Computer-Automated Model Selection: Friedman and Schwartz Revisited

• Activity Number: 173
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Neil R. Ericsson*+
• Companies: Federal Reserve Board
• Keywords: Autometrics ; dynamic specification ; error correction ; model design ; model selection ; money demand

Abstract

Using annual data from Friedman and Schwartz (1982), Hendry and Ericsson (1991a) developed an empirical model of the demand for broad money in the United Kingdom over 1878-1970. Model selection was a central issue in assessing the merits of Hendry and Ericsson's model, so the current paper re-evaluates that model with Autometrics, a recent third-generation algorithm for computer-automated model selection. Hendry and Ericsson's model is remarkably robust to the model selection path, as characterized through variations in the algorithm's settings for target size, pre-search testing, fixity of regressors, impulse indicator saturation, representation of the general model, and choice of dependent variable. This paper also assesses the empirical merits of Autometrics, using it to improve upon Hendry and Ericsson's model.