Vector Autoregressive Model Estimation Under Mixed Frequency Data: An Estimable Equation Approach (Part 1)

Vector Autoregressive Model Estimation Under Mixed Frequency Data: An Estimable Equation Approach (Part 1)

A consistent estimator of VAR (Vector Autoregressive) model is suggested with mixed frequency data, which may be practically useful. It is possible by an ‘estimable equation’ derived from the VAR model, of which coefficients are functions of autoregressive coefficients. In this equation, the dependent and explanatory variables of this equation are all observable with finite order. However, the explanatory variables and error terms of moving average type are correlated with each other. Thus the autoregressive coefficients may be estimated by the GMM (generalized method of moments) using lagged variables as instruments. The GMM estimators are superconsistent under the model nonstationarity and the identification condition is suggested which may be readily checked from the data structure. The Monte Carlo experiments showed the superior performance from other methods especially when the model becomes near nonstationary.