자산 가격결정 실증 연구의최근 동향과 시사점 (A Review of Recent Studies on Empirical Asset Pricing)

서지정보

· 발행기관 : 한국금융학회
· 수록지 정보 : 금융연구 / 34권 / 3호 / 1 ~ 31페이지
· 저자명 : 장지원

초록

이 논문은 최근 10년간 자산 가격결정 분야에서 기대수익률의 횡단면에 대한 실증 연구가 어떻게진행되어 왔는지 검토하고, 향후 연구의 진행 방향에 대한 시사점을 도출한다. 전반부는 여러횡단면 이상현상들과 이를 설명하기 위해 경쟁적으로 등장한 새로운 요인 모형들을 다룬다.
후반부에서는 이상현상의 범람과 고차원성에 관한 논제를 다룬다. 구체적으로, 데이터 스누핑, 다중 가설검정, 거래 비용, 차원 축소 및 모형 선택의 문제에 대한 최근 연구 내용을 소개하고그 관련성을 설명한다.

영어초록

This paper reviews recent studies on empirical asset pricing, focusing on papers published during the last decade. Over the past 50 years, what determines the cross-section of the expected stock returns is one of research questions that have received the most attention in the asset pricing field. While Fama and French (1993) three-factor model has been known to well summarize the cross-section of stock returns as of the 1990s, a wide variety of cross-sectional anomalies have accumulated since then and remained unexplained by the three-factor model. Given a “zoo” of hundreds of characteristics known as significant predictors of the cross-section of stock returns in the literature, the need for an alternative factor model that can account for more anomalies has been increasingly clear. At the same time, researchers have been confronted by multidimensional challenges in empirical studies.
In the first half, I review well-known cross-sectional anomalies whose robustness have been well verified by subsequent studies and several new factor models recently proposed to accommodate those cross-sectional patterns. These alternative factor models can be classified into two categories, one that combines the investment or profitability factors with the existing market, size, and value factors and the other that consists of mispricingrelated factors motivated by the behavioral concepts that a couple of common factors drive mispricing and anomalies.
In the second half, I survey recent studies that address the “factor zoo” and high-dimensionality. I present four important issues: (i) data snooping, (ii) multiple hypothesis testing, (iii) trading costs, and (iv) dimension reduction and model selection. Several papers provide out-of- sample tests or replication studies to examine the possibility of numerous anomalies being a result of data snooping. Multiple hypothesis testing methods become popular recently as a formal and statistical precaution against the data snooping bias. Meanwhile, the need for different empirical methods is rapidly increasing to address the high-dimensional nature of empirical tasks. Many recent studies attempt new statistical techniques of dimension reduction to find a parsimonious factor model that performs the best out-of-sample.
Although the number of new factors is increasing and the use of complex statistical techniques is inevitable, researchers should bear in mind that the end goal of asset pricing studies is to understand economic mechanisms behind asset pricing factors. Economic interpretability is the most important particularly for factors discovered by a purely statistical approach.

참고자료

· 없음