Affine jump-diffusion term structure models (AJTSMs) are recently receiving much attention in mathematical finance, because they often lead to a tractable analysis of the price distribution functions. This thesis concerns three aspects of mathematical finance, when applied to certain classes of AJTSMs. The first aspect concerns the pricing problem in the special case when the underlying process Xt is a Continuous-Time Markov Chain. For exotic options, where the claims are time or path dependent, prices can only be estimated by Monte-Carlo simulation, in most cases. We show that this computation is simplified by conditioning first on the number Nt,T of the jumps of the chain. A recursion is proposed to compute the expected discounted payoff given Nt,T=k; Monte Carlo is then used to average out the result over the distribution of Nt,T=k. This leads to a variance reduction by conditioning. We present results of numerical tests which indicate that the method often outperforms plain vanilla Monte Carlo for different kinds of claims. The second aspect concerns the calibration of a financial model by parameter estimation, when the underlying, a finite state Markov chain, is only partially observed through noisy asset prices. Here, we assume that the jumps of the asset price occurring at the jump-times of the Markov chain are observable as well. Such a model is a special case of the class of models treated in [FR10b]. Their parameter estimation can be addressed via the EM algorithm, following the approach by [EAM08] which, in the case of discrete-time chains, involves the Kalman filter. We extend this approach to to the case of CTMCs via the use of the Wonham filter. Our main contribution is the numerical approximation of the filters and smoothers in the EM algorithm. We compare the classical Euler and Milstein schemes to a new scheme, inspired by [PR10a], that we call a quasi-exact solution and is related to the splitting-up method of [BGR90] and [Gla92]. We prove that such a scheme is of strong convergence order at least 0.5, hence it performs at least as well as the Euler scheme. We present numerical evidence indicating that in fact, in certain cases the method outperforms both the Euler and the Milstein scheme. The third aspect concerns a unified framework for equity and credit risk modeling, with applications to risk management. Here we treat an affine jump-diffusion model with a single jump-to-default, where the default time is a doubly stochastic random time with intensity driven by an underlying affine factor process. This approach allows for flexible interactions between the defaultable underlying asset price, its stochastic volatility and the default intensity, while maintaining full analytical tractability. We characterize all risk-neutral measures which preserve the affine structure of the model and show that risk management as well as pricing problems can be dealt with efficiently by shifting to suitable survival measures. As an example, we consider a jump-to-default extension of the Heston stochastic volatility model. [FR10b] R. Frey and W. J. Runggaldier, Pricing credit derivatives under incomplete information: a nonlinear filtering approach., Finance and Stochastics 14 (2010), no. 4, 495–526. [PR10a] E. Platen and R. Rendek, Quasi-exact approximation of hidden markov chain filters., Communications on Stochastic Analysis 4 (2010), 129–142. [BGR90] A. Bensoussan, R. Glowinski, and A. Rascanu, Approximation of the zakai equation by the splitting up method, SIAM Journal of Control and Optimization 28 (1990), no. 6, 1420–1431. [Gla92] F. Le Gland, Splitting-up approximation for spde’s and sde’s with application to non-linear filtering, in: Stochastic Partial Differential Equations and Their Applications, Charlotte 1991, B. L. Rozovskii and R. B. Sowers, editors, Lecture Notes in Control and Information Sciences 176 (1992), 177–187.

Le strutture a termine affine con diffusione a salti (AJTSMs) stanno recentemente ricevendo molta attenzione in finanza matematica, perché spesso è semplice analizzare le funzioni di distribuzione ad esse associate. Questa tesi riguarda tre diversi aspetti della finanza matematica, applicati su certe classi di AJTSMs. Il primo aspetto riguarda il problema del prezzaggio, nel caso particolare in cui il processo sottostante Xt sia una Catena Markoviana a Tempo Continuo (CTMC). Per opzioni esotiche, dove il “claim”, cioè il “payoff” del derivato è dipende dal tempo oppure dalle traiettorie, solitamente i prezzi devono essere stimati attraverso simulazioni di tipo Monte Carlo. Mostriamo che, quando si condiziona prima sul numero Nt,T=k dei salti della catena, il calcolo di questa stima si semplifica. Viene proposta una ricorsione per calcolare il valore atteso del “payoff” scontato, dato Nt,T=k; in seguito si calcola il valore atteso del “payoff” rispetto alla distribuzione di Nt,T=k attraverso un metodo Monte Carlo. Questo condizionamento comporta una riduzione della varianza. Presentiamo i risultati di vari test numerici, che indicano che, per diversi tipi di “claims”, il metodo proposto supera spesso un semplice “vanilla” Monte Carlo. Il secondo aspetto riguarda la calibrazione, cioè la stima dei parametri di un modello finanziario, dove il processo sottostante (una Catena Markoviana finita) è solo parzialmente osservabile tramite i prezzi corrotti del titolo. In questo lavoro, assumiamo che anche i salti del prezzo del titolo corrispondenti ai tempi dei salti della catena Markoviana siano osservabili. Questo è un caso particolare della classe di modelli trattati in [FR10b]. I loro parametri possono essere stimati mediante l’algoritmo “expectation-maximization” (EM), seguendo l’approccio di [EAM08], che, nel caso delle catene a tempo discreto, coinvolge il filtro di Kalman. Estendiamo questo approccio al caso CTMC, usando invece il filtro di Wonham. Il contributo principale di questa parte della tesi è l’approssimazione numerica dei filtri e degli “smoothers” dell’algoritmo EM. Confrontiamo i classici metodi di Eulero e di Milstein con una nuova strategia, simile a [PR10a], che chiamiamo “soluzione quasi-esatta” e che è anche collegata al metodo di “splitting-up” di [BGR90] e [Gla92]. Dimostriamo che tale schema ha un ordine di convergenza forte di almeno 0.5 e che pertanto è almeno tanto efficace quanto lo schema di Eulero. Presentiamo alcuni risultati numerici che indicano che, di fatto, in certi casi il nuovo metodo converge più velocemente di entrambi i metodi di Eulero e di Milstein. Il terzo aspetto riguarda un quadro unificato per la modellazione del rischio di “equity” e “credit”, con applicazioni alla gestione del rischio. Trattiamo un AJTSM di un’azione con un’unica discontinuità (“jump-to-default”), dove il tempo di fallimento dell’azione è un tempo aleatorio doppiamente stocastico con intensità determinata da un sottostante processo affine. Questo approccio permette una piena trattabilità analitica pur lasciando flessibilità nel definire le interazioni tra il prezzo dell’azione fallibile, la volatilità stocastica e l’intensità del fallimento. Infine caratterizziamo tutte le misure di rischio neutrale che conservano la struttura affine del modello e mostriamo che sia la gestione del rischio che i problemi del prezzaggio possono essere trattati in modo efficiente passando a misure di sopravivenza appropriate. Come esempio, estendiamo il modello di volatilità stocastica di Heston considerando la possibilità di un “jump-to-default”. [FR10b] R. Frey and W. J. Runggaldier, Pricing credit derivatives under incomplete information: a nonlinear filtering approach., Finance and Stochastics 14 (2010), no. 4, 495–526. [PR10a] E. Platen and R. Rendek, Quasi-exact approximation of hidden markov chain filters., Communications on Stochastic Analysis 4 (2010), 129–142. [BGR90] A. Bensoussan, R. Glowinski, and A. Rascanu, Approximation of the zakai equation by the splitting up method, SIAM Journal of Control and Optimization 28 (1990), no. 6, 1420–1431. [Gla92] F. Le Gland, Splitting-up approximation for spde’s and sde’s with application to non-linear filtering, in: Stochastic Partial Differential Equations and Their Applications, Charlotte 1991, B. L. Rozovskii and R. B. Sowers, editors, Lecture Notes in Control and Information Sciences 176 (1992), 177–187.

Aspects of Affine Models in the Pricing of Exotic Options and in Credit Risk / Montes, Juan Miguel. - (2014 Jan 24).

Aspects of Affine Models in the Pricing of Exotic Options and in Credit Risk

Montes, Juan Miguel
2014

Abstract

Le strutture a termine affine con diffusione a salti (AJTSMs) stanno recentemente ricevendo molta attenzione in finanza matematica, perché spesso è semplice analizzare le funzioni di distribuzione ad esse associate. Questa tesi riguarda tre diversi aspetti della finanza matematica, applicati su certe classi di AJTSMs. Il primo aspetto riguarda il problema del prezzaggio, nel caso particolare in cui il processo sottostante Xt sia una Catena Markoviana a Tempo Continuo (CTMC). Per opzioni esotiche, dove il “claim”, cioè il “payoff” del derivato è dipende dal tempo oppure dalle traiettorie, solitamente i prezzi devono essere stimati attraverso simulazioni di tipo Monte Carlo. Mostriamo che, quando si condiziona prima sul numero Nt,T=k dei salti della catena, il calcolo di questa stima si semplifica. Viene proposta una ricorsione per calcolare il valore atteso del “payoff” scontato, dato Nt,T=k; in seguito si calcola il valore atteso del “payoff” rispetto alla distribuzione di Nt,T=k attraverso un metodo Monte Carlo. Questo condizionamento comporta una riduzione della varianza. Presentiamo i risultati di vari test numerici, che indicano che, per diversi tipi di “claims”, il metodo proposto supera spesso un semplice “vanilla” Monte Carlo. Il secondo aspetto riguarda la calibrazione, cioè la stima dei parametri di un modello finanziario, dove il processo sottostante (una Catena Markoviana finita) è solo parzialmente osservabile tramite i prezzi corrotti del titolo. In questo lavoro, assumiamo che anche i salti del prezzo del titolo corrispondenti ai tempi dei salti della catena Markoviana siano osservabili. Questo è un caso particolare della classe di modelli trattati in [FR10b]. I loro parametri possono essere stimati mediante l’algoritmo “expectation-maximization” (EM), seguendo l’approccio di [EAM08], che, nel caso delle catene a tempo discreto, coinvolge il filtro di Kalman. Estendiamo questo approccio al caso CTMC, usando invece il filtro di Wonham. Il contributo principale di questa parte della tesi è l’approssimazione numerica dei filtri e degli “smoothers” dell’algoritmo EM. Confrontiamo i classici metodi di Eulero e di Milstein con una nuova strategia, simile a [PR10a], che chiamiamo “soluzione quasi-esatta” e che è anche collegata al metodo di “splitting-up” di [BGR90] e [Gla92]. Dimostriamo che tale schema ha un ordine di convergenza forte di almeno 0.5 e che pertanto è almeno tanto efficace quanto lo schema di Eulero. Presentiamo alcuni risultati numerici che indicano che, di fatto, in certi casi il nuovo metodo converge più velocemente di entrambi i metodi di Eulero e di Milstein. Il terzo aspetto riguarda un quadro unificato per la modellazione del rischio di “equity” e “credit”, con applicazioni alla gestione del rischio. Trattiamo un AJTSM di un’azione con un’unica discontinuità (“jump-to-default”), dove il tempo di fallimento dell’azione è un tempo aleatorio doppiamente stocastico con intensità determinata da un sottostante processo affine. Questo approccio permette una piena trattabilità analitica pur lasciando flessibilità nel definire le interazioni tra il prezzo dell’azione fallibile, la volatilità stocastica e l’intensità del fallimento. Infine caratterizziamo tutte le misure di rischio neutrale che conservano la struttura affine del modello e mostriamo che sia la gestione del rischio che i problemi del prezzaggio possono essere trattati in modo efficiente passando a misure di sopravivenza appropriate. Come esempio, estendiamo il modello di volatilità stocastica di Heston considerando la possibilità di un “jump-to-default”. [FR10b] R. Frey and W. J. Runggaldier, Pricing credit derivatives under incomplete information: a nonlinear filtering approach., Finance and Stochastics 14 (2010), no. 4, 495–526. [PR10a] E. Platen and R. Rendek, Quasi-exact approximation of hidden markov chain filters., Communications on Stochastic Analysis 4 (2010), 129–142. [BGR90] A. Bensoussan, R. Glowinski, and A. Rascanu, Approximation of the zakai equation by the splitting up method, SIAM Journal of Control and Optimization 28 (1990), no. 6, 1420–1431. [Gla92] F. Le Gland, Splitting-up approximation for spde’s and sde’s with application to non-linear filtering, in: Stochastic Partial Differential Equations and Their Applications, Charlotte 1991, B. L. Rozovskii and R. B. Sowers, editors, Lecture Notes in Control and Information Sciences 176 (1992), 177–187.
24-gen-2014
Affine jump-diffusion term structure models (AJTSMs) are recently receiving much attention in mathematical finance, because they often lead to a tractable analysis of the price distribution functions. This thesis concerns three aspects of mathematical finance, when applied to certain classes of AJTSMs. The first aspect concerns the pricing problem in the special case when the underlying process Xt is a Continuous-Time Markov Chain. For exotic options, where the claims are time or path dependent, prices can only be estimated by Monte-Carlo simulation, in most cases. We show that this computation is simplified by conditioning first on the number Nt,T of the jumps of the chain. A recursion is proposed to compute the expected discounted payoff given Nt,T=k; Monte Carlo is then used to average out the result over the distribution of Nt,T=k. This leads to a variance reduction by conditioning. We present results of numerical tests which indicate that the method often outperforms plain vanilla Monte Carlo for different kinds of claims. The second aspect concerns the calibration of a financial model by parameter estimation, when the underlying, a finite state Markov chain, is only partially observed through noisy asset prices. Here, we assume that the jumps of the asset price occurring at the jump-times of the Markov chain are observable as well. Such a model is a special case of the class of models treated in [FR10b]. Their parameter estimation can be addressed via the EM algorithm, following the approach by [EAM08] which, in the case of discrete-time chains, involves the Kalman filter. We extend this approach to to the case of CTMCs via the use of the Wonham filter. Our main contribution is the numerical approximation of the filters and smoothers in the EM algorithm. We compare the classical Euler and Milstein schemes to a new scheme, inspired by [PR10a], that we call a quasi-exact solution and is related to the splitting-up method of [BGR90] and [Gla92]. We prove that such a scheme is of strong convergence order at least 0.5, hence it performs at least as well as the Euler scheme. We present numerical evidence indicating that in fact, in certain cases the method outperforms both the Euler and the Milstein scheme. The third aspect concerns a unified framework for equity and credit risk modeling, with applications to risk management. Here we treat an affine jump-diffusion model with a single jump-to-default, where the default time is a doubly stochastic random time with intensity driven by an underlying affine factor process. This approach allows for flexible interactions between the defaultable underlying asset price, its stochastic volatility and the default intensity, while maintaining full analytical tractability. We characterize all risk-neutral measures which preserve the affine structure of the model and show that risk management as well as pricing problems can be dealt with efficiently by shifting to suitable survival measures. As an example, we consider a jump-to-default extension of the Heston stochastic volatility model. [FR10b] R. Frey and W. J. Runggaldier, Pricing credit derivatives under incomplete information: a nonlinear filtering approach., Finance and Stochastics 14 (2010), no. 4, 495–526. [PR10a] E. Platen and R. Rendek, Quasi-exact approximation of hidden markov chain filters., Communications on Stochastic Analysis 4 (2010), 129–142. [BGR90] A. Bensoussan, R. Glowinski, and A. Rascanu, Approximation of the zakai equation by the splitting up method, SIAM Journal of Control and Optimization 28 (1990), no. 6, 1420–1431. [Gla92] F. Le Gland, Splitting-up approximation for spde’s and sde’s with application to non-linear filtering, in: Stochastic Partial Differential Equations and Their Applications, Charlotte 1991, B. L. Rozovskii and R. B. Sowers, editors, Lecture Notes in Control and Information Sciences 176 (1992), 177–187.
Derivative Pricing, Exotic Options, Credit Risk, Affine Models, Markov Processes, Stochastic Filtering, Monte-Carlo
Aspects of Affine Models in the Pricing of Exotic Options and in Credit Risk / Montes, Juan Miguel. - (2014 Jan 24).
File in questo prodotto:
File Dimensione Formato  
montes_juan_miguel_tesi.pdf

accesso aperto

Tipologia: Tesi di dottorato
Licenza: Non specificato
Dimensione 3.41 MB
Formato Adobe PDF
3.41 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/11577/3423536
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact