Predicting stock market responses by debiasing the financial analyst consensus

Few researchers have explored the use of crowd aggregation techniques for aggregating financial analyst forecasts. Therefore, this thesis investigates a relatively new technique in the financial analyst environment: The Contribution Weighted Model (CWM), proposed by Budescu and Chen (2014). It is common knowledge that the average of all analyst forecasts, the analyst consensus, is often biased. This model attempts to mitigate the effect of biases by identifying expertise amongst the analysts and assigning weights to forecasts based on a total contribution score. The question ‘To what extent is it possible to more accurately forecast earnings than the analyst consensus using the contribution weighted model of Budescu and Chen (2014) and to use the difference to predict stock market responses?’ will be answered. This research extracts forecast data from the Institutional Brokers’ Estimate System (I/B/E/S) for analyst forecasts in the sample period of January 2011 to December 2017. Firstly, accuracy of the CWM is compared to the analyst consensus and adjusted versions of the CWM. Results of statistical analyses indicate the CWM is more accurate than the analyst consensus, but not more accurate than the adjusted versions. Secondly, the difference between the CWM and the analyst consensus is used to proxy for earnings surprises. Accuracy of the CWM does not seem to be sufficiently high to function as a relevant proxy, as no relation seem to exist between predicted earnings surprises and actual earnings surprises using the full sample. Thirdly, there does not seem to be a relation between predicted earnings surprises and stock market responses. Also, imposing multiple benchmarks does not seem to have an effect on the predictability of market responses. Further research could focus on refining the input of the CWM and investigating the drivers of market responses.