Many studies have documented a lower income tax burden

Transcription

1 Tax Burden of U.S. Corporations Having Substantial Foreign Ownership An Analysis of the Relative U.S. Tax Burden of U.S. Corporations Having Substantial Foreign Ownership Abstract - We compare the tax-paying behavior of U.S. firms substantially influenced by foreign-domiciled firms with other U.S. firms. Because public information is lacking on U.S. firms whollyowned by foreign investors, we concentrate on publicly held firms with significant, but not 100 percent, foreign ownership. Public financial statements are analyzed rather than Internal Revenue Service data. We find firms with significant foreign ownership pay less tax than other U.S. firms, but find no support for the hypothesis that the reduced tax burden is attributable to income manipulation. Our evidence suggests that foreign investors select U.S. targets that are (or become) less profitable than their industry counterparts. Michael Kinney Department of Accounting, College of Business Administration, Texas A&M University, College Station, TX Janice Lawrence School of Accountancy, College of Business Administration, University of Nebraska, Lincoln, NE National Tax Journal Vol. LIII, No. 1 INTRODUCTION Many studies have documented a lower income tax burden for U.S. operations of foreign companies relative to other U.S. firms. Explanations offered for this circumstance include transfer pricing manipulations, differences in ownership concentration/control between foreign and domestic firms, real differences in operating costs (such as those related to agency issues) between domestic and foreign firms, differences in competitive strategies, and differences in the maturity levels of foreign-owned and other U.S. firms. Of these explanations, taxable income manipulation through transfer pricing arrangements is dominant. THIS STUDY In this study, we examine the relative tax burdens of foreign-influenced corporations. Like prior studies, we compare the taxes of U.S. firms that are substantially influenced by foreign investors with those of other U.S. firms. Unlike prior studies, we analyze public financial statement data rather than Internal Revenue Service (IRS) data. Public information is lacking on U.S. operations wholly owned by foreign investors; therefore, we analyze publicly held U.S. firms that have significant, but less than 100 percent, foreign ownership. We select the year in which a foreign investor first acquires a stake in a U.S. corporation as the focal point to assess 9

2 NATIONAL TAX JOURNAL changes in the tax-paying status of the U.S. corporation. For a sample of U.S. firms that become investees of foreign entities (hereafter, F-corp sample), we measure tax burden both before and after the investment event. Likewise, we measure tax burdens of other U.S. firms that were investment targets of domestic, rather than foreign, investors (hereafter, D-corp sample). Our empirical analysis provides no support for the income manipulation (transfer pricing) hypothesis. Although we find, consistent with many other studies, that F-corp firms have lower tax burdens than D-corp firms following the foreign investment, we find no support for the income manipulation explanation. Instead, our evidence suggests that foreign investors select U.S. targets that are less profitable than their industry counterparts. The balance of this paper is arranged in three parts. Next, we briefly review prior studies that have addressed the question of whether, and explanations as to why, foreign-controlled U.S. firms, compared to other U.S. firms, have lower tax burdens. Then, we discuss our design and empirical results. We summarize the study in the final section. PRIOR STUDIES A study sponsored by the Subcommittee on Oversight of the House Ways and Means Committee (Turro, 1990) examines the taxpaying history of 36 foreign-owned companies. Collectively, these firms accounted for U.S. sales of $35 billion in The investigation discovered that some of the companies included in the study had never paid any U.S. income taxes even though they had been operating in the US for years. The authors conclude that U.S. tax obligations are avoided by having U.S. subsidiaries pay inflated prices for goods and services acquired from the foreign parent. Similar results are found by Goldberg (1990) and Grubert and Mutti (1991). Grubert and Mutti find evidence of multinational company (MNC) income shifting that is consistent with a taxminimization strategy. High profits are reported in low-tax countries and low profits are reported in high-tax countries. 1 Goldberg finds that foreign-controlled companies, compared to U.S.-controlled companies, report smaller revenue figures and pay less tax as a percent of revenues. Harris et al. (1993) find further evidence that MNCs are able to avoid taxes through transfer pricing arrangements and that the effective U.S. tax rate of a MNC is sensitive to the tax rates in the jurisdictions of its foreign subs. In particular, firms with foreign subs in high-tax countries have higher U.S. tax burdens than U.S. firms with foreign subs in low-tax countries. Grubert et al. (1993) examine 1987 U.S. tax data for 600 firms in which at least 50 percent of the voting stock is foreign owned. The study compares the tax-paying behavior of this group of firms to 4,000 other U.S. firms. The authors find that the foreign-controlled firms have significantly lower levels of taxable income relative to total assets than do other U.S. firms. The authors potentially attribute one-half of the differential to tax-related manipulations. Grubert (1996) extends the prior study using more recent Statistics of Income (SOI) data and finds that less than one-half of the differential is attributable to potential income manipulation. In a contrary finding, Collins et al. (1997) examined foreign-controlled firms (in the wholesale trade industry) that paid little tax over long time horizons. This study attempted to determine if the low tax burdens of these companies could be attributed to income manipulation. The authors found no evidence of income manipulation in their comparison of the 1 For the sake of expediency, we use the terms low-tax and high-tax to refer to low and high effective tax rates. 10

3 Tax Burden of U.S. Corporations Having Substantial Foreign Ownership sample firms to control firms in the industry. The authors concluded that (unspecified) systematic differences between the foreign-controlled and other U.S. firms are more likely than income manipulation to account for the low tax burdens of the foreign-controlled firms. Other Causes of Tax Burden Differences The lower U.S. tax burdens of foreign firms may be a function of real differences in operating costs and, hence, profitability, between domestic and foreign firms operating in the United States. For example, such cost differences may arise from additional agency or transaction costs incurred by foreign firms as a result of cultural and language differences that affect contracts or by the need for extensive monitoring systems. Alternatively, foreign firms may typically have a higher level of ownership concentration/control that mitigates the agency problem and allows them to more aggressively pursue taxminimization strategies; the related increased feasibility of tax-minimization strategies results in lower tax burdens for the foreign firms (Scholes and Wolfson, 1992). Another possibility is that the lower U.S. tax burden of foreign firms is attributable to differences in competitive strategies between U.S. and foreign firms. Most popular among this set of arguments is that there are timing differences in investment outlay and return recognition between the foreign and U.S. firms. For example, it is argued that Japanese firms invest to generate less current return and greater future return than do U.S. firms. Consequently, current income and income taxes, relative to current investment, are lower for Japanese firms. Finally, it has been suggested that lower tax burdens of U.S. operations of foreign firms are a function of the maturity of the business under foreign control compared to benchmark U.S. businesses. This socalled greenfield start-up effect argues that foreign operations included in tax burden studies have not reached the same level of operational maturity as the benchmark U.S. businesses. Taxes assessed on these operations are below those of U.S. counterparts because of the high start-up costs and low revenues that characterize new businesses. EMPIRICAL ANALYSIS Relative to other studies, ours is unique in that it utilizes a comparative analysis of financial statement data rather than SOI or other IRS data. 2 Empirically, our first step is to identify appropriate samples for analysis. All direct investments in U.S. companies by foreign firms are identified through examination of Mergers and Acquisitions rosters for the 15-year period beginning in 1975 and ending in The investments of interest in this study are those in which a foreign firm acquires an initial equity interest in a U.S. firm of at least ten percent. Hunt (1986) cites examples of several empirical studies that have utilized a similar cutoff for indicating managerial influence of acquired firms. 3,4 Firms meeting our 2 Spooner (1986) provides an informative discussion of advantages and disadvantages of financial statement data relative to SOI data for measuring effective tax rates. There are two distinct advantages of financial statement data that pertain to our study. First, we can follow a specific firm over a period of time and observe its taxpaying behavior before it became a foreign investee. Second, we can develop better controls for size, industry, and other determinants of tax burden. 3 Although we might normally think of 50 percent ownership as an important threshold for organizational control in a tax setting, casual evidence suggests control may be effected with a smaller ownership percentage. For example, H.R. 5270, The Foreign Tax Rationalization and Simplification Act of 1992, treats 25 percent as the threshold for control. 11

4 NATIONAL TAX JOURNAL selection criterion are included in the F- corp sample. 5 By selecting established firms that are not wholly owned by foreign investors, we have a control for the agency cost explanation of lower tax burdens of foreignowned U.S. business. Presumably, an acceptable agency cost control system is in place at the time the foreign investment is initiated. 6 It is important to observe the taxpaying behavior of the U.S. firm before it became subject to influences of any foreign investor. Therefore, only transactions that represent an initial equity investment in a U.S. firm are included in the sample. The equity ownership of each U.S. firm is checked to ascertain that no firm in the study has an equity stake (of ten percent or greater) held by any foreign investor. 7 Finally, we examine the proxy statements for all sample firms in the second year following the foreign investment to identify discussions of transactions occurring between the U.S. firm and the foreign investor. 8 Our intent is to confirm that some business is transacted between the foreign firm and its U.S. investee similar to the manner business is conducted between a wholly-owned subsidiary and its parent. If no such discussions are found, the investment is assumed to be passive and the firm is eliminated from the sample. We examine financial data for a fouryear event window beginning with the year preceding the year of investment. 9 Data for some firms are not available throughout this interval, most often due to subsequent 100 percent acquisition and delisting of the firms. Missing data result in the elimination of such firms from the sample. Next, we search stock ownership information for the year preceding the year of acquisition and eliminate any firm in which a prior foreign ownership stake of ten percent or more is found. In addition, we eliminate any firm in which an ownership interest of at least ten percent is not maintained for at least two years subsequent to the year of acquisition. 10 This is 4 In most cases, the investment examined represented the largest single holding of the investee s common stock. Further, in all cases, the investment examined represented the largest new acquisition of stock during the entire four-year event window. 5 We recognize that total (100 percent) equity acquisitions will not be included in the F-corp sample. When firms become wholly-owned investments, they are normally delisted from stock exchanges and cease publication of financial statements. Additionally, in the case of complete acquisitions, measurement of U.S. tax burdens can be muddled by a revaluation of assets an adjustment permitted under generally accepted accounting principles (GAAP). By examining foreign investments in U.S. companies that constitute less than complete acquisitions, we avoid this potential source of bias. Stock acquisitions of less than ten percent of the outstanding shares are not included in this study, because the ability of the acquiring firm to influence managerial decisions is less certain. 6 However, we do note that following the investment by the foreign firm, it is typical that the foreign firm acquires one or more slots on the board of directors and places one or more people in executive officer positions. 7 We do not exclude transactions in which foreign firms obtain an initial equity position (of ten percent or more of outstanding shares) in the post 1974 period and then later make additional investments. We do exclude transactions in which at least a ten percent acquisition is made in the period, but later disinvestment causes the equity stake of the foreign firm to fall below ten percent of outstanding common shares. The rationale for excluding these firms is that we want to exclude any investments that appear to be temporary or of a passive nature. 8 For example, if the investment occurred in 1983, the financial data for 1985 would be reviewed. 9 The financial data for the F-corp sample is gathered from Compustat, Lotus One Source, Compact Disclosure, CRSP, and annual reports. 10 Specifically, the annual reports, proxy statements, and forms 13-D filed with the Securities and Exchange Commission are reviewed. The form 13-D is filed when ownership changes of five percent or more occur. The proxy statements list substantial and beneficial ownership interests of five percent or more and the annual reports include financial statement footnotes indicating substantial acquisitions of equity interest during the year. 12

5 Tax Burden of U.S. Corporations Having Substantial Foreign Ownership the final screen applied to identify our F- corp sample. 11 Matching Firms To effectively evaluate the tax-paying behavior of the F-corp sample, an appropriate reference group is identified. Conceptually, our approach is to compare the tax-paying behavior of the F-corp sample, over time, to a sample of firms that is similar to the F-corp sample in percent of total ownership turnover and in ownership concentration. Thus, we center our analysis on the investment event. The investment event for a matching firm must occur in the same calendar quarter as the investment is made in its F-corp sample mate. We examine the Mergers and Acquisitions roster for the appropriate quarter to identify all U.S. firms in which an interest is acquired during the quarter by a U.S. investor. The firm selected for the D-corp sample is the one having a percentage of ownership turnover that most nearly matches the ownership turnover of the matching F-corp firm. 12 One of the advantages of examining these samples to evaluate tax-paying behavior is that they are not dominated by greenfield startups. Further, an examination of the history of the samples indicates that they are of nearly equal maturity A summary of sample screening results follows. TOTAL FIRMS Firms identified by Mergers and Acquisitions as having an initial foreign equity investment > ten percent 167 Minus: Firms with data not available on Compustat 71 Preliminary sample size 96 Minus: Firms eliminated due to changes in investment 26 Final sample size If two or more potential matches are equally suited based on this criterion, the one that best reflects the F-corp firm in terms of size and industry is selected. The mean (median) percentage of stock acquired in the F-corp stock acquisition transaction is 29 (22). The mean (median) percentage of stock acquired in the D-corp stock acquisition transaction is 27 (22.8). Neither a t-test nor a nonparametric Wilcoxon test indicated a significant difference, at the two-tailed 0.05 level, between the samples in the percent of ownership turnover. 13 The foreign and domestic investees are matched on date of equity investment, so the intervals of time following the initial investment are identical for both samples. To further examine the mix of acquired firms, we identify the date of incorporation or founding of the foreign investees and calculate the age of the firm at the time of the equity investment. A summary tabulation is presented here. In main, these are established firms, with approximately 60 percent being established more than ten years before the investment period. LENGTH OF PERIOD BETWEEN INCORPORATION (OR FOUNDING) AND EQUITY INVESTMENT IN SAMPLE FIRMS Years F-Corp Firms D-Corp Firms Over Unable to determine 6 2 Total Information gathered from Ward s Business Directory of U.S. Private and Public Companies and Moody s Industrial Manual. 13

6 NATIONAL TAX JOURNAL Similarity of ages of firms in the samples provides a control for tax burden differences that are attributable to firm maturation effects. 14 Event-Time Method Using the date of acquisition as the event year (year 1), this study is conducted in event time. This approach facilitates an examination of tax-paying status while controlling for confounding calendar year effects such as length of investment period and differing exchange rates. The year preceding the event year is referred to as the base year (year 0). If no differences in tax burdens are associated with the differences in domiciles of the investors (i.e., domestic versus foreign), we expect the tax burden measures for the F-corp and D-corp samples to be similar both prior to and subsequent to the investment. 15 By measuring tax burden prior to the ownership turnover, we establish a baseline for comparing post-investment tax burden measures. Before assessing relative tax burdens, we test for size differences between the samples; no statistically significant size difference is found for sales, total assets, or net income. 16 Statistical Analysis In Panel A of Table 1, we present measures of tax burden for each sample in event years 0 and 3; in Panel B, we present differences in the tax burden measures. The year 0 difference in tax burdens is not 14 We also compare the firms in the F-corp sample to their industry counterparts on Compustat. For each firm in the F-corp sample, and the counterpart D-corp firm, the year of equity investment is determined and data for all firms in Compustat for this year are gathered. Means are calculated for several variables (sales, total assets, tax burden, leverage, and income) for all firms in the same four-digit industry code (Compustat variable DNUM) as the sample firms. For each variable, the difference between the value of the sample firm and the mean of all its industry counterparts is calculated. Cross-sectional tests of the differences indicate that the F-corp firms are smaller in terms of both sales and total assets, have a lower tax burden, have about the same level of financial leverage, and have a slightly lower net income (as a percentage of total assets). COMPARISON OF F-CORP FIRMS TO ENTIRE COMPUSTAT DATABASE OF INDUSTRY COUNTERPARTS Compustat Industry Mean Value Minus Sample Firm Value T-Value Sign Rank Mean Median (Pr > T ) (Pr > = S ) Millions of U.S. Dollars Sales (0.2463) (0.0129) Assets ( ) ( ) Net Income (Scaled by assets) ( ) (0.0857) Tax Expense (Scaled by assets) ( ) (0.0024) Debt leverage (0.9023) ( ) 15 There are many ways to measure tax burden. We measure tax burden as current taxes divided by total assets. This definition is consistent with most prior studies we have cited. Although some studies of tax burden have scaled taxes by income, we avoid this approach because income itself may be affected by any manipulations used to reduce taxes. 14

7 Tax Burden of U.S. Corporations Having Substantial Foreign Ownership TABLE 1 MEASURES OF TAX BURDEN ACROSS TEST AND CONTROL SAMPLES Panel A Tax Burden Measures F-Corp F-Match D-Corp D-Match Mean Median Mean Median Mean Median Mean Median Event year Event year Panel B Measure of Tax Burden Differences F-Corp Minus D-Corp Mean Median Event year Event year * 0.004** Panel C Measure of Tax Burden Differences F-Corp Minus F-Match D-Corp Minus D-Match Mean Median Mean Median Event year * * 0.011** Event year * 0.004** *(**)Indicates significant difference at the 0.05 level based on a t-test (sign rank test). Tax burden definition: Total current tax divided by total assets [Compustat: (Item 16 Item 50)/Item 6] statistically significant. Accordingly, we conclude that prior to the investment, the F-corp and D-corp samples have similar tax burdens. 17 Next, we compare tax burdens in the second year following the year of acquisition, event year 3. This particular event year is selected because new investors have had a sufficient opportunity to effect contractual changes that affect tax burden, but it is still close enough to the investment date that we can reasonably infer observed differences are caused by the ownership turnover event. 16 The following are size measures for the four samples used in the study for event year 0. None of the differences between the F-corp and D-corp samples, F-corp and F-match samples, and D-corp and D-match samples are statistically significant at the 0.05 level (based on a t-test and nonparametric Wilcoxon test). F-Corp Sample F-Match Sample D-Corp Sample D-Match Sample Millions of U.S. Dollars Net sales Mean Standard Deviation , , ,789.8 Median Total assets Mean Standard Deviation 1, , , ,649.3 Median Net income Mean Standard Deviation Median We also attempt to determine whether the tax burdens of these two samples are typical relative to the population of foreign-controlled U.S. corporations. The only publicly accessible information about tax payments 15

8 NATIONAL TAX JOURNAL The data in Table 1 suggest that the measures of tax burden have declined for F-corp firms relative to the measures reported for year 0. However, the tax burden measures for the D-corp firms have changed little relative to the numbers reported for year 0. Furthermore, as reported in Panel B, the tax burdens of the F-corp firms are significantly lower than those of the D-corp firms by year 3. The difference is statistically significant at the 0.05 level in year 3 but is not significantly different from zero in year 0. We now consider the possibility that our prior comparisons have offered no control for many variables that purportedly affect firms tax burdens and that are likely to differ between the two samples. Two variables, which have been shown in prior research to affect relative explicit tax burdens, are industry and capital intensity (Stickney and McGee, 1982). To develop controls for these variables, we identify two additional samples of firms. 18 One additional sample is selected to reflect the F-corp sample in terms of industry and size (F-match). Firms in this sample are selected based on a match of four-digit SIC code, book value of total assets, and net sales in the year preceding the foreign investment. 19 The two size measures are equally weighted in the selection process. This sample is selected from firms listed in the annual Compustat database. A fourth sample (hereafter referred to as D-match) is likewise selected to match the D-corp sample on the dimensions of size and industry. Our analysis now involves four, nonoverlapping samples of firms. Empirically, we now compare the F-corp and D-corp samples relative to their respective matched samples. 20 In Table 1, Panel A, we present the mean and median tax burden measures for the match samples in years 0 and 3. In Panel C of Table 1, we present differences in tax burden measures between the F-corp and D-corp samples and their sizeand-industry matched samples. We conduct tests to determine whether the differences vary significantly from zero. For year 0, the F-corp sample has a lower tax burden than the F-match sample; the difference is statistically significant (parametric test only). Similarly, the D-corp sample has a significantly lower tax burden than the D-match sample in year 0. The finding of a lower tax burden in the F-corp sample relative to the F-match sample tends to be mildly contrary to earlier studies that have found the preacquisition profitability of foreign-conof foreign-controlled U.S. corporations is SOI data. Because the SOI data cannot be examined on an event year basis, we use SOI data from The year 1987 was selected because it is the weighted average year in our sample. For 1987, one percent of the total number of U.S. corporations were foreign controlled ( controlled is defined as direct or indirect ownership of 50 percent or more of voting stock as of the end of the tax year). These foreign-controlled corporations accounted for over seven percent of the receipts, and over six percent of the assets, reported on U.S. corporation income tax returns. For these corporations, the ratio of total income tax expense to total assets was This ratio is slightly higher than those reported for the F- corp sample in Table 1 and lower than the counterpart ratio of the D-corp sample. The tax burden mean for the F-corp sample can be contrasted with other domestic corporations. For these corporations, the ratio of income tax expense to assets was 0.01, which is very similar to the level reported for the D-corp sample in Table 1. Our source: Hobbs, J. Domestic Corporations Controlled by Foreign Persons, In Statistics of Income, Bulletin 10 No. 1 (Summer, 1990): We explicitly provide a control only for industry class. However, it is reasonable to presume that industry class is also an effective control for capital intensity. 19 Industry class was determined by the Compustat variable DNUM. Total assets and sales were measured, respectively, by Compustat annual data items I6 and I Both a t-test and Wilcoxon test confirm that no significant size differences exist between the F-corp and D- corp samples and their respective size and industry match samples (at the two-tailed 0.05 level). 16

9 Tax Burden of U.S. Corporations Having Substantial Foreign Ownership trolled U.S. firms to be nearly equal to the average U.S. company (Grubert et al., 1993). When the measures of tax burden in year 3 are compared to those in year 0, they are lower for all samples. This result occurs, in part, because year 0 falls before 1986 and year 3 falls after 1986 for many sample firms. 21 Also noteworthy is that the tax burden of the F-corp sample is significantly lower than that of the F-match sample in year 3. However, by year 3, there is no longer a statistically significant difference in the tax burdens of the D-corp and D-match samples. We conclude that tax burdens of the F-corp and D-corp samples move from year 0 to 3 in opposing directions relative to their match samples: F-corp tax burdens decline, relatively, and D-corp tax burdens increase, relatively. Next, we develop a multivariate regression model. The intent is to examine the change in intrafirm tax burden from event year 0 to 3 for the four samples. We include variables in the model to proxy for specific types of operational and financing changes that managers might make which reduce tax burdens without affecting economic profit generation. We also include a variable to capture effects of the Tax Reform Act of The model follows. [1] Tax-ch = α + β 0 Tax 0 + β 1 Code-ch + β 2 Debt-ch + β 3 PPE-ch + β 4 Temp-ch + β 5 Size + β 6 Foreign + β 7 Acquisition + β 8 Foreign*Acquisition + β 9 Foreign*Debt-ch + β 10 Foreign*PPE-ch where + β 11 Foreign*Temp-ch + β 12 Foreign*Size + β 13 Acquisition*Debt-ch + β 14 Acquisition*PPE-ch + β 15 Acquisition*Temp-ch + β 16 Acquisition*Size + β 17 Foreign*Acquisition*Debt-ch + β 18 Foreign*Acquisition*PPE-ch + β 19 Foreign*Acquisition*Temp-ch + β 20 Foreign*Acquisition*Size Tax-ch is the tax burden in event year 3 less the tax burden in event year 0; Tax 0 is the tax burden in year 0; Code-ch is coded 1 if the period from event year 0 to 3 includes 1986 (Tax Reform Act of 1986) and is coded 0 otherwise; Debt-ch is change in long-term debt, scaled by assets, from event year 0 to 3; PPE-ch is change in gross property, plant, and equipment, scaled by assets, from event year 0 to 3; Temp-ch is change in deferred taxes, scaled by assets, from event year 0 to 3; Size is the log of sales in event year 0; Foreign is coded 1 if the firm is in the F- corp or F-match samples and coded 0 otherwise; and Acquisition is coded 1 if the firm is in the F-corp or D-corp samples and coded 0 otherwise. Tax 0 is included in the model to control for mean reversion of the tax burden 21 Effective tax rates after 1986 would possibly be lower due to the Tax Reform Act of

10 NATIONAL TAX JOURNAL measure over time and to proxy for any firm-specific characteristics that affect tax burden that are not captured by other independent variables. Code-ch is included in the model to capture statutory tax rate effects of the Tax Reform Act of We expect a negative coefficient for this variable. Debt-ch is included in the model to capture tax burden effects of a change in leverage. We expect the coefficient of this variable to be negative. PPE-ch is intended to capture tax burden effects attributable to a change in the level of depreciable assets. We expect the coefficient of this variable to be negative. Other than increasing debt and the level of depreciable assets, tax burdens can also be affected by changes in rates of depreciation, depletion, or amortization. Tempch is a variable that is included in the model to capture timing effects of expense recognition. This variable is expected to have a negative coefficient. Although we have no expectation as to its coefficient, Size is included in the model to capture any unspecified size effects. This model allows each sample to have unique intercept and slope coefficients. We have no interest in the estimated slope coefficients per se; our interest is in the intercept terms. A statistically significant intercept is taken as evidence that the changes in the contractual variables (e.g., change in debt level) fail to account for the changes in tax burden. The primary intercept variable of interest in the model is Foreign*Acquisition. It captures the extent to which significant differences in tax burdens for the F-corp and other samples persist in the presence of the explanatory power of the other contractual variables. Based on the previous analyses, we expect the coefficient for this variable to be negative, unless the change in contractual variables can fully account for the reduction in tax burden of the F-corp sample occurring between year 0 and 3. The results of the regression are presented in Table 2. Because only one estimated coefficient is central to the purpose of developing the regression model, we limit our discussion of the estimated model. The Foreign*Acquisition variable is statistically significant and is negative, indicating that even after accounting for the typical types of contractual changes managers make to reduce taxes, a significant, unexplained reduction in tax burden from event year 0 to 3 persists for the F-corp sample. Thus, the relative reduction in tax burdens of the F-corp sample is not explained by the usual types of contractual changes that reduce taxes without negatively affecting the firm s profit generating capacity (i.e., its market value). Test of Market Returns As a final analysis, we attempt to determine whether the changes in tax burdens between the F-corp and D-corp samples are related to an underlying change in profit generation. If a change in profitability is the cause of the reduced tax burden of the F-corp sample, relative to the D-Corp sample, the profit effect should be mirrored in market returns to equity. Using data from the Center in Research in Securities Prices (CRSP), we accumulate daily market returns 22 for three years (750 days) subsequent to the event date for the F-corp and D-corp samples. 23 In Panel A of Table 3, we provide the three-year cumulative returns for the F- corp and D-corp samples. The data indicate that the relative reduction in tax burden of the F-corp sample is associated with a reduction in profitability rather 22 We delete any observations with more than 50 days of missing data. All returns are adjusted for the effects of cash and stock dividends and stock splits. 23 The three-year return period was selected to correspond in event time to the end of event year 3. 18

11 Tax Burden of U.S. Corporations Having Substantial Foreign Ownership TABLE 2 REGRESSION ANALYSIS OF YEAR 0 TO 3 TAX BURDEN CHANGE Model: Tax-ch = α + β 0 Tax 0 + β 1 Code-ch + β 2 Debt-ch + β 3 PPE-ch + β 4 Temp-ch + β 5 Size + β 6 Foreign + β 7 Acquisition + β 8 Foreign * Acquisition + β 9 Foreign * Debt-ch + β 10 Foreign * PPE-ch + β 11 Foreign * Temp-ch + β 12 Foreign * Size + β 13 Acquisition * Debt-ch + β 14 Acquisition * PPE-ch + β 15 Acquisition * Temp-ch + β 16 Acquisition * Size + β 17 Foreign * Acquisition * Debt-ch + β 18 Foreign * Acquisition * PPE-ch + β 19 Foreign * Acquisition * Temp-ch + β 20 Foreign * Acquisition * Size Variable Intercept Tax 0 Code-ch Debt-ch PPE-ch Temp-ch Size Foreign Acquisition Foreign * Acquisition Foreign * Debt-ch Foreign * PPE-ch Foreign * Temp-ch Foreign * Size Acquisition * Debt-ch Acquisition * PPE-ch Acquisition * Temp-ch Acquisition * Size Foreign * Acquisition * Debt-ch Foreign * Acquisition * PPE-ch Foreign * Acquisition * Temp-ch Foreign * Acquisition * Size R 2 = 0.43; adjusted R 2 of model = 0.38 N = 280 Variable definitions: see Table 3. Parameter Estimate T-Statistic P-Value than a manipulation of taxable income. The three-year market return is significantly lower for the F-corp sample than for the D-corp sample. To further explore the relationship between market returns and change in tax burdens, we regress the three-year market returns on the change in tax burdens from event year 0 to 3 for the F-corp and D-corp samples. We include the year 0 tax burden as a control variable. Our empirical model is as follows: [2] Return = α + β 1 Tax-ch + β 2 Tax 0 + β 3 F-corp + β 4 F-corp *Tax-ch + β 5 F-corp*Tax 0. F-corp is a dummy variable taking the value 1 if the sample firm is a member of the F-corp sample and taking the value 0 otherwise. Our expectation is that, if the change in tax burden results from manipulation rather than from a change in profitability, then the change in tax burdens should better explain market returns for the D-corp sample than for the F-corp sample. Hence, we expect β 1 to be significantly positive and β 4 to be significantly negative if the relative change in tax burden of the F-corp sample is attributable to income manipulation rather than to a decline in real profitability. We expect β 1 to be significantly positive and β 4 to be insignificantly different from zero if 19

12 NATIONAL TAX JOURNAL TABLE 3 ASSOCIATIONS BETWEEN MARKET RETURNS ON COMMON EQUITY AND CHANGES IN TAX BURDENS Panel A Three-Year Cumulative Market Return on Common Equity F-Corp D-Corp Mean cumulative market return 9.43% 41.47% The difference in market returns between the F-corp and D-corp samples is significant at the 0.10 level (0.01 level) based on a two-sample t-test (Wilcoxon rank sum test). Model: Panel B Regression of Three-Year Cumulative Return on Equity on Change in Tax Burden Return = α + β 1 Tax-ch + β 2 Tax 0 + β 3 F-corp + β 4 F-corp * Tax-ch + β 5 F-corp * Tax 0 Variable Intercept Tax-ch Tax 0 F-corp F-corp * Tax-ch F-corp * Tax 0 R 2 = 0.31; Adjusted R 2 of model = 0.28 Parameter Estimate T-Statistic P-Value Variable definitions: Return is the three-year cumulative market return on equity; Tax-ch is the tax burden measure in event year 3 less the tax burden measure in event year 0; Tax 0 is the tax burden measure for event year 0; F-Corp is a dummy variable that is coded 1 if the firm is a member of the F-corp sample and is coded 0 otherwise; Code-ch is coded 1 if the period from event year 0 to 3 includes 1986 (Tax Reform Act of 1986) and coded 0 otherwise; Debt-ch is change in long-term debt, scaled by assets, from event year 0 to 3; PPE-ch is change in gross property, plant, and equipment, scaled by assets, from event year 0 to 3; Temp-ch is change in deferred taxes, scaled by assets, from event year 0 to 3; Size is the log of sales in event year 0; Foreign is coded 1 if the firm is in the F-corp sample and is coded 0 otherwise; Acquisition is coded 1 if the firm is in the F-corp or D-corp samples. relative reduction in tax burden of the F- corp reflects a decline in real profitability. The regression results appear in Panel B of Table 3. The results indicate that the change in tax burden explains the market returns equally well for both samples. There is no structural difference in the relationship between tax burdens and market returns between the two samples as indicated by the insignificant interaction terms. Thus, our results are similar to those of Collins et al. (1997) as is our conclusion... that unidentified systematic differences between foreign- and U.S.-controlled companies, rather than income shifting, account for the [tax] differential 20 between foreign- and U.S.-controlled companies... (p. 80). In the following section, we summarize our study and identify future research possibilities. SUMMARY In this study, we attempt to determine if there is empirical support for the common perception that foreign-owned U.S. firms pay lower levels of taxes than other U.S. firms relative to the levels of profit they generate. This study contributes to the literature in several ways. First, compared to prior studies, we utilize a differ-

13 Tax Burden of U.S. Corporations Having Substantial Foreign Ownership ent database and different empirical proxies. Second, we analyze the data in event time. Third, our approach controls for generic effects associated with ownership turnover. Although it was not an explicit objective, the paper contributes in a fourth way. The firms analyzed in this study have foreign ownership interests that are substantially below traditional thresholds for studies in foreign-control tax effects. Similar to prior studies, we find that U.S. firms that become investees of foreign firms or individuals have a reduction in their tax burdens relative to other U.S. firms. We depart from prior studies in explaining this outcome. We find no evidence that this result is attributed to effects of ownership concentration, greenfield start-up effects, or agency costs relating to management control. More importantly, we find implausible the explanation that income manipulation accounts for the reduction in tax burden of the foreign-influenced sample. Our results indicate the foreign-influenced sample experienced a real decrease in profitability relative to a benchmark sample of firms. Because we find the reduction in profitability to be associated with a reduction in market return on equity, we conclude that the profit reduction is not a mere facade to reduce tax payments the equity market prices it and, hence, it is real. Thus, the most likely explanation of the reduced tax burden experienced by foreign-influenced firms is that the foreign investors are less successful than are domestic investors in identifying more profitable U.S. firms for investment. Accordingly, our conjecture is that foreign investors have an information disadvantage in evaluating investment targets. Future research should evaluate this effect in the context of prior studies that have attributed the reduced tax burden of foreign-controlled U.S. firms to income manipulation. 21 Acknowledgments The authors wish to thank Robert Trezevant, Wanda Wallace, Bill Cready, Bill Wempe, and workshop participants at Texas A&M University and the University of Nebraska for helpful comments on earlier drafts of this paper. Special thanks go to the editors and anonymous reviewers of the National Tax Journal for their many insightful comments and to Mike Wilkins for his expert aid in computing stock market returns. REFERENCES Collins, Julie H., Deen Kemsley, and Douglas A. Shackelford. Transfer Pricing and the Persistent Zero Taxable Income of Foreign-Controlled U.S. Corporations. The Journal of the American Taxation Association 19 Suppl. (1997): Goldberg, Fred. Goldberg Provides Income Evidence for Pickle s Section 482 Hearing on Foreign- Controlled Companies Taxes. Tax Notes International Doc (July 10, 1990): Grubert Harry. More on the Low Taxable Income of Foreign-Controlled Companies in the United States: Update of the 1991 Paper. U.S. Treasury Department Working Paper. Washington, D.C.: U.S. Treasury Department, Grubert, Harry, and John Mutti. Taxes, Tariffs and Transfer Pricing in Multinational Corporate Decision Making. The Review of Economics and Statistics 73 No. 2 (1991): Grubert, Harry, Timothy Goodspeed, and Deborah Swenson. The Taxable Income of Foreign Controlled Companies in the U.S.: New Empirical Evidence. In Studies in International Taxation, edited by A. Giovannini, R. G. Hubbard, and J. Slemrod, Chicago: The University of Chicago Press, 1993.

14 NATIONAL TAX JOURNAL Harris, David, R. Morck, Joel Slemrod, and Bernard Yeung. Income Shifting in U.S. Multinational Corporations. In Studies in International Taxation, edited by A. Giovannini, R.G. Hubbard, and J. Slemrod, Chicago: The University of Chicago Press, Hunt, Herbert, III. The Separation of Corporate Ownership and Control: Theory, Evidence, and Implications. Journal of Accounting Literature 5 (1986): Scholes, Myron, and Mark Wolfson. Taxes and Business Strategy (A Planning Approach). Englewood Cliffs: Prentice-Hall, Spooner, Gillian. Effective Tax Rates from Financial Statements. National Tax Journal 39 No. 3 (September, 1986): Stickney, Clyde P., and Victor E. McGee. Effective Corporate Tax Rates: The Effect of Size, Capital Intensity, Leverage, and Other Factors. Journal of Accounting and Public Policy (1982): Turro, John. U.S. Congressional Committee Blasts Foreign Firms or Tax Dodging. Tax Notes International (August 1, 1990):