Vector Autoregressive Models for Multivariate Time Series Analysis; Macroeconomic Indicators in Ghana

Abstract

This study investigated the relationship, the percentage contribution of endogenous shocks and the direction of causality between real gross domestic product, exchange rate, foreign direct investment and unemployment rate in Ghana. It employed the multivariate Johansen co-integration test via vector auto-regressive model and the vector error correction model, to examine both long-run and short-run dynamic relationships respectively, between the selected macroeconomic variables for the period 1991-2016. The dynamic interactions between the variables were studied with Granger causality tests, impulse response functions, and forecast error variance decompositions. Augmented Dickey-Fuller (ADF) test indicated that all the variables were stationary after their first differencing, thus variables are integrated of order one, I (1). The diagnostic tests on the model residuals revealed that the models were adequate, valid and stable. The Trace test statistic of the Johansen cointegration test indicated one cointegrating relationship indicating long run relationship among the variables. Granger Causality analysis indicated a uni–directional causal relationship between real GDP and FDI. It also showed that FDI Granger-causes all of the other variables. The results revealed the positive effect and sensitivity of the FDI variable in determining the activities pertaining to real GDP, exchange rate, and unemployment rate and vice versa in the Ghanaian economy.

Keywords: Vector Autoregression Model (VAR), Multivariate Time Series, Macroeconomic Variables, Cointegration, Granger causality, Impulse response function, Forecast error variance decomposition,