Persistent Political Divides, Electoral Volatility and Citizen Involvement: The Freezing Hypothesis in the 2004 European Election

Abstract

This paper identifies possible micro-mechanisms for the operation of Lipset and Rokkan's freezing hypothesis and suggests that their effects do not disappear altogether with the decline of cleavage politics but are sustained by any persistent social or attitudinal divide between the electorates of different parties. A multi-level analysis of survey data from the 2004 European Election Study supports the expectation that political involvement should be consistently higher and volatility lower than otherwise expected among citizens who are predisposed to support particular parties because of their enduring attitudinal and social characteristics. This paper argues that this fact powerfully biases the choices of established parties towards appealing to those citizens who vote in a way that maintains existing political divides among groups in the electorate. This provides a new explanation of why the mobilisation of enduring social and attitudinal divides in the electorate makes party systems reflect past divides even when the conflicts that gave rise to them have lost some or all political relevance, for instance, because of a shift from the national to the European electoral arena. The analysis also provides additional insights into why European elections fail to produce a European party system and why sources of political participation and interest vary across countries.

Acknowledgements

This paper was written while the first author held a Karamanlis Fellowship at the European University Institute in Florence. Comments from two anonymous reviewers and participants at paper presentations at the 2007 ECPR Joint Sessions of Workshops and at the Robert Schuman Centre for Advanced Studies, as well as conversations with Mark Franklin greatly helped in clarifying the argument. However, only the authors are responsible for the views presented and any remaining error.

Notes

1. The data and its technical documentation are publicly available from www.europeanelectionstudies.net. Throughout our analysis, individual respondents are weighted by an adjusted version of the country-specific weight variables deposited with the EES data sets, with the adjustment equalising the weighted sample size for each country.

2. Of the 25 member states of the EU at the time of the 2004 election, Malta did not participate in the EES survey; and the surveys in Belgium, Lithuania, Luxembourg, and Sweden omitted the questions on vote probabilities and thus could not be included in this analysis.

3. For most countries the number of input variables was somewhat smaller because only Italians were asked to evaluate as many as 13 parties and not all the socio-demographic variables were available for all national samples.

4. For this purpose we used the Amelia 2 software of Honaker et al. (2007). The multiple imputation produced five data sets and the parameter estimates presented in our paper are based on averaging the estimates across the five data sets using Rubin's method (cf. King et al. 2001). The dataset used at the imputation stage included the 48 variables plus a group of additional attitude variables that seemed helpful in predicting missing values on the former, and included the q15, q16, q17, q18, q19, q21, q23, q24, q27, q28, q29, q30a, q31, and q32 variables from the publicly available EES data file.

5. We reckon that the full psychological chain of causation from Persistent Pull to political involvement and party loyalty runs – largely though presumably not entirely – through variables like party identification or the total utility differential of the individuals between the parties. The chain between our Persistent Pull variable and the latter must, as simple algebra could show, also run through the standard deviation of the E(Utility_j ) _k terms. However, we could not use the latter as a measure of Persistent Pull because that, just like the strength of party identification, is necessarily influenced – partly but not entirely through the b _0j constants – by such transient factors as momentary popularity differences between the individual parties. These transient factors may well influence involvement and party switching but – unlike any direct or indirect effect of our Persistent Pull variable on the latter – have no consequence for stabilising the packages of commitment, which is what the focus of our analysis is.

6. Technical details of this analysis are available from the authors.

7. Hence the dots in Figure 1 plot the sum of Persistent Pull and a random variable (with a variance of 0.0064) against the sum of Involvement and a random variable (variance = 0.0064); and Figure 2 plots the sum of Persistent Pull and a random variable (variance = 0.0009) against the sum of Volatility and a random variable (variance = 0.0009).

8. We would not be able to fit multi-level models where 25+ variables – including such closely correlated ones as age and age-squared – all have random effects across 20 cases. Allowing the 25 variables to have only fixed effects, in turn, would understate their possible country-specific effects and not control adequately for the possibility that they may be the common cause of Persistent Pull on the one hand, and political involvement/electoral volatility on the other.

9. Re-estimating all fixed-effects models discussed in the paper as random-coefficients models that allow for cross-national variance in the effect of all individual-level variables did not alter the substantive findings reported and consistently confirmed the cross-national invariability of the effects in question.

10. The effect of POLARISATION is in an unexpected direction but this is irrelevant here.

11. The estimates in Figures 3 to 4 were derived from aggregate level analyses using Zelig 3.1 (see Imai et al. 2007), with the national means of Volatility regressed on PERSISTENT_PULL, MONTH and MONTH_SQ, and the national means of Involvement on PERSISTENT_PULL, POLARISATION and EFF_N_OF_PARTIES, respectively.

12. In further analyses available from the authors on request, we find the observed individual-level effects of Persistent Pull appear to run through the strength of party identification. However, as explained in our theory section and note 5, party identification per se can only generate freezing effects as long as it is itself dependent on Persistent Pull. Therefore the results of relevance for our analysis of freezing are those that do not include either this intervening variable or the individuals' total utility differential between the parties in the model.

13. In other words, from a score of –1.2 to 1.2 on the Persistent_Pull variable, on which the observed minimum is –1.32 and the maximum is 3.27.