Repatriation Taxes, Internal Agency Conflicts and Subsidiary-level Investment Efficiency

Transcription

1 WU International Taxation Research Paper Series No Repatriation Taxes, Internal Agency Conflicts and Subsidiary-level Investment Efficiency Harald J. Amberger Kevin S. Markle David M. P. Samuel Editors: Eva Eberhartinger, Michael Lang, Rupert Sausgruber and Martin Zagler (Vienna University of Economics and Business), and Erich Kirchler (University of Vienna)

2 Repatriation Taxes, Internal Agency Conflicts, and Subsidiary-level Investment Efficiency Harald J. Amberger WU Vienna University of Economics and Business Kevin S. Markle** University of Iowa David M. P. Samuel University of Wisconsin-Madison WU Vienna University of Economics and Business November 7, 2018 We thank Tobias Bornemann, Amanda Carlson, Lisa De Simone (discussant), Scott Dyreng, Eva Eberhartinger, Alex Edwards, John Gallemore, Jeff Gramlich (discussant), Reinald Koch (discussant), Becky Lester, Dan Lynch, Stacie Laplante, Pete Lisowsky, Karsten Müller, Ben Osswald, Anh Persson (discussant), Leslie Robinson, Silke Ruenger, Dan Shaviro, Max Todtenhaupt (discussant), Aruhn Venkat, Martin Zagler, and participants at the 2018 Hawai i Accounting Research Conference, the 2018 ATA Midyear Meeting, the EAA Annual Congress 2018, the 2018 London Business School Accounting Symposium, the 8 th Conference on Current Research in Taxation in Muenster, the University of Illinois Accounting Brown Bag, the University of Wisconsin-Madison Accounting Workshop, Boston College, WU Vienna, and the Tax Reading Group at UC Irvine for helpful comments. We thank Gwen Yu for sharing the table to match Datastream data to Orbis data. Amberger thanks the University of Graz for financial support. Samuel thanks the Gies College of Business at the University of Illinois at Urbana-Champaign for their kind hospitality when working on this paper. Amberger and Samuel also gratefully acknowledge financial support from the Austrian Science Fund (FWF): W 1235-G16. ** Corresponding author. 108 John Pappajohn Business Building, W322 Iowa City, Iowa

3 Repatriation Taxes, Internal Agency Conflicts, and Subsidiary-level Investment Efficiency November 7, 2018 Abstract Using a global sample of multinational corporations (MNCs) and their foreign subsidiaries, we find that repatriation taxes impair subsidiary-level investment efficiency. Consistent with internal agency conflicts between central management of the MNC and the local management of its foreign subsidiary representing the economic channel, we find that this effect is concentrated in subsidiaries whose decision-making process has less involvement by central management, that are subject to weak monitoring, and where the MNC is flexible in deferring repatriations. We confirm our results and establish a likely causal relationship using natural experiments in the UK and Japan, which both eliminated repatriation taxes from their international tax systems in Our paper provides timely empirical evidence to inform expectations for the effects of a recent change to the U.S. international tax laws, which eliminated repatriation taxes from most of the future foreign earnings of U.S. MNCs. Keywords: repatriation tax, agency, investment, internal capital JEL classification: H21, H25, F23, G31

4 1. Introduction When the earnings of a foreign subsidiary of a multinational corporation (MNC) are repatriated as a dividend to the parent, they may be taxed in the parent s country. These repatriation taxes represent a friction that may distort the allocation of internal capital. 1 Prior research has shown that repatriation taxes affect the financing and investment choices made by the central management of the MNC. For example, the average firm holds more foreign cash (Foley et al., 2007) and engages in value-destroying foreign merger and acquisition (M&A) activity (Edwards et al., 2016; Hanlon et al., 2015; Harford et al., 2017) when its foreign earnings face repatriation taxes. In this paper, we extend the extant literature by investigating whether repatriation taxes affect the efficiency of investment decisions made by the local management of a foreign subsidiary. We find robust empirical evidence to support our hypothesis that subsidiary-level investment efficiency is decreasing in repatriation taxes. We then examine the channel through which this effect works and find that internal agency conflicts between the central management of the MNC and the local manager of the foreign subsidiary drive the effect. Absent frictions within its internal capital market, the central management of the MNC (the principal) will efficiently allocate capital by picking winners among its subsidiaries (Stein, 1997, 2003). 2 If the manager of a foreign subsidiary (the agent) is compensated based on the performance of the subsidiary, the incentives of the principal and the agent are aligned and the manager tries to maximize return in order to maximize the subsidiary s capital allocation (Stein, 2002). However, 1 Payments from a foreign subsidiary to its parent often trigger taxes levied by the subsidiary country (e.g., withholding tax), the parent country (e.g., corporate income tax), or both. Because the payment of tax is triggered by a repatriation of funds, all such taxes are referred to as repatriation taxes. In this paper, we focus on the corporate income tax that is paid by the parent in its home country when it repatriates the earnings of its foreign subsidiaries as a dividend. 2 As documented by prior research, agency conflicts between central management of an MNC and its shareholders, for example due to the CEO empire building, may cause central management to make value-destroying investments. Holding those agency conflicts constant, central management will still aim to maximize firm value when allocating capital across subsidiaries in order to maximize the value of the empire (Scharfstein and Stein, 2000). 1

5 internal agency conflicts between the central management of the MNC and the local manager of a foreign subsidiary will arise if there is information asymmetry and goal incongruence between the two (Jensen and Meckling, 1976). Where such conflicts exist, if frictions in the internal capital market of the MNC cause cash to be left in the foreign subsidiary, the manager of the foreign subsidiary has the opportunity to increase her personal welfare by, for instance, investing in lowvalue projects or by growing the subsidiary beyond its optimal size (Jensen, 1986; Harford, 1999). As repatriation taxes incentive an MNC to leave cash in its foreign subsidiaries (Foley et al., 2007), these taxes represent such a friction that results in the subsidiary receiving capital without competing for it (Beyer et al., 2017; Laplante and Nesbitt, 2017). In addition, the ability of central management to mitigate agency concerns by increasing the leverage of the subsidiary (Jensen, 1986) is confounded because repatriation taxes exacerbate the costless reduction of equity in the subsidiary. If the subsidiary manager s investment decisions are driven more by self-interest and less by the interest of central management, investment will be less efficient. As such, we predict that repatriation taxes impair investment efficiency, and that internal agency conflicts drive the effect. To test this prediction empirically and to capture subsidiary-level investment efficiency, we regress subsidiary-level investment in fixed assets (our proxy for capital expenditures) on local growth opportunities (Badertscher et al., 2013; Shroff et al., 2014). We then test whether the sensitivity of investment to growth opportunities is lower for those subsidiaries whose earnings are subject to repatriation taxes. We use a research design with extensive fixed effects that allows us to compare the investment efficiency of a foreign subsidiary whose earnings are subject to repatriation taxes to a counterfactual subsidiary in the same foreign country-industry whose earnings are not subject to these taxes (Figure 1). Thus, we identify the effect of repatriation taxes on investment efficiency from cross-subsidiary and cross-time variation in repatriation taxes 2

6 within each country-industry. We use unconsolidated financial statement data from Bureau van Dijk s Orbis database for the years 2007 to 2014 and construct a sample of operating subsidiaries in 37 countries. The subsidiaries are owned by MNCs domiciled in 56 countries, some of which levy repatriation taxes for all or part of our sample period (e.g., U.S., UK, Japan, India, and South Korea), and some that do not levy repatriation taxes throughout (e.g., Germany, France, Switzerland, and Italy). This global sample provides variation in repatriation taxes within country-industries and over time and thus constitutes a powerful setting for examining their effect on investment efficiency. We find that repatriation taxes are negatively associated with subsidiary-level investment efficiency. The effect is also economically significant: for the average subsidiary whose earnings face repatriation taxes, we observe that a one standard deviation change in local growth opportunities is associated with a 0.12 percent smaller change in investment compared to a subsidiary whose earnings are not subject to these taxes. When evaluated at the sample mean investment of 0.45 percent, 3 this effect translates into a 27.1 percent lower sensitivity of investment to local growth opportunities. We interpret this finding, which holds across multiple specifications and robustness tests, as evidence that investment efficiency is decreasing in repatriation taxes. To examine whether internal agency conflicts are the channel through which repatriation taxes affect investment efficiency, we next examine whether the effect is concentrated in subsidiaries whose decision-making processes have less involvement by central management. If central management is less involved, it is less able to monitor the subsidiary s investment decisions and internal agency conflicts are more likely to arise (e.g. Björkman et al., 2004). We split the 3 We scale investment (i.e., the annual change in fixed assets) by lagged total assets. When scaling by lagged fixed assets rather than total assets, we obtain a mean investment of 14.8 percent. 3

7 sample based on three different proxies for central management involvement. First, we divide our sample based on industry membership because a parent that operates in a different industry than its subsidiary has less knowledge of the subsidiary s operations and is less involved in the decisionmaking process. Second, we split our sample based on the nationality of the subsidiary managers. If a subsidiary s managers have a different nationality as the parent, the managers were not sent from central management leading to less involvement. Third, we divide our sample based on ownership distance from the parent. A subsidiary that has multiple ownership tiers between itself and its ultimate parent is less likely to have central management involved in its decision making. In all of these tests, we find the negative effect of repatriation taxes on investment efficiency only in the subsample of subsidiaries whose central management is less involved in the decision-making process, consistent with internal agency conflicts driving the effect. We further corroborate these results by examining whether the effect of repatriation taxes on investment efficiency is concentrated in subsidiaries that are weakly monitored by the parent. Similar to less involvement by central management, weak monitoring provides a self-interested subsidiary manager with opportunities for inefficient investment. We first split our sample into partially-owned subsidiaries and wholly-owned subsidiaries because free-riding by minority shareholders on the monitoring effort of the majority shareholder (i.e., the parent) impairs central management s monitoring incentives (Ang et al., 2000). Second, we divide our sample based on the quality of institutions and corporate governance mechanisms in the subsidiary country since lower institutional quality makes monitoring more costly. The results of these tests are consistent with our expectations: the negative effect of repatriation taxes on investment efficiency is concentrated in the subsamples of subsidiaries subject to high monitoring costs and, therefore, a low level of monitoring. This provides further evidence that the effect in our main tests is driven by internal agency conflicts. 4

8 To explore the impact of repatriation taxes across different types of MNCs, we test whether their effect depends on the MNC s flexibility to defer repatriation. We predict that the negative effect of repatriation taxes on investment efficiency will be concentrated in subsidiaries of MNCs that have greater flexibility in timing repatriations. These MNCs are more likely to hold cash abroad, which increases the opportunities for self-interested subsidiary managers to invest inefficiently. To test this, we split our sample based on the operating cash flows of the MNC; MNCs with high internal capital have more flexibility in timing their repatriations (De Simone and Lester, 2018). Consistent with this prediction, we find that repatriation taxes have a negative effect on investment efficiency only in the subsample of subsidiaries that are owned by an MNC with high flexibility. This result suggests that a lower likelihood of repatriating foreign earnings intensifies the negative effect of repatriation taxes on investment efficiency. Finally, we provide evidence that the effect of repatriation taxes on investment efficiency is likely causal by exploiting two natural experiments provided by tax reforms in the UK and Japan. Both countries repealed repatriation taxes on foreign earnings in 2009, providing an exogenous shock to repatriation taxes that allows us to conduct a difference-in-differences (DiD) analysis (Arena and Kutner, 2015; Xing, 2018). 4 We benchmark the investment efficiency of foreign subsidiaries of British and Japanese MNCs against that of foreign subsidiaries of U.S. MNCs, whose earnings are subject to repatriation taxes throughout the sample period. We again require subsidiaries to operate in the same foreign country-industry. While the level of investment decreased as a result of the reform, we find an increase in investment efficiency for subsidiaries of British and Japanese MNCs. Thus, repealing repatriation taxes improves investment efficiency, 4 The DiD design appears to provide the cleanest identification of the effects we are interested in. However, because these settings have some limitations (see Section 6.1Erreur! Source du renvoi introuvable.), we present these tests as supplemental rather than primary. 5

9 consistent with the effect found in our main tests being causal. We contribute to multiple streams of research. First, we advance research on the economic consequences of repatriation taxes (e.g. Foley et al., 2007; Nessa, 2016; Blouin et al., 2017; Gu, 2017). Prior research documents that repatriation taxes facilitate value-destroying M&A activity (Edwards et al., 2015; Hanlon et al., 2015; Harford et al., 2017), which is typically the responsibility of central management and thus susceptible to agency conflicts between an MNC s central management and its shareholders (Graham et al., 2005). Our study, in contrast, is the first to explicitly link repatriation taxes to subsidiary-level investment decisions. 5 Our findings suggest that internal agency conflicts are an economic channel through which repatriation taxes negatively affect subsidiary-level investment decisions and cause efficiency losses. In this regard, our results also offer an investment-based explanation for the valuation discount of foreign cash holdings (Yang, 2014) and variation in the earnings persistence of foreign cash changes (Chen et al., 2018). Second, we add to the literature on internal capital markets (Williamson, 1975; Shin and Stulz, 1998; Rajan et al., 2000; Stein, 2003; Glaser et al., 2013; Beyer et al., 2017). Prior research has examined how MNCs mitigate agency conflicts between the central management of the MNC and its foreign subsidiaries by strategically assigning decision rights (Antràs et al., 2009), by restricting the subsidiary s capital budget (Bernardo et al., 2004), or by drawing on external information to monitor subsidiary managers (Shroff et al., 2014). We extend these findings by showing that repatriation taxes reduce the investment efficiency of foreign subsidiaries. As the effect is concentrated in subsidiaries with less involved central management and in subsidiaries that are costly to monitor, our findings suggest that repatriation taxes induce a trade-off between 5 In supplemental tests in their respective papers, Hanlon et al. (2015) document a positive association between repatriation taxes and the level of foreign capital expenditures, and Harford et al. (2017) find that high foreign cash firms are more likely to make value-reducing foreign capital expenditures. Because both of these tests are done at the MNC level, they do not provide direct evidence of the effect or repatriation taxes on subsidiary-level investment efficiency. Our paper, therefore, contributes novel evidence for the economic channel through which repatriation taxes could lead to efficiency losses. 6

10 agency and monitoring costs. Finally, our findings inform expectations about the effects of changes to the U.S. international tax system. The recently-enacted U.S. tax law eliminates repatriation taxes on most future foreign earnings. Our results indicate that this change removes a source of agency costs borne under the previous system. Thus, U.S. MNCs, and potentially their shareholders, should benefit from efficiency gains. Host countries of subsidiaries, however, may be negatively affected by lower subsidiary-level investment when U.S. (or other) repatriation taxes are reduced. Consequently, our findings should be of interest to policymakers, both in the U.S. and abroad. 6 7 The remainder of this paper proceeds as follows: Section 2 gives an overview of related research and develops our hypothesis. In Section 3, we discuss our empirical design. We present our baseline results in Section 4, and our subsample tests in Section 5. In Sections 6 and 7, we present our additional analyses and robustness tests. Section 8 concludes. 2. Prior Research and Hypothesis Development 2.1 Economic Effects of Repatriation Taxes The home country of an MNC has the right to levy domestic tax on the earnings of foreign subsidiaries. All countries set international tax laws that fall somewhere on the spectrum between 6 On December 22, 2017, President Trump signed the Tax Cuts and Job Act (TCJA) into law. Provisions in the new law reduce most repatriation taxes on existing earnings held abroad, and eliminate U.S. tax on future foreign earnings of U.S. MNCs. Our empirical results, however, remain relevant to policymakers, both as a basis for prediction of the effects of the law change, and because repatriation taxes will remain a salient and important fiscal tool for tax authorities under all international tax regimes. This is so for two reasons. First, withholding taxes on cross-border payments of dividends, interest, and royalties share the same character as repatriation (income) taxes in that they can be deferred. Second, repatriation taxes are a common tool used in the base erosion prevention measures implemented by countries (e.g., as an element of Controlled Foreign Corporation (CFC) Rules), so significant portions of foreign earnings are likely to remain subject to repatriation taxes. 7 The expected negative effect of the TCJA on subsidiary-level investment could be partly offset by the Global Intangible Low Taxes Income (GILTI) and Foreign Derived Intangible Income (FDII) provisions. GILTI is a U.S. tax on excess foreign return that could incentivize capital expenditures abroad while FDII is a U.S. tax deduction for excess domestic return that could discourage domestic investment. Similarly, the Base Erosion Anti-Abuse Tax (BEAT) provision, which is a minimum tax on intra-firm payments to non-u.s. subsidiaries, could provide an incentive to relocate production facilities abroad. 7

11 full exemption (i.e., the home country exempts foreign earnings from domestic tax) and full double-taxation (i.e., the home country immediately levies domestic tax on foreign earnings and allows no credit for the foreign taxes paid). Those that are closer to full exemption are usually grouped in a territorial category, and those nearer the other end are grouped in a worldwide category. 8 In reality, exceptions and provisions in the tax regimes of all major countries result in them falling at different points all along the spectrum. For example, until 2018, the U.S. used a worldwide tax system, but allowed the repatriation tax (i.e., the U.S. tax liability on foreign earnings, net of credit granted for foreign taxes paid) to be deferred until foreign earnings were repatriated to the U.S. parent as a dividend. In contrast, when the home country has a territorial tax system and fully exempts foreign earnings from domestic tax, the repatriation tax is zero. 9 Several studies suggest that repatriation taxes provide an incentive for MNCs to defer repatriation and to hold cash abroad. 10 Foley et al., (2007), for instance, document that repatriation taxes lead to higher foreign cash holdings of U.S. MNCs. This effect became stronger when Congress started deliberating another repatriation tax holiday in 2008 (De Simone et al., 2017). Laplante and Nesbitt (2017) examine motives to hold cash abroad and find that repatriation taxes among other reasons drive the amount of cash held abroad. These taxes account for 42 percent of the cash differential between U.S. MNCs and domestic firms (Gu, 2017). 8 Territorial tax systems are also referred to as exemption or source-based systems. Worldwide tax systems are also referred to as credit or residence-based systems. We use the terms territorial and worldwide in this paper. 9 Some countries, such as Italy or Germany, exempt 95 percent of foreign earnings (i.e., tax 5 percent). In addition, several countries impose repatriation taxes when certain conditions are met. France, for instance, taxes all foreign earnings that are repatriated from a CFC located in a country with an effective tax rate that is 50 percent lower than the current French corporate income tax rate of percent. In such a case, France grants a credit for foreign taxes paid, which resembles a worldwide system. We follow Markle (2016) and treat the worldwide/territorial distinction as binary by classifying a country as territorial if it exempts 95 percent or more of foreign earnings. 10 Hartman (1985) shows theoretically that repatriation taxes do not affect an MNC s decision to repatriate foreign earnings when the tax is constant over time and all foreign earnings will eventually be repatriated. In reality, repatriation taxes do affect repatriation decisions because expected repatriation taxes vary over time (e.g., due to tax holidays; see Altshuler et al., 1994; De Waegenaere and Sansing, 2008). Moreover, an MNC may be able to use foreign earnings for domestic purposes without triggering repatriation taxes (e.g., by domestically borrowing against passive assets held by a foreign subsidiary; see Altshuler et al., 1994). 8

12 Deferring repatriation of foreign earnings could also lead to a financial accounting benefit. U.S. GAAP Accounting Principles Board Opinion No 23 states that firms do not have to accrue a deferred tax expense for foreign earnings deemed as permanently reinvested. Although increasing pre-tax income, firms do not record a U.S. tax expense on these earnings, which provides an additional incentive to defer repatriation (Graham et al., 2011). 11 Holding cash abroad to avoid repatriation taxes, however, creates internal capital market frictions (Beyer et al., 2017). Campbell et al. (2014) find that investors place a valuation discount on foreign cash holdings. In examining the sources of this discount, Harford et al., (2017) show that a combination of repatriation taxes, internal financing frictions, and agency costs contribute to the lower value of foreign cash holdings. This result is consistent with Yang (2014), who reports a lower marginal value for foreign compared to domestic cash, and Chen et al. (2018), who find a lower earnings persistence of cash changes whenever foreign growth opportunities are low and repatriation taxes are high. The effect of repatriation taxes on investment decisions has been predominantly examined in the context of M&A decisions made by the central management of an MNC. 12 Building on a model by Klassen et al. (2014), Edwards et al. (2016) find that repatriation taxes reduce the investment opportunity set of MNCs, reducing the profitability of foreign M&A deals. 13 Similarly, Hanlon et al. (2015) document that repatriation taxes lead to a higher likelihood of acquiring 11 A similar rule exists in IAS 12, which is the prevailing accounting standard for MNCs resident outside the U.S. 12 Another stream of research suggests that repatriation taxes might adversely affect total payouts, external financing, and total investment. Nessa (2016) reports a negative effect of repatriation taxes on payouts which is concentrated among MNCs that are unable to distribute dividends without triggering repatriation taxes. Similarly, Arena and Kutner (2015) study the repeal of repatriation taxes in the UK and Japan in 2009 and find that foreign cash holdings decreased and MNCs initiated larger total payouts to shareholders after the reform. Ma et al. (2017) examine consequences for external financing and find that repatriation taxes are associated with higher loan spreads. 13 Edwards et al. (2016) suggest that, under certain circumstances, lower-return acquisitions made with foreign cash holdings might be economically optimal for the MNC as a whole because investing the pre-repatriation-tax earnings abroad leads to a higher return than investing the after-repatriation-tax earnings domestically. 9

13 foreign rather than domestic targets where shareholders react negatively to the announcement of foreign deals. Harford et al. (2017) show that MNCs subject to high repatriation tax costs incur negative capital-market reactions when announcing foreign capital expenditure and acquisition plans. Both Hanlon et al. (2015) and Harford et al. (2017) speculate that agency conflicts between shareholders and the central management over how to deploy foreign cash holdings may be causing the negative investor responses, but neither study tests this assertion empirically. Complementing these studies of foreign acquisitions, Blouin et al. (2017) examine the effect of repatriation taxes on domestic investment. Their results suggest that repatriation taxes reduce the sensitivity of domestic investment to domestic growth opportunities Agency Conflicts and Investment Decisions In the absence of agency conflicts, firm-level investment is a function of the ratio between the market value of assets and their replacement costs (Tobin, 1969) and managers invest until the marginal benefit of investment equals the marginal cost (Yoshikawa, 1980; Hayashi, 1982; Abel, 1983). Consequently, managers invest exclusively in projects with positive net present value and return excess cash to their capital providers. Such investment behavior maximizes shareholder value and is thus considered economically efficient. Information asymmetry and goal incongruence between the central management of an MNC and the local manager of a foreign subsidiary can create internal agency conflicts in the form of adverse selection or moral hazard (Desai et al., 2007) that distort the internal capital allocation (Stein, 1997) and affect the investment decisions of a foreign subsidiary. For instance, a selfinterested subsidiary manager could hold back private information to increase her subsidiary s 14 A related stream of research examines the effect of repatriation taxes on the volume and direction of M&A. Feld et al. (2016) find that the repeal of repatriation taxes in Japan and the United Kingdom led to an increase in outbound acquisitions of firms located in either of these countries. Similarly, Bird et al. (2017) show that U.S. targets with sizable cash holdings are more likely to be acquired by foreign MNCs not subject to repatriation taxes. 10

14 capital allocation. More internal capital allocated to the subsidiary provides the subsidiary manager with opportunities to increase her personal welfare by investing in low-value projects or by growing the subsidiary beyond its optimal size (Jensen, 1986). Similarly, the subsidiary manager could enjoy a quiet life (Bertrand and Mullainathan, 2003) as competing in the internal capital allocation process becomes less necessary. Informational frictions within an MNC also impair central management s ability to verify information revealed by the subsidiary manager and to assess whether the subsidiary is engaging in value-maximizing investment. Consistent with this reasoning, Shroff et al. (2014) find that foreign subsidiaries operating in more transparent country-industries invest more efficiently. Thus, external information alleviates internal agency conflicts by enabling central management to monitor the investment decisions of its subsidiaries. De Simone et al. (2018) find that investment efficiency is decreased when the subsidiary is part of an MNC that engages in aggressive taxmotivated income shifting impairing the internal information environment. In summary, the two relevant streams of research show that repatriation taxes can negatively affect firm outcomes, and that an MNC s investment decisions can be susceptible to agency conflicts. We bring these two streams together in developing our hypothesis in the next section. 2.3 Baseline Hypothesis: Repatriation Taxes and Investment Efficiency In expectation, deferring the repatriation of foreign earnings reduces the present value of repatriation tax payments by exploiting variation in tax rates over time due to tax holidays or tax reform. The choice to leave cash abroad, however, provides a subsidiary with capital without its managers having to compete for it in the internal capital market. In addition, repatriation taxes exacerbate the costless reduction of equity in the subsidiary. This increases monitoring costs for the central management as alleviating agency concerns by increasing the leverage of the subsidiary 11

15 is more difficult (Easterbrook, 1984; Jensen, 1986). As a result, subsidiary managers face greater opportunities to invest cash in projects that create personal benefits while being of low value to shareholders (Harford, 1999). Blanchard et al. (1994) provide empirical evidence that this occurs by showing that managers who receive a cash windfall maximize personal welfare by investing in economically suboptimal projects rather than distributing excess cash to shareholders. In the presence of internal agency conflicts, the investment decisions of a subsidiary manager are driven more by self-interest than by the interests of central management. Thus, subsidiary managers reap personal benefits by investing cash retained in the subsidiary due to repatriation taxes in projects that do not maximize shareholder value, leading to less efficient investment. In contrast, for an MNC that does not incur repatriation taxes when bringing foreign earnings back to the parent (as under a territorial tax system), repatriations bear no additional (tax) cost, which reduces the opportunities for a subsidiary manager to reap personal benefits. Given these differences, we expect that investment efficiency will be lower if a subsidiary s earnings are subject to repatriation taxes. This leads to our baseline hypothesis, stated in the alternative: H1: Subsidiary-level investment efficiency is decreasing in the repatriation taxes on a foreign subsidiary s earnings. There is support for the null hypothesis because repatriation taxes are triggered when foreign earnings are returned to the parent. Repatriation taxes are generally not paid when earnings remain foreign or are moved to another subsidiary. To the extent that central management is able to efficiently reallocate cash among its subsidiaries, we may not observe the hypothesized effect. 15 Similarly, since we argue that agency conflicts drive the effect on investment efficiency, the 15 The Tax Increase Prevention and Reconciliation Act of 2005 (TIPRA), for instance, enables U.S. MNCs to relocate foreign cash holdings to foreign subsidiaries where internal agency conflicts a less likely to arise without triggering repatriation taxes (Murphy, 2017). Since our data covers financial years as of 2006 (see Section 3), we are unable to test whether this reform had a mitigating effect in our setting. 12

16 relation will not hold if the parent is able to effectively monitor the subsidiary s investment decisions. In this regard, Shroff et al. (2014) show that external information facilitates monitoring and Bloom et al. (2012) find that improved information technology reduces information asymmetries. If such mechanisms effectively alleviate information asymmetries, we may not find an effect of repatriation taxes on subsidiary-level investment efficiency. 3. Research Design, Data, and Sample Selection 3.1 Research Design To test whether investment efficiency is decreasing in repatriation taxes, we estimate the following subsidiary-level OLS regression:!"#$%&'$"& (,* =, - +, / +, 0 +, * + 3 -,/, -,/ 56 -,/,* ,/,* $:;&<;= -,>,* + 3? 9$:;&<;= -,>,* 56 -,/,* + (1) (,* + 3 K 5;F$"&HI"&FIJ% K 0,* + L (,* where i is the subsidiary. t is the year. c is the subsidiary country. j is the subsidiary industry. p is the parent. g is the parent country.!"#$%&'$"& (,* is the subsidiary s annual change in net fixed assets scaled by lagged total assets ,/,* is the price-to-earnings ratio (PE-ratio) in the subsidiary s country-industry-year (one-digit ICB industry). 9$:;&<;= -,>,* is a measure for repatriation taxes, calculated as: 1. (continuous) the difference in statutory corporate tax rates between countries g and c. 16!"#$%&'$"& (,* measures the change in net fixed assets from year t-1 to t and thus resembles a subsidiary s capital expenditures net of annual depreciation charges (net investment in fixed assets). In contrast to Shroff et al., 2014 (2014), we do not use annual changes in total assets as our dependent variable because, consistent with prior research on investment efficiency (Biddle et al., 2009; Biddle and Hilary, 2006), we are interested in real investment of a subsidiary, which is more directly measured with changes in net fixed assets. 13

17 2. (indicator) 1 (0) if country g operates a worldwide (a territorial) tax system in year t. 17 SubsidiaryControls is a vector of subsidiary-level controls. ParentControls is a vector of parent-level controls., - are subsidiary country fixed effects., / are subsidiary industry fixed effects., 0 are parent fixed effects., * are year fixed effects., -,/ are subsidiary country-industry fixed effects. Appendix A provides an overview of all variable definitions. Equation (1) is a regression of investment on local growth opportunities (e.g. Badertscher et al., 2013; Shroff et al., 2014), in which 3 7 captures the sensitivity of subsidiary-level investment to local growth opportunities (investment efficiency). Consistent with efficient investment responding to local growth opportunities, we expect 3 7 to be positive. 18 To draw conclusions about the effect of repatriation taxes on investment efficiency, we extend this model by adding 9$:;&<;= -,>,* and interacting it with 56 -,/,*. Our coefficient of interest is 3?, which captures the investment efficiency of a subsidiary whose earnings are subject to repatriation taxes relative to a counterfactual whose earnings do not bear these taxes. A positive (negative) coefficient on 3? indicates higher (lower) investment efficiency. Consistent with our prediction that repatriation taxes lead to investment that is less aligned with local growth opportunities, we expect 3? to be negative. We include a series of fixed effects in Equation (1). First, we add interactions of fixed effects 17 We collect data on tax systems and corporate tax rates from EY Corporate Tax Guides. The continuous variable is equal to zero if the difference is negative or if the parent resides in a territorial tax system. For the indicator variable, we follow Markle (2016) and classify the tax system in which an MNC resides as territorial or worldwide. 18 We include interactions between each country-industry fixed effect and 56 -,/,*. As 3 7 captures the relation between investment and local growth opportunities for the country-industry that is excluded from the regression, we do not tabulate and interpret 3 7 in our main tests. However, we test for and find a positive relation between investment and local growth opportunities in our sample (see Column 1 of Table 4). 14

18 for each subsidiary country-industry (, -,/ ) with 56 -,/,*. 19 This approach controls for time-invariant country-industry characteristics, such as financial-sector development or the strength of property rights (Claessens and Laeven, 2003) that could affect the investment efficiency of all subsidiaries in a given country-industry. Further, including these fixed effects rules out the possibility that country-industry-level measurement error in 56 -,/,* might bias investment efficiency (Erickson and Whited, 2000). As we allow the sensitivity of subsidiary-level investment to local growth opportunities to vary by country-industry, we identify the effect of repatriation taxes on investment efficiency from cross-subsidiary and cross-time variation within each country-industry. For the example outlined in Figure 1, 3? captures the investment efficiency of a Polish subsidiary of a U.S. MNC relative to a Polish subsidiary of an Austrian MNC in the same industry while controlling for the effect of unobserved country-industry characteristics. Second, we include fixed effects for each subsidiary country (, - ) to control for countrylevel factors, such as legal or regulatory regimes, and for each subsidiary industry (, / ) to control for industry-level factors, such as investment adjustment costs. Third, we add parent fixed effects (, 0 ) to capture time-invariant characteristics of the MNC. By including these fixed effects, we control for the investment effects of agency conflicts between an MNC s central management and its shareholders, which prior research has argued drive the relation between repatriation taxes and value-destroying M&A activity (Hanlon et al., 2015). Fourth, we add year fixed effects (, * ) to absorb the effect of economic shocks or the business cycle on the investment behavior of subsidiaries. Vector SubsidiaryControls captures subsidiary-level characteristics associated with 19 The subsidiary country-industry fixed effects are based on one-digit ICB industries. 15

19 investment (e.g., Cummins et al., 1996; Baker et al., 2003). We include the subsidiary s return-onassets (9IO (,* ) to control for internally generated funds available for investment (Faulkender and Petersen, 2012). ADP$ (,* is the natural logarithm of total assets and captures differences in investment opportunities (Carpenter and Petersen, 2002) as well as in the allocation of decision rights between the parent and the subsidiary (Robinson and Stocken, 2013). We include <;"QDCDJD&G (,* to control for the stock of fixed assets (Biddle and Hilary, 2006). Finally, we add the total amount of credit provided by the banking sector as a percentage of GDP in the subsidiary country (RI'$%&DSHF$ED& -,* ) to control for bank monitoring (Shroff et al., 2014). Vector ParentControls includes parent-level characteristics that might affect subsidiarylevel investment. We include <I&;J5;F&DSD:;&DI" 0 as the sum of direct and indirect participation of the parent in the subsidiary. Parents set their ownership in subsidiaries to align incentives between the central management and subsidiary managers and to facilitate monitoring (Antràs et al., 2009). We also include the parent s cash-flow-to-total-assets ratio (H;%hUJIV5;F$"& 0,* ) as MNCs use internal capital markets to finance the investments of foreign subsidiaries (Shin and Stulz, 1998; Arena and Kutner, 2015) Data and Sample We collect subsidiary-level unconsolidated financial statement data and parent-level consolidated financial statement data from Bureau van Dijk s Orbis database. Our dataset covers the years 2006 to As several variables, such as!"#$%&'$"& (,*, are annual changes, our final sample effectively covers the years Following Shroff et al., (2014), we re-construct an 20 In supplemental tests, we include the subsidiary s cash ratio as a proxy for internally generated funds available for investment. One concern with adding this variable to the baseline model is that the cash ratio could be a mediator or collider control (Gow et al., 2016; Swanquist and Whited, 2018) because subsidiaries whose earnings are subject to repatriation taxes at the same time report higher cash holdings. To address this concern, we run our regressions without control variables and find similar results (untabulated). 16

20 MNC s holding structure and identify directly- and indirectly-held subsidiaries using ownership information available in Orbis. 21 Indirect shareholdings in our sample are subsidiaries held by intermediate subsidiaries that are located across up to four different countries. As repatriation taxes only apply in a cross-border context, we drop domestic subsidiaries. 22 Further, we require a parent to hold a total participation in a subsidiary of more than 50 percent to ensure that central management has control over a subsidiary and its decision to distribute a dividend. 23 We follow prior research and exclude subsidiaries in the financial and utility sectors as well as non-operating financial holdings due to unique investment patterns (e.g. Badertscher et al., 2013). Moreover, we require non-missing values for all variables in Equation (1). To avoid denominator effects and to mitigate the influence of outliers in our dataset, we require total assets, operating revenue, and fixed assets of at least US$10,000 and eliminate observations in the bottom and top 1.5 percent of the variable distribution (similar to Becker et al., 2013). 24 We obtain a final sample of 48,470 subsidiary-years. Depending on the fixed-effects structure applied and data available for our subsample tests, the sample size varies slightly across specifications. Appendix B provides an overview of the sample selection procedure. To measure local growth opportunities, we follow Bekaert et al. (2007) and Shroff et al. (2014). We collect Datastream s equity indices and construct aggregate PE-ratios per country- 21 One limitation of the Orbis database is that ownership information is stale and reflects the status of the last year in the dataset. This feature might lead to measurement error as we could classify a firm that was acquired by an MNC in the final year of the dataset as being a foreign subsidiary throughout the entire sample period. As repatriation taxes only apply in a cross-border context, such ownership changes would bias against finding a significant effect of repatriation taxes on investment efficiency. Consequently, the effect size obtained from our baseline regression model likely constitutes a lower bound estimate. 22 Data on domestic subsidiaries of U.S. MNCs is not available so that we cannot test investment behavior of U.S. foreign subsidiaries against their domestic counterparts. 23 Our results are similar when requiring lower thresholds for a parent s total participation. 24 We randomly checked outliers and found that they either result from obvious errors in the database or from one-time changes in the subsidiary s asset structure likely initiated by the central management (e.g., initial investments or final disinvestments). Because we are interested in continuous investment decisions made by a subsidiary s local management, we truncate rather than winsorize. 17

21 industry-year in which we observe subsidiary-level investment. 56 -,/,* is an intuitive measure for local growth opportunities as a higher ratio of share-price-to-earnings for firms in an industry suggests that investors expect stronger industry-level growth. Another benefit of using countryindustry-year-level PE-ratios is that they are exogenous to each individual subsidiary in our sample. As PE-ratios require data of public (listed) firms in a country-industry-year, the private (unlisted) subsidiaries in our sample do not enter the calculation of 56 -,/,*. 3.3 Descriptive Statistics Table 1 reports the number of parent-year and subsidiary-year observations in our sample by country. We observe the largest number of parents residing in large, developed countries, such as the U.S., Japan, Germany, France, Sweden, and the UK. We observe similar variation in the distribution of subsidiaries, with the UK, France, Germany, Italy, and Spain contributing the most observations. A sizable proportion of subsidiaries, however, reside in Eastern Europe countries, such as in Poland and the Czech Republic, and in Asian countries, such as South Korea and China. Table 2 presents descriptive statistics. In Panel A, we present information for the full sample. The mean annual change in fixed assets amounts to 0.45 percent of total assets. The average PEratio is 17.8, consistent with the values reported in Shroff et al., (2014). On average, subsidiaries report a return-on-assets of 5.3 percent, total assets of US$64 million, and hold 23.6 percent of their total assets in fixed assets. The average amount of credit provided by the banking sector is percent of GDP. With respect to parent-level controls, we find that the average cash-flowto-assets ratio is 8.5 percent and parents hold an average total participation of 95.1 percent in their subsidiaries. In Panels B and C, we split the sample based on whether the subsidiary s earnings are subject to repatriation taxes. We test for differences between subsamples and do not find differences in 18

22 means for our dependent variable (!"#$%&'$"& (,* ), the PE-ratio (56 -,/,* ), and the total participation of the parent (<I&;J5;F&DSD:;&DI" 0 ). For the remaining variables, means differ between subsamples, which indicates the need for controls in our multivariate analysis. In Panel D, we present information on the group structures of the MNCs in our sample. In total, our sample includes 10,629 unique subsidiaries that are controlled by 2,714 unique parents. Each parent in our sample holds, on average, 3.9 subsidiaries, which is equivalent to 17.9 subsidiary-year observations per parent. In Panel E, we show information on 9$:;&<;= -,>,* by subsidiary country-industry. Consistent with the industry classification used to determine 56 -,/,*, we present this information for one-digit ICB industries. Values for means and standard deviations suggest that 9$:;&<;= -,>,* generally varies within country-industries, which is the result of i) the tax systems operated by the parent countries and ii) differences in the corporate income tax rates between parent and subsidiary countries in our sample. Tax systems of the parent countries vary due to tax reforms during our sample period. 25 The difference in statutory corporate tax rates varies because of tax rate changes in either the parent or the subsidiary country. The mean within-country-industry repatriation tax is 2.51 percent while the average standard deviation is equal to 3.64 percent. We use the fixedeffects structure described above to exploit this variation in our research design. 4. Empirical Results 4.1 Repatriation Taxes and Subsidiary Cash Holdings In order for the subsidiary manager to have the opportunity to make self-interested 25 The UK and Japan both switched from a worldwide to a territorial tax system and thus repealed repatriation taxes in We exploit these tax reforms as natural experiments in our supplemental tests (see Section 6). 19

23 investment decisions, at least some of the earnings must be retained in the subsidiary as cash. Prior research has shown that aggregate foreign cash holdings are associated with repatriation taxes (Foley et al., 2007), but it has not been shown that the effect manifests in operating subsidiaries. If central management is able to reallocate cash among its foreign subsidiaries, it is possible that excess cash will not be available to the managers of operating subsidiaries whose earnings are subject to repatriation taxes. To examine this, we plot the mean subsidiary cash ratio in year t conditional on whether the subsidiary s earnings are subject to repatriation taxes. Figure 2 suggests that subsidiaries whose earnings are subject to repatriation taxes hold a significantly higher amount of cash than subsidiaries whose earnings are not subject to these taxes. The difference in cash holdings between the two groups ranges from 1.5 to 4 percent of total assets and is statistically significant in all sample years (all p < 0.01). We supplement this graphical evidence with a multivariate test. We estimate Equation (1) using the subsidiary cash ratio (H;%h9;&DI (,* ) as a dependent variable. As reported in Table 3, Column 1, we find that the average subsidiary whose earnings are subject to repatriation taxes has a 0.64 percentage point higher cash ratio than a subsidiary whose earnings are not subject to repatriation taxes. 26 When evaluated at the mean cash ratio of percent of total assets (Panel A of Table 2), this effect is equal to 4.6 percent higher cash holdings (= 0.64/13.92). In Column 2, we modify our regression and replace the fixed-effects structure in Equation (1) with subsidiary country-industry-year fixed effects. We still find a positive and significant coefficient on 9$:;&<;= -,*. 27 These tests show that central management does not fully reallocate foreign cash 26 We calculate this effect using the mean 9$:;&<;= -,>,* of subsidiaries whose earnings are subject to repatriation taxes (7.86 percent; see Panel C of Table 2): x 7.86 = percentage point higher cash ratio. 27 In Equation (1), we include country-industry fixed effects instead of country-industry-year fixed effects because 56 -,/,* is measured at the country-industry-year level. In this test, we are able to include country-industry-year fixed effects because 9$:;&<;= -,* is a subsidiary-level measure that varies within each country-industry-year. 20

24 holdings but leaves at least some of the cash in operating subsidiaries whose earnings are subject to repatriation taxes. 4.2 Growth Opportunities and the Level of Investment Before testing our baseline hypothesis, we next examine whether the level of subsidiary investment is positively associated with the 56 -,/,*, our measure for local growth opportunities. If the PE-ratio captures local growth opportunities (Baker et al., 2003; Shroff et al., 2014), we expect to observe a positive relation between investment and 56 -,/,*. We estimate Equation (1) without 9$:;&<;= -,* and the interactions of country-industry fixed effects with 56 -,/,* in order to obtain a coefficient estimate for 56 -,/,*. As expected, the coefficient on 56 -,/,* in Column 1 of Table 4 is positive and significant, consistent with the PE-ratio of the subsidiary s country-industry-year being a valid proxy for the subsidiary s local growth opportunities. 4.3 Repatriation Taxes and Investment Efficiency We next estimate Equation (1) to test our baseline hypothesis. In Column 2 (3) of Table 4, we include our binary (continuous) measure for repatriation taxes. 28 The coefficient on 9$:;&<;= -,>,* is insignificant for both measures, indicating that subsidiaries whose earnings are subject to repatriation taxes do not exhibit a different level of investment than subsidiaries in the same country-industry whose earnings do not face these taxes. However, the coefficients on the interaction of repatriation taxes and growth opportunities (9$:;&<;= -,>,* 56 -,/,* ) are negative and significant when using the binary measure in Column 2 and the continuous measure in Column 3. These results suggest that the investment behavior of a subsidiary whose earnings are 28 As we rely on the tax system of the parent country to construct the binary measure for 9$:;&<;= -,>,*, variation in repatriation taxes within an MNC is limited and we do not include parent fixed effects in this test. 21

25 subject to repatriation taxes is less in line with local growth opportunities, i.e. less efficient. In economic terms, the coefficient on the interaction (9$:;&<;= -,>,* 56 -,/,* ) in Column 3 suggests that, compared to a subsidiary whose earnings are not subject to repatriation taxes, a one standard deviation change in local growth opportunities is associated with a 0.12 percent smaller change in investment for the average subsidiary whose earnings are subject to these taxes. 29 When evaluated at the sample mean investment of 0.45 percent, this effect is equal to a 27.1 percent lower sensitivity of investment to local growth opportunities. Taken together, these results support our hypothesis that investment efficiency is decreasing in repatriation taxes. 5. Subsample Results 5.1 Identifying Internal Agency Conflicts as a Channel for Lower Investment Efficiency We next provide support for our assertion that internal agency conflicts drive this relation by testing whether the effect is concentrated in subsidiaries susceptible to these conflicts. First, we split our sample based on the extent to which central management is involved in the decision making of the subsidiary. Second, we split our sample based on the level of monitoring by central management Central Management Involvement Prior research shows that when central management is less involved in decision-making process of a foreign subsidiary, it is less able to monitor the subsidiary s investment decisions (e.g., Grinblatt and Keloharju, 2001). A self-interested subsidiary manager, in turn, is more likely 29 We calculate this effect based on the mean 9$:;&<;= -,>,* of subsidiaries whose earnings are subject to repatriation taxes (7.86 percent; see Panel C of Table 2) and the standard deviation of 56 -,/,* for the full sample (7.81; see Panel A of Table 2): x 7.89 x 7.86 = percent smaller change in investment. The mean 9$:;&<;= -,>,* of subsidiaries whose earnings are not subject to repatriation taxes is 0 by definition. 22

26 to consume private benefits and to invest inefficiently when she is monitored less strongly. To test this empirically, we split our sample based on three different proxies for central management s involvement in the decision making of a subsidiary (Björkman et al., 2004). First, we split our sample into subsidiaries that operate in an industry different from their parent s and those that operate in their parent s industry. Central management of a parent operating in a different industry has less knowledge about the subsidiary s operations and investment opportunities (Goodman et al., 2014), and is, therefore, less involved in the investment decisions of the subsidiary. To identify industry affiliation, we use 1-digit NACE industry codes available in the Orbis database. Second, we divide the sample based on the nationality of the subsidiary s management team. Expatriate managers that have been sent from the parent to the foreign subsidiary diminish crossborder frictions, such as language or cultural differences, which improves information transfer from the subsidiary to the central management. Thus, if the parent does not send managers to the subsidiary, we expect central management to be less involved in the subsidiary s decision-making process. We exploit manager-level information provided by the Orbis database and classify the central management as less involved if the number of a subsidiary s managers from the parent country is below the MNC-wide median. 30 Third, we split the sample into subsidiaries that are owned directly by their parents and those that are owned directly. If a parent directly controls a foreign subsidiary (first-tier subsidiary), cross-border frictions are low, and information transfer from the subsidiary to the parent should be smooth. In contrast, if a subsidiary has multiple tiers between itself and the parent, cross-border 30 Orbis provides manager-level information for the board of directors, senior managers, and executives. We use information available for 45,083 subsidiary-years in our sample to identify expatriate managers. We classify a manager as expatriate if she has the same nationality as the parent and count the number of expatriate managers for each subsidiary. Based on that, we compute the MNC-wide median number of expatriate managers per subsidiary. 23

27 frictions increase, and central management will be less involved in the subsidiary s investment decisions. Empirically, we classify a subsidiary as first tier if the parent holds a direct participation of more than 50 percent. We present results in Table 5. In line with our predictions, coefficients on 9$:;&<;= -,>,* 56 -,/,* are negative and significant for subsamples with lower central management involvement and insignificant if the parent is strongly involved in the subsidiary s decision-making process. These results hold for splitting our sample based on industry affiliation (Columns 1 and 2), the nationality of the subsidiaries management team (Columns 3 and 4), and whether subsidiaries are held directly by their parents (Columns 5 and 6). Overall, these findings indicate that the negative effect of repatriation taxes on investment efficiency is concentrated in foreign subsidiaries where central management is less involved in the subsidiary s decision-making process. We interpret these results as evidence that internal agency conflicts are a channel through which repatriation taxes lead to lower investment efficiency Degree of Monitoring Similar to involvement in the subsidiary s decision-making process, weak monitoring by central management also increases opportunities for inefficient investment (e.g., Bloom et al., 2012; Shroff et al., 2014). In this section, we examine two settings in which we expect high monitoring costs and, therefore, a low level of monitoring. If lower investment efficiency is the result of internal agency conflicts, the relation between repatriation taxes and investment efficiency should be concentrated in subsidiaries that are weakly monitored by central management. First, we explore partially-owned subsidiaries because central management faces lower incentives to monitor the investment decisions of these subsidiaries. The presence of minority shareholders in these subsidiaries reduces central management s incentives to monitor since 24

28 minority shareholders free ride on the monitoring effort of the majority shareholder (i.e., the parent) (Ang et al., 2000). We classify a subsidiary as partially-owned if the total shareholding percentage of the parent is less than 100 percent. Second, we examine the quality of institutions and corporate governance mechanisms of the subsidiary country as they shape monitoring costs (Asiedu and Esfahani, 2001). If the subsidiary is located in a country with low-quality institutions and corporate governance mechanisms, central management faces high monitoring costs, leading to weaker monitoring of the subsidiary s investment decisions. To construct a country-level measure, we draw on the 2015 Control of Corruption Index from the World Bank s Worldwide Governance Indicators Project. We classify subsidiary countries with an index score below the sample median as having low-quality institutions and corporate governance mechanisms. 31 We present regression results in Table 6. Consistent with our expectations, the coefficients on 9$:;&<;= -,>,* 56 -,/,* are negative and significant in subsamples with weak monitoring. In contrast, we find insignificant coefficients on 9$:;&<;= -,>,* 56 -,/,* for wholly-owned subsidiaries and for subsidiaries located in countries with high-quality institutions and corporate governance mechanisms. These results indicate that weak monitoring by the central management drives the association between repatriation taxes and investment efficiency. In other words, central management trades off the cost of internal agency conflicts with the costs of monitoring. Collectively, the results in this section are consistent with an agency-based explanation. Thus, internal agency conflicts between the central management of an MNC and the local 31 For more information, please refer to: We lose some observations as the measure is not available for all countries of our sample. Low corporate governance countries are: Bulgaria, Croatia, Czech Republic, Greece, Hungary, Italy, Malta, Poland, Romania, Russia, Slovenia, Spain, and Turkey. High corporate governance countries are: Austria, Belgium, Cyprus, Denmark, Finland, France, Germany, Ireland, Luxembourg, Netherlands, Norway, Portugal, Sweden, and the UK. We crosscheck our results with a similar measure compiled by Transparency International and find that the results do not change. 25

29 management of a subsidiary over how to invest cash held abroad are an economic channel through which repatriation taxes result in lower subsidiary-level investment efficiency. 5.2 Heterogeneity in the Effect: Flexibility to Defer Repatriation In this section, we test whether the effect of repatriation taxes on investment efficiency varies across MNCs. MNCs with sufficient internal capital are flexible in timing repatriations and less regularly return foreign earnings to the parent in order to fund domestic investment or payouts (Altshuler and Grubert, 2003; Nessa, 2016). This results in higher cash holdings abroad (Albring et al., 2011) and increases the opportunities of self-interested subsidiary managers to consume personal benefits. Consequently, we expect the effect of repatriation taxes on investment efficiency to be concentrated in these firms. To identify such MNCs, we use consolidated financial statement data and calculate an MNC s operating-cash-flow-to-total-assets ratio in year t. We then sort MNCs into terciles within each parent country-industry and classify MNC-years in the lowest tercile as having low internal capital and being less flexible to defer repatriation. 32 This approach is consistent with findings in De Simone and Lester (2018), who show that MNCs rely on foreign cash to fund domestic operations if the otherwise available internal capital is insufficient. Thus, MNCs with high operating cash flows relative to their country-industry peers will be more flexible in deferring repatriations and less frequently repatriate earnings from their subsidiaries. We present results in Table 7. The coefficient on 9$:;&<;= -,>,* 56 -,/,* is insignificant for MNCs with a low operating-cash-flow-to-total-assets ratio (Column 1) and negative and significant coefficient on 9$:;&<;= -,>,* 56 -,/,* for the subsample of MNCs less likely to 32 For example, we compare all U.S. MNCs in the oil industry, so an MNC s flexibility to defer is measured relative to its peers within the same tax regime and the same industry. 26

30 repatriate, i.e., MNCs with high operating cash flows (Column 2). These results suggest that the negative effect of repatriation taxes on investment efficiency depends on the flexibility to defer repatriation of foreign earnings. In other words, when the MNC needs to repatriate cash, which is a form of monitoring by central management, subsidiary-level investment efficiency increases. 6. Tax Reforms in the UK and in Japan 6.1 Institutional Setting and Research Design In 2009, both the UK and Japan reformed their international tax systems and switched from worldwide to territorial taxation. The aim of the reform in the UK was to increase the country s attractiveness as a business location and to prevent UK MNCs from relocating abroad. Similarly, the Japanese reform was intended to foster the repatriation of foreign earnings and to promote domestic investment (Arena and Kutner, 2015). Because of this reform, repatriations of MNCs residing in the UK are fully exempt from repatriation taxes from 2009 onward while the tax reform in Japan offers a 95 percent repatriation tax exemption for Japanese MNCs. Neither country altered its corporate income tax rates contemporaneously with the reform. 33 We follow Arena and Kutner (2015) and Xing (2018) and treat these tax reforms as a quasinatural experiment to support a causal interpretation of our baseline findings. In our setting, we expect investment efficiency of foreign subsidiaries of British and Japanese MNCs to increase after reform eliminates repatriation taxes. To test this, we use a difference-in-differences (DiD) research design with foreign subsidiaries of British and Japanese MNCs as the treatment group. We compare the investment behavior of these treatment subsidiaries to a control group of foreign 33 In the UK, a corporate income rate of 28 percent was effective from April 1, This tax rate was further reduced to 26 percent effective from April 1, The Japanese corporate income tax consists of a federal tax, supplemented by local taxes. These taxes led to an average corporate income tax rate of 41 percent, which was lowered to 37 percent for tax years beginning after April 1,

31 subsidiaries of U.S. MNCs whose earnings faced repatriation taxes throughout the entire sample period. The U.S. also did not alter its 35 percent corporate income tax rate during our sample period, allowing us to attribute all differences to changes in repatriation taxes. The quasi-natural experiment provided by these reforms appears to provide the cleanest identification of the effects we are interested in. However, we present this DiD analysis as supplementary rather than primary due to two limitations of the setting. First, both reforms took place during the financial crisis, which could affect the investment behavior of subsidiaries. Although the subsequent analysis allows us to control for the uniform effect of the financial crisis on the treatment and the control group, we are unable to capture differential effects as data limitations prevent us from excluding the crisis years and comparing the investment behavior after the reform to pre-crisis years (e.g., ). Second, the Japanese tax reform was widely discussed in the media, potentially resulting in anticipation effects (Xing, 2018). In expectation of the costless repatriation of foreign cash holdings after the reform, subsidiary managers of Japanese MNCs could have faced greater incentives to invest inefficiently in the year prior to the reform. To test our prediction, we modify Equation (1) and estimate the following OLS regression separately for foreign subsidiaries of British MNCs and foreign subsidiaries of Japanese MNCs:!"#$%&'$"& (,* =, - +, / +, ,/, -,/ 56 -,/,* ,/,* $WIF' - + 3? 5I%& * + 3 X 9$WIF' - 5I%& * + 3 Y 9$WIF' ,/,* + 3 Z 5I%& * 56 -,/,* + 3 [ 9$WIF' - 5I%& * 56 -,/,* + (2) (,* + 3 K 5;F$"&HI"&FIJ% K 0,* + L (,* We include indicator variables for the treatment group (9$WIF' - ) and for years after the tax reforms (5I%& * ). 9$WIF' - is equal to 1 if the subsidiary is owned by an MNC residing in the UK 28

32 or Japan, respectively, and 0 for subsidiaries of U.S. MNCs. 5I%& * is equal to 1 for years after the tax reform. As the reforms became effective in both countries in 2009, 5I%& * is equal to 1 for years after The interaction term 9$WIF' - 5I%& * captures the change in the level of investment from the pre period to the post period of subsidiaries of UK and Japanese MNCs, respectively, relative to the subsidiaries of U.S. MNCs. We expect a negative coefficient on 9$WIF' - 5I%& *, which suggests that subsidiaries of British and Japanese MNCs invest less after the reforms (Arena and Kutner, 2015). Our main coefficient of interest is the treatment effect conditional on local growth opportunities. To this end, we interact 9$WIF' - 5I%& * with 56 -,/,* and expect a positive coefficient on the interaction. This result would indicate that, after the reform, investment of subsidiaries of British and Japanese MNCs is more sensitive to local growth opportunities relative to investment of subsidiaries owned by U.S. MNCs; i.e., investment efficiency increases because of the reform. Consistent with Equation (1), we include subsidiary country and subsidiary industry fixed effects, parent fixed effects, and the interactions of subsidiary country-industry fixed effects with 56 -,/,*. 34 Using the DiD specification, we thus identify the effect of the tax reforms by comparing the investment efficiency of treatment subsidiaries to the investment efficiency of control subsidiaries in the same country-industry. This approach again mitigates concerns that unobserved subsidiary country or subsidiary industry characteristics might affect investment efficiency. 6.2 Regression Results In Table 8, we present results for the DiD estimations based on foreign subsidiaries of British (Columns 1 and 2) and Japanese MNCs (Columns 3 and 4), respectively. In these tests, we limit 34 We do not include year fixed effects in the DiD estimation because they are highly collinear with the 5I%& * variable. Results are qualitatively similar when including year fixed effects. 29

33 the pre (post) period to the year 2008 (2009). The coefficient on 9$WIF' - 5I%& * is negative and significant in Column 1 and negative but marginally insignificant in Column 3 (p = 0.103). This result is in line with Arena and Kutner (2015) and suggests that subsidiaries of British and Japanese MNCs reduced the level of investment in response to the reform. Coefficients on 9$WIF' - 5I%& * 56 -,/,* are positive and significant in both columns. Corroborating our baseline findings, these results suggest that the investment efficiency of subsidiaries of British and Japanese MNCs relative to subsidiaries of U.S. MNCs increased after reform removed repatriation taxes. In Columns 2 and 4, we assess whether the parallel trend assumption holds in our settings, i.e., whether subsidiaries in the treatment and control groups exhibit similar trends in investment efficiency prior to the reform. We drop all observations in the post period and extend the pre period to cover the years 2007 and Moreover, we define a placebo treatment for the year Thus, in this test, 5I%& * is equal to 1 for the year If the parallel trend assumption holds, coefficients on 9$WIF' - 5I%& * 56 -,/,* should be insignificant. In Column 2, we find insignificant coefficients on 9$WIF' - 5I%& * and on 9$WIF' - 5I%& * 56 -,/,*, providing evidence for parallel pre-reform trends in the level of investment and in investment efficiency of subsidiaries of British and U.S. MNCs. With respect to subsidiaries of Japanese MNCs, however, the coefficient on 9$WIF' - 5I%& * 56 -,/,* (9$WIF' - 5I%& * ) is negative (positive) and significant in Column 4. Thus, prior to the reform, trends in the level of investment and in investment efficiency of subsidiaries of Japanese MNCs differ from subsidiaries of U.S. MNCs. This result indicates that, relative to subsidiaries of U.S. MNCs, subsidiaries of Japanese MNCs increased the level of investment but decreased investment efficiency from 2007 to This investment behavior is consistent with anticipation effects due to public discussions preceding the tax reform (Xing, 2018). Thus, managers of subsidiaries of Japanese MNCs faced 30

34 incentives to invest inefficiently in the year before repatriation taxes were repealed. 35 We conduct several additional tests to further validate our results from the DiD estimations. First, we extend the pre and the post period to include two years each and obtain results consistent with our initial findings. Second, we conduct placebo tests in which we assign the tax reforms to random years other than Third, we run our tests on a sample of foreign subsidiaries of MNCs located in countries without a similar tax reform, such as Germany or France, and compare their investment behavior to subsidiaries of U.S. MNCs. In all tests (untabulated), the coefficients on 9$WIF' - 5I%& * 56 -,/,* are insignificant, indicating that it is unlikely that we capture a random, non-tax reform effect in our tests. These results suggest that the repeal of repatriation taxes, while leading to a lower level of investment, resulted in an increase in investment efficiency. As the negative effect of repatriation taxes on investment efficiency decreases once repatriation taxes are abolished, we conclude that repatriation taxes have a likely causal effect on subsidiary-level investment efficiency. 7. Robustness Tests We conduct several additional tests to assess the robustness of our findings and present results in Table 9. First, we test whether the effect of repatriation taxes on investment efficiency is incremental to the effect on foreign M&A activity as documented in prior research (Hanlon et al., 2015; Harford et al., 2017). To test this, we add to Equation (1) an indicator variable (\&O (,* ) with the value of one if a subsidiary or its parent engages in M&A activity in the subsidiary country in the current or the previous year. 36 As reported in Column 1, we continue to find a negative and 35 Such anticipation effects are unlikely among foreign subsidiaries of British MNCs because the tax reform in the UK was proposed and ultimately enacted during December 2008 (Arena and Kutner, 2015). 36 We obtain M&A data from Bureau van Dijk s Zephyr database. 31

35 significant coefficient on 9$:;&<;= -,>,* 56 -,/,* when controlling for an MNC s M&A activity. In Column 2, we drop subsidiary-years with M&A activity and results are again consistent with our baseline findings. These results support our conclusion that repatriation taxes have an incremental effect on subsidiary-level investment. In Column 3, we adjust our measure of subsidiary-level investment and add back annual depreciation on fixed assets. These expenses reduce the book value of fixed assets on the subsidiary balance sheet and potentially affect our measure of investment. By adjusting!"#$%&'$"& (,* for depreciation, the variable measures annual gross investment in fixed assets. Although reducing the sample size, results are unchanged. Moreover, the coefficient on 9$:;&<;= -,>,* 56 -,/,* has a magnitude comparable to our baseline results (Table 4). Thus, differences in depreciation on fixed assets across subsidiaries do not affect our baseline results. In Column 4, we include the subsidiary cash ratio (H;%h9;&DI (,* ) as an additional control variable for a subsidiary s internal funds available for investment. As discussed, this variable could be a mediator (or collider) control because subsidiaries that face repatriation taxes report higher cash holdings in our sample. However, after including H;%h9;&DI (,*, the coefficient on 9$:;&<;= -,>,* 56 -,/,* remains negative and significant. In Column 5, we replace parent fixed effects with subsidiary fixed effects to capture unobserved time-invariant characteristics of the subsidiary. 37 This analysis requires variation in repatriation taxes within a subsidiary. In our sample, such variation stems from changes in the tax system of the parent country (e.g., in the UK and Japan) as well as from changes in corporate income tax rates in either the parent country or the subsidiary country. The coefficient of interest 37 Because ownership data are static, the parent fixed effects are subsumed by the subsidiary fixed effects. 32

36 remains negative and significant in Column 5, providing comfort that the omission of a subsidiary characteristic is not driving our baseline results. Lastly, we limit subsidiary-years subject to repatriation taxes to foreign subsidiaries of U.S. MNCs. In line with our baseline results, the coefficient on 9$:;&<;= -,>,* 56 -,/,* in Column 6 is negative and significant. Interestingly, we find a positive and significant coefficient on 9$:;&<;= -,>,*. This result, which is consistent with a supplemental test by Hanlon et al., (2015), suggests that the level of investment of subsidiaries of U.S. MNCs is increasing in repatriation taxes. Our findings extend their result by showing that repatriation taxes also reduce the investment efficiency of subsidiaries of U.S. MNCs. 8. Conclusion Using a global sample of MNCs and their foreign subsidiaries, we show that repatriation taxes impair the efficiency of investment decisions made by the local management of a foreign subsidiary. This effect is concentrated in subsamples of subsidiaries, consistent with internal agency conflicts being the driver. Specifically, the effect occurs when central management is less involved in the decisions of the subsidiary, when the monitoring of the subsidiary is weak, and when the MNC is more flexible in deferring repatriation. We also provide support for a causal interpretation of our findings by examining international tax reforms in the UK and Japan that repealed repatriation taxes on earnings of foreign subsidiaries in Using these quasi-natural experiments in a DiD research design, we find that the investment efficiency of subsidiaries increased after repatriation taxes were eliminated. Our study provides timely empirical evidence on the economic consequences of repatriation taxes as the U.S. recently enacted legislation that eliminates repatriation taxes on most of the future foreign earnings of U.S. MNCs. Our results suggest that this reform may lead to efficiency gains 33

37 for U.S. MNCs and their shareholders. The subsidiary country, in contrast, may bear negative economic consequences in the form of lower investment. These results should be of interest to policymakers in the U.S. and in the countries in which U.S. MNCs operate. 34

38 APPENDIX A: VARIABLE DEFINITIONS Variable CashRatio CashFlowParent DomesticCredit Investment GrossInvestment M&A Definition Cash and cash equivalents of subsidiary i in year t scaled by total assets of subsidiary i in year t. Operating cash flow of parent p in year t scaled by total assets of parent p in year t. Amount of credit provided by the banking sector of country c scaled by the Gross Domestic Product (GDP) of country c in year t. Fixed assets of subsidiary i in year t less fixed assets of subsidiary i in year t-1 scaled by total assets of subsidiary i in year t-1. Fixed assets of subsidiary i in year t less fixed of subsidiary i in year t-1 plus depreciation expenses of subsidiary i in year t and scaled by total assets of subsidiary i in year t-1. Indicator variable with the value of one if parent p or subsidiary i engage in M&A activity in subsidiary country c in year t or year t- 1, and zero if parent p and subsidiary i do not engage in M&A activity in subsidiary country c in year t or year t-1. PE Price-to-earnings ratio for industry j in country c in year t. Industry j is based on one-digit ICB industries. We compute the PE-ratio for year t as the median of the end of month values obtained from Datastream. RepatTax (continuous) RepatTax (indicator) Continuous measure for RepatTax in year t calculated as the statutory corporate income tax rate in country g less the statutory corporate income tax rate in country c. We set RepatTax equal to zero if parent p resides in country g with a territorial tax system. In addition, we set RepatTax equal to zero if parent p resides in a country with a worldwide tax system but the statutory corporate income tax rate in country g is lower than the statutory corporate income tax rate in country c. Indicator variable with the value of one if parent p resides in country g with (1) a worldwide tax system and (2) a higher corporate income tax rate than country c. RepatTax takes on the value of zero if parent p resides in country g with (1) a territorial tax system or (2) a lower corporate income tax rate than country c. We calculate RepatTax for year t. 35

39 RoA Profit after tax of subsidiary i in year t scaled by total assets of subsidiary i in year t. Size Natural logarithm of total assets of subsidiary i in year t. Tangibility Fixed assets of subsidiary i in year t scaled by total assets of subsidiary i in year t. TotalParticipation Total direct and indirect participation of parent p in subsidiary i. Additional Variables for DiD Analysis Reform Post Indicator variable with the value of one if parent p resides in UK or Japan, respectively. Reform takes on the value of zero if parent p resides in the U.S. Indicator variable with the value of one if year t is equal to the year Post takes on the value of zero if year t is equal to the year

40 APPENDIX B: SAMPLE SELECTION Sample Selection Observations (subsidiary-years) Data from Bureau van Dijk s Orbis database after dropping subsidiaries without or limited financial statement information, subsidiaries exclusively filing consolidated financial statements, domestic subsidiaries, and subsidiaries without NACE and merged ICB codes After dropping subsidiaries where the parent holds a total participation 50 percent After dropping subsidiaries in the financial sector (NACE ), the utility sector (NACE ), and financial holdings (NACE 7010) After dropping financial years < 2006 and > 2014 due to data for these years being not available or being not comprehensive After dropping observations with missing or negative values for total assets, operating revenue, fixed assets, or cash and cash equivalents 561, , , , ,870 After dropping observations with total assets, operating revenue, and fixed assets < US$10, ,975 After dropping observations with missing tax data 116,077 After dropping observations with missing values for Investment, PE, RoA, or 57,475 CashFlowParent After dropping observations with values for Investment, PE, RoA, Size, Tangibility, or CashFlowParent that are in the bottom and top 1.5% of the 48,470 variable distribution Final sample 48,470 37

41 References Abel, A.B., Optimal investment under uncertainty. The American Economic Review 73, Albring, S.M., Mills, L.F., Newberry, K.J., Do debt constraints influence firms sensitivity to a temporary tax holiday on repatriations? Journal of the American Taxation Association 33, Altshuler, R., Grubert, H., Repatriation taxes, repatriation strategies and multinational financial policy. Journal of Public Economics 87, Altshuler, R., Newlon, T.S., Randolph, W., Tax effects on income repatriation by US multinationals: Evidence from panel data. Mimeograph, US Department of the Treasury, Office of Tax Analysis. Ang, J.S., Cole, R.A., Lin, J.W., Agency Costs and Ownership Structure. The Journal of Finance 55, Antràs, P., Desai, M.A., Foley, C.F., Multinational Firms, FDI Flows, and Imperfect Capital Markets. The Quarterly Journal of Economics 124, Arena, M.P., Kutner, G.W., Territorial tax system reform and corporate financial policies. The Review of Financial Studies 28, Asiedu, E., Esfahani, H.S., Ownership Structure in Foreign Direct Investment Projects. The Review of Economics and Statistics 83, Badertscher, B., Shroff, N., White, H.D., Externalities of public firm presence: Evidence from private firms investment decisions. Journal of Financial Economics 109, Baker, M., Stein, J.C., Wurgler, J., When does the market matter? Stock prices and the investment of equity-dependent firms. The Quarterly Journal of Economics 118, Becker, B., Jacob, Marcus, Jacob, Martin, Payout taxes and the allocation of investment. Journal of Financial Economics 107, Bekaert, G., Harvey, C.R., Lundblad, C., Siegel, S., Global Growth Opportunities and Market Integration. The Journal of Finance 62, Bernardo, A.E., Cai, H., Luo, J., Capital Budgeting in Multidivision Firms: Information, Agency, and Incentives. The Review of Financial Studies 17, Bertrand, M., Mullainathan, S., Enjoying the quiet life? Corporate governance and managerial preferences. Journal of Political Economy 111, Beyer, B., Downes, J., Rapley, E.T., Internal capital market inefficiencies, shareholder payout, and abnormal leverage. Journal of Corporate Finance 43, Biddle, G.C., Hilary, G., Accounting quality and firm-level capital investment. The Accounting Review 81, Biddle, G.C., Hilary, G., Verdi, R.S., How does financial reporting quality relate to investment efficiency? Journal of Accounting and Economics 48,

42 Bird, A., Edwards, A., Shevlin, T., Does US foreign earnings lockout advantage foreign acquirers? Journal of Accounting and Economics 64, Björkman, I., Barner-Rasmussen, W., Li, L., Managing knowledge transfer in MNCs: The impact of headquarters control mechanisms. Journal of International Business Studies 35, Blanchard, O.J., Lopez-de-Silanes, F., Shleifer, A., What do firms do with cash windfalls? Journal of Financial Economics 36, Bloom, N., Sadun, R., Van Reenen, J., The organization of firms across countries. The Quarterly Journal of Economics 127, Blouin, J.L., Krull, L.K., Robinson, L.A., The Location, Composition, and Investment Implications of Permanently Reinvested Earnings. Working Paper. Campbell, J.L., Dhaliwal, D.S., Krull, L.K., Schwab, C.M., U.S. Multinational Corporations Foreign Cash Holdings: An Empirical Estimate and Its Valuation Consequences. Working Paper. Carpenter, R.E., Petersen, B.C., Is the Growth of Small Firms Constrained by Internal Finance? The Review of Economics and Statistics 84, Chen, N., Chiu, P.-C., Shevlin, T., The Persistence and Pricing of Changes in Multinational Firms Foreign Cash Holdings. Working Paper. Claessens, S., Laeven, L., Financial Development, Property Rights, and Growth. The Journal of Finance 58, Cummins, J.G., Hassett, K.A., Hubbard, R.G., Tax reforms and investment: A cross-country comparison. Journal of Public Economics, Proceedings of the Trans-Atlantic Public Economic Seminar on Market Failures and Public Policy 62, De Simone, L., Klassen, K.J., Seidman, J.K., The Effect of Income-shifting Aggressiveness on Corporate Investment. Working Paper. De Simone, L., Lester, R., The Effect of Foreign Cash Holdings on Internal Capital Markets and Firm Financing. Working Paper. De Simone, L., Piotroski, J.D., Tomy, R.E., Repatriation Taxes and Foreign Cash Holdings: The Impact of Anticipated Tax Policy. Working Paper. De Waegenaere, A., Sansing, R.C., Taxation of International Investment and Accounting Valuation. Contemporary Accounting Research 25, Desai, M.A., Foley, C.F., Hines Jr, J.R., Dividend Policy inside the Multinational Firm. Financial Management Easterbrook, F.H., Two agency-cost explanations of dividends. The American Economic Review 74, Edwards, A., Kravet, T., Wilson, R., Trapped cash and the profitability of foreign acquisitions. Contemporary Accounting Research 33, Edwards, A., Schwab, C., Shevlin, T., Financial constraints and cash tax savings. The Accounting Review 91,

43 Erickson, T., Whited, T.M., Measurement Error and the Relationship between Investment and q. Journal of Political Economy 108, Faulkender, M., Petersen, M., Investment and capital constraints: repatriations under the American Jobs Creation Act. Review of Financial Studies 25, Feld, L.P., Ruf, M., Scheuering, U., Schreiber, U., Voget, J., Repatriation taxes and outbound M&As. Journal of Public Economics 139, Foley, C.F., Hartzell, J.C., Titman, S., Twite, G., Why do firms hold so much cash? A taxbased explanation. Journal of Financial Economics 86, Glaser, M., Lopez De Silanes, F., Sautner, Z., Opening the Black Box: Internal Capital Markets and Managerial Power. The Journal of Finance 68, Goodman, T.H., Neamtiu, M., Shroff, N., White, H.D., Management Forecast Quality and Capital Investment Decisions. The Accounting Review 89, Gow, I.D., Larcker, D.F., Reiss, P.C., Causal Inference in Accounting Research. Journal of Accounting Research 54, Graham, J.R., Hanlon, M., Shevlin, T., Real Effects of Accounting Rules: Evidence from Multinational Firms Investment Location and Profit Repatriation Decisions. Journal of Accounting Research 49, Graham, J.R., Harvey, C.R., Rajgopal, S., The economic implications of corporate financial reporting. Journal of Accounting and Economics 40, Grinblatt, M., Keloharju, M., How distance, language, and culture influence stockholdings and trades. The Journal of Finance 56, Gu, T., US multinationals and cash holdings. Journal of Financial Economics Forthcoming. Hanlon, M., Lester, R., Verdi, R., The effect of repatriation tax costs on US multinational investment. Journal of Financial Economics 116, Harford, J., Corporate Cash Reserves and Acquisitions. The Journal of Finance 54, Harford, J., Wang, C., Zhang, K., Foreign Cash: Taxes, Internal Capital Markets, and Agency Problems. The Review of Financial Studies 30, Hartman, D.G., Tax policy and foreign direct investment. Journal of Public Economics 26, Hayashi, F., Tobin s marginal q and average q: A neoclassical interpretation. Econometrica: Journal of the Econometric Society 50, Jensen, M.C., Agency costs of free cash flow, corporate finance, and takeovers. The American Economic Review 76, Jensen, M.C., Meckling, W.H., Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, Klassen, K.J., Laplante, S.K., Carnaghan, C., A Model of Multinational Income Shifting and an Application to Tax Planning with E-Commerce. The Journal of the American Taxation Association 36,

44 Laplante, S.K., Nesbitt, W.L., The relation among trapped cash, permanently reinvested earnings, and foreign cash. Journal of Corporate Finance 44, Ma, Z., Stice, D., Wang, D., US Worldwide Taxation Policy and Bank Loan Contracting. Working Paper. Markle, K., A Comparison of the Tax-Motivated Income Shifting of Multinationals in Territorial and Worldwide Countries. Contemporary Accounting Research 33, Murphy, F., Do Firms Alter Foreign Organizational Structure in Response to Changes in U.S. International Tax Policy? Evidence From TIPRA Working Paper. Nessa, M.L., Repatriation tax costs and US multinational companies shareholder payouts. The Accounting Review 92, Rajan, R., Servaes, H., Zingales, L., The cost of diversity: The diversification discount and inefficient investment. The Journal of Finance 55, Robinson, L.A., Stocken, P.C., Location of decision rights within multinational firms. Journal of Accounting Research 51, Scharfstein, D.S., Stein, J.C., The Dark Side of Internal Capital Markets: Divisional Rent- Seeking and Inefficient Investment. The Journal of Finance 55, Shin, H.-H., Stulz, R.M., Are internal capital markets efficient? The Quarterly Journal of Economics 113, Shroff, N., Verdi, R.S., Yu, G., Information environment and the investment decisions of multinational corporations. The Accounting Review 89, Stein, J.C., Chapter 2 - Agency, Information and Corporate Investment, in: Constantinides, G.M., Harris, M., Stulz, R.M. (Eds.), Handbook of the Economics of Finance, Corporate Finance. Elsevier, pp Stein, J.C., Information Production and Capital Allocation: Decentralized versus Hierarchical Firms. The Journal of Finance 57, Stein, J.C., Internal capital markets and the competition for corporate resources. The Journal of Finance 52, Swanquist, Q.T., Whited, R.L., Out of Control: The Use and Misuse of Controls in Accounting Research. Working Paper. Tobin, J., A general equilibrium approach to monetary theory. Journal of Money, Credit and Banking 1, Williamson, O.E., Markets and Hierarchies: Analysis and Antitrust Implications. Collier Macmillan Publishers, Inc., New York. Xing, J., Territorial tax system reform and multinationals foreign cash holdings: New evidence from Japan. Journal of Corporate Finance 49, Yang, S., The disclosure and valuation of foreign cash holdings. Working Paper. Yoshikawa, H., On the" q" Theory of Investment. The American Economic Review 70,

45 42

46 FIGURES AND TABLES Figure 1: Empirical Approach Note: This figure illustrates our empirical approach. Assume we have two foreign subsidiaries located in Poland that operate in the same industry. The earnings of the subsidiary with the Austrian parent (Sub 1) are not subject to repatriation taxes because Austria applies a territorial tax system. Sub 2 has a U.S. parent and the subsidiary s earnings are therefore subject to repatriation taxes of 16 percent (U.S. corporate income tax of 35 percent less a tax credit for the Polish corporate income tax of 19 percent) since the U.S. applies a worldwide tax system (until 2018). By including interactions of country-industry fixed effects and growth opportunities in our regression model, we empirically compare whether the sensitivity of investment to local growth opportunities (investment efficiency) of these two Polish subsidiaries varies conditional on repatriation taxes. Thus, we identify the effect of repatriation taxes on investment efficiency from cross-subsidiary and cross-time variation in repatriation taxes within the same countryindustry. 43