Previous comparative research into the determinants of voting using aggregate data has suffered from two limitations: it relied predominantly on country-level data; and it seldom ventured beyond a consideration of one or two types of elections. In order to overcome these shortcomings, we use an original dataset in which data are aggregated to sub-national units; and include examples of national, sub-national, and supra-national elections. A total of 66 elections between 1995 and 2008 are included, drawn from ten Western European countries: Belgium, England, Finland, France, Greece, Ireland, Italy, the Netherlands, Spain, and Sweden. For each country, the same sub-national geographical units are used for all election types, allowing the direct comparison of the effects of our selected institutional and socio-demographic variables. We find that the effects of the institutional determinants of voting are substantially and systematically reduced as the salience of the election type increases. For the socio-demographic variables, no such systematic relationship with salience is found. However, for some variables, the direction of effect is the opposite for European Parliament elections to that found for Municipal and Lower House elections, and supports the idea that EP elections differ sufficiently from sub-national, second-order elections to justify their ‘third-order’ classification. When we turn our attention to the effects of the socio-demographic variables in five individual countries, we find that the results are often consistent across different types of elections, and for all five countries. However, we also find that the effects of some variables have different effects in different countries. In these cases, we suggest explanations which relate turnout differences to wider political and social factors.