Caplin, Andrew :: Economic Data Engineering

authors

Caplin, Andrew

url

https://www.nber.org/papers/w29378

Paper about the exciting new frontier of domain-specific ontologies for economic analysis. In the vein of the neuroeconomics project, it centers a "theory-led" approach. I seriously doubt its having engaged with the ungrounded model semantics problem properly, but it does take the trouble to name it, give it a subheading ("Fundamental Identification"), and give me a citation for it. So it might be worth reutrning to, maybe even as a source for precise delineations of viewpoints that I see the semantics problem in.

In formal terms, even if the econometrician observes actions, states, and prizes, both the decision maker’s utility function and the posteriors about the state of the world are subjective (this was highlighted by Machlup, 1993, in dif- ferentiating between the contents of a transmitted message and comprehension of it). The path forward that data engineering suggests is to follow Samuelson’s lead and to conceptualize an ideal data set that allows beliefs and utilities to be separately identified. In fact this is precisely what Block and Marschak, 1959, proposed when introducing stochastic choice data into economic analysis. While modeling randomness in utility, they were intimately aware of psycho- logical models dating back to Thurstone, 1927, 1931, and Luce, 1956, 1958, in which stochasticity in choice derives from the imperfect ability to discriminate between “percepts”. For that reason they were powerfully struck by the identi- fication problem as between beliefs and utility, or “information and desirability” as they phrased it (Block and Marschak, 1959 p. 1.6).