Streetlight Effects in the Study of Employer Collusion in the Labor Market.

This paper discusses how research and regulation of anti-competitive behavior in the labor market may suffer from a streetlight effect: looking for evidence where the light is already shining. Drawing from a text corpus of 151,708 franchise documents assembled from public records, I demonstrate gaps in knowledge about the extent and variety of anti-competitive language in this domain. Using new open-source methods and data to classify the unstructured text, I demonstrate a variety of language with anti-competitive effects, and illustrate patterns in their use over time. I provide evidence that hundreds of "naked" no poach and noncompete clauses continue into the present moment, despite their legally suspect nature. To shine a new light on this growing area of research, this paper reports on and releases to the public: the text corpus, the software used to build data, a knowledge base of rules to detect anti-competitive clauses, and an open-source machine learning classifier to detect no poach clauses. [ABSTRACT FROM AUTHOR]