Using a sample of privately held C corporations and S corporations from the motor carrier industry during 198492, we assess the effect of the 1986 Tax Reform Act on the amount of corporate income shareholders of privately held C corporations shifted to their personal tax bases. We estimate that the C corporations shifted a mean of $130,587 taxable income each year to shareholders (representing 29percent of their mean accounting earnings before income shifting) after the 1986 tax law change. The C corporations used deductible managerial compensation and rent expense, but not interest expense, to shift income to shareholders.
1. Introduction
Most tax research is devoted to publicly traded corporations (Shackelford and Shevlin 2001). Due to data limitations, there is little research on tax planning in privately held taxable corporations. Although both publicly traded firms and privately held firms have incentives to avoid corporate taxation, privately held firms have more opportunities to avoid taxes because share ownership is highly concentrated and the owners are usually employees, lessors, or creditors of the same firm (Nagar, Petroni, and Wolfenzon 2001). Using firmlevel data from the motor carrier industry during 198492, we estimate the amount of corporate income shifted to shareholder tax bases due to the 1986 Tax Reform Act (hereafter TRA86) by privately held taxable corporations. The motor carrier industry offers a rich source of publicly available data on privately owned firms as trucking companies were required to file annual reports that contain detailed financial and operating data during the period of our study.
TRA86 significantly altered the tax rate structures of individual and corporate taxpayers. As a result, we expect that privately held taxable motor carriers shifted income to shareholders after TRA86 to minimize the combined corporate and shareholder tax liabilities. To identify the amount of taxinduced income shifting, we compare the amount of pretax corporate income (deflated by total revenues) for a sample of 6,839 privately held C corporations and S corporations (hereafter referred to as Ccorps and S corps, respectively) over the period 198492. Although both types of motor carriers are highly concentrated in stock ownership, earnings from Ccorps are subject to immediate corporate taxation and additional future shareholder taxation when dividends are distributed or shareholder capital gains are realized. In contrast, because earnings from Scorps are taxed only once but immediately at the shareholder level, Scorps have no incentive to otherwise shift income to shareholders. As a result, Scorps are used as a benchmark to detect the amount of taxinduced income shifting for Ccorps.
Consistent with income shifting, we find that Ccorps reported lower pretax accounting rates of return than Scorps after TRA86. We find that Ccorps used tax deductible managerial compensation and rent expenses (but not interest expense) to shift income to shareholders after TRA86. Furthermore, Ccorps that faced high tax costs of retaining corporate earnings shifted more income to shareholders after TRA86. Consistent with the hypothesis that costs of income shifting increase with the number of shareholders, we find that Ccorps with one single shareholder reported lower corporate rates of return after TRA86 than Ccorps with multiple shareholders. Our estimates indicate that the Ccorps shifted a mean of $130,587 taxable income per year (or 29 percent of the mean accounting earnings before income shifting) to shareholders after TRA86.
This study expands our understanding of tax planning in privately held Ccorps. Although taxinduced income shifting to shareholders by privately held Ccorps is anticipated, the actual amount of income shifting has not been well documented. Using aggregate Statistics of Income (SOI) data published by the Internal Revenue Service, Wilkie, Young and Nutter (1996) report that the ratio of “deductible dividends” (i.e., rental, interest and officers’ compensation expenses) to distributable income (net income plus deductible dividends) increased for small relative to large Ccorps following TRA86. Gordon and Slemrod (2000) investigate the effect of changes in personal relative to corporate tax rates on the aggregate taxable labor compensation reported by individuals in the top half of the labor income distribution each year. Gordon and Slemrod estimated that a one point increase in the spread between personal and corporate tax rates raises labor income by 3.2% and lowers reported corporate rate of return on assets by 0.147%. Due to data limitations, Wilkie et al. (1996) and Gordon and Slemrod (2000) do not control for nontax factors correlated with tax rate changes. Furthermore, examinations of changes over time in the aggregate income distribution are unable to directly trace the shifting of income by specific firms to its owners.
Ke (2001) uses individual firm data from the property and liability insurance industry to provide direct evidence that from the period 198892 to 199396, privately held managementowned insurers used tax deductible compensation expense to shift less income from corporate to shareholder tax bases due to the significant increase in individual tax rates in 1993. This study extends Ke (2001) in several respects. First, we examine a different time period when individual tax rates decreased rather than increased relative to corporate tax rates. The expected income shifting in our study and Ke are in opposite directions. Therefore, our findings should provide further assurance that Ke’s results are not due to omitted nontax effects. Second, we examine a nonfinancial industry. Because Ke’s results are based on a sample of heavily regulated financial companies, it is unclear whether his findings can be extended to other industries. Finally, we examine both the aggregate magnitude and specific mechanisms of taxinduced income shifting. Ke (2001) examines only deductible compensation.
The magnitude of taxinduced income shifting from corporate to shareholder tax bases provides useful information for assessing the excess burden (or inefficiency) of income tax changes.[1] The excess burden of a tax increase results from taxpayers’ behavioral responses to the tax increase and equals the difference in the actual tax revenues collected and the hypothetical tax revenues collected from a lump sum tax (i.e., a tax that does not affect taxpayers’ behaviors) that gives taxpayers the same utilities (Stiglitz 1988, 113). Thus, a critical input needed to calculate the excess burden is how and how much taxpayers respond to marginal tax changes. Prior tax research commonly ignores income shifting between corporate and personal tax bases in computing the excess burden of individual or corporate tax changes (see e.g., Feldstein 1995). However, as Gordon and Slemrod (2000) indicate, to the extent that the magnitude of income shifting between corporate and personal tax bases is significant, the existing excess burden calculations of individual tax increases are overstated. Our study provides direct evidence on the magnitude of income shifting.
2. Research Hypotheses
2.1. TRA86 and Income shifting from Corporate to Shareholder Tax Bases
Shareholders are subject to double taxation on their investment income from C corps. Corporate earnings are first taxed at the corporate rate when earned and at the shareholder rate when aftertax corporate earnings are distributed to shareholders in the form of dividends or when shareholders sell their stocks and realize capital gains.
To avoid the double taxation on their investment income, shareholders could distribute corporate earnings to themselves in the form of “deductible dividends”, i.e., payments that are deductible by the firm and (often) taxable to shareholders. Deductible dividends essentially convert a Ccorp to an Scorp, which is subject to only shareholder tax. Shareholders of privately held Ccorps are often employees, lessors, or creditors of the same firms. As a result, they have the flexibility to avoid the double taxation by paying themselves generous compensation, interest, rent or any other forms of deductible expenses. A direct consequence of paying deductible dividends to shareholders is that corporate accounting rates of return will be lower than otherwise.
The incentive to avoid the double taxation depends on the difference in current individual and corporate marginal tax rates (denoted τi and τc, respectively) and the annualized effective tax rate on future capital gains from the sale of the stock (denoted τa) (see Scholes et al. 2002, 7273). If corporate earnings were shifted to shareholders and R is the beforetax corporate rate of return, the annual aftertax rate of return to shareholders would be
If corporate earnings were retained within the Ccorp, the earnings would be subject to double taxation and the annual aftertax rate of return to shareholders would be
Thus, shareholders prefer income shifting if (a)(b) is greater than zero. That is,
Prior to TRA86, the top corporate and individual statutory tax rates were 46% and 50%, respectively; After TRA86, the two tax rates dropped to 34% and 28%, respectively.[2] In addition, the maximum longterm capital gains tax rate increased from 20% to 28% after
TRA86. Thus, τa increased after TRA86. Because shareholders of privately held firms are H1 eteris paribus, corpss beforetax accounting rates of return declined after TRA86. 21 eteris paribus, corpss taxdeductibce compensation to officers, interest expense, and rent expense increased after TRA86.
Because accounting rates of return and taxdeductible expenses could change even absent tax changes, we use a sample of Scorps in the same motor carrier industry to control for nontax trends over time when testing the above two hypotheses. As explained previously, shareholders of Scorps are taxed only once on the earnings from the Scorps; thus, they have no need to shift income from corporate to shareholder tax bases.
2.2. Constraints of income shifting
Income shifting from corporate to shareholder tax bases is the easiest when a firm is one hundred percent owned by one shareholder because both deductible dividends and investment income accrue to the same person. Income shifting becomes more complicated when there are multiple shareholders with different marginal tax rates or when some shareholders are not employees, lessors, or creditors and thus cannot receive deductible dividends from the firm. In these cases, deductible dividends to one shareholder represent a direct loss to another who does not receive deductible dividends in proportion to her stock ownership. As a result, we also examine H1 and H2 separately for firms owned by one shareholder and firms owned by multiple shareholders.
2.3. Income shifting from corporate to nonshareholders
Up to this point our analysis has focused on income shifting between Ccorps and their shareholders. As indicated by Scholes et al. (2002), Ccorps could also shift income to taxpayers who are not shareholders (e.g., nonshareholder employees) to minimize their combined tax liabilities. Such shifting could also affect corporate rates of return, managerial compensation, interest and rent expenses, to the extent that these expenses are not paid to shareholders only. However, such income shifting should not materially affect our hypotheses above, since the amount of income shifting between Ccorps and nonshareholders is likely limited. For example, the amount of income shifting between a Ccorp and its nonshareholder employees is limited to the economic fair value of the labor services provided.[3] In addition, shifting income between the firm and its non shareholders is more difficult to implement because there are more nonshareholders than shareholders and variation in marginal tax rates is greater for nonshareholders than shareholders. Even if income shifting between the Ccorp and its nonshareholders is feasible, it is likely to be limited to the two years before and after significant tax law changes due to incentive problems. During the years with no changes in marginal tax rates, the tax benefit of intertemporal income shifting between Ccorps and non shareholders is minimal.
To eliminate any intertemporal income shifting between Ccorps and nonshareholders around tax law changes, we also estimated our regressions excluding the years around TRA86 (i.e., 198688) and obtained similar conclusions. Thus, our empirical findings cannot be attributed to income shifting from Ccorps to non shareholders.
3. Sample Selection
This study used financial and operation data reported on Motor Carrier Annual Reports (MCAR) that were filed annually with the Interstate Commerce Commission (ICC). We obtained these data from Martin Labbe Associates in electronic format. Several previous studies (e.g., Schipper et al. 1987, Enis and Morash 1985) used the same data. For the period of this study, all motor carriers of property with annual operating revenues of at least $1 million were required by law to file annual reports and base their financial reporting on “The Uniform System of Accounts” (Title 49 CFR §1249.2) (American Trucking Association 1991). We chose the period 198492 simply because the corporate and individual tax rates changed significantly as a result of TRA86. We do not extend our analysis beyond 1992 because most trucking firms stopped filing information with ICC.[4]
We started with an initial sample of 11,151 motor carriers (firm years) that contain no missing data, are not classified as household goods carriers and report zero household goods operating revenues or expenses. Because household goods carriers followed a different reporting format, we excluded them from our sample. We excluded
406 partnerships and sole proprietorships (firm years) because the owners of these organizations assume unlimited liabilities while both Ccorp and Scorp have limited liabilities. Our results were similar if the partnerships and sole proprietorships were included in the control sample. We also excluded 934 firm years because they were publicly traded or owned by a publicly traded firm. We used Dun and Bradstreet’s America’s corporate Famicies (19851989 editions) and the CRSP files (SIC code 421) to determine whether a trucking firm is public or private.
As the electronic data source we use does not directly disclose whether a privately held trucking company is a Ccorp or Scorp, we used the following convention to classify a firm as either a C or Scorp. Four distinctive attributes of Scorps are: (1) they pay no federal income taxes; (2) they are allowed to issue only one class of stock; (3) they must have no more than 35 shareholders during our sample period; and (4) they cannot be owned by foreign or corporate shareholders. We used the first three attributes to identify the Scorps in our sample. We could not use the fourth attribute due to a lack of data. Specifically, we assume that a motor carrier is an Scorp over the entire period
198492 if it reported only one class of stock, zero current and deferred federal corporate income taxes, and the maximum number of officers in the firm was no more than 35 throughout the entire period198492. We do not have data on the actual number of shareholders. However, Nagar, Petroni, and Wolfenzon (2001) find that managers of privately held corporations are usually shareholders. Thus, we used the number of officers as a proxy for the number of shareholders in the firm. Data on the number of officers for the period 198487 are unavailable. Thus, we use the maximum number of officers during 198892 as a proxy for the number of shareholders. This procedure
yielded 1,268 Scorp years and 8,543 Ccorp years.[5]
Previous research indicates that many small Ccorps were converted to Scorps after the 1986 Tax Act (see e.g. MacKieMason and Gordon 1997). To make sure that our sample of Ccorps does not contain such converted Scorps, we eliminated from the C corp sample those firms that did not report current and deferred federal income taxes during 198892.[6] This restriction reduced the Ccorp sample by2,168 firm years.[7]
Finally, we required each firm to have at least one observation in each of the two periods
198487 and 198892. The final sample contains 1,028 Scorp years and 5,811 privately held Ccorp years, representing 151 unique Scorps and 780 unique Ccorps.
Our entity classification criteria could misclassify some firms. As Scorps have no incentive to shift income to shareholders, misclassifying Scorps to be Ccorps reduces the power of our regression tests. Misclassifying Ccorps to be Scorps should not affect our inference. This is because our classification criteria (1) and (2) ensure that the misclassified Ccorps were likely to have low marginal tax rates and Ccorps with low marginal tax rates had little incentive to use deductible dividends to shift income to shareholders. As a result, like Scorps, these misclassified Ccorps could be used as benchmark to identify the magnitude of income shifting for the Ccorps with high marginal tax rates (see Section 5.3 for direct evidence).
To directly verify the accuracy of our classification criteria, we also obtained the filing status (i.e., C vs. S corp) of our sample firms from the hard copies of motor carrier annual reports from the Bureau of Transportation Statistics of the Department of Transportation for years 1996 and 1997. The annual reports stopped disclosing the filing status after 1997; hard copies for the years prior to 1996 are not available. Since our sample period is much earlier than 1996, any discrepancy between our ownership classification and that from the hard copies could be due to the election or revocation of Scorp status after our sample period rather than errors in our classification.
Of the 931 trucking firms in our sample (780 Ccorps and 151 Scorps determined based on our classification criteria), we obtained the filing status for 358 firms (38%) from the hard copies.[8] Of the 358firms, 326 firms were Ccorps and 32 Scorps during
198492 using our classification criteria. Of the 326 Ccorps determined based on our
classification criteria, 260 (80%) were Ccorps and 66 Scorps according to the hard copies of annual reports for years 199697. These statistics suggest that our Ccorps are classified with reasonable accuracy. In addition, we believe that the discrepancy between our classification scheme and the hard copies for the 66 firms is more likely due to the change in filing status from C to S after our sample period. The reason is that the 66 firms reported positive current federal income taxes for twothirds of the sample years 8492 and thus could not be Scorps during our sample period.
Of the 32 Scorps determined using our classification criteria, 15 (47%) were S corps and 17 Ccorps according to the hard copies of annual reports for years 199697. We are not sure whether the discrepancy between our classification scheme and the hard copies for the 17 firms is a result of misclassification or revocation of the Scorp election after our sample period. However, even if the 17 firms were indeed misclassified, the misclassification would still not affect our inference because, as argued above, these firms were likely to have low marginal tax rates and thus had little incentive to shift income to shareholders. Based on the above analysis, we conclude that our ownership classification is reasonably reliable and would not bias our inference.
4. Descriptive Statistics
We first report univariate statistics for the Ccorps and Scorps over 198492 in Table 1. In terms of total revenues or total assets, the Ccorps are significantly larger than the Scorps. GEN denotes carrier type and is defined as one if it is a general freight carrier and zero otherwise. Although the trucking firms are classified into seven different categories (e.g., general freight, bulk commodities, building materials, and etc.), general freight is the most common type. The samples of Ccorps and Scorps contain a similar percentage of general freight carriers. However, we control for the seven carrier types in later regression analyses.
ROR is net income before taxes divided by total revenues. An alternative deflator is total assets. We favored total revenues over total assets as a deflator because revenues are neutral regarding a carrier’s mix of leased versus owned revenue equipment.
Furthermore, ROR is one minus the operating ratio. The operating ratio is operating expenses divided by revenues and is an important performance measure used to evaluate transportation companies. Over the entire sample period 198492, the Ccorps were more profitable than the Scorps. This result is not surprising because Scorps had been traditionally used to pass business losses to shareholders(e.g., Plesko 1994).
The last three variables in Table 1, OFFSAL, INTEREST, and RENT, represent tax deductible compensation to officers and managers, interest expense, and equipment rent expense respectively, all expressed as a percentage of total revenues. INTEREST is small relative to OFFSAL and RENT for both corporate entities. There is no significant difference in OFFSAL and RENT between the two types of corporations over 8492, but INTEREST is significantly higher for the Ccorps than for the Scorps.
5. Empirical Results
5.1. Return on revenues (ROR)
Before investigating the exact mechanisms Ccorps used to avoid corporate taxes, we first provide evidence on the effect of TRA86 on the total magnitude of income shifting by the Ccorps. After observing significant fluctuations in corporate profits from the seventies to eighties, some tax scholars (e.g., Gordon and Slemrod 2000) speculated that the changes in corporate profits were due to tax law changes. Although we cannot examine corporate profits for the entire economy, we provide firmlevel direct evidence that changes in corporate profits for our sample of Ccorps were at least partially due to taxinduced income shifting. We use the following regression model to test our H1 (for simplicity, firm and time subscripts are omitted):
Because TRA86 went into full effect in 1988, we bifurcated the sample period with YR88_92. The results were similar if the transition year 1987 were deleted or treated as part of the period post TRA86. We predict that the coefficient on YR88_92*PRIVC is negative as a result of TRA86 (H1).[9] Although anecdotal evidence suggests that Ccorps had incentives to retain income before 1988, we do not predict the coefficient on PRIVC because it may also capture nontax differences for which we could not control between the two types of entities.
The remaining variables are control variables. YR88_92 captures nontax trends over the two periods. STATE controls for any statespecific fixed effects. Ln(REV) and LINES control for firm size and motor carrier type, respectively. The financial performance of motor carriers that haul specialized commodities such as petroleum, automobiles, building materials, and agricultural products is tied to the economic conditions of the segments that they serve (Wilson 1980). In contrast, the performance of general freight carriers hinges on conditions in the economy as a whole. Furthermore, specialized carriers face greater competition from railroads in that they tend to haul larger shipments of “bulkier” and lowerunit value commodities (Boyer 1977). Prior research (e.g., Enis and Morash 1993, Morash and Enis 1982) uses size and freight type to control for economic performance in the motor carrier industry.
The OLS regression result of ROR for the full sample is reported in Table 2. Throughout the paper we use Cook’s (1977) distance statistics to delete regression outliers and compute tstatistics based on the method of Rogers (1993), which allows heteroskedasticity and any type of correlation for observations of the same firms but assumes independence for observations of different firms. For brevity, we have omitted the coefficients on LINES and STATE. Consistent with our prediction, the coefficient on YR88_92*PRIVC is significantly negative, suggesting that Ccorps shifted income to shareholders after TRA86. The magnitude of the coefficient suggests that the mean taxable income shifted to shareholders by Ccorps during the postTRA86 period amounted to approximately $130,587 per year. [10] This amount represents 29 percent of the mean annual accounting earnings before income shifting.[11] Since the mean ROR for the Ccorps during 198487 was 2.84 percent, the taxinduced percentage change in ROR from the period 198487 to 198892 was 28.17 percent (i.e., the coefficient on YR88_92*PRIVC divided by 2.84 percent). Assuming the marginal tax rates for the C corps and their shareholders were at the top statutory tax rates, the difference in individual and corporate tax rates changed from 4 percent to 6 percent from 198487 to 198892, representing a decline of 250 percent (i.e., 10%/4%). Thus, the implied elasticity of ROR with regard to change in the difference in corporate and individual marginal tax rates is 11.27 percent (i.e., 28.17%/250%). That is, for a ten percent decrease in the spread between individual and corporate tax rates, Ccorps’ ROR declines by 1.13 percent.
The coefficient on PRIVC is significantly positive, suggesting that Ccorps were more profitable than Scorps during 198487. Although this result is consistent with the prediction that Ccorps had the incentive to retain earnings before TRA86, the coefficient on PRIVC should be interpreted with caution because it may also reflect nontax differences between the two types of corporations. The negative coefficient on YR88_92 suggests that corporate profits declined from 8487 to 8892 independent of TRA86. This evidence suggests that the entire decline in corporate profits during this period cannot be solely attributed to TRA86. The positive coefficient on ln(REV) suggests that larger firms were more profitable during the period of our study.
5.2. Income shifting mechanisms
To provide more direct evidence on income shifting from corporate to shareholder tax bases after TRA86, we examine three expense accounts which are believed to be used to shift income by privately held firms: tax deductible compensation for officers and managers, interest expense, and equipment rental expense (Jones 2002). Ideally we should examine only the portion of the above three accounts paid directly to shareholders of the firm, but such information is not available. We believe that using the three total expense accounts will add noise to our tests but should not create any obvious biases. The following regression model is used to examine each of the three expense accounts:
The other variables are defined as before. The coefficient on YR88_92*PRIVC is expected to be positive if Ccorps used the three expense items to shift income to shareholders after TRA86 (H2). If Ccorps retained excessive corporate income before TRA86, the coefficient on PRIVC should be negative. However, due to potential omitted fixed effects, the coefficient on PRIVC should be interpreted with caution. ln (REV), LINES, and STATE serve the same purposes as in regression model (1). RORb controls for any performancerelated effects.
The last three columns of Table 2 report the regression results for the three expense accounts using the full sample. As expected, the coefficient on YR88_92*PRIVC is significantly positive for deductible compensation and rental expense but not for interest expense. The regression coefficients suggest that the Ccorps used deductible compensation and rental expense to shift a mean of $88,756 and
$134,630 taxable income respectively to shareholders after TRA86.[12] Although we do not find evidence that Ccorps used interest expense to shift income, this result may not be generalizable to other settings because debt financing was very low for our sample firms.
5.3. Variation in Ccorps’ tax incentive to use deductible dividends to shift income
The hypotheses H1 and H2 are derived under the assumption that that all Ccorps have the same tax incentive to shift income to shareholders. However, as formula (c) in section 2 indicates, the incentive to shift income from corporate to shareholder tax bases is reduced as the corporate marginal tax rate (τc) declines, ceteris paribus. Furthermore, if a firm does not generate positive current income nor reports a substantial balance in accumulated earnings and profits, regular shareholder dividends are preferred to deductible dividends because the former is likely tax free to shareholders (i.e., a return of capital, which reduces shareholders’ tax basis). To test the implications of the above discussion, we employ the following regression model for the sample of Ccorps. Since Scorps are not subject to corporate tax, we excluded them from this analysis.
+ control variables+ ν;
The dependent variable in model (3) The dependent variable in model (3) is ROR or one of the three expenses as defined in model (2). Control variables refer to those included in model (1) or (2). DUMM is defined as one if a firm’s retained earnings at the beginning of the year scaled by total revenues in the previous year is above the median of the whole Ccorp sample, and zero otherwise.[13] We deflate retained earnings by revenues to control for size effects. We use DUMM as a proxy for Ccorps’ tax incentive to shift income to shareholders using deductible dividends for two reasons. First, firms with low levels of retained earnings are more likely to face low current corporate marginal tax rates because only profitable firms can amass significant retained earnings. Consistent with this assumption, we find that the Spearman rank correlation between DUMM and current year federal income tax payment scaled by total revenues is 15 percent for our sample of Ccorps and highly significant. Second, assuming retained earnings are a reasonable proxy for earnings and profits (a tax term), low or negative retained earnings imply that Ccorp shareholders can receive future payments from the Ccorp tax free (i.e., a return of capital). Accordingly, tax deductible dividends now may be less advantageous if the Ccorp’s current marginal tax rate is low relative to shareholders’ current marginal tax rates. Therefore, we hypothesize that Ccorps with DUMM equal to one are more likely to shift income to shareholders using deductible dividends than Ccorps with low DUMM, ceteris paribus. That is, the coefficient on YR88_92*DUMM should be negative for the regression of ROR and positive for the regressions of OFFSAL, INTEREST, and RENT. We do not make a prediction on DUMM because it also reflects nontax differences for which we cannot control.
Table 3 reports the OLS regression results of model (3) for dependent variables ROR, OFFSAL, INTEREST, and RENT.[14] As expected, the coefficient on YR88_92*DUMM is significantly negative in the regression of ROR. The coefficient on YR88_92*DUMM is significantly positive in the regressions of OFFSAL and RENT, but not significant in the regression of INTEREST. Overall, these results provide further support that the documented results in Table 2 are a result of Ccorps’ taxmotivated income shifting to shareholders.
5.4. Sensitivity Analyses
5.4.1. Firm Size
As indicated in Table 1, Ccorps are significantly larger than Scorps. Thus, a concern one may have is that the coefficient on YR88_92*PRIVC in Table 2 could proxy for the uncontrolled difference in firm size across Ccorps and Scorps. To deal with this concern, Table 4 repeats all the regressions in Table 2 using a sizematched sample. The sizematched sample includes all Scorp observations from Table2 but only those C corps whose median total revenues over 198492 were below the median of the entire C corp sample. The median total revenues for the Ccorps and Scorps in the sizematched sample are very similar ($3.13 million vs. $3.21 million). Except for the regression of INTEREST, the coefficients on YR88_92*PRIVC are still consistent with the predictions and significant. Thus, the results in Table2 are not due to any uncontrolled difference in firm size between C corps and S corps.
5.4.2. Firm Fixed Effects
Regression models (1) and (2) could have omitted unobservable variables correlated with PRIVC and YR88_92*PRIVC. Without controlling for these omitted variables, the regression coefficients from the OLS regressions will be inconsistent. Although we cannot control for the unobservable variables that change over time, any constant unobservable effects can be controlled using the firm fixed effects regression method. As a result, we also reestimated the regressions in Table 2 using fixed effects regression (results not tabulated). Consistent with the results in Table 2, the coefficient on YR88_92*PRIVC remains significantly negative for the ROR regression (coefficient=0.005, onetailed p value=0.036) and significantly positive in the OFFSAL and RENT regressions (the coefficient (onetailed p value) is 0.005 (<0.001) and 0.004 (0.055), respectively). The coefficient on YR88_92*PRIVC is insignificant for the INTEREST regression.
5.4.3. Alternative Specification
Due to missing observations, not all firms have equal number of observations in the pre and post TRA86 period. Thus, it is possible that a subset of firms is over represented in one period but underrepresented in the other period. Therefore a concern one may have is that the coefficient on YR88_92*PRIVC could reflects the changing mix of the sample firms over the two periods. To address this concern, we compute the mean value of each regression variable for the pre and post TRA86 period separately. Then we estimate the regression models (1) and (2) using the mean values of each variable. Because each firm is represented in both periods, the coefficient on YR88_92*PRIVC should not be due to the changing mix of the sample over time. However, a cost of this approach is the reduced power of our test due to the averaging of the variables and the reduced sample size (1,862 vs. 6,839).
The OLS regression results using this approach are reported in Table 5 for the full sample. In spite of the test’s reduced power, the coefficients on YR88_92*PRIVC for the ROR and RENT regressions are in the expected directions and remains marginally significant (p<0.10, one tailed). The coefficient on YR88_92*PRIVC for the OFFSAL regression is still positive but insignificant at the 10% onetailed significance level. Furthermore, for the sizematched sample, the coefficient on YR88_92*PRIVC is as predicted and marginally significant (p<10%, onetailed) for the regressions of ROR, OFFSAL, and RENT and insignificant for the INTEREST regression (results not tabulated). Therefore, we conclude that the pooled regression results in Table 2 are not due to the changing mix of the sample firms over the two periods.
5.5. Constraints of income shifting
As we argued in Section 2.2, privately held Ccorps with one shareholder should have more flexibility than those with multiple shareholders to shift income to their shareholders. We do not have information on the number of shareholders, but we have data on the number of officers for our sample firms over the period 198892. Nagar, Petroni, and Wolfenzon (2001) report that shareholders of privately held firms are commonly managers of the same firms. Thus, we make a simplified assumption that all officers in our sample are shareholders. Because we do not have data on the number of officers for the period 198487, we use the median number of officers for each firm over
198892 as a proxy for the number of shareholders over the period 198492. Due to missing values, this proxy is available for only 6,224 firm years. We create a dummy variable N_SH that equals one if the median number of shareholders over 8492 is more than one, and zero otherwise. The mean (median) N_SH is 0.80 (1) for the Ccorps sample and 0.69 (1) for the Scorps sample.
Table 6 reports the OLS regression results of model (1) and (2) by allowing the coefficients on YR88_92, PRIVC, and YR88_92*PRIVC to vary with N_SH. The coefficients on PRIVC and YR88_92*PRIVC identify the taxinduced income shifting for the Ccorps with only one shareholder, whereas the coefficients on N_SH*PRIVC and N_SH*PRIVC*YR88_92 identify the incremental income shifting for the Ccorps with more than one shareholder. The coefficient on YR88_92*PRIVC is significantly negative in the ROR regression and significantly positive in the regressions of OFFSAL, INTEREST, and RENT, suggesting that Ccorps with one shareholder shifted income to the shareholder using all three accounts. As predicted, the coefficient on N_SH*PRIVC is significantly negative for the ROR regression and significantly positive for the OFFSAL and RENT regressions. As predicted, the coefficient on N_SH*PRIVC*YR88_92 is significantly positive for the ROR regression and significantly negative for the RENT regression. The coefficient on N_SH*PRIVC*YR88_92 is insignificant for the OFFSAL regression. These results are consistent with the hypothesis that privately held Ccorps have more ability to shift income to their shareholders when there are fewer shareholders. We reached similar inference for a sizematched sample of Ccorps and Scorps.
6. Summary and Conclusions
Using a sample of privately held Ccorps and Scorps from the motor carrier industry during 198492, we examine how shareholders avoid double taxation on their investment income from Ccorps in response to TRA86. Prior to TRA86, the top corporate and individual tax rates were 46% and 50%, respectively. TRA86 reduced the top corporate and individual tax rates to 34% and 28%, respectively. In addition, the capital gains tax rate for individuals increased after TRA86. As a result, Ccorps were expected to shift income to shareholders after TRA86. Consistent with the prediction, we find that, relative to a sample of Scorps, which are not subject to corporate taxation, the Ccorps reported lower corporate profits after TRA86. We determine that the mechanisms the Ccorps used to shift income to shareholders are deductible compensation and rental expense (but not interest expense). Furthermore, we find that C corps that faced high tax costs of retaining corporate earnings shifted more income to shareholders after TRA86. Our regression results suggest that the Ccorps shifted a mean of $130,587 taxable income per year (or 29 percent of their mean accounting earnings before income shifting) to shareholders after TRA86. Further analyses suggest that C corps with only one shareholder had a better ability to shift income than those with multiple shareholders.
Our empirical results suggest that taxes can significantly alter the business activities of privately held Ccorps. Because privately held corporations represent a significant sector of the economy, our results have important implications for predicting the tax revenue effect and efficiency loss/gain of future tax changes. In particular, our findings suggest that future researchers need to take into account the income shifting between corporate and personal tax bases in computing the excess burden of future individual or corporate tax changes.
References
American Trucking Association. 1991. Accounting service for motor carriers.
Alexandria, VA: ATA.
Boyer, K. D. 1977. Minimum rate regulation, modal split sensitivities, and the railroad problem. Journac of Pociticac Economy 85 (3): 493512.
Cook, R. D. 1977. Detection of influential observations in linear regression.
Technometrics 19, 1518.
Enis, C. R., and E. A. Morash. 1985. Accounting for public policy actions: The case of motor carrier deregulation. Abacus 21 (March): 6383.
. 1993. Infrastructure taxes, investment policy, and intermodal competition for the transportation industries. Journac of Economics and Business 45 (February): 6989.
Feldstein, M. 1995. The effect of marginal tax rates on taxable income: a panel study of the 1986 tax reform act. Journac of Pociticac Economy 103, 551572.
Gordon, R. H., and J. Slemrod. 2000. Are ‘real’ responses to taxes simply income shifting between corporate and personal tax bases? In J. Slemrod ed., Does Atcas Shrun? The Economic onsequences of Taxinn the Rich. New York: Russel Sage Foundatioin.
Jones, S.M. 2002. Principces of Taxation for business and investment pcanninn. Irwin/McGrawHill, New York, NY.
Ke, Bin. 2001. Taxes as a Determinant of Managerial Compensation in privately held insurance companies. The Accountinn Review 76: 655674.
MacKieMason, J., and R. Gordon. 1997. How much do taxes discourage incorporation?
Journac of Finance 52, 477505.
Morash, E. A., and C. Enis. 1982. The effects of motor carrier deregulation: A stock market perspective. Transportation Research Forum 23 (1): 265271.
Nagar, V., Petroni, K., and D. Wolfenzon. 2001. Preventing squeeze outs in close corporations. University of Michigan working paper.
Omer, T. C., G. A. Plesko, and M. K. Shelley. 2000. The influence of tax costs on organizational choice in the natural resource industry. Journac of American Taxation Association 22, 3855.
Plesko, G., 1994, Corporate taxation and the financial characteristics of firms, Public Finance Quartercy, Vol 22: 311334.
Rogers, W.H. 1993. sg17: Regression standard errors in clustered samples. Stata Technical Bulletin 13, 1923.
Schipper, K., R. Thompson, and R. L. Weil. 1987. Disentangling interrelated effects of regulatory changes on shareholder wealth: The case of motor carrier deregulation. Journac of Law and Economics 30, 67100.
Scholes, M., M. A. Wolfson, M. Erickson, E. L. Maydew, and T. Shevlin. 2002. Taxes and business strateny1 a pcanninn approach. 2nd edition. PrenticeHall Inc. Englewook Cliffs, New Jersey.
Shackelford, D., and T. Shevlin. 2001. Empirical tax research in accounting. Journac of Accountinn and Economics 31, 321388.
Stiglitz, J. E. 1988. Economics of the public sector. 2nd edition. New York: W. W. Norton.
Wilkie, T. J., J. C. Young, and S. E. Nutter. 1996. Corporate business activity before and after the tax reform act of 1986. Statistics of Income Buccetin 16, 3245.
Wilson, G. W. 1980. Economic Anacysis of Intercity Freinht Transportation. Bloomington, IN: Indiana University Press.
Table 1. Summary Statistics for the samples of Ccorps and Scorps
The samples of Ccorps and Scorps have 5,811 and 1,028 firm years, respectively, over the period 1984 1992. ASSETS is total assets of the firm. REV is total sales revenues of the firm. GEN is one if a firm is a general freight carrier, and zero otherwise. ROR is total accounting profit before taxes divided by total revenues. OFFSAL, INTEREST, and RENT are deductible compensation to officers and managers, interest expense, and equipment rent expense, respectively, all expressed as a percentage of total revenues. The values in the cells of columns two and three represent the mean, median, and standard deviation, respectively.
Table 2. Full sample OLS regression results on the effect of TRA86 on income shifting from corporate to shareholder tax bases
The table reports the regression results of models 1 and 2 above using the full samples of 5,811 Ccorp years and 1,028 Scorp years over the period 19841992. The dependent variable for model 2 is OFFSAL, INTEREST, or RENT. ROR is total accounting profit before taxes divided by total revenues. OFFSAL, INTEREST, and RENT are deductible compensation to officers and managers, interest expense, and equipment rent expense, respectively, all expressed as a percentage of total revenues. YR88_92 is one for the years 198892 and zero otherwise. PRIVC is one for Ccorps and zero otherwise. YR88_92*PRIVC is the interaction of YR88_92 and PRIVC. REV is total revenues of the firm. RORb is ROR before the dependent variable. STATE are dummies denoting each firm’s state of domicile. LINES represents dummies denoting six of the seven types of motor carriers. The coefficients on variables LINES and STATE are omitted for brevity. The regression of INTEREST excluded 58 unique firms (390 firm years) that reported no interest expenses over the entire period 198492. Outliers were deleted using Cook’s (1977) method. The standard errors in parentheses are adjusted for heteroskedasticity and serial autocorrelation using STATA’s cluster command (Rogers 1993). *, ** denote significant levels of 5% and 1%, respectively, onetailed if there is a prediction and twotailed otherwise.
Model1 
Model2 

Hypothesis ROR 
Hypothesis 
OFFSAL 
INTEREST 
RENT 

YR88_92 
0.007 
0.001 
0.001 
0.006 

(0.003)* 
(0.002) 
(0.001) 
(0.003)* 

PRIVC 
0.016 (0.004)** 
0.004 (0.002) 
0.001 (0.001) 
0.016 (0.004)** 

YR88_92*PRIVC 
– 0.008 
0.005 
0.000 
0.008 

(0.003)** 
(0.002)** 
(0.001) 
(0.003)** 

LN(REV) 
0.004 (0.001)** 
0.003 (0.001)** 
0.001 (0.000)** 
0.004 (0.001)** 

RORb 
0.245 (0.014)** 
0.038 (0.006)** 
0.970 (0.005)** 

CONSTANT 
0.041 (0.017)* 
0.103 (0.012)** 
0.008 (0.005) 
0.039 (0.017)* 

N 
6658 
6645 
6284 
6661 

R–SQUARED 
0.071 
0.353 
0.102 
0.964 
Table 3. Ccorps’ differential tax incentives to use deductible dividends to shift income to shareholders In the pre and postTRA86 periods
The table reports the regression results of the above model for the sample of 5,452 Ccorp years over the period 198492. 359 firm years were lost due to missing values on retained earnings. Thedependent variable is ROR, OFFSAL, INTEREST, or RENT. The control variables refer to those included in model (1) for the regression of ROR and those included in model (2) for the regressions of OFFSAL,INTEREST, and RENT. ROR is total accounting profit before taxes divided by total revenues. OFFSAL, INTEREST, and RENT are deductible compensation to officers and managers, interest expense, andequipment rent expense, respectively, all expressed as a percentage of total revenues. YR88_92 is one for the years 1988 92 and zero otherwise. DUMM is one if a firm’s retained earnings at the beginningof the year scaled by total revenues in the previous year is more than the median of the whole sample of Ccorps, and zero otherwise. YR88_92*DUMM is the interaction of YR88_92 and DUMM. REV istotal revenues of the firm. RORb is ROR before the dependent variable. STATE are dummies denoting each firm’s state of domicile. LINES represents dummies denoting six of the seven types of motorcarriers. The coefficients on variables LINES and STATE are omitted for brevity. The regression of INTEREST excluded 42 unique firms (256 firm years) that reported no interest expenses over the entireperiod 198492. Outliers were deleted using Cook’s (1977) method. The standard errors in parentheses are adjusted for heteroskedasticity and serial autocorrelation using STATA’s cluster command (Rogers1993). *, ** denote significant levels of 5% and 1%, respectively, onetailed if there is a prediction and twotailed otherwise.
Dependent Variables
Hypothesis 
ROR 
Hypothesis 
OFFSAL 
INTEREST 
RENT 

YR88_92 
–0.009 
0.004 
0.000 
0.008 

(0.002)** 
(0.001)** 
(0.001) 
(0.002)** 

DUMM 
0.025 
0.003 
–0.003 
–0.027 

(0.002)** 
(0.002) 
(0.001)** 
(0.002)** 

YR88_92*DUMM 
– 
–0.012 (0.003)** 
0.003 (0.002)* 
–0.001 (0.001) 
0.013 (0.003)** 

LN(REV) 
0.005 (0.001)** 
–0.002 (0.001)** 
0.001 (0.000)** 
–0.005 (0.001)** 

RORb 
0.260 (0.017)** 
0.044 (0.007)** 
0.972 (0.004)** 

CONSTANT 
–0.061 (0.016)** 
0.068 (0.011)** 
–0.008 (0.005) 
0.083 (0.018)** 

N 
5,284 
5,317 
5,070 
5,289 

R–SQUARED 
0.114 
0.386 
0.144 
0.971 
Table 4. Sizematched sample OLS regression results on the effect of TRA86 on income shifting from corporate to shareholder tax bases
The table reports the regression results of models 1 and 2 above using sizematched samples of 2,906 C corp years and 1,028 Scorp years over the period 198492. The sizematched sample includes all Scorps from Table 2 and the Ccorps whose median total revenues over 198492 were less than the median of the entire Ccorp sample. The dependent variable for model 2 is OFFSAL, INTEREST, or RENT. ROR is total accounting profit before taxes divided by total revenues. OFFSAL, INTEREST, and RENT are deductible compensation to officers and managers, interest expense, and equipment rent expense,respectively, all expressed as a percentage of total revenues. YR88_92 is one for the years 198892 and zero otherwise. PRIVC is one for Ccorps and zero otherwise. YR88_92*PRIVC is the interaction ofYR88_92 and PRIVC. REV is total revenues of the firm. RORb is ROR before the dependent variable. STATE are dummies denoting each firm’s state of domicile. LINES represents dummies denoting six of the seven types of motor carriers. The coefficients on variables LINES and STATE are omitted for brevity. The regression of INTEREST excluded 51 unique firms (347 firm years) that reported no interest expenses over the entire period 198492. Outliers were deleted using Cook’s (1977) method. The standard errors in parentheses are adjusted for heteroskedasticity and serial autocorrelation using STATA’s cluster command (Rogers 1993).
*, ** denote significant levels of 5% and 1%, respectively, onetailed if there is a prediction and twotailed otherwise.
Model 1 Model 2
Hypothesis 
ROR 
Hypothesis 
OFFSAL 
INTEREST 
RENT 

YR88_92 
–0.007 
0.002 
–0.001 
0.009 

(0.003)* 
(0.002) 
(0.001) 
(0.003)** 

PRIVC 
0.022 (0.005)** 
–0.005 (0.003)* 
–0.001 (0.001) 
–0.020 (0.004)** 

YR88_92*PRIVC 
– 
–0.010 (0.004)** 
0.005 (0.002)* 
–0.001 (0.001) 
0.008 (0.004)* 

LN(REV) 
0.009 
–0.010 
0.000 
–0.006 

(0.002)** 
(0.002)** 
(0.001) 
(0.003)* 

RORb 
0.216 (0.019)** 
0.030 (0.006)** 
0.943 (0.008)** 

CONSTANT 
–0.121 
0.214 
0.005 
0.086 

(0.039)** 
(0.028)** 
(0.010) 
(0.040)* 

N 
3831 
3836 
3511 
3839 

R–SQUARED 
0.098 
0.336 
0.099 
0.952 
Table 5. OLS regression results on the effect of TRA86 on income shifting from corporate to shareholder tax basesusing the mean value of each variable in the pre and postTRA86 periods
The table reports the regression results of models 1 and 2 above using the mean value of each variable in the pre and postTRA86 periods. The sample includes 1,560 Ccorp years and 302 Scorp years.The dependent variable for model 2 is OFFSAL, INTEREST, or RENT. ROR is total accounting profit before taxes divided by total revenues. OFFSAL, INTEREST, and RENT are deductible compensation toofficers and managers, interest expense, and equipment rent expense, respectively, all expressed as a percentage of total revenues. YR88_92 is one for the years 198892 and zero otherwise. PRIVC is onefor Ccorps and zero otherwise. YR88_92*PRIVC is the interaction of YR88_92 and PRIVC. REV is total revenues of the firm. RORb is ROR before the dependent variable. STATE are dummies denoting eachfirm’s state of domicile. LINES represents dummies denoting six of the seven types of motor carriers. The coefficients on variables LINES and STATE are omitted for brevity. The regression of INTERESTexcluded 58 unique firms (116 firm years) that reported no interest expenses over the entire period 198492. No outliers were deleted using the Cook’s (1977) method. The standard errors in parenthesesare adjusted for heteroskedasticity and serial autocorrelation using STATA’s cluster command (Rogers 1993). *, ** denote significant levels of 5% and 1%, respectively, onetailed if there is a prediction andtwotailed otherwise.
Model 1 Model 2
Hypothesis 
ROR 
Hypothesis 
OFFSAL 
INTEREST 
RENT 

YR88_92 
0.005 
0.004 
0.003 
–0.005 

(0.013) 
(0.004) 
(0.002) 
(0.012) 

PRIVC 
0.020 (0.005)** 
–0.004 (0.003) 
–0.002 (0.001) 
–0.018 (0.005)** 

YR88_92*PRIVC 
– 
–0.019 
0.005 
–0.003 
0.016 

(0.014) 
(0.005) 
(0.002) 
(0.012) 

LN(REV) 
–0.002 
–0.005 
0.001 
0.005 

(0.004) 
(0.002)* 
(0.001) 
(0.004) 

RORb 
0.279 (0.101)** 
0.050 (0.025)* 
0.912 (0.034)** 

CONSTANT 
0.077 
0.119 
–0.004 
–0.095 

(0.070) 
(0.024)** 
(0.011) 
(0.067) 

N 
1862 
1862 
1746 
1862 

R–SQUARED 
0.056 
0.359 
0.087 
0.916 
Table 6. OLS regression results on the effect of TRA86 on income shifting from corporate to shareholder taxbases by the number of shareholders
The table reports the regression results of models 1 and 2 above using the full samples of 5,292 Ccorp years and 932 Scorp years over the period 19841992, allowing the key variables to vary with thenumber of shareholders. 615 firm years were lost due to missing values on the number of shareholders. The dependent variable for model 2 is OFFSAL, INTEREST, or RENT. ROR is total accounting profitbefore taxes divided by total revenues. OFFSAL, INTEREST, and RENT are deductible compensation to officers and managers, interest expense, and equipment rent expense, respectively, all expressed as apercentage of total revenues. YR88_92 is one for the years 198892 and zero otherwise. PRIVC is one for Ccorps and zero otherwise. YR88_92*PRIVC is the interaction of YR88_92 and PRIVC. REV is totalrevenues of the firm. RORb is ROR before the dependent variable. N_SH is one if the median number of officers over
198492 (a proxy for the number of shareholders) is more than one, and zero otherwise. N_SH*PRIVC, N_SH*YR88_92, and N_SH*PRIVC*YR88_92 are interaction terms. STATE are dummies denoting eachfirm’s state of domicile. LINES represents dummies denoting six of the seven types of motor carriers. The coefficients on variables LINES and STATE are omitted for brevity. See Tables 1 and 2 for othervariable definitions. The coefficients on variables LINES and STATE are omitted for brevity. The regression of INTEREST excluded 58 unique firms (390 firm years) that reported no interest expenses over theentire period 198492. Outliers were deleted using Cook’s (1977) method. The standard errors in parentheses are adjusted for heteroskedasticity and serial autocorrelation using STATA’s cluster command(Rogers 1993).
*, ** denote significant levels of 5% and 1%, respectively, onetailed if there is a prediction and twotailed otherwise.
Model 1 Model 2
Hypothesis 
ROR 
hypothesis 
OFFSAL 
INTEREST 
RENT 

YR88_92 
0.011 
–0.002 
–0.004 
–0.009 

(0.008) 
(0.003) 
(0.001)** 
(0.008) 

PRIVC 
0.027 
–0.012 
0.000 
–0.026 

(0.007)** 
(0.004)** 
(0.002) 
(0.007)** 

YR88_92*PRIVC 
– 
–0.026 (0.009)** 
0.008 (0.004)* 
0.003 (0.002)* 
0.025 (0.009)** 

Ln(REV) 
0.005 (0.001)** 
–0.005 (0.001)** 
0.001 (0.000)** 
–0.005 (0.001)** 

N_SH 
0.011 
0.002 
0.001 
–0.009 

(0.007) 
(0.005) 
(0.002) 
(0.007) 

N_SH*PRIVC 
– 
–0.016 (0.008)* 
0.011 (0.005)* 
–0.001 (0.002) 
0.013 (0.008)* 

N_SH*PRIVC*YR88_92 
0.020 (0.010)* 
– 
–0.003 (0.005) 
–0.002 (0.002) 
–0.017 (0.009)* 

N_SH*YR88_92 
–0.018 
0.004 
0.003 
0.015 

(0.009)* 
(0.004) 
(0.002) 
(0.009) 
[1 ]See Gordon and Slemrod (2000) for an excellent discussion of the policy implications of income shifting
between corporate and personal tax bases.
[2] The top individual statutory tax rate increased to 31% after 1990, but it was still lower than the top
corporate tax rate. None of our empirical results was affected if the sample period was limited to 198490.
[3 ]An effective tax strategy publicly traded firms could use to minimize the joint tax liabilities of the firm and its employees is to issue employees stock options (e.g. Scholes et al. 2002). Because shares of privately held firms are not frequently traded, this strategy is rarely used by privately held firms.
[4] In 1994 the annual reporting requirement changed to mandate that all Class I carriers (revenues of at least of $10 million) file annual reports on Form M1 and Class II carriers (revenues of at least $3 million) file annual reports on Form M2 (a scaled down version of Form M1). Class III carriers (revenues under $3 million) were exempt from filing. The $10 and $3 million thresholds are for base year 1994 and are inflation adjusted annually. Motor Carrier Annual Reports were filed with the Intestate Commerce Commission (ICC) up until the ICC Termination Act of 1995 (49 U.S.C. 11145). This Act transferred the collection of these annual reports to the Bureau of Transportation Statistics (BTS) of the Department of Transportation. In 1998, Form M, a much simpler annual report, replaced Forms M1 and M2. Vistronix, a private contractor for BTS, maintains Motor Carrier Annual Reports from 1996 to the present.
[5] Six states (CT, LA, MI, NJ, NC, and TN) did not recognize the Scorp form for state income tax purposes
during our sample period. Omer et al. (2000) finds that firms in these states are less likely to select the Scorp form. Thus, as a sensitivity check, we also eliminated the Scorps domiciled in the six states from our regression analysis and obtained similar conclusions.
[6] §1374 requires Scorps that were converted from Ccorps before 1987 to pay corporate tax on the net recognized builtin gains inherited from the Ccorps during a 10year period, starting from the first year of the S election. Therefore, if a C corp made an S election and recognized builtin gains during the period
198892, it would be misclassified as a Ccorp instead of as a CtoS conversion using our entity classification criteria. Such misclassification would reduce the power of but not bias our empirical tests.
[7] If our Ccorp sample contains only those firms that anticipated low profits post TRA86 because those who anticipated high profits post TRA86 converted to Scorp, we may also observe declines in corporate profitability after TRA86, as predicted by H1. To check whether this potential selfselection bias explains our finding, we compared ROR for our Ccorp sample and the excluded CtoS conversion sample. We find that ROR was significantly higher for our Ccorp sample in both time periods. So selfselection is not a
problem for our Ccorp sample.
[8] The annual reports of our sample firms could be missing in years 199697 for several reasons. First, after 1993, only motor carriers with operating revenues of at least $3 million were required to file annual reports, while the threshold for the years before 1994 was $1 million (see footnote 4). Second, the compliance level for filing annual reports declined after 1993. Third, some trucking firms were liquidated, merged and acquired after our sample period.
[9 ]We reached similar conclusions if we allowed the coefficient on PRIVC to vary with each year and test our H1 using the average coefficient on PRIVC for the two periods.
[10] This number was calculated as the product of the coefficient on YR88_92*PRIVC and mean total revenues for the Ccorps during 198892.
[11] As a sensitivity check, we also included lagged ROR and its interaction with PRIVC in regression model
(1). An advantage of including lagged ROR is that any effects unrelated to taxes will be controlled effectively, but a disadvantage is that it may eliminate the predicted tax effect if Ccorps responded to TRA86 quickly. The regression coefficient on YR88_92*PRIVC after including lagged ROR and its interaction with PRIVC is 0.006 and highly significant (onetailed p value=0.009).
[12 ]The combined magnitude of income shifting here is higher than the magnitude reported in section 5.1. This is likely due to using different regression models.
[13] Choosing a cutoff ranging from the 40th percentile to 60th percentile of the retained earnings over total revenues yields similar results.
[14 ]We also estimated model (3) using the smaller sample of 277 Ccorps identified from the hard copies of motor carrier annual reports and obtained similar conclusions.
Previously published by the Pennsylvania State University Smeal College of Business Administration, July, 2002
Written by: Charles R. Enis Bin Ke*