After the seminal paper of Engle (1982) and the generalization provided by Bollerslev (1986), studies on time-varying volatility abound in the literature. Researchers proposed several extensions to capture the so-called stylized facts, i.e., the empirical regularities in the series of financial variables like asset returns. One is related to the long-memory behavior, that led to the development of component models to capture, in a parsimonious way, this complex dependence structure. Furthermore, component models based on the MIDAS filter can capture the effect of variables sampled with a lower frequency, such as economic variables, on conditional variance. Then, we can analyze the relationship between financial volatility and economic conditions. So,within this Ph.D. thesis, in the first chapter, we briefly review several univariate and multivariate volatility models, highlighting their drawbacks and the improvements we want to provide with the models we propose in the following two chapters. Indeed, univariate MIDAS models cannot immediately capture bursts of volatility due to the smoothness of their long-run component. Then, in the secondchapter, we propose a new MIDAS model with a markovian dynamic in the short-run component to detect abrupt shifts in the average level of the series. Empirical results indicate that taking into account both abrupt shifts in the average level and economic source of volatility improves the in-sample performance of the model than competitive ones. In the multivariate framework, component models based on the Cholesky decomposition provide us with a covariance matrix that is sensible to the order of the assets. For this purpose, in the third chapter, we propose a multivariate component model that is invariant to asset ordering with a substantial gain in the estimation time. We provide also a specification, based on the Hadamard exponential function, with time-varying and asset-pair specific parameters. Both in-sample and out-of-sample analyses indicate that the proposed model outperforms the competitive ones.

Component Volatility Models: A MIDAS Approach

SCAFFIDI DOMIANELLO, Luca
2022-02-28

Abstract

After the seminal paper of Engle (1982) and the generalization provided by Bollerslev (1986), studies on time-varying volatility abound in the literature. Researchers proposed several extensions to capture the so-called stylized facts, i.e., the empirical regularities in the series of financial variables like asset returns. One is related to the long-memory behavior, that led to the development of component models to capture, in a parsimonious way, this complex dependence structure. Furthermore, component models based on the MIDAS filter can capture the effect of variables sampled with a lower frequency, such as economic variables, on conditional variance. Then, we can analyze the relationship between financial volatility and economic conditions. So,within this Ph.D. thesis, in the first chapter, we briefly review several univariate and multivariate volatility models, highlighting their drawbacks and the improvements we want to provide with the models we propose in the following two chapters. Indeed, univariate MIDAS models cannot immediately capture bursts of volatility due to the smoothness of their long-run component. Then, in the secondchapter, we propose a new MIDAS model with a markovian dynamic in the short-run component to detect abrupt shifts in the average level of the series. Empirical results indicate that taking into account both abrupt shifts in the average level and economic source of volatility improves the in-sample performance of the model than competitive ones. In the multivariate framework, component models based on the Cholesky decomposition provide us with a covariance matrix that is sensible to the order of the assets. For this purpose, in the third chapter, we propose a multivariate component model that is invariant to asset ordering with a substantial gain in the estimation time. We provide also a specification, based on the Hadamard exponential function, with time-varying and asset-pair specific parameters. Both in-sample and out-of-sample analyses indicate that the proposed model outperforms the competitive ones.
28-feb-2022
File in questo prodotto:
File Dimensione Formato  
PhD Thesis - Luca Scaffidi Domianello.pdf

accesso aperto

Tipologia: Tesi di dottorato
Licenza: Creative commons
Dimensione 3 MB
Formato Adobe PDF
3 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11570/3221495
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact