Political Parties and Higher Education Spending: Who Favours Redistribution?

Abstract

A nation's endowment with human capital is an important source of economic prosperity, yet education systems as well as the amount of public spending differ both between and within industrialised countries. Traditional approaches in comparative political economy explain education spending from a perspective in which leftist parties favour human capital formation. In contrast, recent approaches claim that – with regard to public financing of higher education in stratified education systems – the basic assumptions of partisan theory rather lead to the opposite hypothesis. In such systems, a pattern of reverse redistribution emerges, under which electoral incentives let right-wing parties favour increases in higher education budgets. This article tests both claims within the decentralised German education system. Its encompassing empirical strategy provides clear support for the latter hypothesis in the 16 German states between 1992 and 2003. The results imply that the partisan composition of government and preferences for redistribution continue to matter.

Acknowledgements

We are indebted to Prof. Dr. Markus Freitag for support in drawing up the original version of this paper. We benefited further from valuable comments of the anonymous referees and of the convenors and participants of the panel ‘The Comparative Political Economy of Education and Training’ at the ECPR General Conference in September 2009, as well as from feedback received during the MPSA conference in April 2009. Replication data are available upon request. Authors appear in alphabetical order.

Notes

1. Obviously, classical partisan theory assumes class-based voting, an assumption that is challenged in the increasingly differentiated literature on the determinants of individual voting behaviour (for an overview see Evans 2000). But while other factors seem to have gained relevance in recent decades, belonging to a social class remains a relevant predictor for individual party preferences, not least in Germany (e.g. Elff and Rossteutscher 2011; Van der Brug 2010; Pappi and Shikano 2002).

2. We refer to the share of the so-called ‘Grundmittel’ category which captures the expenditure on universities, university hospitals, advanced technical colleges, art schools and the German Research Foundation less any income of these institutions (Bund-Länder Kommission 2006: 7; Wolf 2006: 20).

3. We refrain from a conceptual distinction between the CDU and the Bavarian CSU because of both their programmatic similarities and their close alliance (cf. Lijphart 1999: 69) but also because CSU party shares occur in only one of the 16 investigated units (Oberndorfer and Steiner 2006).

4. We are grateful for a comment by Jon Lauglo who argued that such a demand variable interferes with the political process the partisan variables are intended to capture. While this might be true, not controlling for short-term demand fluctuations entails the risk that government priorities for higher education spending simply follow functional needs and we possibly overstate the importance of partisan differences. In order to provide for a more conservative test of partisan influences, we thus decided to keep this control.

5. Following common practice both economic wealth – measured as the GDP per capita – as well as population size are transformed by their natural logarithm to capture their theoretically non-linear influence.

6. We thank Aline Schniewind for providing access to the socio-economic data as well as Prof. Markus Freitag and Prof. Adrian Vatter for granting access to the data on government composition (Freitag and Vatter 2008). Further we appreciate the support of the German Federal Statistical Office in providing and handling the budgetary data.

7. For estimation, the xtfevd ado (version 2.00) by Thomas Plümper has been used. A current debate questions the advantages of FEVD in large samples, but recognises the advantage ‘in smaller samples where the large sample concept of consistency does not dominate … a trade-off between bias and efficiency in which FEVD often appears to be better than either FE or HT [Hausman Taylor]’ (Breusch et al. 2010: 3).

8. Due to the Prais-Winston transformation, we lose the 16 states in 1992 and the number of cases available for estimation is reduced to 176 in this specification.

9. Multicollinearity is a problem as party competition largely equals a zero-sum game: a large share of the success of one major party can be often explained by the losses of other major parties and vice versa.

10. Detailed robustness results are available from the authors upon request.