Democratic governments are generally fairly responsive towards their citizens. At the same time, studies demonstrated that they are not always equally responsive to all citizens: some people are better represented than others. One element in explaining this inequality is representatives’ accurate information about what citizens want. Whether using their own background or citizens’ participation as a source to learn about people’s preferences, representatives’ views may be biased exactly because of who they are or who the participants are. Regardless, the outcome would be that representatives might reflect some citizens’ opinions better than others. Here we examine whether representatives’ knowledge about their party voters’ preferences is correct and unbiased. Using data from the first round of the Panel of Elected Representatives in Norway and matching opinion data from the Norwegian citizen panel, we examine whether legislators’ assessments are correct and whether it reflects different groups in society equally. We find that representatives are quite accurate in assessing their voters’ preferences, overall. Moreover, we find that legislators reflect preferences of men better than that of women. Further, representatives’ own preference strongly predicts their accurateness: the more negative they are about an issue, the more likely they are to underestimate support for it.