Electoral campaigns are shaped by non-policy forces before any platform is announced. Candidates observe these conditions not directly but through pre-platform opinion polls, which aggregate the effect of these conditions into a single signal of competitive standing. I develop a spatial model with probabilistic voting in which two policy-motivated candidates observe candidate-specific polling signals before simultaneously choosing platforms. Vote-seeking intensity depends on perceived baseline electoral support, separating the true electoral environment from candidates' perceptions. When signals are accurate, the disadvantaged candidate moderates more than the advantaged candidate, reversing the standard valence competition prediction. When polling information is inaccurate, equilibrium platforms diverge from what the true environment warrants: common misperception inflates or deflates moderation symmetrically, while differential misperception produces platform asymmetry without any underlying electoral asymmetry. A platform gap of identical magnitude can therefore reflect either a genuine structural advantage or a one-sided information failure. These results imply that electoral competition delivers less representational discipline than the standard account assumes.
Zeynep Ozkan (Fri,) studied this question.