Abstract
The 2007-2009 financial crisis have shown the importance of monitoring systemic risk in the financial system. In this regard, this thesis analyses systemic risk in the South African insurance sector and its potential effects on the South African economy. Thus, the first empirical chapter of this thesis examines the extent to which individual insurers are systemically important in South Africa. To this end, we use a dynamic mixture copula model to estimate the marginal expected shortfall of five South African insurance companies (Sanlam, Santam, Discovery, Liberty, and Momentum). Using a sample of daily equity prices spanning the period from November 2007 to June 2020, our results reveal that in South Africa, Sanlam is the most systemically important insurer, followed by Discovery, and that Santam contributes little to systemic risk in the South African insurance sector. Thus, there is a need for regulators to implement stricter regulatory measures such as higher capital and loss absorbency requirements for systemically important insurance companies in South Africa.
In the second empirical chapter, this thesis focuses on the link between systemic risk in the insurance sector and the economic activity in South Africa. In doing so, we use the equity prices of five insurers in South Africa and the South African volatility index to construct a variety of systemic risk measures of the South African insurance sector. The sample period spans from November 2007 to June 2020 and contains the 2007-2009 financial crisis data. We then employ the principal component quantile regression to aggregate the systemic risk measures into a composite stress index of systemic risk for the whole South African insurance sector. Finally, we assess the ability of the proposed index to forecast future economic downturns in South Africa. Our findings indicate that the proposed index possesses the ability to significantly predict future economic downturns in South Africa. The results of this study suggest that regulators must develop an analysis of systemic risk in the insurance sector with particular attention to its effect on the real economy. Moreover, our index can be used by regulators as a tool to monitor and mitigate systemic risk in the insurance sector.
In the last empirical chapter, we investigate the performance of different quantile regression-based Machine Learning models in predicting the lower tail risk of South African GDP growth. These models include the Kernel quantile regression, the quantile regression forest, the quantile regression neural network, the gradient boosting model, and the traditional quantile regression model used as a benchmark model. We assess the performance of the different models using the Coefficient of
vii
Determination and the Root Mean Square Error. We employ the equity price data of five insurers, the US$/ZAR exchange rate, the South African volatility index, and a mixture of short- and long-term interest rates, spanning the period January 2008 to June 2020. We find that quantile regression-based Machine Learning models perform well in predicting the lower tail risk of South African GDP growth than the traditional quantile regression model. However, the Kernel quantile regression model is the best in forecasting the South African GDP growth lower tail risk. These results suggest that regulators and risk managers should employ alternative prediction models to select the ideal strategy for managing systemic risk in the South African insurance sector.
Keywords: dynamic mixture copula, marginal expected shortfall, systemic risk, insurance sector, quantile regression, macroeconomy, machine learning, systemic risk, forecasting risk.