Why constitutional conventions fail to change the constitutional status quo

Why do voters reject constitutional proposals drafted by conventions they themselves elected? We study a two-stage constitutional process in which voters first elect a legislator to draft a new constitution and subsequently approve or reject it in a referendum. The model highlights how imperfect electoral signals generate asymmetric information between voters and legislators. Its principal novelty lies in reversing the conventional direction of asymmetric information—placing uncertainty on the side of legislators rather than voters. We show that informational distortions are most severe at intermediate levels of voter polarization: when polarization is high, electoral separation allows legislators to infer voter preferences accurately, whereas moderate polarization increases misinterpretation and rejection risk. Moreover, learning from noisy electoral signals can backfire. Legislators who rationally update their beliefs may be less likely to secure approval than those who strategically ignore electoral information. Surprisingly, referenda may fail to improve voter welfare even when all actors prefer reform. These findings offer a new perspective on recent constitutional failures by emphasizing interactions between polarization, information, and strategic learning.