What is Really Good for Long-Term Growth? Lessons from a Binary Classification Tree (BCT) Approach

Abstract

Although the economic growth literature has come a long way since the Solow-Swan model of the fifties, there is still considerable debate on the real' or deep determinants of growth. This paper revisits the question of what is really important for strong long-term growth by using a Binary Classification Tree approach, a nonparametric statistical technique that is not commonly used in the growth literature. A key strength of the method is that it recognizes that a combination of conditions can be instrumental in leading to a particular outcome, in this case strong growth. The paper finds that strong growth is a result of a complex set of interacting factors, rather than a particular set of variables such as institutions or geography, as is often cited in the literature. In particular, geographical luck and a favorable external environment, combined with trade openness and strong human capital are conducive to growth.