1 8 S e p t e m b e r V o l u m e 8 3 1

Transcription

1 FUNDS ON FRIDAY b y G l a c i e r R e s e a r c h 1 8 S e p t e m b e r V o l u m e We are very aware of the fact that stock market corrections are often triggered by news flow either positive or negative. Some of the drivers of stock market corrections can result from investor perceptions triggered, for example, by policy pronouncements by politicians. At times, stock market prices may react to data flow, such as economic indicators. Other markets, such as money markets, forex markets or offshore equity markets, may all have an effect either positive or negative on our local stock market. Which begs the question: Statistically speaking, which data releases or market indicators best explain JSE share price developments? Unfortunately, many economic and market data time series are often correlated meaning they influence one another or a range of market factors. So one would hardly ever have a data series moving in isolation and then for this factor alone to affect stock prices. Usually there is some causality or sequence of events at work. For instance, inflation may start to pick up as a result of credit demand which has outstripped the growth in GDP. This causes a tightening in monetary policy hence interest rates are hiked. This causes the rand to strengthen. In the absence of other external market developments, these factors will usually have an impact on stock prices. But this impact will vary depending on the magnitude of the relative changes. Higher interest rates are often seen as stock market negative while a stronger rand may be perceived as positive. This article will attempt to statistically measure the strength and direction (positive or negative) of the impacts of a number of factors on JSE share prices (the ALSI) using simple regression analysis. Page 1

2 Where possible, logarithms of data were used in order to interpret changes in the explanatory variables vis-a-vis changes in the dependent variable. Price behaviour data are also often exponential, despite deflating the time series with a price index, such as the CPI, making the logarithms of the series a better option to use in regression analysis. In all graphs that follow, the ALSI is on the vertical (y) axis. This graph shows the correlation between the nominal ALSI price index and the nominal earnings index of the ALSI. The regression coefficient (R-square) is quite high at around 92%, although some clustering of values above and below the regression line can be observed. The problem is that this good correlation may be due, to a large extent, to inflation. With inflation stripped out of the price indices, the R2 value remains fairly high, with the variation in real earnings explaining around 87% of the variation in the All-share price index. Investors are of course hoping for rising real earnings because this would be indicative of a possibility of higher profits, higher dividends and higher share prices in future. Page 2

3 The fact that real share prices are positively correlated with short term interest rates may come as somewhat of a surprise. The reason is that investors are perceived to be trading off the risk / return characteristics of the different asset classes. Therefore, if interest rates rise, equity investments should be sold off in favour of cash. However, the regression equation does not bear this out. Perhaps higher interest rates are seen to be inflation positive; hence growth positive over the longer term and also positive for equities. The regression between bonds and equity prices looks more in line with what is expected: Higher bond yields (lower bond prices) is generally thought to be positive for equity prices and vice versa. Higher bond yields are often an indication that state finances are under pressure because of a higher fiscal deficit and hence a bigger need to for the state to obtain loan finance. Higher bond yields are often also indicative of rising inflationary pressures. Page 3

4 A yield curve is usually calculated as the difference between short-term and long-term interest rates. A normal or positive yield curve prevails when long-term bond yields are higher than short-term interest rates. This is normal in the sense that investors would want to be compensated for the risk of having an exposure to a long-term investment. When inflation is high or rising, an anti-inflationary monetary policy stance may cause higher short-term interest rates and perhaps an inverse yield curve. Regression of the yield curve on share prices in the South African context, did not show a strong correlation between these two variables. The real effective exchange rate (REER) is the trade weighted average exchange rate after relative inflation rates of all countries have been taken into account. A rising REER would imply that the rand, compared with the currencies of our trading partners, is appreciating after inflation has been stripped out from the calculation. The regression shows a slightly negative slope, suggesting that a stronger currency will lead to some downward pressure on share prices. However, note the clustering of values above and below the regression line. Page 4

5 The clustering of x-y values referred to in the graph above, would indicate that the earlier part of the regression (which started in 1995) may have had a different regression result compared with the latter period (after around 2005). This would indeed appear to have been the case when one observes the graph to the right. The period shows a materially stronger R2 value (indicating that the change in the REER explains, on its own, some 24% of the change in the ALSI). The reason why the REER impact might be getting stronger, is to be found in the fact that many big South African firms have dual offshore listings and operations abroad. A weaker rand is actually then good news for South African shareholders in these companies. These shares are therefore regarded as randhedgers. The clustering of x-y values referred to in the graph above, would indicate that the earlier part of the regression (which started in 1995) may have had a different regression result compared with the latter period (after around 2005). This would indeed appear to have been the case when one observes the graph to the right. The period shows a materially stronger R2 value (indicating that the change in the REER explains, on its own, some 24% of the change in the ALSI). The reason why the REER impact might be getting stronger, is to be found in the fact that many big South African firms have dual offshore listings and operations abroad. A weaker rand is actually then good news for South African shareholders in these companies. These shares are therefore regarded as randhedgers. Page 5

6 Using data between 1995 and 2015, it would appear that there is a relatively strong correlation between offshore markets (with the New York S&P500 being used as proxy) and the local ALSI. However, again note the clustering of lower (i.e. earlier) ALSI values and later ALSI values. The period is shown in the next graph. As is the case with the exchange rate, using only the period , indicates a much stronger relationship between S&P500 prices and the ALSI. The result suggests that as much as 81% of the change in the ALSI can be explained by changes in the S&P500. Unfortunately GDP data are only available on a quarterly basis. To regress the ALSI on domestic GDP data, the latter was interpolated to arrive at monthly data values for the GDP. It would nevertheless appear that the correlation is quite strong. As is to be expected, a better performing economy will have quite a strong impact on the ALSI. Page 6

7 The impact between the ALSI s real earnings index and the GDP is even stronger a very good fit indeed. But this probably does not come as a surprise since the GDP shows value of all final goods and services produced within a particular period in South Africa. Listed companies form an important part of this production of goods and services. Conclusion GDP, company earnings and offshore market prices (in particular the S&P500), would appear to be some of the more important drivers of local share prices. Other asset prices (cash and bonds), as well as the exchange rate, and seem to be of lesser importance in explaining changes in JSE share prices. Page 7

8 Glacier Research would like to thank Christo Luüs for his contribution to this week s Funds on Friday. Christo Luüs BCom Econometrics (Hons), MBL Christo is the economist at Third Circle Asset Management. He studied econometrics/economics at RAU (now University of Johannesburg) and an MBL through Unisa. He worked at Absa (initially United Building Society) for 23 years between 1998 and 2007 as chief economist. He started his own economic consultancy firm in 2008 conducting mainly economic impact assessments for large projects. He is currently involved in macro-economic modelling and research as well as financial market and sectoral research. He has been an economic and markets advisor for Third Circle Asset Management since He has co-authored a number of books on personal finance and financial instruments and is currently the chairman of the South African Institute of Financial Markets. Page 8