In this work we focused on the problem of multi-asset derivatives pricing in a local volatility framework. As a matter of fact several models have been proposed during the last years to improve the limitations of the Black & Scholes theory, and among them, Local volatility plays certainly a significant role thanks to its capability of reproducing the volatility smile consistently with the market. The main issue concerning multi-asset derivatives is to determine the dependence structure between the assets and to correlate them. Before approaching directly to this problem, we first needed to reproduce a realistic dynamics for the assets in time and to imply their probability distributions from market data. Therefore, we solved the Fokker-Planck PDE to obtain the time density evolution of the considered financial assets. After that, to accomplish the fundamental task of correlating the assets, we chose to use Copulas. We preferred this approach with respect to other popular methods lying within the Black and Scholes framework such as the lognormal "Moment matching“ and the multi-dimensional brownian motions, because we were able to overcome their main limitations such as constant volatility and the assumption of lognormal dynamics for the basket underlyings. In addition to that, another important advantage of copulas is that they not only overcome an important limitation of multivariate normal distributions which restrict the association between margins to be linear (as measured by the covariance), but also allow to deal separately with the asset marginal distributions and their dependence structure. In the work presented, we decided to focus on elliptical copulas and implemented a Gaussian and a Student’s t model. Finally, we verified our methodology in repricing some quoted single-asset and multi-asset Autocallable certificates and compared our results with market prices.
In questo lavoro ci siamo concentrati sul problema del pricing dei derivati multi-asset nel contesto di un modello a volatilità locale. Negli ultimi anni sono stati proposti diversi modelli per migliorare i limiti della teoria di Black & Scholes, e tra questi, la Volatilità Locale gioca certamente un ruolo significativo grazie alla sua capacità di riprodurre lo smile di volatilità in modo consistente con il mercato. Il problema principale relativo ai derivati multi-asset è identificare la struttura di dipendenza tra i sottostanti e correlarli. Prima di affrontare direttamente questo problema, abbiamo innanzitutto cercato di riprodurre la dinamica temporale degli asset considerati ed estrarre le loro distribuzioni di probabilità dai dati di mercato. Per fare ciò abbiamo risolto l’equazione a derivate parziali di Fokker-Planck. Dopodiché, per studiare la relazione di dipendenza tra gli asset, abbiamo scelto di utilizzare le Copule. Abbiamo preferito questo approccio rispetto ad altri metodi descritti nella letteratura come il “Moment matching“ o l’utilizzo di moti browniani multi-dimensionali, perché ci ha premesso di poterne superare i limiti principali come l’assunzione di volatilità costante e di dinamica lognormale per i sottostanti del basket. In aggiunta a ciò i modelli di copula superano un’importante limitazione delle distribuzioni normali multivariate in quanto quest’ultime riescono a catturare solo una dipendenza lineare tra le marginali. Tali modelli presentano infine il vantaggio di poter affrontare separatamente il problema modellistico delle distribuzioni marginali e della loro struttura di dipendenza. Nel lavoro presentato, abbiamo deciso di concentrarci sulla categoria delle copule ellittiche e abbiamo implementato un modello Gaussiano e t-Student. Infine, abbiamo verificato la nostra metodologia nel prezzare alcuni certificati Autocallable single-asset e multi-asset e abbiamo confrontato i nostri risultati con i prezzi quotati sul mercato.
Pricing of multi-asset equity products with a local volatility model
SALERNO, ELENA MARIA
2017/2018
Abstract
In this work we focused on the problem of multi-asset derivatives pricing in a local volatility framework. As a matter of fact several models have been proposed during the last years to improve the limitations of the Black & Scholes theory, and among them, Local volatility plays certainly a significant role thanks to its capability of reproducing the volatility smile consistently with the market. The main issue concerning multi-asset derivatives is to determine the dependence structure between the assets and to correlate them. Before approaching directly to this problem, we first needed to reproduce a realistic dynamics for the assets in time and to imply their probability distributions from market data. Therefore, we solved the Fokker-Planck PDE to obtain the time density evolution of the considered financial assets. After that, to accomplish the fundamental task of correlating the assets, we chose to use Copulas. We preferred this approach with respect to other popular methods lying within the Black and Scholes framework such as the lognormal "Moment matching“ and the multi-dimensional brownian motions, because we were able to overcome their main limitations such as constant volatility and the assumption of lognormal dynamics for the basket underlyings. In addition to that, another important advantage of copulas is that they not only overcome an important limitation of multivariate normal distributions which restrict the association between margins to be linear (as measured by the covariance), but also allow to deal separately with the asset marginal distributions and their dependence structure. In the work presented, we decided to focus on elliptical copulas and implemented a Gaussian and a Student’s t model. Finally, we verified our methodology in repricing some quoted single-asset and multi-asset Autocallable certificates and compared our results with market prices.File | Dimensione | Formato | |
---|---|---|---|
2018_12_Salerno.pdf
solo utenti autorizzati dal 05/12/2021
Descrizione: Testo della tesi
Dimensione
3.27 MB
Formato
Adobe PDF
|
3.27 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/144393