The Key Inequality Indicators Forecasting Economic Growth Under Heterogeneity and Nonlinearity: A Machine Learning Approach

A fundamental question in social sciences is whether inequality facilitates or hinders economic growth. Before finding the answer, it is necessary to establish the type of inequality indicators that holds greater significance, while controlling for the heterogeneity of the countries. This research proposes multiple novel approaches utilizing the recent advances in Machine Learning to determine which inequality measure for each group of countries is the key index forecasting growth. A dataset comprising a panel of 150 countries spanning the period 1980 to 2020 has been employed. To account for heterogeneity, clustering and feature importance issues, the K-Means the XGBoost methods are used. The results show that while for a majority of developed and developing countries, wealth inequality is the most influential factor, for a group of pre-communists and underdeveloped, income inequality indicators are more strongly associated with growth. However, wealth inequality has been found to be significant across all groups of countries worldwide. JEL Classification: E01, O15, O47.