EMMA NEUMAN 2016:11. Performance and job creation among self-employed immigrants and natives in Sweden

EMMA NEUMAN 2016:11 Performance and job creation among self-employed immigrants and natives in Sweden

Performance and job creation among self-employed immigrants and natives in Sweden Emma Neuman a Abstract This paper uses individual panel data for all self-employed in the retail and service sectors to study how immigrant and native owned firms perform and contribute to job creation in Sweden. In particular, we use an individual-fixed effects model to explore how self-employment outcomes among immigrants and natives evolve with years in self-employment. The results show that native men have higher earnings and profit level from selfemployment activities than immigrant men. Immigrant men converge towards the earnings and profit levels of native men as self-employment experience increase, but do not reach parity. On the contrary, immigrant women catch up with the earnings and profit levels of native women after between 3 to 5 years in business. Turnover is highest for men, and in firms owned by non-european immigrants, independently of length of business experience. Immigrant firms, and in particular those owned by non-european immigrants, contribute more to job creation than firms owned by natives. As length of business experience increases immigrants hire additional persons to their firms to a higher extent than natives. JEL-classification: J15; L25; L26 Keywords: Immigrants; Job creation; Performance; Self-employment a Centre for Labour Market and Discrimination Studies, Linnaeus University, SE-351 95 Växjö, emma.neuman@lnu.se. The author thanks Lina Aldén and Mats Hammarstedt for useful comments and suggestions. Financial support from the project Obstacles and possibilities for self-employment among immigrants in the retail and trade sector funded by the Jan Wallander and Tom Hedelius Foundation is greatly appreciated.

1 Introduction Immigration has increased rapidly in many countries and the labour market integration of immigrants in the host countries has received the attention of researchers and politicians. In Sweden the share of the population born abroad nearly doubled between 1970 and 2000 (Statistics Sweden, various years). 1 Over time the migration flows have also shifted from consisting mainly of labour-force migrants from Europe, to be more likely to comprise refugees and tied-movers from outside of Europe. It is well known that in Sweden, as in other countries, immigrants perform worse on the labour market in comparison to natives. Self-employment has often been suggested as a way to increase labour market participation among the foreign born. In several countries immigration policies have been constructed to facilitate immigration of potential entrepreneurs (Schuetze and Antecol, 2006; Fairlie and Lofstrom, 2015). In Sweden the Swedish government proposes increased selfemployment rates among the newly arrived immigrants as a way to better utilise their competencies (SOU 2015/16:100). In previous research it has been shown that immigrants tend to be self-employed to a higher degree than natives in many developed countries (see e.g. Borjas, 1986; Fairlie and Meyer, 1996; Fairlie and Robb, 2007; Robb and Fairlie, 2009 for studies on the US, see e.g. Clark and Drinkwater, 2000, 2010; Hammarstedt, 2001; Constant and Zimmermann, 2006; Andersson and Hammarstedt, 2011 for studies on European countries). Immigrants and natives appear to enter self-employment partly due to different factors (for an overview see e.g. Simoes et al., 2016). For instance, immigrants might face discrimination by employers in paid labour market work, and this pushes them into self-employment (Clark and Drinkwater, 2000). Furthermore, there is reason to expect that immigrants face more difficulties at start-up of their firms than do natives. Research has shown that immigrants and minorities face problems, such as attaining capital and understanding laws and regulations, when they start a new business, (e.g. Blanchflower et al., 2003; Blanchard et al., 2008; Asiedu et al., 2012; Aldén and Hammarstedt, 2016). It is also possible that poor language skills in the host country language restrict immigrant self-employment (Bates, 1999). We would therefore expect that immigrant firms are less successful and grow slower than native firms in the start-up phase. However, it is still unknown whether the potential differences in performance and job creation between immigrants and natives remain as the owners gain self-employment experience. Until now few studies have investigated how immigrants and natives self-employment outcomes develop with business experience. Lofstrom (2011) studies how the earnings among self-employed immigrants and natives in the US evolve with years in business. He finds that the earnings of self-employed immigrant men increase with years in business and at a faster rate than for native men. Among women, self-employed immigrants have a similar earnings growth as self-employed natives initially, but after about 5 years immigrants earnings decrease with additional years in business. In this paper we contribute to this literature by studying how immigrant and native owned firms in Sweden grow, both in terms of more employees and better performance, as they gain selfemployment experience. In contrast to Lofstrom (2011) we will focus on a broader set of performance measures, including earnings, turnover, and profit and also study job creation. Firms started and operated by immigrants 1 A detailed description of the history of migration to Sweden is provided by Lundh and Ohlsson (1999). 1

contribute to the host economy, not the least by increasing employment and income from self-employment (Fairlie and Lofstrom, 2015). However, there is a need for further research on the contribution of immigrant firms to job creation (e.g. Kerr and Kerr, 2016; Fairlie and Lofstrom, 2015). The paper makes use of Swedish longitudinal register data for the years 1998 to 2007. This data provides information on individual and firm characteristics for all self-employed natives and immigrants in Sweden. We will restrict the focus to self-employed in the retail and service sectors in Sweden. These sectors are of particular interest for immigrants since many immigrants have established their businesses there (Andersson and Hammarstedt, 2011). 2 Literature review 2.1 The self-employment performance of immigrants and natives The self-employment performance among immigrants to different countries has been extensively studied. 2 Several studies have investigated differences in performance between self-employed immigrants and natives. In this literature it emerges that the self-employment experience of immigrants differs across host countries as well as for different immigrant groups. In Sweden, non-european immigrants have lower earnings and profit from their firms in comparison to their native counterparts, while the self-employment performance of immigrants originating from Europe is more similar to native performance (Hammarstedt, 2006; Andersson and Hammarstedt, 2011; Andersson Joona, 2011). Evidence for the US shows that immigrant businesses-owners have lower earnings in comparison to self-employed natives (Lofstrom, 2002). However, Asian-owned businesses, which are mainly immigrant owned, have higher sales and profits in comparison to natives (Robb and Fairlie, 2009). In any study of immigrant earnings is it important to take into account the duration of stay in the host country. A large literature on immigrant earnings assimilation in the wage sector has shown that immigrants tend to have lower earnings than natives initially, but the gap reduces as time in the host country increases (see e.g. Chiswick, 1978; Borjas, 1985). A few papers expand the immigrant assimilation literature to the self-employment sector. Lofstrom (2002, 2011) finds that the earnings of self-employed immigrants in the US converge or even surpass the earnings of self-employed natives as they age. In Canada and Australia it takes longer time for self-employed immigrants to reach earnings parity with natives in comparison to the US (Schuetze and Antecol, 2006). Furthermore, the literature investigating important factors for job creation among the self-employed is rather scarce (Henley, 2005; Fairlie and Miranda, 2016). In general men tend to hire more employees in their firms in comparison to women (van Praag and Cramer, 2001; Fairlie and Miranda, 2016; Burke et al., 2002; Henley, 2005). Moreover, there are relatively few studies on ethnic differences in job creation. Andersson and Hammarstedt (2011) find that in Sweden firms owned by non-european immigrants tend have more employees and higher turnover than natives and European immigrants firms. In contrast, Henley (2005) finds that ethnic minorities in the UK have a lower likelihood to hire employees to their firms in comparison to natives. Furthermore, recent evidence from the US shows that immigrant and native business-owners tend to hire employees to the same extent if differences in individual characteristics are taken into account (Fairlie and 2 For a review of the literature on immigrant business performance see Fairlie and Lofstrom (2015). 2

Miranda, 2016). Although, among firms having survived three years in business in the US immigrant businessowners employ additional personnel at a faster rate than natives (Kerr and Kerr, 2016). 2.2 The role of self-employment experience It has been established in the literature that a high level of human capital is of great importance for successful self-employment. For the self-employed, age, education, and experience can be considered as the most relevant components of human capital. 3 Older persons have longer labour market experience, more wisdom and knowledge of institutions, better social and business networks, and more financial capital making them more likely to enter self-employment and to be successful in it (Cowling et al, 2004; Constant et al, 2007). Furthermore, it is well-known that individuals to a high degree start businesses in the same type of industry as the one where they previously worked in paid employment. Shane (2003) means that one reason might be that the industry-specific information and skills gathered when working raise the expected value of exploitation and increase individuals effort at start-up. In addition to the general experience attained along with ageing and working, specific experience from self-employment spells might prove valuable for self-employment performance. A person with experience of being self-employed might have learnt about business opportunities and attained skills in selling, negotiating, leading, planning, decision-making, problem-solving, organising, and communicating (Cowling et al, 2004; Parker, 2009; Shane, 2003). Against this background it seems reasonable to expect that self-employment performance increase with age, labour market experience, and experience from spells of self-employment. Up to now, few studies have distinguished the impact of experience in paid employment from that of experience in self-employment on individuals performance in self-employment (Van der Sluis et al., 2008). In general, the empirical evidence for that previous experience from being self-employed enhances performance as self-employed is mixed (Burke et al., 2002; Shane, 2003; Van der Sluis et al., 2008). Storey (1994) argues that more experience from self-employment might capture both that individuals have learnt skills relevant for successful self-employment, and in addition that individuals have previously been selfemployed and failed because of lack in managerial ability. The mixed empirical results on the role of selfemployment experience for self-employment performance might be due to the lack of distinction between experience from current and previous spells of self-employment. Previous literature has focused mainly on differences between immigrants and natives in self-employment performance over the life-cycle, i.e. related to aging. However, differences in native and immigrant business performance might also emerge because of variation in length of experience of self-employment. For instance, immigrants might face additional difficulties in the business start-up process. Research has shown that immigrants and minorities face problems, such as attaining capital and understanding laws and regulations, when they start a new business, (e.g. Blanchflower et al., 2003; Blanchard et al., 2008; Asiedu et al., 2012; Aldén and Hammarstedt, 2016). It is also possible that poor language skills in the host country language restrict immigrant self-employment (Bates, 1999). We would therefore expect that immigrant firms are less successful and grow slower than native firms initially. Few studies have investigated whether the potential differences in performance between immigrants and natives remain as the owners gain self-employment experience. In a cross-sectional study on Germany Constant et al. (2007) find that longer business tenure is positively correlated with 3 For an overview on human capital and self-employment see e.g. Parker (2009). 3

immigrants earnings, but not related to natives earnings. To our knowledge Lofstrom (2011) is the only study which answers this question with help of individual panel data. Lofstrom finds that the earnings of low-skilled self-employed immigrant men grow faster than for native men with years in business. The results for low-skilled women show that self-employed immigrant women have a slower earnings growth than self-employed native women. Moreover, in an analysis of the returns to self-employment it is important to take into account differences in survival rates, not least to avoid non-random selection out of self-employment which could also be related to performance. It has been shown that immigrant firms do not survive for as long time as native firms (e.g. Fairlie and Meyer, 1996; Bates, 1999; Fairlie and Robb, 2007; Andersson Joona, 2010). In this paper we use individual panel data, which makes it possible to follow individuals and their firms over time. This enables estimations of the returns to self-employment and taking into account unobservable time-constant individual factors. 3 Data 3.1 Sample construction In this paper we use register data from the longitudinal data base LISA developed by Statistics Sweden. The data contains yearly information on individual, demographic, and firm characteristics for all individuals in Sweden older than 16 years. We restrict our sample to individuals working in the retail, trade or service sector and who have been registered as self-employed in at least one year during the period of our study, 1998 to 2007. 4 To be registered as self-employed earnings from self-employment need to be the individual s main source of income. 5 Only persons who entered self-employment in the period of our study are included, and thus individuals who were self-employed during all nine years are excluded. The reason is that the data does not include information on years of experience from self-employment prior to 1998, and this implies that information on years in selfemployment is missing for individuals who remained self-employed all nine years. The included individuals can still have experience from earlier spells of self-employment. In the estimations the individual fixed effect model will control for the effect of such previous experience. This implies that the return to experience will not be biased by previous experience, but the estimated gap at start-up might be biased by omitted previous selfemployment experience. Furthermore, in order to capture persons with high attachment to the labour market (e.g. not part-time retired persons or students) we include only individuals aged 25 to 64 years. In total the number of individual-year observations amounts to 431,371, with 155,081 unique entrants. About 24 per cent are immigrants and a majority (around 61 per cent) are men. Individuals are defined as immigrants if they were born in a country other than Sweden and as natives if they were born in Sweden. 4 About 0.6 per cent of the sample is self-employed in limited liabilities. Controlling for whether the firm is a private firm or a limited liability has no impact on the results. The results are not displayed, but available from the author upon request. 5 If individuals have income from both wage-employment and self-employment Statistic Sweden defines them as self-employed if the income from self-employment is higher than the income from wage-employment. Selfemployment income is up-weighted with the factor 1.6 since income from self-employment is underestimated in relation to the number of hours spent in this activity. 4

3.2 Descriptive statistics Table 1 presents the performance of self-employed immigrants and natives by years in self-employment, separately for men and women. The firms performance is measured by yearly income from self-employment, yearly firm turnover (income from sales net of taxes and discounts), and yearly firm profit (income minus costs, excluding deductions related to interest rates and taxes). In addition, the owner s hiring decisions are described by the average number of employees at the firm and the share of firms having at least one employee. It appears that immigrants have more employees than natives and their firms are bigger in terms of turnover, but at the same time their profit and earnings are lower than natives. Moreover, men tend to hire more persons than women, and men s firms are more successful. All groups improve their performance as years in self-employment increase, but not many persons stay in self-employment for the full eight years. This highlights the importance of utilizing individual fixed effects, i.e. making comparisons within the individual, which will make non-random selection out of self-employment caused by time-invariant factors like ability or motivation less of a concern. Table 1: Performance by years since start-up Native men Immigrant men Native women Immigrant women At year of start-up Number of employees 0.24 0.35 0.22 0.28 Share with employees 9.44 19.44 8.93 14.11 Yearly real earnings(1000 SEK) 107.96 75.13 72.06 57.03 Yearly real net turnover(1000 SEK) 66.99 77.18 53.08 63.27 Yearly real operating profit/loss(1000 SEK) 20.77 13.76 13.89 11.33 Observations 69,032 25,509 48,193 12,347 After 4 years Number of employees 0.36 0.60 0.24 0.32 Share with employees 15.06 29.45 12.06 16.25 Yearly real earnings(1000 SEK) 159.25 118.60 113.99 96.34 Yearly real net turnover(1000 SEK) 86.22 103.02 51.90 68.39 Yearly real operating profit/loss(1000 SEK) 26.06 19.30 17.22 15.80 Observations 12,087 5,029 9,021 1,908 After 8 years Number of employees 0.41 0.68 0.26 0.23 Share with employees 18.32 29.26 14.80 15.93 Yearly real earnings(1000 SEK) 171.02 135.50 123.76 103.05 Yearly real net turnover(1000 SEK) 96.92 88.26 52.30 51.04 Yearly real operating profit/loss(1000 SEK) 27.19 21.69 18.40 15.21 Observations 1,212 475 1,034 182 1 svensk krona (SEK) US Dollar 0.15 in 2007.Earnings, profit, and turnover are expressed in real values, i.e. adjusted for changes in the price level. Table 2 displays some descriptive statistics for our sample. It emerges that natives have somewhat higher educational attainment and are about two years older than immigrants. Almost half of the native men are active in financial and business services, while immigrant men are over-represented among hotel and restaurant owners. Among women, near 40 per cent of natives are active in personal and other services, while immigrants 5

are more evenly spread in a number of different sectors. On average men and native women have been running their own firms for around 1.6 years, while immigrant women are slightly less experienced with about 1.5 years. Immigrants are more likely to be married and have more children than natives. The average length of residence in Sweden is about 18 years for immigrant women, and about one year shorter for immigrant men. Table 2: Summary statistics Native men Immigrant men Native women Immigrant women Primary education (%) 16.88 25.08 13.03 19.97 (37.45) (43.35) (33.66) (39.98) Secondary education (%) 49.12 43.45 53.77 44.69 (49.99) (49.57) (49.86) (49.72) Post-secondary or tertiary education (%) 33.77 28.29 33.07 32.58 (47.29) (45.04) (47.05) (46.87) Age 44.84 42.18 43.85 42.12 (11.25) (9.13) (10.58) (9.50) Business line (%) Trade and repair of motor vehicles 7.69 4.55 0.87 0.73 (26.64) (20.85) (9.29) (8.52) Wholesale and commission trade 6.73 4.41 3.42 4.35 (25.06) (20.53) (18.16) (20.40) Retail trade 9.83 18.03 14.69 19.33 (29.77) (38.44) (35.40) (39.49) Transportation and warehousing 10.80 16.52 1.96 2.35 (31.04) (37.14) (13.87) (15.14) Postal and telecommunication 0.38 0.18 0.09 0.07 (6.18) (4.27) (3.08) (2.63) Financial and business services 46.80 16.73 35.42 25.26 (49.90) (37.33) (47.83) (43.45) Hotels and restaurants 3.65 31.28 5.59 20.44 (18.74) (46.37) (22.97) (40.33) Personal and other services 14.12 8.29 37.96 27.47 (34.82) (27.57) (48.53) (44.63) 1.61 1.66 1.67 1.50 (1.80) (1.81) (1.85) (1.74) Married (%) 46.94 66.40 50.83 63.79 (49.91) (47.23) (49.99) (48.06) Number of children 0.659 1.189 0.833 1.067 (0.990) (1.277) (1.043) (1.169) Years since migration 16.99 18.08 (10.38) (11.77) Observations 190,107 72,975 136,147 32,142 Displayed are sample means and standard deviations within parentheses. 4 Empirical analysis 4.1 Self-employment performance Our aim is to study the relative success of self-employed immigrants to self-employed natives and especially how this is affected by self-employment experience. From the descriptive statistics it emerges that performance differs across groups, native men have the highest earnings and profit and immigrant men have highest turnover and most employees. However, part of the group differences could be explained by individual characteristics 6

such as education, age, family composition, or business line. To take into account the differences in observable characteristics we estimate the following OLS regression separately for men and women. yy "# = XX "# ββ + ρρ ) tt "# +ρρ + tt "# + + ττ # +γγ # + θθ # + εε "# (1) where y it is the performance of individual i in year t. As measures of performance we will use the logarithm of individuals yearly earnings, and their firm s yearly turnover and profit. This implies that we exclude observations with values equal to or below zero. 6 The matrix X it includes individual characteristics (a quadratic function of age, educational level, labour earnings, marital status, number of children, and a quadratic function of years since migration for immigrants (set to zero for natives) ). t it is a trend that measures the effect of years in self-employment, specified as a quadratic function. In addition, the regression includes year fixed effects, τ t, sector fixed effects, γ t, and region fixed effects, θ t. 7 All independent variables are allowed to vary for immigrants and natives, implying that they are interacted with an indicator variable for being foreign born. ε it is the error term. Although equation (1) includes a large number of individual characteristics it is still possible that omitted variables, such as motivation, talent, and ability, affect the self-employment performance. To reduce this bias we will estimate regressions where we include individual fixed effects, α i, which control for all individual unobservable factors that are constant over time. Furthermore, the individual fixed effects implicitly control for selection out from self-employment that is related to time-constant factors, however, it is still possible that timevarying factors might affect self-employment exit of natives and immigrants differently. We estimate the following equation by OLS: yy "# = αα " +XX "# ββ + ρρ ) tt "# +ρρ + tt "# + + ττ # +γγ # + θθ # + εε "# (2) where all variables are defined as above. The year fixed effects are dropped in the fixed effects estimations, since it is not possible to include both age and year of observation simultaneously as they are perfectly correlated for each individual. We include the yearly local unemployment rate to reduce the bias from the omitted year dummies (Barth et al, 2004). For comparative reasons, we will estimate equation (1) without year fixed effects and controlling for the yearly local unemployment rate instead. Furthermore, a drawback with including individual fixed effects in the model is that this makes it impossible to identify the initial performance gap at year of start-up. In the figures presented below, the plots based on regressions including individual fixed effects will have the starting values imposed as the sample means for each group. Our main purpose is to estimate differences in the return to experience in self-employment between immigrants and natives. However, the 6 6,081 observations are excluded in the earnings regressions, 25,459 observations are excluded in regressions for profit, and 8,378 observations are excluded in the turnover regressions. As a robustness check we have estimated regressions without logarithmic values and these are presented in Table A1 in the appendix. These results are in line with the baseline results, although the differences between immigrants and natives returns to experience tend to be smaller. 7 A region is defined as a county and Sweden contains 21 counties. We have investigated whether the results are robust to specifying years in self-employment and the other time variant variables (age and years since migration) as indicator variables instead of continuous variables. This had little impact on the results and does not change the main conclusions drawn. The results are not displayed, but available from the author upon request. 7

performance gap at start-up is of interest especially for how the results can be interpreted in terms of immigrant assimilation. The initial performance gap incorporates differences in observed individual characteristics, but does not take into account potential selection into self-employment. If natives and immigrants entering selfemployment differ in terms of e.g. previous self-employment experience, talent, or ability, this can bias the estimated difference in performance levels. In other words, the differences in levels should be interpreted with caution. In table 3 the results for the regressions of yearly earnings from self-employment are displayed. From column (1) it emerges that for native women earnings increase with about 31 per cent with one additional year of selfemployment experience. The yearly return to experience is about 4 per cent lower for immigrant women. In column (2) the year fixed effects are omitted and the local unemployment rate is included and this has little effect on the results. Including individual fixed effects in column (3) reduces the returns to experience substantially and there is no longer a statistically significant difference between natives and immigrants. It is likely that a large part of the observed returns to experience is due to the fact that those with high ability, motivation and talent stay in self-employment and the unsuccessful leave. However, when taking time-constant factors such as entrepreneurial ability into account, experience still increases earnings, and in a similar magnitude for natives and immigrants. The results for men show that native men have an earnings increase of about 28 per cent for each additional year in self-employment, and immigrant men have between 1 and 3 per cent lower returns to self-employment (columns (4) and (5)). In line with the results for women, including individual fixed effects reduces the returns to experience and there is no longer any statistically significant difference between immigrants and natives. General experience attained with ageing also appears beneficial for the self-employed. Being one year older generates an increase in earnings of about 20 per cent for women and immigrant men, and around 25 per cent for native men (columns (3) and (6)). In figure 1 the predicted earnings from the regressions in table 3 are plotted against years in self-employment. The graph to the left includes year fixed effects (column (1) and (4)) and the graph to the right includes individual fixed effects (column (3) and (6)). At start-up, the predicted earnings are lower for women than for men, and immigrants predicted earnings are lower than natives. As experience increase all groups increase their earnings and the graph without individual fixed effects shows a rather high earnings growth and no catch-up of immigrants. Adding individual fixed effects implies that the earnings growth is reduced substantially and immigrant women catch up with native women after about 5 years in self-employment. For men, natives still earn more than immigrants 8 years after start-up. 8

Table 3: OLS Regressions for Yearly Self-employment Earnings Women Men VARIABLES (1) (2) (3) (4) (5) (6) 0.313*** 0.326*** 0.106*** 0.281*** 0.291*** 0.094*** (0.005) (0.005) (0.006) (0.004) (0.004) (0.005) Immigrant* -0.038*** -0.026** 0.014-0.034*** -0.014** 0.012 (0.012) (0.011) (0.014) (0.007) (0.007) (0.009) squared -0.029*** -0.029*** -0.013*** -0.027*** -0.026*** -0.012*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Immigrant* squared 0.005*** 0.003* -0.002 0.003** 0.001-0.004*** (0.002) (0.002) (0.002) (0.001) (0.001) (0.001) Age 0.095*** 0.094*** 0.206*** 0.112*** 0.110*** 0.252*** (0.004) (0.004) (0.009) (0.003) (0.003) (0.008) Age squared -0.001*** -0.001*** -0.002*** -0.001*** -0.001*** -0.002*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Immigrant*Age -0.045*** -0.044*** 0.009-0.061*** -0.063*** -0.044** (0.010) (0.010) (0.031) (0.007) (0.007) (0.018) Immigrant*Age squared 0.001*** 0.001*** 0.000 0.001*** 0.001*** 0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Years since migration 0.021*** 0.021*** 0.023 0.010*** 0.012*** 0.047*** (0.003) (0.003) (0.018) (0.002) (0.002) (0.007) Years since migration squared -0.000*** -0.000*** -0.001*** -0.000-0.000-0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Individual fixed effects No No Yes No No Yes Year fixed effects Yes No No Yes No No Local unemployment rate No Yes Yes No Yes Yes R-squared 0.160 0.154 0.079 0.162 0.155 0.088 Prediction at sample means 6.572 6.572 6.572 6.913 6.913 6.913 Number of individuals 59,718 59,718 59,718 93,388 93,388 93,388 Number of observations 165,596 165,596 165,596 260,054 260,054 260,054 Robust standard errors, clustered on individuals, in parentheses. ***p<0.01 **p<0.05 *p<0.1 Dependent variable is logarithm of yearly self-employment earnings. All regressions include controls for education, marital status, number of children, and income from wage-employment. In addition, region and sector fixed effects are included. Without individual fixed effects With individual fixed effects LnEarnings 6 6.5 7 7.5 LnEarnings 6 6.5 7 7.5 Immigrant women Native women Immigrant men Native men 95% CI Figure 1: Predicted yearly earnings by years in self-employment 9

Table 4 shows the results from the estimations of yearly profit from self-employment. Profit increases with additional years in business, and the yearly return is around 20 per cent for native women (column (1) and (2)) and men (column (3) and (4)). Among women, self-employment experience has a similar impact on the profit level of immigrant and native firms. Immigrant men have about 2 per cent lower return to self-employment experience in comparison to native men as indicted by the results in column (4), but this result is not robust to excluding the year dummies and adding the local unemployment rate in column (5). Controlling for individual fixed effects reduces the returns to experience to about 5 per cent for women and 3 per cent for men. Furthermore, the fixed effects regressions indicate that immigrant women (men) gain about 5 (1) per cent additional profit from one more year of experience than do native women (men). Moreover, being one year older generates an increase in profit of around 20 per cent (columns (3) and (6)). In figure 2 the regression results from table 4 are used to plot the predicted profit against years in selfemployment. At start-up, men have higher predicted profit than women, and natives have higher predicted profit than immigrants. When including individual fixed effects it appears that immigrant men approaches but do not fully catch up with the profit level of native men, while immigrant women reach native women s profit level after about 4 years of business experience. Table 4: OLS Regressions for Yearly Profit from Self-employment Women Men VARIABLES (1) (2) (3) (4) (5) (6) 0.233*** 0.233*** 0.047*** 0.197*** 0.201*** 0.031*** (0.005) (0.005) (0.005) (0.004) (0.004) (0.004) Immigrant* 0.003 0.014 0.045*** -0.021*** -0.001 0.013* (0.011) (0.011) (0.012) (0.007) (0.007) (0.008) squared -0.023*** -0.021*** -0.006*** -0.020*** -0.019*** -0.004*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Immigrant* squared -0.000-0.002-0.004*** 0.001-0.001-0.004*** (0.002) (0.002) (0.001) (0.001) (0.001) (0.001) Age 0.083*** 0.082*** 0.185*** 0.097*** 0.095*** 0.203*** (0.004) (0.004) (0.009) (0.003) (0.003) (0.007) Age squared -0.001*** -0.001*** -0.002*** -0.001*** -0.001*** -0.002*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Immigrant*Age -0.061*** -0.061*** -0.000-0.062*** -0.064*** -0.016 (0.010) (0.010) (0.027) (0.007) (0.007) (0.017) Immigrant*Age squared 0.001*** 0.001*** 0.000 0.001*** 0.001*** 0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Years since migration 0.026*** 0.026*** 0.022 0.018*** 0.020*** 0.038*** (0.003) (0.003) (0.014) (0.002) (0.002) (0.008) Years since migration squared -0.000*** -0.000*** -0.001*** -0.000** -0.000*** -0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Individual fixed effects No No Yes No No Yes Year fixed effects Yes No No Yes No No Local unemployment rate No Yes Yes No Yes Yes R-squared 0.126 0.123 0.054 0.127 0.120 0.059 Prediction at sample means 4.831 4.831 4.831 5.166 5.166 5.166 Number of individuals 57,632 57,632 57,632 90,572 90,572 90,572 Number of observations 157,524 157,524 157,524 248,748 248,748 248,748 Robust standard errors, clustered on individuals, in parentheses. ***p<0.01 **p<0.05 *p<0.1 Dependent variable is logarithm of yearly firm profit. All regressions include controls for education, marital status, number of children, and income from wageemployment. In addition, region and sector fixed effects are included. 10

Without individual fixed effects With individual fixed effects LnProfit 4.5 5 5.5 LnProfit 4.5 5 5.5 Immigrant women Immigrant men 95% CI Native women Native men Figure 2: Predicted yearly profit by years since in self-employment The regression results for yearly turnover are displayed in table 5. In line with the results for profit and earnings, turnover increases with about 19 per cent for each additional year in self-employment. Immigrant men have around 3 per cent lower returns to experience than native men (columns (4) and (5)). If time constant individual factors are taken into account the increase in turnover form having longer business experience decreases for both men and women. The fixed effects estimations for firm turnover provide no evidence for that the return to years in self-employment differs for immigrants and natives. Furthermore, being one year older generates an increase in turnover of close to 20 per cent (columns (3) and (6)). Figure 3 displays the predicted turnover based on the regressions presented in table 5. In the left panel estimates without individual fixed effects are displayed and it appears that immigrants at start-up have higher turnover than natives, but among women natives catch up with immigrants after about 7 years of experience. However, when individual fixed effects are included the gaps observed at start-up are of about similar magnitude after 8 years of experience. This is in contrast to the results for earnings and profit, where a large part of the initial gaps between immigrants and natives are closed 8 years after start-up. In general, the results for the performance measures indicate that immigrants gain more from longer self-employment experience in comparison to natives. Noteworthy is also the fact that failing to incorporate time-invariant individual factors severely bias the results, and would lead to the misleading conclusion that immigrants have lower a return to experience than natives. 11

Table 5: OLS Regressions for Yearly Firm Turnover Women Men VARIABLES (1) (2) (3) (4) (5) (6) 0.198*** 0.198*** 0.038*** 0.185*** 0.187*** 0.049*** (0.005) (0.005) (0.005) (0.004) (0.004) (0.004) Immigrant* -0.009-0.001 0.011-0.038*** -0.024*** -0.009 (0.011) (0.011) (0.010) (0.007) (0.007) (0.007) squared -0.020*** -0.018*** -0.004*** -0.020*** -0.018*** -0.004*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.000) Immigrant* squared -0.001-0.003-0.002* 0.001 0.000-0.002*** (0.002) (0.002) (0.001) (0.001) (0.001) (0.001) Age 0.083*** 0.082*** 0.169*** 0.102*** 0.101*** 0.194*** (0.004) (0.004) (0.008) (0.003) (0.004) (0.007) Age squared -0.001*** -0.001*** -0.002*** -0.001*** -0.001*** -0.002*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Immigrant*Age -0.057*** -0.057*** 0.014-0.077*** -0.078*** -0.013 (0.010) (0.010) (0.021) (0.007) (0.007) (0.014) Immigrant*Age squared 0.001*** 0.001*** -0.000 0.001*** 0.001*** 0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Years since migration 0.031*** 0.032*** 0.015 0.034*** 0.035*** 0.024*** (0.003) (0.003) (0.010) (0.002) (0.002) (0.006) Years since migration squared -0.000*** -0.000*** -0.000-0.000*** -0.001*** -0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Individual fixed effects No No Yes No No Yes Year fixed effects Yes No No Yes No No Local unemployment rate No Yes Yes No Yes Yes R-squared 0.212 0.209 0.057 0.182 0.176 0.074 Prediction at sample means 5.823 5.823 5.823 6.263 6.263 6.263 Number of individuals 60,005 60,005 60,005 93,235 93,235 93,235 Number of observations 165,075 165,075 165,075 258,278 258,278 258,278 Robust standard errors, clustered on individuals, in parentheses. ***p<0.01 **p<0.05 *p<0.1 Dependent variable is logarithm of yearly firm turnover. All regressions include controls for education, marital status, number of children, and income from wageemployment. In addition, region and sector fixed effects are included. Without individual fixed effects With individual fixed effects LnTurnover 5.5 6 6.5 7 LnTurnover 5.5 6 6.5 7 Immigrant women Immigrant men 95% CI Native women Native men Figure 3: Predicted yearly turnover by years in self-employment 12

4.2 Job creation among the self-employed From the society s point of view not only is it important that businesses generate income and are profitable, but they also have the potential to generate jobs for others. Therefore, we will study hiring decisions among immigrant and native business owners and how these change as time in self-employment increases. We will estimate three outcomes with the same specifications as described above in equation (1) and (2). The outcomes considered are the firm s number of employees, the likelihood of having employees, and the probability of hiring additional persons to the firm. All regressions will be estimated using OLS, implying that we use a linear probability model for the two last outcomes. The results from the regressions of the number of employees at the firm are displayed in table 6. It emerges that both men and women hire more persons as they gain more business experience. With and additional year of selfemployment experience the number of employees increase with about 0.02 persons from the mean of 0.24 for women and around 0.06 persons from the mean of 0.34 for men. It appears to be no statistically significant differences between natives and immigrants as regards how many more they hire when self-employment experience increases. It emerges that older persons have more employees at their firms, being one year older is associated with about 0.06 more employees for women and native men, and about 0.22 additional employees for immigrant men (columns (3) and (6)). In figure 4 the firm s predicted number of employees is plotted against years in self-employment. It appears that immigrant men tend to have almost twice as many employees as the other groups. Immigrant women have more employees than native women initially, but after eight years native women instead tend to have more persons employed than immigrant women. However, the average number of employees is never more than one person, so many of the firms might never have any employees. Therefore, we also consider the likelihood of having employees and how this changes with experience in table 7. The predicted probability of having employees (evaluated at the sample means) is about 12 per cent for women and 15 per cent for men. With an additional year of self-employment this probability increases with between 1 to 2 percentage points. The regression results do not indicate that the rate at which the probability of having employees increases with experience is different for immigrants and natives. In figure 5 it appears that immigrants have a higher probability of having employees independently of how long experience in self-employment they had, and the gaps in the predicted probability between immigrants and natives remain quite the same over all eight years. 13

Table 6: OLS Regressions for Number of Employees at Firm Women Men VARIABLES (1) (2) (3) (4) (5) (6) 0.023** 0.024*** 0.010** 0.065*** 0.069*** 0.049*** (0.009) (0.007) (0.005) (0.019) (0.018) (0.016) Immigrant* 0.012 0.011 0.008-0.010-0.009-0.036 (0.014) (0.013) (0.011) (0.021) (0.020) (0.022) squared -0.001-0.001 0.000-0.007*** -0.007*** -0.003 (0.001) (0.001) (0.001) (0.003) (0.003) (0.002) Immigrant* squared -0.004** -0.003* -0.002 0.002 0.004 0.001 (0.002) (0.002) (0.002) (0.003) (0.003) (0.003) Age 0.017*** 0.017*** 0.061*** 0.025*** 0.024*** 0.065*** (0.005) (0.005) (0.008) (0.007) (0.007) (0.009) Age squared -0.000*** -0.000*** -0.001*** -0.000*** -0.000*** -0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Immigrant*Age -0.015-0.015 0.047-0.016-0.017 0.153*** (0.010) (0.010) (0.035) (0.010) (0.010) (0.042) Immigrant*Age squared 0.000 0.000-0.001 0.000 0.000-0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Years since migration 0.012*** 0.012*** -0.000 0.024*** 0.024*** -0.017 (0.003) (0.003) (0.009) (0.002) (0.002) (0.012) Years since migration squared -0.000** -0.000** 0.000-0.000*** -0.000*** 0.001 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Individual fixed effects No No Yes No No Yes Year fixed effects Yes No No Yes No No Local unemployment rate No Yes Yes No Yes Yes R-squared 0.009 0.009 0.001 0.010 0.010 0.003 Prediction at sample means 0.24 0.24 0.24 0.34 0.34 0.34 Number of individuals 60,972 60,972 60,972 94,676 94,676 94,676 Number of observations 168,289 168,289 168,289 263,082 263,082 263,082 Robust standard errors, clustered on individuals, in parentheses. ***p<0.01 **p<0.05 *p<0.1 All regressions include controls for education, marital status, number of children, and income from wage-employment. In addition, region and sector fixed effects are included. Without individual fixed effects With individual fixed effects Number of employees.2.4.6.8 Number of employees.2.4.6.8 Immigrant women Immigrant men 95% CI Native women Native men Figure 4: Predicted number of employees at firm by years in self-employment 14

Table 7: OLS Regressions for Probability of Having Employees at Firm Women Men VARIABLES (1) (2) (3) (4) (5) (6) 1.625*** 1.666*** 0.933*** 1.671*** 1.787*** 1.295*** (0.132) (0.129) (0.135) (0.121) (0.119) (0.128) Immigrant* 0.097 0.282-0.022 1.242*** 1.503*** 0.396 (0.354) (0.347) (0.361) (0.274) (0.270) (0.300) squared -0.127*** -0.116*** -0.061*** -0.162*** -0.139*** -0.096*** (0.024) (0.023) (0.019) (0.023) (0.022) (0.019) Immigrant* squared -0.122* -0.098-0.033-0.267*** -0.213*** -0.180*** (0.065) (0.063) (0.051) (0.049) (0.048) (0.042) Age 0.403*** 0.393*** 2.492*** 0.651*** 0.615*** 3.497*** (0.108) (0.108) (0.249) (0.093) (0.093) (0.220) Age squared -0.006*** -0.006*** -0.024*** -0.008*** -0.008*** -0.033*** (0.001) (0.001) (0.003) (0.001) (0.001) (0.002) Immigrant*Age -0.410-0.417 0.745-0.819*** -0.890*** 3.815*** (0.281) (0.282) (0.701) (0.230) (0.231) (0.664) Immigrant*Age squared 0.004 0.004-0.000 0.006** 0.007** -0.018*** (0.003) (0.003) (0.008) (0.003) (0.003) (0.006) Years since migration 0.423*** 0.418*** 0.769*** 0.947*** 0.968*** 0.428 (0.090) (0.090) (0.209) (0.067) (0.067) (0.346) Years since migration squared -0.006*** -0.006*** -0.016*** -0.014*** -0.014*** -0.018*** (0.002) (0.002) (0.005) (0.001) (0.001) (0.005) Individual fixed effects No No Yes No No Yes Year fixed effects Yes No No Yes No No Local unemployment rate No Yes Yes No Yes Yes R-squared 0.114 0.112 0.021 0.122 0.116 0.034 Predicted probability at sample means 11.59 11.59 11.59 15.46 15.46 15.46 Number of individuals 60,972 60,972 60,972 94,676 94,676 94,676 Number of observations 168,289 168,289 168,289 263,082 263,082 263,082 Robust standard errors, clustered on individuals, in parentheses. ***p<0.01 **p<0.05 *p<0.1 All regressions include controls for education, marital status, number of children, and income from wage-employment. In addition, region and sector fixed effects are included. The coefficients are reported in percentage points. Without individual fixed effects With individual fixed effects obab ty a g e p oyees 10 15 20 25 30 35 Probability having employees 10 15 20 25 30 35 Immigrant women Immigrant men 95% CI Native women Native men Figure 5: Predicted probability of having employees at firm by years since start-up 15

Finally we consider the probability of hiring more persons (independently of the firm s number of employees in the previous period). The results from the regressions with the dependent variable being the probability of hiring are displayed in table 8. The predicted probability of hiring is about 3 per cent for women and 5 per cent for men at the sample means. Increasing self-employment experience by one more year implies an increase by around 2 percentage points of hiring for natives and an about twice as large increase for immigrants. From figure 6 we can see that the probability of hiring follows an inverted u-shaped pattern for all groups. Initially, the likelihood of hiring increases but after about four years it starts to decrease. When individual fixed effects are included the probability of hiring is always higher for immigrants than for natives. Table 8: OLS Regressions for Probability of Hiring more Employees to Firm Women Men VARIABLES (1) (2) (3) (4) (5) (6) 2.242*** 2.259*** 2.203*** 2.509*** 2.559*** 2.493*** (0.073) (0.072) (0.089) (0.070) (0.070) (0.085) Immigrant* 1.585*** 1.582*** 1.390*** 3.331*** 3.318*** 2.893*** (0.197) (0.195) (0.243) (0.170) (0.168) (0.207) squared -0.263*** -0.257*** -0.259*** -0.298*** -0.291*** -0.275*** (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Immigrant* squared -0.247*** -0.227*** -0.166*** -0.462*** -0.419*** -0.400*** (0.036) (0.035) (0.035) (0.031) (0.031) (0.032) Age 0.095** 0.088** 0.969*** 0.123*** 0.109*** 1.450*** (0.044) (0.044) (0.145) (0.041) (0.041) (0.131) Age squared -0.002*** -0.002*** -0.010*** -0.002*** -0.002*** -0.016*** (0.000) (0.000) (0.002) (0.000) (0.000) (0.001) Immigrant*Age 0.055 0.056 0.948** -0.230** -0.258** 2.135*** (0.117) (0.117) (0.429) (0.108) (0.108) (0.437) Immigrant*Age squared -0.001-0.001-0.012** 0.001 0.002-0.015*** (0.001) (0.001) (0.005) (0.001) (0.001) (0.004) Years since migration 0.053 0.049 0.426*** 0.231*** 0.233*** -0.275 (0.034) (0.034) (0.126) (0.029) (0.029) (0.235) Years since migration squared -0.001-0.001-0.004-0.004*** -0.004*** 0.000 (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) Individual fixed effects No No Yes No No Yes Year fixed effects Yes No No Yes No No Local unemployment rate No Yes Yes No Yes Yes R-squared 0.040 0.038 0.017 0.058 0.055 0.026 Prediction at sample means 3.47 3.47 3.47 5.25 5.25 5.25 Number of individuals 60,972 60,972 60,972 94,676 94,676 94,676 Number of observations 168,289 168,289 168,289 263,082 263,082 263,082 Robust standard errors, clustered on individuals, in parentheses. ***p<0.01 **p<0.05 *p<0.1 All regressions include controls for education, marital status, number of children, and income from wage-employment. In addition, region and sector fixed effects are included. The coefficients are reported in percentage points. 16

Without individual fixed effects With individual fixed effects obab ty g e p oyees 0 5 10 15 20 Probability hiring employees 0 5 10 15 20 Immigrant women Immigrant men 95% CI Native women Native men Figure 6: Predicted probability of hiring additional employees to firm by years since in self-employment 4.3 Heterogeneity Previous research has shown that self-employment performance among immigrants in Sweden differ depending on whether they are from European or non-european countries (e.g. Andersson Joona, 2011; Hammarstedt, 2006; Andersson and Hammarstedt, 2011). In order to test for this we have estimated the regressions as specified in equation [2], but distinguished between immigrants from European and non-european countries. The predicted earnings, profit, turnover and probability of having employees are plotted in figure 7 for European immigrants, non-european immigrants, and natives separately. The solid lines are the predictions for women, while the dashed lines display men s predicted outcomes. As regards earnings and profit, women from non- European countries earn the least to start with but experience a steeper growth than natives and European immigrants as they get more experienced. However, after eight years there are no statically significant differences between the groups among women. Similarly, earnings and profit are lowest for men from non- European countries at start-up of the firms. Male business owners from non-european countries increase their earnings and profit more than the other groups as they gain experience and they eventually reach the level of native men. On the contrary, European immigrant men have a lower return to experience and do not reach the profit and earnings level of native men. Furthermore, the turnover and probability of having employees are much higher for non-european immigrants than for European immigrant and natives, irrespectively of gender. The firms of non-european immigrants continue to be bigger and employ more persons as business experience increases. 17