The Utility of Impulse Response Functions in Regional Analysis: Some Critical Issues

Regional scientists have long been interested in measuring the effects of various external and internal stimuli on a regional economy. Measuring the actual size and timing of exogenous and endogenous impacts has been of special interest, as numerical or estimation techniques allow regional actors (governments, business, and others) to make policy-type probability statements and actions in response to changes to these stimuli. Recently, the use of vector autoregressive (VAR) models and, consequently, impulse response functions has become increasingly popular. This paper will closely examine the VAR methodology and its assumptions and will address the types of empirical issues that arise from actual regional implementation. The issues of stationarity, model specification and selection, order determination, and impulse responses are discussed.