Comparison of Univariate and Bivariate Time Series Forecasts of Nigerian Stock Exchange variables

Abstract

The study examined univariate and bivariate time series forecast of Nigerian stock exchange variables: All Share Index (ASI) and the External Reserves (ER) which comprise of monthly value from 1985 to 2018 of them both. It filled the lacuna by explicitly modeling and forecasting stock returns in Nigeria using the univariate ARIMA and bivariate VAR models. The monthly and yearly means plots were done, to have a better understanding of the series behaviours. The order of the regular autoregressive and moving avenge model that is necessary to adequately represent the time series model was determined. The series plots showed that ASI series is integrated of order 1 without seasonality while ER series is integrated of order 1 with the seasonality of order 12. A suitable ARIMA and VAR Model were obtained for both series using model selection criteria (MSC) and the models were used to generate forecasts. The univariate and bivariate model forecasts were compared and the result shows that the bivariate model is better to predict the two series than the univariate model from the result of forecast accuracy measures (i.e. MAPE and MSE).