Following the recent recession, major global economies are still experiencing weak recoveries. The likelihood that the global economy may experience a double-dip recession driven by poor performance by advanced economies stresses the need for predicting the behaviour of leading indicators such as stock returns and equity premium. An understanding of market behaviour helps in guiding both policy and trading decisions. The main objective of this thesis is to assess the predictability, spillovers and determinants of stock returns in South Africa.
Stock returns are determined by a number of financial and macroeconomic variables including valuation ratios (price-earnings ratio and price-dividend ratio), payout ratio, interest rates, the term spread, stock returns of South Africa‟s major trading partners, the inflation rate, money stock, industrial production and the employment rate, world oil production, the refiner acquisition cost of imported crude oil, global activity index, industrial stock returns and financial stock returns. A number of econometric models are used in investigating the determinants, predictability and spillovers of the stock returns – including; predictive regressions using in-sample and out-of-sample test statistics (t-statistics, MSE-F and the ENC-NEW, , utility gains, forecasting encompassing test); exponential smooth-transition autoregressive; Monte Carlo simulations; data-mining-robust bootstrap procedure; in-sample general-to-specific model selection, bootstrap aggregating, combining method (simple averages, discounting, clusters, principal components, Bayesian regression methods under the Gaussian and double-exponential prior); sign restriction VAR and a TVP-VAR model specification with stochastic volatility.
The results show that firstly, the stock returns are determined by certain financial and macroeconomic variables (assessing both the statistical and economic significance). Secondly, South African stock returns react differently to different types of oil shocks – suggesting that the cause of the oil price shock is crucial in determining policy. The combination model forecasts, especially the Bayesian regression methods, outperform the benchmark model (AR(1)/random walk model). Further, the analysis does not only show evidence of significant spillovers to consumption and interest rate from the stock market, but, more importantly, it also highlights the fact that these effects have significantly varied over time.