06 June 2011

Uncertainty in Food Insecurity: What Do We Really Know?

Following my post yesterday on the NYT discussion of food and climate change, a fascinating paper came my way (thanks KH!) from the International Food Policy Research Institute that challenges conventional wisdom on recent trends in the impact of commodity price spikes on food availability to the world's poor.

Everyone knows, me included, that far more people have been thrown into food scarcity by the recent food crisis, right?  Yesterday's NYT article repeated this conventional wisdom:
Those price jumps, though felt only moderately in the West, have worsened hunger for tens of millions of poor people, destabilizing politics in scores of countries, from Mexico to Uzbekistan to Yemen.
Derek Headey of the IFPRI says maybe not.  In a provocative paper titled, Was the Food Crisis Really a Crisis? Heady takes issue with model-based calculations of food insecurity and asks if self-reported indicators are less bad (see Headey writing VoxEu for a brief discussion of the paper's main points).

Headey characterizes the two leading approaches to calculating food insecurity as follows:
[T]he FAO uses minimum energy requirements as a “hunger line” and then estimates the proportion of people falling below that line based on estimates of the total number of available calories in the country and a lognormal distribution of calories estimated from income data from household surveys.
USDA uses calorie–income elasticities based on cross-country data on per capita calorie availability (as per the FAO) and per capita income, along with income distribution data from the World Bank. It then incorporates these elasticities into a partial equilibrium global trade and production model that includes elements like a food demand function.
His paper explains in detail that both of these methods (as well as several others in practice and in the literature) have severe weaknesses, concluding:
We think a fair assessment of these hunger models is that they are far too crude to reliably predict the impact of access shocks, such as a rise in international food prices. . . [V]irtually all of these simulation exercises asked a very specific question of their models: “What would happen to poverty if food prices went up, and only food prices went up?” However, this ceteris paribus assumption definitely does not apply to the period 2005–2008.
Headey then employs an alternative methodology to assess food insecurity, self-reported data from the Gallup World Poll, noting that it too has serious weaknesses:
Since self-reported indicators of welfare inevitably have flaws (see Section 3 below), the question is not whether our approach is imperfect but whether it is more or less imperfect than alternative approaches.
After conducting a wide range of tests and sensitivity analyses the paper concludes provocatively that fewer people may have been in food insecurity during the food crisis than before it, explaining that the "all else equal" factors were not so equal:
This paper has explored the usefulness of the Gallup World Poll indicators of self-reported food insecurity and hunger for assessing global food insecurity patterns and trends. In this concluding section we overview the strengths and weaknesses of these data, and summarize our main findings regarding trends in the two indicators of interest. To reiterate the main findings, our main result is that in 2007/08—the food crisis period—there were fewer people reporting trouble affording food than in 2005/06. We are hesitant to say exactly how many, though two of our most conservative estimates suggest that global food insecurity fell by 60–90 million people, although these would be lower-bound estimates if the trends in China and India were somewhat stronger than a 2–3 percentage point reduction in food insecurity assumed or predicted in these scenarios. Certainly the fantastic growth rates and muted food inflation in these two countries could warrant a strong downward trend. Of course this conclusion does not mean that the global food crisis did not hurt. On the contrary, it hurt poor people in many countries, particularly in Africa. Yet our main finding is that the food crisis had a very limited impact in the most populous countries, thus casting into doubt existing estimates of global trends in food insecurity and hunger.

This last point is particularly important because all existing simulation-based estimates of the impacts of the food crisis omit China, and many omit other large countries. Yet our results suggest that strong economic growth prevented the surge in international food prices from resulting in a genuine global crisis. Moreover, the fact that populous countries tend to be wary of heavily relying on international cereal markets—and the fact that many large countries also imposed export restrictions to protect domestic prices (Headey 2011a)—prevented them from experiencing significant food inflation. However, on this last point we add a note of caution. The events of 2005–2008 are not necessarily a good predictor of food price impacts in 2010/11. While countries like China and India are still growing rapidly, a notable difference in the current crisis (2010/11) is that some of these large countries are now experiencing quite rapid food inflation (although not yet rice price inflation). Hence the global impact of the current crisis could potentially be significantly worse than that of the 2007/08 crisis.
When I think about it, this makes a lot of sense.  In November, 2010 Dani Rodrik asked whether high food prices are good or bad for poverty, and provided this answer:
It depends on whether the poor are selling or buying, of course.

High food prices benefit poor farmers who are net food sellers, and hurt poor food consumers in urban areas. Low food prices have the opposite effects. In each case, the net effect on poverty depends on the balance between these two effects. But you would hardly know it from reading what NGOs and international organizations have produced on the topic.
Headey concludes a short discussion of his paper with the following bit of wisdom:
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.
One of the frustrations of practicing policy analysis is that you are reminded on a daily basis that we are never as smart as we think we are, and things that we thought we knew for sure, sometimes just ain't so.


  1. I agree on the uncertainty surrounding assessments (let alone predictions) of the impact of world food prices on household food insecurity or malnutrition in various countries. Too many variables are at play.

    Analyses available so far from household surveys and ex post estimates for 2007-2010 suggest a small transient increase in the overall rate of undernourishment (as measured by FAO). At world level its prevalence was pushed temporarily up by about 1.5 percentage points at the peak of the world food price surge, i.e. by about 11% over previous values of about 13%. The figures varies across countries but not as much as feared. Part of the increase in 2007-09 was reversed in 2010, according to FAO.

    Also poverty was only moderately affected: surveys in various countries show only small increases in the rate of poverty in several countries as a result of food inflation. The local effects of the international financial crisis may have been larger, and their relative magnitudes are difficult to disentangle.

    One reason for the moderate impact is that raw export or import (FOB or CIF) prices of food commodities like wheat or rice represent just a limited proportion of the finished products' final retail price (e.g. the price of the wheat grain required to make one pound of noodles equals about 15% of the retail price of noodles in typical developing-country conditions). The biggest impact is for crops that are directly consumed without much processing or long-distance marketing, as in maize-tortilla-based diets in Central American countries. So far, most "food riots" and related troubles have appeared mostly in those countries. These problems are in some cases worsened by gross distortion of relative prices (e.g. undervalued currencies making local food much cheaper than international levels, thus making imports much more expensive), as was seen in several African countries.

  2. My previous post referred to the existing assessment but not much to the uncertainty surrounding them.
    Regarding the FAO assessment, it is normally released in their annual SOFI report on The State of Food Insecurity in the World (http://www.fao.org/publications/sofi/en/) as a 3-year average, based on national figures on food availability and updating (if needed) the variance of the country-specific log-normal distribution. The latest figure of this kind (SOFI 2010) is based on data from 2005-07.

    Starting in SOFI 2008 FAO published only a preliminary estimated (or rather a forecast) for the current year (i.e. 2008, 2009 and 2010 in the three latest issues). Thus the recent annual figures (after 2007) are not based on a detailed country by country assessment, but on some (as yet undisclosed) method, and thus must be subject to subsequent corrections when actual data are received and analyzed (the usual lag is three years). These annual forecast figures, on the other hand, are not strictly comparable to previous three-year averages of actual ex post data.
    Referring to the absolute numbers instead of the rates of prevalence is also misleading since total world population is increasing, especially in developing countries, and therefore the number may go up while the prevalence falls or remains stable. The only reasonable measuring rod is the percentage prevalence. Otherwise, we should have to conclude that there is more unemployment now in the US than during the Great Depression, since there are more people unemployed, and that the US are now more stricken by poverty (since it has more people under the poverty line) than during most of its past history.
    Estimates based on household surveys, on the other hand, are usually obtained through econometric simulation (i.e. estimating poverty or food demand under two different price levels while keeping all the rest constant). This is not the same as an actual net impact, since an increase in food prices would cause a shift in the whole relative price structure (including wages), at varying speeds depending on countries and types of goods and jobs. These feedbacks are usually negative, i.e. they attenuate the marginal or differential impact of a single variable (such as food prices), with a lag that in a great many countries is relatively short (a matter of some months). Thus the annual, let alone triennial, net impact is much less than suggested by the econometric exercises. Other things have the irritating habit of not remaining equal.
    And finally, as remarked in my previous comment, the rise of an international food commodity does not translate into a proportional increase in the retail price of the corresponding food product (e.g. bread or noodles versus wheat grain). An impact there exists, but is normally much attenuated. Besides, staple food forms the bulk of food expenditure in only a few extremely poor countries (say Haiti). In most, a large majority of the expenditure (as distinct from the calories) corresponds to other foodstuffs (dairy products, vegetables, tubers, fruit, meat, fish and so on) that are more expensive on per-calorie terms. Those other products' price may also rise, echoing the rise in staple foods, but they are less sensitive (especially tubers like potatoes or manioc, which are generally non tradeable and may substitute for cereals).

  3. The first words of the second paragraph of my previous comment, which are:

    Starting in SOFI 2008 FAO published only

    should be:

    Starting in SOFI 2008 FAO published also.

    Sorry for this erratum.

  4. last paragraph: "new" should be "knew"

  5. Highly recommend the World Bank's recent collection of essays, "Food Prices and Rural Poverty," ed. Aksoy & Hoekman, particularly Jo Swinnen's article "Food Prices and Poverty: How the Food Crisis Changed Policy Perspectives," which asks why both long-term falling and periodically rising are both always seen as bad for the global poor.

  6. This is really interesting. I've noticed that IFPRI reports often refer to the "food crisis of 2008" to justify more funding for agricultural research.