Abstract
The study utilises a time series econometric approach to determine if there is a link and causal relationship between the stock exchange and the real economy of South Africa. Various hypothesis testing procedures against a set of six predetermined macroeconomic variables for South Africa were adopted. To address the main research question and objectives of the dissertation, a five-step econometric procedure is adopted to test Granger causal directionality and cointegration amongst the selected dependent and independent variables over the short and long term. The time frame considered for the research is set at a 25-year period using quarterly data providing 104 observations per variable. The study uses two models to establish the link and causal relationships. The two models created each have differing dependent variables but are tested against the same set of independent variables. The dependent variables selected are real GDP and the stock market (FTSE/JSE ALSI). The two models are effectively compared, and similar results are concluded from the analysis with regards to cointegrated relationships. Cointegrated relationships have been found and from the various econometric testing processes employed, the link between the stock market and GDP has been established using a set of macroeconomic variables. The critical differences in the two models are the strength of the statistically significant relationships formed between the dependent and independent variables. Model 1 uses GDP as a dependent variable with stock market being one of the dependent variables. Model 2 uses stock market as the dependent variable as GDP as one of the independent variables. The strength of the relationship between GDP and the stock market is less statistically significant in Model 1 than in Model 2. This result has indicated that GDP has a larger impact on the stock market as per the results of Model 2 than what the stock market has on GDP in Model 1. The study also analysed causal linkages using short- and long-term Granger causal testing. The short run results indicate that using GDP as a dependent variable does not form any short run causal relationship. Using the stock market as the dependant variable would yield short term causal relationships in the short run. In the long run Granger assessments causal relationships are found between the stock market and GDP. The results would indicate a link and long-term causal relationship between the stock market and real GDP.
vi
Keywords:
Vector Error Correction Model (VECM), Granger causality, coefficient vector, variance decomposition, impulse response function (IRF), cointegration.