The purpose of the study reported here was to investigate how characteristics of analysts affect their forecast errors. Previous research has found positive serial correlation in forecast errors, which can be attributed to underreaction to new information, especially to bad news. The relationship between an analyst's behavior and that analyst's characteristics is not clear, however, because most previous work was based solely on consensus estimates. By using detailed historical data, I found a stronger serial correlation among the herd-to-consensus analysts (that is, the group with a small average distance between their forecasts and the consensus forecast) than among other analysts. Moreover, average distance to consensus itself has a positive serial correlation, and it may be attributed to an analyst's personality (optimistic or pessimistic). I found strong positive serial correlation in the average distance to consensus among the herd-to-consensus analysts. These results show that herd-to-consensus analysts submit earnings estimates that are not only close to the consensus but are also strongly affected by their personalities.