Application of modern portfolio theory to the Russian state bond market

The behaviour of the Russian state bond market is analyzed. Attention is mainly paid to short-term fluctuations and efficiency of short-term investments. Analysis of return time series has shown that there exists a significant autocorrelation, and that distribution of random fluctuations is non-Gaussian. It predetermines a choice of forecasting schemes. The most efficient ones appear to be non-linear. The efficiency was checked not only by the traditional statistical indices by direct numerical experiments where various types of predictors were used as basic elements of decision rules. The decision algorithms have included the solution to the modified optimal portfolio problem where the forecasts were used as expected returns and the covariance matrix was estimated via forecasting errors.