Group-specific Effects of Inter‐regional Mobility on Earnings – A Microdata Analysis for Germany

Abstract

Lehmer F. and Möller J. Group-specific effects of inter-regional mobility on earnings – a microdata analysis for Germany, Regional Studies. The paper analyses the relationship between regional mobility and earnings for different groups of workers. Using a large panel microdata set the authors find negative earnings differentials of movers in the year before migration and strong evidence for significant wage gains through mobility. A decomposition of Blinder/Oaxaca type reveals different group-specific rewards effects suggesting a positive post-mobility wage differential of movers over the incumbent workforce for some groups irrespective of the region of destination. The existence of a general wage growth effect of mobility appears to be robust and cannot be explained by the time-invariant part of unobserved heterogeneity.

Lehmer F. et Möller J. L'impact sur les revenus la mobilité interrégionale des groupes spécifiques des groupes sur les revenus – une analyse des données microéconomiques pour l'Allemagne, Regional Studies. Cet article cherche à analyser le rapport entre la mobilité régionale et les salaires pour divers groupes de travailleurs. A partir d'un puissant ensemble de données microéconomiques provenant d'une enquête par panel, il est évident que les écarts des salaires des migrants l'année qui précède la migration s'avèrent négatifs et qu'il y a de fortes preuves en faveur d'importants gains des salaires à cause de la mobilité. Une décomposition du type Blinder/Oaxaca laisse voir des gains qui varient suivant le groupe, ce qui laisse supposer pour certains groupes un écart des salaires positif pour les migrants suite à la migration par rapport à la main- d'oeuvre autochtone, quelle que soit la zone d'accueil. Il semble que la présence d'un effet généralisé de mobilité sur la croissance des salaires est solide et ne s'explique pas par la partie de l'hétérogénéité non-observée qui ne varie pas dans le temps.

Mobilité interrégionale Ecarts des salaires interrégionaux Hétérogénéité non-observée Propension à la correspondance des scores

Lehmer F. und Möller J. Gruppenspezifische Effekte der interregionalen Mobilität auf das Einkommen – eine Mikrodatenanalyse für Deutschland, Regional Studies. In diesem Beitrag wird die Beziehung zwischen regionaler Mobilität und Einkommen für verschiedene Gruppen von Arbeitnehmern analysiert. Anhand eines umfangreichen Satzes an Panel-Mikrodaten ermitteln wir negative Einkommensdifferentiale von Ortwechslern im Jahr vor der Migration sowie starke Anzeichen für signifikante Einkommensverbesserungen durch Mobilität. Bei einer Dekomposition des Typs Blinder/Oaxaca zeigen sich verschiedene gruppenspezifische Entlohnungseffekte, was für einige Gruppen unabhängig von der Zielregion auf ein positives Einkommensdifferential nach Mobilität von Ortwechslern im Vergleich zu den übrigen Arbeitnehmern schließen lässt. Die Existenz eines generell einkommenssteigernden Effekts der Mobilität erscheint robust und lässt sich nicht durch den zeitlich invarianten Teil einer nicht beobachteten Heterogenität erklären.

Interregionale Mobilität Migrationsbedingte Lohnunterschiede Nicht beobachtete Heterogenität Propensity score matching

Lehmer F. y Möller J. Efectos de la movilidad interregional en los ingresos de grupos específicos: un análisis con microdatos de Alemania, Regional Studies. En este ensayo analizamos las relaciones entre la movilidad regional y los ingresos para diferentes grupos de trabajadores. Con ayuda de un gran panel de microdatos observamos diferencias negativas en los ingresosde personas que se desplazaron en el año antes de la migración y pruebas ólidas de importantes ingresos adicionales gracias a esta movilidad. Una descomposición del tipo Blinder/Oaxaca muestra diferentes efectos de gratificaciones para determinados grupos lo que indica un diferencia positiva de sueldos después de la migración en los trabajadores beneficiados para algunos grupos independientemente de la región de destino. La existencia de un efecto de crecimiento general de ingresos causado por la movilidad parece ser sólida y no puede explicarse por la parte de tiempo invariable de heterogeneidad no observada.

Movilidad interregional Diferenciales de sueldo por migración Heterogeneidad no observada Propensity score matching

Keywords:

• Inter-regional mobility
• Migration wage differentials
• Unobserved heterogeneity
• Propensity score matching

JEL classifications:

• J61
• R23

Acknowledgements

The authors are grateful to two anonymous referees for valuable suggestions for improving the paper. They would also like to thank the participants of a session on regional labour markets at the annual meeting of the German Economic Association (Verein für Socialpolitik) and Alisher Aldashev for very helpful comments. The authors alone are responsible for any remaining errors. Financial support from the German Science Foundation (Deutsche Forschungsgemeinschaft, Projekt MO 523/3-2) is gratefully acknowledged.

Notes

1. The standard human capital model of migration predicts that workers migrate when the discounted value of real income available at a potential destination exceeds that at the origin by more than the costs of moving (Sjaastad, 1962).

2. Topel and Ward (1992) state that job search plays a crucial role for wage growth; they estimate that about one-third of overall wage growth in the first decade of working life can be attributed to job switching.

3. Some older studies dealing with the determinants affecting the probability of migration are Da Vanzo (1978) and Herzog and Schlottmann (1981) for the USA. For a survey of other relevant studies see Greenwood (1975, 1985).

4. Spatial differences in productivity are crucial for explaining spatial wage differentials. Empirical studies in this context typically find a statistically significant positive relationship between density measures of economic activity and productivity (e.g. Ciccone and Hall, 1996; Harris and Ioannides, 2000). This supports the results of previous studies focusing on the positive effects of city population or industry employment on productivity (e.g. Sveikauskas, 1975; Segal, 1976; Moomaw, 1981, 1985; Henderson, 1986).

5. In this data source gross daily earnings are calculated as average over the observed employment period for each person. The notions wages and earnings are used synonymously throughout this paper.

6. The Nomenclature of Territorial Units for Statistics (NUTS) was developed by EUROSTAT. It is a standard for referencing the administrative division of countries for statistical purposes. NUTS-3 refers to county or district level, respectively.

7. For an overview of region types according to this classification see Table A1 in the Appendix.

8. See, for instance, Kemper (2004) for an exploration of migration patterns in Western and Eastern Germany.

9. This definition does not differentiate between migration and commuting. In analogy to the distinction made by Eliasson et al. (2003, p. 831), the definition of movers in this paper includes the following categories: (1) workers who change their region type of residence and the region type of workplace; (2) workers who do not change their place of residence, but start commuting to a different type of region; and (3) commuters who do not change their place of residence, but change the region type where the workplace is located. Because our definition of mobility is based on region type, our concept of mobility is predominantly related to the first category. Note that adjacent regions are in many cases of the same type.

10. The described differences are robust within the sample period 1990 to 1997.

11. The measure of concentration is calculated as: 100*share of movers of this category in total movers divided by the share of movers and stayers of this category in total workers.

12. Hunt (2004) states that those results are strongly influenced by a special group of movers. Workers who migrate from one state to another without changing the employer are more highly educated than stayers.

13. Here and in the following potential experience in years is measured as age minus average duration of education minus 6. For low-skilled workers without an upper secondary education we assume 10 years as the average educational period, for low-skilled workers with an upper secondary education 13 years, for skilled workers 12.5 and 15 years respectively, for highly-skilled workers holding a polytechnic type of degree 16 years and for highly skilled alumni of a university 18 years.

14. The potential work experience is categorized as follows: low experience: 0–9 years; medium experience: 10–19 years; high experience: 20 or more years. In order to avoid problems with cell sizes being too small, the authors aggregated the BBR-region types and firm size categories (see Tables A1 and A2 in the Appendix).

15. For some years in the sample the share of movers is even under-represented in this region type. In 1994 and 1992, for example, the share of mobile workers in RT1 was just 49%, while more than 50% of all stayers worked in this region type.

16. See Mincer (1974).

17. Here we included a category ‘firm size missing’.

18. All workers except for low-skilled male and female workers are considered to be qualified. All interactions are defined for the linear and quadratic experience variable.

19. The study calculated corresponding estimates for all successive pairs of years in the sample. It turns out that the findings are sufficiently robust over time. In order to save space, the results in the following are for the most recent years only. The results for other pairs of years are available from the authors on request.

• 20. The differential of low-skilled male workers relative to the average in the economy, for example, is obtained as:

• where ω _n denotes the share of category n workers in total employment. The skill differentials of workers in categories n(2, 3,…, 6) are re-calculated according to the formula:

• A corresponding procedure was applied to the coefficients of firm size and region type category variables as well.

21. An explanation of the Blinder–Oaxaca (1973) type decomposition technique is given in the Appendix.

22. The differentiation by gender and firm size is neglected to keep the table readable.

23. This is in accordance with the findings of Yankow (2003) for the USA. He points to the fact that this group of migrants searches for immediate wage gains, while highly educated young migrants invest in their human capital.

24. Note that the overall differential between highly-skilled movers and stayers in RT1 is negative, while the differential is positive for all experience groups. This is due to the fact that experience (or age) of movers and stayers differs markedly. Typically the group of young or not experienced workers is clearly over-represented in the group of movers. The fact that this group earns significantly less than the high-experience group explains the negative difference (−2.53) for the category RT1/highly skilled.

25. A further objection against this method is that only workers are considered who are employed before and after moving. If participation and employment rates vary systematically over types of regions, these results cannot be generalized to the whole working age population. Pekkala and Tervo (2002) present an approach which explicitly takes account of this selectivity issue. Their approach requires instruments which are not available in our data set. However, we checked the existence of a possible influence of the type of region on employment and participation rates. A scatter plot between population density and employment or participation rates across 439 German NUTS-3 regions shows no significant relationship. Therefore, the authors feel confident that this possible source of bias in these results is not substantial.

26. For an overview of recent developments of this approach see Cobb-Clark and Crossley (2003) or Smith and Todd (2005).

27. See Roy (1951) and Rubin (1974).

28. The basic idea goes back to the seminal contribution of Rosenbaum and Rubin (1983).

29. The results of the probit model are not documented in the paper, but are available on request from the authors. For the calculation of the matching model they used the Psmatch2 Stata module (Version 3.0.0) by Leuven and Sianesi (2003).

30. The authors conducted the usual diagnostics on the success of matching without finding any clues for questioning the results. The common support assumption is fulfilled in our case. After matching, the differences in characteristics between movers and controls are statistically insignificant.

31. The results are available from the authors on request.