Party-switching Between Elections is Influenced by Polarization, not the Number of Parties
Yves Dejaeghere and Ruth Dassonneville
In 1979 Mogens Pedersen published a seminal paper in which he analyzed the impact of party-system variables on inter-electoral volatility. By means of an analysis of aggregate data from over 100 elections, he concluded that the number of parties increases volatility. Although Pedersen tested his hypothesis with the best data available at the time, they were actually not ideal to do so, as he indicated himself that his hypotheses implied ‘a test on the basis of individual level data’ (Pedersen, 1979: 16).
Switching parties between two elections is clearly something that needs to be investigated at the individual level. Using raw election results does not allow the researcher to know how many people actually switched from one party to another or from abstention to voting. As a result, the ideal data to investigate what party-system variables drive party-switching (and not turnout or the vote choice of new voters) are comparative individual-level survey data.
Only recently, such data have become available, as a recall question asking the respondents their previous electoral choice was added to the Comparative Study of Electoral Systems (CSES) questionnaire as of the second wave. Even though we have to take into account the limitations of recall data, they allow operationalizing vote switching at an individual level. Making use of such an operationalization, we can investigate what elements of the party-system influence the probability that individual voters switch parties between elections? We investigate this by means of an analysis of almost 30,000 voters in 33 elections.
Our paper, A comparative investigation into the effects of party-system variables on party switching using individual-level data, published in Party Politics, has just two central hypotheses. First, that voters will switch more if they are offered a larger choice set. We hence predict context with more ‘effective’ parties to incite party switching. The second hypothesis relates to the ideological diversity in party choice. If parties are very far from each other in ideological terms, switching from one party to another entails a rather large ideological switch, which seems less likely. If on the contrary multiple choice options are ideologically closeby, switching should become a more likely option. The second hypothesis in the paper is therefore that a higher level of polarization will decrease the probability of party switching.
While analyzing the impact of these party-system variables, we also take into consideration important individual-level predictors of switching. Most importantly we control for political sophistication, disaffection with politics, gender, level of education and age, which have all been linked to volatility or party-switching in previous research.
In a first analysis we replicated the aggregate level analysis of the Pedersen paper for the 33 elections in our CSES sample. Our conclusion is the same as what Pedersen found in his 1979 paper: the number of parties influences volatility, while polarization is not related significantly to volatility. Using the individual-level CSES data in a subsequent multilevel analysis, however, we find that polarization (not the number of parties) is the most important factor influencing party-switching.
What could be the explanation for these contrasting results? We argue that the difference is related to the level of analysis of both sets of models. When looking at volatility, measured at the aggregate level, this measure includes both people going from non-voting to voting (or vice versa) as well as party switching. When looking at party-switching at an individual level, by contrast, we only take into account the behaviour of the second group. These groups might be very different, explaining differences in outcomes of the analysis. We confirm this expectation by means of an additional individual-level analysis in which we included voters that went from non-voting to voting as ‘switchers’ as well. In line with the aggregate level analysis, these results indicate that volatility is higher as the number of parties increases.
In summary, our paper shows that the party-system variable influencing volatility broadly conceived is the number of parties, as Pedersen has shown. By contrast, when we investigate what causes individual voters to switch parties, the conclusion has to be this is a consequence of polarization, and not the mere number of options available to voters.
Yves Dejaeghere is senior researcher and lecturer at the University of Antwerp and is affiliated with the Centre for Citizenship and Democracy at the KU Leuven as a lecturer on political communication and social movements.
Ruth Dassonneville is an Assistant Professor at the Département de science politique of the Université de Montréal.