IJSC

Predictive Autoregressive Models of the Russian Stock Market Using Macroeconomic Variables - Pages 2439-2449

N.G. Bagautdinova, E.I. Kadochnikova and A.N. Bakirova

DOI: https://doi.org/10.6000/1929-4409.2020.09.296

Published: 30 December 2020


Abstract: This article evaluates the relationship of macroeconomic variables of the domestic market with the stock index on the example of the Moscow exchange and selects forecast specifications based on an integrated auto regression model - the moving average. The methods that have been used are included in integrated auto regression-moving average model with exogenous variables and seasonal component, Box&Jenkins approach, auto-arima in R function, Hyndman & Athanasopoulos approach, and maximum likelihood method. The results demonstrate that the inclusion of external regressors in the one-dimensional ARIMAX model improves its predictive characteristics. Time series of macro-indicators of the domestic market – the consumer price index, the index of the output of goods and services for basic activities are not interrelated with the index of the Moscow exchange, with the exception of the dollar exchange rate. The positive correlation between the Moscow exchange index and macro indicators of the world economy - the S&P stock index, the price of Brent oil, was confirmed. In models with minimal AIC, a rare presence of the MA component was found, which shows that the prevailing dependence of the stock market yield on previous values of the yield (AR component) and thus, better predictability of the yield. It has shown that for stock market forecasting, "manual" selection of the ARIMA model type can give better results (minimum AIC and minimum RMSE) than the built-in auto.arima algorithm in R. It is shown that from a practical point of view, when selecting forecast models, the RMSE criterion is more useful for investors, which measures the standard error of the forecast in points of the stock index.

Keywords: Macroeconomics, stock market, autoregression model, forecast error.

Download

Submit to FacebookSubmit to TwitterSubmit to LinkedIn