The Analyst Anomaly: Investor Bias Toward Analyst Estimates Despite Contrarian Alternative Models Forecasts

Abstract

This Master's thesis explores the weight US investors assign to analyst estimates compared to alternative earnings forecasting methods. Despite frequent inaccuracies and bias in analyst forecasts, they're widely used for investment decisions, prompting this study's key question: “How do US investors weight analyst estimates versus alternative forecasting methods, and can predictive strength be improved by combining these methods?” The thesis develops a novel model merging three alternative forecasting methods, demonstrating that analysts' forecast errors can be minimized when adjusted by these methods, particularly in certain situations. Investment strategy tests showed investors tend to overly rely on analysts' estimates, particularly undervaluing forecasts derived from time series industry growth data. The model's utility is demonstrated via a combined z-score model, identifying firms likely to experience predictable earnings surprises relative to the mean consensus. This produced statistically significant risk-adjusted abnormal returns (14.6% FF3 alpha annually), suggesting investors systematically undervalue alternative forecasting methods when diverging from analysts' estimates. The strategy returns are improved even more after applying the model on high NOA firms, firms where analysts’ characteristics making up the mean consensus suggest that the consensus is unlikely to be correct or firms wehre the mean consensus experiences high volatility of estimates. For those firms, analysts’ are even more predictable in terms of their errors, however investors’ fail to adjust for it and underweight alternative models even more. The findings find however that analysts’ predictable errors are deteriorate at the tails of alternative models’ optimism. These in thesis were associated with significant transitions of the firms’ strategy (certain corporate actions), for which findings indicate that weight should be tilted back to analysts’ forecasts. Therefore, investors are less likely to overweight the analysts’ estimates for the firms in such situations. Contrary to the semi-strong form of the efficient market hypothesis, the study finds significant risk-adjusted returns from the model's strategy, indicating the market isn't semi-strong efficient. The thesis offers a new perspective on earnings forecasting, revealing investor biases and providing a more accurate model for future research. This could encourage more balanced use of alternative forecasting models, offering regulators new insights and practitioners a practical tool to counteract optimism in consensus analysts’ earnings predictions. The thesis contributes significantly to existing knowledge, offering a fresh lens for efficient capital allocation through earnings forecasting.