Research

Published Papers

  • Evaluating the Impact of South Korea’s SME Support Policies Using the Economic Shock of Covid-19 with Seunghoon Lee. Published in Global Economic Review, Feb. 2026, pp. 1-22. [Paper]. [Working Draft]
    Abstract

    This study evaluates the impact of South Korea’s small and medium enterprise (SME) support policies on firm survival and performance by leveraging the exogenous shock of the COVID-19 pandemic. Using administrative data from the SMEs Integrated Management System (SIMS), we analyze the effectiveness of various SME support programs, includi$$ng technology, human resources, export, domestic sales, startup, and management support. To address endogeneity, we exploit the natural experiment created by COVID-19; to mitigate selection bias, we restrict the sample to firms that received financial support in 2019 and examine how the timing of support (across 2019 quarters) affects outcomes. If firms receiving support closer to the pandemic show greater improvements, this suggests a policy effect rather than selection bias. Our results indicate that technology and domestic sales support are significantly associated with higher employment, revenue, and firm survival. We also examine complementarity and substitutability among support policies, finding significant interactions for certain combinations.

Working Papers

  • Difference-in-Differences with fine treatment timing and coarse outcomes: Theoretical justification and an empirical application to Korean SME support programs
    Abstract

    Administrative program records routinely timestamp treatment entry at a fine frequency (e.g., the month a firm or an individual enters a government subsidy program) while the outcomes researchers care about (e.g., the firm’s or individual’s taxable returns) are reported at a sparser frequency, for example, only once a year. A common compromise collapses entry into a “treated this year” dummy, discarding the within-year variation in exposure, or to focus only on the observation treated at the beginning of the coarser period. These two methods are either biased or inefficient, since the former leads to biased estimates of the full-year average treatment effect on the treated, and the latter leads to discarding observations whose treatment effect gives some information about the full-year effect. I show that the 2×2 difference-in-differences that discards the within-year variation in exposure identifies an exposure-weighted average of the fine treatment-effect schedule and is tilted toward short exposures. To correct for the bias, I propose an estimating strategy that identifies the fine-period effect schedule first and then recovers the full-year effect by a roughness-penalized smoothing estimator with a Bayesian reading whose posterior for the full-year effect widens in exactly one direction. Furthermore, when entry timing is as-good-as-random given participation and the magnitude of the additional effect strictly decreases with exposure (Lee and Park, 2026), the fine effect schedule can be still recovered and a selection-robust lower bound on the full-year effect can be computed together. Applying the foregoing methodology to the Korean SIMS SME database, I convert an informal heuristic in Lee and Park (2026) into a quantitative, selection-bias-robust floor on the first-year effect of the support programs.

Work in Progress

  • Dutch Disease or Dutch Blessing? Resource Booms and Educational Decisions