Financial markets adjust continuously to forces such as monetary policy decisions and economic or geopolitical events. In such an environment of uncertainty, accurately forecasting future volatility becomes crucial. This thesis develops forecasting models that apply reconciliation techniques to produce coherent and potentially more accurate volatility forecasts than those delivered by the HAR model. The approach exploits an intraday decomposition of realized volatility and the hierarchical structure induced by it. The empirical analysis is based on twenty exchange rate series. The study proceeds in several stages: high-frequency data are first cleaned; an exploratory analysis of returns is then carried out; realized volatility is decomposed; and finally, forecasting models are estimated. The aim is to evaluate both predictive accuracy and the effectiveness of alternative reconciliation strategies. Forecast performance is assessed using loss functions such as MSE and QLIKE, and formally compared through the Diebold–Mariano test.
I prezzi di mercato variano quotidianamente in risposta a dinamiche quali decisioni di politica monetaria ed eventi economici o geopolitici. In tale contesto di incertezza, stimare la volatilità futura diventa fondamentale. Questo elaborato di tesi si propone di implementare modelli che utilizzino tecniche di riconciliazione per ottenere previsioni coerenti e, potenzialmente, più accurate rispetto a quelle ottenute con il modello HAR. A tal fine si sfrutta una decomposizione intragiornaliera della volatilità realizzata e la struttura gerarchica che essa implica. Il lavoro empirico, condotto su venti serie di tassi di cambio, si articola in più fasi. In primo luogo si effettua la pulizia dei dati ad alta frequenza. Successivamente viene condotta un'analisi esplorativa della serie dei rendimenti. Si procede poi con la decomposizione della volatilità realizzata e, infine, si implementano modelli di previsione, con l'obiettivo di valutare l'accuratezza e l'efficacia delle strategie di riconciliazione. Le performance previsionali dei modelli vengono confrontate mediante criteri MSE e QLIKE e tramite il test di Diebold e Mariano.
Previsione della volatilità realizzata tramite decomposizione giornaliera: un’analisi sui tassi di cambio
CASAGRANDE, MARTINA
2024/2025
Abstract
Financial markets adjust continuously to forces such as monetary policy decisions and economic or geopolitical events. In such an environment of uncertainty, accurately forecasting future volatility becomes crucial. This thesis develops forecasting models that apply reconciliation techniques to produce coherent and potentially more accurate volatility forecasts than those delivered by the HAR model. The approach exploits an intraday decomposition of realized volatility and the hierarchical structure induced by it. The empirical analysis is based on twenty exchange rate series. The study proceeds in several stages: high-frequency data are first cleaned; an exploratory analysis of returns is then carried out; realized volatility is decomposed; and finally, forecasting models are estimated. The aim is to evaluate both predictive accuracy and the effectiveness of alternative reconciliation strategies. Forecast performance is assessed using loss functions such as MSE and QLIKE, and formally compared through the Diebold–Mariano test.| File | Dimensione | Formato | |
|---|---|---|---|
|
Casagrande_Martina.pdf
accesso aperto
Dimensione
870.29 kB
Formato
Adobe PDF
|
870.29 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/20.500.12608/98981