(In progress) We estimate the impact of international conflict on bilateral trade relations and firm values using an incident of politically motivated boycott in Korea. In particular, based on the synthetic control method and product-level supply chain data, we explore how the boycott affected individual firm values and trade patterns of goods.

(Working paper) Can machine learning methods safely apply to time series data? Recent literature in asset pricing utilize machine learning methods in order to deal with a large number of pricing factors. Penalized (regularized) regressions such as LASSO or Ridge, are widely adopted among them for its simplicity. However, K-fold cross-validation, which is typically used to find the hyperparameters for ML regressions, may not yield consistent results if there exists serial correlation in data. In this paper, I explore the extent to which serial correlation can affect the choice of regularization parameters and features.

(Published) We analyze whether a firm’s corporate social responsibility (CSR) plays a significant role in promoting its market value in an emerging market, namely Korea. We employ environmental, social, and corporate governance (ESG) scores to evaluate CSR performances and examine their effect on firm valuation.

This paper examines time-variation and significance of ‘alpha’ in asset pricing models by adopting kernel smoothing regression and simultaneous confidence bands. Using returns data of actively managed mutual funds in U.S., I show that time-series of alphas of mutual funds are significantly different from zero, implying that popular asset pricing models such as Fama-French five-factor are not able to explain ‘fund manager skills.’