APPLICATION OF THE LONG MEMORY MODELS FOR RETURNS FORECASTING IN THE FORMATION OF INVESTMENT PORTFOLIOS

Abstract

This paper continues research within the framework of the scientific direction in econophysics at the Department of Information Systems and Mathematical Methods in Economics of the faculty of Economics of PSU. This paper describes the method of formation investment portfolios of four assets based on forecasted returns obtained using long memory econometric models and tests the hypotheses that the optimization of portfolio structure by forecasted returns obtained using such models (by the example of ARFIMA) allows to improve portfolio characteristics in comparison to the optimization by historical returns. Different variants of portfolios of four financial instruments were formed to test the method and test the hypotheses. The study obtained the following results. Portfolio parameters do not deteriorate, on average, when optimized by forecasted data and, in some cases, they improve because the optimizer identifies the most profitable assets more often and gives them more weight. The optimizer is better at identifying the most profitable assets based on the forecasted returns than the least risky ones because the autoregressive models predict the trend of the index rather than its volatility. Finally in the paper there are formulated the possible directions for further research: improving the methodology, namely, performing preliminary fractal analysis of series, imposing stricter restrictions on risk, using other forecasting models, rebalancing the portfolio; conducting research on data from the U.S. stock market, which is certainly more developed in comparison with Russia; using stock indices as a benchmark for assessing the effectiveness of portfolios.