International Debt Shifting: The Value-Maximizing Mix of Internal and External Debt

Abstract

We study the capital structure of multinationals and expand previous theory by incorporating international debt tax shield effects from both internal and external capital markets. We show that: (i) multinationals’ firm value is maximized if both internal and external debt are used to save tax; (ii) the use of internal and external debt is independent of each other; and (iii) multinationals have a tax advantage over domestic firms, which cannot shift debt across international borders. We test our model using a large panel of German multinationals and find that internal and external debt shifting are of about equal importance.

Keywords:

• Corporate Taxation
• Multinationals
• Capital Structure
• International Debt Shifting
• Tax Avoidance

JEL Classifications:

• H25
• G32
• F23

Acknowledgments

We are grateful to Carsten Bienz, Roger Gordon, Heinz Herrmann, Harry Huizinga, Erling Røed Larsen, Tore Leite, Martin Ruf, Georg Wamser, and Alfons Weichenrieder, as well as to participants at the annual conference of the Western Economic Association International in Portland, the Norwegian Research Forum on Taxation in Moss, and to participants at seminars in Bergen, Oslo, and Tilburg for very helpful suggestions. Special thanks go to the people at Deutsche Bundesbank, in particular to Alexander Lipponer, for their invaluable support and for the hospitality of their research center. Remaining errors are ours.

Notes

1 The various costs and benefits of debt have been reexamined in empirical studies by van Binsbergen et al. (2010) and Korteweg (2010). They estimate that the net benefits of debt amount to 5.5% of firm value and 3.5% of a firm’s book value, respectively.

2 An example of a company using this strategy is the Formula One business, where several highly internal-debt loaded firms under the umbrella of the Delta Topco Holding are paying 15% interest to an internal bank located on the Channel Island of Jersey, which is well known as a tax haven. See Sylt and Reid (2011).

3 The mechanism of such external debt shifting was first identified and analyzed in Huizinga et al. (2008). Their analysis, however, neglects the simultaneous use of internal debt.

4 A notable exception is Mackie-Mason (1990) who avoids the lack of tax-rate variation by focusing on whether a firm is near tax exhaustion.

5 In contrast, Park et al. (2013) only find little indication that taxes trigger a distinct financial policy in US multinationals relative to their domestic counterparts.

6 Extending the analysis in Huizinga et al. (2008), Sorbe et al. (2017) provide evidence that external debt shifting increases total debt within the whole MNC.

7 It should be noted that Huizinga et al. (2008) discuss internal debt in an extension to the empirical analysis. In order to explore the robustness of their results, they utilize the tax rate differential between the parent company and an affiliate. They do not find a significant effect of this variable and conclude that their main result is not affected by the incentive to use internal debt. As will become clear in the next section, it is the tax rate differential to the lowest-taxed affiliate that is best suited as a variable to measure the use of internal debt.

8 See Hovakimian et al. (2004) and Aggarwal and Kyaw (2010) for some overviews on factors affecting the optimal capital structure.

9 See Gertner et al. (1994) for a discussion on internal debt and how it relates to external debt and equity. Chowdhry and Coval (1998, pp. 87f) and Stonehill and Stitzel (1969) argue that internal debt should in fact be seen as tax-favored equity.

10 For similar approaches and a more detailed discussion of costs on internal debt, see, e.g., Mintz and Smart (2004), Fuest and Hemmelgarn (2005), and Schindler and Schjelderup (2016).

11 The “trade-off” theory of capital structures balances bankruptcy costs with returns from the tax shield. See, for instance, Graham (2000), who estimates a tax-shield value (before personal taxes) close to 10% of the value of the firm.

12 Gopalan et al. (2007) find that business groups do in fact support financially weaker firms in the group in order to avoid default. They also find that bankruptcy by a group firm gives negative spillovers to other affiliates in the form of a significant drop in external financing, investments, and profits, and an increase in the bankruptcy probability.

13 Note, however, that this assumption does not imply that the costs of debt do not reduce potential withholding and repatriation taxes.

14 It can be shown that from the viewpoint of a shareholder in an MNC, maximizing profits of the MNC after global corporate taxation and maximizing the net pay-off on equity investment after opportunity costs and personal (income) taxes yield identical results under mild assumptions.

15 Indeed, this seems to explain why countries such as Belgium, Luxembourg, and the Netherlands attract so many financial coordination centers of MNCs (see, e.g., Ruf and Weichenrieder, 2012, table 4). All these countries have special tax rules for financial operations that lead to very low effective tax rates. The absence of source taxes on dividends between EU member states then makes it easy to shift income streams tax free across affiliates.

16 See the discussion of the cost function on page §.

17 A full documentation is given by Lipponer (2009).

18 From 2002 onwards, only companies and investors with more than three million euros in total assets are required to report to the Deutsche Bundesbank.

19 Aggarwal and Kyaw (2010) find a relationship between debt in MNCs and political risk.

20 When the weighted tax difference variable cannot be constructed (see restriction 4 in Table 1), it is usually because the parent company assets share or tax data are missing.

21 We have computed liabilities as “liabilities” plus “other liabilities.” It is not entirely clear what is included in the variable “other liabilities,” but accruals for pensions are one example. Whether “other liabilities” are included has little effect on the empirical results.

22 Buettner and Wamser (2013) do not count loans from the parent companies in the internal debt ratio, as they claim such loans cannot be given for tax reasons. Our model does not justify such a choice, but our findings are robust to excluding parent debt. It should be noted that the amount of parent debt is substantial. The average share of internal debt in total debt is 27%. Two thirds of this is parent debt, and one third is internal debt from other affiliates. The high level of lending from German parent companies is puzzling, given the high tax level in Germany. One explanation may be agency costs, as pointed out by Dischinger et al. (2014); another may be frictions in the external capital market, as pointed out by Desai et al. (2004).

23 Note that earning stripping rules where not introduced until after the end of our sample period. We return to other aspects of this measurement problem in Section 7.3.

24 We are grateful to Martin Ruf for providing the data to us electronically. For all countries, the rates reflect the general corporate tax rates, including average or typical local taxes.

25 This is the variable that corresponds to the variable called “tax incentive to shift debt” in Huizinga et al. (2008).

26 Main summary statistics by country are available in the working paper version of this article (Møen et al. 2011, appendix A.2).

27 Note that the error in the net internal lending variable only affects the descriptive statistics and not the regressions.

28 In terms of specification, this interaction is not strictly necessary, since observations with NLS_pit=0 by definition also have zero maximum tax difference.

29 This is because the NLS dummy is not part of our structural model. It is included because the total debt regression equation is the sum of the external debt regression equation, which is relevant for all affiliates, and the internal debt regression equation, which in our theory model is not relevant for observations from the lowest-taxed affiliates (NLS_pit=0). These units should be internal banks and therefore not have internal debt. When pooling the predicted internal banks with the other affiliates, the NLS dummy allows the observations of predicted internal banks to have a separate intercept.

30 The nonsignificant and close-to-zero coefficient in the external debt regression (see Table 6) suggests that this extra debt is internal debt.

31 Note that the specification in Mintz and Weichenrieder (2010, ch. 5) includes affiliate-specific fixed effects, while we control for fixed effects at the group level. Affiliate fixed effects will absorb a lot more of the variation in the tax variables than group fixed effects, and the two specifications are therefore not directly comparable. The specification used by Jog and Tang (2001) also includes variables that may reduce the bias.

32 A ten percentage point tax change is not uncommon in the early years of our sample, although large tax reductions are far more common than tax increases.

33 The semi-elasticity implied by the coefficient reported in Desai et al. (2004) is taken from Buettner and Wamser (2013).

34 Inspired by Buettner et al. (2011), we have also tried to interact the loss carryforward dummy with the tax variables and to run separate regressions for firms with and without loss carryforwards. Buettner et al. find that loss carryforwards reduce the firms’ tax elasticity with respect to the debt-to-asset ratio. We do not find any clear-cut evidence for this effect in our sample.

35 The semi-elasticity implied by the coefficient reported in Desai et al. (2004) is taken from Buettner and Wamser (2013).

36 See also DeAngelo and Masulis (1980).

37 In the world sample, we only include affiliates that belong to corporate groups that have all affiliates located within the 68 countries covered. Without this restriction, the sample size increases by about 25,000 observations. As compared to columns (7) and (8) of Table 8, the coefficients on the weighted tax difference variable then fall notably and become insignificant. The other coefficient stays about the same.

38 The data source is the Annual Reports of International Activity to the Norwegian Directorate of Taxes (“Utenlandsoppgaven”). No reliable source for the effective rate for German companies was found.

Additional information

Funding

Financial support from the Research Council of Norway, Grant No. 267423, and the Deutsche Forschungsgemeinschaft is gratefully appreciated.