We must find a better way of monitoring the impact of food insecurity, argues Derek D. Headey.
Food prices have risen sharply in recent years and look set to remain high and volatile for several years to come — a global food crisis that raises concern for the welfare of the world's poor.
For poor people who produce much less food than they consume, a sharp rise in food prices amounts to a sizeable decline in disposable income and sharp reductions in both the size and the quality of their diet.
But has food insecurity really risen in recent years? The answer, according to international institutions such as the UN Food and Agriculture Organization (FAO), is a resounding 'yes'. Yet survey data throw this conclusion in doubt, raising questions about how to measure the impact of food crises.
The FAO claimed that around 75 million people faced hunger in the 2007–08 food price crisis, and World Bank researchers estimated that 105 million people were forced into absolute poverty, living on just US$1 per day.
These two quite different indicators of food insecurity are broadly consistent. Even so, we should be cautious about what these simulation results really tell us.
In fact, the FAO found that its global hunger model was incapable of gauging the impacts of higher prices on calorie intake because it did not incorporate food prices.
As a result, the FAO had to rely on a US Department of Agriculture (USDA) trade model that excluded middle-income countries, such as Brazil, China and Mexico, and used reduced food imports as the means of estimating changes in calorie supply in response to higher prices.
Yet the USDA's own historical data for 2007–08 cast serious doubt on the accuracy of this model's predictions: at a regional level, calorie supply almost universally went up in 2007 and 2008. 
The World Bank's models were conceptually better equipped because they used household-level data on income sources, expenditures and food prices. But what comes out of a simulation is largely a direct result of what goes in. And in terms of the model's key assumptions, what typically goes in is just a change in food prices.
Self-assessed food insecurity
If the results from simulations are suspect, why not just ask people if they had problems affording food? Using the Gallup World Poll, a well-established pollster in the United States and elsewhere, I have been able to track self-reported food insecurity in a large number of developing countries from 2005 onwards. 
The main result is that self-assessed food insecurity was actually lower in 2008 — the height of the food crisis — than in 2006, and by a large number: roughly 100 million people.
This certainly contradicts the messages sent out by the FAO, the World Bank and others. But the picture isn't entirely rosy.
A closer reading of the numbers shows a sharp divergence between countries with falling levels of food insecurity, such as China, India and Indonesia, and those with rising food insecurity, many of which are in Africa or Latin America.
Since large Asian countries make up the bulk of the world's poor, what happens in these countries is critical for global trends. But food inflation in these countries was fairly modest, and they experienced considerable economic growth, which more than offset the adverse impacts elsewhere.
Can we do better?
The Gallup data offer a different picture to that of the simulation models used by the World Bank and the FAO, but the unfortunate reality is that none of these indicators of food insecurity is satisfactory.
Even in the best of times, it is often difficult to say confidently whether hunger or poverty is higher in one country or another. And self-assessments could obviously suffer from a number of biases.
To gauge the impact of economic shocks on food insecurity, we need to do even better. We need more regular welfare surveys that go beyond calorie and income indicators to look at diets as a whole. This is because poor people can react to crises in ways that are not always obvious beforehand.
In the Indonesian financial crisis of 1998, for example, rice prices rose sharply but poor people generally did not consume less rice. Why? Because rice was still the cheapest source of calories available. Instead they cut back on dairy products, eggs and meat. So although calorie intake stayed roughly the same, anaemia rates among Java's children rose from 52 per cent to 68 per cent in just 18 months. 
The good news is that there is now a stronger impetus to improve our measurement of food insecurity. For example, there is growing awareness of the importance of measuring nutrition outcomes at the individual level (such as wasting, stunting and anaemia), rather than simply gauging calories and incomes at the household level.
It is also more widely recognised that we need to observe diets as directly as possible, and researchers are experimenting with improving estimates of food consumption and dietary diversity. These indicators generally seem to be stronger predictors of actual nutrition outcomes; the hope is that they can also capture the unpredictable dietary changes made by poor people in times of crisis.
There is now renewed interest in making such improvements. The FAO will host a symposium on measuring food security in January 2012, the first such event for 10 years.  It is crucial to reach a consensus on how the international community can better monitor food security in a more volatile world.
Derek D. Headey is research fellow in the Development Strategy and Governance Division of the International Food Policy Research Institute (IFPRI), headquartered in Washington DC, United States.
 Aid Watch FAO senior economist responds on "made-up world hunger numbers" (2010)
 International Food Policy Research Institute Was the Global Food Crisis Really a Crisis? (IFPRI 2011)
 Block, S. A. et al. Macro shocks and micro outcomes: child nutrition during Indonesia's crisis. Economics and Human Biology 2, 21–44 (2004)
 International Scientific Symposium on Food & Nutrition Security Information: From valid measurement to effective decision-making (2012)