On a VAR estimation and its identification problems

Non-restricted Vector Autoregressive Regression models are widely used to obtain information on the economy. Some restrictions, however, should be imposed on the structural parameters for satisfying identification conditions. This paper compares the traditional methods, signifying the role and economic implications of cointegrations proposed by Engel and Granger in identifying structural shocks of VAR models. In recent years, cointegrations have received attentions in terms of nonstationarity, but their theorems can be utilized to find out permanent innovations to the economy. The authors applied these identification conditions to the VAR estimation on the Japanese economy. The variance decompositions found the important relationships with the permanent real shocks and expectations. [In Japanese.]