Multilevel Growth Modeling using R: A Case of GDP influencing Factors

Abstract

Analysis of longitudinal data always remained a point of consideration among the researchers due to its complex nature. The analysis of longitudinal data becomes complex when data collected from the same individuals over the different time intervals. One of the techniques to overcome these types of issues is called Multilevel Growth Curve modeling. This study used a panel data of twenty four different countries comprising the data on two independent variables (Gross National Income and Gross National Expenditure) with one dependent variable (Gross Domestic Product) over the period of twenty four years ranging from 1991 to 2014 which was collected from world development indicators. Three different Unit Root tests i.e. Augmented Dickey-Fuller (ADF) choi Z statistic, Im et al. (1997) and ADF Fisher Chi-square, provided the result that the variables used in this study are all stationary at level-1. Study utilized the approach provided by Johansen panel integration and proved the long run relationships for all the ten cases. Error Correction Model (ECM) does not show any short run relationships of GDP with Gross National Income and Gross National Expenditure. The results of Multilevel Growth models showed that all the fitted models were good as independent variables explained around 98% variation in GDP. The results also exposed that both GLM and LMER provide the same results of the parameter estimated in the models.  Based on the results we concluded that Gross National Income and Gross National Expenditure are the key factors to the growth of Gross Domestic Product of any country over long term.