posted on 2017-03-02, 23:36authored bySakutukwa, Tutsirai
This dissertation consists of three essays focusing on linking the capital market with the macroeconomy, and use this knowledge in forecasting some macroeconomic variables.
The first essay analyses whether implied stock market volatility could be used to forecast both stock market returns and future output. While addressing the potential issues of autocorrelation and heteroskedasticity due to overlapping data, we examined the forecasting information in implied stock market volatility for seven developed economies (Belgium, France, Germany, Japan, the Netherlands, the UK, and the US). In-sample results showed that implied stock market volatility had forecasting power for stock returns and output, however the forecasting power of the information content differed by country. By splitting the sample, as dictated by structural break tests, we showed that the output forecasting information content in implied stock market volatility was stronger in recent years than earlier years. We further explored the out-of-sample forecasting properties of implied stock market volatility for Germany and the US, and showed that the forecasting information was stronger for Germany. However, the out-of-sample predicting power of output by holding period stock returns performed relatively better for the US than for Germany. In addition, combined forecasts out-performed any single predictor alone.
The second essay tests the in-sample and out-of-sample forecasting of output growth rates using the composite leading business cycle indicator (CLI) for South Africa, and then compares the results to the traditional yield curve approach. Compared to both the yield curve and the benchmark model (lagged output growth), the CLI contained more in-sample forecasting information. Using squared forecasting error ratios of the benchmark, the yield curve, and the CLI models, together with an econometric approach that addressed the upward bias present in a nested model (overlapping models), we found evidence that the CLI outperformed both the yield curve and the benchmark model in the out-of-sample forecasting. In addition, as shown by the diagnostic plots, the CLI was more consistent and stable in out-of-sample forecasting than the yield curve. However, compared to the benchmark model, the yield curve’s forecasting ability was more accurate.
In the third essay, the main objectives are to evaluate the intra-relationship between the volatility of financial variables and to investigate their link with macroeconomic volatility for six industrialised countries (Belgium, France, Germany, Japan, the UK, and the US). The results of the analysis indicated that implied stock market volatility (VIX) led both the Treasury bill rate volatility (TB) and term structure volatility (TS). Even though TB led TS, this link was relatively weak. On the relationship between finance and the macroeconomy, our empirical results were positioned between the findings of Schwert (1989) and Diebold and Yilmaz (2008). The relationship between VIX and inflation was statistically significant in all countries but the UK and France. In addition, uncertainty in an economy measured by real output growth volatility and inflation rate volatility could be predicted by VIX. The results showed that there was a statistically significant relationship between TB and the volatilities of real GDP growth rate for all countries, except for Japan and the US. Conversely, the relationship between TB and inflation rate volatility was weak for all countries.