Research

1.  Investments and the Technology of Skill Formation: The Role of Children’s Time Allocation [Draft]

Abstract: The literature on children’s human capital production emphasizes the role of parental investment, but it leaves the role of children's own time allocations understudied. To examine this potentially essential mechanism, I extend the dynamic skill production model of Cunha et al. (2010) to incorporate both parents' composite investment and children's time allocation. Estimating the model on longitudinal microdata from South Korea, I find distinct roles for parent and child inputs in cognitive and socio-emotioanal skill formation. Parental investment functions as a generalized input across both skill domains. On the other hand, child time operates as a targeted input, revealing significant trade-offs: academic time serves as a primary driver of cognitive development at the cost of socio-emotional growth, while non-academic time exhibits the opposite pattern. 

Presented at: Southern Economic Association Annual Conference (2025); Michigan State University Empirical Micro Lunch Seminar (2026)

1.  Skill Production Technology and Children's Time Allocation: Evidence from U.S. Data

Abstract:  

2. Teacher–Student Match Effects: Structured Shrinkage over a Matching Surface (joint work with Soo Jeong Lee)

Abstract: Standard teacher value-added models impose common effects across students, while unrestricted heterogeneous models can be unstable in sparse teacher–student demographic cells. We develop a structured regularization framework for teacher-specific match heterogeneity. The approach represents teacher effectiveness through a dyadic surface capturing race and gender match effects and their complementarity, while allowing unrestricted student-type effects to deviate from this surface. Penalization shrinks only these deviations, continuously linking the structured and unrestricted estimators, with tuning selected by leave-cohort-out prediction. Monte Carlo simulations show that structured shrinkage improves stability under sparse demographic support while retaining flexibility when heterogeneity departs from the dyadic structure. An application to North Carolina administrative data studies teacher–student matching patterns, teacher-specific match heterogeneity, and out-of-sample prediction.