This book discusses the nature of exogeneity, a central concept in standard econometrics texts, and shows how to test for it through numerous substantive empirical examples from around the world, including the UK, Argentina, Denmark, Finland, and Norway. Part I defines terms and provides the necessary background; Part II contains applications to models of expenditure, money demand, inflation, wages and prices, and exchange rates; and Part III extends various tests of constancy and forecast accuracy, which are central to testing super exogeneity.
About the Series
Advanced Texts in Econometrics is a distinguished and rapidly expanding series in which leading econometricians assess recent developments in such areas as stochastic probability, panel and time series data analysis, modeling, and cointegration. In both hardback and affordable paperback, each volume explains the nature and applicability of a topic in greater depth than possible in introductory textbooks or single journal articles. Each definitive work is formatted to be as accessible and convenient for those who are not familiar with the detailed primary literature.

This overview examines conditions for reliable economic policy analysis based on econometric models, focusing on the econometric concepts of exogeneity, cointegration, causality, and invariance. Weak, strong, and super exogeneity are discussed in general; and these concepts are then applied to the use of econometric models in policy analysis when the variables are cointegrated. Implications follow for model constancy, the Lucas critique, equation inversion, and impulse response analysis. A small money-demand model for the United Kingdom illustrates the main analytical points. This paper then summarizes the other articles in this special section of the Journal of Business and Economic Statistics on "Exogeneity, Cointegration, and Economic Policy Analysis."

Robustness and fragility in Leamer's sense are defined with respect to a particular coefficient over a class of models. This paper shows that inclusion of the data generation process in that class of models is neither necessary nor sufficient for robustness. This result holds even if the properly specified model has well-determined, statistically significant coefficients. The encompassing principle explains how this result can occur. Encompassing also provides a link to a more common-sense notion of robustness, which is still a desirable property empirically; and encompassing clarifies recent discussion on model averaging and the pooling of forecasts.