Greens in government: the distributive policies of a culturally progressive force

Abstract

Given the crucial role of Green parties in modern party systems, surprisingly little is known about the impact that Green parties in governments have on distributive issues, on the spending or the revenue side of distributive politics. Providing such insights is the purpose of this study. Based on an encompassing cross-national data set of 34 countries from 1970 to 2015, this article examines the impact of Green government participation on three dimensions of distributive policy making: social consumption, social investment and taxation policies. It is found that the inclusion of Green parties in national governments leads to higher spending on social investment, while the status quo prevails regarding social consumption and taxation. Nonetheless, as procurers of centre-left majorities, Greens in government prevent retrenchment on social consumption and decreasing corporate and top marginal income taxes. This work bears relevance for the study of party politics, the reconfiguration of welfare states and the literature on coalition dynamics.

Keywords:

• Green parties
• government coalition
• social investment
• welfare state
• fiscal policies

Acknowledgements

Previous versions of the manuscript were presented at ‘Political Economy in the Computational Age Workshop’ at the Hertie School of Governance, Berlin in June 2018. We would like to thank the participants of the workshop for the instructive discussion. In addition, we thank Björn Bremer and the reviewers of West European Politics for valuable and constructive comments. We further thank Taiwo A. Ahmed for excellent research assistance.

Disclosure statement

No potential conflict of interest was reported by the authors

Notes

1 Rihoux and Rüdig (2006) emphasise that at least when Green parties enter government, they must take a stance on distributive issues.

2 But see Hamilton (2002) on how environmentalism and nationalism may overlap and Galbreath and Auers (2009) on the specific overlapping of environmentalism and nationalism in Latvia.

3 The ‘Matthew-effect of social investment’ states that social investment policies are to the particular advantage of the middle class that use educational and childcare facilities disproportionally (Bonoli et al. 2017).

4 Included countries: Australia, Austria, Belgium, Bulgaria, Canada, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Luxembourg, Latvia, Lithuania, Luxembourg, the Netherlands, New Zealand, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, Switzerland, and United Kingdom.

5 The classification of housing in the consumption/investment distinction is contested. On the one hand, housing fulfils the ‘buffer’ function of social investment for those in need (see Hemerijck 2017 on the function of social investment). On the other hand, housing can be considered as social compensation, as this passive benefit helps to cover the costs of shelter. We keep this ambiguity in mind when interpreting the results.

6 Data for Bulgaria, Croatia, Cyprus, Latvia, Lithuania, Malta and Romania (not covered by the OECD) is taken from Eurostat (2019).

7 Equally, there is an ongoing debate in the social policy literature about the classification of different family policies (Häusermann 2018; Morel et al. 2012). Parental leave tends to be considered as part of reconciliation-oriented social investment while family allowances are more often classified as passive consumption benefit. As our data does not allow us to disentangle the difference, we keep the ambiguity of spending on families in mind when discussing the results.

8 Ronchi’s approach improves the comparability of education spending data and divides public spending categories on education by target groups. Accordingly, we can interpret the measurements as per capita spending. Alternative measures such as public education spending as a percentage of GDP from the United Nations Educational, Scientific and Cultural Organization (UNESCO) are alternatively used in the online appendix. The latter two data sets have two main drawbacks with regard to our purposes. They are unfortunately very fragmentary (up to 50% missing in our sample) and use classification approaches which are less comparable across countries and time.

9 An alternative explanation to changes in distributive politics is changes in public opinion. The idea here is that the same electoral preferences that lead to Green parties’ government participation affect the positions of other parties as well and therefore affect the distributive agenda of the government directly. Folke (2014) addresses this issue with a Regression Discontinuity Design (RDD) that rules out the influence of public opinion change and establishes the causal effect of government participation of small parties on a number of issues directly. In our large N, cross-national setting, we cannot implement such a design approach. Electoral rules do not change sufficiently frequent in a cross-national perspective to apply a RDD, in particular since we are interested in the executive representation of Green parties where changes in the formal rules are even less frequent.

10 We assume that changing electoral support and ministerial distribution across parties are important for our argument and hence we follow the cabinet definition of Döring and Manow (2019) – cabinet and opposition parties in parliament with their seat strength at each instance of government formation. Note: new cabinets are defined for (I) any change in the set of parties holding cabinet membership; (II) any change in the identity of the prime minister; (III) any general election; (IV) any substantively meaningful resignation.

11 Interestingly, both articles propose a cabinet specification for the exact opposite reason. A precise discussion of under- and overestimation of partisan effects using cabinet periodisation is beyond the aim of this study.

12 However, the procedure fails to balance the second and third moment once we shift the periodisation from country-years to cabinets. The reason is simple: cabinet periods reduce a lot of variance in the controls and thus reduce the chance of exact balance by definition. Accordingly, regression weights are based on balanced means for all models using cabinet periodisation.

13 We specify first-order autocorrelation.

14 All models entail the same controls, only the lagged dependent variable is adjusted to the respective dependent variable.

Additional information

Notes on contributors

Leonce Röth

Leonce Röth is a postdoctoral researcher at the Cologne Center for Comparative Politics, University of Cologne. Major research interests are the measurement and implication of political ideologies as well as the analysis of partisan rationales and dynamics in decentralised political systems. [[email protected]]

Hanna Schwander

Hanna Schwander is Full Professor and Chair of Political Sociology and Social Policy at the Humboldt University, Berlin. Located at the intersection between comparative politics, political sociology and political economy, her research is guided by her interest in how post-industrial transformations of welfare states, labour markets, and societies affect various aspects of political life. [[email protected]]