CHAPTER 3: A CHRONOLOGICAL REVIEW OF THE LITERATURE
3.3 The Transition to the Application of Cointegration Techniques
3.3.5 Hancocks (2010)
Hancocks (2010) conducted a similar study to the study by Moolman and du Toit (2005) in which a multivariate analysis was conducted in order to determine the magnitude of the effects on the All-Share, Financial, Mining and Retail sectors of the JSE by a series of macroeconomic variables. The sample period ranged between 1996 and 2008. In order to analyse the various relationships, the author employed Johansen’s approach to test for cointegration between the macroeconomic variables and the stock price index in question. The Vector Error Correction
Model (VECM) was used to establish the magnitude of the effects of each cointegrating vector and to determine the speed of adjustment coefficients in each case. The author hypothesized that the different macroeconomic variables would have varying effects on each particular index, and used the example that a change in interest rates would be more likely to have more of an impact on the retail and financial indices than on the resources sector, while a change in the Rand/US$ exchange rate would have the opposite effects.
The macroeconomic variables included in this study, for which the relationship with the All Share Index were analysed are listed as follows: money supply, the Rand/US$ exchange rate, long and short-term interest rates and finally inflation. The most applicable of these relationships to the analysis of the capacity for equities to act as a hedge against inflation is the relationship between the inflation rate and the All Share Index. In the study by Hancocks (2010) inflation was measured by the 12-month rate of change in consumer prices between 07-1995 and 11-2008, using the Consumer Price Index (CPI). Hancocks (2010) notes that between 1995 and 2001 inflation in South Africa exhibited a consistent downward trend until the onset of the US global economic crisis and terror attacks. Post 2001 the inflation rate exhibited a sharp increase due to rising oil prices and a decline in the Rand/US$ exchange rate, which was caused by decreases in foreign investment as a result of increased risks of global economic instability.
Following this inflationary spike, the Rand/US$ exchange rate proceeded to strengthen considerably and when combined with the adoption of an inflation targeting framework, resulted in a decline in inflation and stabilised the rate of inflation until mid-2007. At this point inflation rose to well above the 6% inflationary target band, an increase Hancocks (2010) attributes to increasing global inflationary pressures and rising global oil and food prices. This last paragraph highlights the relative volatility of the South African inflation rate and its vulnerability to international economic shocks.
Hancocks (2010) hypothesized that the relationship between inflation and stock returns is dependent on the current level of inflation in the South African economy. The author states that the relationship may exhibit contradictory movements, where in some cases stock returns increase alongside inflation and in others experiences a decrease when inflation increases simultaneously. This finding is explained by stating that if the inflation rate is within the target band and a rise in the money supply resulted in an increase in rate of inflation, but it remained within the target band, an increase in stock returns would occur as interest rates would not be adjusted by the Reserve Bank – leading to increases in stock prices caused by the increase in aggregate demand. On the other hand, should the inflation rate exceed the target band, interest
rates would be adjusted in order to reduce inflation to target levels leading to a decrease in share prices (Hancocks, 2010).
Using a standard Johansen’s approach to test for cointegration, Hancocks (2010) began by establishing the order of integration of the series. The Johansen’s procedure found evidence of the existence of cointegration, due to the potential for the presence of multiple cointegrating vectors. Finding the presence of cointegrating vectors, the Vector Error Correction Model (VECM) was estimated, which allows for the restriction of the long-run behaviour of the endogenous variables in order to allow for convergence on their cointegrating relationships, while measuring the short-run adjustments that occur within the system. Following the VECM test the author conducted impulse response tests, which describe the effect shocks to a variable would have on the variables in a system. Finally, a variance decomposition test shows the effect of each variable on the other variables included in the model.
The ADF and KPSS tests indicating that the variables were integrated of the first order, I(1).
In this particular study it was found that the various information criterion indicated the use of different and conflicting lag lengths as being optimal. Hancocks (2010) used the criterion starting from the smallest selected lag order and sequentially increased the lags until cointegration was found with serially uncorrelated residuals. In this case the optimal VAR lag order was found to be 4 lags. Evidence was discovered to support at least one cointegrating relationship between the variables. The VECM results indicated that the money supply and the exchange rate variables had a positive and significant relationship with the All Share Index.
Hancocks (2010) discovered a negative and significant relationship between inflation and the All Share Index and the financial sector index; however, it was discovered that the relationship between inflation and the retail sector was positive. It should be noted at this point that the inflation coefficient was -9.819778, indicating that the relationship between inflation and the All Share Index would be nearly 1:10. This indicates that for a 1% increase in inflation, the All Share Index would decrease by 10%, a level that does not seem possible in reality. For example, should an exogenous shock occur that increases inflation by 3%, a shift that has not been uncommon in the last 30 years in South Africa considering its inflationary volatility, stock returns would decrease by a marked 30%. This relationship therefore seems to be far too high to occur in reality, especially considering that the majority of other studies have determined results between 0 and 2.5% in the extreme case, for a wide range of countries (Alagidede and Panagiotidis, 2010). The error correction term indicated efficient adjustment of the All Share
Index, of 14.4% per year towards the long-run-equilibrium. The findings of the negative relationship over the period between inflation and the All Share Index are consistent with the theory of the Proxy Hypothesis, in that inflation decreases asset prices, although as previously noted, the relationship determined in the study exceeds expectations by too great a factor to be consistent with the previous literature.