Political participation is socially distorted. Groups with high socio-structural resources participate in politics more strongly than groups with low socio-structural resources. This phenomenon of social stratification has been proven above all for elections. But recent research has shown that socio-structurally biased participation also becomes obvious when studying direct democracy (referenda). When comparing these biases across a large number of local referenda in Germany, however, we find remarkable differences in the degree of social stratification. The question we want to answer in this paper therefore is, how this variance can be explained and how unequal participation might be reduced. Based on data for 64 referenda that took place in German cities (> 100.000 inh.) between 2000 and 2019, prior research has also shown that increasing turnout does not automatically decrease social bias. After discussing different hypotheses on macro factors that might affect the degree ofunequal participation , we will test whether social stratification can be reduced when referenda are initiated top-down as political parties then are more active in mobilizing citizens to cast their ballots.