The influence of the market on inflation, not the other way around

ABSTRACT

The study of return prediction is fundamental to investors. However, inconclusive evidence exists as to whether returns on the South African (SA) stock market may be explained by movements in SA or international macroeconomic variables. This study investigates integration between macroeconomic variables and the JSE ALSI. Using a monthly dataset from 1995–2016, the study is able to update the determination of integration relationships and reduce the ‘noise’ prevalent in prior research. Unit root, correlation, integration, causality and a vector error correction model were applied. The study identified that the ALSI was statistically significant in explaining SA inflation. The direction and significance of this relationship is of interest to investors and financial economists. If the ALSI has a predictive relationship with inflation, then market performance could impact the decisions made to raise or drop the Repo rate. In addition to the determination of the integration relationships, this study informs researchers on the efficiency and predictability of the SA market.

KEYWORDS:

• macroeconomic variables
• JSE
• FTSE
• integration
• VECM
• interest rates
• exchange rate
• inflation

Introduction

Finance revolves around how information is absorbed into the market and how it can be used to predict future returns. The study of return prediction is fundamental to investors and past research has focused efforts on determining whether asset markets are efficient. Fama (1965) used the Efficient Market Hypothesis theory (EMH) to provide a framework to better understand the relationship between market information and market returns. Asset pricing models such as the Capital Asset Pricing model (CAPM; Sharpe, 1964) and the Arbitrage Pricing Theory model (APT; Ross, 1976) have assisted investors in identifying and taking advantage of mispriced securities. Stock price movements occur within an economy and thus should be affected by movements in other market factors, namely macroeconomic variables. Integration analyses provides insight into both the predictability of returns and the directional relationship and timing over which variables interact with each other.

Currently, inconclusive evidence exists as to whether returns on the South African (SA) stock market may be explained by movements in SA or international macroeconomic variables. Therefore, the following research question is put forward: how are local and global macroeconomic variables integrated with the All Share Index (ALSI), and what is the direction and significance of any relationships?

This study adds to the body of research by exploring the relationship between the stock market and selected macroeconomic variables: SA and US Inflation, SA and US interest rates, rand USD exchange rate, SA Money supply and the FTSE all share index. Monthly data for a 20-year period from July 1995 to December 2016 is used. This is a more recent dataset than in prior studies and the majority of prior studies tested quarterly data. The increased frequency of the data in this study reduces ‘noise’ prevalent over longer periods which may have compromised findings of prior research. The use of unit root, correlation, integration, causality and a vector error correction model also allowed for robust statistical testing.

The study identified that the macroeconomic variables do not have statistically significant explanatory power for the ALSI. However, the ALSI was statistically significant in explaining South Africa inflation. The direction and significance of this relationship is of interest to investors and financial economists. If the ALSI has a predictive relationship with inflation then market performance could impact the decisions made to raise or drop the Repo rate.

Understanding integration and Granger causal flows is also of interest to financial economists who would seek to describe behaviour from a policy perspective (Alexander, 2008). Providing more conclusive evidence in this field will enable increased confidence in understanding policy impacts and assist in making more informed investment decisions. Repo rates are used to manage inflation in South Africa and thus the relationship between inflation and markets is relevant. It is also important to understand what the impact of US economic factors has on the JSE as it could impact foreign investment and economic growth in South Africa.

Literature review

An investor invests with the objective of earning a return which consists of two parts: dividend returns and capital appreciation. Both are influenced by company exposure to systematic (macroeconomic) and unsystematic (firm-specific) risk (Sharpe, 1964). The ability to earn above-market returns is in the interest of investors and, consequently, effort has been focused on finding predictive measures.

To earn returns on the stock market the investor attempts to predict movements in stock prices or take advantage of mispriced securities. Thus, the investor needs a means to determine the true value of an asset. A share in a company is considered to have an intrinsic value at any point in time, which is dependent on the company’s future expected earnings, which, in turn, are affected by the economic environment in which the company operates (Fama, 1965). This evaluation implies that stock prices and changes in those stock prices should be related to movements in macroeconomic variables. An investor’s ability to take advantage of the influence of macroeconomic variables on stock prices is conditional on the level of efficiency within the market.

Asset pricing theories

The EMH (Fama, 1965) helps illustrate whether it is possible to predict returns on investments given the level of efficiency in the market. The EMH is closely linked to both CAPM (Sharpe, 1964) and APT (Ross, 1976) which provide a framework for determining that equilibrium asset value.

Central to CAPM is the notion of risk having two constituents: systematic and unsystematic risk. Systematic risk is common to all assets (Sharpe, 1964). It is determined by general influences in the market, over which the investor has no control; thus, market changes due to macroeconomic variables may inform asset risk. It follows that a relationship between macroeconomic variables and asset prices should exist. While CAPM attempted to find asset prices based on risk and return, APT suggests that assets should be fairly priced, due to the opportunity for arbitrage within the market (Shanken, 1992).

There has been much research into the factors that influence pricing and the relationship between different assets. This is important as accurate pricing models aid investors’ ability to earn returns in the stock market. However, there is limited consensus as to whether returns are predictable, presenting the need for this study. The predictability of stock returns has important implications for capital asset allocation and the ability to test for market efficiency.

As developed markets generally have higher levels of liquidity, stronger economic policy and efficient capital flows they should exhibit a level of market efficiency as defined by Fama (1970). Thus, there is the expectation that the increased efficiency will result in fewer relationships between the market and macroeconomic variables and, consequently, less predictability in the stock market. Despite this expectation, such relationships have been identified.

Real economic activity has long been considered a driver of stock market returns. Fama (1990) theorised that the relationship between current stock returns and future production levels provided information about future cash flows for an investor. This relationship was further confirmed by Schwert (1990) who performed similar testing on an increased sample size in the US. Drawing on both studies there appears to be value creation opportunities in utilising economic growth data when analysing stock returns.

Inter-linkages between different markets and economic factors

Since the seminal study by Grubel (1968) which investigated inter-linkages between 11 geographically and economically diverse countries, this area of research has become one of the most extensively discussed topics of financial literature. The combination of predictive variables and global market integration has led to combined testing of macroeconomic factors globally with stock markets. To test the validity of long run relationships between stock markets and macro-economic variables Granger (1986) and Engle and Granger (1987) proposed using cointegration techniques. This method was then applied to a number of studies in Europe (Mukherjee & Naka, 1995; Cheung & Ng, 1998; Nasseh & Stauss, 2000; McMilliam, 2001; Chaudhuri & Smile 2004). The development of this testing with accompanying findings are discussed separately across developed and emerging markets and finally, South Africa.

Developed markets

Wong, Khan, and Du (2005) examined long-term and short-term equilibrium relationships between the major stock indices and selected macroeconomic variables in Singapore and the US between 1982 and 2005. Using Johansen’s cointegration, the results indicated a long run equilibrium relationship between the major stock indices and interest rates and money supply in the Singaporean market, but not in the US. Similarly, Humpe and Macmillan (2009) used the Johansen procedure and a vector error correction model (VECM) in Japan and North America. Despite the similarities in the variables and markets, the results differed. Humpe and Macmillan (2009) found no long run equilibrium relationship with money supply. This might be owing to the longer time period (1965–2005) examined by Humpe and Macmillan (2009). Testing these same relationships, but over a later time period (2008–2014), Jareño and Negrut (2016) noted further differences. This indicates that despite both Japanese and US markets being considered developed, there is a lack of consistency in significance of the relationships.

These outcomes may be due to the differences in the type of tests performed, the period covered, or because the examined markets, being developed, should exhibit higher levels of efficiency. According to the EMH (Fama, 1965) this would result in little to no opportunity for predictability. As developing markets grow and undergo technological changes it is likely that they will become more efficient. This increased efficiency may change the way the market responds to macroeconomic events and, thus, it is likely that there will be a difference in the ability of these factors to be predictors of returns.

Emerging markets

Inconsistencies are also noted within emerging markets. An early study over the period 1976–1997 was performed by Muradoglu, Taskin, and Bigan (2000). Using Granger causality, 19 emerging markets were examined, and the results were different for each of the countries. Looking at a more recent period, Barakat, Elgazzar, and Hanafy (2016) examined the stock markets of Egypt and Tunisia (countries not included in the study by Muradoglu, Taskin, and Bigan (2000)). Using both Johansen cointegration and Granger causality they concluded that interest rates, exchange rates, CPI and money supply all have either a long run relationship or causal relationship with the stock markets.

In India, Trivedi and Behera (2012) found significant relationships between equity prices on the Bombay stock exchange and selected macroeconomic variables. This analysis was performed using Johansen cointegration and the VECM framework. Using the same method of analysis, Tripathi (2014) used data from 1997–2011 and found significant short and long run relationships with inflation rates, interest rates and the exchange rate. In both studies, interest rates held significance, indicating that movements in the Indian stock exchange may be linked to monetary policy (Tripathi, 2014). This suggests that individual countries might have unique considerations that are not controlled for in meta-analyses.

South Africa

Considering SA and its neighbours, Jefferis and Okeahalam (2000) examined the impact of both domestic and foreign macroeconomic factors on returns in the SA, Zimbabwean and Botswanan stock markets using Johansen cointegration method and an error correction model. The international factors selected were trade related as these variables were deemed most likely to impact international markets. The findings were different for each of these markets with only domestic economic factors being common among countries. The impact of foreign macroeconomic factors varied depending on the size, market orientation and openness of the economy, and the size and liquidity of the stock exchange (Jefferis & Okeahalam, 2000). While the foreign economic variables were considered important, the ones related to trade rather than capital flows were of most importance. This indicates little integration between these three capital markets.

SA has had relatively tight exchange rate controls in capital account transactions in the past (Moolman & Du Toit, 2005). Thus, it would be expected that real exchange rates would correlate with the market, given the relative openness of the SA economy in terms of international trade (Jefferis & Okeahalam, 2000). Jefferis and Okeahalam (2000) found this relationship to hold true. More recently, Chinzara (2011) showed that volatilities in short-term interest rates and exchange rates were of highest significance and that the strength of this relationship was heightened during periods of increased volatility (the period assessed included the 2008 financial crisis).

Similar to the review of literature in developed markets, differences in results are noted when the frequency of the data and the stage of the economic cycle tested are different. Bonga-Bonga and Makakaule (2010) performed a smoothed transition regression model and while international market variables were found to hold a significant relationship with the SA market, they were not identified as predictors of returns. Bonga-Bonga and Makakaule (2010) used weekly data from 1988 to 2006 which resulted in a large dataset. Moolman and Du Toit (2005) also noted that the strength of the relationship between stock returns and macroeconomic variables was dependent on the stage in the economic cycle that the market was in.

Khumalo (2013) looked at the impact of select variables on stock prices using an auto-regressive distributed lag model and VECM. Using quarterly data between 1980 and 2010, Khumalo (2013) concluded that there was a strong relationship between all variables and the stock prices and specifically a negative relationship between stock price and inflation. Further assessment by Phiri (2016) showed that inflation has a non-linear impact on stock returns. Phiri (2016) investigated the asymmetric cointegration between inflation and the JSE using monthly data from 2003–2014 using a momentum threshold autoregressive model.

Finally, Shawtari et al. (2016) examined the long-term relationship between the SA stock market and selected macroeconomic variables between 1988 and 2010 using Johansen cointegration and a VECM. The inflation rate, rand/dollar exchange rate, real economic activity and money supply all showed a significant relationship with stock prices. However, real economic activity was found to be the most significant determinant of stock prices.

Despite the evidence of relationships between macroeconomic variables and stock returns, predictability within the SA economy also shows mixed results. There is some consistency between variables such as interest rates and exchange rates, but other variables are less consistent. This study will explore these relationships further, providing evidence on a more recent data set with high frequency, so as to add value and insight to policy holders and investors.

Global markets do not operate in isolation and with the growing speed of globalisation, it is important to consider the impact global market factors have on each other. These relationships are important for an investor to understand when making investment decisions. Chinzara and Aziakpono (2009) analysed the returns and volatility linkages between the SA stock market and leading world equity markets. There were both returns and volatility links between SA and the Australian, Chinese and North American stock markets. Thus, there is some evidence of the impact of global markets on returns in the SA market (Chinzara, 2011). This relationship will be further explored in this study through the inclusion of proxies for both macroeconomic activity and stock market returns in international markets.

Research methodology

The literature suggests that there is limited consensus as to whether macroeconomic variables are predictors of stock market returns. The objective of this study is to examine whether the selected macroeconomic factors can be classified as predictors of returns on the Johannesburg Stock Exchange (JSE), represented by the ALSI. This study aims to add to the body of literature and provide clarity to the mixed findings by increasing the scope of testing by using monthly data and including both local and international macroeconomic variables. The research question that this study aims to answer is:

How are local and global macroeconomic variables integrated with the All Share Index (ALSI), and what is the direction and significance of any relationships?

Research data

This study covers the period between June 1995 and December 2016. A period of this length (20 years) allows for the examination of the relationship between macroeconomic variables and the stock market over time. The period includes data relating to the financial crisis of 2008. The monthly data consisted of 258 data points over the period, allowing for robust statistical testing.

The dependent variable is the JSE ALSI, which was chosen as a proxy for the SA stock market. SA and international (US) macroeconomic variables were used to assess whether they are significant in the predictability of the returns on the JSE. As the US economy plays a leading role in movements in global stock markets, the inclusion of US variables considers the impact of the global economy on the SA stock market.

It is expected that the same hypothesised relationship will exist between the dependent variable and both the SA and international macroeconomic variables due to potential contagion between the markets (Chinzara & Aziakpono, 2009). The direction of movement is also hypothesised to be the same due to linkages between economies as a result of globalisation (Bekiros, 2014). While the direction of impact is expected to remain the same, it is expected that the size of the impact is likely to differ.

The following independent variables were selected: 1) Interest Rates (SA and US), 2) Inflation (SA and US), 3) SA Money Supply, 4) Rand/Dollar Exchange Rate, and 5) FTSE Index.

Interest rates

The prime lending rate is used as a measure of SA interest rates. This rate underpins borrowings from commercial banks and is seen as a market reference rate (South African Reserve Bank, 2009). Following this same reasoning, the US interest rate is also represented by the prime lending rate. This is the rate at which commercial banks provide loans to preferred borrowers (US Government, 2013).

Consequently, changes in interest rates bring about changes in stock market demand, which affects asset prices. A rising interest rate is likely to result in falling stock prices, causing a substitution effect of stocks for interest-bearing securities (Jefferis & Okeahalam, 2000). Interest rates (both SA and US) are hypothesised, therefore, to have a negative relationship with the dependent variable.

Inflation

Inflation is represented by the Consumer Price Index (CPI) which measures the prices paid by consumers for a representative basket of goods (Federal Reserve Bank, 2016). Changes in inflation impact the economy and, consequently, the stock market. Inflation is hypothesised to have a negative relationship with the JSE ALSI. When an economy experiences inflation, it often coincides with the tightening of monetary policies (Hamzah, Maysami, & Howe, 2004). This results in lower money supply, effectively resulting in lower disposable income for consumers and, consequently, lower expected cash flows for companies. Thus, rising inflation is expected to drive stock prices downwards (Shawtari & Salem, 2016).

South African money supply

Money supply is the amount of money that is in circulation within an economy (Talla, 2013). It consists of all notes and coins that are in circulation, excluding notes and coins held by commercial banks. This determination is relevant because it represents the money that flows freely within the market and, thus, is theorised to be closely linked to movements in the market. An increase in money supply is hypothesised to positively affect stock prices: when monetary policy is expansionary, the economy is positively stimulated (Talla, 2013). Consequently, consumers are likely to have increased disposable income which may result in an increase in demand for stocks, ultimately increasing stock prices (Trivedi & Behera, 2012).

Rand/Dollar exchange rate

The rand/dollar exchange rate represents the units of SA rand required to purchase a dollar. When the rand appreciates against the dollar, this movement impacts economic activity: the level of imports into the market increase due to imports becoming relatively cheaper, exports are decreased because they are more expensive, and the domestic value of international investments is reduced because they are worth less in rand terms (Hsing, 2016). It is hypothesised that there will be a negative relationship between the rand/dollar exchange rate and the JSE ALSI.

FTSE All Share index

The FTSE is a capitalisation weighted index that comprises all listed companies on the London Stock Exchange. This exchange was included in the analysis due to the prevalence of SA companies that have listed on the London Stock Exchange or may be affected by movements in international stock markets. Many SA-based companies have chosen to have dual listings in an effort to increase shareholder value and access to international sources of finance (Carmody, 2002). It is hypothesised, therefore, that a positive relationship will exist.

Table 1 summarises the variables used in this study. It details the source of the data and provides respective variable names to be used.

Table 1. Variables used for analysis.

Variable	Variable Name	Source
SA Stock market	ALSI	Iress
SA Interest rate	SAint	South African Reserve Bank
SA Inflation	SAinf	StatsSA
SA Money supply	MS	South African Reserve Bank
Rand US Dollar exchange rate	$/R	Iress
US Interest Rate	USint	Iress
US Inflation	USinf	Iress
FTSE Index	FTSE	Iress

Research process

To answer the research question, the data was analysed to determine whether a relationship exists between the variables of interest. In order to establish whether a relationship exists and whether it is asymmetric and significant; unit root, correlation, integration, causality and a vector error correction model were applied.

In order to test correlation and causality the data is required to be stationary (Ouma & Muriu 2014). Sharma and Seth (2012) outline over 50 studies that have used the Augmented Dickey-Fuller (ADF) test to establish stationarity in econometric data. If we reject the null hypothesis, we can conclude that the series is indeed integrated of order 1, rather than having a higher order of integration, implying the data is stationary (Alexander, 2008).

Correlation is used to test the co-movements of different indices returns and economic variables in relation to each other (Alagidede, Panagiotidis, & Zhang, 2011). This gives an indication of the strength and direction of the movements between the two variables. Johansen (2007) discusses that correlation does not always mean there is a causal relationship and thus further testing in the form of cointegration and causality is required to assess the relationship correctly.

Cointegration is a situation in which the regression of two or more nonstationary time series may not result in a spurious regression (Gujarati, 2011). When data are found to be integrated it indicates that there is a long-term dependency between the two variables, with that relationship not being spurious. The Johansen cointegration procedure is a vector autoregression (VAR) based test using the methodology developed in Johansen (Johansen, 1991). The null hypothesis for the cointegration tests is that the two variables are not cointegrated. Cointegration models are then constructed in two phases. Phase one examines the association of long-term equilibrium between the variables. Phase two is called a vector error correction model (VECM) which is based on a linear regression analysis of returns (Piesse & Hearn, 2002). The VECM can be interpreted as the movement in one variable is due to the movement from another variable, and a prior period error, which is a proxy for a long run movement due to past disequilibrium (Ashanapalli & Doukas, 1993).

Once an appropriate lag length is determined, cointegration will be used to determine whether the macro-economic variables and the stock market display a stationary process of a shared linear trend. Changes in lag lengths have significant effects on the results on the Johansen cointegration test (Emerson, 2007). The Hannan-Quinn criterion (HQC) and Schwarz information criterion (SIC) were used to select the optimal lag length due to the size of the sample being relatively large (Liew, 2004). The optimal lag length for the Vector Auto Regression (VAR) model is critical as the results from the VAR model are dependent on the model specifications. Liew (2004) found that while different tests for lag length might offer different results, applying a test improves the Johansen results.

Once the VECM has been defined, it may to be used to test the Granger causality flows. Granger causality measures the lead-lag relationship between the two indices (Alexander, 2008). The Granger causality test runs two tests: The first checks whether there is a predictive relationship between the two indices and the second tests whether one index causes the movement in the other index (Collins & Biekpe, 2003). This would assess whether there is an endogenous or exogenous relationship between the ALSI returns and the other variables.

If a relationship is established using the above procedures, a VECM is derived and a regression is run to test the model. If the model is significant and correcting, a Wald test is performed to test if there is a short-term relationship between the variables. A LM test and CUSUM test are performed to ensure that the model is free from serial correlation and dynamically stable.

Results

This section presents the results of each of the tests outlined in the research process to determine if statistically significant relationships exist between the ALSI returns and the variables examined.

Stationarity test

The ADF unit root test was run over the selected variables, to ensure they were non-stationary. The same tests were run on the first differences to test that they were stationary. As would be expected the variables themselves were all found to be non-stationary and their respective first differences stationary. This is as expected.

Correlation

A correlation test was run on the first differences to give an indication of the strength and direction of the variables.

Table 2. Correlations results between variables.

	FDALSI	FDFTSE	FDMS	FDSAinf	FDR$	FDSAint	FDUSinf
FDFTSE	0.5868	1.000	−0.106	−0.074	−0.259	0.063	0.069
FDMS	−0.0657	−0.1058	1.000	0.002	0.010	−0.011	0.012
FDSAinf	−0.0390	−0.0743	0.0025	1.000	0.050	−0.406	0.071
FDR$	−0.2870	−0.2588	0.0105	0.0501	1.000	0.086	−0.138
FDSAint	−0.0202	0.0631	−0.0115	−0.4058	0.0858	1.000	−0.027
FDUSinf	0.0090	0.0688	0.0124	0.0715	−0.1384	−0.0273	1.000
FDUSint	0.1517	0.2500	−0.0864	0.2833	−0.2330	−0.1931	0.1857

Note: FD indicates the first difference of the variables was used when assessing correlation.

The returns on the FTSE showed a substantial 0.586 positive correlation with returns on the ALSI, which indicates a relationship and possible integration. The only other correlation greater than 0.3, was found between the South Africa inflation rate and South African interest rate. This relationship is negative 0.406, which is expected, given that the Reserve Bank uses the repo rate as a tool for containing inflation (SARB, 2019).

Lag selection for cointegration

An appropriate lag length is crucial when running the correlation and the VECM. The results from running a VAR lag order selection criteria yield the following results for SIC and HQC:

Table 3. Lag length selection criterion results.

Lag	SIC	HQC
0	101.4516	101.3742
1	73.8245	73.0496
2	73.6622*	72.1900*
3	74.6472	72.4777
4	75.8555	72.9886
5	77.2646	73.7003
6	78.6729	74.4113
7	80.0564	75.0974
8	81.4493	75.7930
9	82.6119	76.2582
10	83.7385	76.6874
11	84.5734	76.8249
12	85.1780	76.7322
13	85.8167	76.6735
14	86.7797	76.9392
15	87.7305	77.1926
16	88.4743	77.2390

Note: The ‘*’ indicates the optimal lag length per the different criterion tested.

As these tests are best for large samples and both yield a lag of two, a lag length of two was applied in the testing below.

Cointegration test results

The Johansen cointegration procedure was performed on the monthly data. The South African stock was the dependent variable in each test. The cointegration test’s null hypothesis is that the variables are not cointegrated. If the null hypothesis is rejected this indicates that there is a long run relationship between the two variables.

Table 4. Results from the Johansen cointegration procedure.

Variable	Trace statistic	Prob	Max-Eigen statistic	Prob	Normalised cointegrating coefficients
FTSE	15.0641	0.058	14.4935	0.046*	−16.1180
MS	21.8402	0.0048*	14.1882	0.0514	−0.1389
SAinf	19.1792	0.0133*	19.1553	0.0078*	−96232.88
SAint	8.8911	0.3755	8.8430	0.2995	8379.054
USinf	5.8085	0.718	4.9997	0.7419	−672.7823
USint	6.3344	0.656	6.2294	0.5839	9931.535
$/R	10.3505	0.2547	10.3435	0.1903	−5870.761

Note: The critical value for the Trace statistic is 15.49471 and for the Max-Eigen value 14.2646. The t-statistic was calculated from the normalized cointegrating coefficients. The critical value for the t-statistic is 1.645. The ‘*’ indicates variables that have a significant cointegration relationship with the ALSI at a 5% level.

Where there are conflicting results between the Trace statistic and the Max-Eigen, the results of the Trace statistic are relied upon, as this is deemed a more powerful test (Serletis & King, 1997). When applying the above, we see that money supply and SA inflation have a significant relationships with ALSI returns at a 5% level and the FTSE does not have a statistically significant relationship at a 5% level when looking at the Trace statistic. From the normalised cointegration coefficients we can see the direction of the relationships. For money supply and inflation there is a positive relationship. The positive relationship with money supply aligns with the previous findings of Shawtari et al. (2016) and Trivedi and Behera (2012), but is contrary to the findings of Khumalo (2013). For inflation, this is contrary to prior research where a negative relationship was found by Khumalo (2013), Shawtari et al. (2016) and Phiri (2016).

Granger causality

The Granger causality test was run on the first differences. The test was only run on the variables where a cointegration relationship was identified. The following results were obtained:

Table 5. Granger causality test results.

Granger causality test null hypotheses	F Statistic	Prob
MS does not Granger cause the ALSI	0.0887	0.9152
SAinf does not Granger cause the ALSI	1.0224	0.3612
ALSI does not Granger cause the MS	0.2946	0.7451
ALSI does not Granger cause the SAinf	4.0333	0.0189*

Note: The ‘*’ indicates that the null hypothesis is rejected and there is a significant causal relationship at a 5% level.

The only statistically significant relationship is between the ALSI and inflation. However, the test states that the direction of the relationship is that the ALSI has a causal effect on inflation in South Africa. This relationship is in line with expectations; however, the direction is contrary to the findings noted above.

Vector error correction model

Table 6. VECM results between the ALSI (dependent) and MS (independent).

VECM coefficients with ALSI as the dependent and MS independent variable
	Coefficient	Std. Error	t−Statistic	Prob.
C(1)	−0.0016	0.0077	−0.2109	0.8331
C(2)	0.1835	0.064	2.8676	0.0045
C(3)	−0.0132	0.0644	−0.2053	0.8375
C(4)	0.0048	0.0177	0.2691	0.7881
C(5)	−0.0045	0.0178	−0.2536	0.8
C(6)	146.9372	62.23	2.3612	0.019

Note: C(1) represents the error correction term. C(2) and C(3) is the change dependent variable lagged by one period and two periods respectively. C(4) and C(5) is the independent variable lagged by one and two periods respectively. C(6) is the constant in the equation.

A vector error correction model was built on the cointegrated variables and the derived system tested for significance. This gives more information than Granger causality on how the variables are related.

Money supply

The following equations were derived for the relationship between the ALSI and money supply variable:$\mathrm{\Delta }ALS{I}_t=-0.1389\left(ALS{I}_{t-1}-0.03202M{S}_{t-1}-8933.371\right)+0.1835\mathrm{\Delta }ALS{I}_{t-1}-0.013223\mathrm{\Delta }ALS{I}_{t-2}+0.0048\mathrm{\Delta }M{S}_{t-1}-0.0045\mathrm{\Delta }M{S}_{t-2}+146.9372$ $\mathrm{\Delta }ALS{I}_t=-0.0016\left(ALS{I}_{t-1}-0.1389M{S}_{t-1}-8933.371\right)+0.1835\mathrm{\Delta }ALS{I}_{t-1}-0.013223\mathrm{\Delta }ALS{I}_{t-2}+0.0048\mathrm{\Delta }M{S}_{t-1}-0.0045\mathrm{\Delta }M{S}_{t-2}+146.9372$

When applying a least squared regression to the above model, it returns the following t-statistics and probabilities:

The vector error correction equation, represented by C(1), is correcting as it has a negative coefficient. However, the model is not significant given the low test statistic and p-value. It is also interesting to note that the short run relationship of money supply and the ALSI, represented by C(4) and C(5) are also not significant. This confirms the Granger causality results that there is no causal relationship. This does not align with the expectations and finding noted above.

Inflation

Table 7. VECM results between the SAinf (dependent) and ALSI (independent).

VECM coefficients with SAinf as the dependent and the ALSI independent variable
	Coefficient	Std. Error	t−Statistic	Prob.
C(1)	−0.0431	0.0136	−3.1800	0.0017
C(2)	0.4754	0.0611	7.7779	0.0000
C(3)	0.0806	0.0617	1.3048	0.1932
C(4)	−0.0000	0.0000	−1.1444	0.2536
C(5)	0.0001	0.0000	2.2057	0.0283
C(6)	−0.0034	0.0335	−0.1025	0.9184

Note: C(1) represents the error correction term. C(2) and C(3) is the change dependent variable lagged by one period and two periods respectively. C(4) and C(5) is the independent variable lagged by one and two periods respectively. C(6) is the constant in the equation.

The cointegration and causality testing found that the ALSI Granger caused South African inflation. The model was built with the ALSI being the explanatory variable. The following equation was derived:$\mathrm{\Delta }SAin{f}_t=-0.0431\left(SAin{f}_{t-1}-0.00001ALS{I}_{t-1}-5.8250\right)+0.4754\mathrm{\Delta }SAin{f}_{t-1}+0.0806\mathrm{\Delta }SAin{f}_{t-2}-0.00004\mathrm{\Delta }ALS{I}_{t-1}+0.0001\mathrm{\Delta }ALS{I}_{t-2}-0.0034$When examining the least squared regression on the model, the results are as follows:

The cointegration model represented by C(1) can be seen to be correcting as it has a negative coefficient, and is statistically significant at a 5% level given the test statistic and p-value. It is also interesting to note the short run relationship as indicated between South African inflation and the ALSI (represented by C(5)) which shows the ALSI lagged by two months is a significant variable in explaining the movements in the South African inflation rate. To confirm this, a Wald chi-squared test was run on C(4) and C(5). As the p-value of the Wald test is greater than 5% we can conclude there is no short run causality between the ALSI and Inflation. This finding aligns with the relationship identified when applying the Granger causality test.

Conclusion and recommendations

This study examined the relationship between select local and US macroeconomic variables and the SA stock market. The macroeconomic variables were selected based on financial and economic theory, yet the results of the testing identified misalignment between theory and practice. Monthly data points were used as higher frequency data is preferred to lower frequency and it captures dynamic interactions occurring more regularly (Chinzara & Aziakpono, 2009).

This study asked the following research question: How are local and global macroeconomic variables integrated with the All Share Index (ALSI), and what is the direction and significance of any relationships? This study identified money supply and inflation as having cointegration relationships. However, after running a VECM and Granger Causality the only relationship found was between South African inflation and the ALSI. The ALSI was found to have a statistically significant relationship in explaining South African inflation. With the nature of the relationship being positive, this presents new findings which differ from the prior research performed by Khumalo (2013), Shawtari et al. (2016) and Phiri (2016). Each of these studies covered different time periods, and had less data points, because they used quarterly data or examined shorter periods.

This finding provides insight into the relationship between stock markets and inflation and could be useful from an economic policy perspective. Repo rates are used to control inflation. The current understanding is that the relationship between markets and inflation is from inflation to markets. However, this study shows this relationship to be in the opposite direction. This could be useful to consider as a factor for predicting future inflation, and thus useful in identifying if one can raise or reduce the repo rate without inflation leaving its target. These implications are far-reaching and further research is recommended on the topic.

Acknowledgement

This publication is based on research that has been supported in part by the University of Cape Town Research Committee's Open Access Journal Publication Fund.