Elsevier

Economics Letters

Volume 109, Issue 3, December 2010, Pages 179-182
Economics Letters

Returns to schooling in self-employment

https://doi.org/10.1016/j.econlet.2010.05.014Get rights and content

Abstract

This paper challenges the idea that returns to schooling in self-employment are similar to those in wage work by establishing a non-linear relationship with very low returns for most educational levels in self-employment. We conclude that previous log-linear specifications in the self-employment literature have missed significant variation in the true relationship between education and earnings.

Introduction

This paper challenges the idea that returns to schooling in self-employment are similar to the returns to schooling in wage employment using detailed data from the Danish labor market. We show that the estimated average return to schooling of 6.5% per year hides substantial differences across different educational levels. Specifically, we find a highly non-linear relationship with very low returns to most educational levels in self-employment. The exceptions are a few specialized graduate levels of education (doctors and lawyers). Returns to schooling in wage employment, on the other hand, are found to be well approximated by the standard log-linear relationship.

Heterogenous returns to education for wage workers have been considered by, e.g., Heckman et al. (2006). While our results indicate that heterogeneous returns are likely to be much more important for self-employed than for wage workers, this issue has not received much attention in the self-employment literature. Most studies of the returns to schooling in self-employment rely on the log-linear specification, with some studies including one or at most a few dummies for educational attainment; see van der Sluis et al. (2008). In their recent meta-analysis, they report an average estimated return of 6.1% per year of education. While this corresponds well with the average return found in the present paper, our detailed dummy specification for educational length and type reveals that the large returns are concentrated among the longest educational levels, while shorter education carry very limited returns in self-employment. We conclude that previous log-linear specifications in the self-employment literature have missed significant variation in the true relationship between education and earnings for self-employed.

Section snippets

Data

We use data from the Integrated Data Base for Labor Market Research (“IDA”) which contains register data for all Danish residents since 1980. The data provide detailed information about occupations, earnings and experience, as well as background information about education and family characteristics. For more information about the IDA data, see Abowd and Kramarz (1999).

Occupations (self-employed, wage employed, etc.) are defined from an individual's primary labor market status in the last week

Results

We estimate a Mincer-type specification:lnYi=fEdui+βXi+uiwhere Yi is the annual surplus or wage of individual i, f(Edui) is a function of the educational attainment of individual i, and Xi contains other individual characteristics. ui is a random error. Throughout, we estimate Equation (1) by OLS.

In our benchmark version, f(Edui) is simply years of education, i.e., a log-linear relationship between annual surplus and years of education is assumed. Results from this specification are contained

Conclusion

This paper has revealed very diverse returns to different types of education in self-employment and very limited returns to a large number of these. Large returns are concentrated among the longer educational levels, and in particular within the medical and social sciences. Shorter education levels carry much smaller (if any) returns. Furthermore, this heterogeneity in educational returns does not seem to reflect a general feature of the Danish labor market, as the traditional log-linear

Acknowledgements

We thank participants at the CEBR conference on “Entrepreneurship: Occupational Choice and Financing” and at the Arne Ryde symposium “The Economics of Higher Education and the Education of Economists” for helpful comments. Jonas Helth Lønborg provided efficient research assistance. We thank the National Agency for Enterprise and Housing for financial support for this project. Sørensen gratefully acknowledges financial support from the Tuborg Foundation. The usual disclaimer applies.

References (5)

There are more references available in the full text version of this article.

Cited by (21)

  • Entrepreneurial ecosystem and the quality and quantity of regional entrepreneurship: A configurational approach

    2021, Journal of Business Research
    Citation Excerpt :

    Scholars agree that a high level of education correlates positively with the probability of being involved in new ventures (Davidssona & Honigb, 2003; Delmar & Davidsson, 2000): Education improves the ability of potential entrepreneurs to identify entrepreneurial opportunities. A high level of education is also linked to the success of startups (Iversen et al., 2010; Van Der Sluis et al., 2008): It increases the likelihood of entrepreneurial success by providing potential entrepreneurs with a high level of scientific and business-related knowledge. Moreover, a high level of human capital means an abundance of technical workers and experienced managers that can significantly enhance the competitiveness of new ventures (Andersson & Koster, 2011; Qian, 2013).

  • Curvilinear effects of openness and agreeableness on the imaginative capability of student designers

    2014, Thinking Skills and Creativity
    Citation Excerpt :

    Research to date has examined only the linear relationships between design imagination and designer personality (Liang, Chang, & Hsu, 2013; Lin et al., 2013). However, overlooking the existence of curvilinearity can lead to erroneous conclusions regarding the characteristics and magnitude of associations (Iversen, Malchow-Moller, & Sorensen, 2010), which includes associations between personality and its outcomes (Vasilopoulos, Cucina, & Hunter, 2007). This limitation suggests that additional studies on the subject may be beneficial.

  • Mentoring receipt and personality: Evidence for non-linear relationships

    2014, Journal of Business Research
    Citation Excerpt :

    However, the presence of curvilinear relationships is within reason. Ignoring the presence of curvilinearity can lead to erroneous conclusions over the nature and magnitude of associations (e.g., Iversen, Malchow-Moller, & Sorensen, 2010; Trevor, Gerhart, & Boudreau, 1997), and that includes associations between personality and its outcomes (Vasilopoulos, Cucina, & Hunter, 2007). These limitations suggest that additional investigations on the subject may be beneficial.

  • The returns to education for opportunity entrepreneurs, necessity entrepreneurs, and paid employees

    2013, Economics of Education Review
    Citation Excerpt :

    Next, we consider the possibility that the returns to education are not constant, which means that log wages are a nonlinear function of years of education, and that such nonlinearities may differ between paid employees, opportunity and necessity entrepreneurs. Based on data from Denmark, Iversen, Malchow-Møller, and Sørensen (2010) estimate non-linear OLS earnings regressions for the self-employed and find that their returns to education are small for most educational levels and large only for higher educational levels (they do not account for endogeneity of education or non-random selection, however). To detect possible nonlinearities in Germany, in specification (S8) we repeat the consistent FE estimations with quadratic years of education and the corresponding interactions with the opportunity and necessity entrepreneurship dummies.

  • Do serial entrepreneurs run successively better-performing businesses?

    2013, Journal of Business Venturing
    Citation Excerpt :

    First, any omitted variables would need to be correlated with lagged performance for any bias in γ to occur (Greene, 2003). Second, many determinants of performance, such as ability (Hartog et al., 2010) and education (Iversen et al., 2010) are time-invariant and so are differenced away in the estimation of Eq. (2) [see Section 4.3 above]. Third, it is possible to conduct some robustness checks which we turn to next.

View all citing articles on Scopus
View full text