Abstention is a key issue for any representative democracy. Turnout has a direct impact on the input legitimacy of the democratic system, but also eventually on the output legitimacy, that is on the policies that are designed and implemented. This issue has long sparked debate about the determinants of abstention in elections and its consequences both for politics and for policies. Yet, a multi-level approach is often missing. While the second-order elections theory has shown that different levels of elections mean different levels of turnout, it has been criticized theoretically for its so-called nationalist – i.e. in favor of national elections – bias and methodologically because it typically assesses elections not held on the same day. The last elections held in Belgium offer a fertile ground of investigation for multi-level abstention as the regional, federal and European elections were organized the very same day, under the rule of compulsory voting. This paper aims at presenting and discussing the real abstention rates at the three levels of elections, as well as for the local elections. The 2014 PartiRep voter survey offers an original way to explore the potential abstention – that is electoral participation in the absence of compulsory voting – as the respondents were not asked the typical single question about their potential electoral participation if compulsory voting was to be abolished, but a fourfold question asking them whether they would always, often, sometimes or never vote at the local, regional, federal and European elections. The results will show that there is a significant amount of the voters who would vote differently: 437 of them (22%) would always vote at one – level of – election, but not always at the other ones. The paper seeks to assess this potential multi-level abstention in light of socio-demographic and political (namely party) variables. This contribution will thus shed new light on the issue of abstention – both real and potential – in its multi-level dimension.