An Aggregation Method for Stock Market Recommendations

In this thesis a research is conducted to and a good aggregation method for stock market advices. This aggregation method is an improvement of the original aggregation method developed in de Market Advices Aggregation System by (Stibbe, 2007). The improvement lays in the fact that an under- writer is assigned an adjusted weight, which is based on the experience of the underwriter (amount of advices issued) and the smoothed performance figure. In addition, the method is extended with a bias correction method, which tries to correct the aggregation method by correcting for affiliated un- derwriters. The correction method gives different weights to underwriters who have done the Initial Public Offering for the stock which advices are aggregated. Two types of experiments are conducted: experiments in which the bias of underwriters is tested and experiments in which the improved aggregation method is tested against the original method. Experiments are executed and results are collected over a period of 48 months, between January 2003 and December 2006. The returns of the aggregated advices are calculated for a period of one day, one week, one month and half a year. According to the results, the improved aggregation method performs signifficantly better on the short run, while on the large run, there is no signifficant improvement observed anymore. In regards to the correction method for bias, the results are adjusted for the market capitalization of the stocks (small-cap, mid-cap, large-cap). In all cases, underwriters who have done the Initial public Offering add informational value to the aggregation method and should not be given less weight. However, at large cap companies on the short run, some extreme positive differences showed up when excluding the underwriters who have done the IPO for the stock.