ScholarWorks@중앙대학교

초록

This study examines whether investor sentiment extracted from Korean financial news is priced as a systematic risk factor in the stock market. Unlike previous domestic studies that mainly used sentiment as a predictive variable, this paper integrates sentiment directly into an asset pricing framework by extending the Fama-French five-factor model. We construct a firm-level sentiment score from 5.47 million news headlines collected from 19 major newspapers between 2001 and 2021, using the KR-FinBERT model, a deep-learning language model fine-tuned for Korean financial texts. The monthly sentiment index is computed as the difference between the probabilities of positive and negative tone and aggregated across firms to form a market-wide sentiment factor. Based on KOSPI- and KOSDAQ-listed firms, portfolios are sorted into deciles by sentiment, and the return difference between the most positive and most negative groups (Positive−Negative) yields average excess returns of 1.77% (equal-weighted) and 0.34% (value-weighted), with statistical significance for the former. Regression analyses further confirm that the sentiment factor remains significant after controlling for the market, size, value, profitability, and investment factors. The sentiment effect is most pronounced among small-cap firms, suggesting that it captures behavioral risks associated with information asymmetry and investor attention. When the sentiment factor (PMN) is incorporated into the traditional five-factor framework, it exhibits a positive risk premium comparable to that of the profitability factor, without deteriorating the model’s explanatory power. Overall, the results indicate that investor sentiment functions as a priced behavioral factor in the Korean stock market. By demonstrating that textual sentiment derived from news can systematically explain return variations, this study provides empirical evidence that bridges behavioral finance and traditional factor modeling, offering new implications for portfolio construction and market efficiency research.

키워드