Working Paper, Federal Reserve Bank of Atlanta 99-4
This paper reports the results of 15 experimental asset markets designed to investigate the effect of optimistic forecast bias on market behavior. Each market is organized as a double oral auction in which participants trade a single-period asset with uncertain value. Traders are informed of the asset value distribution and, prior to trading, given the opportunity to acquire a forecast of the asset's period-end value. The degree of forecast bias is manipulated across experimental sessions so that in some sessions the forecast contains a systematic, upward (low or high) bias. We conduct sessions with inexperienced and experienced traders. The results suggest that market prices are supportive of a full revelation unbiased price in the unbiased markets and the experienced, low-bias markets. The results from the low-bias markets indicate that as long as traders have sufficient experience with such forecasts, asset prices reflect the debiased forecasts. In contrast, we find no evidence that high-bias forecasts are reflected in market prices, regardless of experience. We also find that the demand for forecasted information persists over time, but it is greater in the unbiased and low-bias conditions than in the high-bias condition. Finally, we provide little evidence that the net profit (that is, net of the information cost) of informed and uninformed traders differs, regardless of bias condition or experience level.