This thesis investigates the approach to predict analyst forecast errors and uses the results to describe the relationship between the predictable components of analyst forecast errors and abnormal returns in markets. Besides, this study introduces a new application of machine learning in finance. I follow the approach of Hughes, Liu, and Su (2008) to predict analyst forecast errors in a first stage and use the pre- dictions in a second stage to set up trading strategies. In the first stage, a neural network outperformed the traditional linear model to predict analyst errors, ques- tioning the assumption of a linear relationship between the errors and the set of predictor variables. The results in the second stage provide evidence that there is a small common error analysts and the market share in forming expectations. By us- ing the neural network predictions, I obtained sightly significant profitable trading strategies on a yearly basis. This result might be due to the inability of both parties to fully reflect recent earnings information leading to an underreaction in expec- tations. However, this common error is overwhelmed by inefficiencies specific to analysts or the market as the significance in returns vanishes for the more extensive monthly time series. On the analyst side, an explanation could be the optimism bias and in general analysts being not as efficient as the market. Moreover, Hong and Stein (1999) deliver with the concept of the "momentum traders" a further driver specific to the market that increases the difference in expectations. Concluding, the results in this study fosters the theory that analysts might not be a good proxy for market expectations
«
This thesis investigates the approach to predict analyst forecast errors and uses the results to describe the relationship between the predictable components of analyst forecast errors and abnormal returns in markets. Besides, this study introduces a new application of machine learning in finance. I follow the approach of Hughes, Liu, and Su (2008) to predict analyst forecast errors in a first stage and use the pre- dictions in a second stage to set up trading strategies. In the first stage, a n...
»