Richard King, my highly numerate colleague, grapples with the confusion surrounding the FAO’s hunger numbers.
Global hunger numbers must be among the most widely quoted and over-interpreted of all the indicators at development wonks’ and 
campaigners’ disposal. ‘One billion people (one in seven of the world’s population) go to bed hungry’ is a compelling headline and is used variously to argue for more effective social protection mechanisms, increased investment in smallholder agriculture, and effective measures to curb food price volatility. All are urgent causes, and all are worthy of a punchy ‘killer fact’ or two.
In fact, so powerful is the draw of a current big-scary-number that until recently the World Bank’s website displayed a ticking ‘hunger clock’, which extrapolated the (then) latest hunger estimate from the UN’s Food and Agriculture Organization (FAO) and portrayed a situation getting graver by the second. In April the same clock was displayed several metres above the heads of Washington pedestrians in the lead-up to the Bank’s Spring Meetings. But for all the power of a big number, and all the urgency of the underlying situation, is this metric anything more than smoke and mirrors?
Over at Global Dashboard, David Steven has succinctly summarised what we don’t know about how many people are hungry. He notes that the FAO’s estimates for the number of undernourished people in 2009 and 2010 have been withdrawn and no figures for 2011 have been estimated. As David puts it:
“In the midst of the first ever global food crisis… the lights have been turned off. 837m people were probably hungry four to six years ago. Maybe. That might have gone up above a billion, or perhaps it didn’t. Hunger is either resurgent or it isn’t.”
Whilst it is frustrating as an advocate to have no current global data with which to bash people over the head, this isn’t all bad news.
First, the FAO should be applauded for revoking their problematic 2009 and 2010 estimates.
In probable response to political pressure for data relating to the food price crisis of the time, these guesstimates departed from the usual estimation method and leaned heavily on a USDA trade model that was based on expectation of global economic collapse in 2009. Largely as a result of this economic conjecture, the number of ‘hungry people’ topped one billion for the first time. As it transpired, the sky didn’t fall in, economic projections improved, and the 2010 hunger projections consequently returned below the billion mark to 925 million people. To leave this model-based yo-yoing in the historical record would be deeply misleading, and I am pleased to see these figures scrubbed out.
Second, with the lights currently out and FAO statisticians back at the drawing board, the rest of us should take the chance to pause and consider how we have been using the FAO’s Prevalence of Undernourishment (PoU) indicator (the proper term for the 1 billion number). In truth, this has never been suited to tracking hunger in anything like real-time, and has been routinely over-interpreted.
How is such a global figure arrived at? By asking people if they’re hungry? By sampling people’s nutritional intake? No. As the otherwise excellent technical background paper for the recent Committee on World Food Security round table on monitoring food security rather abstrusely notes,
“The calculation is an exercise in model-based statistical inference: A probability distribution model is assumed for the annual average dietary energy intake of a representative individual in the population and its parameters are estimated on the basis of the best available data. Required data include: (a) the total availability of food in the population, (b) the demographic structure of the population (by sex and age-classes), (c) information on the distribution of food access within the population, and (d) a normative level of minimum dietary energy requirements to set a lower bound of adequate nutrition. Once the probability distribution is characterized and the threshold is set, the proportion of the population that is likely suffering from chronic food deprivation, PoU, is estimated as the probability mass that falls below the threshold.”
Right. In other words, no hungry people were harmed (or even consulted) in the making of this number. The calculation relates not to any real person, household, or other group experiencing any actual hunger but instead to a ‘representative’ individual within a population. Equally, because the reference period is a year, the calculation is not equipped to capture the impacts of acute food shocks due to factors such as price volatility, or climatic events, or even due to regular, seasonal, waxing and waning of hunger.
And because the calculation is based on the food available to a whole population, rather than actual access to food, it doesn’t account for intra-household inequalities, or food wasted or spoilt in the home, or even food produced for subsistence consumption. The calculation is also very sensitive to errors in food balance sheets (used to calculate the total availability of food), which measure food consumption from a food supply perspective. There are several weaknesses of a food balance sheets approach, not least that if changes to food storage are not accurately captured (which is difficult to do) this throws off the rest of the calculation. This is why hunger estimates (until 2008) were averaged across three years, to try to minimise any distortions from year-to-year inaccuracies in storage data. Couple these limitations with the time taken to collect other food distribution parameters from nationally representative household surveys, and it’s soon apparent that this FAO measure is poorly suited to the timely tracking of current food insecurity.
So where next? Following the CFS Round Table, the Committee endorsed a proposal to create a richer suite of core food security indicators and strongly recommended that the FAO improves its current measure of undernourishment. According to the FAO’s State of Food Insecurity in the World 2011 report:
FAO will make several adjustments, including in the estimation of how changes in food access due to changes in income and food prices affect undernourishment. Work is also underway to improve the construction of food balance sheets. A large number of household expenditure surveys are being processed to provide improved estimates of the distribution of food consumption within a country. FAO’s measures of undernourishment will also be complemented with a number of other indicators intended to better capture the multifaceted nature of food insecurity.
But all this will take time to overhaul, and will likely still result in indicators that are more suited to measuring recent chronic food insecurity rather than current acute hunger. For that, we may have to turn to more subjective indicators, such as those in the Gallup World Poll surveys recently analysed by IFPRI, in which people were asked: “Have there been times in the past 12 months when you did not have enough money to buy the food that you or your family needed?” (yes or no). This is an imperfect alternative, not least because ‘food’ and ‘need’ are more abstract than counting calories and are likely to be interpreted differently depending on respondents’ location. And, because the question is asked of the household unit as a whole, this still fails to capture gender-based inequalities in food insecurity. If nothing else, let’s hope that the FAO review finally gives us an indicator that helps us to track this particularly crucial dimension of hunger…]]>