VAR的误差分析.pdfVIP

POLICY INFERENCE USING VAR MODELS R. W. HAFER and RICHARD G. SHEEHAN* There has been relatively little systematic investigation of the sensitivity of policy inferences derivedfrom VAR models to changes in the lag structure. W e investigate this issue using a simple macro model consisting of output, prices, money and interest rates. Using six different lag length selection criteria that vary the bias-eficiency tradeoff, we compare the policy inferences derived jivm the different estimations of our VARmodel. Tlze evidence shows that policy recommendations are quite sensitive to changes in the lag structure. 1. INTRODUCTION borders.2 Although debate over the al- Macroeconomics has a long tradition of leged atheoretical nature of VARs contin- examining the response of macroeconomic ues-for example, see Cooley and LeRoy variables to shocks in alternative policy 119851, Learner [1985] and Braun and measures. Much recent work in this area Mittnik [1985]-VAR models are fre- has been based on the empirical regulari- quently used for policy analysis. ties between variables found in vector au- How the dynamic specification of a toregressive (VAR) models. The VAR re- VAR model is generated and how the dy- sults in turn have been used to investigate namics may influence the resulting policy the impact of changes in different policy recommendation is the subject of this variables. VAR modeling procedures have study. While modeling t