
United States 1,363 1,509 1,526 1,490 1,511 1,502 1,483 10384
Vietnam 166 198 234 216 232 250 254 1550
Notes: This table shows the number of foreign affiliates in each country where
information on dividends is available.

45
Table A2: Total Amount of Dividends Paid by Foreign Affiliates in Each Country
Country 2007 2008 2009 2010 2011 2012 2013 Total
Argentina 254 537 111 263 178 202 362 1907
Australia 38244 22975 45261 73528 75737 37609 56572 349926
Austria 254 198 820 818 286 653 1068 4097
Belgium 8776 4623 18386 13563 15229 13689 17049 91315
Brazil 35276 44702 19984 16407 28952 46415 35029 226765
Canada 18852 14856 4616 8033 7775 6935 35963 97030
Cayman Islands 518 1199 291 1075 1919 6360 241 11603
Chile 4260 2621 3684 1227 1841 1557 1884 17074
China 248464 211628 395335 207479 228844 291873 268910 1852533
Czech Republic 350 1482 856 1318 1276 1182 2642 9106
Denmark 190 151 78 19 34 753 920 2145
France 5182 6960 5675 2852 6673 14274 19028 60644
Germany 34338 27969 19491 19382 15994 17208 14578 148960
Hong Kong 46518 70922 46402 51985 56954 72523 86041 431345
Hungary 577 897 10 94 1287 455 469 3789
India 17593 16755 52336 4536 4533 4638 4963 105354
Indonesia 24686 41438 43226 23368 65098 54056 54097 305969
Ireland 616 488 669 390 336 23107 600 26206
Italy 2536 1844 6074 3870 6314 2905 4347 27890
Korea, Rep. 26103 21976 20409 33952 27710 33053 35404 198607
Malaysia 21602 21634 25471 17687 28140 21546 44148 180228
Mexico 7807 8199 3816 4556 5994 7741 6474 44587
Netherlands 39743 61971 13152 45194 64468 176712 138746 539986
New Zealand 1792 1361 1134 1127 509 722 756 7401
Panama 586 675 1125 3935 2356 479 721 9877
Philippines 14504 6980 8477 12999 20528 8715 17561 89764
Poland 5988 9507 1377 1473 1355 5156 402 25258
Portugal 206 98 138 442 293 35 0 1212
Russia 3160 1355 119 106 174 764 1006 6684
Singapore 78875 64243 34570 81821 92474 83043 61895 496921
South Africa 1970 1287 1054 1138 1441 1195 1204 9289
Spain 3054 1272 2467 561 7931 3348 5702 24335
Sweden 513 933 295 1815 258 251 754 4819
Switzerland 172 504 114 255 697 3988 4892 10622
Taiwan 49221 48656 34512 29397 37130 41110 48258 288284
Thailand 119200 122224 234928 150043 138659 259295 191525 1215874
Turkey 939 405 6082 383 392 126 113 8440
United Arab Emirates 2765 1710 2940 2356 2104 1605 2203 15683
United Kingdom 21814 41255 29620 12711 18058 15034 30937 169429
United States 297084 109597 335061 197554 491199 254041 293698 1978234
Vietnam 12064 15693 12790 6148 9709 8385 13714 78503

46
Table A3: Mean of Affiliate Average Tax Rates in Each Country
Country 2007 2008 2009 2010 2011 2012 2013
Argentina 0.292 0.294 0.314 0.21 0.256 0.349 0.334
Australia 0.21 0.197 0.212 0.206 0.263 0.245 0.236
Austria 0.11 0.104 0.135 0.222 0.168 0.209 0.172
Belgium 0.229 0.207 0.17 0.176 0.224 0.255 0.243
Brazil 0.274 0.287 0.24 0.25 0.281 0.274 0.269
Canada 0.258 0.233 0.218 0.235 0.235 0.214 0.211
Cayman Islands 0 0.022 0.035 0.089 0.15 0.075 0.07
Chile 0.095 0.109 0.148 0.112 0.148 0.172 0.15
China 0.093 0.102 0.105 0.152 0.175 0.185 0.189
Czech Republic 0.106 0.119 0.08 0.097 0.124 0.082 0.123
Denmark 0.186 0.18 0.14 0.138 0.184 0.186 0.215
France 0.246 0.26 0.216 0.213 0.252 0.315 0.252
Germany 0.287 0.244 0.228 0.24 0.238 0.255 0.246
Hong Kong 0.121 0.104 0.091 0.121 0.13 0.131 0.134
Hungary 0.13 0.128 0.159 0.21 0.095 0.189 0.171
India 0.278 0.257 0.213 0.235 0.204 0.185 0.192
Indonesia 0.317 0.294 0.266 0.26 0.26 0.255 0.281
Ireland 0.088 0.078 0.062 0.068 0.116 0.102 0.145
Italy 0.497 0.377 0.358 0.396 0.448 0.455 0.377
Korea, Rep. 0.188 0.167 0.132 0.157 0.175 0.158 0.167
Malaysia 0.191 0.183 0.159 0.174 0.176 0.19 0.189
Mexico 0.258 0.2 0.253 0.306 0.323 0.32 0.312
Netherlands 0.205 0.173 0.167 0.156 0.152 0.161 0.161
New Zealand 0.257 0.192 0.252 0.371 0.202 0.263 0.189
Panama 0.009 0.009 0.006 0.025 0.015 0.006 0.004
Philippines 0.226 0.221 0.184 0.206 0.224 0.23 0.217
Poland 0.137 0.143 0.122 0.184 0.234 0.203 0.131
Portugal 0.163 0.151 0.381 0.326 0.339 0.447 0.306
Russia 0.254 0.18 0.302 0.226 0.29 0.266 0.177
Singapore 0.14 0.151 0.126 0.132 0.112 0.104 0.099
South Africa 0.263 0.212 0.205 0.265 0.245 0.199 0.22
Spain 0.242 0.161 0.201 0.176 0.202 0.18 0.195
Sweden 0.202 0.212 0.145 0.137 0.207 0.238 0.257
Switzerland 0.129 0.112 0.133 0.173 0.169 0.158 0.155
Taiwan 0.235 0.238 0.234 0.193 0.177 0.195 0.183
Thailand 0.179 0.166 0.153 0.191 0.203 0.165 0.143
Turkey 0.213 0.154 0.117 0.143 0.102 0.093 0.139
United Arab Emirates 0 0.016 0.002 0 0.001 0 0.001
United Kingdom 0.212 0.192 0.189 0.207 0.202 0.2 0.208
United States 0.262 0.23 0.21 0.242 0.25 0.265 0.272
Vietnam 0.053 0.042 0.043 0.074 0.093 0.088 0.132
Notes: The sample for this table consists of foreign affiliates where information
on dividends is available. To bind the average affiliate tax rates between zero
and one, we dropped observations with negative corporate tax payments and
those with nonpositive pretax profits but positive tax payments. The average
affiliate tax rate is set equal to zero if the tax payment is zero and the pretax
profit is nonpositive, and one if the average tax rate is greater than one. The
average affiliate tax rates are also winsorized at the top and bottom 1% to
mitigate the influence of outliers.

47
Table A4: Dividend Equation Including the Stock of Retained Earnings (Scaled
by Sales)
(Dividend/Sales)ijct
(1) (2) (3) (4) (5) (6)

Rijct 0.0316*** 0.0135*** 0.0134*** 0.0036 0.0040* 0.0040*

(0.0031) (0.0028) (0.0028) (0.0024) (0.0024) (0.0024)
Rijct ∗ DEt 0.0038 0.0095*** 0.0094*** 0.0042* 0.0042* 0.0042
(0.0035) (0.0032) (0.0032) (0.0025) (0.0025) (0.0025)
Pijct -0.0841*** -0.0732*** -0.0728*** 0.0007 0.0020 0.0021
(0.0059) (0.0056) (0.0056) (0.0019) (0.0019) (0.0019)
DEt ∗ Pijct 0.0021 -0.0041 -0.0041 0.0027 0.0018 0.0019
(0.0070) (0.0066) (0.0066) (0.0022) (0.0022) (0.0022)
Rijct ∗ Pijct 0.0243*** 0.0274*** 0.0273*** -0.0038 -0.0076 -0.0077
(0.0089) (0.0084) (0.0084) (0.0059) (0.0059) (0.0060)
Rijct ∗ DEt ∗ Pijct -0.0186* -0.0193** -0.0188* -0.0042 -0.0020 -0.0019
(0.0103) (0.0097) (0.0097) (0.0066) (0.0066) (0.0066)
D
wct -0.1553*** -0.1369*** -0.1379*** -0.0268 -0.0278 -0.0274
(0.0432) (0.0408) (0.0409) (0.0238) (0.0237) (0.0238)
D
DEt ∗ wct -0.0179 -0.0229 -0.0234 -0.0204** -0.0200** -0.0223**
(0.0255) (0.0241) (0.0241) (0.0093) (0.0093) (0.0094)
D
Rijct ∗ wct 0.1155*** 0.1148*** 0.1129*** 0.0132 0.0093 0.0097
(0.0224) (0.0210) (0.0210) (0.0108) (0.0111) (0.0112)
D
Rijct ∗ DEt ∗ wct -0.0500* -0.0493** -0.0484** -0.0211** -0.0183* -0.0187*
(0.0256) (0.0240) (0.0240) (0.0102) (0.0104) (0.0105)
R
wct -0.0061 0.0139 0.0155 -0.0150 -0.0121 -0.0132
(0.0474) (0.0447) (0.0448) (0.0218) (0.0218) (0.0219)
R
DEt ∗ wct -0.0428 -0.0370 -0.0376 0.0000 -0.0014 -0.0006
(0.0379) (0.0358) (0.0358) (0.0148) (0.0147) (0.0148)
I
wct -0.0291 -0.0354 -0.0323 -0.0214 -0.0209 -0.0208
(0.0637) (0.0600) (0.0600) (0.0272) (0.0271) (0.0274)
I
DEt ∗ wct 0.0369 0.0292 0.0282 0.0158 0.0144 0.0135
(0.0303) (0.0285) (0.0286) (0.0114) (0.0113) (0.0114)
τct 0.0035 0.0296 0.0263 0.0096 0.0143 0.0136
(0.0296) (0.0280) (0.0281) (0.0112) (0.0112) (0.0113)
DEt ∗ τct -0.0155 -0.0087 -0.0078 -0.0116 -0.0116 -0.0120
(0.0205) (0.0194) (0.0194) (0.0084) (0.0083) (0.0084)
Observations 47,390 47,369 47,051 47,390 47,369 47,051
R-squared 0.0087 0.0145 0.0146
Affiliate Control Variables No Yes Yes No Yes Yes
Parent Control Variables No No Yes No No Yes
Year Dummies Yes Yes Yes Yes Yes Yes
Country & Industry Dummies Yes Yes Yes No No No
Affiliate Fixed Effects No No No Yes Yes Yes
Tobit or OLS Tobit Tobit Tobit OLS OLS OLS
Notes: All specifications contain a constant term. DEt is a dummy variable that is equal to one if t ≥ 2009
and zero otherwise. Rijct is the stock of retained earnings in year (t − 1) scaled by sales in year (t − 1).
See Table 5 for the definitions of all other variables. Affiliate control variables include Profitabilityijct
and Sales Growthijct . Parent control variables include (Parent Net Profit/Assets)ijc(t−1) and (Parent
Debt/Assets)ijc(t−1) . Standard errors are in parentheses. Standard errors are clustered by affiliate in the
OLS specifications. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively.

48
Table A5: Mean of Repatriation Tax Rates in Each Country
Country 2007 2008 2009 2010 2011 2012 2013
Argentina 0.136 0.125 0.37 0.37 0.37 0.369 0.369
Australia 0.221 0.236 0.02 0.02 0.02 0.019 0.019
Austria 0.317 0.321 0.12 0.12 0.12 0.119 0.119
Belgium 0.193 0.222 0.07 0.07 0.07 0.069 0.069
Brazil 0.158 0.161 0.02 0.02 0.02 0.019 0.019
Canada 0.17 0.198 0.07 0.07 0.07 0.069 0.069
Cayman Islands 0.407 0.391 0.02 0.02 0.02 0.019 0.019
Chile 0.336 0.316 0.37 0.37 0.37 0.369 0.369
China 0.335 0.328 0.12 0.12 0.12 0.119 0.119
Czech Republic 0.318 0.315 0.12 0.12 0.12 0.119 0.119
Denmark 0.262 0.259 0.02 0.02 0.02 0.019 0.019
France 0.185 0.183 0.02 0.02 0.02 0.019 0.019
Germany 0.147 0.198 0.17 0.17 0.17 0.169 0.169
Hong Kong 0.32 0.331 0.02 0.02 0.02 0.019 0.019
Hungary 0.309 0.305 0.02 0.02 0.02 0.019 0.019
India 0.15 0.178 0.19 0.19 0.186 0.181 0.181
Indonesia 0.132 0.154 0.12 0.12 0.12 0.119 0.119
Ireland 0.339 0.354 0.02 0.02 0.02 0.019 0.019
Italy -0.031 0.067 0.12 0.12 0.12 0.119 0.119
Korea, Rep. 0.255 0.27 0.07 0.07 0.07 0.069 0.069
Malaysia 0.251 0.261 0.02 0.02 0.02 0.019 0.019
Mexico 0.203 0.239 0.02 0.02 0.02 0.019 0.019
Netherlands 0.246 0.266 0.07 0.07 0.07 0.019 0.019
New Zealand 0.171 0.23 0.17 0.17 0.17 0.169 0.169
Panama 0.399 0.399 0.12 0.12 0.12 0.119 0.119
Philippines 0.212 0.22 0.12 0.12 0.12 0.119 0.119
Poland 0.31 0.281 0.12 0.12 0.12 0.119 0.119
Portugal 0.272 0.285 0.22 0.22 0.235 0.269 0.269
Russia 0.185 0.256 0.17 0.17 0.17 0.169 0.169
Singapore 0.302 0.294 0.02 0.02 0.02 0.019 0.019
South Africa 0.178 0.223 0.02 0.02 0.02 0.069 0.069
Spain 0.191 0.273 0.12 0.12 0.12 0.119 0.119
Sweden 0.239 0.226 0.07 0.07 0.07 0.069 0.069
Switzerland 0.306 0.32 0.12 0.12 0.12 0.019 0.019
Taiwan 0.209 0.206 0.27 0.22 0.22 0.219 0.219
Thailand 0.256 0.268 0.12 0.12 0.12 0.119 0.119
Turkey 0.254 0.279 0.12 0.12 0.12 0.119 0.119
United Arab Emirates 0.407 0.393 0.02 0.02 0.02 0.019 0.019
United Kingdom 0.224 0.242 0.02 0.02 0.02 0.019 0.019
United States 0.166 0.196 0.02 0.02 0.02 0.019 0.019
Vietnam 0.365 0.375 0.02 0.02 0.02 0.019 0.019
Notes: The sample for this table consists of foreign affiliates where information
on dividends is available. The repatriation tax rate (Repatriation Taxijct )
is the tax costs of remitting one dollar of dividends from foreign affiliate i
D
to its parent j: Pijct if t ≤ 2008 and 0.05τHt + wct if t ≥ 2009, where
Pijct = (τHt − τ̂ijct )/(1 − τ̂ijct ), τHt is Japan’s statutory corporate tax rate,
D
τ̂ijct is the affiliate average tax rate, and wct is the withholding tax rate on
dividends.

49