Sunday, May 22, 2011

Food crisis: Simulation versus self-reporting

Derek Headey doubts the existing literature and analysis on food crisis and argues that the negative impacts of higher food prices in 2007-08 was more-than-compensated by economic growth (and insufficient coverage of China and India), something the existing models don’t take into account. Here is a paper (Was the global food crisis really a crisis?) by Headey.

Heady shows that global self-reported food insecurity fell during 2005 and 2008, with 60 million to 250 million fewer food-insecure people. The main reasons for this are rapid economic growth and very limited food price inflation in the world’s most populous countries, particularly China and India. Hence, he argues that food insecurity outcome shown by simulations do not match up with self-reported food insecurity level by countries. So, if the existing estimates are incorrect, then are the policy interventions designed to counter food insecurity misplaced?

Estimates of global trends in self-reported food insecurity, 2005/06 to 2007/08 (million)
Estimation scenarios Estimated change
Raw results, 70 countries -408
Raw results, 70 countries, plus assumptions for 16 omissions -326
Raw results, 68 countries, after excluding China and India 9
Raw results, 69 countries, after excluding China -132
Raw results, China and India trends adjusted by error margins -250
Raw results, China and India reductions=3 percentage points -63
Predicted change with econometric model, 88 countries -87

Source: Headey (2011) estimates from Gallup World Poll data.


In a new IFPRI discussion paper I show that these simulations suffer from serious flaws, and that their results are largely contradicted by self-reported food insecurity trends from the Gallup World Poll. The poverty simulations are often quite nice studies, but they are partial equilibrium studies with no wage adjustments, no changes in other commodity prices (like fuel, cotton, coffee, minerals), and no changes in incomes (which were growing all around the developing world from 2000-2008). Hunger simulations have more fundamental problems. Basically they count calorie availability, but the problem with a food or a financial crisis is that it is an access shock, not a production shock. Hence the FAO had to rely on a USDA model (which included reduced "calorie imports") to provide estimates of changes in hunger during the crisis. Yet in my paper I show that USDA's estimates of calorie availability (from early 2008) seem to be contradicted by USDA's own historical data on cereal availability.


[Note that self-reported food insecurity varies from country to country: in Colombia it is 67 percent, but in Nepal it is just 9 percent. The figures discussed here are global level data, not country-specific.]