Thread by @rohlamba

1/ Inverse Selection, new paper with @MarkusEconomist and Carlos Segura-Rodriguez.

With the advent of Big Data, AI and Machine Learning, should we think about standard screening models, eg. used for insurance or credit markets, differently?

papers.ssrn.com/sol3/papers.cf…

2/ Most models in information economics assume that customers have an informational advantage. Hence, the principal, e.g. the insurance company, faces an adverse selection problem, which it tries to mitigate by offering a menu of screening contracts to potential customers.

3/ While customers might still have private information about some of their characteristics, big data allows insurance companies to develop superior aggregate information, using new statistical tools to better infer correlates about the characteristics and the ultimate risk.

4/ In other words, the principal here can “invert” the mapping from characteristics to risks through an informational and technical advantage. Thus, big data and AI transform many adverse selection problems to what we call “inverse selection” problems.

5/ This paper studies a new version of the classical Rothschild and Stiglitz [1976] insurance model where the insurer (principal) has access to some statistical info that allows it to better predict the eventual risk faced by the agent once the agent reveals her private info.

6/ 3 types of results are presented, described in abstract above. At high level, they speak to rise of data brokers, the perils of market concentration with advent of big data, importance of consumer activism & their understanding of info, & merits of a public data repository.

7/ Here is a short blog about some of the ideas in the paper:

genre.com/knowledge/blog…

8/ Personally, it has been rewarding experience working with @MarkusEconomist, who in addition to being a remarkable economist, has been a teacher and mentor for many years, and Carlos, who is supremely thoughtful economist, friend & colleague, at the Central Bank of Costa Rica.