Recent studies have found that politicians misperceive their constituents’ policy preferences, ideology, and socio-demographic characteristics. These studies focus almost exclusively on perceptions of large masses of constituents: the residents in a politician’s electoral district, the supporters of the politician’s political party, or citizens in general. While these are theoretically important groups, politicians themselves often emphasize that they are focused on contact with and information from particular constituents rather than aggregate signals. As a result, politicians may be more accurate when trying to assess the ideology or opinions of individual constituents.

In this paper, we survey Canadian local politicians (N~800) and measure how well they know specific kinds of constituents. Using a novel pre-registered experiment, we ask them to estimate the ideological self-placement and policy preferences of constituents whose specific characteristics vary by age, gender, racial identity, education, immigrant status, duration in the community, home ownership, and party identification. We then measure the accuracy of these perceptions using data from a large survey (N~6,500) of the general public that allows for precise estimates of different types of constituents. This research design allows us to answer three questions on elite perceptual accuracy: (a) whether politicians have more accurate perceptions of some constituents than others; (b) whether politicians have accurate perceptions of differences across demographic groups; and (c) whether politicians’ tendency to overestimate constituents’ conservatism applies consistently across demographic groups.