We will consider two models, extensions of the GARCH model, in particular the Realized GARCH and the Heavy GARCH, an empirical estimate of the conditional variance will be made with these models on a dataset containing the prices of Microsoft Open, High, Low, Close and Volume, therefore opening, highs, lows, closing and volumes, relating to the years 2018-2023, comparing these models with the GARCH model as a benchmark, to see if the two models fit empirically better to high frequency data, of 10 seconds specifically , compared to the GARCH model. GARCH models for calculating volatility typically use daily squared returns, which, however, are not suitable for estimating current conditional volatility, as they do not capture rapid changes and jumps, this is due to the fact that squared returns daily returns contain less information than higher frequency returns. To estimate the conditional variance, with high frequency data, the realized volatility, estimated via realized variance, the bipower variation or the realized kernel, is included in the models. Two models that make use of realized variance are the HEAVY model of Shephard and Shephard (2010), which is a specification of the Multiplicative Error Model (MEM) of Engle and Gallo (2006), and the Realized GARCH model.
Considereremo due modelli, estensione del modello GARCH, in particolare il Realized GARCH e l’Heavy GARCH, verrà fatta una stima empirica della varianza condizionata con tali modelli su un dataset contenente i prezzi di Microsoft Open, High, Low, Close e Volume, quindi di apertura, massimi, minimi, di chiusura e i volumi, relativi agli anni 2018-2023, confrontando tali modelli con il modello GARCH come benchmark, per vedere se i due modelli si adattano empiricamente meglio ai dati in alta frequenza, di 10 secondi nello specifico, rispetto al modello GARCH. I modelli GARCH per il calcolo della volatilità tipicamente utilizzano i rendimenti al quadrato giornalieri, che, però, non sono adatti alla stima della volatilità condizionata corrente, in quanto non catturano i rapidi cambiamenti e i salti, ciò è dovuto al fatto che i rendimenti al quadrato giornalieri contengono meno informazione rispetto ai rendimenti con frequenza più elevata. Per la stima della varianza condizionata, con dati ad alta frequenza, si include nei modelli la volatilità realizzata, stimata tramite varianza realizzata, la bipower variation o la kernel realizzata. Due modelli che fanno uso della varianza realizzata sono l’HEAVY model di Shephard e Shephard (2010), che è una specificazione del Multiplicative Error Model (MEM) di Engle e Gallo (2006), e il modello Realized GARCH.
Modelli GARCH aumentati con dati in alta frequenza: Heavy e Realized GARCH
GRAVILI, COSIMO MARCO
2023/2024
Abstract
We will consider two models, extensions of the GARCH model, in particular the Realized GARCH and the Heavy GARCH, an empirical estimate of the conditional variance will be made with these models on a dataset containing the prices of Microsoft Open, High, Low, Close and Volume, therefore opening, highs, lows, closing and volumes, relating to the years 2018-2023, comparing these models with the GARCH model as a benchmark, to see if the two models fit empirically better to high frequency data, of 10 seconds specifically , compared to the GARCH model. GARCH models for calculating volatility typically use daily squared returns, which, however, are not suitable for estimating current conditional volatility, as they do not capture rapid changes and jumps, this is due to the fact that squared returns daily returns contain less information than higher frequency returns. To estimate the conditional variance, with high frequency data, the realized volatility, estimated via realized variance, the bipower variation or the realized kernel, is included in the models. Two models that make use of realized variance are the HEAVY model of Shephard and Shephard (2010), which is a specification of the Multiplicative Error Model (MEM) of Engle and Gallo (2006), and the Realized GARCH model.File | Dimensione | Formato | |
---|---|---|---|
Gravili_CosimoMarco.pdf
accesso aperto
Dimensione
842.72 kB
Formato
Adobe PDF
|
842.72 kB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/71264