Implications of higher global food prices for poverty in low-income countries

Authors


  • Note: Data and programs for replication of this article's main results will be posted online alongside the article, at the publisher's website.

  • Data Appendix Available Online

  • A data appendix to replicate main results is available in the online version of this article. Please note: Wiley-Blackwell, Inc. is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

*Corresponding author: Tel.: +202-473-3853; fax: +202-522-1159. E-mail address:wmartin1@worldbank.org (W. Martin).

Abstract

In many poor countries, the recent increases in prices of staple foods have raised the real incomes of those selling food, many of whom are relatively poor, while hurting net food consumers, many of whom are also relatively poor. The impacts on poverty will certainly be very diverse, but the average impact on poverty depends upon the balance between these two effects, and can only be determined by looking at real-world data. Results using household data for 10 observations on nine low-income countries show that the short-run impacts of higher staple food prices on poverty differ considerably by commodity and by country, but that poverty increases are much more frequent, and larger, than poverty reductions. The recent large increases in food prices appear likely to raise overall poverty in low-income countries substantially.

1. Introduction

Since 2005, the world has experienced a dramatic surge in the price of many staple food commodities. The price of maize increased by 80% between 2005 and 2007, and has since risen further. Many other commodity prices also rose sharply over this period: milk powder by 90%, wheat by 70%, and rice by about 25%. Prices rose substantially more in early to mid 2008. Although most food prices had fallen from their peak levels by early November, they remained above 2005 levels, with most forecasts suggesting an extended period of relatively high prices.

Despite widespread concern about the impacts of high food prices on poor people and on social stability, little hard information appears to be available on actual impacts on poor people. The overall impact on poverty rates in poor countries depends on whether the gains to poor net producers outweigh the adverse impacts on poor consumers. Whether higher food prices improve or worsen the situation of particular households depends importantly on the products involved; the patterns of household incomes and expenditures; and the policy responses of governments (World Bank, 2008).

Existing analyses tell us that the impacts of higher food prices on poverty are likely to be very diverse, depending upon which commodity prices change, the structure of the economy (Hertel and Winters, 2006; Ravallion and Lokshin, 2005) and the distribution of net buyers and net sellers of food among low-income households (Aksoy and Izik-Dikmelik, 2008). While most studies find that higher prices of staple foods raise poverty in the short run (Barrett and Dorosh, 1996; Minot and Goletti, 2000; and Ravallion, 1990), it is certainly possible for increases in at least some food prices to transfer income to some poorer households (Aksoy and Izik-Dikmelik, 2008). Only with careful examination of outcomes at the household level is it possible to tell whether changes in the prices of specific staple foods will help or hurt poor people.

A particular reason for concern about the impacts of high food prices on poor countries arises from the fact that the poorest people spend roughly three-quarters of their incomes on staple foods (Cranfield et al., 2007). On the other hand, the incomes of farm households—frequently one of the poorest groups in low-income countries—may be increased by higher commodity prices (Hertel et al., 2004). However, the benefits of higher food prices to poor farm households may be less than they first appear, since these benefits depends not on what they produce, but on their net sales of these goods.

In this study, we attempt to address the main implications of higher food prices on poverty following the methodologically simple, yet data-intensive, approach of calculating the short-run impacts on households' income and costs of living following the changes in food prices. We do so using household surveys containing at least a thousand households in each of nine low-income countries, for which the data on consumption and production of the main food commodities are available. Besides calculating the change in household real income, we also estimate the impact of food prices on poverty rates and poverty gaps.

We consider two experiments in this article. First, we estimate the importance of small changes in the prices of individual commodities on poverty rates in each of our sample countries. We do so by conducting a stylized simulation in which we increase individual prices by 10% to assess the effect of a small change in the price of each commodity. Second, we estimate the impact of the actual food price changes between 2005 and 2007 on poverty in our sample of countries. In the initial analysis, we assume that changes in international prices are fully transmitted into domestic markets, and consider only the direct impacts of commodity price changes. We also consider scenarios under which commodity price changes lead to changes in the wage rate for unskilled labor—the other key source of income for most low-income households identified in Ravallion (1990).

2. Methodology

In this study, we consider some of the welfare and poverty impacts of changes in the world prices of key staple food commodities. Basic food commodities that are traded globally and are important for small farmers and consumers in the developing world include wheat, rice, dairy products, maize, sugar, beef, and poultry. We analyze the impacts of changes in the prices of these commodities using household-level data for nine low-income countries.

To undertake the analysis, we needed information on households' production, purchases, own-consumption, and sales of these agricultural products in order to calculate the net sales of each commodity at the household level—a need that limited us to countries with high-quality household surveys containing this information. This need also precluded using many of the models and databases prepared for the Hertel and Winters (2006) volume, since many of these did not include detailed information on enough of the commodities central to our analysis.

To assess the household impacts of changes in commodity prices, we use a very simple methodology based on Singh et al. (1986) and Deaton (1989; 1997, p. 185). We represent the impact of price changes on an individual household, using an expenditure function to characterize household consumption and factor supply behavior and a profit function to represent household production activities through unincorporated enterprises such as family farms. This yields a simple expression for the welfare impacts of small price changes that involves multiplying the price changes resulting from trade reform by the shares of income and expenditure affected by these price changes. Where a household consumes its own output of, for instance, staple foods, the share of the good's production in total output value is offset by its share in consumption so the only effect on income is through the household's net sales of the good.

Our framework is partial equilibrium in that we consider only the direct impacts of changes in the prices of staple foods on households, except that, like Ravallion (1990), we also consider the potential impacts on poor households through induced changes in the wage rate for their net sales of unskilled labor. We ignore changes in the returns to skilled labor on the grounds that these returns make an extremely small contribution to the incomes of the poor. Our focus is on real price changes, and we ignore the potentially serious costs (Easterly and Fischer, 2001; Ravallion and Datt, 2002) imposed on the poor through market disequilibria created by inflation. This omission seemed reasonable given that the average rate of inflation in developing countries2 has been very moderate by historical standards until very recently. It averaged 5% in the period from 2000 to 2007 before rising to 7% by January 2007 and 8% in February 2008. However, to the extent that commodity price changes are contributing to inflation increases in developing countries, there may be some important additional costs—the Ravallion and Datt (2002, p. 389) estimate suggests that a rise in the inflation rate of 3 percentage points would increase the poverty headcount by 1.2 percentage points.

When we take into account impacts through changes in unskilled wage rates, we use national versions of the GTAP model to assess the short-run impacts of changes in commodity prices on wage rates—essentially measuring the Stolper–Samuelson relationship between the prices of goods and the prices of factors. Given our short-run focus, we specify the standard specific-factors model of production, with capital fixed in each sector. Under our standard labor market closure, we assume that unskilled labor is mobile between agricultural and nonagricultural employment so that the benefits of an increase in demand for unskilled labor are shared between agricultural and nonagricultural workers. This assumption of mobility between agricultural and nonagricultural labor is supported by studies such as Bertrand and Squire's (1980) classic analysis of rural–urban labor mobility in Thailand. To assess the robustness of our results to this assumption, we examine in Section 6 a second labor market closure where the mobility of labor between agricultural and nonagricultural labor markets is very imperfect.

Consistent with many other studies considering the impacts of price changes on poverty (e.g., Chen and Ravallion, 2004), we consider only the first-order impacts of food price changes on the poor. This seems reasonable because the behavioral responses underlying the second-order effects we consider would need to be particularly large to change the direction of the effects we consider.3 Demand elasticities for staple foods generally appear to be low (Tyers and Anderson, 1992), and the scope for reducing consumption of staple foods in total when the prices of all these goods rise together is even less.4 Most of the scope for making large quantity adjustments is on the supply side, where the scope for adjustment is less when many product prices rise together, and where major adjustments can take an extended period. The fact that commodity booms are typically short-lived (Deaton and Laroque, 1992) further reduces the likelihood of large second-order impacts overwhelming the first-order impacts on which we focus—unless producers believe that higher prices will be sustained for a longer period, they are unlikely to make the investments needed for substantial supply responses.

We also omit from consideration in all but our analysis from 2005 to the first quarter of 2008 any policy measures that might insulate poor net buyers of food from increases in the prices of staple foods. Clearly, such policies could reduce the impact of an increase in domestic prices of staple foods on poor consumers. However, the details of such programs are extremely important for their impact, with potentially substantial risks of incomplete coverage of the poor. Detailed studies of the operation of such policies would be important in studies of individual countries. The results presented in this study indicate the effects on poverty in the absence of functioning schemes of this type.

In all of the analyses reported in this article, we assume full transmission from world market prices to the prices facing producers and consumers, a finding supported by Mundlak and Larson (1992). The resulting poverty impacts will certainly be sensitive to this assumption. However, the nature of the impact will depend on the specific situation. In some countries, where many poor consumers are in areas strongly integrated with world markets, and net sellers are to some degree insulated, the impacts on poverty may be more adverse than under our assumption. By contrast, in other countries, consumers may be more insulated than producers, perhaps by consuming goods that are more strongly differentiated from imported goods, and the impacts on poverty may be more favorable than under our assumption. The extent and nature of price transmission should be taken into account when considering the actual impacts in particular countries.

We calculate the real income level after the change in food prices by adding the change in households' real incomes to their initial income levels in order to find their real incomes after the price changes. By comparing the new achievable level of expenditure with the established poverty-line level of expenditure for each country, we are able to use individual survey records to identify and count the number of households in poverty and the gap between their income level and the poverty line after the change in policy. We then cumulate these shocks to see the impact on poverty, and compare these impacts with those prevailing when wages are also able to adjust. Finally, we use these data to calculate two complementary poverty measures (Ravallion and van der Walle, 1991)—the poverty headcount and the poverty gap. The first measures the percentage of people falling below the poverty line and the second the average percentage of the poverty line by which the incomes of the poor fall below the poverty line relative to the poverty line.

To reiterate the main points of our methodology, we apply commodity price and wage changes to individual households' observed net sales of staple foods to calculate the changes in revenues and the costs of expenditure. Even though we do not consider any changes in the quantities consumed or sold, we estimate changes in unskilled wage rates from food price changes since wages are an important source of income for many poor households.

3. The data

For this study, we needed household survey information on supply, demand, and net sales of food products—a constraint that narrowed the range of countries we could consider. We sought relatively poor countries from different regions, with different relationships between poverty and location, and between poverty and sources of income household; and with different net trade positions in agriculture. Because of the extremely rapid structural change in the Vietnamese economy, we examined data for two different periods.5

The sample sizes for each of the 10 country-periods for which were able to obtain suitable, high-quality survey data are listed in Table 1. These were Bolivia in 2005, Cambodia in 2002–3, Madagascar in 2002, Malawi in 2004, Nicaragua in 2001, Pakistan in 1998–9, Peru in 2003, Vietnam in 1998 and 2004, and Zambia in 1998. For our initial poverty rates and poverty gaps, we used the standard “dollar-a-day”6 expenditure-based measures of poverty from the 2007 World Bank World Development Indicators (WDI). We ranked the households in our surveys by income and placed the poverty line at the income level that reproduced the poverty rate in the (frequently different) year used in the WDI. This process retains the survey information on the sources and distribution of income, while updating the poverty line to the most recent year for which information is available without the need to adjust the household incomes in the surveys in any way.

Table 1. 
Sample sizes for the household surveys
 RuralUrbanTotal
Bolivia (2005) 1,750 2,335 4,085
Cambodia (2003)11,990 2,99414,984
Madagascar (2001) 2,036 3,039 5,075
Malawi (2004) 9,840 1,44011,280
Nicaragua (2001) 1,790 2,211 4,001
Pakistan (1998–1999)10,254 5,90916,163
Peru (2003) 1,437 3,578 5,016
Vietnam (1997–1998) 4,269 1,730 5,999
Vietnam (2004) 6,938 2,250 9,188
Zambia (1998) 8,373 8,17016,543
Total60,87831,45692,334

Because of the generally higher poverty rates in rural than in urban areas, the greater frequency of net seller households in rural areas, and the focus of this study on food prices, we categorized each household as either rural or urban based on the classification used in the national survey. For lack of information on urban versus rural costs of living, we have ignored these differences, which are potentially important for the level of poverty, but much less so for the changes in welfare levels on which we focus.

The initial poverty headcount and poverty gap numbers are presented in the first column of Table 2. The initial national poverty headcounts varied considerably between countries, from 75.8% in Zambia to 12.5% in Peru with the other six countries widely distributed in the intervening range. In almost all countries, the poverty headcount was higher in rural areas, with 41.7% of rural people poor on average, as against 24.3% of urban people.

Table 2. 
Initial $1 per day poverty rates and impacts of a 10% price increase on poverty, % and % points change
  InitialBeefDairyMaizePoultryRiceSugarWheatAll
NWWWNWWWNWWWNWWWNWWW NWWWNWWWNWWW
  1. Note: NW = no wage impacts included; WW = results with wage impacts; average excludes Vietnam, 1998.

BoliviaRural40.90.20.20.0−0.1−0.1−0.10.10.10.20.20.20.10.30.30.50.4
Urban 9.90.20.20.00.00.00.00.10.10.00.00.10.00.20.20.60.6
Total23.20.20.20.00.00.00.00.10.10.10.10.10.10.20.20.50.5
CambodiaRural38.7−0.3−0.30.00.00.00.00.0−0.10.60.50.10.10.00.00.30.1
Urban15.70.00.00.00.00.00.00.00.00.50.40.00.00.00.00.50.2
Total34.1−0.2−0.30.00.00.00.00.0−0.10.50.50.00.00.00.00.30.1
MadagascarRural76.80.20.10.00.00.00.00.00.01.71.40.20.10.00.01.91.4
Urban50.40.5−0.20.20.20.00.00.10.01.20.70.1−0.10.30.31.80.2
Total61.00.4−0.10.10.10.00.00.00.01.41.00.10.00.20.21.80.7
MalawiRural23.30.00.00.00.00.50.50.00.00.00.00.10.10.00.00.60.5
Urban 3.70.00.00.00.00.30.20.00.00.00.00.00.00.00.00.40.3
Total20.80.00.00.00.00.50.40.00.00.00.00.10.10.00.00.50.5
NicaraguaRural61.10.10.00.20.2−0.2−0.20.20.10.40.40.20.20.40.41.51.4
Urban32.20.20.20.60.60.10.10.50.50.50.50.20.20.20.32.72.5
Total45.10.10.10.40.40.00.00.40.30.40.50.20.20.30.32.12.0
PakistanRural20.80.00.0−0.1−0.30.00.00.00.0−0.1−0.10.00.0−0.1−0.1−0.1−0.3
Urban10.40.00.00.20.20.00.00.00.00.00.00.10.10.40.40.80.8
Total17.00.00.00.0−0.10.00.00.00.0−0.1−0.10.00.00.10.10.30.1
PeruRural20.7−0.3−0.40.00.0−0.1−0.1−0.1−0.10.0−0.10.00.00.10.1−0.2−0.3
Urban9.2−0.1−0.10.00.00.00.00.00.00.00.00.00.00.10.10.00.0
Total12.5−0.1−0.10.00.00.00.00.00.00.00.00.00.00.10.1−0.1−0.1
Vietnam, 2004Rural20.9−0.1−0.10.00.0−0.1−0.1−0.2−0.3−1.0−1.10.00.00.00.0−1.4−1.5
Urban7.60.00.00.00.00.00.0−0.1−0.10.20.00.00.00.00.00.20.0
Total17.7−0.1−0.10.00.0−0.1−0.1−0.2−0.3−0.7−0.80.00.00.00.0−1.0−1.1
Vietnam, 1998Rural21.6−0.1−0.10.00.00.00.0−0.1−0.2−0.8−1.00.00.00.00.0−0.9−1.1
Urban8.10.00.00.00.00.00.00.1−0.20.20.20.10.10.10.10.30.2
Total17.7−0.1−0.10.00.00.00.0−0.1−0.2−0.5−0.70.00.00.00.1−0.6−0.7
ZambiaRural72.20.00.00.00.00.80.80.20.20.10.10.00.00.00.01.11.1
Urban79.50.10.10.10.10.20.20.10.00.00.00.00.00.00.00.60.5
Total75.80.10.00.00.00.50.50.20.10.00.00.00.00.00.00.80.8
AverageRural41.70.0−0.10.00.00.10.10.00.00.20.10.10.10.10.10.40.3
Urban24.30.10.00.10.10.10.10.10.10.30.20.10.00.10.10.80.6
Total34.10.00.00.10.00.10.10.10.00.20.10.10.10.10.10.60.4

The poverty gap numbers presented in the first column of Table 3 are defined as the gap between the average income of the poor and the poverty line, as a share of the poverty line, times the poverty headcount (Ravallion, 1992). On average, this gap was 18.3% for rural people, 10.6% for urban people, and 14.6% overall. In Madagascar, it was 30.4% overall and in Zambia 44.7%, and in the country with the lowest poverty gap, Pakistan, it was 4.0%. In all cases but Zambia, the poverty gap was larger in rural than in urban areas.

Table 3. 
Initial $1 per day poverty gap and impacts of a 10% increase in prices, % and % points change
  InitialBeefDairyMaizePoultryRiceSugarWheatAll
NWWWNWWWNWWWNWWWNWWWNWWWNWWWNWWW
  1. Note: NW = no wage impacts included; WW = results with wage impacts; average excludes Vietnam, 1998.

BoliviaRural15.20.00.00.00.0−0.1−0.10.00.00.00.00.10.00.20.20.30.2
Urban2.80.00.00.00.00.00.00.00.00.00.00.00.00.10.10.20.2
Total8.10.00.00.00.00.00.00.00.00.00.00.00.00.10.10.20.2
CambodiaRural11.4−0.1−0.10.00.00.00.00.00.00.60.60.00.00.00.00.60.5
Urban4.30.00.00.00.00.00.00.00.00.20.20.00.00.00.00.20.1
Total10.0−0.1−0.10.00.00.00.00.00.00.60.50.00.00.00.00.50.4
MadagascarRural43.0−0.2−0.40.00.00.00.0−0.1−0.10.80.70.10.10.10.10.70.4
Urban22.00.1−0.30.00.00.00.00.0−0.10.60.30.10.00.10.11.00.1
Total30.40.0−0.40.00.00.00.00.0−0.10.70.50.10.00.10.10.90.2
MalawiRural7.70.00.00.00.00.20.20.00.00.00.00.00.00.00.00.20.2
Urban1.10.00.00.00.00.10.10.00.00.00.00.00.00.00.00.10.1
Total6.80.00.00.00.00.20.20.00.00.00.00.00.00.00.00.20.2
NicaraguaRural25.10.00.00.10.10.00.00.10.00.30.30.20.20.10.10.80.8
Urban10.60.10.00.20.20.10.10.10.10.20.20.10.10.10.10.90.8
Total17.10.00.00.20.20.10.10.10.10.20.20.20.20.10.10.90.8
PakistanRural4.80.00.00.0−0.10.00.00.00.00.00.00.00.00.10.10.10.0
Urban2.40.00.00.10.10.00.00.00.00.00.00.00.00.10.10.20.2
Total4.00.00.00.00.00.00.00.00.00.00.00.00.00.10.10.10.1
PeruRural8.3−0.2−0.20.00.00.00.00.00.00.00.00.00.00.00.0−0.2−0.2
Urban4.00.00.00.00.00.00.00.00.00.00.00.00.00.00.00.0−0.1
Total5.2−0.1−0.10.00.00.00.00.00.00.00.00.00.00.00.0−0.1−0.1
Vietnam, 2004Rural6.20.00.00.00.00.00.0−0.1−0.1−0.3−0.30.00.00.00.0−0.4−0.4
Urban2.10.00.00.00.00.00.00.00.00.00.00.00.00.00.00.00.0
Total5.20.00.00.00.00.00.00.0−0.1−0.2−0.20.00.00.00.0−0.3−0.3
Vietnam, 1998Rural6.40.00.00.00.00.00.00.0−0.1−0.3−0.30.00.00.00.0−0.4−0.4
Urban2.00.00.00.00.00.00.00.00.00.10.10.00.00.00.00.10.1
Total5.20.00.00.00.00.00.00.00.0−0.2−0.20.00.00.00.0−0.2−0.3
ZambiaRural43.50.00.00.00.00.50.50.20.20.00.00.00.00.00.00.80.8
Urban45.80.10.10.10.10.10.00.10.00.00.00.10.10.00.00.50.3
Total44.70.10.00.00.00.30.30.10.10.00.00.10.00.00.00.70.6
AverageRural18.3−0.1−0.10.00.00.10.10.00.00.20.10.10.10.00.00.30.3
Urban10.60.00.00.00.00.00.00.00.00.10.10.00.00.00.00.30.2
Total14.60.0−0.10.00.00.10.10.00.00.10.10.00.00.00.00.30.2

4. Experiments

4.1. Impacts of 10% price increases

In the first experiment, we consider two analyses in a small, open-economy setting. The first examines the direct impact of hypothetical 10% increases in the prices of individual products on both the poverty headcount and the poverty gap. The second considers the impact of the same changes, taking into account their effects on wage rates for unskilled workers.

The specification of 10% price increases in experiments one and two do not imply that food prices are expected to rise by 10% (or any other amount). Rather, we examine the impact of a uniform change in the prices of these goods to get an indication of the direction and relative order of magnitude of effects for particular commodities. Correctly interpreted, the results of these experiments provide vitally important information because they summarize all the relevant information in our household survey database on the sources of income and the patterns of expenditure for each household. If, for instance, a large number of poor people in a particular country are net sellers of wheat, this experiment will tell us how much a 10% increase in the price of wheat will raise their incomes, and whether the gains to these households are large enough to offset losses to poor net-buying households so that national (and rural/urban) poverty rates fall.

In these initial analyses, we are able to sidestep the vexed question of substitutability relationships between international and domestic commodity prices by examining the direct impact of a 10% change in the domestic price. The results without wage impacts from these experiments may be more relevant in the short term since, as shown in Ravallion's (1990) study of Bangladesh rice pricing policy, the short-run effect on wages of food price changes may be considerably smaller than the long-run impact.

4.2. Impacts of actual 2005–2007 price increases

In the second scenario, we consider the impact of the recently observed food price increases on poverty. The scenario again includes the same set of agricultural commodities as before with the exception of beef and sugar, whose prices did not change much between 2005 and 2007. For the remainder of the commodities, we obtained the average annual prices from the FAO website and calculated rough percentage changes over this two-year period. The imprecise and rounded estimates are the result of the existence of various varieties of the same commodity and different prices at various points of sale. However, because the prices of similar commodities changed at roughly the same rate, our scenario broadly captures the overall development of the relevant food commodities in the reference period. The price changes considered in this scenario are shown in Table 4.

Table 4. 
World price scenario for 2005–2007
 Price change %
  1. Source: FAO.

Beef 0
Dairy90
Maize80
Poultry15
Rice25
Sugar 0
Wheat70

5. Results

5.1. Impact of a 10% increase in food prices

In this analysis, we first considered the direct impacts of changes in the prices of the commodities, and then the combined effect of changes in the prices of commodities and induced changes in wages.

5.1.1. Commodity price impacts

The results presented in Table 2 show the impacts on poverty rates of raising the prices of particular goods by 10% from their initial levels. These results are shown first for the individual commodities and then for all commodities considered. Corresponding results are presented in Table 3 for the poverty gap. Fig. 1 shows results for four country-periods that highlight some key findings.

Figure 1.

Key impacts of commodity price increases on poverty.

The results presented in Table 2 show that the impacts of changes in product prices on poverty differ greatly between products and countries. In Bolivia, wheat and beef have the largest impacts on poverty rates. For wheat, the increase in poverty is greater in rural than in urban areas. For all products except maize, poverty would increase. For maize, the reduction in poverty is concentrated in rural areas, and is very small relative to the increases in poverty associated with rising prices for other commodities.

In Cambodia, the commodity for which price changes have the greatest impact on poverty is the staple food, rice, for which an increase of 10% raises national poverty rates by 0.5 percentage points—and increases poverty in both rural and urban areas. Beef price increases lower rural and national poverty rates because significant numbers of poor people produce beef but consume relatively little of it. Fig. 1 highlights the importance of these two products for Cambodia.

In Madagascar, the poverty rate is much more sensitive to the price of these staple foods. Increases in the price of rice raise poverty substantially, a result consistent with the importance of rice in the purchases of the poor in urban and rural areas. Increases in the price of beef raise poverty because many urban people near the poverty line are net buyers of beef. The effect of a 10% increase in the prices of the goods considered is to raise poverty by 1.8 percentage points.

For Malawi and Zambia, the most influential individual commodity is maize, for which both urban and rural households are net buyers so that an increase in its price raises poverty both in rural and in urban areas. Increasing the price of maize by 10% would raise poverty in Malawi and Zambia by 0.5% and 0.8% in rural areas, 0.3% and 0.2% in urban areas, and 0.5% nationwide in both cases. When the prices of all staples increase by 10% in Malawi and Zambia, poverty rises by 0.6% and 1.1% in the rural areas, 0.4% and 0.6% in the rural areas for the total increase of 0.5% and 0.8%, respectively.

Nicaragua is like Madagascar in having much higher poverty impacts from increases in the prices of staple foods than other countries, but in Nicaragua the adverse effects cumulate from almost all commodities, rather than coming from just one or two staples. If the prices of all goods considered rose by 10%, the poverty rate would rise by 2.1%. In contrast with Madagascar, however, a large number of products contribute to this overall increase, with rice, dairy products, poultry, and wheat having larger impacts than other commodities. The adverse impacts on poverty tend to be smaller in rural areas, and increases in the price of maize reduce rural poverty, while raising it nationally.

In Pakistan, there are sharp contrasts between the impacts on rural and urban poverty. Increases in the prices of rice, dairy products, and wheat lower poverty in rural areas, while raising it in urban areas, and have very small impacts on national poverty rates. Increases in the prices of all goods lower rural poverty slightly, but the increase in urban poverty outweighs this and generates an increase of 0.3 percentage points in the national poverty rate.

In Peru, the effects of most commodity price changes on poverty are smaller than in most other countries. Beef, for which both urban and rural households are net sellers, has the largest absolute impact, with an increase in beef prices lowering the poverty rate by 0.3 percentage points in rural areas and 0.1 in urban. Since both urban and rural households are net sellers of wheat, increases in its price raise poverty slightly so the impact of increases in all of the commodities considered is an extremely small reduction in poverty of 0.1 percentage points.

In Vietnam, both in 1998 and in 2004, the largest single commodity impact was through the price of rice, for which a 10% price increase cut rural poverty by 0.8 percentage points in 1998 and 1.0% in 2004. Urban poverty rises by 0.2 percentage points in each case, but the overall effect was a decline of 0.5 percentage points in the national poverty rate using 1998 data and 0.7% using 2004 data. These results support the finding of Edmonds and Pavcnik (2005) that increases in rice prices would reduce poverty, and are more optimistic than the finding by Minot and Goletti (2000, p. 13 and p. 64) that an increase in rice prices would lower rural poverty rates, but raise national poverty rates slightly in Vietnam. Key influences on these findings appear to be the fact that Vietnam is an exporter of rice; the relatively egalitarian distribution of land in Vietnam; and the absence of a large class of poor landless laborers (Ravallion and van der Walle, 2008). In both years, increases in the price of poultry reduce poverty, and the size of this favorable impact doubled between 1998 and 2004, from 0.1% to 0.2%.

The overall average poverty impacts presented at the bottom of Table 2 show that, on average, increases in the prices of all commodities considered would increase poverty. The increases in urban poverty are generally larger than those in rural areas—unsurprisingly, since it is less likely that urban households will be net sellers of these goods. For all commodities except beef, price increases raise both rural and urban poverty.

The results for changes in the poverty gap in Table 3 complement the estimates for the poverty rate by taking into account the depth of poverty. As for the poverty headcount, there are substantial differences in the effects by commodity and between rural and urban areas. However, the poverty gap numbers tend to be less volatile than the poverty rate impacts and, particularly, less affected by the initial poverty rate. In terms of the overall average, the picture is very similar to that discussed above for the poverty rate. Increases in the prices of all of these staple foods increase the poverty gap more in urban than in rural areas, but increase the poverty gap in both, and raise the national poverty rate. The one exception at the commodity level is beef, for which price rises reduce the poverty gap both in rural areas and nationally, albeit by a small amount.

In Bolivia, the impacts on the poverty gap arise from a number of commodities, including wheat, rice, and sugar although the impacts through the price of wheat are by far the largest. By contrast, in Cambodia and in Madagascar, the overall effects are dominated by rice, as would be suggested by Fig. 1. Ten percent increases in the price of rice in these two countries raise the poverty gap by 0.6 and 0.7 percentage points, contributing more than the total increase in the poverty gap. In Nicaragua, as in Bolivia, a wide range of commodities contribute to the increase in the poverty gap, with rice and dairy products having the largest impact. An increase in the price of wheat has the most marked adverse impact on the poverty gap in Pakistan, with a 10% increase in the price of wheat increasing the poverty gap by 0.1.

Peru is quite different in that the poverty gap falls when the prices of these commodities rise, with most of the reduction of 0.1 percentage points coming from beef. In Vietnam the poverty gap falls when prices of all commodities considered rise, an effect dominated—as was the case with the poverty rate—by rice. In Malawi, increases in virtually all commodity prices raise the poverty gap, with the most influential case being maize, which contributes 0.2 to the overall national increase in the poverty gap of 2 percentage points.

5.1.2. Commodity price and wage impacts

In Tables 2 and 3 we also consider the impact of changes in commodity prices accompanied by changes in unskilled wage rates. The results in Table 2 show that labor market effects can be important in determining the impact of price changes on poverty rates. When the prices of all goods rise by 10%, the resulting average increase in rural poverty falls from 0.4% when wage impacts are ignored, to 0.3% with wage impacts included. For some countries, such as Nicaragua, the incorporation of wage impacts has very little impact. For others, such as Pakistan, the effect on the poverty rate is sharply reduced when all commodities are considered. However, in no case is the sign of the overall effect changed.

At the individual commodity level, there is somewhat more diversity in the effect of including wage rate impacts. In a few cases wage rate impacts are sufficient to change the direction of the effect on national poverty rates. This is the case for beef in Madagascar, where the poverty impact changes from 0.4 to −0.1 percentage points. But these effects are very small and increases in the prices of all of these staple goods continue to raise national poverty rates.

The results in Table 3 show that the impact of including wage rate impacts on the poverty gap is generally small, with the average increase in the poverty gap remaining constant at 0.3 percentage points in rural areas, and changing from 0.3 to 0.2 percentage points at the national level. The only notable exception is Madagascar where the national poverty gap increases by 0.2 percentage points instead of 0.9 when wage impacts are considered.

5.2. Impact of the 2005–2007 global food price increases on poverty

In the second experiment, we consider the developments in global food prices in 2005–2007 and apply these changes to our sample of households in order to estimate the impact on poverty in the absence of policy responses. Based on the FAO data on price developments, we implement the scenario shown in Table 4. Following Mundlak and Larson (1992), we assume that price transmission between world and domestic markets is approximately complete under normal market circumstances and specify a unitary elasticity of price transmission.

We first consider a scenario with wage impacts ignored. The main poverty rate change results of this scenario are shown in Table 5. As the table suggests, the average impact of the past global food price increase results in a rise in poverty of 3.0 percentage points. The increase among the urban households is greater, at 3.6 percentage points, while rural poverty rises by 2.5 percentage points.

Table 5. 
Poverty rate impacts of 2005–2007 global food price increases, % and % points change
  InitialBeefDairyMaizePoultryRiceSugarWheatAll
NWWWNWWWNWWWNWWWNWWWNWWWNWWWNWWW
  1. Note: NW = no wage impacts included; WW = results with wage impacts; average excludes Vietnam, 1998.

BoliviaRural40.90.00.00.30.0−0.8−0.80.20.10.20.20.00.02.52.51.91.6
Urban 9.90.00.00.50.30.00.00.10.10.20.20.00.01.11.12.12.0
Total23.20.00.00.40.2−0.3−0.30.10.10.20.20.00.01.71.72.01.8
CambodiaRural38.70.00.0−0.10.00.00.00.0−0.11.41.20.00.00.10.11.51.3
Urban15.70.00.00.00.20.00.10.1−0.11.20.80.00.00.10.11.41.0
Total34.10.00.0−0.10.00.00.00.0−0.11.41.10.00.00.10.11.51.3
MadagascarRural76.80.00.00.30.30.20.2−0.1−0.12.62.30.00.00.50.63.53.1
Urban50.40.00.00.80.60.30.30.10.02.51.40.00.01.71.75.63.9
Total61.00.00.00.60.50.20.30.00.02.51.80.00.01.21.34.73.6
MalawiRural23.30.00.00.00.04.33.90.00.00.00.00.00.00.20.24.64.1
Urban 3.70.00.00.00.02.82.50.00.00.00.00.00.00.10.13.33.0
Total20.80.00.00.00.04.13.80.00.00.00.00.00.00.20.24.44.0
NicaraguaRural61.10.00.02.52.3−0.5−0.50.30.21.11.10.00.01.61.64.24.2
Urban32.20.00.04.24.12.62.60.60.51.71.70.00.02.12.110.710.5
Total45.10.00.03.43.31.21.20.50.41.41.40.00.01.81.97.87.7
PakistanRural20.80.00.00.2−0.9−0.2−0.20.00.0−0.2−0.30.00.01.92.12.71.9
Urban10.40.00.02.82.50.00.00.10.10.10.10.00.02.72.76.56.1
Total17.00.00.01.10.3−0.1−0.10.00.0−0.1−0.10.00.02.22.34.13.4
PeruRural20.70.00.0−0.4−0.5−0.6−0.6−0.1−0.1−0.1−0.10.00.00.10.1−1.0−1.0
Urban 9.20.00.00.0−0.1−0.1−0.10.00.00.00.00.00.00.10.10.10.0
Total12.50.00.0−0.1−0.2−0.2−0.20.00.00.00.00.00.00.10.1−0.2−0.3
Vietnam, 2004Rural20.90.00.00.10.0−0.8−0.8−0.3−0.4−1.9−2.10.00.00.00.0−2.8−3.1
Urban 7.60.00.00.00.00.00.0−0.1−0.10.30.20.00.00.00.00.30.1
Total17.70.00.00.10.0−0.6−0.6−0.3−0.3−1.4−1.60.00.00.00.0−2.0−2.3
Vietnam, 1998Rural21.60.00.00.10.0−0.5−0.5−0.2−0.3−1.4−1.70.00.00.30.3−1.8−2.2
Urban 8.10.00.00.1−0.2−0.1−0.10.1−0.21.00.80.00.00.30.31.30.9
Total17.70.00.00.1−0.1−0.4−0.4−0.1−0.3−0.7−1.00.00.00.30.3−0.9−1.3
ZambiaRural72.20.00.00.40.46.46.30.40.40.10.10.00.00.00.07.47.4
Urban79.50.00.00.50.41.81.50.20.10.10.10.00.00.00.12.52.3
Total75.80.00.00.40.44.14.00.30.20.10.10.00.00.00.05.04.9
AverageRural41.70.00.00.40.20.90.90.00.00.40.30.00.00.80.82.52.2
Urban24.30.00.01.00.90.80.80.10.10.70.50.00.00.90.93.63.2
Total34.10.00.00.70.50.90.90.10.00.50.30.00.00.80.93.02.7

The table also suggests that even though the average impact of recent food price developments is adverse, two countries—Vietnam and Peru—would likely have benefited from reductions in rural poverty. In the case of Vietnam 2004, the reductions in rural poverty are large enough to reduce overall poverty, even though urban poverty rates rise slightly. Looking at the poverty impacts of individual commodities, we can see that in the case of Peru rural households benefit from the rising price of almost all commodities with the exception of wheat. In the contrasting case of Vietnam, essentially all of the poverty reduction comes from the rise in the price of rice.

The country where poverty would have been most adversely impacted by the rise in food prices appears to be Nicaragua, especially its urban households. The overall poverty rate in Nicaragua would have risen by 7.8% under this scenario, while the urban poverty rate would have risen by 10.7 percentage points.

Table 6 presents the corresponding changes in poverty gaps following our simulation. Its results are broadly consistent with the poverty rate results in Table 5. The poverty gap rises in all countries but Peru and Vietnam. In Peru, the major source of the reduction in the poverty gap is the increase in prices of maize. By contrast, in Vietnam, the most important determinant of the result is wheat. In all other country-periods, the poverty gap rises.

Table 6. 
Poverty gap impacts of 2005–2007 global food price increase, % and % points change
  InitialBeefDairyMaizePoultryRiceSugarWheatAll
NWWWNWWWNWWWNWWWNWWWNWWWNWWWNWWW
  1. Note: NW = no wage impacts included; WW = results with wage impacts; average excludes Vietnam, 1998.

BoliviaRural15.20.00.00.20.1−0.3−0.30.00.00.10.10.00.01.31.31.41.2
Urban 2.80.00.00.10.10.00.00.00.00.10.10.00.00.50.50.80.8
Total 8.10.00.00.20.1−0.1−0.10.00.00.10.10.00.00.80.81.10.9
CambodiaRural11.40.00.00.00.00.00.00.0−0.11.81.70.00.00.00.01.81.8
Urban 4.30.00.00.00.10.00.00.0−0.10.60.40.00.00.00.00.60.5
Total10.00.00.00.00.00.00.00.0−0.11.61.50.00.00.00.01.61.5
MadagascarRural43.00.00.00.20.10.20.2−0.1−0.12.11.90.00.00.40.42.92.6
Urban22.00.00.00.40.20.10.10.0−0.11.70.90.00.00.80.83.12.1
Total30.40.00.00.30.20.10.20.0−0.11.91.30.00.00.60.73.02.3
MalawiRural 7.70.00.00.00.02.32.10.00.00.00.00.00.00.10.12.42.2
Urban 1.10.00.00.00.00.90.80.00.00.00.00.00.00.00.01.00.9
Total 6.80.00.00.00.02.12.00.00.00.00.00.00.00.10.12.22.0
NicaraguaRural25.10.00.01.71.60.30.30.10.00.70.70.00.00.80.93.83.7
Urban10.60.00.01.81.81.11.10.10.10.50.50.00.00.80.85.04.9
Total17.10.00.01.81.70.80.70.10.10.60.60.00.00.80.84.54.4
PakistanRural 4.80.00.00.40.10.00.00.00.00.00.00.00.00.90.91.51.3
Urban 2.40.00.00.70.60.00.00.00.00.00.00.00.00.70.71.71.6
Total 4.00.00.00.50.30.00.00.00.00.00.00.00.00.80.81.61.4
PeruRural 8.30.00.0−0.1−0.1−0.2−0.20.00.00.00.00.00.00.10.1−0.1−0.2
Urban 4.00.00.00.00.0−0.1−0.10.00.00.00.00.00.00.00.0−0.1−0.1
Total 5.20.00.00.00.0−0.1−0.10.00.00.00.00.00.00.10.1−0.1−0.1
Vietnam, 2004Rural 6.20.00.00.00.0−0.3−0.3−0.1−0.1−0.6−0.60.00.00.00.0−0.8−0.9
Urban 2.10.00.00.00.00.00.00.00.00.10.00.00.00.00.00.10.0
Total 5.20.00.00.00.0−0.2−0.2−0.1−0.1−0.4−0.50.00.00.00.0−0.6−0.7
Vietnam, 1998Rural 6.40.00.00.00.0−0.2−0.2−0.1−0.1−0.5−0.60.00.00.10.1−0.6−0.7
Urban 2.00.00.00.00.00.00.00.00.00.30.30.00.00.10.10.40.3
Total 5.20.00.00.00.0−0.1−0.10.0−0.1−0.3−0.30.00.00.10.1−0.3−0.4
ZambiaRural43.50.00.00.40.46.16.10.20.20.10.10.00.00.00.07.17.0
Urban45.80.00.00.50.51.20.90.20.10.10.10.00.00.00.12.01.6
Total44.70.00.00.50.43.73.50.20.20.10.10.00.00.00.14.64.4
AverageRural18.30.00.00.30.20.90.90.00.00.50.40.00.00.40.42.22.1
Urban10.60.00.00.40.40.40.30.00.00.30.20.00.00.30.31.61.4
Total14.60.00.00.40.30.70.70.00.00.40.30.00.00.40.42.01.8

When wage impacts are also considered in our calculations, the results become less adverse for poverty outcomes, as unskilled wages rise in response to the increase in commodity prices, but the signs of the effects are generally not reversed. By including wage impacts, the average increase in the poverty rate drops from 3.0 to 2.7 percentage points. Similarly, the urban poverty change drops from 3.6 to 3.2 and rural poverty from 2.5 to 2.2 percentage points. The impact on the poverty gap follows the same pattern of ameliorating the results without changing their conclusions.

Table 5 also allows us to consider the importance of each commodity price change on the simple average across countries of changes in poverty. The prices of maize and wheat appear to have played the most prominent role in the poverty increases, contributing 0.9 percentage points. The price increases for rice and dairy added 0.3 and 0.5 percentage points. The overall impact of the price of chicken is zero because it as often increases poverty as it decreases it. Turning to the poverty gap numbers in Table 6, we see that the commodities with the greatest leverage on the results are maize, wheat, and rice, which raise the poverty gap by 0.7%, 0.4%, and 0.3%, respectively.

6. Robustness checks

The results of our robustness checks are presented and discussed in detail in Appendix B of the online version and only briefly summarized here. The first robustness check was to investigate possible nonlinearities in the relationship between the size of the price change and the poverty outcome, to see whether the results obtained could be scaled up or down to gain a broad indication of the effects of other price changes. The results of this analysis suggest that the results can be scaled up or down to give a reasonably reliable indication of the effects of reasonably small price changes.

Our second robustness check examined different possible settings of the poverty rate. Given the wide range of possible different poverty lines we needed to be sure that our results were not strongly sensitive to the choice of poverty line. As expected, the level of the poverty line affected the magnitude of the percentage point changes in poverty rates for changes in staple food products. However, only in one case, Pakistan, was there a change in the sign of the effect, and only when raising the poverty line from its initial 17% to over 65% of the population.

Our final robustness check considered the impact of moving from a labor market in which unskilled labor is reasonably mobile between rural and urban employment, and a strongly segmented labor market specification based on parameters estimated for China, where there are formal barriers to mobility such as the residence permit system. Under the strongly segmented labor market assumption, rural unskilled wages rose much more than urban wages in response to increases in agricultural prices, and rural poverty rates rose less with a single labor market. However, urban poverty rose by more, and the impact on national poverty rates was essentially the same, as shown in Fig. 2.

Figure 2.

Simple-average poverty impacts under three labor market assumptions.

7. Concluding remarks

Our study has attempted to shed some light on the linkages between higher global food prices and poverty. By applying a simple approach of calculating the first-order welfare changes of households covered in ten detailed surveys, it was able to assess the impact of higher food prices on poverty in our sample of nine low-income countries. Even though the methodology employed in the study is simple, it is powerful enough to give us a good understanding of the underlying mechanisms and the expected magnitude and direction of change of the poverty rate of the developing countries resulting from the changes in the global food prices.

The findings of the study suggest that the overall impact of higher food prices on poverty is generally adverse. Although there are variations by commodity and by country, poor people generally appear to be net consumers of food and as such tend to be hurt by higher food prices. This conclusion is much more obvious for urban households where farming is much less important. Even though many rural households gain from higher food prices, the overall impact on poverty remains negative.

These findings are reinforced by the results of the second simulation undertaken in this study, where we calculated the impact of the observed increases in the global food prices between 2005 and 2007. Again, we discovered that the average impact of this development was to increase poverty for a majority of the countries covered in our sample, mainly due to the negative impact of higher wheat prices, followed by the prices of rice, dairy, and maize. There was considerable variation among countries and the types of households in both the impacts of a given commodity price change, and in the effect of the particular constellation of price changes considered over the 2005–2007 period. While there were a few cases where higher commodity prices lowered rural poverty, in most cases poverty—even rural poverty—increased, and the sample average poverty impact was clearly adverse.

A number of robustness checks are reported in detail in Appendix B of the online version. The first of these examines the effect of changing the size of the shock. A second considers the sensitivity of the results to changes in the poverty line. A third examines the implications of a segmented labor market, with strong resistance to movement of unskilled labor between agricultural and nonagricultural employment. These robustness checks lead to some changes in specific results, but leave unchanged the broad conclusion that increases in prices of most of the staple foods considered will much more frequently increase, rather than reduce, poverty in low-income countries.

Our analysis includes some households that are net buyers and some which are net sellers of the staple foods considered. While it is possible that higher prices of staple foods could lower poverty by raising the incomes of some poor farmers, this effect was, in most cases we considered, offset by adverse impacts on poor households that were net buyers of food.

A key current policy question is the impact of the current commodity price surge on global poverty. While detailed information on a large group of countries is not available, we can use the analysis developed in this article to guide some back-of-the-envelope calculations through the first quarter of 2008. For this period, we felt it necessary to adjust these price changes for the decline in the dollar relative to the currencies of our sample countries, and for inflationary effects. We also reduced the price changes to adjust for barriers to transmission of changes between domestic and international prices, using 66% as a “guesstimate” of the average value of this parameter7 for low-income countries.

With these assumptions, we can assess the impacts on poverty in each sample country taking into account both increases in commodity prices and estimated impacts on unskilled wage rates relative to other prices. The simple average of the estimated effects on national poverty rates (U.S.$1/day) in this nine-country sample is an increase of 4.5 percentage points. Applying this average result to all low-income countries translates into an increase in the poverty headcount of 105 million people (out of the low-income population of 2.3 billion). Alternatively, as the rate of poverty reduction has averaged 0.68% annually since 1984, a 4.5% increase in the poverty headcount corresponds to a loss of almost seven years of poverty reduction.

While any such back-of-the envelope calculation must be treated with great caution, there seem to be good reasons to be concerned about the potentially adverse impacts of large changes in world food prices for poverty in poor countries. The high shares of staple foods in the expenditures of poor people increase their vulnerability to food price rises, while the limited share of output marketed by small, subsistence farmers reduces their benefits. There are many possibilities for mitigation of these poverty impacts—many of which are currently being explored by governments and the development community—but there are also risks that the full costs could be even greater, particularly if the surge in food and energy prices is transmitted into higher overall inflation rates.

Footnotes

  • 1

    The findings, interpretations, and conclusions expressed in this article are entirely those of the authors. A longer version of this article providing additional details and robustness checks is available online, at http://econ.worldbank.org.

  • 2

    Data were kindly provided by Hans Timmer.

  • 3

    Considering the net trade position of a household, or a country, reversal of the sign of the welfare impact requires that the net trade position be more than reversed when only the direct impacts of the price changes are considered. We relax this assumption in the empirical analysis by considering impacts through changes in the wage rate for unskilled labor.

  • 4

    Although small changes in the consumption of near-subsistence households may generate large proportional changes in their net sales.

  • 5

    We would like to thank Carolyn Turk and Martin Rama for making available these data.

  • 6

    According to this definition, a person is poor if he/she consumes less that 1.08 USD in 1993 Purchasing Power Parity terms. The years for the poverty lines by country were (Bolivia, 2002; Cambodia, 1997; Madagascar, 2001; Malawi, 2004; Nicaragua, 2001; Pakistan, 2002; Peru, 2002; Vietnam, 2002; and Zambia, 2002–2003).

  • 7

    While the evidence on the value of this parameter is limited, Dawe (2008) found an average of 51% transmission from world to domestic prices between 2003(Q4) and 2007(Q4) for a sample of Asian countries, including countries such as India and the Philippines, where policies have typically sought to insulate domestic prices from world market price shocks.

Ancillary