In order for electoral accountability to work, citizens must attribute policy successes and failures to the officeholder that was responsible for the policy. While scholars understand that complex institutions and citizens’ cognitive limitations can endanger accurate attributions, the interaction of these challenges is not well understood. When moving from simple to complex institutions, different psychological perspectives make different predictions about citizens’ attribution: heuristic decision-making would suggest that citizens’ beliefs become random but remain unbiased while directional motivated reasoning would predict that citizens’ beliefs decrease in variance but become biased. I test and disentangle these predictions using a 2-by-2 design where I vary the extent to which policy-making is complex—as operationalized by the possibility of a veto by a third party—, and the existence of identity-based incentives to engage in motivated reasoning.