Born in the last decade, cryptocurrency markets combine technological innovation with unprecedented volatility. Accurate risk measurement of this highly speculative asset class is therefore vitally important. The following study addresses the limitations of traditional GARCH models, which assume a single variance equation, by presenting the two-state Markov-Switching GARCH (MS-GARCH) regime-switching approach. The analysis conducted on the daily returns of Bitcoin (BTC) and Ethereum (ETH) in the period 2018-2025 aims to identify and distinguish between periods of low volatility and high volatility. While the results of the in-sample estimates show that MS-GARCH models offer a better fit than traditional GARCH models, the performance of the out-of-sample forecasts reveals that the additional complexity does not translate into statistically significant predictive improvement when compared to single-regime benchmarks. It should be noted that for Ethereum, although the Diebold-Mariano test is not statistically significant in comparison, an improvement in forecast error metrics was found. The Markov-Switching approach proved to be a solid choice for describing risk states and obtaining more information on the characteristics of cryptocurrencies. At the same time, from a forecasting perspective, the results of the study show that traditional GARCH models achieve equivalent results in attempting to predict these rapidly growing markets.
Nati nell'ultimo decennio, i mercati delle criptovalute affiancano all'innovazione tecnologica una volatilità senza precedenti. Di vitale importanza è dunque una misurazione accurata del rischio di questa asset class altamente speculativa. Il seguente studio affronta il limite dei modelli GARCH tradizionali , i quali assumono un'unica equazione della varianza , presentando l'approccio a cambiamento di regime Markov-Switching GARCH (MS-GARCH) a due stati. L'analisi condotta sui rendimenti giornalieri di Bitcoin (BTC) ed Ethereum (ETH) nel periodo 2018-2025 si pone l'obiettivo di identificare e distinguere tra periodi di bassa volatilità ed alta volatilità. Se da una parte i risultati delle stime in-sample mostrano che i modelli MS-GARCH offrono un fit superiore rispetto ai GARCH tradizionali, dall'altra le performance delle previsioni out-of-sample rivelano che la complessità aggiuntiva non si traduce in una superiorità predittiva statisticamente significativa nel confronto con i benchmark a regime singolo. È opportuno sottolineare che per Ethereum , seppure il test Diebold-Mariano non sia statisticamente significativo nel confronto, si è riscontrato un miglioramento nelle metriche di errore di previsione. L'approccio Markov-Switching è risultata una scelta solida per descrivere gli stati di rischio ed ottenere maggiori informazioni sulle caratteristiche delle criptovalute. Al tempo stesso in ottica di forecasting, dai risultati dello studio, i modello GARCH tradizionali raggiungono risultati equivalenti nel tentativo di prevedere questi mercati in rapida ascesa.
Analisi dei Regimi di Volatilità nei Mercati delle Criptovalute attraverso Modelli Markov Switching GARCH
ZAPODEANU, MATTEO STEFANO
2024/2025
Abstract
Born in the last decade, cryptocurrency markets combine technological innovation with unprecedented volatility. Accurate risk measurement of this highly speculative asset class is therefore vitally important. The following study addresses the limitations of traditional GARCH models, which assume a single variance equation, by presenting the two-state Markov-Switching GARCH (MS-GARCH) regime-switching approach. The analysis conducted on the daily returns of Bitcoin (BTC) and Ethereum (ETH) in the period 2018-2025 aims to identify and distinguish between periods of low volatility and high volatility. While the results of the in-sample estimates show that MS-GARCH models offer a better fit than traditional GARCH models, the performance of the out-of-sample forecasts reveals that the additional complexity does not translate into statistically significant predictive improvement when compared to single-regime benchmarks. It should be noted that for Ethereum, although the Diebold-Mariano test is not statistically significant in comparison, an improvement in forecast error metrics was found. The Markov-Switching approach proved to be a solid choice for describing risk states and obtaining more information on the characteristics of cryptocurrencies. At the same time, from a forecasting perspective, the results of the study show that traditional GARCH models achieve equivalent results in attempting to predict these rapidly growing markets.| File | Dimensione | Formato | |
|---|---|---|---|
|
Zapodeanu_Matteo_Stefano.pdf
accesso aperto
Dimensione
2.67 MB
Formato
Adobe PDF
|
2.67 MB | 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/20.500.12608/99013