Revised article published in Journal of Money, Credit and Banking vol. 35, no. 4 (August 2003): 557-589.
Reproduced here with permission from the Ohio State University Press.

This paper quantifies the importance of heterogeneity in regional housing markets for the conduct of monetary policy using a new model called a heterogeneous-agent VAR (HAVAR), which generalizes conventional macro VARs. The HAVAR model integrates a national monetary authority and financial market with regional housing markets, imposing exact aggregation. Monetary policy is transmitted to the national to regional markets via the mortgage rate. Although the HAVAR model is based on linear regional VARs, its aggregate impulse responses exhibit two nonlinearities: (1) time variation, stemming from aggregation over heterogeneous regions; and (2) state dependence on initial economic conditions in regions. Thus, the effects of monetary policy on the economy depend on the extent and nature of regional heterogeneity, which vary over time. Using longitudinal data for a subsample of detailed U.S. regions, we estimate the effects of time variation and state dependence on the dynamic responses of the HAVAR model. The estimated model provides plausible and tangible explanations for ?long and variable? lags in monetary policy. To provide a policy-relevant illustration, we show how coastal housing booms influence the efficacy of monetary policy. Revised September 2001.