Nonlinear ARDL Approach and the Housing Market in the U.S.

Abstract

This study investigates the existence of linear cointegration, nonlinear cointegration or no cointegration between house prices and fundamentals in the U.S. states over the period of 1975Q1-2014Q3. I employ Autoregressive Distributed Lag (ARDL) model by Pesaran et al. (2001) to test for linear cointegration and Nonlinear Autoregressive Distributed Lag (NARDL) model by Shin et al. (2014) to test for nonlinear cointegration between house prices and fundamentals. Decomposing fundamentals into positive and negative components in the nonlinear ARDL model allows me to study the nature of impacts of income and/or mortgage rates on house prices. By using these methods (ARDL and NARDL), I can also estimate both short-run and long-run impacts of fundamentals on house prices. Moreover, estimating a bivariate model that captures the sole impact of income on house prices lets me check not only causality but also asymmetric causality from income to house prices. My main findings show that fundamentals have short-run effects on house prices in all states. Moreover, cointegration between house prices, and income and/or mortgage rate exists in 34 states. Investigating the sole impact of income on house prices determines that not only is there a long-run equilibrium relationship between house prices and income but also income Granger causes house prices in 46 states. The Granger causality turns out to be asymmetric in 18 states of the United States.