VoxEU Column Development Environment Poverty and Income Inequality

Was the global food crisis really a crisis? Simulations versus self-reporting

Have rising food prices hurt the world’s poorest? The broad consensus, which is based on simulation analyses, is that they have. This column begs to differ. Self-reported food insecurity data from the Gallup World Poll contradict the consensus, and this column argues that the FAO and the World Bank need to do a much better job of measuring the impacts of higher food prices on the poor.

Have rising food prices hurt the poor or helped them? So far everything we know about this topic comes from simulation analyses, all of which suggest that rising prices have substantially raised global poverty or hunger (e.g. Aksoy and Hoekman 2010). Despite this broad consensus, simulation analyses suffer from serious methodological limitations.

Indeed, my recent research using trends in self-reported food insecurity data from the Gallup World Poll contradicts the findings of simulation studies (Headey 2011). Of course, neither approach is ideal and there is still a lot of uncertainty about what has really happened to global poverty. The only thing that is certain is that we need to do a much better job of measuring the impacts of higher food prices on the poor.

The limitations of simulation studies

Simulation studies fall into two types:

  • poverty simulations
  • hunger simulations

Poverty simulations model the effect of a real food price shock on household disposable income based on the extent to which households are net food consumers or net food producers. These studies almost invariably suggest that higher food prices raise poverty, often even in rural areas. Global estimates extrapolated from a nine-country World Bank study by Ivanic and Martin (2008) suggested that 105 million people would be thrown into poverty by higher food prices (at the $1/day level). A subsequent 73-country World Bank study estimated that global poverty rose by around 160 million people, 90 million of whom were rural (de Hoyos and Medvedev 2009). And of course, there are a number of country studies that frequently find rising poverty on the back of higher food prices.

Hunger simulations instead use the concept of adequate calorie availability, although this approach is very poorly suited to modelling price and income shocks (or “access shocks” as they are commonly called). For example, the Food and Agriculture Organisation (FAO) was previously criticised for not finding any impact of the 1997-1998 Asian financial crisis on hunger, precisely because that crisis was an access shock and not a production shock. Hence in the more recent food crisis – which is also an access shock – the FAO instead chose to rely on the US Department of Agriculture’s (USDA 2009) hunger model for low-income countries, which models food price impacts primarily via their effect on food imports. With that model USDA (and hence FAO) estimated that around 80 million people were thrown into hunger during the 2008 food crisis. Another World Bank study, using an extended FAO-type methodology, also estimated that 63 million people were thrown into hunger by the two crises (Tiwari and Zaman 2010).

What happens if not all else is equal?

While all of these multi-country studies consistently show that rising food prices hurt the poor, there are good grounds to question whether global food insecurity has really increased in recent years.

For one thing, the basic assumptions are very artificial. They invariably assume that food prices rise while all else is equal. Anyone who has followed global economic events over recent years knows full well that everything else was not equal. Fuel prices have risen dramatically as have most other non-food commodity prices. So the commodity boom as a whole obviously had ambiguous effects on poor countries. Whilst most import fuel, many export other commodities.

More generally the fact that household incomes are held constant in these simulations creates an artificial context because the 2000s were a decade of very strong economic growth in most developing regions and because wages can adjust to higher food prices. For example, Mason et al. (2011) show that food prices undoubtedly hurt the urban poor in Zambia and Kenya but that the urban poor were still better off in 2009 than at any time between 1994 and 2003.

Simulation studies also fail to account for substitution effects to cheaper calorie sources. And many of these studies omit populous countries, particularly China, where economic growth was particularly strong and food inflation particularly low.

Finally, the aforementioned hunger simulation conducted by USDA appears to be contradicted by the most recent USDA data on food availability. Since hunger is measured as an insufficient availability of calories in a country, hunger can only be influenced by domestic production shocks or by a reduction in net food imports. Yet Table 1 shows that the availability of major cereals did not decline by large margins in any world region.

Table 1. Availability of major cereals in 2007/08 and 2008/09 relative to 2005/06 (% change)

 

 
Region
Maize
Wheat
Rice
Any major declines?
 
2007/08
2008/09
2007/08
2008/09
2007/08
2008/09
Caribbean
1.1%
0.9%
7.7%
1.4%
10.0%
3.6%
no
Central America
13.4%
13.4%
-3.0%
-3.3%
3.3%
5.2%
wheat only
South America
4.8%
9.6%
2.3%
2.7%
0.6%
4.6%
no
East Asia
16.3%
18.5%
-0.1%
-0.5%
-0.7%
3.6%
no
South Asia
-4.7%
10.8%
9.6%
4.8%
6.6%
8.0%
maize only
Southeast Asia
12.5%
20.8%
3.6%
4.5%
5.1%
4.5%
no
Sub-Saharan Africa
5.0%
16.3%
-11.0%
3.8%
5.2%
10.5%
wheat only
North Africa
15.0%
30.0%
6.0%
9.7%
1.7%
15.6%
no
Middle East
9.2%
9.2%
1.4%
3.4%
-1.1%
2.7%
no
Former USSR
5.7%
11.7%
-0.4%
0.1%
-2.4%
-6.7%
rice only
Other Europe
-4.1%
-4.1%
-4.6%
-5.5%
16.2%
6.8%
maize & wheat
European Union
-10.3%
0.7%
-0.7%
3.0%
20.1%
10.3%
maize only
North America*
40.0%
56.4%
1.9%
1.1%
4.5%
5.5%
no

 

Source: Author’s estimates from US Department of Agriculture PS&D Online Database. Notes: Data generally run from July in year t to June in year t + 1. Note that all data are aggregate. *In the case of maize and wheat we have used non-feed consumption data, which includes industrial uses such as biofuels. This explains the sharp increase in North American maize consumption.

Strengths and weaknesses of self-reported food insecurity data

Given these limitations of simulation studies, it is surely worth asking what alternative methodologies have to say about global trends in food insecurity. My work makes use of the Gallup World Poll surveys conducted in a large number of developing countries both prior to (2005/06) and during the food crisis (2007/08). Gallup (2011) asks a highly useful question in this context: Have you or your family had any trouble affording sufficient food in the last 12 months? I take the percentage of respondents who answer yes to this question as a measure of national food insecurity.

Inevitably, there are obvious caveats to any self-reported indicator, as well as some not-so-obvious caveats discussed at greater length in my paper. Among the obvious caveats are issues related to test-retest reliability, differing definitions of “sufficient food”, biases related to political factors (such as fear of government) or cultural factors (such as the shame associated with measuring poverty), and general measurement error. But while we cannot rule out these problems, it is not obvious these issues would bias trends in the variable of interest, except in special circumstances (such as increasing fear of governments). A second issue relates quite specifically to the surveys conducted in China, which reported an unusually large decline in food insecurity.1 This is obviously a concern because China is large enough to have a dramatic effect on any estimates of global trends in food insecurity.

Despite these concerns, there are some important strengths with this self-reported data. The most important of these is that trends in self-reported food insecurity are robustly explained by food price inflation and per capita economic growth rates, as one would expect. Regression 2 in Table 2, for example, shows that the elasticity of food insecurity with respect to per capita income is -2.10 in low-income countries, a level commensurate with elasticities from the poverty and growth literature, while the elasticity with respect to food prices (the food CPI) is 0.74 in low-income countries. Both elasticities are significant at the 1% level. The results also show that these effects decline substantially in middle- or upper-income countries where there are fewer poor people and where food is a much smaller component of household budgets. These results therefore provide some indicative evidence that the survey measure of food insecurity is picking up the kind of effects that we would want it to.

Table 2. Are changes in self-reported food insecurity explained by economic growth and food inflation?

Regression
1
2
Dependent variable
Change in food insecurity
Percent change in food insecurity
Number of countries
107
107
Number of observations
254
254
Sample
All
All
Constant
0.54
6.81
Economic growthb (low income)
-0.70***
-2.10***
Food inflationc (low income)
0.30***
0.74***
Growth*middle income
0.63*
2.13*
Inflation*middle income
-0.24#
-0.77#
Growth*upper income
0.62**
1.43#
Inflation*upper income
-0.33**
-1.69
 
 
 
R-square
0.37
0.43

Source: See the working paper for source details. Notes: These are ordinary least squares regressions with fixed effects. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively, and # indicates marginal insignificance at the 10% level. a. The dependent variable is measured as the change in food insecurity between month M in year Y and the previous survey (Mt-1 and Yt-1). b. Economic growth is the% change in GDP per capita between the two years in which the surveys were conducted. c. Food inflation is the% change in the food consumer price index (CPI) between the month of the survey and the month of the previous survey, where the food CPI in any given month is actually the maximum food CPI in the previous 12 months. d. Low income is defined as a 2005 GDP per capita of less than $5,000 PPP, middle income as $5,000–13,000, and upper income as greater than $13,000. Note that by this definition China is defined as a low-income country.

So what does the self-reported food insecurity data show?

A striking feature of the self-reported food insecurity data is that it suggests that global food insecurity went down from 2005/06 (the pre-crisis period) to 2007/08 (the food crisis period). Moreover, self-reported food insecurity appears to have gone down by a large number, at least on the same order as the estimates of poverty and hunger increases reported by the World Bank, USDA, and the FAO. This is true irrespective of the assumptions made about China, India, and 16 countries with missing data for 2005/06 (Table 2).

Table 3. Alternative estimates of global food insecurity trends

Estimation scenarios
Estimated change in global food insecurity, 2005/06 to 2007/08
Raw results, 70 countries
-408 million
Raw results, 70 countries, plus assumptions for 16 omissions
-326 million
Raw results, 69 countries, after excluding China
-132 million
Raw results, China and India trends adjusted by error margins
-250 million
Raw results, China and India reductions=3 percentage points
-63 million
Predicted change with econometric model, 88 countries
-87 million

Source: Author’s estimates from Gallup (2011) World Poll data.

Regional data also tell an equally positive story. Table 3 shows averages by various regions as well a group of the ten most populous developing countries. Populous countries stand out for how much self-reported food insecurity decline (and not just in China). Yet

Table 4 also shows that although self-reported food insecurity went down in most regions, parts of Africa and Latin America are prominent exceptions.

Table 4. Regional trends in self-reported food insecurity (% prevalence)

Developing region
# obs.
2005/06 (pre-crisis)
2007/08
(food-crisis)
2008/09 (early financial crisis)
Big countries*
9
33.1
26.7
29.1
Sub-Saharan Africa
14
55.8
54.6
57.2
 West Africa, coastal
4
48.5
51.3
58.0
 West Africa, Sahel
5
59.6
49.2
55.2
 Eastern & southern Africa
5
57.8
62.8
58.6
Latin America & Caribbean
15
33.2
36.4
35.7
 Central America, Caribbean
7
38.4
41.4
40.3
 South America
8
28.6
32.0
31.6
Middle East
3
19.7
26.0
21.3
Transition countries
13
31.9
30.2
34.6
 Eastern Europe
6
21.8
19.7
25.8
 Central Asia
7
40.6
39.1
42.1
Asia
12
30.6
28.3
29.7
 East Asia
7
33.3
29.3
30.4
 South Asia
5
26.8
26.8
28.6

Source: Author’s calculations from Gallup (2011 #1492) self-reported food insecurity prevalence rates. Note: *“Big countries” includes China, India, Indonesia, Brazil, Pakistan, Bangladesh, Nigeria, Mexico, and Vietnam.

Cautious conclusions and caveats

The conclusions we draw from these results obviously need to be very cautious but the weaknesses of simulations results and their contradiction by the Gallup data suggest that institutions such as the World Bank and FAO need to rethink their strategy towards measuring food security. The FAO should consider abandoning the “counting calories” approach. In the first place it is extremely difficult to estimate calorie requirements and availability in any given country, and the inability of this approach to gauge the impact of price or income shocks is really a fatal flaw. As for poverty simulations, more sophisticated approaches would be useful for exploring channels of impact, but modellers need to go beyond the back-of-the-envelope approaches thus far used. We need much more research into household adaptation to rising food prices, as well as into the critically important macroeconomic channels of strong economic growth and a more general commodity boom.2 As things currently stand, there is a huge degree of uncertainty about what has really happened to the world’s poor in the recent years.

References

Aksoy, M Ataman and Bernard Hoekman (2010), “Food prices and rural poverty”, VoxEU.org, 8 October.
Arndt, C, R Benfica, N Maximiano, AMD Nucifora and JT Thurlow (2008), “Higher fuel and food prices: impacts and responses for Mozambique”, Agricultural Economics, 39(2008 supplement): 497–511.
de Hoyos, R and D Medvedev (2009), Poverty Effects of Higher Food Prices: A Global Perspective, Policy Research Working Paper 4887, The World Bank, Washington DC.
Gallup (2011), Gallup World Poll. Gallup, accessed 21 January.
Headey, Derek (2011), “Was the Global Food Crisis Really a Crisis? Simulations versus Self-Reporting”, IFPRI Discussion Paper 01087.
Mason, Nicole M, TS Jayne, Antony Chapoto, Cynthia Donovan (2011), “Putting the 2007/2008 global food crisis in longer-term perspective: Trends in staple food affordability in urban Zambia and Kenya”, Food Policy, 36(2011):350-367
Ivanic, M and W Martin (2008), “Food prices and food security”, VoxEU.org, 21 November.
Tiwari, S and H Zaman (2010), The impact of economic shocks on global undernourishment, Policy Research Working Paper Series with number 5215, World Bank, Washington DC.
USDA (2009), Food Security Assessment, 2008-09, US Department of Agriculture (USDA) Economic Research Service (ERS).


1 The sharp drop in China – from 32% in 2006 to 16% in 2008 – may be related to an overestimate in 2006 since the food affordability question followed an income question (i.e. respondents may have been primed to answer yes). I thank Angus Deaton for pointing this out to me.

2 In this regard the paper by Arndt et al (2008) is an excellent template for future work on the macroeconomic channels by which higher commodity prices influence poverty.

1,365 Reads