This thesis applies a Mixture of Distributions (MDH) model with common factors to a historical series of stock returns and their trading volume. The dataset considered covers a ten-year period, from 7 March 2011 to 3 March 2021, and concerns stocks of the Dow Jones Industrial Average (DJIA). The model used represents an extension of the approach introduced by Tauchen and Pitts (1983), including a common information component that explains interactions between stocks due to general news, such as monetary and fiscal policies. For the identification of latent variables and common factors, a three-step iterative procedure was implemented: the initial estimation of information intensity using the conditional moment method, followed by the use of the Expectation-Maximisation (EM) algorithm to obtain estimates of factorial loadings, and finally a recursion step until the convergence of the likelihood criterion. The empirical results show that the inclusion of common factors significantly improves the model's ability to explain the co-movement between volatility and volume observed in financial markets. The positive correlation between volatility and volume, theoretically predicted by the MDH model, is confirmed in the data. Finally, a Monte Carlo simulation completes the validation of the approach. The thesis is therefore structured as follows: in the first chapter, the standard MDH model and its most relevant extensions are presented; in the second chapter, the MDH model with common factors is analysed and explored in depth; in the third chapter, the EM algorithm is presented; and in the fourth chapter, the process and results of the experiment are shown.
Questa tesi applica un modello di Mixture of Distributions (MDH) con fattori comuni a una serie storica di rendimenti azionari e dei relativi volumi di scambio. Il dataset considerato copre un periodo di dieci anni, dal 7 marzo 2011 al 3 marzo 2021, e riguarda i titoli del Dow Jones Industrial Average (DJIA). Il modello utilizzato rappresenta un’estensione dell’approccio introdotto da Tauchen e Pitts (1983), includendo una componente di informazione comune che spiega le interazioni tra i titoli dovute a notizie di carattere generale, come politiche monetarie e fiscali. Per l’identificazione delle variabili latenti e dei fattori comuni, è stata implementata una procedura iterativa in tre passaggi: la stima iniziale dell’intensità dell’informazione tramite il conditional moment method, seguita dall’impiego dell’algoritmo di Expectation-Maximization (EM) per ottenere le stime dei carichi fattoriali, e infine una fase di ricorsione fino alla convergenza del criterio di verosimiglianza. I risultati empirici evidenziano come l’inclusione dei fattori comuni migliori significativamente la capacità del modello di spiegare la co-movimentazione tra volatilità e volume osservata nei mercati finanziari. La correlazione positiva tra volatilità e volume, prevista teoricamente dal modello MDH, trova conferma nei dati. Infine, una simulazione Monte Carlo completa la validazione dell’approccio. La tesi è pertanto strutturata come segue: nel primo capitolo vengono presentati il modello MDH standard e le sue estensioni più rilevanti; nel secondo capitolo viene analizzato e approfondito il modello MDH con fattori comuni; nel terzo capitolo si espone l’algoritmo EM; nel quarto capitolo si mostrano il processo e i risultati dell’esperimento.
COMMON FACTORS IN TRADING VOLUME AND VOLATILITY
MASTRIA, ALEX
2024/2025
Abstract
This thesis applies a Mixture of Distributions (MDH) model with common factors to a historical series of stock returns and their trading volume. The dataset considered covers a ten-year period, from 7 March 2011 to 3 March 2021, and concerns stocks of the Dow Jones Industrial Average (DJIA). The model used represents an extension of the approach introduced by Tauchen and Pitts (1983), including a common information component that explains interactions between stocks due to general news, such as monetary and fiscal policies. For the identification of latent variables and common factors, a three-step iterative procedure was implemented: the initial estimation of information intensity using the conditional moment method, followed by the use of the Expectation-Maximisation (EM) algorithm to obtain estimates of factorial loadings, and finally a recursion step until the convergence of the likelihood criterion. The empirical results show that the inclusion of common factors significantly improves the model's ability to explain the co-movement between volatility and volume observed in financial markets. The positive correlation between volatility and volume, theoretically predicted by the MDH model, is confirmed in the data. Finally, a Monte Carlo simulation completes the validation of the approach. The thesis is therefore structured as follows: in the first chapter, the standard MDH model and its most relevant extensions are presented; in the second chapter, the MDH model with common factors is analysed and explored in depth; in the third chapter, the EM algorithm is presented; and in the fourth chapter, the process and results of the experiment are shown.| File | Dimensione | Formato | |
|---|---|---|---|
|
tesi.pdf
accesso aperto
Dimensione
1.28 MB
Formato
Adobe PDF
|
1.28 MB | Adobe PDF | Visualizza/Apri |
È consentito all'utente scaricare e condividere i documenti disponibili a testo pieno in UNITESI UNIPV nel rispetto della licenza Creative Commons del tipo CC BY NC ND.
Per maggiori informazioni e per verifiche sull'eventuale disponibilità del file scrivere a: unitesi@unipv.it.
https://hdl.handle.net/20.500.14239/30289