Investigating Wagner's Law for Nigeria: A Robust Estimate Between ARDL and Engle-Granger Tow-Step Cointegration and Error Correction Models

Abstract

This study titled “Investigating Wagner’s Law for Nigeria: A Robust Estimate Between ARDL and Engle-Granger Two-Step Cointegration and Error Correction Models” examined whether there is empirical evidence that Wagner’s law holds in the Nigeria economy using time series annual data over the period from 1981 to 2015 for Nigeria. Two techniques were applied; the Engle-Granger two-step and Autogressive distributed lag (ARDL). The two approaches were used to test for cointegration and estimation of error correction in order to validate the existence of long-run relationship and short-run dynamics of the variables. In particular, the study keeps a special focus to examine the validity of five versions of Wagner’s hypothesis, which support the existence of long-run relationship between government expenditure and economic growth. Real Gross Domestic Product (RGDP) was used as a proxy for economic growth in the study. Based on the result findings, the co-integration test found a long-run relationship between real gross domestic product and real government expenditure by ARDL techniques for all the five versions considered and only Goffman model for Engle-granger approach. Error correction models showed a desirable negative sign for the estimated coefficients of the error correction terms in the five versions for both techniques. However, the error correction terms estimates were insignificant for Engle-Granger techniques but were all significant with ARDL versions. Diagnostic test on the residuals of the models built using the two techniques revealed that Engle-Grange technique was not void of heteroskedasticity thereby violating one of the important assumptions of the classical linear regression model. Comparing the results of cointegration by both approaches to Johansen cointegration, it was revealed that the cointegrations results by ARDL models were consistent since Johansen cointegration also revealed one cointegrating equation in all the models. Also, the results from Toda-Yamamoto granger causality showed that, the real gross domestic product does granger cause real government expenditure for the five versions of Wager’s law which coincided with the long-run results by ARDL error correction terms for all the versions of the law. Hence, evidence that Wagner’s Law is valid for Nigeria economy. However, based on the result of this study, it is recommended among others that, the use of autoregressive distributed lag (ARDL) model by Statisticians and Econometricians should be encouraged and if Engle-Granger two-step approach will be use, the diagnostic tests should by performed, to see to the fulfillment of all the desirable properties of a good model.