This Selected Issues paper highlights that cautious monetary and fiscal polices of South Africa during 1997 resulted in a return of financial investor confidence and capital inflows during 1997 and through April 1998. These policies helped the South African economy emerge successfully from the exchange market pressures of 1996 and weather the contagion from the East Asian crisis in the second half of 1997. Throughout 1997 and up until May 1998, inflation and market interest rates fell considerably, net international reserves increased, and the net open forward position of the Reserve Bank was reduced sharply.


This Selected Issues paper highlights that cautious monetary and fiscal polices of South Africa during 1997 resulted in a return of financial investor confidence and capital inflows during 1997 and through April 1998. These policies helped the South African economy emerge successfully from the exchange market pressures of 1996 and weather the contagion from the East Asian crisis in the second half of 1997. Throughout 1997 and up until May 1998, inflation and market interest rates fell considerably, net international reserves increased, and the net open forward position of the Reserve Bank was reduced sharply.

III. Inflation, Money Demand, and Purchasing Power Parity in South Africa

A. Introduction

113. Inflation has fallen considerably in South Africa during the 1990s. Annual growth in the consumer price index fell from 16.2 percent in 1991 to 9.4 percent in 1993 and further to 6.8 percent in 1997. However, developments in broad money (M3) have not followed the same pattern. Although the annual growth rate in M3 fell from 13.8 percent in 1991 to 5.9 percent in 1993, it has since increased to 17.3 percent in 1997. During the 1970s and 1980s, fluctuations in inflation and money growth were even larger (Figure 10). The contrasting developments in inflation and money growth during the 1990s have led analysts to question whether there exists a stable relationship between money growth and inflation, i.e., whether money demand is stable, and whether it is appropriate for the Reserve Bank to use M3 as an intermediate target for its monetary policy decisions.48,49 As a consequence of the perceived uncertainty in the relationship between money and prices, and as a means of enhancing the transparency and credibility of monetary policy decisions, the possibility of adopting an explicit inflation targeting regime in South Africa has been discussed, mirroring the monetary policy regime in countries such as New Zealand, United Kingdom, and Canada.

Figure 10.
Figure 10.

South Africa: Inflation, money growth, exhange rate, and trade developments, 1970-97

Citation: IMF Staff Country Reports 1998, 096; 10.5089/9781451840940.002.A003

Source: South African Reserve Bank, Quarterly Bulletin

114. At the same time, the nominal exchange rate has fluctuated widely in South Africa (Figure 10), For example, in recent years, the nominal effective exchange rate depreciated by 21.5 percent in 1996, appreciated by 6.0 percent during the first half of 1997, before it depreciated again by 6.4 percent in the second half of the year.50 Although the presence of trade sanctions reduced imports as a share of GDP during the 1980s and early 1990s (Figure 10), it is likely that movements in foreign prices and the nominal exchange rate have also contributed to inflationary developments in South Africa.

Figure 11.
Figure 11.

South Africa: Impulse responses of inflation to a unit shock in a monetary variable

(Bivariate VAR models including 4 lags)

Citation: IMF Staff Country Reports 1998, 096; 10.5089/9781451840940.002.A003

115. The purpose of this empirical study is to simultaneously examine the long-run and short-run relationships between prices, the exchange rate, and various monetary aggregates in South Africa. This includes a search for a stable money demand relationship, and an examination of the purchasing power parity (PPP) hypothesis. In the course of doing this, the leading indicator properties of different aggregates for inflation are studied, and the issue of a potential structural break in the data since 1994—the starting year of the successful political transformation of the economy and the lifting of sanctions—is tentatively examined.

116. The results indicate that (i) there exists a sensible and stable money demand type of relationship between M3, nominal income, and long-term interest rates, as well as a long-run relationship between foreign prices, the nominal effective exchange rate, and domestic prices; (ii) temporary deviations from the long-run equilibrium in the money market and, in particular, in the external market are important for short-run inflation fluctuations; and (iii) there are indications of a structural break in the data sometime around 1994, implying that a given rate of money growth is likely to lead to an inflation rate that is about 2-3 percentage points less per year after 1994, ceteris paribus.

117. The rest of this section is organized as follows: Section B briefly reviews some earlier studies related to inflation, money demand, and PPP in South Africa, before it lays out the theoretical and methodological approach in the current study, and discusses some data issues. Section C presents the empirical results, while Section D summarizes the findings and concludes.

B. Background, Methodology, and Data

118. Previous empirical studies have either focused exclusively on the demand for money or the degree of pass-through from exchange rate movements to inflation. In a previous Selected Issues paper (SM/96/109), it was argued tentatively that narrow money (notes and coin in circulation outside the banking system) might warrant a more prominent role in the monetary policy framework, as this aggregate is a fair leading indicator for inflation, and as the demand for narrow money appears to be stable. However, on the first point, DeJager and Ehlers (1997) show in an extensive study that growth in M3 is a better and more stable indicator for future inflation rates than narrow money, and that M3 has a consistent negative relationship with interest rates. Hum (1991) has examined the demand for M3 in South Africa, using data for the period 1965-90, and finds that the long-run effects of prices, real income, and interest rates are significant and of sensible magnitudes. This contrasts with Moll (1997) who argues that the results for long-run money demand are more robust when the models include a narrow definition of money.

119. The issue of PPP and whether there is a full pass-through from exchange rate movements to domestic prices, have been discussed recently by Tsikata (1998) and Subramanian (1998). Although the results are sensitive to the choice of price aggregates and sample period, indications are that the effective nominal depreciation of the rand during the 1990s is almost fully reflected in higher prices of imported goods.

120. The first step in the empirical section below includes a general examination of the leading indicator properties of various monetary aggregates for inflation. This is motivated by two reasons: (i) in the context of an inflation target regime, it is important to find a set of indicators that contain information about future inflation rates,51 as the lags between changes in the monetary aggregates (or instruments) and their effect on inflation typically are relatively long; and (ii) this exercise can possibly cast some light on which money aggregate and interest rate might be suitable to include in the money demand framework. The leading indicator properties for inflation were examined with the so-called “Granger-causality” tests, and by inspecting the associated impulse-response functions.52 Hence, the following equation was estimated:


where p refers to the underlying price level, and x denotes variables such as money and credit, interest rates, exchange rates, wages, or output (see Appendix for details)53. The variable x is said to “Granger-cause” p if the null hypothesis of β1 = β2 = … = βk = 0 is rejected.

121. The theoretical underpinning for the study of money demand and the PPP relationship is standard. Hence, assume that the domestic price level in the economy, p, is a linear combination of the price level for tradeable goods, PT and nontradeable goods, pNT, respectively, i.e., p = αpr + (I-α)pNT, where 0≤α≤l. The PPP hypothesis implies that the price of the tradeable good is determined in the world market and equal to pT = p*-e in domestic currency, where e is the nominal (effective) exchange rate (defined as foreign currency per domestic currency), and p* represents foreign prices. The price of nontradeables is then determined by adding an assumption of equilibrium in the money market, i.e., where real money supply equals real money demand: ms-p = md(y, i), with real money demand assumed to be a positive function of real income, y, and a negative function of the nominal interest rate, i54 Thus, we end up with a system of 6 interrelated variables [p, p* e, m, y, and i], where economic theory suggests that two long-run relationships could be found; one between domestic prices, foreign prices, and the nominal exchange rate, and another between domestic prices, money, real income, and the nominal interest rate. While we would expect both the real exchange rate and real money demand (or, alternatively, velocity) to be fairly stable in the long-run, we would also expect temporary deviations from these two long-run equilibriums to affect future fluctuations in prices such that the long-run equilibriums are restored.

122. In addition to these considerations, a dummy variable for the period 1994-97 was added to the model in an attempt to identify a possible structural break associated with the economic effects of the political transformation that took place in the early 1990s. This transformation, as well as some important economic structural reforms, could arguably have affected both the long-run money demand relationship and the real exchange rate, since it led to both some financial deepening (as nonwhite households gained access to formal banking services to a larger extent), as well as a strong increase in foreign competition which in turn could have had a one-off effect on the domestic price level.55

123. The empirical analysis was carried out using quarterly data between 1970:1 and 1997:4. The underlying consumer price index (CPIu) was used rather than the headline consumer price index (CPI) throughout the study. By using CPIu, which excludes highly volatile food prices and housing costs from the CPI, it was expected that the signal to noise ratio would improve in the estimations. To further examine whether broad money or narrow money is more closely related to inflation, three different monetary aggregates were used in the analysis: NC, notes and coin in circulation outside the banking sector; M1, consisting of NC plus cheque and demand deposits; and M3 (broad money) consisting of M1 plus medium-and long-term deposits. In addition, when estimating the money demand relationship, two alternative interest rates were used: 3-month t-bill rates, and 10-year government bond yields. Traditional unit root tests (not reported) indicated that all series were integrated of order 1, i.e., the series were nonstationary in levels but stationary in first-differences.

124. The nonstationarity of the data together with the notion that none of the variables a priori can be regarded as exogenous (possibly with the exception of foreign prices), suggested that an appropriate methodology would be to start with a nonstructural vector auto regression model (VAR), and use Johansen’s cointegration tests to examine whether there existed any long-run relationship among the variables. As a second step, economic theory (as described above) was used for identification, turning the empirical model into a structural VAR, and specific cointegrating vectors—related to the PPP and money demand hypothesis—were estimated and tested. Hence, to allow for dynamic interaction among the variables in the system, the two long-run relationships (as suggested by theory) were estimated separately but simultaneously.

C. Results

125. The results from the Granger-causality tests are reported in Table 4. The tests were conducted for the whole sample period and for the subperiods 1972:2–1984:4 and 1985:1–1997:4. Although the results are not very impressive when the whole sample period is considered, the subsample regressions show that the overall results are driven by a very poor relationship between the variables during the first subperiod. However, for the later subperiod (Table 4b), a number of reasonable leading indicators for inflation were found, including narrow money aggregates (M1 and, possibly, NC), interest rates and interest spreads,56 the nominal exchange rate, foreign prices, wages, and output fluctuations.57 The associated impulse response functions for some selected aggregates are displayed in Figure 11. They show that movements in broad money contain information about the underlying inflation rate further into the future than narrow money, and that a depreciation of the rand indicates a pick up in underlying inflation within two quarters. An increase in the output gap tends to be followed initially by a fall in underlying inflation, but this fall is more than reversed in the subsequent 3-6 quarters, possibly indicating that output shocks on average are driven by fluctuations in aggregate demand.58

Table 4.

Leading Indicators and Underlying Inflation

(Marginal significance levels) 1/

article image

The p-values are the marginal significance levels for a Wald-test (Chi^2(k)) of the hypothesis that the estimated coeffients for the relevant variable are jointly zero. Shaded numbers indicate rejection of the hypothesis of no Granger-causality of a monetary indicator at the 10-percent level. All regresssions include seasonal dummy variables.

The interest rate indicators include the level of the interest rates (3-months t-bills and 10-year government bonds), and the spread against the comparable U.S. interest rate. Hence, these indicators are not in first-difference form.

Gap stands for output gap. This indicator is not in first-difference form.

126. Shifting to the long-run relationships among the variables, the results from the cointegration tests are summarized in Table 5. The number of cointegrating vectors was estimated using the Johansen (1988) procedure. It is well-known that cointegration tests in the Johansen setting are sensitive to the lag-length of the VAR. Although it is common to include 4 lags in the VAR when quarterly data are used, common information criteria tests59 indicated that only 2 lags should be included. However, inspection of the standard diagnostic tests suggested that a VAR-order of 3 or 4 was more appropriate. Hence, the results are reported with 2, 3, and 4 lags included in the VAR, respectively.

Table 5.

Results from the structural VAR-model 1/

article image

The VAR also include (unrestricted) seasonal dummy variables and a time dummy for the period 1994:1-1997:4.

Number of cointegrating vectors based on Johansen’s Trace statistic and the maximum eigenvalue statistic, respectively, at the 5-percent significance level. * and ** indicate rejection of the LR-test at the 5-percent and 1-percent significance level, respectively.

127. The economic model suggests that 2 cointegrating vectors should be found, and the cointegration tests typically picked up 2-3 stationary vectors, but the results varied between 0 and 4 vectors depending on the number of lags included in the model, as well as on the choice of monetary aggregate and interest rate (see column 2 in Table 5). Despite these somewhat inconclusive results, restricted cointegration tests were performed under the assumption of the presence of two cointegrated vectors. The parameters in the restricted model were constrained to test whether the two stationary vectors could be represented by the two long-run relationships discussed above. The results were in general supportive of the theoretical arguments: the hypothesis that one of the cointegrating vectors includes only the variables p, p* and e (column 3, Table 5) was rejected in only 3 of the 18 specifications,60 and the hypothesis that one cointegrating vector includes only the variables p,m, GDP and i (column 4, Table 5) was rejected in only 4 of the 18 specifications. However, the joint test of the two hypotheses was typically rejected when M1 or NC was included in the model, but it was not rejected when M3 was included in the model (column 5, Table 5).

128. Turning to the parameters of the cointegrating vectors, it can be noted that the estimated parameters for the nominal effective exchange rate and foreign prices have the expected signs. However, the joint hypothesis of the coefficients being equal to -1 and 1, respectively, (so-called “strict PPP”) was almost always rejected (not reported). Nevertheless, it is interesting to notice that the joint movements in the nominal effective exchange rate and foreign prices seem to be almost fully reflected in domestic prices in the long-run, in the sense that the sum of the estimated parameters (in absolute values) is relatively close to 2. A possible explanation for these results is that domestic price-setters sometimes hesitate to adjust domestic prices in line with exchange rate fluctuations—perhaps because they regard these fluctuations as temporary. This would explain a coefficient of less than one for e. But since this behavior would erode competitiveness in the long-run, the price-setters compensate by increasing domestic prices by slightly more than a corresponding increase in international prices.

129. The estimated coefficients in the money demand relationship have the expected signs and are of a plausible magnitude when M3 is included in the model. The estimated coefficient on M3 is between 0.9 and 1.1, and the coefficient on real income is between 0.6 and 2. Constraining the coefficient on M3 to equal 1 yields an income elasticity for real money demand of 1.8 (where the VAR includes 3 or 4 lags and the interest rate variable is government bond yields), but the standard errors are fairly large, and the joint hypothesis of the coefficients on M3 and GDP being equal to 1 and -1, respectively, could not be rejected. The estimated coefficient for the nominal interest rate is positive when the government bond rate is used but negative when the t-bill rate is used. A possible explanation for this result could be that the own deposit rate for broad money is more closely related to the short-term t-bill rate than the long-term government bond rate. However, the magnitudes of the estimated coefficients are quite small, and a likelihood ratio test shows that the interest rate can often be excluded from the cointegrating vector, indicating that M3, prices, and real income form a cointegrating vector, i.e., velocity is stationary.

130. The results were less encouraging when narrow money (M1 or NC) was included in the model: the estimated coefficient on real income was quite unstable and often had the wrong sign, and when the coefficient on M1 or NC was constrained to equal 1, the estimated income elasticities become even more implausible. Consequently, a sensible and stable long-run money demand relationship for narrow money could not be established.

131. To further test the robustness of the results regarding the real demand for broad money and PPP, an error-correction model for short-run fluctuations in the underlying inflation rate was estimated. The hypothesis was that in the case of either a disequilibrium in the money market or a deviation from PPP, some adjustment would have to take place in the subsequent period(s) to restore long-run equilibrium, and a good candidate for such an adjustment would be the domestic price level. Hence, a regression was estimated with the underlying inflation rate as an independent variable on lagged residuals from the two restricted cointegration vectors (the error-correction terms)61 and lagged observations of the underlying inflation rate, quarterly growth rates in money and real output, and a linear combination of the nominal exchange rate and foreign prices. As the focus was shifted to the short-run properties of the data, the regressions were estimated both for the whole sample period, and for the shorter subperiod 1985:1–1997:4.

132. The results broadly confirmed the hypothesis regarding inflation adjustments (Table 6); the estimated coefficients on the error-correction terms always came in with the expected negative signs, indicating that temporary deviations from the long-run equilibrium in the money market and from the PPP equilibrium result in short-run inflation adjustments which tend to restore the long-run equilibria. For example, the results indicate that if the real effective exchange rate depreciates (above its long-run equilibrium level) during a quarter due to an unexpected increase in foreign prices or a nominal depreciation of the rand, underlying inflation will increase in the subsequent quarter to partly offset the real depreciation. Likewise, for a given level of real income and nominal interest rates, a positive shock to M3 will result in a pickup in underlying inflation in the subsequent quarter. Although the magnitudes of the error-correction coefficients appear quite small (for example, the estimated coefficient of -0.04 for the PPP error-correction term indicates that 16 percent of any deviation from the long-run equilibrium is restored within 1 year), they are similar to what is typically found for other countries.62 It can also be noted that the estimated coefficient on the error-correction term from the money demand relationship is only significant in the subperiod 1985–97, which is consistent with the results from the Granger-causality tests that the money-price relationship is more reliable during this period. Finally, the estimated coefficient for the 1994–97 dummy variable has the expected negative sign and is significant in most specifications, indicating that for a given rate of money growth, the underlying annual inflation rate after 1993 would be about 2–3 percentage points less than during the period 1972–1993, everything else equal, thereby suggesting an important process of financial deepening in recent years.

Table 6.

Short-Run Fluctuations in Underlying Inflation, Results from the Error-Correction Model 1/

article image

* and ** indicate significance at the 10-percent and 5-percent level, respectively, of either the t-test for a single coefficient or of a Wald-test for the sum restriction. All regressions include seasonal dummies, k is the number of lags, and n is the number of additional independent variables.

The coefficients for Dum9497 are multiplied by 100.

k lags of the variable were included. The reported coefficient is the sum of the estimated coefficients for the k lags.

GAP is the output gap. This variable is not in first difference form.

133. The results further show that there is marked persistence in the inflation rate, as lagged inflation enters significantly. However, the other lagged variables were typically nonsignificant, implying that past observations of these variables were less important, once one controlled for the dynamics of inflation and the deviations from the long-run PPP and money demand relationships. In addition to the monetary aggregates mentioned above, a number of alternative aggregates—which potentially could affect short-run fluctuations of inflation—were included in different specifications of the error-correction model. These included growth in nominal wages and unit labor costs, other measures of the output gap, and measures of the fiscal balance. However, these variables were also nonsignificant once one controlled for the inflation dynamics and the error-correction terms.

134. The fit of the error-correction model was reasonably good (see Figure 12), especially when the focus was on the subperiod 1985–97. Figure 12 also plots the dynamic forecasts for 1997 using equations (1) and (5) in Table 6. Although the forecast overestimates the inflation rate in the first quarter of the year (due to an estimated large coefficient on the first-quarter dummy variable), the actual inflation outcomes in the following three quarters are well within the forecast confidence interval

Figure 12.
Figure 12.

South Africa: Actual values, fitted values, and forecasts of the underlying inflation rate.

Citation: IMF Staff Country Reports 1998, 096; 10.5089/9781451840940.002.A003

D. Summary and Conclusions

135. The main results in this study are that two long-run stable relationships involving domestic prices can be identified: one between domestic prices, M3, real income, and long-term interest rates, indicating that demand for broad money is stable in the long-run; and one between domestic prices, foreign prices, and the nominal effective exchange rate, indicating that the real effective exchange rate is stable in the long-run. The results further show that temporary deviations from these two long-run equilibria (in the money market and in the external market) are important for short-run inflation fluctuations. Finally, there are indications of a structural break in the data sometime around 1994, implying that, ceteris paribus, a given rate of money growth is likely to lead to an inflation rate that is about 2–3 percentage points less per year after 1994.

136. An interesting aspect of the results is that even though the South African economy has undergone a number of important structural changes during the period under review—including long periods of trade sanctions, the presence of the financial rand system and widespread exchange controls on residents, different monetary policy regimes (see below), and considerable swings in the terms of trade—the long-run relationships between the examined macroeconomic and financial aggregates are fairly stable, and consistent with economic theory. In this context, it is perhaps not surprising that it is the broadest measure of money that seems to work better in the long-run, although fluctuations in narrow money aggregates might be useful for short-run inflation projections.

137. Notwithstanding the long-run results, the results regarding the short-run fluctuations are strengthened when the analysis focuses on the subperiod 1985–97, as the fit of the inflation model improves considerably. This is probably explained in part by the shift in the monetary policy regime that took place in the early 1980s, when the Reserve Bank moved to a system of indirect control of money supply. This shift implied a greater scope for the use of short-term interest rates, rather than changes in cash and liquid asset requirements combined with credit ceilings and interest rate controls, thereby enhancing the responsiveness of monetary aggregates to macroeconomic developments. In addition, the opening up of the economy during the 1990s—including the trade reforms and liberalization of capital controls—has probably made the domestic economy more responsive to fluctuations in international markets.

138. To sum up, two general policy conclusions emerge from this study: (i) the result regarding the PPP hypothesis implies that any attempts to enhance competitiveness by actively depreciating the nominal exchange rate will be eventually dissipated through higher domestic inflation, and (ii) the result regarding the stability of money demand means that growth rates in broad money cannot be ignored when monitoring and forecasting inflation developments, although the quantitative effect of money growth on inflation appears to have declined during the last few years.

139. To further understand the dynamic interplay between various macroeconomic aggregates in South Africa, it would be interesting to expand the current framework to include different types of shocks (aggregate demand and supply shocks, or real and nominal shocks, etc.), and trace, for example, the simultaneous response of output and inflation. The structural VAR model used in this paper is a suitable framework for such an exercise, and Moll (1997) has taken interesting steps in that direction.


  • DeJager, C.J. and R. Ehlers (1997), “The relationship between South African monetary aggregates, interest rates and inflation—a statistical investigation”, Departmental memorandum, Economics Department, South African Reserve Bank.

    • Search Google Scholar
    • Export Citation
  • Ericsson, N. (1998), “Empirical modeling of money demand”, Empirical Economics, forthcoming.

  • Hum, A.S. (1991), “Interest rates, inflation and the stability of the demand for M3 in South Africa”, Greek Economic Review, 13, 251268.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund (1996), South Africa—Selected Economic Issues (SM/96/109) Chapter VI, “Narrow monetary aggregates”.

  • Johansen, S. (1988), “Statistical analysis of cointegrating vectors”, Journal of Economic Dynamics and Control, 12, 231254.

  • McDonald, R. (1995), “Long-run exchange rate modeling—a survey of the recent evidence”, IMF Staff Papers, 42, 437489.

  • Moll, P. (1997), “Money, interest, prices, and income in a less developed country”, mimeo, World Bank.

  • Subramanian, A. (1998), “South Africa—pass-through revisited”, mimeo, IMF.

  • Tsikata, Y. (1998), “Liberalization and trade performance in South Africa”, mimeo, World Bank.


All data series are from the South African Reserve Bank, Quarterly Bulletin, except for the U.S. interest rates which are taken from IMF, International Financial Statistics. The following variables are included in the study:

CPIu: Underlying consumer price index. This index equals total CPI excluding “food and nonalcoholic beverages”, “home owner’s cost” and “value added tax”.

NC: Notes and coin outside the banking system.

Ml: NC plus checking deposits.

M2: Ml plus short- and medium-term deposits.

M3: M2 plus long-term deposits.

PSC: Credit extension to the private sector.

3m: Interest rate on 3-months t-bill.

10y: Interest rate on 10-year government bonds.

3mS: Spread between interest rates on 3-months t-bills in South Africa and the U.S.

10yS: Spread between interest rates on 10-year bonds in South Africa and the U.S.

US$: Nominal exchange rate; Rand per US dollar.

NEER or E: Nominal effective exchange rate including (weights in brackets) U.S. dollar (51.7), British pound (20.2), Deutsche mark (17.2), and Japanese yen (10.9).

P*: Effective producer price index in foreign countries, including the same four countries and weights as when calculating the NEER.

Wage: Remuneration per worker in the nonagricultural sectors.

ULC: Unit labor costs in the nonagricultural sectors.

GDP: Real gross domestic product, 1990 prices, seasonally adjusted.

Gap: Output gap; calculated by using the Hodrick-Prescott filter (with λ=400) on real GDP.


Strictly speaking, an unstable relationship between money growth and inflation does not necessarily imply that money demand is unstable, as the latter would be expected to vary with fluctuations in other variables, such as real income and nominal interest rates.


The Reserve Bank has announced annual guidelines for growth in broad money since 1986 (6-10 percent in 1997). However, actual growth in M3 has substantially exceeded the guideline range during the last few years, and the authorities have on several occasions announced that the Reserve Bank in practice is guided by developments in a number of different indicators including, various price indices, the shape of the yield curve, the nominal exchange rate, and the output gap.


The Reserve Bank’s measurement of the nominal effective exchange rate is used throughout the paper; an increase in the effective exchange rate means an appreciation of the rand. See Appendix for details.


An inflation target regime basically involves a two-step approach: First, the authorities project the future inflation rate under current policies by studying indicators such as current and past values of money growth, interest rates, exchange rates, and output activity. Together, these “leading indicators” are used to evaluate the inflation-momentum in the economy which is compared with the inflation target. Second, if projected inflation is outside the target, the authorities pursue an appropriate policy mix using available instruments to correct for the discrepancy. The empirical analysis in this paper is related to the first of these steps, i.e., it can be regarded as a preliminary attempt to find useful leading indicators of inflation.


It should be noted that the “Granger-causality” concept has a different interpretation from what is usually meant by causation. The intuition behind Granger-causality is simply that if an event y is the cause of an event x, then y must precede x in time, and by using past observations of y in addition to past observations of x, the forecast of x should improve.


As is conventional in the literature, all variables are expressed in natural logarithms, except for the nominal interest rate. Hence, the first difference of a variable equals the percentage change from the previous quarter.


More precisely, economic theory suggests that demand for money depends on the opportunity cost of holding money. Although the opportunity cost for holding cash is larger when the nominal interest rate is higher, it is ambiguous whether broader definitions of money are positively or negatively related to the nominal interest rate, as broader money typically is interest bearing. Ideally, one should control for any differences between the “own” deposit interest rate and the interest rate on alternative assets (such as government securities). See Ericsson (1998) for further discussions.


The dummy was included in the model without restricting it to the cointegrating vector, implying that the average growth rates of the variables can change at the time of the structural change while the cointegrating vectors remain unchanged.


The Granger-causality tests should be interpreted with particular caution when (long-term) interest rates are used as an indicator. It is well-known that aggregates which reflect forward-looking behavior often are good predictors of various macroeconomic time series. Obviously, this does not mean that these variables “cause” fluctuations in other variables; it rather reflects that the market uses current information efficiently.


The same tests were performed using inflation rather than underlying inflation. The results (not reported) were similar to those in Table 4, although foreign prices and wages seemed to contain less predictive information about future inflation than for underlying inflation, whereas the opposite was true for M3.


If the output gap is due to a positive aggregate supply shock, one could expect a (temporary) fall in inflation, whereas the opposite would be true when the output gap is the result of a shock to aggregate demand.


Both the Akaike Information Criterion and the Schwarz Bayesian Criterion were examined.


By not constraining the coefficients to equal 1 (in absolute values), the test allows for various fixed costs, such as transportation and menu costs, to vary over time and across countries. This test is sometimes referred to as a simple test of PPP.


The error-correction terms were derived from the restricted VAR model which included M3, and interest rates on government bonds (each with 3 lags), with the coefficients on prices and M3 constrained to -1 and 1, respectively.


For example, McDonald (1995) finds that the speed of adjustment of a deviation from PPP is about 2 percent per month for bilateral U.S. dollar, German mark, and Japanese yen exchange rates.

South Africa: Selected Issues
Author: International Monetary Fund
  • View in gallery

    South Africa: Inflation, money growth, exhange rate, and trade developments, 1970-97

  • View in gallery

    South Africa: Impulse responses of inflation to a unit shock in a monetary variable

    (Bivariate VAR models including 4 lags)

  • View in gallery

    South Africa: Actual values, fitted values, and forecasts of the underlying inflation rate.