基于多变量动态模型的产出缺口估算

Estimating Output Gap:A Multivariate Dynamic Model Approach

本文运用多变量动态模型系统下的Beveridge-Nelson分解方法和贝叶斯Gibbs抽样估计,估算了1985年1季度至2008年2季度期间中国的产出缺口,并且与传统的单变量估计方法测算的结果在统计属性和对货币政策调节的预测效果方面进行了比较。实证结果表明,不同产出缺口的统计属性存在差别,并且只有基于多变量系统测算的产出缺口对货币政策具有显著预测效果。这说明多变量模型估计出的产出缺口更全面地考虑了经济产出与其他相关变量的互动效应,含有的信息更为丰富,从而对宏观政策调整具有更重要的参考价值。

This paper constructs a multivariate dynamic model to estimate China output gap over 1985Q1 -2008Q2 by using Bevefidge-Nelson decomposition and Bayesian Gibbs sampling methods. The paper compares statistical nature and forecasting effects on monetary policy of the output gap based on the multivariate system with those of the output gap measures based on univariate models. The empirical results show that different measures of the output gap differ from each other not only in terms of their statistical properties, but also in terms of their forecasting effects on monetary policy. In particular, only the measure based on the multivariate system manifests significant forecasting effects on monetary policy. This result indicates that the output gap estimated by the multivariate system takes into account dynamic interactions between the output and other relevant variables, contains more information, and hence is more important for policy adjustments.