Human Capital Externalities in Western Germany

Abstract

Using panel data and employing instrumental variables we show that regional wage differences across German regions are partly attributable to localized human capital externalities. This finding is stable across different indicators for regional aggregate education and robust to agglomeration, wage curve, price level and amenity effects. A comparison of our results with Moretti's findings for the USA suggests that national labour market institutions influence the distribution of wage gains from aggregate regional education among workers of different educational backgrounds. An analysis by sector reveals that human capital externalities are generally more pronounced in manufacturing than in the service sector.

Externalités des ressources humaines en Allemagne de l'Ouest

Résumé En utilisant des données pluridimensionnelles et en faisant usage de variables instrumentales, nous sommes en mesure de démontrer que les différences salariales régionales dans les différentes régions d'Allemagne sont attribuables, en partie, à des externalités en ressources humaines localisées. Cette conclusion est stable dans les différents indices pour l’éducation globale à l’échelon régional, et solide sur le plan de l'agglomération, des courbes salariales, des niveaux de prix, et des facteurs d'agrément. Une comparaison entre nos résultats et les conclusions de Moretti, pour les États-Unis, indique que les institutions nationales du marché du travail influent sur la distribution des gains salariaux découlant de l’éducation régionale globale parmi des travailleurs provenant de différents milieux éducatifs. Une analyse par secteur révèle que les externalités de ressources humaines sont généralement plus prononcées dans le secteur industriel que dans le secteur tertiaire.

Factores exógenos del capital humano en Alemania Occidental

Resumen Utilizando datos de panel y empleando variables instrumentales, demostramos que las diferencias salariales interregionales alemanas pueden atribuirse en parte a factores exógenos de capital humano confinados. Esta conclusión es estable entre los diferentes indicadores del acumulado regional de educación y consistente ante los efectos de la aglomeración, curva salarial, nivel de precios y servicios públicos. Una comparación de nuestros resultados con las conclusiones de Moretti relativas a los EE.UU., sugiere que las instituciones asociadas con el mercado laboral nacional ejercen influencia sobre la distribución de las ganancias salariales del acumulado de educación regional entre los trabajadores de los diferentes niveles de estudio. Un análisis por sector revela que los factores exógenos de capital humano son, en términos generales, más pronunciados en el sector de producción que en el de servicios.

Keywords:

• Human capital externalities
• regional wage differences

JEL classification::

• D62
• J24
• J31
• O15

Notes

1. Excellent surveys of this literature are provided by Moretti (2004a) and Duranton (2006).

2. Südekum (2008) is the only study to address the issue of human capital externalities in Germany, although not with respect to regional wages. Südekum shows that regional human capital endowments positively influence employment growth of low-skilled workers, a finding which indicates that human capital externalities may arise through complementarities between workers of different skill levels.

3. The central prediction from both models is that ‘unskilled workers benefit from an increase in the share of educated workers in the city even in the absence of any spillovers (γ=0), but the effect on the wage of skilled workers depends on the magnitude of the spillovers’ (Moretti, 2004b, p. 179).

4. Technical colleges (‘Fachhochschulen’) are supposed to be more practically oriented than universities and entry requirements are slightly below those for universities.

5. Possible values are ‘no formal education’ (9 years), ‘degree from Volks-/Haupt-/Realschule and subsequent vocational training’ (13 years), ‘grammar school without vocational training’ (13 years), ‘grammar school with vocational training’ (16 years), ‘degree from a technical college’ (18 years), ‘and university degree’ (20 years).

6. We use prices for sold land of all types, rather than prices for building land only, since the data quality is much better; both types of land prices display a correlation of 0.967.

7. The 10% of workers earning wages above this threshold, which increases from year to year, are free to choose to either pay the maximum amount of social security payments, or to leave the public system and insure privately.

8. In order to keep the panel structure of the data we have drawn a 10% sample of workers and then added information on all available years for these workers.

9. Counties are equivalent to NUTSIII regions and are either made up of a single large city (Kreisfreie Stadt) or of an administrative unit of several smaller cities or towns (Landkreis).

10. Figure 2 indicates that the share of workers holding a degree from a university or a technical college is on average much larger in Eastern Germany than in Western Germany. This is a legacy from socialist times when a large fraction of each generation went to technical schools (‘Fachschulen’) after having completed secondary school; in our dataset, degrees from these schools count as degrees from universities or technical colleges. However, due to a large-scale devaluation of educational degrees in Eastern Germany at the time of reunification, degrees from these technical schools can, in practice, not be regarded as being equivalent to university degrees in Western Germany. Consequently, since the level of individual education serves as our core variable throughout the analysis we omit Eastern German regions altogether so as to avoid inconsistent results.

11. We use a test of seemingly unrelated regressions, because a Hausman (1978) test is prone to under reject the hypothesis of exogeneity when used on clustered data (see Baum et al., 2003).

12. See Kelejian (1971), Newey (1990) and Carroll et al. (2004) on the use of polynomials in first stage regressions.

13. We thank one anonymous referee for emphasizing this important point.

14. These 14,983 observations represent 9% of the observations in the sample; of these, about 7,000 belong to highly qualified and about 8,000 to non-highly qualified workers; hence, similar to findings by Büttner & Rässler (2008), about 42% of highly qualified and 6% of non-highly qualified workers earn wages above the threshold of maximum social security payments.