Unit Root Testing with Neural Network Nonlinearity: An Application to GDP Per Capita
Abstract
This study investigates the existence of the unit root hypothesis in the GDP per capita of African countries. We apply the nonlinear unit root testing framework of the autoregressive neural network ADF (ANN-ADF) to the panel SUR Dickey-Fuller method. This method combines the distinct properties of individual time series with the wider panel structure, producing a more robust tool for empirical economic analysis. Various empirical applications with GDP per capita across several countries is carried out. Using GDP per capita data from several countries, we discover that most of them do not have unit roots. This implies mean-reverting behaviour and suggesting economic stability in the region. Consequently, result of the panel SUR Dickey-Fuller techniques outperform the
traditional unit root testing methods.
This result does not only contribute to the understanding of economic growth in Africa but also highlight the importance of sophisticated econometric methods in capturing the complexities of real-world data.
Keyword: Unit root, GDP per capita, African countries, Neural network, SUR-ADF