Tax progressivity and self-employment: evidence from Canadian provinces
- First Online:
- Cite this article as:
- Ferede, E. Small Bus Econ (2013) 40: 141. doi:10.1007/s11187-011-9350-7
- 322 Views
We examine the effects of personal income tax progressivity—in the sense of rising marginal income tax rate—on self-employment. The impacts of income tax progressivity on self-employment depend on the relative effects of taxing success and the presence of tax evasion opportunities. Empirical estimates using Canadian provincial data for the period 1979–2006 indicate that there is a negative association between income tax progressivity and self-employment. This suggests that the adverse impact of income tax on entrepreneurial risk-taking outweighs the tax evasion opportunities for the self-employed. An important implication of our results is that a reduction in income tax progressivity encourages self-employment. The empirical estimates are robust to the various sensitivity checks.
KeywordsSelf-employmentTax progressivityOccupational choice
There is a growing interest in the effects of tax policies on self-employment in both academic and political circles. The self-employed are generally viewed as a key source of employment and growth in an economy (OECD 2001). Consequently, governments often design tax policies in an attempt to promote entrepreneurial activities by the self-employed. What is the effect of personal income tax progressivity on self-employment? Can governments encourage self-employment through income tax cuts?
In most previous studies self-employment is used as a common proxy for entrepreneurship. As a result a number of studies that examine entrepreneurship focus on self-employment and treat self-employment as synonymous with entrepreneurship. However, only a few studies investigate the effects of personal income tax progressivity on entrepreneurship. Gentry and Hubbard (2000) analyze the impact of tax progressivity on entrepreneurial entry. They show that under a progressive tax system, entrepreneurs pay substantial taxes when they are successful, but when entrepreneurs incur losses, their tax savings are minimal. As a result they argue that a progressive tax schedule can be viewed as a tax on “success” and discourages entrepreneurial entry if entrepreneurs are risk-neutral. They also find supporting empirical evidence using US household panel data. Keuschnigg and Nielsen (2004) also explore the effects of a progressive income tax system on entrepreneurial risk taking in the presence of moral hazard. Their analysis indicates that since the expected income from risky entrepreneurial activities is higher than wage income, progressive taxation redistributes from entrepreneurs to workers, imposes a net tax on the risky activity, and as a result discourages entrepreneurial entry. Cullen and Gordon (2007), on the other hand, find a result contrary to conventional wisdom. They argue that entrepreneurs can deduct business losses under the personal income tax system and therefore the government shares the risk taking of entrepreneurs. They show that a reduction in the personal tax rate reduces the risk sharing of governments (as this reduces the tax saving that can be obtained by deducting business losses) and discourages entrepreneurial risk taking.1 The above studies however ignore the tax evasion opportunity by the self-employed.
A number of empirical studies have examined the effects of taxes on various proxies for self-employment. Like the theoretical analyses, the results of previous empirical studies are also mixed.2 Robson and Wren (1999) for OECD, Folster (2002) for Sweden, Harvey (2004) for the United States, Torrini (2005) for OECD countries, and Mooij and Nicodeme (2008) for European countries find that taxes affect self-employment or entrepreneurship negatively. Bruce and Mohsin (2006), on the other hand, find a positive effect using US data. Although there is a vast literature on the relationship between taxation and self-employment, there are very few studies that focus on Canada. For example, Schuetze (2002) explores the extent of tax avoidance among the self employed. While Schuetze (2000) focuses on self-employment of males, Kuhn and Schuetze (2001) examine self-employment in men and women separately. Kamhi and Leung (2005) also investigate the trends in self-employment and the various factors that influence the total self-employment rate in the country. However, none of these studies provides a detailed investigation of the effects of personal income tax progressivity on self-employment or entrepreneurial activity.
The principal objective of this study is to examine the effects of personal income tax progressivity on self-employment in Canada. The study also fills the gap in the literature by providing empirical evidence as to how tax progressivity influences self-employment. Tax progressivity has two opposite effects on self-employment. First, for the self-employed, income taxation is based to a large extent on voluntary compliance to the tax laws. As a result the self-employed have a great opportunity to avoid and evade taxes relative to wage and salary employees.3 Thus an increase in income tax progressivity, in the sense of rising marginal income tax rate, symbolizes a greater opportunity for tax avoidance and evasion and this encourages self-employment. Second, an increase in income tax progressivity reduces the return of successful entrepreneurs and dents their risk-taking potentials. This ultimately discourages entrepreneurial activities. Thus the impact of income tax progressivity on self-employment depends on the relative effects of the adverse impact of progressivity on success and the tax evasion opportunity for the self-employed.
We use aggregate panel data from ten Canadian provinces over the period 1979–2006. Our empirical findings suggest that personal income tax progressivity discourages self-employment implying that the adverse impact of higher tax progressivity on success outweighs the tax evasion opportunity that may be available to the self-employed. The empirical results are also quite robust to various sensitivity tests. The policy implication of this study is that a reduction in the marginal personal income tax rate that lowers the degree of tax progressivity encourages self-employment.
The remaining part of this paper is organized as follows. In Sect. 2 we discuss the possible theoretical relationship between self-employment and income tax progressivity. The empirical analysis is presented in Sect. 3. In Sect. 4, we conduct robustness checks to our empirical results. Section 5 concludes.
2 Analytical framework
In this section we briefly discuss the theoretical relationship between income tax progressivity and self-employment.4 Suppose there are two groups of individuals: the self-employed and workers. Empirical stylized facts show that the majority of the self-employed do not hire workers (i.e. own account self-employment).5 Thus following Parker (1999), we assume that the self-employed (entrepreneurs) work for themselves but they do not hire workers. Entrepreneurial activity is a risky occupation and, as in Gentry and Hubbard (2000), we assume that it can yield a higher return if successful or zero otherwise. Working for others, on the other hand, yields a riskless income (which is the wage rate). In order to focus on tax progressivity—rather than risk aversion—as the determinant of occupational choice we assume that individuals are risk neutral. Thus individuals choose between a safe job and a risky one based on the after-tax return from the two occupations.
In Canada, as in many other countries, income from self-employment is taxed as any ordinary personal income at the applicable personal income tax rates after deductions for business expenses. As extensively discussed in the literature, there is a greater tax evasion opportunity for the self-employed than for workers. We assume that it is difficult to evade and avoid taxes in the safe job but there is a tax evasion and avoidance opportunity in the self-employed sector. When the self-employed evade taxes, there is a probability of detection through auditing. If caught through auditing, tax evaders pay the full tax liability and a certain percent of the amount of tax evaded as penalty. Thus in a progressive income tax system, the benefit of evading taxes increases with the tax progressivity. In such theoretical set up, equilibrium in occupational choice requires that the expected after-tax income from self-employment must be equal to the after-tax return from the safe job (employment); see Evans and Jovanovic (1989). That is, individuals’ occupational choice depends on the after-tax returns from the two jobs. Such analytically tractable simple model highlights the relationship between income tax progressivity and self-employment rate.
Our model shows that in the absence of tax evasion opportunity, an increase in tax progressivity discourages self-employment. That is the greater the tax progressivity, the lower the rate of self-employment. Thus, as Gentry and Hubbard (2000) explain, the progressive income tax has a detrimental effect on entrepreneurship and can be viewed as a “success tax”. The reason for this is straightforward. When the tax structure becomes more progressive the government claims a higher share of the return from successful entrepreneurs (which assume the risk of earning a very low income if not successful). This discourages the risk taking behaviour of the self-employed as the government shares the reward but not the loss from the risky activity.
In the presence of tax evasion opportunity, however, the relationship between tax progressivity and self-employment becomes more complicated. This is because it depends on the relative strength of the negative effect of the “success tax” and the positive effect of the tax evasion opportunity. Other things remaining constant, if the tax evasion opportunity from being a self-employed is stronger than the adverse impact of taxing success at a progressive rate, an increase in tax progressivity raises the self-employment rate. On the other hand, if the tax evasion opportunity is lower (say because of a high level of tax enforcement, penalty, higher frequency of auditing, etc.) then higher tax progressivity acts as a tax on successful entrepreneurs and discourages self-employment. Thus, the impact of tax progressivity on self-employment is generally an empirical issue. Consequently, in the following section we explore this issue empirically using aggregate panel data from ten Canadian provinces over the period 1979–2006.
3 Empirical analysis
3.1 The data
We use annual aggregate data from ten Canadian provinces over the period 1979–2006. The main sources of our data set are Statistics Canada database (CANSIM) and Finances of the Nation (formerly National Finances) published by the Canadian Tax Foundation. The data on marginal personal income tax rates and corporate income tax rate come from various issues of Finances of the Nation. Data on self-employment, total employment, unemployment, implicit tax rate and relative self-employment income are from CANSIM. We provide a brief description of the data and definitions of the variables used in the empirical analysis in Appendix Table 5.
Summary statistics, 1979–2006
Coefficient of residual income progression (RIP)
Average income tax rate
Marginal income tax rate
Corporate tax rate (small businesses)
Real minimum wage rate ($)
Correlations between main variables
Real minimum wage rate
Corporate income tax rate
Average income tax rate
Real minimum wage rate
Corporate income tax rate
Average income tax rate
Real minimum wage rate
Corporate income tax rate
Average income tax rate
We measure the degree of personal income tax progression with the coefficient of residual income progression (RIP) from Musgrave and Musgrave (1989) and Dahlby (2008). RIP depends on the marginal and average income tax rates. More specifically, RIP is measured as the ratio of one minus the marginal income tax rate to one minus the average income tax rate. Since marginal and average income tax rates vary across income tax brackets, it is difficult to come up with a single RIP value that is suitable for an empirical analysis. We would ideally like to construct RIP for each income tax bracket using the statutory marginal and average tax rates corresponding to the various tax brackets. However, there are no data on provincial average income tax rates corresponding to the various tax brackets. Thus, in order to circumvent this problem, we rely on the statutory marginal income tax rate (federal and provincial) that is applicable to the average income of single persons (or unattached individuals). The average income is the same for all provinces. Thus our RIP variable captures only statutory tax rather than income differences across provinces. See for example Van Ewijk and Tang (2007) for a similar approach. We also use the implicit average income tax rate for single persons as a measure of an overall average tax rate. Then we use the two variables to generate RIP for all provinces. In our sensitivity analysis, we also experiment with other alternative measures of tax progressivity.
3.2 Empirical specification
We are interested in the coefficient α1. Our theoretical framework suggests that the sign of α1 may be positive or negative, depending on the relative strength of the tax evasion opportunity in self-employment and the adverse impact of greater tax progressivity on risk-taking. If the former dominates, we expect α1 to be negative. On the other hand, the coefficient will be positive if the latter outweighs.
There are factors other than personal income tax progressivity that may influence the self-employment rate. Our set of control variables includes neighbouring provinces’ RIP, the average personal income tax rate, the combined (provincial and federal) corporate income tax rate that is applicable to small businesses, the real minimum wage rate, the relative income from self-employment, and the unemployment rate. Most of these control variables are commonly used in the empirical literature; see for example Blau (1987), Kamhi and Leung (2005), Bruce and Deskins (2006) and Bruce and Mohsin (2006). The unemployment rate is included to capture business cycle fluctuations. As in Blau (1987), we also include the real minimum wage rate to capture possible wage rigidity.
Labour is a mobile factor. Thus, other things remaining constant, the self-employed can move from one province to another if there is a significant variation in the personal income tax system. In fact, the literature on horizontal tax competition suggests that due to the mobile nature of labour, tax progressivity in other provinces can affect self-employment rate. That is when neighbouring provinces raise their income tax progressivity, the self-employed may move from the higher–tax (and more progressive tax) jurisdiction to a lower progressive tax jurisdiction. Thus to account for this potential mobility of the self-employed across provinces, we include the weighted average—weighted by GDP—tax progressivity (RIP) of neighbouring provinces. If the self-employed move from a higher to lower progressive tax province, we expect the coefficient of neighbours’ RIP to be negative.
Some of the self-employed are incorporated businesses. For this kind of self-employment the corporate income tax rate may be relevant. Thus, as in Mooij and Nicodeme (2008), we control for the corporate tax rate that is applicable to small businesses. Our theoretical model indicates that the relative income between self-employment and wage employment is relevant. Thus we also control for the relative income. We expect that a higher relative income from self-employment encourages entrepreneurial activity. Note also that while the self-employment rate depends on the relative income, our simple model also shows that the relative income in turn depends on the self-employment rate. We will come back to this issue later in the empirical analysis.
3.3 Regression results
Self-employment regression, 1979–2006
Average personal income tax rate
Corporate income tax rate
Real minimum wage rate
Overidentification test (p-value)
We begin with OLS estimation of the own-account self-employment rate on just the provincial RIP and the neighbours’ RIP in column 1. We control for time and provincial fixed-effects in all our regressions and the reported t-statistics are based on robust standard errors.7 In this regression, the coefficient of RIP is positive and statistically significant. This indicates that greater tax progressivity (lower RIP) discourages self-employment rate. That is as the tax rate becomes more progressive, the adverse impact of income tax on entrepreneurial risk taking outweighs the tax evasion opportunity for the self-employed.
The estimated quantitative effect of income tax progressivity on self-employment is also of significant magnitude. The coefficient estimate implies that a one percentage point increase in the marginal income tax rate (say from the mean value 0.28–0.29), holding the average income tax rate constant, leads to a decrease in the self-employment rate by about 0.19 percentage point. Thus our basic estimation result suggests that a decrease in income tax progressivity is associated with an increase in the rate of self-employment as measured by the own-account self-employment rate.
In column 2, we control for the average personal income tax rate to capture the income tax burden. As expected, the coefficient of average tax rate is negative and statistically significant indicating that higher income tax burden discourages self-employment. Our variable of interest, RIP, is still positive and statistically significant suggesting that the basic result is robust to the inclusion of the average income tax rate. The implication of this is that both the overall tax burden and tax progressivity matter for self-employment. Recall that RIP is measured as the ratio of one minus marginal tax rate to one minus average tax rate. Thus the estimated results indicate that, given the marginal tax rate, a one percentage point increase in the average tax rate (say from the mean value 0.15–0.16) leads to a decrease in the self-employment rate by about 0.29% points.
Following most of the related empirical studies in the literature, in column 3 we control for some of the factors that may influence self-employment. In particular, we include the unemployment rate, the relative self-employment income, corporate tax rate that is applicable to small business and minimum wage rate as control variables. Kamhi and Leung (2005) also include some of these control variables in their analysis. The coefficient of RIP is still positive and statistically significant indicating that tax progressivity affects the self-employment rate after we control for other determinants of self-employment.
So far we have used ordinary least square estimation on the assumption that the explanatory variables are exogenous. However, our theoretical model indicates that relative self-employment income is endogenous as it depends on the self-employment rate. Furthermore, the unemployment rate can also be potentially endogenous. The Hausman test for endogeneity also suggests that these variables are endogenous. It is well known that if the right hand side variables are endogenous, results from OLS may be biased. To address this issue we use the instrumental variable (IV) estimation method in column 4 treating relative income and the unemployment rate as endogenous variables.
A common empirical challenge in studies of self-employment is what instrument to use for the endogenous variables. A valid instrument should be fairly correlated with the endogenous variables but not with the self-employment rate. In studies based on aggregate data such as ours it is generally very difficult to come up with such instruments. One solution is to use lagged values of the endogenous variables as instruments since they are predetermined. However, the use of lagged values may be plagued by the potential problem of collinearity with the dependent variable if there is some persistence in the data. Thus to circumvent this potential problem, we instrument the relative income ratio with the two-period lagged value of the log of corporate profit to GDP ratio. The rationale for the use of this instrument is that generally when business conditions improve in the province (as measured by the increase in the corporate sector profit) the self-employment income also rises. It is known that when the unemployment ratio rises government spends more on unemployment benefits. Thus we use contemporaneous and one-period lagged values of the unemployment benefit to working age population ratio as instruments for the unemployment rate. We check the validity of these instruments using a standard Hansen over-identification test.
Column 4 shows the main results of this paper and we focus our analysis on this regression. The over-identification test does not reject the validity of our instruments for unemployment rate and relative income. Once again our variable of interest, RIP, is positive and highly significant. That is, a higher degree of personal income tax progressivity (lower RIP) reduces the return for successful entrepreneurs and discourages self-employment. Note also that the numerical magnitudes of the coefficient estimates of these variables are higher (in absolute value) under the IV method. Our coefficient estimate indicates that a one percentage point increase in the marginal income tax rate (say from 0.28 to 0.29), given the average income tax rate, leads to a fall in the self-employment rate by about 0.35 percentage point.8 Alternatively, a one standard deviation higher marginal income tax rate, given the average income tax rate, leads to a reduction in the self-employment rate by about 1.3% points.
How do our results compare with those of previous studies? Due to differences in methodology and the types of tax rates or tax progressivity measures used, it is not possible to make direct comparisons with earlier results. Our coefficient estimates for tax progressivity (RIP) are lower than those of Mooij and Nicodeme (2008), which is methodologically the closet to our study. Their study however is not based on aggregate data and their dependent variables are based on number of businesses rather than individuals. Our results can also be compared (albeit indirectly) to those of Folster (2002) and Bruce and Mohsin (2006) which used top personal income tax rate. While our empirical results are close to estimates obtained by Folster (2002) they are higher than those of Bruce and Mohsin (2006).
The control variables have their respective expected signs. A higher corporate tax rate reduces the return of businesses that are incorporated and hence discourages incorporated self-employment. On the other hand, an increase in the corporate tax rate increases the competitive advantage of the majority of the self-employed that are not incorporated (that are subject to the personal income tax rate) and raises the self-employment rate. Our result indicates that the coefficient of the corporate tax rate is negative but it is statistically insignificant. The coefficient of relative self-employment income is positive as expected and statistically significant. The coefficient of the unemployment rate on the other hand is negative but statistically insignificant. Furthermore, as in Blau (1987), an increase in the real minimum wage rate seems to encourage self-employment.
As explained before, there is no general consensus on how to measure the self-employment rate. Thus, as in previous empirical studies, we use other alternative measures of self-employment rate as our dependent variable in columns 5–7. In column 5 we use the ratio of total self-employment to total employment (SEMP 2) as a dependent variable. The results are quite similar to what we find in column 4 suggesting that our results do not depend on the type of self-employment rate that we use.
In column 6, we also use a measure of self-employment rate based on the number of people with income from self-employment. That is, we use the number of individuals with self-employment income as a ratio of total number of recipients of self-employment and wage income (SEMP 3). The regression results are broadly consistent with what we find in column 4. More importantly, despite the use of a different measure of self-employment rate, the qualitative effect of tax progressivity on self-employment does not change much.
Another measure of self-employment used in our analysis is based on the number of individual tax-filers. As in Harvey (2004) and Bruce and Mohsin (2006), we use the number of tax filers with self-employment income as a ratio of total number of tax-filers (SEMP 4). The data on tax-filers are available only for the period 1989–2006. The regression result is given in column 7. Despite a significant fall in the number of observations, the qualitative and quantitative effects of income tax progressivity on self-employment are surprisingly similar to those found previously. More importantly, the coefficient of RIP is still positive and statistically significant. The magnitude of the coefficient estimate is slightly lower than those of our preferred regression of column 4. While the difference in the number of observations may limit our ability to make direct quantitative comparison, the result suggests that tax evasion opportunities are more important for this measure of self-employment.
In sum, our empirical results suggest that there is a robust positive association between RIP and self-employment. That is, the results imply that high tax progressivity discourages self-employment. Thus a general reduction in the marginal personal income tax rates (which reduces tax progressivity) encourages self-employment.
4 Sensitivity analysis
Robustness checks, 1979–2006
Without time dummies (IV)
Alternative progressivity measures
Average personal income tax rate
Our basic specification includes time effects. Including province invariant time effects may remove the yearly variation in tax rates. To check the sensitivity of our results to the presence of yearly dummies, in column 3 we exclude time dummies and re-estimate the model over the same period. The results again indicate that there is a strong positive association between RIP and self-employment rate.
It is well known that measuring the degree of tax progressivity is difficult. This is particularly true in analyses based on aggregate (rather than individual) data on taxation and income. The use of the RIP in our empirical analysis is consistent with our theoretical framework. However, to check the robustness of our empirical results we also use other alternative measures of tax progressivity in Table 4. More specifically, in column 4a we use a coefficient of RIP that is constructed using the marginal tax rate that is applicable to the average self-employment income and the average tax rate of singles (or unattached individuals). Furthermore, in column 4b we use a coefficient of RIP that is based on the (unweighted) average marginal tax rates of all income brackets and implicit tax rates of unattached individuals. In column 4c, the RIP measure is based on the top marginal personal income tax rate and the average tax rate of unattached individuals. In all cases the coefficient of RIP is positive and statistically significant. Thus our empirical result is robust to the use of differently constructed RIP measures.
In summary, our empirical results suggest that tax progressivity has a statistically significant negative association with self-employment. This result is quite robust to the various robustness checks. The implication of this is that the adverse impact of higher tax progressivity on self-employment outweighs the opportunity to evade taxes.
The effects of taxes on entrepreneurial activities have attracted a lot of attention recently. A number of previous studies indicate that taxes can affect self-employment in various ways. While some of the studies show that entrepreneurs’ decision to invest and hire can be significantly affected by the tax system, others explain why taxes can affect entry into entrepreneurial activities. In Canada income from self-employment is subject to personal income (after deducting for business expenses). For this reason, as compared to income from employment, the self-employed have a higher opportunity to avoid and evade taxes. These opportunities become greater as the degree of tax progressivity increases. Self-employment is inherently a risky occupation. As a result, a higher degree of tax progressivity reduces the return of successful entrepreneurs and discourages risk taking. Thus it is interesting to explore the net effects of personal income tax progressivity on the rate of self-employment. The main objective of this paper has been to examine whether an increase in tax progressivity encourages or impedes self-employment.
Theoretically, income tax progressivity has two opposite effects on self-employment. An increase in the income tax progressivity reduces the return of successful entrepreneurs. This discourages self-employment. On the other hand, the return from tax evasion and avoidance will be higher as tax progressivity rises. This in turn encourages self-employment. Thus the net effect of tax progressivity on self-employment is ambiguous. It all depends on the relative strength between the positive effects of tax evasion opportunity and the adverse effects of taxing success.
In order to examine the relationship between income tax progressivity and self-employment, we employ aggregate provincial data from Canada over the period 1979–2006. We use the own-account self-employment rate as our measure of self-employment. However, we also experimented with other related measures of self-employment. We measure the degree of tax progressivity by the coefficient of residual income progression. Our empirical analysis suggests that a reduction in marginal tax rates that lowers the degree of tax progressivity encourages self-employment rate. This implies that the negative effect of income tax progressivity on entrepreneurial risk-taking outweighs the tax evasion opportunity for the self-employed. We also conduct sensitivity analysis to check the robustness of the results. In general our empirical results are quite robust to various sensitivity checks.
The tax system may also encourage risk taking by entrepreneurs by treating income from self-employment more favourably than wage income.
The self-employed can also deduct personal consumption from business income (for example, business use of automobiles).
The details of the theoretical model are available upon request.
See Kamhi and Leung (2005) for a detailed analysis of self-employment and its various components in Canada.
Complete data set on the share of the agriculture sector in the own-account self-employment rate is not available.
Ideally we would prefer to use cluster-robust standard errors. However, due to the small number of clusters in our dataset (just ten provinces) cluster robust standard errors would be downward biased. See Wooldridge (2003) and the reference contained therein.
Given the average income tax rate, a one percentage point increase in the marginal income tax rate leads to a fall in the value of RIP by about 0.0035 (say from 0.8462 to 0.8427).
Funding for this project was provided by Donner Canadian Foundation. I would like to thank Bev Dahlby for his comments on an earlier version of this paper.