| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.77 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Desde o trabalho de Black e Scholes [1] que estabeleceu a base moderna para a valorização de derivados, ficou claro que a hipótese de volatilidade constante é incompatível com a evidência empírica (ver [2, 3, 4, 5, 6]). O crash de 1987 expôs de forma particularmente aguda essas limitações e impulsionou o desenvolvimento de modelos de volatilidade estocástica, nos quais tipicamente a variância segue um processo com reversão à média. Surgiram então propostas como Hull–White [2] e, mais tarde, o modelo de Heston [7], cuja tração decorre da tratabilidade analítica e de fórmulas quase fechadas para opções Europeias. Não obstante, estudos empíricos evidenciam restrições desses modelos clássicos (por exemplo, na reprodução do sorriso de volatilidade) e motivam alternativas mais flexíveis, mantendo ao mesmo tempo boas propriedades de calibração e precificação. Neste contexto, o modelo –hipergeométrico de volatilidade estocástica (HG), introduzido por Fonseca e Martini [8], surge como uma alternativa não afim que procura contornar limitações de estruturas afins, garantindo positividade da volatilidade instantânea, permitindo correlação entre preço e volatilidade e preservando a tratabilidade para fins de precificação [9, 10]. Partindo desta motivação, o presente trabalho foca-se em dois problemas fundamentais sob a dinâmica HG: O problema de portfolio com consumo em horizonte finito para um investidor com utilidade de potência e o problema da precificação de opções Europeias quando os parâmetros do modelo são estimados com incerteza.
This thesis studies asset allocation and option pricing under the non-affine -hypergeometric stochastic volatility model (HG), a tractable framework that ensures strictly positive instantaneous volatility. The first contribution solves the finite-horizon portfolio problem with consumption with power utility when the risky asset follows HG volatility. After posing the associated stochastic control problem, we transform the Hamilton–Jacobi–Bellman (HJB) equation to a semilinear PDE and represent its candidate solution via an implicit Feynman–Kac formula. On a suitably weighted function space, the induced integral operator is shown to be a contraction and by Banach’s Fixed-Point Theorem this yields existence and uniqueness of the solution. A verification argument then identifies the value function and the optimal feedback policies for consumption and portfolio weight. We implement a numerical scheme to approximate the semilinear HJB and compute the policies. The second contribution develops pricing bounds for European options when HG parameters are statistically estimated and thus uncertain. Parameter uncertainty is encoded via a compact set of admissible perturbations. For each admissible control, the cost process satisfies a linear BSDE and by the comparison theorem we obtain tight upper and lower price bounds as solutions to linear BSDEs with pointwise optimized drivers. Empirically, we calibrate the volatility factor using high-frequency (∼1 minute) data. The realized variance provides a daily proxy for volatility, enabling maximum-likelihood estimation of the latent dynamics and the price–volatility correlation. The admissible uncertainty set is derived from the covariance of the estimators, and the bounding BSDEs are solved numerically.
This thesis studies asset allocation and option pricing under the non-affine -hypergeometric stochastic volatility model (HG), a tractable framework that ensures strictly positive instantaneous volatility. The first contribution solves the finite-horizon portfolio problem with consumption with power utility when the risky asset follows HG volatility. After posing the associated stochastic control problem, we transform the Hamilton–Jacobi–Bellman (HJB) equation to a semilinear PDE and represent its candidate solution via an implicit Feynman–Kac formula. On a suitably weighted function space, the induced integral operator is shown to be a contraction and by Banach’s Fixed-Point Theorem this yields existence and uniqueness of the solution. A verification argument then identifies the value function and the optimal feedback policies for consumption and portfolio weight. We implement a numerical scheme to approximate the semilinear HJB and compute the policies. The second contribution develops pricing bounds for European options when HG parameters are statistically estimated and thus uncertain. Parameter uncertainty is encoded via a compact set of admissible perturbations. For each admissible control, the cost process satisfies a linear BSDE and by the comparison theorem we obtain tight upper and lower price bounds as solutions to linear BSDEs with pointwise optimized drivers. Empirically, we calibrate the volatility factor using high-frequency (∼1 minute) data. The realized variance provides a daily proxy for volatility, enabling maximum-likelihood estimation of the latent dynamics and the price–volatility correlation. The admissible uncertainty set is derived from the covariance of the estimators, and the bounding BSDEs are solved numerically.
Descrição
Tese de doutoramento em Matemática (Análise Matemática), Universidade de Lisboa, Faculdade de Ciências, 2026.
Palavras-chave
–hypergeometric stochastic volatility optimal control option pricing Hamilton-Jacobi-Bellman -hipergeométrico volatilidade estocástica controlo ótimo precificação de opções Hamilton–Jacobi–Bellman
