Abstract
This thesis is an empirical investigation of an assessment of sovereign credit rating on the South African economy. This thesis applies classification techniques to analyse Sovereign Credit Ratings in their original symbol format without converting them to numerical values as adopted by previous literature (Bennell, Crabbe Thomas & Ap Gwilym, 2006; Kumar & Haynes, 2003; & Kräussl, 2005), by doing so, this study overcomes the errors that come with the conversion of symbol ratings into whole numbers. The application of outlook and microeconomic variables makes the study unique as previous studies focused only on macroeconomic indicators.
The study aimed at developing a forecasting model that predicts Sovereign Credit Ratings with utmost precision to assist governments to prevent credit downgrades and promote financial stability. Different econometric and machine learning models (Naïve bayes, Random Forest, logistic regression, principal component analysis, structural vector autoregressive, and stepwise regression) are used for this purpose. Quarterly data from 1999 to 2018 of macroeconomic variables and categorical Sovereign Credit Ratings from major credit rating agencies such as Fitch, Moody’s, and Standard & Poor’s were collected from Trading Economics data on country ratings, the South African Reserve Bank (SARB) data, Quantec Easy data, Statistics South Africa (Stats SA) and Thomson Reuters.
The findings of the study are enormously useful. Firstly, these findings suggest that household debt to disposable income ratio, exchange rates, and inflation are the most important variables for estimating and classifying credit ratings. Secondly, the findings show that Machine learning models generate higher forecast precision and classify credit ratings better than traditional econometric models. Thirdly, the findings show that improvements in economic indicators like Real Effective Exchange Rates, Gross Domestic Product Growth,
v
Household debt to disposable income ratio, and Consumer Price Index Headline result in favourable movements of ratings. These findings suggest that governments, authorities, central banks, and policy makers should try to maintain positive exchange rates movements, work to raise GDP growth, stabilise inflation, and boost credit systems to households to boost aggregate demand. Fourthly, the study finds that credit downgrades are likely to emanate from other non-quantitative issues like governance, corruption, and political instability, etc. Therefore, the study recommends that the South African Reserve Bank adopts a consistent macroeconomic policy framework that fosters a plethora of aspects, like a flexible exchange rate system, low stable inflation, sustainable monetary stability, and fiscal restraint, and constantly monitor the financial system through macro-prudential analysis of the corporate bonds and securities markets to build a strong financial market infrastructure that boosts the economic environment in which intermediaries operate.
Keywords: Sovereign Credit Rating, Rating Outlook, Financial Stability, Macroeconomic variables, Machine Learning, Linear Regression Model, Structural Vector Auto-Regression, Naïve Bayes, Random Forest, and Logistic Regression Model.