Much of the recent debate about tax avoidance has centered on cross-border tax planning by large multinationals. The results above suggest such planning is likely only a partial
explanation for the decreasing trend in effective tax rates. However, from an empirical perspective, it is challenging to measure multinational activity precisely (e.g., Donohoe et al., 2012). In this section, we employ several variants of MNE to ensure that our conclusions are not sensitive to a measurement issue regarding firms’ multinationality.
We first repeat our interacted model regression specified in equation (3) using six different measures of multinational activity. Specifically, we include, one-by-one, the following alternative MNE variables: (1) an indicator variable for whether or not the firm records foreign sales in its geographic segment data; (2) an indicator variable for whether or not the firm
discloses at least one subsidiary in a foreign country in Exhibit 21 of Form 10-K; (3) the ratio of foreign sales disclosed in the geographic segment data to total sales; (4) the ratio of pretax foreign income to total pretax income; (5) the ratio of the absolute value of pretax foreign income to sales; (6) the log of the total number of foreign countries in which a firm discloses a significant subsidiary in Exhibit 21 of its 10-K. The results of this additional analysis are reported in Table 9. Overall, the analysis yields similar results to those reported in column 1 of Table 7. Specifically, across all specifications, the coefficient on TIME is negative and
significant, with values ranging from -0.526 to -0.582 and all the t-statistics have values greater in magnitude than 8.56. The interaction coefficient on X*TIME is significant in only two of the specifications (both using variation in the same underlying data from Exhibit 21 of Form 10-K).
32
Thus, using different measures of multinational activity, we find that increases in multinational activity explain only a portion of the decline in effective tax rates.
6.2. Sensitivity of results to financial accounting losses and net operating loss carryforwards
Because CASH ETRs are well-defined only when pretax income is positive, in our main tests discussed above we eliminate observations where the denominator of the CASHETR, pre- tax accounting income, is not positive. However, perhaps eliminating loss firm-years from the sample yields results that are somehow not representative of the population of firms. Another potential issue related to tax losses in particular is the ability to utilize current-period tax losses against taxable profits in other periods (i.e., tax loss carryforwards/carrybacks).37 This can reduce taxes paid in those other periods, resulting in a temporal shift in tax rates that could influence trend in effective tax rates we document.
To examine the effect of accounting and tax losses on our results, we conduct several tests, which are reported in Table 10. First, we estimate Eq. (3) on a constant subsample of firms that have positive pretax income in each of the 25 years we study. The resulting sample is 4,200 observations corresponding to 168 firms. Results in Model 1 of Table 10 show that even for consistently profitable firms, CASH ETR exhibits a reliable downward trend. Second, we loosen the sample restriction so that firms are required to always report a pretax profit, but are not required to exist in the database for the entire 25 years of the sample period. The resulting sample is 16,035 observations corresponding to 1,281 unique firms that never report a loss. The results in Model 2 of Table 10 once again demonstrate a reliable negative trend in CASH ETR. Third, we drop all firms that ever report a net operating loss carryforward in the database. The resulting sample is 16,347 observations corresponding to 1,498 unique firms. Once again, the
37 See Erickson, Heitzman, and Zhang (2013) for a discussion and tests of tax loss carryback incentives and
33
results again show a negative trend in CASHETR. Overall, the results in Table 10 suggest that it is unlikely that firms with losses in some years that offset tax payments in other years are driving the downward time trend in CASHETR.38
6.3 Examination of the largest multinational and domestic firms
There are certain firms—including GE, Google, and Apple—that are repeatedly mentioned in the popular press for their tax strategies.39 In this section, we examine whether these highly visible firms have a different over-time pattern in effective tax rates. Specifically, we track the effective tax rates for the 25 largest multinationals and 25 largest domestic firms over our sample period. The 25 largest firms are measured by total assets as of the end of 2012. For multinationals, the 25 largest firms include such firms such as Apple, Berkshire Hathaway, GE, Google, and Microsoft; for domestic only firms, it includes firms such as Kroger, CVS, Macy’s, Union Pacific, and Lowe’s.
Figure 10 shows the trend in effective tax rates for these two groups of firms. The figure shows that the trend in effective tax rates is declining over time for these firms. Moreover, the trend is similar for the multinational group as it is for the domestic group. From this analysis, we conclude that the most visible, newsworthy firms appear to have experienced a similar declining trend in effective tax rates as the average firm in the sample.40
6.4. Comparison of U.S. multinational firms to foreign firms
38 In addition, we run a separate test where we retain all observations, including firms with pretax losses. Because
traditional ratio-based effective tax rate measures are difficult to interpret when the denominator is negative, we instead estimate regressions of the form: 𝑇𝑋𝑃𝐷𝑖𝑡 = 𝛼0+ 𝛼1𝑃𝐼𝑖𝑡+ 𝛼2𝑃𝐼𝑖𝑡∗ 𝑇𝐼𝑀𝐸𝑡+ ∑ 𝛼𝑘 𝑘𝐶𝑂𝑁𝑇𝑅𝑂𝐿𝑘𝑖𝑡+ 𝜖𝑖𝑡, where 𝑇𝑋𝑃𝐷 is cash taxes paid and 𝑃𝐼 is pretax income. We find that the coefficient 𝛼2, which captures the time
trend on the estimated effective tax rate, remains negative and significant. We also estimate the model allowing the coefficients to vary for observations with a pretax loss. In those observations we find that the effective tax rate is declining over time for profitable observations, and that the effective tax rate is near zero and unchanging over time for loss observations.
39 See, for example “G.E.’s Strategies Let It Avoid Taxes Altogether” by David Kocieniewski, New York Times,
March 24, 2011; “How Apple Sidesteps Billions in Taxes” by Charles Duhigg and David Kocieniewski, New York Times, April 28, 2012; and “Google’s 2.4% Rate Shows How $60 Billion Lost to Tax Loopholes” by Jesse Drucker, Bloomberg, October 21, 2010.
34
The primary focus of this study is on the effective tax rates of U.S. multinational firms compared to those of purely domestic U.S. firms. In this subsection, we briefly compare the time trend in GAAP effective tax rates of U.S. multinational firms to that for firms from five large foreign economies: Germany, France, UK, Brazil, and Japan. We use GAAP effective tax rates because most foreign firms do not disclose cash taxes paid, nor is it common for them to disclose the current and deferred components of tax expense separately.
We present results from the analysis in Figure 11. The analysis illustrates that the effective tax rates of U.S. firms have fallen over time, but that the effective tax rates of firms located in major foreign economies have also dropped, sometimes more dramatically than those of U.S. firms. The most dramatic drop is observed in German firms. The average German firm reported a GAAP effective tax rate of over 50 percent in 1989, but less than 30 percent in 2012. Thus, consistent with Markle and Shackelford (2012), the evidence suggests that currently the GAAP effective tax rates of U.S. multinationals are higher than those of companies in many other countries. This is also consistent with the recent trend in which U.S. multinational firms undertake inversions so that the new parent entity is in a foreign country, usually one with territorial taxation and lower rates than the U.S.
We note several caveats to consider when interpreting the results in Figure 11. First, most foreign countries have substantially reduced their statutory corporate tax rates over the sample period whereas the U.S. statutory tax rate has remained essentially constant. Second, the U.S. has retained a worldwide tax system with deferral, along with an accounting exception that allows firms to not record deferred tax expense on the residual tax owed to the U.S. on foreign earnings of subsidiaries if the earnings are indefinitely reinvested. Many other countries have moved to territorial taxation regime in which there is no home country residual tax and thus makes the
35
rates computed less comparable. For example, Germany, the U.K., and Japan have all moved towards territorial taxation. With these caveats noted, one way to view the decline in effective tax rates for U.S. firms is as necessary adaptation in the face of the declines in effective tax rates in the rest of the world, coupled with the adoption of more favorable territorial taxation by many of the largest economies.