A History of Polyvalent Structural Parameters: the Case of Instrument Variable Estimators

This paper investigates the rise and fall of the IV method in macro-econometric models and its subsequent revival in micro-econometric models. The key findings are: (i) the IV method implicitly breaks the contemporaneously circular causality postulated in a simultaneousequation model (SEM) by redefining the conditional variable concerned as a suboptimal conditional expectation of it; (ii) the IV method falls out of favour in macro-econometrics mainly because of lack of empirical validations for such redefinitions; (iii) the IV method wins its popularity in micro-econometrics by its capacity to produce multiple suboptimal conditional bexpectations of the latent conditional variables of interest under the disguise of an SEM consistent estimator; nevertheless, (iv) such suboptimal conditional expectations give rise to the insurmountable difficulty of credibly interpreting the IV-based parameter estimates, especially in the case of prognosticated omitted variable bias. The findings highlight the methodological drawback of the estimator-centric strategy of textbook econometrics.