文件名称：Machine-learning Techniques in Economics_New Tools for Predicting Economic Grow
文件大小：2.04MB
文件格式：PDF
更新时间：2021-01-18 14:44:32
Machine learning
In this book, we develop a Machine Learning framework to predict economic growth and the likelihood of recessions. In such a framework, different algorithms are trained to identify an internally validated set of correlates of a particular target within a training sample. These algorithms are then validated in a test sample. Why does this matter for predicting growth and business cycles, or for predicting other economic phenomena? In the rest of this chapter, we discuss how Machine Learning methodologies are useful to economics in general, and to predicting growth and recessions in particular. In fact, the social sciences are increasingly using these techniques for precisely the reasons we outline. While Machine Learn- ing itself is not a new idea, advances in computing technology combined with a recognition of its applicability to economic questions make it a new tool for economists (Varian 2014). Machine Learning techniques present easily interpret- able results particularly helpful to policy makers in ways not possible with the standard sophisticated econometric techniques. Moreover, these methodologies come with powerful validation criteria that give both researchers and policy makers a nuanced sense of confidence in understanding economic phenomenon. As far as we know, such an undertaking has not been attempted as comprehen- sively as here. Thus, we present a new path for future researchers interested in using these techniques. Our findings should be interesting to readers who simply want to know the power and limitations of the Machine Learning framework. They should also be useful in that our techniques highlight what we do know about growth and recessions, what we need to know, and how much of this knowledge is dependable. Our starting point is Xavier Sala-i-Martin’s (1997) paper wherein he summarizes an extensive literature on economic growth by choosing theoretically and empiri- cally ordained covariates of economic growth. He identifies a robust correlation between economic growth and certain variables, and divides these “universal” correlates into nine categories. These categories are as follows: