Job-to-job transitions and the job satisfaction puzzle: a diff-in-diff analysis for eleven EU countries

ABSTRACT

This paper studies the impact of different types of job-to-job transitions (from salaried employment to self-employment, from self-employment to salaried employment, and within salaried employment) on job satisfaction. Considering the three types of job transition, allows us to separate the pure mobility effect from the type of employment effect. We design an identification strategy based on the diff-in-diff approach. This is possible because our panel data allow us to compare the same individuals before and after job-to-job transitions occur. Our findings indicate that individuals who move from salaried employment to self-employment increase their overall job satisfaction more than workers who carry out other types of job transitions, while the impact of job changes on satisfaction in other job domains is mixed. We also find heterogeneous interactive effects of previous unemployment spells and self-perceived skill mismatch with job transitions depending on the type of transition. While skill mismatch is more important in the transition from salaried employment to self-employment, previous unemployment is more important for job changes within salaried employment.

KEYWORDS:

• Job satisfaction
• job-to-job transitions
• self-employment
• salaried employment
• unemployment

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ Some related studies published very recently also use the ECHP (Burke et al. 2018; Millán, Millán, and Román 2018; van Stel et al. 2018; Justo, Congregado, and Román 2021; Albiol-Sánchez, Díaz-Serrano, and Teruel 2021).

² These authors used cross-country panel data from 22 OECD countries for the period 1972–2007 to find a causal relationship between entrepreneurship and the business cycle.

³ We focus on the private sector since transitions in the public sector have certain particularities in comparison with the private sector (more bureaucratic nature of the labour). For instance, the transitions to self-employment in the public sector are remarkably scarce (Özcan and Reichstein 2009) and have a different nature (Demircioglu and Chowdhury 2021).

⁴ See Tables A1 and A2 in the Appendix for a description and a descriptive statistic of the variables used in the empirical analysis, respectively.

⁵ This model can be estimated in Stata using the command feologit (Baetschmann et al. 2020).

⁶ We decided to recode the variable in this way for two reasons. First, marginal effects for (y = 5) and (y = 6) were practically identical. Second, the marginal effects for lower levels of satisfaction (1, 2, 3 and 4) reported the opposite sign that the highest (5 and 6) in all models.

⁷ Table A2 in the Appendix shows the full set of estimated coefficients.

⁸ Both Hanglberger and Merz (2015) and Binder and Coad (2016) used the German Socioeconomic panel (SOEP) covering the period 1984–2009 and 1984–2010, respective. Millán, Millán, and Román (2018) and Justo, Congregado, and Román (2021), used the same data we use in here.

⁹ The full set of estimated coefficients is reported in Table A3 in the Appendix.

¹⁰ In this subsection we do not carry out separate estimations by country, since in some countries the sample of movers is too small to provide meaningful results.

¹¹ Belloc, Molina, and Velilla (2022) used the Working Conditions Survey of 2015 for 35 countries.

¹² The full set of estimated coefficients is reported in Table A5 in the Appendix. We have also carried out separate estimates for the Northern and Southern countries, not reported here, but we did not find a differentiated pattern regarding the interaction of the job transitions with skill mismatches and previous unemployment spells.

¹³ Unemployment is a dummy variable that takes the value 1 if individuals answer “yes” to the following question: “Existence of an unemployment period before current job”. Since in our sample we only consider those workers who were employed (either in salaried or self-employment) in t-1 and t, unemployment has occurred between survey periods t and t-1. This that is, between the previous job in t-1 and the new job in t. This circumstance implies that the unemployment spells experience between t-1 and t have a duration of less than one year.

Additional information

Funding

The authors acknowledge the financial support from the Spanish Ministry of Science and Innovation (grant # RTI2018-094733-B-I00) and ”La Caixa” Foundation (grant # 2014ACUP0130)