The two tax reforms provide two dimensions along which the treatment and control groups can be distinguished. First, the legislation changes were only applicable to tradesmen as opposed to liberal professionals. Second, only individuals with gross income from self- employment above a certain threshold were affected by the reforms. Table 5 sketches the definition of the treatment and control groups.
Table 5: Treatment and control groups
Potential Gross Self-Empl. Income Below Threshold
Potential Gross Self-Empl. Income Above Threshold Potential Tradesmen
(Gewerbetreibende) control group treatment group
Pot. Liberal Professes-
sionals (Freiberufler) control group control group
The distinction between tradesmen and liberal professionals in the German tax law has its origins in the 19th century. It was sometimes justified by the opinion that tradesmen used capital more intensively in production than liberal professionals.28 Bach, Broer, and Fossen (2008) illustrate empirically that production structures differ little between tradesmen and liberal professionals nowadays.29 The self -employed can be identified as tradesmen or liberal professionals by the profession they report, since the liberal professions are defined by an official catalogue in the German income tax law. In some special cases this catalogue is
27
The difference is not only a formal one. It determines the level of personal risk, which is a prominent char acteristic of an entrepreneur. Furthermore, from a public viewpoint, the distinction is important as it determines taxes, social security obligations and benefit entitlements.
28
The Federal Constitutional Court (1977) cast serious doubt on this argument. In 2003, the German federal government also stated that there was no reason to distinguish between tradesmen and liberal professionals, but its initiative to eliminate this distinction in the tax law failed to pass parliament (German Bundestag 2003).
29 Dent ists and medical specialists (liberal professionals) typically work with higher capital intensity than
software firms (tradesmen), for example. Law firms (liberal professionals) often have more employees than crafts businesses (tradesmen).
rendered more precise by court ruling. Table A 1 lists the professions classified as liberal. Additionally, I assign a label “potential tradesman” or “potential liberal professional” to persons who are currently dependently employed or not working. This is necessary to determine who would benefit from the tax rate reduction if he or she decided to become self- employed. Again, the reported professions determine the classification. Someone is labelled as a “potential liberal professional” if the person reports a profession that would be a liberal profession if he or she were self-employed, no matter if the person is actually self-employed or not.30
Individuals who are unemployed and those not participating in the labour force usually do not report a profession. Hence, the data do not provide obvious criteria to determine if someone is likely to become either a tradesman or a liberal professional. The person is somewhat arbitrarily labelled as “potential tradesman”. However, as it is unlikely that the income of formerly unemployed or not working individuals will be above the relevant threshold in the first year of starting up their own business, they are part of the control group irrespective of their classification as trade smen or liberal professionals.
Apart from distinguishing between tradesmen and liberal professionals, the second dimension for separating out the treatment group requires determinin g if an individual’s gross income from self-employment is or would be above the threshold defined by the income tax law. As the threshold first set in 1994 was reduced in 1999 and 2000, I create two dummy variables: one indicating if an individual’s income would be above the threshold prevailing in 1994, and another one indicating if it would be above the 2000 threshold. Since the two steps of the second reform were part of the same act, the temporary 1999 threshold is not treated separately.
30
For example, a physician can be employed in a hospital, unemployed/not working, or self-employed. In the latter case he or she would be a liberal professional in the sense of the law. Thus, all physicians are interpreted as “potential liberal professionals”.
Income from se lf-employment of the self-employed is observed directly.31 For the dependently employed, unemployed and not working individuals, I estimate the counter- factual income they would earn if they decided to be self-employed. Mincer type regressions of income from self -employment are estimated separately for each year. As explanatory variables I include an individual’s age and its square, the number of children, and dummy variables indicating the type of secondary schooling, professional qualification, gender, marital status, residence in eastern or western Germany, the city size, and a set of industry dummies.32 The estimated earnings equations are used to predict potential income from self- employment for those who are not self-employed.33 In the model of the probability of being self-employed, I use predicted income from self-employment for the self-employed, too, in order to treat all observations in the same way.
The microcensus only provides net income, and only as a categorical variable with 18-24 classes. Thus, I approximate a self-employed person’s net income by the midpoint of his/her net income class and use it as the dependent variable in the regression of self-employment earnings described above. Then, I employ a simple function to compute an approximate gr oss income from the predicted net income, using average income tax rates by income de ciles in Germany as calculated by Bach, Corneo, and Steiner (2005).34
A choice has to be made whether to use real or nominal incomes. As the income threshold defined by the tax law refers to nominal income, at first sight it seems reasonable to use nominal incomes in order to decide who is really affected by the law. If I used nominal incomes, however, people would creep up from the low income group to the high income group with time just due to inflation. Haan and Steiner (2005) showed that this bracket
31 The data do not allow separating out additional wage and salary or capital income of the self-employed. In
principle, this would be necessary because the tax rate limitation only applied to income from trade business. Using the German Socio Economic Panel (SOEP), a representative household panel survey for Germany provided by the German Institute for Economic Research (DIW Berlin), I find that additional wage income of the self-employed was on average only 4.6 % of the income from self -employment in the period 1991-2001. Capital income was on average 17.3 % of self-employment income for tradesmen and 20.6 % for liberal professionals. My inability to subtract capital income from the income of the self-employed may lead to an overestimation of income from self-employment in the income predictions and thus to groups of potential high- income tradesmen and liberal professionals which are somewhat too large. As the effect is similar for the treatment and control groups, I do not expect this to cause a major distortion in the analysis. To be sure, in the empirical analysis a sensitivity check regarding a possible misclassification of income groups is performed (section 3.5.2).
32
The industry dummies are normalised in such a way that the base category represents the average industry effect. This allows accounting for industry effects in the estimation of earnings equations also for currently non- employed people, for whom no industry affiliation is observed, by assigning them the average effect across al l industries.
33 The regression and prediction results are available from the author upon request.
34 This simple tax function cannot account for joint taxation of married couples. I tested the robustness of the
cree ping effect is quite substantial in the German tax system. For the application of this chapter’s estimation method, comparability of the treatment and comparison groups before and after the reforms is crucial. This requires a definition that is consistent and stable over time. For this analysis, it is relevant to distinguish between individuals who expect to find themselves in the high income or the low income group if self-employed. Thus, I decided to deflate all incomes (and also the thresholds defined by the law) using the Consumer Price Index.
Table A 2 in the Appendix presents descriptive statistics for the self-employed in Germany, separated by tradesmen with income above and below the income threshold relevant for the 1999 reform, and liberal professionals. The weighted mean of the self- employment rate in the sample is 6.1 %. The mean characteristics reveal important differences between the various groups of self-employed people: 53.2 % of the liberal professionals have a university degree, but only 11.8 % of the tradesmen above the income threshold and even only 7.2 % of those below the threshold; the reason is that most liberal professions are academic professions, whereas a large share of the tradesmen are craftsmen, shopkeepers etc. Descriptive statistics displaying differences in individual characteristics between self- employed, dependently employed and unemployed/inactive individuals are summarised in Table A 3.
As described above, all individuals who would be tradesmen with an expected income above the threshold if they decided to be self-employed belong to the treatment group of the refor ms. Table A 4 compares characteristics of this group to those of potential tradesmen with lower income and potential liberal professionals. The actual self-employment rate is highest among potential liberal professionals (20.8 %) and lowest among potential tradesmen with low income (4.7 %). The latter is by far the largest group in the sample (89.1 %). Only 2.8 % of the potential tradesmen in the high income group are women. One reason for this low share is that a large share of women in Germany work part-time, which results in relatively low estimates of income for women.