Drifting together or falling apart? The empirics of regional economic growth in post-unification Germany

Abstract

The objective of this article is to address the question of convergence across German districts in the first decade after German unification by drawing out and emphasizing some stylized facts of regional per capita income dynamics. We achieve this by employing nonparametric techniques which focus on the evolution of the entire cross-sectional income distribution. In particular, we follow a distributional approach to convergence based on kernel density estimation and implement a number of tests to establish the statistical significance of our findings. This article finds that the relative income distribution appears to be stratifying into a trimodal/bimodal distribution.

Acknowledgements

The article has been presented at the Annual Meeting of the European Economic Association in Vienna (August 2006). We would like to thank the editor and an anonymous referee for helpful comments on an earlier draft. The usual disclaimer applies.

Notes

¹ The obvious difficulty here is to figure out in the data which countries are in the bad and which ones are in the good equilibrium. Barrier to getting out of such a trap can be the lack of a ‘big push’ (Murphy et al., 1989). Rodrik (1996) has argued that the East Asian miracle may have depended on a state-assisted process of overcoming coordination failure, and a consequent shift between two different equilibrium output levels (or a virtuous circle). It is also worth noting that the possibility of nonuniqueness is discussed informally even in Solow's (1956) original exposition of the neoclassical growth model.

² In this article we add to the contributions of Bianchi (1997), Lopez-Bazo et al. (2004), Corrado et al. (2005), Laurini et al. (2005) and Pittau (2005) testing for ‘two-club’ or ‘twin-peak’ convergence of GDP per capita across countries and EU regions by analysing data which do not overlap with the data of existing papers.

³ Funke and Niebuhr (2005) have demonstrated the existence of two clubs across West German regions prior to unification using threshold estimation techniques.

⁴ We focus on district-level data because state-level data tend to ‘aggregate away’ important differences between smaller geographic entities within the 16 states. For example, in the dataset that we analyse below, the ratio of GDP per capita between the richest (Hamburg) and the poorest state (Sachsen-Anhalt) was 2.63 in 2001, while the corresponding ratio for the richest (Landkreis München) and the poorest district (Mittlerer Erzgebirgskreis) was 7.30. On the other hand, one has to be aware that district-level GDP per capita figures may be affected by a commuting bias. Especially, commuters could overstate GDP per capita in agglomerations and city regions. Hamburg and Berlin are classified as a single region. This was forced on us because of lack of district-level data for both states. We also run the kernel estimates excluding Berlin and Hamburg. Qualitatively, results are unchanged and the pattern is not much affected.

⁵ See http://www.statistik-bw.de/Arbeitskreis_VGR/publikationen.asp

⁶ A ‘mode’ is meant here to be a point on the empirical density estimate around which the tangent to the curve changes its slope from positive to negative.

⁷ In particular, the growth rates of the real GDP per capita over the decade from 1992 to 2001 were negative in 66 districts. Out of these 66, seven districts (Delmenhorst, Landkreis Holzminden, Landkreis Sigmaringen, Landkreis Soltau-Fallingborstel, Landkreis Unterallgäu, Neustadt an der Weinstrasse, Wilhelmshaven) have even experienced two-digit negative growth rates. Following Jones (1998, p. 4) these districts might be labelled ‘growth disasters’.

⁸ This test of multimodality has been used by Bianchi (1997) to test the hypothesis of income convergence for a group of 119 countries between the years of 1970 and 1989. Bianchi (1997) rejects the hypothesis of convergence in favour of the formation of convergence clubs.

⁹ Rescaling is necessary since the kernel estimation artificially increases the variance of the estimate (Efron and Tibshirani, 1993). Since the procedure samples from a smooth estimate of the population, it is called smooth bootstrap.

¹⁰ In our simulations we set the number of bootstrap replications to 3000.

¹¹ For the correct interpretation of the maps it is important to bear in mind that they are not suitable to assess the absolute growth performance of the 439 German districts: in particular, it is not possible to say whether over the last decade the poorest areas caught up with the richest ones or whether some areas got richer or poorer as they switched from a cluster to another. (The reason for that is that the thresholds defining the identified categories have changed over time). Looking at those thresholds–which all rose considerably–it is indeed possible to state that the average German GDP has risen between 1992 and 2001.

¹² Data for Mecklenburg-Vorpommern (18 Kreise) is not available for 1992. However, their per capita income level is found to reside within the lowest income category as soon as these figures become available in 1996.

¹³ In 2001 the share of western districts included in the low-income group over the total number of West German districts rose to 18% (58 out of a total of 326 West German districts) from 10% in 1992.

¹⁴ These results are consistent with the empirical evidence in Bayer and Jüßen (2007) and Villaverde and Maza (2008) showing moderate speed of ß-convergence across German and European regions.

¹⁵ The exact nature of multimodality, however, is still surrounded by some degree of uncertainty. At first glance, it might seem promising to consider a growth model with multiple equilibria in the tradition of Drazen and Azariades (1990), Matsuyama (1991), Azariadis (1996) and Aghion and Howitt (1998, Chapter 10) when trying to explain ‘job-poor’ versus ‘job-rich’ growth experiences. In such models, a country may be trapped in a ‘job-poor’ equilibrium when, in principle at least, an alternative and superior equilibrium is also feasible. However, the recent literature has cast doubts on the robustness of multiple equilibria. Frankel and Pauzner (2000) analyse a two sector model with increasing returns, based upon Matsuyama (1991). They show that if the wage is stochastic and arrives as a Poisson process, the muliplicity property may be eliminated because some of the deterministic equilibria are more robust to perturbations than others. A similar conclusion has been established by Herrendorf et al. (2000) for heterogeneous agents. They show that sufficient heterogeneity of agents will lead to a refinement in the set of observable equilibria and uniqueness in models like that of Matsuyama (1991).