Abstract
Since the transition to democracy in the early 1990s, more than 60 per cent of governments in Central and Eastern Europe have terminated prematurely. This article argues that the character of party system development in the region has facilitated the emergence of a polarised pattern of party competition and that competition for government now takes place in distinct ideological blocs. Parties seek to form governments within these blocs but not across them and therefore there is little incentive to defect from a governing coalition due to the lack of viable alternatives. As a result, more polarised party systems produce more durable governments. The empirical evidence shows that polarisation and ideological diversity of the government are significant indicators of government duration in Central and Eastern Europe. Ideologically compact governments formed within narrow blocs in the party system survive longer than ideologically diverse coalitions that emerge from less polarised party systems.
Acknowledgements
I would like to thank the anonymous reviewers for West European Politics for their helpful comments on a previous version of this article. This research was funded by the Economic and Social Research Council grant number ES/J003425/1.
Notes
1. Based on data from Conrad and Golder (2010), excludes Croatia.
2. Author’s calculations based on data from Austria, Belgium, Denmark, France, Germany, Portugal, Spain, Sweden for all elections to the parliament or lower chamber in each country between 1990 and 2007. Election data accessed via the PARLINE database http://www.ipu.org/parline-e/parlinesearch.asp
3. It should be reiterated that this article covers the period from 1990 to 2006. The global financial crisis which begun in 2008 may well have influenced domestic politics in many East European countries and contributed to a renewal of party system instability. Recent elections (2010–12) in the Czech Republic, Hungary, Poland, Slovakia and Slovenia have seen a dramatic drop in support for parties that have been mainstays of the political scene in their respective countries together with the emergence of new parties that have gained significant parliamentary representation. HZDS in Slovakia, KDU–CSL in the Czech Republic, LDS in Slovenia are all now unrepresented in parliament while the SDL in Poland and MSZP in Hungary suffered a haemorrhaging of electoral support. These developments are not relevant to the data used in this paper but will be for future studies. See Haughton et al. (2011), Krašovec and Haughton (2012), Szczerbiak (2011) and Batory (2010) for analysis of developments following recent elections.
4. In total eight former communist countries in Central and Eastern Europe were considered to be consolidated democracies, the five chosen countries plus Estonia, Latvia, Lithuania. Romania, Bulgaria and the Balkan states were not considered to be consolidated democracies.
5. See Alvarez and Nagler (2004), Dalton (2008) and Maoz and Somer-Topcu (2010) for a discussion of alternative methods of measuring party system polarisation.
6. The inclusion of the ideological diversity of the government as an independent variable in the models presented in this paper leads me to omit a measure of the effective number of parties (ENP) in parliament. Warwick (1995) argues that a greater ENP in parliament necessitates the creation of more ideologically diverse governments and therefore the ENP is directly related to ideological diversity of the government. Indeed, the data used in this analysis show that the ideological diversity of the government and the ENP in parliament are correlated at the p = 0.01 level. To improve model specification only the ideological diversity of the government is included in the models used in this paper.
7. Due to missing data, in some cases monthly unemployment data were generated by linear interpolation of adjacent quarters. GDP per capita is reported annually. In this case, monthly data were generated by cubic spline interpolation from data for adjacent years.
8. Glasgow et al. (2011) acknowledge this in their recent study of government formation in CEE.
9. It should be emphasised that this paper is not a critique of the validity of the CMP data or its coding and collection procedures. The CMP data is a valuable resource for political scientists and has been used in many empirical studies over the years. My concern here is that the standard Left–Right dimension in the CMP data, which is constructed after coding has taken place, is not ideally suited to identifying the ideological positions of political parties in many CEE countries. This is not surprising as it is a variable that was constructed to identify the Left–Right position of parties in the original CMP data which did not cover CEE.
10. A copy of the survey instrument can be found in Savage (2012) and the online appendix to this article (see note 13).
11. Personal contacts of colleagues in the Department of Government at the University of Essex were also drawn upon extensively when compiling the list of potential respondents.
12. The response to the survey may have been hindered at the time by the overlap with the Benoit and Laver and Chapel Hill surveys. Many respondents may have completed one or the other before also being asked to complete a similar survey.
13. The online appendix can be downloaded from http://leesavage.weebly.com/uploads/1/2/0/0/12002471/wep_party_system_polarization_and_government_duration_online_appendix.pdf
14. The policy dimensions included in this analysis are those that Benoit and Laver (2006) found to be most closely associated with Left–Right positions in CEE in their expert survey. These are taxation and public ownership in the Czech Republic and Slovakia and social policy and religion in Hungary, Poland, and Slovenia.
15. This Left–Right dimension has been rescaled to a range from 0 to 10 with 0 equating to the most distant point on the Left and 10 equating to the most distant point on the Right.
16. A more detailed discussion of the Left–Right positions of political parties in CEE, and a further discussion and validation of this expert survey can be found in Savage (2012).
17. The hazard ratio offers a more intuitive interpretation of the PH model as it is similar to the odds ratio produced by logit models. If a covariate is judged to have absolutely no effect on the risk of termination in a specific month then the hazard is one and all hazards are judged against this baseline of no effect. If a hazard ratio is 0.75 then this indicates that that covariate reduces the risk of termination by 25 per cent, while a hazard ratio of 1.65 indicates a 65 per cent increased risk of termination.
18. Such a government would still have been a minority administration but would most likely have been tolerated due to the lack of any other feasible coalition.