Decentralization and electoral swings

ABSTRACT

This paper explores how the uniformity of electoral swings in the district vote within countries is affected by the level of economic and political decentralization. It relies on aggregate data from 3796 districts in 31 Organisation for Economic Co-operation and Development (OECD) countries in two consecutive national elections before and after the Great Recession to show that the more influential regional policies are for individuals’ well-being, the more uniform are electoral swings across districts. This causal mechanism accounting for the effect of decentralization on dynamic nationalization is examined with Internet panel surveys from national elections in Canada and Spain.

KEYWORDS:

• decentralization
• dynamic nationalization
• electoral swing
• Great Recession

JEL:

• D72
• O57
• R50

ACKNOWLEDGEMENTS

The authors thank Michael Lewis-Beck and the participants in the 3rd International Conference on ‘Decentralization after the Great Recession: Fine-tuning or Paradigm Change?’, Santiago de Compostela, Spain, 26–27 October 2017, for very helpful comments.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. For the same reason that, all else being equal, the lower the mean, the lower the standard deviation (SD) around the mean.

2. In Belgium, the pair of selected elections is 2010 and 2014 instead of the June 2007–June 2010 cycle for two reasons: the instability of government coalitions and the fact that the main opposition party in the 2010 election, the Open Flemish Liberals and Democrats (Open VLD), was the second largest party in the government until April 2010. In Italy, we have selected the April 2006 and April 2008 elections instead of the 2008 and 2013 ones as the incumbent from 2011 to 2013 was a government of technocrats led by Mario Monti. In countries using mixed-member electoral systems, we focus on single-member districts. In France, we use the election results for the first round.

3. When using the Pearson correlation coefficient, the assumption of normality is violated in some countries in the sample. For instance, in Luxembourg, Iceland or Slovenia, the number of districts (i.e., observations) is four, six and eight, respectively. We have run the models using Kendall's tau-a correlation coefficient, a non-parametric alternative to the Pearson correlation coefficient. The results are very similar. This is not surprising, as the correlation between the two measures is 0.97.

4. When comparing the Pearson correlation coefficient with the few available indicators of nationalization using a different method, the results are remarkably similar for the countries included in both samples (Morgenstern et al., 2009).

5. The results are similar when using subnational government expenditure as a percentage of gross domestic product (GDP) because the correlation between the two measures is 0.92.

6. See https://read.oecd-ilibrary.org/governance/oecd-regions-at-a-glance-2016_reg_glance-2016-en#page1.

7. The RAI is a measure of the authority of regional governments across 10 dimensions: institutional depth, policy scope, fiscal autonomy, borrowing autonomy, representation, law-making, executive control, fiscal control, borrowing control and constitutional reform. The first five dimensions capture Self-rule, or the authority a regional government exerts within its territory, while the last five dimensions capture Shared-rule, or the authority a regional government or its representatives exerts in the country as a whole. Country scores aggregate scores for each regional tier and individual regional governments in a country.

8. We tested for other specifications of the electoral system variable and the results are remarkably similar.

9. The results do not change appreciably when fragmentation is replaced by segregation (Alesina & Zhuravskaya, 2011). We decided to use the former measure because it is available for all countries in the sample.

10. We tested for other specifications of the economic crisis, in particular the difference in the unemployment rate between the two election years in every count. The source is the OECD. The results do not change appreciably.

11. Whereas the Pearson correlation coefficient for Slovenia is –0.19, the average for the remaining 30 countries is 0.82 (SD = 0.18). When running the model with the RAI, for instance, the Studentized residual for Slovenia is –4.71, while all the residuals for the remaining observations do not exceed ±1.79.

12. See www.electoraldemocracy.com for more information.

13. The specific questions in the survey are the following: ‘How much influence do the policies of the following governments have on the well-being of you and your family: national government?’ and ‘How much influence do the policies of the following governments have on the well-being of you and your family: regional government?’

14. When running a multinomial probit model, which does not assume independence of irrelevant alternatives (IIA), the results are very similar than when using a logit model.

Additional information

Funding

Ignacio Lago acknowledges financial support from the Ministry of Economy and Competitiveness [grant number CSO2017-85024-C2-1-P1 (AEI/FEDER, UE)] and the ICREA under the ICREA Academia programme.