Regression and causation: a critical examination of six econometrics textbooks
Published on Real-World Economics Review RWER, issue no 65, by Bryant Chen and Judea Pearl, September 27, 2013.
Abstract:
This report surveys six influential econometric textbooks in terms of their mathematical treatment of causal concepts. It highlights conceptual and notational differences among the authors and points to areas where they deviate significantly from modern standards of causal analysis. We find that econonometric textbooks vary from complete denial to partial acceptance of the causal content of econometric equations and, uniformly, fail to provide coherent mathematical notation that istinguishes causal from statistical concepts. This survey also provides a panoramic view of the state of causal thinking in econometric education which, to the best of our knowledge, has not been surveyed before … //
… 5. Conclusion:
The surveyed econometrics textbooks range from acknowledging the causal content of the SEM (e.g.Wooldridge, Stock and Watson) to insisting that it is nothing more than a compact representation of a joint distribution (e.g. Ruud). The rest fall somewhere in the middle, attempting to provide the model with power to answer economic questions but unwilling to accept its causal nature; the result is ambiguity and confusion. Nowhere is this more evident than in the text by Hill, Griffiths, and Lim in which definitions of the model parameters conflict with stated assumptions of the model. Other textbooks (e.g. Greene) are more careful about avoiding contradictions but their refusal to acknowledge the causal content of the model results in ambiguous descriptions like “autonomous variation”. Finally, even textbooks that acknowledge the role of causality in econometrics fail to provide coherent mathematical notation for causal expressions, luring them into occasional pitfalls (e.g. equating ß with a regression coefficient or some other property of the joint distribution of X and Y) and preventing them from presenting the full power of structural equation models.
The introduction of graphical models and distinct causal notation into elementary econometric textbooks has the potential of revitalizing economics education and bringing next generation economists to par with modern methodologies of modeling and inference … //
… Appendix A and B, Definitions, etc.
(full long text).
Download the whole issue No. 65, 27 September 2013.
See also the Real-World Economics Review Blog.