Detection and correction of heteroscedasticity and its effect on modelling of Nigerian economic data

Abstract

The study centred on detection and correction of heteroscedasticity and its effect on modelling of economic data. Data were collected from CBN Statistical bulletin on five economic variables namely; Gross Domestic Product (GDP), Inflation Rate, Exchange Rate (Exch_rate), Balance of Payment and Government Debt from 1987 - 2017. The data was analysed using the Ordinary Least Square method and variance stability techniques was used to correct heteroscedasticity using log transformation of the dependent variable while Breusch-Pagan, White and HarveyGodfrey tests were used to detect the presence of heteroscedasticity in the variables at 5% level of significance with the aid of E-views 9. The regression model fitted to the data set was GDP = 37242.63 - 256.7745 Inflr – 40.196 Exchr + 252.364 BoP + 3.147 GovD and it showed that only inflation rate and Government debt were discovered to be significant predictors (p < 0.05) while and heteroscedasticity was detected. Result showed that after log transforming the variables, heteroscedasticity was eliminated as shown by Breusch-Pagan, White and HarveyGodfrey test respectively. It is concluded therefore that transforming economic data helps correct and eliminate detected heteroscedasticity in modelled economic variables.