Abstract: International labor migration offers high wages to developing-country workers at scale, but employers and intermediaries are prone to extract more rent from migrants than allowed by law. Policies often seek to protect migrants by regulating the destination labor market, but how do their effects propagate through the labor supply chain? Using a simple model and a sudden increase in the minimum wage for overseas Filipino workers, I show that intermediary market structure matters for how such policies affect equilibrium total migrant surplus. Specifically, migrant protection policies increase intermediaries' costs of operating legally, increasing both exits and incentives for illegal practices. I document that (i) exits of intermediaries that serve multiple destination and origin markets induce spillovers and reduce market access to and from untreated markets, and that (ii) the policy disproportionately drives out illegal intermediaries, who tend to have small market shares and place workers in low-wage jobs.

Abstract: Institutional curriculum design in higher education is often discussed in terms of breadth and depth. While studies often rely on coarse disciplinary classifications or the number of majors, more granularity may be desired if students select courses from a broad range of fields. We develop a method to quantify curricular breadth by applying deep learning and graph theory to text data found in course catalogs. Using 17 years of data from a liberal arts college, we show that mapping the curriculum as a high-dimensional network based on semantic similarity recovers intuitive relationships between courses cutting across disciplinary boundaries. We apply Weitzman (1992)'s diversity index to measure breadth in a fully distance-based and discipline-agnostic way, and show three facts about the college's curriculum: i) differences in content within disciplines and similarities across disciplines are substantial, ii) our breadth metric has good theoretical properties and can be interpreted as a combination of more "primitive" measures, and iii) our breadth metric predicts student enrollment even conditional on the number of courses offered.