Three Essays on Korean Capital Market

Abstract

This dissertaion collects three essays on the topic of Korean capital market. The first essay observes ESG factors mitigating or causing next year’s stock price crash for Korean firms with core interest on chaebols, unique type of Korean conglomerates. Using firm and (non)financial data, I analyze the relation of individual ESG data and future price crash. The major findings are the following. First, I find the negative relationship between governance and stock price crash for Korean firms, implying that their price crashes are mitigated by the improvement of governance (G) factor. On the other hand, in contrast with previous researches, the relationship between social (S) factor and stock price crash for both chaebols and non-chaebols is quite uncertain in the sample due to contrasting effect of its subfactors such as partner and employee relation. Second, I find that the subfactor for governance, stockholder, has negative relationship with Korean firms regardless of their association with chaebols. Lastly, I find that environment (E) score for chaebols has positive relationship with stock price crash, signaling that the endeavor to improve environmental performance is linked to the substantial increase of financial costs, which could be penalized in the perspectives of investors. In contrast, this relationship is negative for non-chaebols, which is in accordance with previous researches. The aforementioned results are robust to controls for year characteristics and the removal of any outliers.

Employing a variety of Machine Learning (ML) algorithms, my second essay predict optimal capital structure of listed firms in Korea, comparing the performance of linear and machine learning models - namely, Multi-regression, LASSO, Random Forest (RF) and Gradient Boosting Regression (GBM). Setting the training and test set as 2003-2014 and 2015-2019, I find that the predicting performance on firm leverage, as measured in and MSE (Mean Square Error) for RF and GBM is much effective than that of LM and LASSO. In particular, the variables with high predictive power are the Market-to-Book ratio, NetPay, Z-score, Profit, and so on. Finally, after estimating the speed of adjustment (SOA) to the optimal capital structure, using the model of Amini et al. (2021), I confirm that RF and GBM are more predictive than LM and LASSO. Lastly, I find that the leverage adjustment speed of non-chaebols is much faster than that of chaebols especially in machine learning models which is due to the debt-dependent characteristic of chaebols.

My third essay provides evidence on the effect of monetary policy shocks on research and development (R&D) in setting with and without firm-specific variables. I identify monetary shocks by orthogonalizing policy rate change with respect to economic forecast information. Using the shock, I examine the responses of the R&D expenses to increase in the short-term interest rate changes. My empirical results prove that R&D investment gradually decreases in response to monetary policy shock. This trend becomes less apparent for chaebols, due to their access to internal financing within affiliates while the R&D investment of non-chaebols to monetary policy shock has decreased at statistically significant level. In particular, non-chaebols with a high Tobin’s-Q and firm size are more responsive to monetary tightening shocks than those with low ones, which are in consistence with previous researches.