High-Frequency Volatility Measurement on Borsa Istanbul: A TTSE-Anchored Comparison of Range Estimators

This study examines how closely low-frequency range-based volatility estimators approximate high-frequency variance proxies, using the Two-Time-Scale Estimator (TTSE) as a microstructure-noise-robust reference instead of the true latent variance. Using Borsa Istanbul index data, daily estimators are evaluated relative to TTSE within a matched framework that distinguishes between intraday variance and total variance, including overnight returns. The results show that low-frequency estimators provide close approximations to high-frequency measures in terms of average accuracy. In particular, the Garman–Klass estimator consistently achieves the lowest loss across indices under both mean square error and quasi-likelihood criteria. Under extreme market conditions, however, extended estimators and the Yang–Zhang model exhibit stronger upper-tail dependence with the benchmark, highlighting the role of overnight variation in capturing volatility spikes. Overall, estimator performance depends on whether the objective is to minimize average error or to capture tail-risk dynamics.