Testing the Financial Accelerator Using a Smooth Transition Regression Model

Willi Semmler
Department of Economics, New School for Social Research, New York
Semmler@newschool.edu

Levent Kockesen
Department of Economics, New York University

Abstract

A Smooth Transition Regression (STR) model is used to empirically examine the plausibility of the financial accelerator mechanism which predicts nonlinear relations between financial and real variables over the business cycle. We employ a macrodynamic model of financial-real interaction and apply an STR methodology to postwar U.S. times series data for firms and households. The STR methodology finds essential nonlinearities in the times series data for firms. The dynamic properties of the estimated model on firms include regime changes and multiple phase dynamics as the variables pass through certain thresholds in the business cycle. Locally unstable but globally bounded fluctuations (limit cycles) as well as asymmetric responses to shocks are detected.We, however, failed to detect essential nonlinearities in the data for households.