Social psychologists have documented a tendency for people to overestimate their similarity to others. I investigate whether financial analysts' forecast errors are consistent with this bias. I model the bias by assuming analysts overestimate the correlation of the private signals they receive about a firm's future earnings. My model predicts a positive relationship between (i) the likelihood of an analyst's revised forecast being too close to his earlier forecast and (ii) the number of analysts issuing forecasts during the time interval between his two forecasts. I empirically confirm this prediction and consider several alternative explanations.