A review of the empirical literature

Fig. 1. Payroll tax rates, 1950–1995. tax code-along with substantial variation at the state and individual levels-provide ample variation that can be used in analyzing individual sensitivity to taxes among potential entrepreneurs.

3. A review of the empirical literature

The few empirical studies that have examined taxes and self-employment have generally considered the effects of the aggregate tax climate using time-series data on self-employment rates. A few have analyzed individual-level cross-sectional data, but most have used aggregated tax variables to avoid the potential endogene- ity of taxes on the individual level; only two of these used individual-level tax information, and neither one tested or corrected for the endogeneity of tax rates. Ž . Beginning with the time-series studies, Long 1982a investigated the effects of income tax rates on the SMSA-level ratio of self-employment to total employment. Using 1970 Census data, he found that a 10 increase in the average marginal income tax rate in an SMSA increased the self-employment rate in that SMSA by 6.4. Further, a US300 increase in average income tax liability increased the self-employment rate by 1. Ž . Ž . Blau 1987 followed Long’s 1982a method and estimated aggregate time- series regressions using US data for 1948 through 1982. His tax variables consisted of two marginal tax rates, however, for incomes of US7000 and US17,000, respectively. While his findings at the higher marginal tax rate supported Long’s, Blau found that increases in lower-bracket marginal tax rates actually reduced the self-employment rate. This empirical puzzle was not ex- plained. Ž . Ž . Parker 1996 more closely resembled Long 1982a , but used a 1959 to 1991 time series of United Kingdom data. Again, higher marginal tax rates were found Ž . to increase self-employment rates. He followed Blau 1987 in using two marginal tax rates associated with two different levels of income, but failed to support Blau’s findings at the lower rate. Ž . Robson and Wren 1998 provided a theoretical model that incorporated both average and marginal tax rates. Their model predicted, and time-series regressions confirmed, that higher average tax rates reduced self-employment rates while higher marginal tax rates increased self-employment. This divergence was at- tributed to the assumptions that self-employment earnings more closely resemble Ž the value of the individual’s marginal product of labor i.e., a higher average tax . rate in self-employment reduces the net reward from working , and that higher marginal tax rates increase the payoff to evading or avoiding taxes by under-re- porting income or increasing business-related deductions. This time-series evidence on taxes and self-employment supports the popular claim that individuals become self-employed in order to avoid paying higher wage-and-salary taxes. However, the most recent time-series evidence, presented Ž . by Fairlie and Meyer 1998 , finds no significant effect of taxes on self-employ- ment rates. Using aggregated individual data from seven decennial Censuses of Ž . Population for 1910 and 1940–1990 , they found that the overall relationship between tax rates and self-employment is, at best, weak during this period. Their paper is an exception, however, as no multivariate analysis was performed to test this conclusion. Cross-sectional evidence generally supports the earlier time-series results. Long Ž . 1982b used disaggregated data from the 1970 Census for males between the ages of 25 and 64. Linear probability models of whether or not one is self-employed regressed on expected income tax liability under wage-and-salary employment indicated economically and statistically significant tax effects. Echoing his aggre- gate findings, an increase in expected wage-and-salary tax liability of 10 was found to increase the probability of being self-employed by 7.4. Ž . Moore 1983 expanded on Long’s research, but focused on the payroll tax. Ž . Indeed, Moore’s was the first and only study to examine both individual income and the payroll taxes using microdata. He used 1978 CPS data for males between 25 and 65 working full time to estimate linear probability models and logits of self-employment status. The effect of the expected wage-and-salary income tax Ž . variable was significant, but much smaller than those estimated by Long 1982a,b . Moore found much larger effects from the payroll tax; a 10 increase in expected wage-and-salary payroll tax liability caused a 5 to 8 increase in the probability of being self-employed. Ž . Schuetze 1998 provided the most recent look at the effects of taxes on self-employment, using aggregate Atax climateB variables in a cross-sectional comparison of the United States and Canada. Using CPS data for 1983 through 1994, he found that a 10 increase in the US average marginal tax rate in one year caused a 2.1 increase in the probability of being self-employed in the following year. 2 While these studies conclude almost universally that higher overall marginal tax rates increase self-employment rates, they do not actually estimate individual-level sensitivity to the relative tax treatment of the self-employed. Two recent papers by Ž . Carroll et al. 1995, 1997 provided evidence of the importance of investigating the tax response using individual-level data. Their investigation of the impact of taxes on decisions by existing entrepreneurs to hire additional workers and make capital investments suggests that the self-employed are indeed cognizant of their own personal tax situations. Specifically, marginal tax rate increases were found to reduce both mean investment expenditures and the probability of hiring employ- ees. While individual-specific taxes have been found to have important effects on the actions of those already in self-employment, little is known about their effects on entry into self-employment. It is the intent of this paper to examine whether or not indiÕidual-leÕel income and payroll taxes affect indiÕidual decisions to become self-employed. More precisely, I wish to estimate the existence of tax effects by examining the expected post-transition tax situations of potential entrepreneurs during the 1980s. Also, in order to correctly gauge the level of tax sensitivity, I focus on individual-specific tax differentials by comparing each person’s tax situation in wage-and-salary and self-employment. 2 Even stronger effects were found in the Canadian data.

4. Data and empirical specification