Abstract

The paper illustrates the following thesis. Time series specific components inherited in econometric models allow to identify shock-error models completely from the observed data. In static models some auto- or crosscorrelation involved in the true exogenous variables is utilized to consistently estimating the model parameters. The appearance of some lag structure in the linear relation of the model allows to weaken or eventually get rid of the requirement for the existence of some autocorrelation within the true regressor variables. This holds too in models with autoregressively correlated errors. The asymptotic distribution of the regression parameter estimator in the unvariate model is shown to be normal. The asymptotic variance of the estimator is small if there is a strong autocorrelation involved in the true regressor.