Incentives to self-employment decision in Sweden

Abstract

The objective of this paper is to empirically examine the determinants of the self‐employment decision, with a particular focus on gender differences and occupational choice by using register‐based individual panel data for the period 2003 to 2006. Individuals choose to move into self‐employment out of three possible initial statuses: paid employment, combiner or inactivity. These groups are of specific interest for the Swedish political agenda. The question that we specifically pose is how a set of socio‐economic factors separately induce men and women in these three statuses to enter self‐employment.

Keywords:

• self‐employment
• entry rate
• gender differences
• predicted earnings differential

JEL Classifications:

• J16
• J23
• J24
• J31
• M38

Acknowledgements

I am thankful to my colleague Stina Eklund for all useful support. Valuable discussions with Henry Ohlsson, Pernilla Andersson and Pontus Braunerhjelm have been helpful. I am very grateful to constructive comments from two anonymous referees.

Notes

1. See Folta, Delmar, and Wennberg (2009) for theoretical rationales and empirical evidence on the combiners’ labour market and their self‐employment entry in Sweden.

2. Baumgartner and Caliendo (2008) evaluate the positive effects of attempts to turn unemployed into self‐employed in the German labour market.

3. Mainly for foreign‐born individuals or for men only.

4. Parker (2009) finds that small firms are a better environment for new business creation than large firms. These new entrepreneurs share certain characteristics that make them choose employment in small firms in the first place.

5. See Rees and Shah (1986), de Wit and van Winden (1989), Taylor (1996) and Hammarstedt (2009).

6. See Lindh and Ohlsson (1996), Blanchflower and Oswald (1998), Johansson (2000) and Nykvist (2008).

7. The setup of the model follows Trost and Lee (1984) and Gyourko and Tracy (1988).

8. The time subscript is suppressed hereafter.

9. Incomes in our study cannot be adjusted for working hours. We assume though that individuals are aware of the long working hours that self‐employment may lead to and that it may be the total income from each status that plays the decisive role for switching.

10. Although the choice of these limits is made on an ad hoc basis, further sensitivity analysis of our estimations do not show any dramatic changes.

11. For a similar definition of combiners see Folta, Delmar, and Wennberg (2009).

12. The reduced form multinomial logit models estimated in the first stage contain all variables. Variables that are thought to affect the status choice but not earnings are excluded from the earnings equations for identification reasons. Capital income variables, intergenerational relations, firm‐size, start‐up subsidies for inactive, days in inactivity, inactive groups and the combiner distinction are thus excluded. Results on transition into wage‐employment, combiner status or inactivity are available upon request.

13. The results are expressed in log‐odds units. The reason for not choosing marginal effects is that they depend on where the independent variables are being evaluated (at the mean, quantiles or other points in the distribution). This in turn may lead to a change in sign for the parameter estimates, which makes them difficult to interpret.

14. Although we do not present a detailed discussion on these results, they are available upon request.

15. A formal test does not, however, support a significant difference between men and women. Performing such a test, which compares coefficients between men and women in binary choice models is not a trivial exercise (see Allison 1999). The outcome of applying it is available upon request.

16. We have no evidence in our data whether the respondents’ partners work in the same business.

17. If, for some reason, unemployment is caused as a result of structural transformations, these workers might still be very productive.

18. As discussed in de Wit and van Winden (1989), the inclusion of industry as an explanatory factor implies that the entry decision is made after the selection of industry. Our data show, however, that about 70% of employees and combiners enter self‐employment within the same industry.

19. Although the level of education among self‐employed is lower compared with employees (Table A2), the entry rate of highly educated employees into self‐employment is higher.