Low- and high-skill migration flows: free mobility versus other determinants

Abstract

We investigate what economic factors drive international migration of workers to France and how their influence varies across different skill levels under restrictive policies and through time as free mobility is implemented. We find that neither incentives nor policy parameters are similar across skill levels. Migration drivers such as a network of compatriots and relative incomes influence the movement of low-skill workers. High-skill individuals however move only according to financial opportunities be they standard of living or returns to skill within a class. We conclude that competition for high-skill workers among Organization for Economic Co-operation and Development (OECD) countries requires more than free mobility to successfully attract high-skill migrants even for a developed country such as France.

Acknowledgements

We thank two anonymous referees for their constructive comments, and Brigitte van Baalen, Vangheli Lakiotis and Florence Miguet from the University of Geneva for their very helpful research assistance. Financial support from SSHRC-INE program, HRSDC and Industry Canada, is gratefully acknowledged.

Notes

¹ The international economics literature has addressed indirectly related issues such as the possible substitution or complementarity between trade and migration flows (see Harris and Schmitt, 2003, for a survey) or Foreign Direct Investment (FDI) and migration flows (Kugler and Rapoport, 2007). A simple model of two-way migration by skill levels is presented by Schmitt and Soubeyran (2006).

² The credential recognition by employers whether objective or subjective may thus be an issue to take into account insofar as it favours immigrants from some source countries.

³ See Clark et al. (2007) for a characterization of the net rate of immigration when relative wage inequality changes with skills at the individual level.

⁴ Cutler et al. (2008) find that ethnic concentration has a positive impact on earnings; Anderson et al. (2009) find that recent immigrants benefit from networks by having higher employment rates and higher earnings when employed. Also, Edin et al. (2003) show that the benefits from the network vary with immigrants’ characteristics and skills in particular. Their results show that the least skilled immigrants benefit the most from networks and ‘for immigrants with high school or more education, there is no evidence of an enclave effect’ (p. 348). It is less easy to find direct evidence about the second theoretical result. However, since our empirical test of the clustering hypothesis is independent of the wage premium, the empirical analysis depends on only one of these two results.

⁵ See EEC (1994, Appendix V, pp. 0325–6). In 1995, Austria, Finland and Sweden became members of the Union and their accession had no new implications for mobility.

⁶ Between 1995 and 2000, the total number of intra-company transferees to France increased by a factor of 2.5 (OECD, 2002, Part I, Table I4). It is likely however that these occupations represent only a small proportion of the high-skill category; nevertheless we do test whether the change had an impact on the magnitude of the flows.

⁷ Estimates show that, in the 1990s, the number ‘indirect entries’ on the labour market was about the same as that of new immigrant workers (Fondation Kastler, 2005).

⁸ The transformation is applied to the remaining small number of zero-observations (9, 4, 8 for low-, intermediate-, high-skill workers, respectively). A detailed description of all variables is provided in Appendix 2.

⁹ Since our theory is about income variation within skill classes which are empirically broadly defined we chose not to use the alternative source of wage provided by the International Labour Organization (ILO) October Enquiry on occupational wages (see, Freeman and Oostendorp, 2002) which would be too specific.

¹⁰ The only exception is America as there is no distinct data for North and South America; however, the distribution is almost even across Northern and Southern sample countries (i.e. 52.9% and 47.1% respectively in 1990 and 54.9% and 45.1% in 1999; Docquier and Marfouk, 2006).

¹¹ To our knowledge the only source for country-specific stock of immigrants in France is Docquier and Marfouk (2006). However, while most sample countries are covered the data is available only for two census years (1990 and 1999). So the gain in cross-sectional quality for the measure occurs at the cost of losing half the time dimension of the panel. Yet the results from the shortened sample with country-specific clustering data are robust qualitatively, i.e. cultural clustering is weaker for skilled workers (results available upon request).

¹² To test whether the large impact of opening up borders with Portugal on low-skill worker in 1992 biases the overall results, we added a dummy for that particular event. It is not significant implying that the FREEMOB dummy does capture the event adequately even for Portugal.

¹³ Note that the results are insensitive to the introduction of a time trend (results available upon request).

¹⁴ The results of the basic specification with a dummy for the exemption interacting with unemployment rate are

p-values are in parentheses.

¹⁵ Note that p(n) does not need to be very high (i.e. p(n) > 1 − (e_h  − e_l )/(w_e  − w_a )) for all migrants to shirk.