Food Inflation in Sub-Saharan Africa
Causes and Policy Implications1

Contributor Notes

Authors’ E-Mail Addresses: EAlper@imf.org, NHobdari@imf.org, AUppal@imf.org

This paper analyzes food inflation trends in Sub-Saharan Africa (SSA) from 2000 to 2016 using two novel datasets of disaggregated CPI baskets. Average food inflation is higher, more volatile, and similarly persistent as non-food non-fuel (NF/NF) inflation, especially in low-income countries (LICs) in SSA. We find evidence that food inflation became less persistent from 2009 onwards, related to recent improvements in monetary policy frameworks. We also find that high food prices are driven mainly by non-tradable food in SSA and there is incomplete pass-through from world food and fuel prices and exchange rates to domestic food prices. Taken together, these finding suggest that central banks in low-income countries with high and persistent food inflation should continue to pay attention to headline inflation to anchor inflation expectations. Other policy levers include reducing tariffs and improving storage and transport infrastructure to reduce food pressures.

Abstract

This paper analyzes food inflation trends in Sub-Saharan Africa (SSA) from 2000 to 2016 using two novel datasets of disaggregated CPI baskets. Average food inflation is higher, more volatile, and similarly persistent as non-food non-fuel (NF/NF) inflation, especially in low-income countries (LICs) in SSA. We find evidence that food inflation became less persistent from 2009 onwards, related to recent improvements in monetary policy frameworks. We also find that high food prices are driven mainly by non-tradable food in SSA and there is incomplete pass-through from world food and fuel prices and exchange rates to domestic food prices. Taken together, these finding suggest that central banks in low-income countries with high and persistent food inflation should continue to pay attention to headline inflation to anchor inflation expectations. Other policy levers include reducing tariffs and improving storage and transport infrastructure to reduce food pressures.

I. Introduction

“Because shocks to commodity price inflation are typically beyond the control of policymakers, hard to predict, and often not sustained, central banks seeking to establish credibility are generally better off setting and communicating their monetary policy in terms of underlying inflation rather than headline inflation.”— IMF, 2011

For a long time, the consensus in the literature has been that central banks should not respond to changes in food and energy prices as such changes tend to be quickly reversed and that they are highly volatile (Motley, 1997; Mishkin, 2007; Kiley, 2008). Indeed, the textbook approach is that monetary policy should ‘look through’ supply shocks such as a sudden increase in food prices (Mohanty, 2014). This view has been internalized by central banks across the world and while most central banks continue to target headline inflation, their policy decisions are often informed by measures of “core” inflation that strip the impact of food and energy prices, which are seen as providing a clearer image of underlying price developments.2

Using inflation measures excluding food and fuel prices to gauge the monetary policy stance is not only prevalent in advanced economies (AEs), but also in a growing number of emerging market economies (EMs) and low-income countries (LICs), including more recently in a number of SSA countries with evolving monetary frameworks. Indeed, central banking in SSA has transformed over the past few decades. Prior to the 1990s, interest rates were largely administratively controlled and central banks were required to abide by the doctrines of the fiscal authorities which largely directed monetary policy to fill government financing gaps. Furthermore, parallel foreign exchange markets were ubiquitous during this period reflecting exchange rate controls in a context of monetary financing of large fiscal deficits. Since then, however, there have been two key developments in SSA. First was the move towards more flexible exchange rate regimes and second was that price stability became the de jure principal aim of monetary policy in most countries. The result of these reforms is that monetary policy frameworks in SSA are now increasingly aligned with those in AEs (Ajakaiye and O’Connell, 2011; IMF, 2014). As part of this alignment, a growing number of central banks in SSA have also emphasized the use of core inflation in a similar way to those in AEs.3

However, while food price inflation is less persistent than non-food price inflation in the AEs, recent evidence shows that this is not necessarily the case for LICs, leading some to argue that food prices should not be excluded from core measures of inflation in these countries (Walsh, 2011; Rangasamy, 2011; Durevall et al., 2013). In addition, a higher share of food in the consumption basket of LICs makes core inflation a poor measure of changes in the cost of living (Furceri et al., 2015; Durevall et al., 2013). Indeed, the share of food in the consumption basket is 40 percent on average in SSA while the equivalent for AEs and EMs is 15 and 30 percent, respectively (Figure I.1)

Figure I.1
Figure I.1

Share of Food in the Consumption Basket

Citation: IMF Working Papers 2016, 247; 10.5089/9781475563115.001.A001

Source: Country statistical sources, WEO database and Haver Analytics.

Against this background, using two novel datasets we compare average food inflation with that of non-food non-fuel (NF/NF) in SSA, as well comparing their volatility and persistence. We also assess whether food inflation drives NF/NF inflation in SSA. This is followed by an assessment of the importance of the pass-through from international food prices, fuel prices, and exchange rates in explaining domestic food inflation. Finally, using highly disaggregated food items, we assess whether these external drivers have heterogeneous effects on the different food items.

Our findings for the period 2000 to 2016 indicate the following:

  • Food inflation has been higher than, more volatile than, and at least as persistent as NF/NF inflation, especially in LICs. In the aftermath of the 2007-2008 international food price surge, from 2009 onwards, food inflation has on average remained higher than NF/NF inflation (1.4 percentage points per year) and more volatile. On average, however, it is significantly less persistent than NF/NF inflation. Our regression results on the drivers of persistence suggest that improvements in macroeconomic management across SSA, and also the adoption of forward-looking monetary policy frameworks in a number of countries are associated with less persistent food inflation relative to NF/NF inflation.

  • Granger causality tests indicate that, across the majority of countries in SSA, food inflation does not cause NF/NF inflation, nor does NF/NF inflation always cause food inflation.

  • Over a third of SSA countries have experienced statistically significant pass-through to domestic food prices from changes in world food prices and in exchange rates. The average pass-throughs are 32 and 17 percent, respectively.

  • Higher food inflation in SSA is most prominent in LICs. Food inflation is higher in fresh (and thus non- tradable) food items. In addition,4 and there are more countries with a statistically significant pass-through from world food prices and exchanges rates to non-fresh food prices.

Taken together, these findings suggest that measures of inflation that exclude food prices would likely underestimate underlying price pressures in SSA and other LICs where food inflation remains significantly high and persistent. In such cases, monetary policy should continue to monitor food inflation and seek to reduce the second-round impact of food price shocks on NF/NF prices. Additionally, by focusing their efforts to establish credibility, central banks may also reduce the persistence of food inflation. Another conclusion is that policies that increase tradability of food items, including, for example, reducing tariffs and improving storage and transportation infrastructure could potentially reduce food inflation pressures in LICs.

The remainder of the paper is organized as follows: Section II describes the two datasets; Section III describes food and NF/NF inflation dynamics, including assessing the drivers of inflation persistence; Section IV explores external and structural drivers of food inflation using pass-through regressions; and Section V concludes with policy implications.

II. Data

We form two novel datasets on consumer prices in SSA countries, both non-seasonally adjusted, by gathering detailed monthly CPI statistics from January 2000 to May 2016 for SSA countries from published reports in country statistical agencies, and IMF staff data based on submissions by country authorities.

  • The first dataset on SSA CPI consists of monthly data on headline CPI and its 12 major subcomponents (Table II.1 during 2000-16 for 41 SSA countries.5 Other than monthly price series for the headline CPI and its subcomponents, we also gathered CPI basket weights for each country during 2000 to 2016. For the remainder of this paper, “food prices” refer to item number 1 in Table II.1 “Food and Nonalcoholic Beverages”; fuel prices refer to item number 4 “Housing, water, electricity, gas and other fuels”; and NF/NF is calculated by subtracting item number 1 and item number 4 from headline CPI after multiplying them with their respective weights and then reweighing the difference.6 There are on average 171 months (over 14 years) of observations with a minimum of 114 (Republic of Congo) and a maximum of 197 (Benin, Botswana, Central African Republic, Ghana, Guinea, Kenya, Madagascar, Nigeria, and Togo) observations.

Table II.1

CPI Subcomponents

article image
  • The second dataset consists of monthly data on highly detailed food subcomponents (up to 91 items) as well as their CPI weights for 20 SSA countries (Annex 2 Table A.3). For comparability to international food prices, we aggregate the food price series using their individual consumption basket weights into 9 categories: bakery, coffee, fish, fruit, meat, milk, oils, sugar, and vegetables. These categories are fairly representative of a typical food basket: on average these 9 categories make up 87 percent of the food basket. There is on average around 8.5 years of monthly data with a minimum of just under 4 years of monthly observations (Ghana) and a maximum of nearly 15 years of monthly observations (Madagascar).

For the purposes of this paper, we classify the SSA countries with 2015 per capita GDP valued by purchasing power parity greater than US$5,000 as middle-income and the remaining as low-income countries.7 Based on this classification, in the first dataset 11 are middle-income and 30 are low-income countries (Annex 2 Table A.2), and in the second dataset 5 are middle-income and 15 are low-income countries (Annex 2 Table A.3).

In addition to the two novel datasets above, the analysis in the following sections of this paper uses a number of other macroeconomic series as well as structural indicators (see Annex 2 Table A.1 for a summary of sources and definitions).

III. Inflation in SSA in the New Millennium

Headline consumer price inflation in SSA has, on average, been on a trend decline over the last two decades, reflecting improved macroeconomic policy management in most countries in the region. Nevertheless, inflation in SSA remained well above the levels prevailing in other regions (Figure III.1) The experience with disinflation in SSA has also varied widely: about a quarter of SSA countries actually experienced increasing inflation trends during 2000-16, and consumer price inflation dispersion in SSA remained much higher relative to AEs or EMs.

Figure III.1:
Figure III.1:

Annual Consumer Price Inflation Across the World (%)

Citation: IMF Working Papers 2016, 247; 10.5089/9781475563115.001.A001

Source: WEO database.

A. Average Inflation

We first derive average quarterly inflation rates from monthly indices for 41 SSA countries. The annualized quarterly inflation rates for the full sample of countries as well as various sub-groupings reveal that food inflation is greater than NF/NF inflation for all main sub-groupings (Table III.1).8 For the full sample of 41 SSA countries, the annual average difference is close to 1.8 percentage points during 2000-16. This is a non-trivial difference: over the 16-year span included in this paper, food inflation has been cumulatively higher by over thirty percentage points relative to NF/NF inflation. While Table III.1 uses a simple country average to derive the values for the different sub-groupings, the key result, that food inflation consistently exceeds NF/NF inflation, is robust to using GDP valued in purchasing power parity weighted averages (Annex 2 Table A.4).

Table III.1

Average Inflation Rates in SSA and Sub-groups

article image
Source: Country statistical sources and authors’ calculations.

The largest difference between food and NF/NF inflation is among the members of the Central African Economic and Monetary Community (CEMAC) and the East African Community (EAC), where the difference on average is about 2½ percentage points per year. MICs turn out to be more similar to AEs9 while LICs have experienced higher food inflation compared to NF/NF inflation on average by around 2.1 percentage points annually.

The wedge between food and NF/NF inflation has narrowed since 2009 (i.e., following the international food price surge of 2007-08), but is still close to 1.4 percentage points annually. Around sixty percent of this decline in the wedge has been driven by lower food inflation. Indeed, inflation for most food sub-groupings experienced a reduction since 2009 with the exception of EAC countries where the difference between food and NF/NF inflation actually increased to just over 3 percentage points per year.

Country-specific average inflation information (Annex 2 Table A.5) reveals that 90 percent of SSA countries from 2000 to 2016 experienced food inflation exceeding NF/NF inflation. For the 2009-16 period, which excludes the 2007-08 international food price surge, the figure drops to just over 70 percent of SSA countries witnessing food inflation exceeding NF/NF inflation. Another way to assess the cross-country variation between food and NF/NF inflation is to just look at the percentage of quarters when food inflation is greater than NF/NF inflation (Figure III.2). About % of countries in SSA have quarterly food inflation greater than NF/NF inflation more than half the time over the period 2000-16. On average in SSA, quarterly food inflation is greater than quarterly NF/NF inflation about sixty percent of the time. These two observations indicate higher persistence of food inflation relative to NF/NF inflation. However, when looking at the 2009-16 period only, over half of the countries saw a decline in the percentage of quarters when food inflation is greater than NF/NF inflation while only a third saw a rise and a tenth saw no change. This suggests that the persistence of food inflation may have declined in the latter period.

Figure III.2
Figure III.2

Share of quarters with Food Inflation exceeding NF/NF Inflation (2000-16)

Citation: IMF Working Papers 2016, 247; 10.5089/9781475563115.001.A001

Source: Author calculations.

We therefore conclude that food inflation has been an important contributor to headline inflation in the vast majority of SSA countries from 2000 to 2016. However, food inflation appears to be a less significant, albeit still important, driver of headline inflation since 2009. This result is likely driven by the exclusion of the international food price surge but other factors such as an increased resilience to food price shocks, including the 2011 drought in the Horn of Africa. More modernized central banks and improved macroeconomic management may have played an important a role.

B. Volatility of Inflation

Using quarterly average inflation, volatility is calculated as the coefficient of variation (i.e., the standard deviation divided by the mean multiplied by 100) for each country in the sample. We then show the median volatility for the same country groupings in SSA (Table III.2).10 We observe that the volatility of food inflation in SSA is over 50 percent higher than that of NF/NF inflation. However, the differences in volatilities of food and NF/NF inflation is driven by LICs as the difference in the volatilities is negligible for MICs. The difference is highest for members of the West African Economic and Monetary Union (WAEMU), but for CEMAC and the oil exporters food inflation is less volatile than NF/NF inflation, on average.11

The difference of the median volatilities of food inflation and NF/NF inflation has increased slightly relative to NF/NF inflation since 2009 in SSA, despite the broad stability in global food prices. This appears to be driven by a nearly 50 percent increase in the difference for LICs.

Table III.2:

Volatility of Inflation Rates in SSA and Sub-groups

article image
Notes: The median volatility is used for each grouping.Source: Country statistical sources and authors’ calculations.

Country-specific inflation volatility (Annex 2 Table A.6) reveals that there is considerable heterogeneity in the volatilities across different countries. Indeed, across both the 2000-16 and 2009-16 periods close to half of the countries have broadly similar volatilities for food and NF/NF inflation but as the table above indicates, the median volatility for food inflation across both periods is higher than that for NF/NF inflation. 12

We therefore conclude that food inflation in SSA is, on average, more volatile than NF/NF inflation during both the 2000-16 and 2009-16 periods, but also note the considerable heterogeneity across countries.

C. Persistence of Inflation

We next turn to persistence of shocks to food and NF/NF inflation in SSA. As discussed earlier, if food price shocks are relatively transitory, this could support the focus on inflation excluding food in assessing the monetary policy stance.13

There are various alternative methods to calculate the persistence of a series. One approach (Andrews and Chen, 1994) to measure persistence is to calculate the sum of the autoregressive coefficients (SARC).

We therefore estimate a country-by-country univariate regression of the following form:

πtz=Σi=1Lβiπtiz+Σi=14ρiQi,t+ɛtz(1)

where πtz represents annualized quarterly inflation for item Z (food or NF/NF) at t. Qit represent quarterly time dummies to control for seasonal variation.

For each country in the sample, we estimate the autoregressive equation for both food and NF/NF inflation, four times with the number of quarterly lags, L, ranging between 1 and 4. The number of lags is selected on the basis of the Akaike Information Criterion and then the β coefficients are summed up to yield the SARC. A higher estimate of the SARC indicates greater persistence. Medians of the SARC estimates across all the countries in our sample reveal that, contrary to conventional wisdom, food inflation is more persistent than NF/NF inflation in SSA over the period 2000-16 (Table III.3). However, from 2009 onwards (i.e., after stripping out the effect of the international food price shock in 2007-08), food inflation is significantly less persistent than NF/NF inflation. This result is driven by the decreasing persistence of food inflation in LICs.

Table III.3:

Persistence of Different Inflation Rates - Whole Sample

article image
Notes: Inflation used for these calculations is QoQ, annualized. The median SARC estimate is used for each grouping.Source: Author calculations

SARC estimates for each country (Annex 2 Table A.7) and Figure III.3 below provide the distribution of these estimates. The distribution of the SARC estimates for food inflation appears to have shifted down and leftwards from the period 2000-16 to 2009-16 (Figure III.3 Panel A). This shows that the decreasing persistence of food inflation is broadly representative across SSA. The persistence estimates of NF/NF inflation show a flattening of the distribution with the mean remaining broadly the same (Figure III.3 Panel B).14

Figure III.3:
Figure III.3:

Distribution of Inflation Persistence (SARC Estimates) across SSA

Citation: IMF Working Papers 2016, 247; 10.5089/9781475563115.001.A001

Notes: Vertical lines represent the means. Inflation used for these calculations is QoQ, annualized.Source: Author calculations

The above finding corresponds with the summary results in Table III.3. That is, in the period from 2000 to 2016, food inflation was at least as persistent as NF/NF inflation (and in many cases more persistent) with around 60 percent of countries experiencing food inflation that is more persistent than NF/NF inflation. But in the 2009-16 period, food inflation has become much less persistent, with close to 70 percent of countries in SSA witnessing a fall in the persistence of food inflation since the 2000-16 period. Indeed, in the 2009-16 period, food inflation is less persistent than NF/NF inflation in over 60 percent of countries.

The shift in the distribution seems to reflect, in part, the exclusion of the international food price surge of 2007-08.15

Given the different distribution in 2009-16, we attempt to identify which factors affect persistence of food inflation relative to NF/NF inflation (i.e., the wedge between the two) during this period. As this period has seen significant institutional development, particularly in monetary policy frameworks (IMF, 2015), we expect that improved monetary policy management (e.g., through inflation targeting) has helped anchor inflation expectations and would therefore reduce the persistence of food inflation relative to NF/NF inflation.

To test our hypothesis, we run a cross-sectional regression on the wedge between the persistence of food and NF/NF inflation on an index that measures macroeconomic management (a higher value on this index indicates that monetary and exchange policies promote price stability and better aggregate demand policy framework including on fiscal policy) and a dummy variable for countries that have inflation targeting. We also include a number of structural variables, such as GDP per capita, a categorical variable capturing the de Jure central bank independence from government financing (Jacome et al (2012), total droughts, logistics index capturing the quality of infrastructure for storage and transportation, as controls (Table III.4).16 We also include the share of China imports to control for the dampening effect on NF/NF inflation from increasing Chinese imports in SSA.

Furthermore, we estimate persistence of food inflation and NF/NF inflation separately with the identical specification. This would allow us to ascertain whether the changes in the wedge are being driven by changes in the persistence of food inflation or NF/NF inflation.

Table III.4:

Regression on the Estimates of the Persistence (SARC) of Food Inflation, NFNF Inflation, and the Wedge in the 2009-16 Period

article image
Standard errors in brackets

p<0.01

p<0.05

p<0.1

The results in the final column confirm our hypothesis that having an inflation target and better macroeconomic management (as measured by the CPIA Macro index) is associated with a smaller wedge between the persistence of food inflation and NF/NF inflation. Results in columns one and two indicate that the decrease in the wedge is driven by the negative correlation between these institutional variables and food persistence. This suggests that improvements in monetary policy management may, through better anchoring of inflation expectations, lead to food price shocks being considered transitory.

Only two of the controls are statistically significant. First, more imports from China are associated with a lower wedge. This is driven by more persistent NF/NF inflation. Given imports from China will be largely nonfood imports, we would expect that increasing the amount of imports from China is likely to increase the persistence of NF/NF inflation. Second, having an exchange rate peg is also associated with a lower wedge. This is driven by less persistent food inflation. Exchange rate pegs can promote price stability (Ghosh et al., 1996) and in a similar way to the effect of having an inflation target, a peg can therefore help anchor inflation expectations leading food price shocks to be considered transitory. After controlling for the CPIA macro, share of Chinese imports, inflation targeting and the exchange rate peg dummies, remaining structural variables were not statistically significant possibly also due to the small sample size or collinearity with CPIA macro and other variables.

We therefore conclude that shocks to food prices were at least as persistent as shocks to NF/NF prices in SSA during 2000-16. However, since 2009, for the majority of countries in SSA, shocks to food prices are less persistent than shocks to NF/NF prices. Furthermore, we find that improved macroeconomic management, in particular, policies promoting price stability and anchoring inflation expectations reduce the persistence of food inflation relative to NF/NF inflation.

D. Transmission Mechanism To and From Food Prices

We next test for potential second-order effects of food inflation in order to ascertain whether food inflation drives NF/NF inflation. Clearly, if food inflation drove NF/NF inflation significantly, then ignoring food inflation could lead to misguided monetary policy. We use Granger-causality tests between food and NF/NF inflation and find that the second-order effects from food inflation is limited.17 Empirical tests suggest that in only a fifth of the sample there is one-way Granger-causality from food to NF/NF inflation (Annex 2 Table A.12). In another fifth of the sample, the Granger causation runs both ways.

The monetary policy framework does not seem to make a difference, with food inflation Granger-causing NF/NF inflation both in countries operating a form of forward-looking monetary policy framework (such as Kenya and South Africa), and in countries with exchange rate pegs (such as Cameroon, Gabon, and Mali, which are part of regional currency unions).

We conclude that neither food inflation nor NF/NF inflation drives the other in SSA.18

Recapping, the findings presented in this section suggest the following for SSA:

  • Average food inflation is consistently higher than NF/NF inflation during 2000-16.

  • Food inflation is on average more volatile than NF/NF inflation throughout 2000-16.

  • Food price shocks have been at least as persistent as NF/NF price shocks on average during 2000-16. However, since 2009, the persistence of food price shocks has declined markedly for the majority of countries in SSA. Findings suggest that improved central bank credibility and monetary policy management have played an important role. Indeed, central banks across have been modernizing and building credibility of late (Ajakaiye and O’Connell, 2011; IMF, 2015).

IV. Drivers of Food Prices in SSA

Having assessed some of the key characteristics of food inflation in SSA, we next assess the external and structural drivers of food prices in SSA during 2000-16. While domestic food prices in SSA countries have been correlated with global food prices, the strength of the correlation seems to have weakened markedly in recent years. Indeed, domestic food prices in SSA have continued to increase at a rapid pace lately, despite international food prices having plateaued since 2010 (Figure IV.1).19 The figure therefore suggests that at least part of the reason for rising food prices is domestically driven. The domestic food price increase has been particularly pronounced in LICs, although this appears to be driven, in part, by movements in exchange rates. Indeed, we found in the previous section that exchange rate pegs could guard against such movements and hence reduce the persistence of food inflation. The next section explores the external drivers systematically.

Figure IV.1:
Figure IV.1:

Domestic and World Food Price Developments

Citation: IMF Working Papers 2016, 247; 10.5089/9781475563115.001.A001

Source: Author calculations.

A. External Drivers of Food Inflation in SSA

This section quantifies pass-throughs from world food and fuel inflation, and exchange rate depreciation to domestic food inflation in SSA. Given the limited time span, the analysis is based on the full sample only. We estimate a country-by-country regression of quarterly domestic food inflation on current and 4 lags of: quarterly world food inflation; quarterly world fuel price inflation; and quarterly nominal exchange rate depreciation20 and control for 4 lags of domestic food inflation. We also include a time trend and quarterly dummies to control for seasonal variation. The regression is estimated using the following equation:

πtDOM=Σi=14βiπtiDOM+Σi=04γiπtiINT+Σi=04δiFtiINT+Σi=04kiEtiUSD+Σi=14ρiQi,t+t+ɛt(2)

where πtDOM is annualized quarterly domestic food inflation at t, πtINT is annualized quarterly world food inflation at time t, FtINT is annualized quarterly world fuel inflation at time t, and EtUSD denotes the annualized quarterly nominal depreciation of domestic currency with respect to the US$ at time t. Qi,t represents quarterly time dummies and t reflects the linear time trend. The pass-throughs from world food inflation, world fuel inflation, and exchange rates to domestic food inflation is computed as the sum of the coefficients on world food inflation, world fuel inflation, and nominal exchange rate depreciations, respectively.21

The average pass-throughs to domestic food inflation from world food inflation, fuel inflation, and exchange rate depreciation were 32, 2, and 17 percent, respectively (Table IV.1).22 Country-specific pass-throughs suggest that over one third of the countries have statistically significant pass-throughs from world food inflation and exchange rate depreciation, while less than ten percent have statistically significant pass-through from world fuel inflation (Annex 2 Table A.8).

Table IV.1:

Pass-Through Coefficients by Type - Whole Sample

article image
Source: Author calculations

The distribution of the pass-through coefficients for world food prices is broadly centered between 0 and 40 percent, with many countries having a higher pass-through than this (Figure IV.2 Panel A). All but one of the pass-through coefficients for world fuel prices below 20 percent (Figure IV.2 Panel B). The pass-through coefficients for exchange rates are similar to world food prices and broadly centered between 0 and 40 percent with many countries having a higher pass-through.

Figure IV.2:
Figure IV.2:

Distribution of Pass-Through Coefficients

Citation: IMF Working Papers 2016, 247; 10.5089/9781475563115.001.A001

Source: Author calculations

These estimated pass-through coefficients, which are similar in magnitude to the results for AEs and EMs,23 suggest that there is incomplete pass-through in SSA from world food and fuel prices, and exchange rates to domestic food prices. We further our analysis by exploring potential structural and domestic drivers of food inflation in the next section.

B. Structural Drivers of Food Inflation SSA

The dataset for 20 SSA countries with detailed information on food weights for nine food subcomponents suggests that higher food inflation is broadly driven by fresh (and non-tradable) food items across SSA, including in both MICs and LICs (Table IV.2). Furthermore, the difference between fresh food inflation and non-fresh food inflation is greatest for LICs at 2.7 percentage points. This also suggests that the wedge between food and NF/NF inflation is being driven to a great extent by fresh food items. Moreover, country-specific data shows that around three quarters of the countries in our second dataset have fresh food inflation greater than non-fresh food inflation (Annex 2 Table A.9).

Table IV.2

Average Inflation for Disaggregated CPI

article image
Source: CPI data from country statistical sources, 20 countries.

Fresh food inflation is also consistently less volatile than non-fresh food inflation (Table IV.3). Country-specific data reveals that for three quarters of the countries in our sample, fresh food inflation is less volatile than non-fresh food inflation (Annex 2 Table A.10).

Table IV.3:

Volatility of Disaggregated Inflation

article image
Source: CPI data from country statistical sources, 20 countries.

Next, we use the pass-through regression estimated in equation (2), except with fewer lags24, on the nine food categories (as well as on the fresh and non-fresh foods composite) for the 20 SSA countries in our second dataset. Two key differences in this regression is that we use country-specific food weights and that domestic food inflation is based solely on the combined inflation of the nine food categories. These nine food categories are broadly representative of the weight of food in the consumption basket and represent about 87% of the food basket on average (Annex 2 Table A.3).

The estimated pass-through from world food and fuel inflation, and nominal exchange rate depreciation to domestic food inflation is 19, 3, and 14 percent, respectively. While these estimates are slightly lower than those from the larger sample the proportion of significant results are higher (Table IV.4). Although of the same magnitude, only 15% of the pass-through coefficients for fresh food were statistically significant (3 countries) while over a third of countries had a statistically significant pass-through to non-fresh foods. The pass-through from exchange rates to fresh food was 10 percent while that to non-fresh food was 24 percent. Both fresh and non-fresh food categories had close to half of the countries possessing a statistically significant pass-through from exchange rates. As in our larger sample, the pass-through from world fuel prices is broadly negligible.25

Table IV.4:

Disaggregated Pass-Through Coefficients by Type

article image
Notes: Numbers in parentheses indicate the proportion of the sample that was significant. For example, in relation to meat, 10% of the world food price pass-through coefficients were significant. Ethiopia was excluded in calculating the average of exchange rate coefficients for total food basket, total fresh, total non-fresh, meat, vegetables, milk, and bakery given it was an outlier. Ghana was excluded in calculating the average of the world food price coefficient for sugar given it was an outlier.Source: Author calculations

The findings in this section suggest that external drivers alone cannot explain all the variation in domestic food inflation in SSA and that structural and domestic drivers also play an important role. Indeed, we find that food inflation has been driven by food that is largely non-tradable (i.e., fresh food). We also find that there is more pass-through to non-fresh food than to fresh food. This suggests incomplete/limited tradability for perishable food items in SSA. This could reflect a number of factors such as shortage of storage capacity for perishable food, and/or inadequate transport infrastructure.

V. Conclusions

This paper describes food inflation trends in SSA from 2000 to 2016 and explores its causes, in particular why it was high and persistent, as well as policy implications.

We find that over the period from 2000 to 2016, food inflation has been on average 1.8 percentage points per year higher than NF/NF inflation; food inflation has been more volatile than NF/NF inflation; and food inflation has been at least as persistent as NF/NF inflation. Excluding the 2007-08 global food price surge (i.e., since 2009) food inflation has, on average, become less persistent. In addition, the differences across the two time periods also masks considerable heterogeneity between different economies in SSA. For example, while on average in SSA food inflation is significantly less persistent than NF/NF inflation from 2009 to 2016, for many countries food inflation has become more persistent.

Our findings also suggest that in countries where central banks have adopted forward-looking monetary policy frameworks, food inflation tends to be less persistent than NF/NF inflation. We also find that improved macroeconomic management is associated with food inflation being less persistent than NF/NF inflation. Moreover, we find that increasing integration with China as captured by the share of imports from China, reduces the persistence of food inflation relative to NF/NF inflation.

Granger causality tests on food and NF/NF inflation do not yield a definitive conclusion on whether food inflation causes NF/NF inflation across SSA. Nevertheless, in around a fifth of the sample of 41 countries, there is a one-way causality from food inflation to NF/NF inflation.

Taken together, these findings suggest that, given the high level of food inflation and its prominence in the consumption basket, using measures of core inflation that exclude food prices to assess the monetary stance would not be appropriate. This is especially the case in countries where persistence of food prices is as high as that of NF/NF prices (currently over a third of SSA countries).

We also documented that over a third of SSA countries have experienced statistically significant pass-through to domestic food prices from world food prices and exchange rates with an average pass-through of 32 and 17 percent, respectively. We find that higher food inflation is largely driven by fresh (and non-tradable) food items, and that more countries have statistically significant pass-through from world food prices and exchanges rates to non-fresh food prices than to fresh food prices.

The analysis in the paper has two important policies implications. First, central banks and fiscal authorities in SSA and other LICs should take food prices into consideration when assessing the appropriate stance of monetary and fiscal policy Second, improving the tradability of food—for example, by reducing tariffs and improving transport and storage infrastructure—and structural measures to increase productivity in the agricultural sector could decrease food price pressures and therefore improve welfare.

The relationship between improved central bank credibility and the persistence of food inflation would be an interesting area to explore further, particularly as central banks in SSA and other LICs have modernized rapidly over the last few decades. Moreover, exploring the demand and supply determinants of food inflation in LICs would also be an important avenue for future research, and help with more precise policy prescriptions for SSA and other LICs.

References

  • Ajakaiye, O. and S. A. O’Connell, 2011, “Central Banking Sub-Saharan Africa: Introduction and Overview,Journal of African Economies, Vol. 20, pp. ii3ii15.

    • Search Google Scholar
    • Export Citation
  • Andrews, D. W. K. and H. Chen, 1994, “Approximately Median-Unbiased Estimation of Autoregressive Models, Journal of Business and Economic Statistics, Vol. 12 (2).

    • Search Google Scholar
    • Export Citation
  • Bernanke, B. S., 2007, “Inflation Expectations and Inflation Forecasting”, Speech at the Monetary Economics Workshop of the National Bureau of Economic Research Summer Institute, Cambridge, Massachusetts.

    • Search Google Scholar
    • Export Citation
  • Biryabarema, E., 2015, “Uganda central bank raises rates to stem inflation pressures,Reuters Africa, [Online]. June 16. [Accessed August 12 2016]. Available at: http://af.reuters.com/article/investingNews/idAFKBN0OW11220150616

    • Search Google Scholar
    • Export Citation
  • Central Bank of Kenya, 2009, “Second Bi-Annual Report of the Monetary Policy Committee,April 2009.

  • Central Bank of Kenya, 2015, “Fifteenth Bi-Annual Report of the Monetary Policy Committee,October 2015.

  • Da Silva Filho, T. N. T. and F. M. R. Figueiredo, 2015, “A Volatility and Persistence-Based Core Inflation,IMF Working Paper, WP/15/8.

    • Search Google Scholar
    • Export Citation
  • Dias, D. and C. Marques, 2010, “Using Mean Reversion as a Measure of Persistence,Economic Modelling, Vol. 27, pp. 226273.

  • Durevall, D., Loening, J. L., and Y.A. Birru, 2013, “Inflation dynamics and food prices in Ethiopia”, Journal of Development Economics, Vol. 104, pp. 89106.

    • Search Google Scholar
    • Export Citation
  • Furceri, D., Loungani, P., Simon, J., and S. Wachter, 2015, “Global Food Prices and Domestic Inflation: Some Cross-Country Evidence,IMF Working Paper, WP/15/133.

    • Search Google Scholar
    • Export Citation
  • Gelbard, E., Deléchat, C., Jacoby, U., Pani, M., Hussain, M., Ramirez, G., Rui, X., Fuli, E., and D. Mulaj, 2015, “Building Resilience in Sub-Saharan Africa’s Fragile States,African Departmental Paper Series, IMF.

    • Search Google Scholar
    • Export Citation
  • Ghosh, A., Gulde, A-M., Ostry, J., and H. Wolf, 1996, “Does the Exchange Rate Regime Matter for Inflation and Growth,Economic Issue Series, IMF.

    • Search Google Scholar
    • Export Citation
  • Gurría, A., 2011, “The food crisis: beyond agriculture,Remarks at the Forum of the Americas, Montreal, Canada.

  • IMF, 2011, World Economic Outlook: Slowing Growth, Rising Risks, September, pp. 101133.

  • IMF, 2014, Regional Economic Outlook: Sub-Saharan Africa Fostering Durable and Inclusive Growth, April, pp. 4359.

  • IMF, 2015, “Evolving Monetary Policy Frameworks in Low-Income and Other Developing Countries,Staff Report.

  • IMF, 2016, Box 3.3: How much Do Global Prices Matter for Food Inflation?” World Economic Outlook, October.

  • Jacome, L., M. Matamorors-Indorf, M. Sharma, and S. Townsend, 2012, “Central Bank Credit to the Government: What Can We Learn from International Practices?IMF Working Paper, WP/12/16.

    • Search Google Scholar
    • Export Citation
  • Kiley, M. T., 2008, “Estimating the common trend rate of inflation for consumer prices and consumer prices excluding food and energy prices,Finance and Economics Discussion Series, 2008-38, Board of Governors of the Federal Reserve System.

    • Search Google Scholar
    • Export Citation
  • King, M., 2010, Letter from the Governor to the Chancellor, May 17 2010.

  • Minot, N., 2014, “Food price volatility in sub-Saharan Africa. Has it really increased?Food Policy, Vol. 45, pp. 4556.

  • Mishkin, F. S., 2007, “Headline versus Core Inflation in the Conduct of Monetary Policy,Speech at the Business Cycles, International Transmission and Macroeconomic Policies Conference, Montreal, Canada.

    • Search Google Scholar
    • Export Citation
  • Mohanty, D., 2014, “Why is recent food inflation in India so persistent?Speech at the Annual Lalit Doshi Memorial Lecture, Mumbai, India.

    • Search Google Scholar
    • Export Citation
  • Motley, B., 1997, “Should Monetary Policy Focus on ‘Core’ Inflation?FRBSF Economic Letter, Number 199711.

  • OECD, 2012, OECD Economic Outlook, Vol. 2012 (2), pp. 197.

  • Rangasamy, L., 2011, “Food Inflation In South Africa: Some Implications For Economic Policy,South African Journal of Economics, Vol. 79 (2), pp. 184-201.

    • Search Google Scholar
    • Export Citation
  • Wakeman-Lynn, J. and P. Drummond, 2008, “IMF Survey: Coping With Food Price Increases in Sub-Saharan Africa,IMF News Article. [Online]. March 20. [Accessed August 12 2016]. Available at: http://www.imf.org/external/pubs/ft/survey/so/2008/CAR032008A.htm

    • Search Google Scholar
    • Export Citation
  • Walsh, J., 2011, “Reconsidering the Role of Food Prices in Inflation, IMF Working Paper, WP/11/17.

Annex I. Nonparametric Estimation of Persistence

We utilize the nonparametric approach for measuring persistence developed by Dias and Marques (2010). Their measure of persistence relies on the idea that there is a relationship between persistence and mean reversion. Broadly speaking, following a shock, persistent series converge to their long-run mean more slowly than less persistent series. Persistent series therefore exhibit a lower level of mean reversion than non-persistent series.

The authors then define persistence, denoted γ, as the unconditional probability of a stationary stochastic process not crossing its mean in period t. Simply put, γ measures how infrequently a particular stationary process crosses its mean. Formally, γ is defined as the following:

γ=P{[(ytμ)>0(yt1μ)][(ytμ)<0(yt1μ)<0]}(1)

Where yt is the series y in period t and μ is its mean.

Given the data generating process (DGP) is unknown, a nonparametric and unbiased estimator for γ is the following26:

γ^=1nT(2)

Where n is the number of times yt crosses the mean during a time interval with T+1 observations.

Unlike using SARC estimates of persistence, γ^ is, by construction, between zero and one. If γ^ is 0.5 then this indicates zero persistence. Whilst above and below 0.5 indicate positive and negative persistence, respectively. Furthermore, SARC estimates require the DGP to follow a pure autoregressive process while, after assuming stationarity, γ is independent of underlying DGP. Therefore, this serves as a useful robustness check to the conclusions reached in Section III.C (i.e., that since 2009, the persistence of food inflation has fallen relative to the persistence of NF/NF inflation).

In Section III.C, we showed that the distribution of SARC estimates for food inflation had shifted leftwards while that of NF/NF had broadly flattened. Indeed, plotting the distribution for the nonparametric estimates of the persistence of food inflation, we find a similar trend for the persistence of food inflation as the distribution has skewed leftwards (Figure A.1 Panel A). Moreover, the figure shows that most of the mass of the distribution is around or to the left of 0.5 indicating that food inflation has little or no persistence in SSA. There does not seem to be an obvious change in the distribution for the NF/NF estimates. Taken together, this suggests that the persistence of food inflation relative to NF/NF inflation has fallen since 2009.

Figure A.1:
Figure A.1:

Distribution of Inflation Persistence (Nonparametric Estimates) across SSA

Citation: IMF Working Papers 2016, 247; 10.5089/9781475563115.001.A001

Notes: Inflation used for these calculations is QoQ, annualized.Source: Author calculations

Annex II. - Tables

Table A.1:

Source and Definitions of Variables

article image
Table A.2:

Definitions of Dummy Variables

article image
Notes: Low income country (LIC) defined as a country with PPP GDP per capita below 5,000 US$
Table A.3:

Country Food Weights Disaggregated (rescaled)

article image
Notes: The nine categories of food weights have been rescaled to ensure each sum to 100. These categories do not represent 100% of the food basket in each country but the final column shows they are highly representative nonetheless.Source: Country statistical sources
Table A.4:

Average Inflation Rates (PPP GDP per capita weighted) in SSA and Sub-groups

article image
Source: Price data from country statistical sources. PPP GDP per capita data from the WEO database.
Table A.5:

Average Inflation Rates by Country

article image
Notes: F-NF/NF refers to food inflation minus non-food non-fuel inflationSource: Country statistical sources