We economists have gotten too good at making theories. In fact, the set of plausible and admissible economic theories is now dense in the space of possible conclusions: For every desired conclusion X, for every ε, there is a degree of theoretical complexity nδ and a chain of theories of increasing complexity T such that:
for all n > nδ -> | T(n) - X | < ε
We have also gotten not too good at but too proficient at econometrics: RCT, regression discontinuity, diff-in-diff, IV, aggressively and preemptively seizing the high ground of the null hypothesis, unconscious and conscious, deliberate and accidental specification searches, plus file-drawer problems. My first econometrics teacher Zvi Griliches said that the CLT was the second most important thing he had to teach. The most important?:
If someone tortures the data enough, they will confess...
Is a theoretical result an interesting constraint on reality or a demonstration of the ingenuity of the researcher? Is an econometric-empirical result a robust finding about the world out there or a demonstration that research assistants desperate to please jet-setting tenured bosses can do amazing things?
This is not to say that things were not worse in the old days. The pasts of many sub-literatures in economics are better understood as careerists on the make taking the line of theoretical least resistance rather than trying to understand the world. The lack of computer power that made it next to impossible to robustly summarize the data meant that arresting anecdotes could not be checked for representativeness and typicality.
But good theory is in the end nothing but distilled and crystallized economic history. It could, after all, be nothing else. And piling more and more computer power on to the analysis of nonhistorical data could only be an intellectual optimum if the world was created ex nihilo the instant of the first date of your panel.
So this is why we are here. Welcome to economic history.
#teachingeconomics #economichistory #highlighted