The effect of income shifting on the implied cost of equity capital: evidence from US multinational corporations

Abstract

This study examines the effect of income shifting on the implied cost of equity capital (ICOE) for US multinational corporations (MNCs). We find that income shifting is significantly positively associated with the ICOE after controlling for corporate tax avoidance and other determinants of the ICOE. On average, a one-standard deviation increase in income shifting is associated with an increase in the ICOE of around 0.37%. Our main results are robust to additional tests of risk-pricing and endogeneity. Furthermore, the association between income shifting and the ICOE is more pronounced for MNCs operating in low-quality information environments and where corporate governance monitoring is inadequate. Overall, our results show that the capital market perceives the income shifting of MNCs as a significantly risky undertaking.

Keywords:

• income shifting
• implied cost of equity capital
• US multinational corporations

JEL Classification:

• G30
• H20

Acknowledgements

We would like to sincerely thank the Editors, Mark Clatworthy, Juan Manuel Garcia Lara and Edward Lee for providing excellent feedback on earlier versions of our paper. We would also like to thank Professor Kevin Holland, in addition to the anonymous Reviewers, for providing helpful comments and suggestions on our paper. All remaining errors are our own.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Financial and taxation arbitrage opportunities are frequently exploited simultaneously and have flow-on effects on cash flows, the cost of capital and corporation value (Rego and Wilson 2012; Balakrishnan et al. 2019).

2 Corporate tax avoidance is defined broadly as the reduction of explicit taxes (Hanlon and Heitzman 2010). This definition conceptually follows Dyreng et al. (2008) and reflects all activities that have any effect on a corporation's explicit tax liability.

3 Hanlon and Heitzman (2010) find that income shifting is a specific form of tax avoidance that MNCs regularly undertake at an international level to significantly reduce their tax liabilities.

4 For example, Klassen and Laplante (2012a) find that MNCs with low average foreign tax rates shifted around $10 billion of income out of the U.S. over the 2005–2009 period relative to the 1998–2002 period.

5 For instance, in their 2012 10-K annual report (p.71), Microsoft state: ‘the reduction from the federal statutory rate from foreign earnings taxed at lower rates results from producing and distributing our products and services through our foreign regional operations centers in Ireland, Singapore, and Puerto Rico, which have lower income tax rates.’

6 Previous studies furnish evidence consistent with this particular tax incentive (Grubert and Mutti 1991; Collins et al. 1998; Klassen and Laplante 2012a, 2012b) and also with investment incentives (Krull 2004).

7 However, Dyreng and Markle (2016) examine the effect of financial constraints on income shifting by MNCs. They show that financially-constrained MNCs shift less income from the U.S. to other countries than do their unconstrained peers. This is due to the higher costs of borrowing from external capital markets for such MNCs, compared to any potential tax benefits received (e.g., deferred repatriation of income).

8 Of course, we note that a corporation's stock returns are used in the computation of its ICOE (i.e., the ICOE of the corporation is the internal rate of return that equates that corporation's stock returns to the present value of its expected future cash flows) (Hou et al. 2012), so corporations’ stock returns as captured in the findings of tax avoidance studies by Hanlon and Slemrod (2009) and Wilson (2009) are linked to previous studies on tax avoidance and the ICOE (Hutchens and Rego 2015; Goh et al. 2016; Dhaliwal et al. 2017).

9 However, a recent cross-country study by Delis et al. on income shifting and tax rate uncertainty shows that income shifting is higher for subsidiaries in countries that have stable corporate tax rates over time. They also find that corporations move from transfer pricing to thin capitalization, which has lower adjustment costs.

10 Income shifting may also be priced into a MNC's cost of equity due to the association between income shifting and firm-specific risk. For example, if there are greater market frictions or local investor biases associated with the stocks of MNCs that tend to engage in more income shifting, this risk may be priced as investors are unable to fully diversify their portfolios (Spiegel and Wang 2005; Fu 2009). However, we find that the positive association between income shifting and the ICOE is driven by systematic risk rather than idiosyncratic risk (see Section 5.3).

11 For example, Tyco shifted pre-tax profits from jurisdictions with high tax rates to its subsidiaries in tax havens with low tax rates using complex income shifting arrangements that led to a large drop in its income tax expense. Managers exploited the opacity of Tyco's income shifting arrangements to conceal their personal resource diverting transactions. In particular, managers’ ability to concoct post-tax profits by income shifting to tax haven subsidiaries masked Tyco's actual profitability and allowed them to divert funds without damaging Tyco's reported financial performance. In due course, the market disclosure of widespread resource diversion over the 1997–2002 period led to a major fall in Tyco's stock price, from around $95 in early 2002 to $14 in the middle of 2002 (a drop of about 85%) (Desai 2005).

12 For instance, Demere and Gramlich (2020) find that income shifting opportunities are associated with a discount on corporation value of 2.2%-3.7% for each standard deviation increase in income shifting opportunities. They report that changes in income shifting are associated with the reversal of returns in future periods, suggesting that investors are uncertain about the benefits of income shifting, leading to volatility in future cash flows.

13 We establish whether the S&P 500 corporations have no overseas subsidiaries by checking their 10-K annual reports over the full sample period.

14 As a robustness check of our baseline ordinary least squares (OLS) regression model results, we re-estimate the models without winsorizing the continuous variables. We find that the regression results still hold (untabulated).

15 These measures and the relevant variables used to calculate them are explained in Appendix A.

16 Appendix B provides a classification of the method we used to adjust for predictable errors in analyst forecasts.

17 As a lower INCS value is a proxy for greater income shifting, we transform INCS by multiplying it by -1 to obtain an increasing measure of income shifting for use in our empirical analysis.

18 Earnings of offshore entities may be repatriated to domestic entities for which tax is then paid. If foreign earnings are deemed to be permanently reinvested in the foreign entity, income taxes do not need to be accrued under ASC 740 Accounting for Income Taxes. Designating foreign subsidiary earnings as permanently reinvested earnings (PRE) increases financial statement earnings for the group (Krull 2004), but current and deferred taxes are not recognized as reducing taxable income, so contribute to the size of the difference between accounting and taxable income attributable to foreign tax rate differentials (and hence, to the size of our INCS variable). INCS measures the average effect of differential tax rates on earnings that can consist of PRE- and non-PRE. All of these forms of earnings contribute to accounting income for the group, but not to taxable income and so are captured by INCS, as large negative adjustments on foreign earnings are an indication that managers have the incentive to shift income to achieve favorable capital management and tax objectives. Large negative adjustments of foreign earnings in ETR reconciliations are suggestive of the availability of favorable tax jurisdictions in which funds may be retained or shifted to.

19 However, it may not be readily apparent to investors how much income offshore is attributable to each of the lower taxed jurisdictions, which produces uncertainty for investors.

20 The difference between tax rates related to offshore jurisdictions and the U.S. statutory tax rate of 35% is a major item recorded in ETR reconciliations in a MNC's 10-K annual report. An adjustment to income tax expense on pre-tax accounting profit is required due to the tax rate differential on non-U.S. income earned by overseas domiciled subsidiaries of MNCs (Markle 2011). A MNC's foreign earnings are usually taxed at a lower rate than the U.S. STR of 35%, and thus yield negative adjustments to total income tax expense. Large absolute adjustments to prima facie income tax expense on accounting profit owing to net differential foreign tax rates could reflect greater incentives for a MNC to engage in income shifting.

21 Table 3 (Panel B) reports the descriptive statistics for the variables used in various validation tests, robustness checks and subsample analysis.

22 As a robustness check of our choice of the FF12 industry classification in our baseline regression model, we replace it with two other industry classifications: (1) the FF48 industry classification, and (2) the two-digit standard industry classification (SIC), and re-estimate our regression models. Our untabulated results are qualitatively similar across the five ICOE proxy measures (r_AVG, r_CT, r_GLS, r_OJ and r_PEG) to those shown in Table 4 (Panels A and B).

23 It could be argued that only the highest income shifting increases the ICOE, so the association between INCS and ICOE may be non-linear. To address this concern, we included a squared term (INCS²) in our baseline regression model. We find that INCS² is not significant in any of the regression models, while INCS remains positively and significantly associated with the ICOE, suggesting a linear association between INCS and the ICOE (untabulated).

24 As a robustness check of including ACCETR as a control variable in our baseline regression model, we replace it with the BTAXGAP proxy measure of tax avoidance and re-estimate our regression models. Our untabulated regression results are qualitatively similar across the five ICOE proxy measures (r_AVG, r_CT, r_GLS, r_OJ and r_PEG) to those presented in Table 4 (Panels A and B).

25 As a robustness check, we calculate the repatriation cost (REPAT_COST) (Foley et al. 2007; LaPlante and Nesbitt 2017) and carry out a subsample analysis. We divide our sample into high and low repatriation costs (based on the median value of REPAT_COST) and re-estimate our baseline regression model. For both subsamples, we continue to observe a positive and significant association between INCS and the ICOE (p < 0.01) (untabulated).

26 The kernel approach of matching uses the weighted averages of all of the observations in the control group to construct the counterfactual outcome (Heckman et al. 1997, 1998). Thus, a major advantage of this approach is that a lower variance is achieved in the matching, as more information is used in the process (Smith and Todd 2005).

27 Although the literature on whether information transparency, quality or asymmetry affects a corporation's ICOE has developed substantially over recent years, there is an ongoing debate over whether information risk is or is not diversifiable, and therefore whether it should be priced (Easley and O’Hara 2004; Lambert et al. 2007; Blanco et al. 2015). The literature has also evolved substantially with considerable research devoted to the issue and the effects of different characteristics of financial reporting quality on ICOE estimates (McInnis 2010; García Lara et al. 2011; Barth et al. 2013; Core et al. 2015). Schreder (2018) provides an overview of the ICOE effects relating to information quantity, precision and asymmetry. Past research yields conflicting arguments and results about whether financial reporting quality should affect the ICOE (McInnis 2010; García Lara et al. 2011; Barth et al. 2013; Core et al. 2015). This literature strongly suggests that to determine cost of capital effects, standard asset pricing tests should be used to establish whether financial reporting quality is a priced factor. The variability in outcomes evident in past studies shows that the association between ex-ante estimates of ICOE and proxies for financial reporting quality is not sufficient to show that financial reporting quality affects the ICOE. Thus, we encourage future research to use the standard asset pricing approach to analyze whether the effect of income shifting on the ICOE is conditional on information environment channels.

28 It is also possible that larger boards are less effective than smaller boards due to coordination issues and director free-riding (Lipton and Lorsch 1992; Jensen 1993). For example, Yermack (1996) finds that smaller boards are positively related to higher corporation value (Tobin's Q).