This Selected Issues paper discusses the policy response by a sample of central banks to the ongoing oil and food price shocks in South Africa, drawing some lessons, which can help put in context developments in the country. The paper discusses first- and second-round effects of “supply shocks,” and attempts to gauge second-round effects in South Africa. The paper also analyzes the factors that have constrained South Africa’s growth since the end of apartheid, by comparing its GDP components and its saving and investment performance with those of a panel of faster-growing countries.


This Selected Issues paper discusses the policy response by a sample of central banks to the ongoing oil and food price shocks in South Africa, drawing some lessons, which can help put in context developments in the country. The paper discusses first- and second-round effects of “supply shocks,” and attempts to gauge second-round effects in South Africa. The paper also analyzes the factors that have constrained South Africa’s growth since the end of apartheid, by comparing its GDP components and its saving and investment performance with those of a panel of faster-growing countries.

I. Monetary Policy and Large Shocks to Relative Prices: International Experience and Implications for South Africa1

1. This note discusses the policy response by a sample of central banks to the ongoing oil and food price shocks (Figure I.1), drawing some lessons which can help put in context developments in South Africa. These shocks amount to an increase in the relative price of those goods; but in certain contexts they may threaten to raise inflation. The shocks and their impact on realized inflation is undermining the stability of inflation expectations and the credibility of central banks in many countries.

Figure I.1.
Figure I.1.

Food and Price Shocks in Perspective

Citation: IMF Staff Country Reports 2008, 347; 10.5089/9781451841060.002.A001

Source: WEO, April and June 2008.

2. Mishkin (2007) argues that an energy price shock may not merit a tightening of the monetary policy stance “as long as the permanent change in relative energy price does not lead to a change in the underlying trend rate in inflation—a crucial assumption.” A key question is thus whether and when that assumption holds true, and in particular, whether it does in South Africa today. We try to answer this question by examining the reaction to oil and food price shocks of underlying inflation and monetary policy in South Africa and a number of other countries.

3. The paper is organized as follows. It starts with a discussion of the first and second round effects of “supply shocks” and attempts to gauge second round effects in South Africa. It follows with a historical look at monetary policy responses to supply shocks in a number of countries after the 1980s. We then examine in some detail a number of recent monetary policy decisions in several inflation-targeting countries to get a sense of what determines central banks responses to these shocks. The final section offers concluding remarks.

A. First and Second Round Effects of Relative Price Shocks

4. First and second round effects may be hard to distinguish in practice. The first round effect of increases in certain food and fuel prices is often taken to be their direct impact on the general price index, although conceptually the indirect cost-push effects on low-margin goods that use food and fuel as inputs should be considered as first round too, given the almost inevitable pass-through. In South Africa, food represents 26 percent of the CPIX basket; gasoline accounts for 5.1 percent of the basket. Using the 2002 input-output matrix, we estimate that the total (direct plus first round indirect) effect of a 1 percent rise in a broad set of food items is a 0.35 percent rise in the cost of consumption.2 In this note we use a narrow definition of first round effects for practical reasons, but even then it is clear that South Africa has a significant exposure to these shocks. Increases in the prices of food and fuel are often seen as “supply shocks”; however, the ongoing shocks are demand driven on a global scale, even if from a single country perspective they look like supply shocks. Thus, in this note we will call them “relative price shocks,” and we will refer to the rise in the prices of other goods as “underlying inflation.”3

5. The second round effects of relative price shocks include their eventual impact on inflation expectations, wage settlements, and price setting in the economy at large. These effects propagate through direct and indirect channels—i.e., inflation expectations might react both to oil price shocks and to headline inflation, which incorporates first round shocks. Second round effects unfold over long periods of time, and may evolve in complex ways. For example, a rise in the price of food will boost CPI directly and raise costs where food is used as an input; but it will also reduce consumers’ purchasing power, depressing demand for other goods (and/or at a later stage).

6. Some evidence of second round effects in South Africa can be obtained by looking at certain correlations. The left-hand chart in Figure I.2. shows correlations between monthly food and nonfood inflation at various time horizons for the period January 2000–March 2008, using seasonally adjusted data. An initial rise in food prices (a shock to relative prices) tends to be followed by upward movements in nonfood inflation, peaking at lags of about six months. The positive correlations are suggestive of pass-through from food to nonfood prices; but they are not too high. The right side chart tries to identify threshold effects by restricting the calculation of correlations to cases in which the change in food prices was above its own sample median. We observe that in this case the correlations are higher earlier, supporting the presence of threshold effects whereby second round effects are more intense when the original shock is larger; we again observe a peak in the sixth month, suggesting semiannual price revisions in South Africa. This simple analysis is not conclusive, though, since food and nonfood inflation might just be responding with different speeds to some third factor, such as exchange rate changes.

Figure I.2.
Figure I.2.

Second Round Effects of Shocks to Food Prices

Citation: IMF Staff Country Reports 2008, 347; 10.5089/9781451841060.002.A001

Sources: Statistics South Africa and IMF staff’s calculations.

7. Further evidence of pass-through from relative price shocks to general inflation was obtained from Granger causality tests. Granger causality tests using one and two lags indicate that food and fuel inflation helps forecast inflation in the rest of the price index in South Africa, but not the other way around. Using more lags in the tests reduces the significance of the results, but in all cases the higher F statistics are those obtained when trying to reject the null that food and fuel inflation does not cause underlying inflation in South Africa.

Table I.1.

Granger Causality Tests

article image
Sample: 1998M01- 2008M03

8. Second round effects from the relative price shocks of the last few years appear to be under way in South Africa. Inflation expectations stand above the upper limit of the inflation target and growth in the CPIX without food and fuel has been accelerating (Figure I.3). Thus, in terms of Mishkin’s “crucial assumption” cited above, it seems that in South Africa underlying trend rates of inflation are being affected by relative price shocks.

Figure I.3.
Figure I.3.

Second Round Effects May Already Be in Train in South Africa

Citation: IMF Staff Country Reports 2008, 347; 10.5089/9781451841060.002.A001

Sources: SARB and Bureau of Economic Research, South Africa.

B. Historical Perspective on Monetary Policy Responses to Relative Price Shocks

9. In this section, we use regression analysis to characterize the behavior of several central banks since the late 1980s, by which time the lessons from the 1970’s and 1980’s oil shocks had been absorbed. In September 2005, in the wake of Hurricane Katrina, which caused major temporary disruptions to the US oil industry, the Reserve Bank of New Zealand (RBNZ) kept its policy rate unchanged with the following statement: “Monetary policy will not attempt to offset the unavoidable first round price effects of the oil spike. However, it will be used to resist any flow-through to ongoing price and wage inflation.” This statement is representative of the view that monetary policy should be concerned only with containing the second round effects of relative price shocks.4 We find that the behavior of the central banks we examine has, on average during the period under study, broadly conformed to that policy view.

10. We estimated policy response functions for Australia, Iceland, New Zealand, Norway, and the UK, and Brazil, Colombia, Chile, Korea, Turkey, and South Africa. These countries now target inflation, although they may not have been inflation targeters during the entire sample period. The basis of all models is a policy rule that allows for differentiated responses to relative price inflation and underlying inflation. A simple model of this type is the following:5


In expression (1), it is the nominal policy interest rate; πtund is underlying inflation in the last 12 months, zt is the nominal depreciation of a country’s currency during the same period, yt is the output gap, and rt is an indicator of relative price movements for key products (food and fuel). Underlying inflation, as noted previously, denotes the rate of inflation excluding food and fuel items. The relative price variable is constructed as the residual from an auxiliary regression of headline inflation against underlying inflation—and thus reflects the component of the change in oil and food prices that is uncorrelated to changes in the prices of all other goods. In richer countries where food has a small weight in the headline price index, underlying inflation should be a more satisfactory predictor of headline inflation, reducing the usefulness of rt (see the appendix for a fuller explanation).6

11. One might expect to see βrel =0 and βund > 0 in equation (1) in cases where inflation expectations are well anchored and underlying inflation is less vulnerable to relative price shocks—that is, where Mishkin’s crucial assumption holds. But his is not something one should expect across the board, because even if a bank is focused on second round effects only, known sensitivity of underlying inflation to relative price shocks could motivate βrel >0.

Table 1.2.

Estimation of Equation 1 for Several Inflation Targeting Countries

article image
Source: IMF staff estimates based on central bank and statistical institute data.Coefficients in boldface are significant at the right-hand-side 5 percent level or better. Regressions constants not shown.

Residuals from a regression of headline inflation on underlying inflation and a constant.

article image
Source: IMF staff estimates based on central bank and statistical institute data.Coefficients in boldface are significant at the right-hand-side 5 percent level or better.

Residuals from a regression of headline inflation on underlying inflation and a constant.

Lag a equals 2 if bimonthly data, and 1 if quarterly data.

12. We estimated several variants of these models, including by replacing πtund with its one-year lead to capture inflation expectations, and replacing rt with a the change in world oil prices measured in local currency.7 The sample periods in some cases start before the adoption of IT in the sample countries (see the appendix for a description of the data and its sources).

13. Equation (1) and its variants may be subject to bias because the right-hand side variables could react to interest rate changes—this is especially clear for the exchange rate. Thus, we ran the ordinary least squares (OLS) regressions replacing the right-hand side variables with their one-period lags. This could also be justified on grounds that some of these variables are observed by policy makers with a lag. We also estimated the regressions with instrumental variables (IV), using as instruments 6-months of lags of our inflation variables and of the exchange rate. Table I.2 shows both OLS and IV results for equation 1; the results are qualitatively similar, but the magnitude of the OLS coefficients appears somewhat more plausible.

Table I.3.

Distribution of the OLS Coefficient on Underlying Inflation

article image
Source: IMF staff’s estimates based on data from central banks.
Table 1.4.

Distribution of the OLS Coefficient on Relative Price Shocks

article image
Source: IMF staff’s estimates based on data from central banks.

14. For the most part, the story seems to hold that central banks in these countries have largely sought to address only second round effects. Tables I.3 and I.4 report the results of OLS regressions for all variants of the basic model, presenting summary information on the two coefficients of highest interest: those on the underlying and relative price inflation variables. A majority of countries and models feature significant responses to underlying inflation, but not to relative price shocks. Regressions for Australia and New Zealand provide the clearest example of this configuration.

15. There are some countries (Colombia, Chile, Iceland, South Africa and the UK) where the policy interest rate responds to relative price inflation in some specifications. In our IV estimate of equation 1, South Africa also displays a positive and significant coefficient on relative price inflation. Two possible explanations could be put forward for these results, which should be seen as suggestive rather than conclusive.

16. Compared to New Zealand or Australia, countries such as South Africa have a higher exposure to food shocks and have in recent memory experienced relatively higher inflation (Figure I.4).8 The central banks in these countries might therefore be more concerned about the effects of relative price shocks on inflation expectations. Thus, results displayed in Tables I.2 and I.4, where these countries’ regressions sometimes show a significant role for relative price shocks, is not surprising.

Figure I.4.
Figure I.4.

Indicators of Inflation in Sample Countries

Citation: IMF Staff Country Reports 2008, 347; 10.5089/9781451841060.002.A001

Sources: Staff estimates based on central banks and statistical institutes data.

17. Also, it may be that the sample period encompasses more than one distinct monetary policy “regime,” whose average policy response function has the features shown above. This might help explain the estimate for the UK. This country has the longest sample in our group, with about half its observations representing the era before the Bank of England was made autonomous and inflation targeting was adopted. In the case of South Africa, monetary policy became more systematic with the adoption of inflation targeting in 2001. Figure I.5 depicts rolling regression estimates of equation 1 for South Africa using 30-month windows. As shown by the chart, the coefficient measuring the policy reaction to relative price shocks diminished and stabilized at a positive level similar to that of the coefficient on underlying inflation after the adoption of inflation targeting. Prior to that, both coefficients are unstable, very large in absolute value, and bear opposite signs.9

Figure I.5.
Figure I.5.

Rolling Estimates of Equation 1 for South Africa

Citation: IMF Staff Country Reports 2008, 347; 10.5089/9781451841060.002.A001

Sources: Statistics South Africa and IMF staff’s estimates.

C. Recent Monetary Policy Decisions Around the World

18. The statements and minutes released to the press by the monetary authorities in our sample countries since mid-2007 indicate that the responses to relative price shocks are not mechanical, but vary according to the relevant context.10 The releases reveal concern over oil and food prices throughout the world. However, these shocks, global in nature and thus a source of inflation risk in virtually all countries, play out differently in each case depending on recent inflation readings, the likely future trajectory of inflation, the state of expectations, the strength of demand, and other factors.11 The three responses we examine are tightening, easing, and lengthening of the policy horizon (another choice is to leave rates unchanged, but we do not examine it since we discuss easing, a more extreme choice).

19. A common action in the last year among the central banks examined for this note was raising the policy rate. Table I.5 presents some detail on a few of these recent tightening decisions. In virtually all cases, the recent rise in food and fuel shocks is mentioned as a source of concern. However, policy tightening occurred in those cases in which the general inflation context raised the vulnerability to those shocks. In particular, tightening occurred where underlying inflation, headline inflation and/or inflation expectations were already a cause for concern, as well as where other inflation pressures were at work, usually as a result of strong aggregate demand or currency depreciation.

Table I.5.

Selected Recent Tightening Decisions

article image
Sources: Various central banks’ press releases and published minutes.

20. There have also been some recent decisions to ease monetary policy despite the global shocks to food and fuel prices (Table I.6). In most such decisions two factors were key: (i) the risk of recession either mitigated or outweighed the inflation risk from relative price shocks, and (ii) trends in headline and underlying inflation seemed reasonably subdued at the time of the easing decisions. An exception was Turkey, where the inflation rate was well above its target; yet, even in that case, an argument was made that previous tightening and softening demand conditions implied that inflation would move down toward the target.

Table I.6.

Selected Recent Easing Decisions

article image
Sources: Various central banks’ press releases and published minutes.

21. In addition to changing their policy rates, some central banks appear to have decided to implicitly or explicitly extend their time horizon for bringing inflation back to target once it is forecast to exceed it. In the recent experience, the marked acceleration of inflation, combined with adherence to a gradual approach to tightening (the usual interest rate increases are still 25 or 50 basis points, with no indication that policy lags are getting shorter), tend to imply a longer time until inflation returns to target. In Chile, for example, the June 2008 decision states the central bank’s commitment to “reduce the current elevated inflation toward 3 percent in the policy horizon” (emphasis added). This choice is consistent with the overall inflation targeting framework.

22. The only case in our sample where an explicit change was made to the inflation targeting framework was Turkey. In June 2008, Turkey’s central bank, in conjunction with the government, announced it would increase its inflation targets for 2009, 2010 and 2011 to 7.5, 6.5, and 5.5 percent, respectively, from 4 percent. Having failed to meet their inflation targets of 5 percent in 2006 and 4 percent in 2007, and with headline inflation pushing into double digit range, the monetary authorities argued that food and energy shocks that were expected to persist in the coming months would make it impossible to meet the inflation target in the near term. They also explained that expectations were becoming increasingly backward-looking, and that the official target was losing effectiveness as an anchor for expectations. Finally, they indicated that their decision to revise the inflation targets for the next several years did “not necessarily mean that monetary policy will be looser in the forthcoming period.” Although this decision may have brought about more realistic targets, and thus may improve monetary policy transparency, it entails a risk that inflation expectations could become unhinged.

23. Where does South Africa’s Reserve Bank stand against this context? The Monetary Policy Committee (MPC) statements emphasize second round effects; but with a high degree of concern over the original shocks themselves. For example, the MPC statement described oil prices as a “threat” after Brent crude reached $70 dollars a barrel in April 2006. Still, like other central banks, the SARB reacted to these shocks only when context, especially ongoing headline and underlying measures of inflation, heightened their impact. Thus, it was not until June 2006 that the SARB started a tightening cycle, motivated by a deterioration in the general outlook for inflation.

24. After its interruption in early 2008, tightening resumed in April 2008, with supply shocks taking on a more prominent place in SARB’s MPC statements. In that release, the SARB noted that the deteriorated inflation outlook reflected the impact of “a series of supply side shocks.” A point highlighted in the press release was the large increase in inflation expectations, which have now breached the inflation target band, as shown earlier in Figure I.3. Essentially, the argument appeared to be that second round effects were under way and needed to be countered. Thus, hikes in oil and food prices were seen as requiring a strong response, as expectations, core inflation, and credibility were at risk.

25. The SARB seems also to have chosen, implicitly, to accept a possibly longer time to return inflation to the target band (Figure I.6). This is especially clear in the June 2008 decision, when following a steep increase in headline inflation, and against expectations of a larger-than-usual interest rate hike, the SARB decided to increase the repo rate by the usual 50 basis points. In its policy statement, the SARB acknowledged that inflation would now take longer to come back within the 3–6 percent target band.

Figure I.6.
Figure I.6.

Implicit Lengthening of the Policy Horizon in South Africa

Citation: IMF Staff Country Reports 2008, 347; 10.5089/9781451841060.002.A001

Sources: SARB, Statistics SA.

D. Concluding Remarks

26. Persistent and large shocks may have significant second round effects in some contexts which could destabilize expectations and hurt central bank credibility. Speaking about oil shocks, Ed Gramlich (2004) said that although some additional unemployment and inflation may have to be accepted, the best response is likely to involve policy rate increases because “[…] the worst possible outcome is for monetary policy makers to let inflation come loose from its moorings.” This viewpoint, rooted in the lessons from the oil shocks of the 1970s and 1980s, finds echo in the policies examined in this note.

27. Inflation targeting central banks generally aim to distinguish between first and second round effects, largely seeking to contain the latter even as the former are accepted; but their concrete actions are context-specific. Such a policy approach involves a relatively more aggressive stance when already high inflation readings and rising inflation expectations complicate the outlook, which is especially likely if the goods whose prices are spiking represent a large proportion of the consumption basket, and if there is a history of high and variable inflation. Similarly, a more aggressive stance is observed when strong demand and other factors generate additional inflation pressures. Although its policy actions are unable to offset first round effects, a central bank can help anchor expectations by signaling its commitment to the inflation target. Precisely because expectations are put at risk by relative price shocks, the expectations channel of monetary policy takes on special importance in those circumstances.

28. The SARB has applied the approach described above to address the ongoing shocks to food and fuel prices by tightening its policy stance; however, it is facing a particularly difficult challenge going forward. Responding to domestic inflation risks, the SARB started tightening its policy stance earlier than other banks, and the recent intensification of food and fuel price shocks has merited additional tightening actions. But these shocks come at a stage in which the SARB might have expected to be ending its tightening cycle. The rise in headline inflation despite previous tightening puts a premium on effective communication of the nature of the ongoing shocks, the reasons for the breaching of the target band, and the role of monetary policy actions. Also, the increasing debt service burden on borrowers who have seen their interest rate increase by 500 basis points since mid-2006 necessitates careful monitoring of rising risks to the quality of loans. In this context, SARB’s recent decisions to tighten policies while allowing a longer time to bring inflation back to target are appropriate and in line with good practice around the world. Looking forward, a further challenge to the SARB’s communication strategy may arise from the change in the CPI weights in 2009, which will modify the way food and fuel prices affect measured inflation in South Africa.


Derivation of the regressor rt

The overall or “headline” price index can be written as h=w u + (1-w) n, where u is the underlying inflation index, w is its weight in the consumption basket, and n is the index for the rest of the goods (those responsible for non-underlying inflation). Then the relationship between the various inflations is


Now, we can write non-underlying inflation πn as consisting of an element which is correlated to underlying inflation, and an element vt which is orthogonal to underlying inflation: πn = a0 + a1 πu + v, where we have omitted time indices for convenience. If underlying and non-underlying inflation move closely, vt will explain a small part of the variation in non-underlying inflation. Then we can rewrite headline inflation as follows:


This expression looks very much like a regression of headline on underlying inflation (the difference between this decomposition and a regression of πh on πu is that the expressions in the first two brackets will be constant by construction). In the text, we have been interested in the residual from such a regression, which is the variable called rt. As we can see, roughly speaking, the importance of the residual from a regression of headline inflation on underlying inflation depends on the weight of non-underlying prices in the consumption basket, 1-w. If w is very close to 1, the regression will have excellent fit and the residual will account for a small part of the variation in headline inflation. Fit also depends on how closely correlated underlying and non-underlying inflations are: if vt explains a small portion of the variance of πn, fit of the headline inflation regression will also be good. Moreover, if a1 is large, some effects of non-underlying inflation on total inflation (and ultimately on policy) will be confounded with the effects of underlying inflation on total inflation (and policy). Hence our interest in the orthogonal component vt, which combined with the weight of goods in the non-underlying index, 1-w, is obtained as a residual rt from the regression of headline on underlying inflation.

Data sources

Oil prices and exchange rates are from the World Economic Outlook and International Financial Statistics, respectively. Other data sources are shown in the table below. Output gaps were estimated applying the Hodrick-Prescott filter to the output indicators.

article image


  • Aoki, Kosuke (2001), “Optimal Monetary Policy Responses to Relative Price Changes,” Journal of Monetary Economics, v.48 (2001), 5580

    • Search Google Scholar
    • Export Citation
  • Clarida, Richard, Jordi Gali and Mark Gertler (1998), “Monetary Policy Rules in practice: Some International Evidence,” European economic Review, v. 42, 1033 -1067

    • Search Google Scholar
    • Export Citation
  • Gramlich, Edward (2004), “Oil Shocks and Monetary Policy,” remarks at the Annual Economic Luncheon, Federal Reserve Bank of Kansas City, September 16, 2004,

    • Search Google Scholar
    • Export Citation
  • Kamps, Cristophe and Christian Pierdzioch (2002), “Monetary Policy Rules and Oil Price Shocks,” Kiel Institute of World Economics Working Paper 1090, 38 p.

    • Search Google Scholar
    • Export Citation
  • Mishkin, Frederic (2007), “Headline versus Core Inflation in the Conduct of Monetary Policy,” Speech at the Conference on Business Cycles, International transmission, and Macroeconomic Policies, HEC, Montreal October 2007

    • Search Google Scholar
    • Export Citation
  • Statistics South Africa (2006), Final Supply and Use Tables, 2002: An Input-Output Framework, Statistics South Africa, Pretoria, 69 pages. (The data can be downloaded from

    • Search Google Scholar
    • Export Citation
  • John B. Taylor (1993), “Discretion versus policy rules in practice,” Carnegie-Rochester Conference Series on Public Policy 39 (1993) 195214

    • Search Google Scholar
    • Export Citation

The press releases, official monetary policy statements and minutes of monetary policy meetings used were obtained from the following central bank websites:










New Zealand:

South Africa:



Prepared by Alfredo Cuevas.


The goods are agricultural products, meats, fish, fruit and vegetables, oils and fats, and several other products. These goods account for 26 percent of household consumption in the 2002 input-output table. See Statistics South Africa (2006). CPIX is CPI excluding the interest on mortgage loans.


In some countries this would be similar to the concept of “core inflation.” However, the exact definition of core inflation varies across countries.


This approach can be interpreted as implicitly targeting underlying inflation, whatever the formal target may be. In fact, there is a small literature on the inflation index one should target in the presence of relative price shocks. These papers often favor targeting indexes that abstract from such shocks, such as “core” or domestic inflation indexes. See for example Aoki (2001) and Kamps and Pierdzioch (2002).


Taylor’s (1993) pioneering work on policy rules had an even simpler rule, including on the right hand side only general inflation and the output gap.


The use of this variable permits us to include in the analysis countries for which indices of food and fuel prices were not available. Were such information available, one could substitute the growth in that index for rt in expression (1). The estimated regression coefficients and their interpretation would be different, but the information contained in the analysis would be the same.


All models were also run with a lag of it on the right hand side to allow for gradual policy reactions as in Clarida, Gali and Gertler (1998). Including the lagged variable reduced the explanatory power of other variables in most regressions. These results are omitted in the interest of brevity.


Note, however, that different societies may not have the same tolerance of inflation, as suggested by the pattern shown in Figure I.4.


Charts for rolling regression coefficients estimated with IV, and also using directly food and fuel inflation instead of rt (not shown) have similar characteristics. The behavior of these coefficients in the early part of our sample is likely due to the strong movements in the interest rate in response to large exchange rate swings observed at that time.


We added Canada to the countries discussed in this section to increase the variety of experiences analyzed.


Also, the pass-through from international to domestic prices is affected by tax and subsidy policies.

South Africa: Selected Issues
Author: International Monetary Fund