This study examines the relationship between the regret premium, investor composition, and investor-specific trading responses in the Korean stock market. Behavioral finance suggests that asset prices may reflect psychological biases. Regret is relevant because investors may compare the performance of the assets they hold with that of alternatives they could have selected. Recent studies define regret (REG) as the difference between the highest return within an industry and the realized return of an individual stock and show that stocks with higher REG subsequently earn higher returns. Although prior research has established this premium, relatively little is known about the investor base in which it is more pronounced or how different investor groups subsequently trade high-REG stocks. The Korean market provides a useful setting because individual investors account for a substantial share of trading, and transaction data permit separate analyses by investor type. If regret-related price effects are associated with behavioral biases, their strength may vary with investor composition. Individual investors may exhibit limited information-processing capacity, attention-based trading, and contrarian behavior, whereas institutional and foreign investors may engage relatively more in information-based trading or arbitrage. This study first confirms the baseline relationship between REG and future returns. It then examines whether the magnitude of the REG premium differs according to investor trading shares. The analysis employs portfolio sorts and Fama–MacBeth cross-sectional regressions that include interactions between REG and the trading shares of individual, institutional, and foreign investors. It also examines subsequent investor-specific responses using net buying ratios and trading shares, investigates the persistence of the REG premium, and compares its strength between the KOSPI and KOSDAQ markets. The empirical results are as follows. First, REG positively and significantly predicts future stock returns. Portfolios with high REG earn higher subsequent returns than those with low REG, and the return spread remains significant after risk adjustment using the CAPM and the Fama–French three-factor model. Fama–MacBeth regressions likewise confirm that REG has positive predictive power for future returns. Second, the magnitude of the REG premium varies systematically with investor trading shares. The premium is larger among stocks with high individual-investor trading shares, whereas it is weaker among stocks with high institutional- or foreign-investor trading shares. Interaction regressions provide consistent evidence: the interaction between REG and the individual-investor trading share is significantly positive, whereas the interaction between REG and the institutional-investor trading share is significantly negative. The interaction between REG and the foreign-investor trading share is also negative, although its significance weakens after control variables are included. Third, subsequent net buying responses differ across investor types. In regressions using investor-specific net buying ratios as dependent variables, individual investors tend to be net buyers of high-REG stocks, whereas institutional investors exhibit significant net selling. Foreign investors exhibit a negative net buying response in univariate analysis, but the result is not robust after controlling for other stock characteristics. The trading-share analysis reveals a related pattern: REG is positively associated with the subsequent trading share of individual investors and negatively associated with the corresponding shares of institutional and foreign investors. Taken together, high-REG stocks are followed by net buying and higher relative trading participation by individual investors, whereas institutional investors exhibit net selling and lower relative participation. Fourth, the REG premium is not limited to the immediately subsequent month. Although it gradually weakens, it remains evident for several subsequent months. Among stocks with low individual-investor trading shares, the premium weakens relatively quickly, whereas among stocks with high individual-investor trading shares, it tends to persist longer. Investor composition is therefore associated with both the magnitude and persistence of the premium. Fifth, the KOSPI–KOSDAQ comparison provides supplementary evidence consistent with the main results. Individual investors account for a significantly higher share of trading in the KOSDAQ market, whereas institutional and foreign investors account for lower shares. The REG premium is significant in both markets but is larger in the KOSDAQ market. Within each market, the predictive power of REG also tends to be more pronounced among stocks with higher individual-investor trading shares. Overall, this study shows that the REG premium is closely associated with investor composition and heterogeneous subsequent trading responses rather than merely representing an average return anomaly. The larger and more persistent premium among stocks with high individual-investor trading shares, together with the subsequent net buying and higher relative participation of individual investors in high-REG stocks, indicates that investor-specific trading behavior should be considered when interpreting regret-related return patterns. The findings do not establish that investor-specific trading behavior causally generates the REG premium. Rather, they document systematic relationships among the premium, the investor base in which it is more pronounced, and subsequent trading responses. This study extends the economic interpretation of regret effects by identifying investor composition as an important conditioning factor associated with their strength and persistence.