Perceptions of peer beliefs about gender norms are shaped through social learning, and inaccuracies in these perceptions can reinforce inequitable norms and impede normative change. Prior studies suggest that individuals, particularly men, tend to overestimate peer support for gender inequality, and that correcting these misperceptions may increase support for more egalitarian norms. However, existing evidence is often drawn from restrictive settings or focuses narrowly on domains such as labor and political participation. In this study, we examine norm misperception in the United States using a nationally balanced sample and an 18-item measure capturing perceived and actual peer agreement across multiple domains, including family, education, violence, and leadership. We find consistent evidence of overestimation, with the largest discrepancies occurring for statements framed in traditionalist terms. Contrary to prior findings, more educated participants were more likely to overestimate peer support, while older participants were more accurate. To explore how these perceptions evolve, we developed an agent-based model in which agents observe peers, interact, and update their beliefs using distinct learning strategies. Simulations show that overestimation tends to persist without external intervention, and that the accuracy of learning varies by strategy. While prestige-based learners exhibit lower overestimation under certain conditions, this advantage is inconsistent, highlighting the limits of education as a reliable proxy for normative insight. Together, these findings underscore the persistence of norm misperception in egalitarian contexts and point to the need for targeted interventions, improved measurement practices, and deeper integration between social learning theory and empirical research on gender norms.