We test whether forecast bias affects household equity trading. We conduct an incentivized forecasting experiment to measure individual-level forecast bias. Our experiment finds that, on average, subjects exhibit extrapolation bias – a tendency to excessively project recent experiences in the same direction. There is substantial heterogeneity, however, and a large minority instead exhibits contrarian bias. We combine the experimental results with administrative data on the subjects’ stock trading. Forecast bias is strongly related to the past returns of stocks investors select to purchase. A one standard deviation increase in forecast bias is associated with purchasing stocks with 5.7 percentage points higher excess return over the prior year. Forecast bias is not associated with higher performance.