A behavioral approach to efficient portfolio formation

Abstract

This paper investigates the portfolio performance of subjective forecasts given in different forms. In constructing the efficient frontier, we base the expectation formation processes on subjective forecasts and human behavior, rather than on past prices. We construct the efficient portfolios first, using point, interval, and probabilistic forecasts. Next, we compare their performance to portfolios constructed using the standard time series data approach. Subjective forecasts are provided by actual portfolio managers who forecast stock prices on a real-time basis. Our first contribution is to show that the portfolio performance of subjective forecasts is superior to those of standard time series modeling. Our second contribution lies in the fact that we use experts as forecasters, professional fund managers with substantive expertise. Our third contribution is that we investigate the expert subjects' forecasts using point, interval, and probabilistic forecasts, which renders our findings robust to the task format.