A Geographically Weighted Regression Analysis of General Election Turnout in the Republic of Ireland

Turnout in the 2002 general election confirm e d a downwar d trend in Irish voter pa rticipation levels, that first becam e evident in the early 1980s and was to culm inate in th e lowest turnout fo r an Irish general election since the foundation of the st ate (Figure 1). Given this context, voter turnout is a topic that requires further analysis within the Irish con t ex t, and two ap proaches to address i ng this top i c ex ists, with each approach having its own distin ctive adv a ntages and dis a dvantages. The surveying approach draws on individual-level data, and has the ad vantage of being able to draw on a wide range of variables, including social-psychological variables, an d m a ke direct inferences about the behaviour, and sources of the behaviour, of individuals. However, questionnaire surveys generally te nd to overestim ate turnout levels, with poten tial bias in su rvey re spon se rates resu lting in the under-rep res e ntation of non-voters in survey sam p les. The second type of data – aggregate data – is subject to problem s arising from the lim ited range of variables that can be accessed in seeking explanations of low turnout (u sually lim ited to data that can be drawn from census analyses) and the cross-lev e l infere nce problem that arises in any attem p t to m a ke inferences about individual-leve l behaviour from aggregate-leve l data (Achen and Shively, 1995). However, the big advantage of aggregate data lies in th e accuracy of the estim a tion of the dependent variable (turnout) and in the potential it offers for spatial and censu s-based analysis of variations in the turnout variable. Lower levels of aggregation allow fo r a larger num ber of cases to be analysed and for more detailed pictures of spatial varia tions in turnout leve ls to be gleaned