Turning Points and Classification

40 Pages Posted: 22 Apr 2019

See all articles by Jeremy Piger

Jeremy Piger

University of Oregon - Department of Economics

Date Written: March 22, 2019

Abstract

Economic time-series data is commonly categorized into a discrete number of persistent regimes. I survey a variety of approaches for real-time prediction of these regimes and the turning points between them, where these predictions are formed in a data-rich environment. I place particular emphasis on supervised machine learning classification techniques that are common to the statistical classification literature, but have only recently begun to be widely used in economics. I also survey Markov-switching models, which are commonly used for unsupervised classification of economic data. The approaches surveyed are computationally feasible when applied to large datasets, and the machine learning algorithms employ regularization and cross-validation to prevent overfitting in the face of many predictors. A subset of the approaches conduct model selection automatically in forming predictions. I present an application to real-time identification of U.S. business cycle turning points based on a wide dataset of 136 macroeconomic and financial time-series.

Keywords: machine learning, nowcasting, business cycles, recession

JEL Classification: E32, E37, C53, C55

Suggested Citation

Piger, Jeremy M., Turning Points and Classification (March 22, 2019). Available at SSRN: https://ssrn.com/abstract=3358631 or http://dx.doi.org/10.2139/ssrn.3358631

Jeremy M. Piger (Contact Author)

University of Oregon - Department of Economics ( email )

Eugene, OR 97403
United States

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
60
Abstract Views
509
rank
431,694
PlumX Metrics