Investigating Regression Parameters Stability of Agricultural and Health Sector on Growth Domestic Product of Nigeria, Using Difference and Return Series

Abstract

The international market's dropping in oil price highlights the necessity for Nigeria's economy to be diversified in order to grow the non-oil sector, of which agriculture and health are subsectors. A large portion of the population of Nigeria earns their living through agriculture, which is well-reported in the literature. In this study, the impact of Nigeria's agriculture and health expenditures on the country's GDP is examined from 1999 to 2021. The CBN statistical bulletin was used to extract yearly data for agriculture and GDP. The Nigerian Real Gross Domestic Product was regressed on health and agricultural expenditure from 1999 to 2021. Four regression models were employed in this study (First difference, First difference of the Logarithm of the series, Return series, Return of the Logarithm of the series). Then, each individual series fluctuation were examined to determine the trend component in the series. The time plots of the data revealed the presence of upward trend in the series and suggestion of non- stationrity. The KPSS and Augmented Dickey-Fuller test conducted on the series indicates that actual and first difference series were non-stationary at the critical values of 5%, while the return and return logarithm series were stationary at 5%. Then, the ideal model was selected based on Information Criteria: AIC and SQC. The Return Logarithm series Regression Model was identified as the robust (best) model to examine the relationship between health and agricultural expenditure on GDP, when the transformed series are non-negative which no spurious regression was produced. The identified model revealed that 1% change in agricultural expenditure causes about 71.6% change in GDP and 1% change in health expenditure causes about 27.7% change in GDP. Hence, this result suggested that return series is the better method of making the series to be stationary before fitting regression model when the transformed series are non-negative.

Keywords

References