This Selected Economic Issues paper examines economic development in South Africa during 1995–96. The paper highlights that in 1995, the economy of South Africa grew by 3.3 percent, the third consecutive year of economic growth, and it is expected to grow between 3½ and 4 percent in 1996. Some aspects of the unemployment problem are addressed in this paper. The paper also focuses on the implications for policy of the steps taken in 1994 and 1995 to establish an outward-oriented economy, after many years of effective autarky.

Abstract

This Selected Economic Issues paper examines economic development in South Africa during 1995–96. The paper highlights that in 1995, the economy of South Africa grew by 3.3 percent, the third consecutive year of economic growth, and it is expected to grow between 3½ and 4 percent in 1996. Some aspects of the unemployment problem are addressed in this paper. The paper also focuses on the implications for policy of the steps taken in 1994 and 1995 to establish an outward-oriented economy, after many years of effective autarky.

VI. Narrow Monetary Aggregates

For some years, the South African Reserve Bank has announced guidelines for broad money, M3, 1/ as part of its apparatus for executing monetary policy. These guidelines were supplemented in 1995 by bank-by-bank guidelines on the growth of credit to the private sector.

The M3 guidelines were substantially overshot in 1994, and both guidelines were overshot in 1995. But even as the official guidelines were being substantially breached, underlying inflation has remained under control and indeed has fallen in the latter part of 1995. This experience is compounded by inflation expectations—implicit in long bond yield differentials—through 1995 which indicated that inflation was anticipated to remain subdued, and even fall. To this extent, market participants implicitly rejected the view that the high growth rates of broad money and credit to the private sector, that have been sustained for a considerable period of time, presaged renewed inflation (see Chapter I). The implication is that market participants doubt the usefulness of these aggregates as indicators of inflationary pressures.

This chapter focusses on one set of issues arising from this, namely the possible role of narrow monetary aggregates in the monetary policy framework. This focus is motivated both by the absence of published academic research in this area on South African data and by the prominent role of similar aggregates in the monetary arrangements of a number of other countries. This chapter does not attempt a formal study of the nominal anchor issue. Such a study is simply not possible given the limited formal research on both narrow and broader aggregates and on the behavior of inflation in South Africa. Instead, this discussion highlights a number of issues concerning the narrow aggregates that may be more fully pursued in further research.

Accordingly, the approach is exploratory rather than definitive. The findings suggest that there may be grounds for a greater use of narrow money aggregates in the monetary policy framework than at present because narrow aggregates appear to meet the twin requirements of a nominal anchor: the demand for narrow money has identifiable components, and there is initial evidence that this aggregate is a leading indicator of inflation. However, this does not imply that narrow money can or should be used for operational purposes in the design of monetary policy in its current form now. As currently measured, it exhibits a high degree of volatility which greatly diminishes its information content. But in view of its inflation predicting properties, consideration might be given to addressing this measurement issue directly.

1. Selection criteria for a nominal anchor

In the context of a floating exchange rate regime, such as applies in South Africa, there are two basic requirements that a nominal anchor should meet: its behavior should be explained by a relatively limited number of factors and should be stable over time with respect to these factors; and it should have demonstrated leading indicator properties of the nominal variable—in particular, inflation—that is the ultimate goal of monetary policy.

If either test is failed, the candidate should be rejected. This dual criteria is intuitive. If the anchor lacks a stable relationship with a few variables (such as interest rates), its behavior cannot be explained, so policymakers cannot know what actions to take to ensure that the anchor remains within a desired target range. But even if its behavior is stable with regard to a few key variables, it would still fail as an anchor if it was not a leading indicator of inflation.

The specific choice of target variable is also critical. While public discourse generally focusses on the headline consumer prices index (CPI), there may be good reasons for monetary policy to focus on a different measure of inflation. The CPI can be affected by changes to indirect taxes, temporary shocks to some of its major components such as food prices, or the way in which housing mortgage costs are reflected in it. The SARB calculates—but does not publish—a measure of “underlying” inflation which corrects for these factors. Other possibilities might be the GDP deflator (though that is only published quarterly and with a longer lag than the CPI), the Producer Price Index, or the domestic demand deflator.

Alternatively, it might be argued that some combination of growth and inflation should form the ultimate target of monetary policy. The implications of this are not pursued here, in part because there is no strong evidence in South Africa of a long-run tradeoff between inflation and output variables. 2/

It is conceivable that no single aggregate or combination of aggregates meets both criteria for selection. In such extreme cases, there would be no guide as to how to hit the inflation target, because there would be no leading indicators of inflationary pressures. This might form a basis for targeting inflation directly, without specifying intermediate targets for other variables. But even those countries that target inflation generally supplement their inflation targets with specific intermediate targets, e.g., MO and M4 in the UK.

Research into aggregates that might be used for intermediate targets in South Africa is limited. It is certainly insufficient to form the basis of a case for an explicit inflation target rooted in the lack of an intermediate target that pass both the selection criteria noted here.

2. Reserve money

One of the potential nominal anchors that has received little attention in South Africa currently is reserve money, despite the focus on this aggregate internationally. This section discusses preliminary econometric analysis of the demand for, and inflation predicting qualities of, reserve money in South Africa.

a. Which parts of reserve money?

Chart 14 breaks down reserve money into its various components—cash outside banks, required reserves with the SARB, excess reserves with the SARB, and cash in till—all reported in constant prices. This exercise is intended to highlight which parts of reserve money are responsible for the principal movements in the aggregate as a whole. 3/ The chart shows that required reserves have changed markedly, notably between 1979 and 1982 when they were raised as part of the attempt to sterilize the inflows emanating from the gold price boom of those years. These changes clearly do not reflect market-based demand for reserve money, but rather the policy decisions of the authorities, and so they should be excluded from consideration of the potential role of reserve money as a nominal anchor.

Chart 14
Chart 14

South Africa: COMPOSITION OF RESERVE MONEY, 1965–95

Citation: IMF Staff Country Reports 1996, 064; 10.5089/9781451840926.002.A006

Source: South African Reserve Bank.

It is debateable whether or not cash in till should also be excluded. The evidence from the chart might suggest that in South Africa’s case, it might be excluded because its trend is so stable. But in addition, changes in that trend would be expected to reflect, at least in part, shifts in banking technology, such as changes in the average distance between bank branches and the nearest SARB cash-distributing branch—i.e., factors that will in general have little to do with inflation.

Excess reserves have commonly been a small portion of reserve money, except from 1989 to 1992, when they were temporarily boosted through a forex scheme in which banks were required to maintain balances with the SARB in order to participate in the scheme. Thus, such excess balances might be excluded from study also.

On these grounds, the component of reserve money on which this chapter focuses is cash in circulation outside the banking system. However, this selection is one of the areas where further study would be warranted to identify whether the results of the tests noted in this chapter would be substantially altered by the addition to cash outside banks of one or all of the parts of reserve money excluded here.

An immediate concern with this aggregate is that its demand may be unstable because it is such a close substitute to noninterest bearing deposit accounts held with banks. Chart 15 shows that while the share of notes and coin outside banks in M1 changes over time, its volatility is perhaps surprisingly low. Thus, it is not immediately obvious that the attempt toestimate the demand for cash outside banks formally is ill-fated

Chart 15
Chart 15

South Africa: SHARE OF NOTES AND COIN OUTSIDE BANKS IN M1, 1965–95

Citation: IMF Staff Country Reports 1996, 064; 10.5089/9781451840926.002.A006

Source: South African Reserve Bank.

b. Identifying the scale variable

The first task was to identify the correct scale variable, i.e., the variable which fundamentally determines the demand for cash in South Africa, as opposed to a variable—such as GDP—that is used to define its velocity. One approach to this task would be to complete a direct survey of households to identify the sources of cash they hold—pay packets, ATMs, etc—and the uses to which that cash is put, such as for retail purchases, spontaneous retail purchases, debt service, purchases by companies, etc. However, such survey data is not available in South Africa.

Accordingly, econometric techniques form the only means of identifying the scale variable. The standard approach to this is to identify a long-run cointegrating vector, using the Johansen technique (See Johansen). The approach employed embodies two simple tests: first, it identifies if there is some linear combination of the series—for cash in circulation, the scale variable being tested, and other series used in the test which might account for changes in cash holdings relative to the scale variable—which yields an error series that is stationary; and second, it checks whether or not the estimated parameters that yield the stationary error series are “sensible”. Unless both tests are passed, the variable being studied is unlikely to be the correct scale variable.

Four potential candidates for the scale variable for notes and coin in circulation outside banks were considered: retail sales, GDP, private consumption, and private nondurable consumption. Chart 16 shows the behavior of notes and coin in circulation outside banks relative to these aggregates. In all cases bar retail sales, these series compare the quarter average for cash outside banks to the scale variable.

Chart 16
Chart 16

South Africa: VELOCITY OF NOTES AND COIN OUTSIDE BANKS, 1965–95

Citation: IMF Staff Country Reports 1996, 064; 10.5089/9781451840926.002.A006

Source: South African Reserve Bank.

The chart shows that these series behave very differently: there is a steep jump in the ratios of nondurable consumption and GDP to cash outside banks from the mid-1970s, with no change in trend in this period in the ratio for retail sales and a slight change for private consumption. The ratios for private consumption and nondurable consumption continue to rise in the 1980s, while those for retail sales and GDP flatten, and even fall somewhat in the case of retail sales.

These differences may be critical to the assessment of the econometric results, and the subsequent discussion of those results refers back to them.

(1) Structural breaks in notes and coin

However, these ratios may be distorted by shifts in the demand for rand cash outside South Africa as the Common Monetary Area countries began to issue their own currencies to circulate alongside the rand, and thus replace it to some extent. These developments represent structural breaks in the series for cash in circulation outside South African banks.

An attempt was made to correct for these breaks. The stock of nonrand cash in circulation in the CMA was taken from IFS. This showed that by the late 1980s, it was equivalent to a little under 3 percent of rand in circulation outside the South African banks. The first point to note is that this ratio is low, and so the structural breaks implied by the introduction of these currencies are not substantial. Nevertheless, the formal cointegration tests were conducted on a break-adjusted series, which corrected for these breaks.

It was not possible, however, to correct for all the breaks in this series, since there may also have been shifts in the demand for rand cash held outside the CMA. It is not clear, however, when shifts in this source of demand for rand were most likely to have taken place, or how large such holdings were (or are) relative to the total rand outside South African banks. Rand currency is used in the region as a means of settlement, even though this breaches CMA foreign exchange regulations, but precisely because of their legal status, information about this stock is anecdotal only. Accordingly, no attempt was made to adjust for these breaks in the series.

(2) Cointegration tests

The cointegration tests were carried out as follows. The break adjusted quarter average series for cash outside banks was seasonally adjusted, and deflated by the deflator for the scale variable being tested—the GDP deflator for GDP, the nondurables deflator for nondurable consumption, etc. The constant price seasonally adjusted scale variable constituted the second series. The data period is from 65:q2 to 95:q4. A VAR lag length of 4 periods was used throughout.

The tests for retail sales were complicated by the lack of the retail sales deflator. The consumer prices index was used as a proxy for this.

A further series was used in each test, a cumulative interest rate term. The motivation for including this series is that it may constitute a useful indicator of shifts in the demand for cash relative to its scale variable. The argument is that the demand for cash not only reflects the opportunity cost—the interest forgone—but also reflects fixed “entry” costs into the banking system. Potential depositors, who may be unfamiliar with banks, may not open bank accounts while nominal interest rates are low because the entry costs of acquiring that familiarity and committing the minimum balance to open a bank account are prohibitive relative to the benefits. But once nominal interest rates rise sufficiently, and are expected to stay at those levels for a sufficient time, it becomes worthwhile for potential depositors to incur those entry costs. This implies that if nominal interest rates rise to historically exceptionally high levels, and then fall back to their historical norms, that the demand for cash will be lower afterwards for given nominal interest rates. The cumulative t-bill rate was used as proxy for this “rachet” feature of behavior, a proxy that has been used with considerable success in models of long-run demand for MO in the UK. 4/ The results of the tests are shown below:

Table 19.

Quarterly Model of Notes and Coin: Long-run Relationships

article image

The co-integration results were weak, just below the 90 percent confidence interval.

The tests show that only private consumption cointegrates with “sensible” parameters, i.e., parameters on the scale variable that can be constrained to 1, with the parameters on the cumulative interest term also correctly signed. This suggests that it is the relevant scale variable.

However, this result should be interpreted with caution. Referring back to Chart above, private consumption May be being identified by this test as the scale variable because the change in the ratio of cash outside banks to it changes less in the late 1970s than it does for either nondurable consumption or GDP.

But these developments may be reflecting the relaxation of the regime of direct controls on the banking system—though many controls were not actually withdrawn totally until the early 1980s—and no account is taken of these developments in the testing procedure described.

This regime shift may have reduced the demand for cash relative to its (true) scale variable. Possible reasons for this include increased returns on bank deposits relative to cash holdings due to falling banking spreads, and decreased credit rationing, raising the benefit to prospective borrowers of relationships with banks as depositors to strengthen their credit standing with those banks. However, the effect on demand for cash of the change in the control regime could equally go the other way, if the termination of credit rationing diminishes the need for a relationship with a bank as a depositor prior to borrowing from it.

The issue here is whether or not a regime of direct controls substantively affects the demand for this particular monetary aggregate, whatever its effects may have been on the demand for other aggregates. This is fundamentally an empirical issue. However, identifying such effects is problematic because there is no readily available proxy for the degree to which the control regime was binding at different times. In this light, studies of the demand for cash in circulation could take one of two approaches to these difficulties: either they could discard all control-era data on the grounds that these may be distorted by unmeasurable effects of the control regime, or they could assert that since a cointegrating relationship has been found for a scale variable that encompasses both the direct control and indirect control eras, that this relationship should be the preferred one.

The difficulties with the first option are that the remaining 15 years of data—from 1980 to 1995—provide insufficient degrees of freedom to distinguish between the various options for the scale variable, and that there are well-known econometric reasons for attempting to model structural breaks where they occur, rather than simply discarding data around them. The difficulty with the second option is that the presumption cannot be independently tested that the control regime had at most a minor effect on the demand for cash, whatever its effects on the demand for other banking aggregates may have been, due to the absence of a good measure of the impact of the controls on the demand for cash. This study will take the second option, on the basis that other studies of demand for cash in economies which shifted from controlled to liberalized regimes have done so without having to model the impact of the control regimes on the demand for cash, carrying the implication that the impact of such regimes on demand for this particular monetary aggregate is limited in practice.

To summarize, there is some evidence that the demand for notes and coin in circulation—break adjusted—can be successfully estimated, with sensible parameter estimates, though as noted, there are a number of avenues that future research could develop to enrich and verify this finding.

Further development of this work would include the estimation of a short-run model of the demand for cash. This is not reported here. The principal reason is that, as discussed below, the way notes and coin in circulation outside the banking system is measured imparts a high degree of volatility to the monthly (and even quarter average) data. This makes an attempt to estimate a short run model somewhat redundant.

c. Inflation predictine properties

The second criterion for identifying a nominal anchor concerns its leading indicator properties. This section reports the results of initial tests of this for break-adjusted cash in circulation outside banks.

The test is conducted in the form of a simple VAR. The dependent variable was the 1-quarter change in the 4-quarter growth rate of the unadjusted CPI. The explanatory variables tested were 4 lags of this same variable, and 4 lags of the 1-quarter change in the 4-quarter growth rate of cash in circulation outside banks. This test simply identifies whether rising annual rates of growth of cash in circulation are followed by rising annual rates of inflation.

This VAR was then tested down, eliminating the insignificant lags of both explanatory variables, using an F-test to establish whether or not the variable deletions were accepted. Deletion of all the lags on cash in circulation lags were accepted by the F-test except that for the first- and fourth-quarter lag. These deletions were rejected at the 1 percent confidence level. This implies that if the four-quarter growth rate of cash in circulation is rising, then inflation will tend to rise in the following quarter, and four quarters later. This constitutes preliminary evidence that this aggregate is a leading indicator of inflation.

However, there are several issues that ought to be addressed in future research into the leading indicator properties of cash in circulation outside banks. One would be to examine the inflation predicting properties of cash in circulation against the SARB’s measure of underlying inflation. A further development would be to test the stability of this finding across a greater number of sub-periods—the structural break test used in this study simply breaks the data set into two, and checks that the parameters are stable across the two sub-sets, but this sub-division could (and should) be considerably extended. It might be interesting to check if the relaxation of monetary controls in the early 1980s affects these leading indicator properties. Third, the test could be carried out testing down from a larger VAR, in which cash in circulation was one of several alternative indicators whose inflation predicting features was being considered. This test would help to identify whether cash in circulation was a better predictor than other indicators, or whether some combination of cash in circulation and other indicators yielded a superior inflation predictor.

The findings reported here suggest that there are grounds for a closer study of cash in circulation from the perspective of its use in a nominal anchor framework: the results are certainly not sufficient to establish the case for it as a principal element of the monetary regime now.

3. Measuring cash in circulation outside banks

One issue that would have to be considered further is the fact that the aggregate currently has a high monthly volatility. This high noise-to-signal ratio in the aggregate is clearly problematic, not least for the tests reported above.

However, this problem appears to derive from the way it is measured. Currently, it is measured at the last working day of each month. But since the daily fluctuations in this aggregate are large, this measurement practice obscures its signal content. The results reported here attempted to tackle the problem by using quarter average data—that is the average on the last working day of each month in a quarter—but this is only partially successful in enhancing the information content of the variable.

A better means of increasing the information content may be to measure it as a within-month period average—say an average of the Fridays of each month, or an average of all days in the month. This approach is taken to the measurement of notes in circulation outside the SARB, 6/ and in this case, it is clear that virtually all of the monthly volatility is eliminated from that series. If further study of the characteristics and behavior of notes and coin in circulation outside the banking system, as suggested above, supports the initial findings reported here, then there might be a case for remeasuring this on a within-month period average basis, and repeating the tests on that data to check once again that the findings are robust to this change in measurement technique. If so, then a strong case might be made for the aggregate to be given greater prominence in the monetary policy structures for South Africa.

4. Conclusion

The results reported here provide preliminary evidence that notes and coin in circulation meet the two key criteria for a nominal anchor: demand for cash appears to be determined principally by private consumption, and movements in cash appear to have leading indicator properties of inflation.

However, further research is necessary to develop and verify these findings, to confirm their robustness, and to measure the performance of other aggregates, including those used more prominently currently such as broad money and credit to the private sector, in a similar manner so that an informative comparison may be made. But these results suggest that narrow aggregates may be worthy of further consideration, and that they may ultimately warrant a more prominent role in the monetary arrangements in South Africa than at present.

References

  • Breedon, F.J, and Fisher P.G., MO: Causes and Consequences. Bank of England Disscussion Paper 1993.

  • Courakis A.S. The Demand for Money in South Africa: Towards a more accurate perspective The South African Journal of Economics Vol 52. 1. 1984.

    • Search Google Scholar
    • Export Citation
  • Hall, S.G., Henry, S.G.B., and Wilcox, J.B., The Long Run determination of the UK Monetary Aggregate. Bank of England Discussion Paper. No 41 1989.

    • Search Google Scholar
    • Export Citation
  • Henry S.G.B., and Peseran B., VAR Models of Inflation Bank of England Disscussion Paper 1993, Vol 33, No 2, pp 231239.

  • Hurn A.S., Interest Rates, Inflation, and the stability of the demand for M3 in South Africa Greek Economic Review Vol. 13 No 2 1991.

  • Hurn A.S., and Muscatelli The long-run properties of the demand for M3 in South Africa The South African Journal of Economics Vol 60. 2. 1992.

    • Search Google Scholar
    • Export Citation
  • Johansen, S Statistical Analysis of Cointegrating Vectors, Journal of Dynamics and control, vol 12 no 2/3, 1988.

  • Whittaker J., The Demand for Money in South Africa: A comment The South African Journal of Economics Vol 53.2. 1985

  • Pretorius, C.J., and M.M. Smal, “A Macroeconomic Examination of the Price-formation Process in the South African Economy,Quarterly Bulletin, South African Reserve Bank, March 1994.

    • Search Google Scholar
    • Export Citation
1/

This comprises notes and coin in circulation outside the banks, cheque and transmission deposits, short-, medium-, and long-term deposits with the banks.

1/

See Pretorius and Smal.

1/

The definition of reserve money may also be extended to include liabilities of the Corporation for Public Deposits, as is the case for data reported in IFS. This extension is not considered here, but could be added in future research.

1/

Hall et al. (1989), and Breedon and Fisher (1992).

2/

The Chi-squared test assesses if the parameter equals 1.

1/

Series 1392m on p S-25 in the March 1996 SARB bulletin.

South Africa: Selected Economic Issues
Author: International Monetary Fund