The role of spatial and skill mismatches: explaining long-term unemployment in Paris

ABSTRACT

In the Paris region, one can observe simultaneously the coexistence of a large and dynamic job pool with long-lasting periods of unemployment. This paradox reveals the importance of skill and spatial mismatch mechanisms, which are often used to explain disparities in local labour market outcomes. This paper uses several spatial models to measure the effects of these two mechanisms on unemployment durations in the Paris region. The results show that both problems affect municipalities close to the centre of Paris, while unemployment situations in municipalities on the fringes of the region are mainly affected by a lack of local employment dynamism.

KEYWORDS:

• unemployment
• spatial regime models
• generalized two stages least squares (GS2SLS)
• skills mismatch
• spatial mismatch

JEL:

• C41
• J64
• R1

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

ORCID

Yannick L’Horty http://orcid.org/0000-0001-6258-8565

Notes

1. The Paris region represents 31% of French gross domestic product (GDP) in 2012, while its population represents only 18% of the total French population (INSEE).

2. Historically, this hypothesis concerns populations considered as disadvantaged on the labour market. Here, we assume that this problem concerns all inhabitants without distinction.

3. That is, under-educated individuals often earn more than those with the same level of education working in jobs with the relevant education, but less than individuals with the level of education actually required in such jobs.

4. We did not employ the 2006 population census because it concerns a period subsequent to that observed in our administrative files.

5. A SIRET number is issued once a business has been registered with the relevant authority.

6. We have verified that other estimation strategies do not modify estimated unemployment-to-work transitions. Coefficients obtained in our case are close to those obtained with a piecewise constant hazard model or a Cox model. In our estimation, parameter α of the Weibull model – which determines the link between unemployment duration and exit rate at every period – is close to 1. This particular case corresponds to the exponential model, where unemployment duration has no effect on instantaneous unemployment exit rate. Then, all the usual generalizations give similar results.

7. By construction, this net duration neutralizes all socioeconomic differences of jobseekers because we imposed the same value of these socioeconomic variables on all municipalities.

8. The French employment zone has been retained as spatial unit because it is a geographical area within which most of the labour force lives and works. It is also an area where firms find the greater part of their labour force.

9. For the lowest diploma level (respectively the highest), we considered as mismatched the employees belonging to the first quartile (respectively the last quartile).

10. Appendix E in the supplemental data onlineverifies the relationship between unemployment duration and commuting time to the Paris centre. With a polynomial regression at order 3, we identify two turning points: unemployment durations decrease from the Paris centre to a commuting time between 30 and 40 min. Durations are relatively stable until the threshold of 60 min and then significantly increase. The pattern is similar globally for the 2001 and 2003 cohorts.

11. The null hypothesis is that of homoskedasticity. We reject the null hypothesis if the test statistic (nR²) is higher than the value obtained in the Chi² table (Wooldridge, 2002, ch. 8).

12. This result was also observed with the 2003 cohort (see Appendix G in the supplemental data online).