It is often a goal of the risk management of a portfolio of interest rate sensitive instruments to minimize the impact of movements in market rates on the value of the portfolio. This can be done by considering the sensitivity of the portfolio to each of the market rates that are used to bootstrap a yield curve. However, this is likely to lead to an excessive amount of trading due to an investment in a large number of hedging securities. As an alternative, we consider using principal components analysis (PCA) to condense most of the variability in the market rates into a much smaller number of risk factors, called the principal components. One can then construct a hedging portfolio so as to make the portfolio immune to shocks in these principal components, and hence to the most common movements in the yield curve. We compare the effectiveness of these two hedging strategies for hedging a portfolio of interest-rate options, both in the absence and presence of transaction costs. We also consider the additional feature of being able to update each hedging methodology on a daily basis and rebalance the hedge portfolios accordingly.

One of the central debates in finance concerns the Efficient Market Hypothesis (EMH)—wherein markets are assumed to be efficient in the absolute sense. However, the possibility of time-varying weak-form market efficiency has received increasing attention in recent years. Under the Adaptive Market Hypothesis (AMH) it is postulated that market efficiency is dynamic, which advocates using models with non-constant coefficients. The concept of evolving efficiency has yielded a Test for Evolving Efficiency (TEE) and following that, a Generalised Test for Evolving Efficiency (GTEE) – both with an associated Kalman filtering (KF) technique. Unfortunately, these methods assume that the inherent stochastic processes are Gaussian despite widespread evidence that many real financial time series are nonGaussian. Unlike the classical KF, modern filters such as the maximum correntropy Kalman filters (MCC-KF) have been shown to be less sensitive to non-Gaussian uncertainties. These filters utilise a similarity measure known as correntropy– which incorporates higher order information than the mean square criterion that is utilised in the classical KF. As a result, they have been shown to improve filter robustness against outliers or impulsive noises. In this paper, the South African and American stock markets are tested for adaptive market efficiency using both the standard KF and the MCC-KF. A simulation study shows that the MCC-KF is a more robust estimator of adaptive efficiency but it less accurately estimates unknown system parameters. The South African stock market is found to be inefficient prior to August 2004 but achieves efficiency thereafter. Testing the S&P500 does not provide evidence of inefficiency in the American stock markets. The GTEE, implemented with the MCC-KF, is selected as the bestperforming test for the S&P500.

No description available.

Volatility modelling is used predominantly in order to explain the volatility smile observed in the market. Stochastic volatility models are mainly used to capture the curvature of a volatility smile while local volatility models generally model the skew. Jackel and Kahl ¨ (2008) present a hyperbolic-local hyperbolic-stochastic volatility (Hyp-Hyp) model which aims to improve upon existing local and stochastic volatility models such as the stochastic alpha, beta, rho (SABR) and constant elasticity of variance (CEV) models. The advantageous features of the Hyp-Hyp model are corroborated by implementing the model. Jackel and Kahl ¨ (2008) investigate the accuracy of a scaled analytical approximation for implied volatility, based on approximations presented by Watanabe (1987) and Fouque et al. (2000), for the Hyp-Hyp model. They use the approximation to derive an expression for the delta of an option. This dissertation analyses the Hyp-Hyp model, as well as the approximation, by deriving expressions for other sensitivities and by investigating the effect of the Hyp-Hyp model parameters on the volatility smile. The accuracy of the analytical approximation for functional forms other than those defined by the Hyp-Hyp model is explored. A derivation of the approximation is undertaken, presenting corrections to the expressions introduced by Kahl (2007) and used by Jackel and Kahl ¨ (2008).

Since its development in 1997, the LIBOR market model has gained widespread use in interest rate modelling, largely owing to its consistency with the Black futures formula for pricing interest rate caps and floors. From its original construction(s), the LIBOR market model specifies a discrete set of forward rates that correspond to a fixed tenor structure, e.g. market tenors. This implies the pricing of interest rate contingent claims is restricted to claims with cashflow dates that coincide with the fixed tenor structure. In this light, several interpolation schemes have been suggested to handle the pricing restrictions, however at the cost of introducing possible arbitrage opportunities. The present dissertation studies four such interpolation schemes, paying particular attention to arbitrage-free interpolation schemes: Piterbarg deterministic interpolation, Schlogl deterministic interpolation, Schlogl stochastic interpolation, and Beveridge-Joshi stochastic interpolation.

Common applications of asset-pricing models in practice rely on recalibrating model parameters periodically for effective risk management. Yet, these model parameters are often assumed to be constant over time, thereby countering the notion of readjusting these values. A possible solution to this problem is to recalibrate at times where observed market prices cannot realistically match model prices based on parameter values at those times. This dissertation aims to test the effectiveness of a possible algorithm which can be used in optimally identifying such times. An overview is provided of the recently proposed particle filter with accelerated adaptation which has demonstrated rapid time detection for changes in parameter values and has been applied to regime-shifting and stochastic volatility models. Numerical and graphical evidence of parameter and volatility estimation will be provided under regime-shifting parameters for the Heston (1993) stochastic volatility model. The filter demonstrates rapid adaptation in estimating parameter values and accurate estimation of the volatility process. Furthermore, we provide a discussion for possible extensions towards a metric for optimal recalibration times.

Regulations require banks to hold a minimum amount of capital for market risk resulting from their trading operations and prescribe two approaches to calculating this minimum capital requirement: (i) a Standardised Approach (SA); and (ii) an Internal Models Approach (IMA). The global financial crisis of 2008 highlighted flaws in the Basel 2 regulatory framework used by banks to calculate market risk capital charges for trading operations. In 2009, Basel 2.5 was introduced to deal with some but not all of the flaws of Basel 2. Both Basel 2 and 2.5 use the Value at Risk (VaR) risk measure as the basis to determine IMA capital charges. From 2022 onwards, Basel 2.5 will be replaced by the Fundamental Review of the Trading Book (FRTB), a new framework for calculating market risk capital charges for trading operations. The FRTB replaces VaR with the Expected Shortfall (ES) risk measure in the IMA and introduces a new SA. This dissertation investigates the impact the FRTB will have on market risk capital charges for portfolios of linear South African interbank interest rate products. Capital charges are calculated for these portfolios under the Basel 2, Basel 2.5 and FRTB regulatory frameworks. A comparison and analysis of the resulting capital charges is then presented.

As a consequence of the 2007 financial crisis, the market has shifted towards a multi-curve approach in modelling the prevailing interest rate environment. Currently, there is a reliance on the assumption of deterministic- or constant-basis spreads. This assumption is too simplistic to describe the modern multi-curve environment and serves as the motivation for this work. A stochastic-basis framework, presented by Mercurio and Xie (2012), with one- and two-factor OIS short-rate models is reviewed and implemented in order to analyse the effect of the inclusion of stochastic-basis in the pricing of interest rate derivatives. In order to preclude the existence of negative spreads in the model, a constraint on the spread model parameters is necessary. The inclusion of stochastic-basis results in a clear shift in the terminal distributions of FRA and swap rates. In spite of this, stochastic-basis is found to have a negligible effect on cap/floor and swaption prices for the admissible spread model parameters. To overcome challenges surrounding parameter estimation under the framework, a rudimentary calibration procedure is developed, where the spread model parameters are estimated from historical data; and the OIS rate model parameters are calibrated to a market swaption volatility surface.

The financial crisis in 2007 highlighted the credit and liquidity risk present in interbank (LIBOR) rates, and resulted in changes to the pricing and valuation of financial instruments. The shift to Overnight Indexed Swap (OIS) discounting and multi-curve framework led to changes in the construction of interest rate zero curves, with the OIS curve being central to this methodology. Developed markets, such as the European (EUR), were able to adopt this framework due to the existence of a liquid OIS market. In the case of the South African (ZAR) market, the lack of such tradeable instruments poses the issue of how to construct or infer the OIS curve. Jakarasi et al. (2015) proposed a method to infer the OIS curve through the statistical relationship between SAFEX ROD and 3M JIBAR. The extension of the statistical relationship used by Jakarasi et al. (2015) to more statistically rigorous models, capable of capturing more information relating to the relationship between the rates, arises from the expected cointegrating relationship exhibited between rates. This dissertation investigates the implementation of such statistical models to infer the OIS curve in the ZAR market.

Includes bibliographical references / Forecasting stock performance has long been one of the primary objectives of financial practitioners. Literature has shown that the classical linear approach to modelling the interactions among company-specific factors and its stock market re- turns in time have become less suited for capturing the movements of the stock market. Hence, attempts to predict the performance of a stock have become associated with additional layers of complexity. This has led to the adoption of non-linear approaches to forecast stock performance. This dissertation explores the performance of some non-linear models in the South African market. These were classification and regression trees (CART), logistic regression and a random forest approach com- pared against a linear regression model. Moreover, a hybrid model between CART and logistic regression was considered. The models fell into two categories (i.e., static and dynamic models). Using a set of classification and portfolio performance metrics it was found that that a dynamic modelling approach outperformed a static approach. Overall, the logistic and linear regression models dominated in terms of performance against the tree-based models and hybrid approaches. The results also demonstrated that a hybrid approach offered an improvement over a stand-alone CART.