| Autor: |
Arnerić, Josip, Čeh Časni, Anita, Šoštarić, Antonio |
| Přispěvatelé: |
Scitovski, R, Zekić-Sušac, M. |
| Jazyk: |
angličtina |
| Rok vydání: |
2016 |
| Předmět: |
|
| Popis: |
In this paper we compare volatility estimates using a combination of open, high, low and close (OHLC) price information against realized volatility estimates using intraday data. The latter estimates can be considered unbiased in the absence of microstructure noise and intraday autocorrelation. Competing estimators using OHLC prices are assessed from a statistical perspective using (a) a set of loss functions to gauge the precision and (b) time series of conditional correlations resulting from a MGARCH(1, 1) model which enables an analysis of the direction and magnitude of the models. The results associated with an application to the Croatian index indicate that only the Parkinson and the Yang and Zhang estimators outperform the simple close-to-open and the close-to-close estimators, which is largely comparable to the standard deviation of returns. By adding a simple extension for overnight returns to the estimators an overall increase in conditional correlations over time is noted and at the same time the unbiased loss functions decrease. Amongst the representative subset of the universe of OHLC volatility estimators analyzed the extended Parkinson’s range based estimator shows the least biased results. The analysis shows consistent results during different time horizons taking into account the increased market volatility during the credit crisis in 2007 and 2008, but also the period with low volatility prior to the crisis. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
