Junkyung Lab conducts research to effectively collect, analyze, and utilize various data generated in the Internet of Things (IoT) environment. In particular, we focus on improving data processing efficiency and solving security and privacy issues by converging edge computing, artificial intelligence (AI), and blockchain technologies. Our goal is to build intelligent, secure, and reliable systems in future IoT environments.

준경 연구실은 사물인터넷(IoT) 환경에서 발생하는 다양한 데이터를 효과적으로 수집, 분석, 활용하기 위한 연구를 수행하고 있습니다. 특히, 엣지 컴퓨팅, 인공지능(AI), 블록체인 기술을 융합하여 데이터 처리 효율성을 높이고, 보안 및 프라이버시 문제를 해결하는 데 중점을 둡니다. 우리의 목표는 미래 사물인터넷 환경에서 지능적이고 안전하며 신뢰할 수 있는 시스템을 구축하는 것입니다.

Loan Terms and Collateral: Evidence from the Bilateral Repo Market

Natural Disasters and Municipal Bonds

Cross-Asset Information Synergy in Mutual Fund Families

Procyclical Credit Rating Policy

Bankruptcy Code, Optimal Liability Structure and Secured Short-Term Debt

Factor-based portfolio optimization

Repo Priority Right and the Bankruptcy Code

Procyclical Credit Rating Policy*

The Role of Margin and Spread in Secured Lending: Evidence from the Bilateral Repo Market

Tricks of the Trade? Pre-Issuance Price Maneuvers by Underwriter-Dealers

Real Effect of Cost of Financing

Understanding Corporate Bond Defaults in Korea Using Machine Learning Models<sup>*</sup>

Liquidation Cascade and Hedging Front-Running: Evidence from the Structured Equity Product Market

Liquidation Cascade and Anticipatory Trading: Evidence from the Structured Equity Product Market

Real Connectivity of Firms and Stock Return Covariance

Essays on Corporate Credit

A Blockchain-based Secure Data Sharing Scheme for IoT Environments

Real-time Anomaly Detection in Industrial IoT using Edge Computing and Deep Learning

Privacy-preserving Data Aggregation in Smart Grid using Federated Learning

An Efficient Data Offloading Strategy for Mobile Edge Computing in 5G Networks

We investigate corporate bond defaults from 1995 to 2020 using hand-collected data from hard-copy publications in Korea. Using an under-sampling method, we construct default prediction models based on machine learning models as well as a logistic model. The empirical results show that the random forest model outperforms the others. However, regardless of the models used, model performance in financial crisis periods is significantly worse than in non-crisis periods. This finding suggests the need for additional information to improve model performance during crises when the default prediction is the most relevant. Furthermore, the dominant predictor of defaults before the Global Financial Crisis was the debt ratio, while subsequently, the coverage ratio has become the most important predictor.

Understanding Corporate Bond Defaults Using Machine Learning Models

This paper studies whether credit rating agencies applied consistent rating standards to U.S. corporate bonds using the period around the financial crisis. Based on estimates of issuing firms' credit quality from a structural model, I find that rating standards are in fact procyclical: ratings are stricter during an economic downturn than an expansion. As a result, firms receive overly pessimistic ratings in a recession, relative to during an expansion. I further show that a procyclical rating policy amplifies the variation in corporate credit spreads, accounting for, on average, 11 percent of the increase in spreads during a recession. In the cross-section, firms with a higher rollover rate of debt, fewer alternative channels to convey their credit quality to the market, and firms with business more sensitive to economic cycles are more affected by the procyclical rating policy.

A parsimonious factor model mitigates idiosyncratic noise in historical data for portfolio optimization. We use market predictors and machine learning to incorporate forward-looking information into expected returns. The combination of the factor model and forward-looking returns improves out-of-sample performance, conforming to the theoretical assumption that the mean and variance correspond to future returns.

Factor-Based Portfolio Optimization

We study secured lending contracts using a proprietary, loan-level database of bilateral repurchase agreements containing groups of simultaneous loans backed by multiple tranches within a securitization. We show that lower-quality loans (i.e., loans backed by lower-rated collateral) have higher margins and spreads. We calibrate a model using collateral asset prices and find that lower-quality loans are riskier despite the higher margins, yet cheaper for the borrower. This finding is consistent with a combination of lender optimism and reaching for yield. We also show that lower-quality loans have longer maturity, consistent with models of rollover concerns with asymmetric information.

Loan Terms and Collateral

We show that structured equity derivatives can cause significant price pressure of the underlying stock upon an event of dramatic payoff change. Moreover, one event causes another: the event cascade amplifies the magnitude of the impact. We find that a single event accounts for a −6.4% return on the event day, and it increases the probability of a subsequent event by 21.3%. Given the negative price impact, traders try to liquidate ahead of each other, exacerbating the degree of price pressure. Our results uncover the chain-reaction and (mis)coordination mechanism in complex derivatives markets that can provoke substantial price shocks.

Liquidation Cascade and Anticipatory Trading

This paper shows that when the bankruptcy code protects the creditors’ rights with no impairments to secured creditors, issuance of debt such as repo with exemption from automatic stay adds no value. When the bankruptcy process admits violations of absolute priority rules or results in collateral impairments to secured creditors, the liability structure includes short-term debt, with safe harbor protection when the pledged collateral satisfies a minimum liquidity threshold. Safe harbor rights lead firms to issue more short-term debt, less long-term debt and increase the long-term spreads.

We study the trading of dealers around new bond issues underwritten by their affiliates using a complete matched record of U.S. bond market transactions, bond issue deals, and underwriter ownership structure from 2005 to 2015. Compared to dealers unaffiliated with the lead underwriter, affiliated dealers pay up to 54 basis points more for the issuer's preexisting bonds---prior to, during, and after the issuance event. We interpret this phenomenon as cross-security price support and, prior to the event, price maneuvers aimed at lowering the reference yield for new issue investors. By examining dealer inventories and profits, we find no support for alternative explanations such as hedging, informed trading, or competitive advantage in market-making.

This paper highlights the importance of synthesizing cross-asset information in mutual fund families. We show that the investment decisions of equity and bond funds on commonly held firms significantly comove if and only if both funds are affiliated with the same fund family. Moreover, we find that when a fund family's bond funds produce information, its equity funds generate better performance on the cross-held firm's stocks than stand-alone equity funds that hold the same stocks but do not have affiliated bond funds. These findings suggest that cross-holdings within a mutual fund family reduce frictions regarding asset market segmentation and facilitate spillovers of nonredundant information across equity and bond investors.

Climate change is increasing the frequency of natural disasters, and the municipal bond market could be particularly vulnerable to this trend. We undertake a comprehensive analysis of how and when natural disasters affect municipal bond returns. We find substantial price effects that materialize gradually in the weeks following a disaster, translating into in-sample investor losses of almost $10 billion. These effects are influenced by a range of factors that point to the mechanisms behind the observed response, including the source of bond revenue, bond insurance, disaster severity, federal disaster aid, and local financial conditions. Our findings have significant implications for investors, policymakers, and anyone concerned with the long-term stability of financial markets.

Jun Kyung Auh

교수님 소개

Jun Kyung Auh

연구실 소개

연구실 키워드

알럼나이가 간 회사

연구실 지원 방법

대우 조건

연도별 논문 인용수

저널 & 학회 논문 출판 수

연구실 논문 목록

Understanding Corporate Bond Defaults in Korea Using Machine Learning Models<sup>*</sup>

Understanding Corporate Bond Defaults Using Machine Learning Models

Factor-based portfolio optimization

연구실 보유 특허

대학별 입시끈

우리 연구실 모집

교수님 소개

Jun Kyung Auh

연구실 키워드

알럼나이가 간 회사

연구실 지원 방법

대우 조건

연도별 논문 인용수

저널 & 학회 논문 출판 수

연구실 논문 목록

Understanding Corporate Bond Defaults in Korea Using Machine Learning Models<sup>*</sup>

Understanding Corporate Bond Defaults Using Machine Learning Models

Factor-based portfolio optimization

연구실 보유 특허

최근 3년간 연구실적 높은 연구실

대학별 입시끈

우리 연구실 모집