Parties make policy in government and in parliament, but we do not know what institution matters the most for policy-making. We use an extensive dataset with information on party manifestoes, parties' bargaining power, and redistribution policies for 20 OECD countries to show empirically that parliamentary composition is a better predictor of redistribution policies than cabinet composition when we take parties' bargaining power into account. On the one hand, we use voting power indices in order to calculate parties' bargaining power in parliaments and cabinets. On the other hand, we also use information on parties' preferences using data from the Comparative Manifesto Project (CMP). We estimate dynamic TSCS models based on legislature and cabinet datasets. The empirical results show that parliaments impact policy through the legislative bargaining power that parties obtain after elections. Opposition parties, therefore, also affect policy because they can have or have had the possibility of forming an alternative government, especially in those systems requiring an investiture vote. Hence, we argue that cabinet composition is not enough to understand the determinants of redistribution policies.