Reducing the Risk of Investing in Stocks

Abstract

Although financial literature presents ambiguous evidence about the predicting value of fundamental and technical variables in stock markets, we find that evolving trading models based on fundamental variables substantially reduce the risk of investing in stocks. This reduction is so generous that the risk-adjusted return obtained following these fundamental variables to trade individual stocks is superior to that obtained by the passive investing in the same individual stocks. However the technical indicators we analyze do not show any predicting value neither in terms of return or risk. We observe the dynamics of individual stock prices’ return and risk in a new framework, the Adaptive Market Hypothesis (AMH) proposed recently by Lo (2004). Using this framework, we examine if there is room to improve investment strategies when adapting them to the potential changing conditions of financial markets or to the investors’ learning process. This adaptation is carried out by quantitative adaptive models driven by evolutionary algorithms (genetic algorithms) that update, over time, the threshold values for fundamental and technical indicators. We find that adaptation improves the risk-adjusted return of investment strategies. We test our trading models using a large sample of companies: non-financial firms with data available in Compustat database which have been listed in the S&P 500 for at least two quarters during the period 1976 - 2006. Our sample consists of 332,700 firm-quarterly observations for fundamental trading systems and 7,157,320 firm-daily observations for technical trading systems. Our models are run using parallel computation executed on 81 computers with a global capacity of 225 GFLOPS or (225x109) FLOPS (floating point operations per second) at the Computational Laboratory of the IT School of Madrid’s Complutense University.