Recession Forecasting Using Bayesian Classification

39 Pages Posted: 15 Aug 2016 Last revised: 15 Sep 2016

Troy Davig

Federal Reserve Bank of Kansas City

Aaron Smalter Hall

Federal Reserve Bank of Kansas City

Date Written: September 7, 2016

Abstract

We demonstrate the use of a Naive Bayes model as a recession forecasting tool. The approach has a close connection to Markov-switching models and logistic regression but also important differences. In contrast to Markov-switching models, Naive Bayes treats National Bureau of Economic Research business cycle turning points as data rather than hidden states to be inferred by the model. Although Naive Bayes and logistic regression are asymptotically equivalent under certain distributional assumptions, the assumptions do not hold for business cycle data. As a result, Naive Bayes has a larger asymptotic error rate, but converges to the error rate faster than logistic regression, resulting in more accurate recession forecasts with limited data. We show Naive Bayes consistently outperforms logistic regression and the Survey of Professional Forecasters for real-time recession forecasting up to 12 months in advance. These results hold under standard error measures, and also under a novel measure that varies the penalty on false signals depending on when they occur within a cycle. A false signal in the middle of an expansion, for example, is penalized more heavily than one occurring close to a turning point.

JEL Classification: C11, C5, E32, E37

Suggested Citation

Davig, Troy and Smalter Hall, Aaron, Recession Forecasting Using Bayesian Classification (September 7, 2016). Federal Reserve Bank of Kansas City Working Paper No. 16-06. Available at SSRN: https://ssrn.com/abstract=2821968

Troy Davig (Contact Author)

Federal Reserve Bank of Kansas City ( email )

1 Memorial Dr.
Kansas City, MO 64198
United States

Aaron Smalter Hall

Federal Reserve Bank of Kansas City ( email )

1 Memorial Dr.
Kansas City, MO 64198
United States

Paper statistics

Downloads
83
Rank
241,506
Abstract Views
279