There is a contradiction between theory and empirics with respect to portfolio allocation in parliamentary democracies. While the canonical model of legislative bargaining predicts the existence of a ‘formateur bonus’, empirical studies show that portfolios are allocated in a manner that favours smaller parties. This article argues that the difference between the empirical pattern and the theoretical predictions can be explained by the vote of no confidence, which provides an incentive for large formateur parties to overcompensate smaller coalition partners in exchange for their sustained support over time. This argument is tested by exploiting variations in the presence of no confidence votes across national and regional levels in France. As predicted, we find that larger formateur parties receive a greater share of portfolios if the vote of no confidence is absent than if it is present.