The Impact of e-Voting on Turnout: Insights from the Belgian Case R´ egis Dandoy FLACSO, Ecuador rdandoy@ulb.ac.be Abstract—Even in countries where compulsory voting is ap- plied, one can observe large variations in turnout across electoral districts. This paper analyses turnout rates and the number of invalid votes in local elections in Belgium and tests whether differ- ences across municipalities can be explained by e-voting. Taking into account various elements related to the electoral system, political competition and socio-demographic factors, multivariate models demonstrate that turnout decreases in municipalities where e-voting is used. This is true for all election years (1994- 2012) and the paper also finds out that the negative effect of e-voting on turnout increases over time. Index Terms—turnout, e-voting, compusory voting, local elec- tions I. I NTRODUCTION The explanation of turnout has been done in numerous studies and for a variety of local, regional, national and supra- national elections, often in a cross-country comparison. Most of these studies tested the impact of institutional variables and of the electoral system on turnout. Yet, these studies seldom analyse electoral participation within one system [15, p. 368], particularly if this system presents some specific institutional or electoral characteristics. E-voting is one of them. Depending on the political context, the socio-demographic characteristics of the voters and on the type of electronic system used, e- voting can significantly affect turnout. But e-voting is not the only variable related to the electoral system that has proven to have an impact on turnout. Com- pulsory voting leads to a higher turnout and Geys concluded that “the effect of compulsory voting on turnout is one of the robust findings in studies that analyse cross-national variation in voter turnout” [15, p. 652]. This effect has been empirically confirmed for national legislative elections (see for example [2]), for presidential elections (see for example [11]) or for regional elections (see for example [21]). The same applies to the proportional system. Following [1], we know that the system of transformation of votes into seats (i.e. mainly proportional representation vs. majoritarian system) has an important effect on turnout. This has been for example proven by previous studies (refer to [2], [11], [16]). Similarly, voting age has a positive impact on turnout [2], meaning that systems where the minimum age for voting is 21 witness a higher turnout than systems where the minimum age is 18. Belgium is the ideal case if one wants to analyse the impact of e-voting on turnout while controlling for most of the effects of the electoral system (compulsory voting, PR and voting age), i.e. keeping them constant over time and across the territory. In this paper, I analyse turnout for local elections in the Walloon region 1 of Belgium. E-voting does not concern all municipalities in this region. The organization of paper vote in some municipalities and of e-vote in some other allows the comparison of their effects across municipalities while keeping constant most of the aspects of the electoral system. This kind of analysis is not possible in countries where e-voting has been implemented in all municipalities. Students of turnout believe that compulsory voting produces homogeneous participation rate, but this is far from being the case in Belgium. Dewachter and Lismont [8] attempted to explain differences in turnout (as well as in the amount of invalid votes) for all local elections in Belgium between 1919 and 1968. More recently, Geys and Heyndels demonstrated that turnout rates vary considerably between Flemish munici- palities in the local elections of 2000, allowing for an empirical analysis aimed at identifying the determinants of turnout [15]. I observe the same phenomenon for the local elections in the 262 Walloon municipalities between 1994 and 2012. Due to compulsory voting, turnout is high in Belgium and one of the highest in the world 2 . The average turnout for the local election in Wallonia was 91.63% in 1994, 90.8% in 2000, 91.92% in 2006 and 87.77% in 2012. But turnout is not only different across time but also different equal across municipalities. In Graphs 1 and 2, I present the distribution of turnout and of invalid votes across the 262 Walloon municipalities. Turnout varies from 98.05% in Bertogne in 1994 to 80.28% in Li` ege in 2012. Compulsory voting produces relatively high turnout rates that vary significantly across the Belgian territory. This paper aims to explain this variation. The study of local elections presents two main advantages. First, the number of municipalities permits a large N study, which is lacking in most studies of turnout at the national level. For example, in one of largest analysis of turnout in national elections for 91 countries between 1972 and 1995, Blais and Dobrzynska [2] managed to deal with ‘only’ 276 elections. In this paper, I cover local elections in 262 municipalities in 1994, 2000, 2006 and 2012, which makes 1048 elections 3 . 1 There are three main regions in Belgium: the Flemish region (308 municipalities), the Walloon region (262 municipalities) and the Brussels region (19 municipalities). 2 See for example the world rankings of www.idea.int/. 3 There is one missing data (the municipality of Ramillies in 2000). I did not include by-elections that occurred in the municipalities of Jurbise in 2001 and Brugelette in March 2013. 29