Abstract We quantify long memory in investor-segregated cash flows in the Korean equity market from 2015 to 2024. Using detrended fluctuation analysis on BUY, SELL, and NET aggregates, we estimate the Hurst exponent both as a full-sample measure and in a 250-day rolling window. All series are heavy-tailed, with complementary cumulative distribution exponents roughly between two and three. Time-shuffled controls return Hurst exponents near one half, indicating that persistence arises from temporal correlations rather than marginal distributions. For BUY and SELL flows, persistence ranks highest for retail investors, intermediate for institutions, and lowest for foreign investors. For NET flows, long-range dependence is lower for all three groups than in their own BUY/SELL baselines. Rolling estimates exhibit clear regime sensitivity, with level shifts around the 2018–2019 tariff episode, the COVID-19 pandemic, and the disinflation period from November 2022 through October 2024. Regressions of daily volatility on the rolling Hurst exponent yield positive and statistically significant coefficients for most groups; notably, the retail NET Hurst exponent shows predictive power for future volatility, whereas the institutional NET Hurst exponent does not. The findings provide a model-light, replicable diagnostic of investor-specific temporal correlations and their regime dependence, challenging a simple noise-trader versus informed-trader.
Gabjin Oh (Tue,) studied this question.