NAS releases report on food security and hunger measurement.

The Committee on National Statistics (CNSTAT), part of the National Academies, today issued its long-awaited report on U.S. food security and hunger measurement. Responding in part to concerns raised by staff at the Office of Management and Budget (OMB), the report criticized the use of the term "hunger" as the severe range of the scale of hardship that the federal government describes with the term "food insecurity."

The panel therefore concludes that hunger is a concept distinct from food insecurity, which is an indicator of and possible consequence of food insecurity, that can be useful in characterizing severity of food insecurity. Hunger itself is an important concept that should be measured at the individual level distinct from, but in the context of, food insecurity.... The panel urges USDA to consider alternate labels to convey the severity of food insecurity without the problems inherent in the current labels.

Currently, the federal food security survey asks 18 questions about symptoms of food-related hardship and, on the basis of the number of affirmative responses, classifies households as "food secure," "food insecure without hunger," and "food insecure with hunger." The food security statistics are heavily used in government to evaluate the success of anti-hunger programs (one might better say the limited success, given that the rate of household food insecurity has been rising for several years). The report points out that the term "hunger" is particularly "evocative" and politically potent. If, as seems possible, the report's recommendations cause a halt for several years in federal efforts to measure how many Americans are "hungry," it will be a significant disappointment to anti-hunger advocates. That, in turn, would remove one of the main constituencies that has supported the fairly expensive survey collection effort.

The report also addresses a number of technical issues. It recommended increasing the attention to the "frequency" of specific hardships -- measuring not just whether anybody in the household skipped meals during the year, for example, but how frequently that symptom of hardship occurred. At the same time, the report recommended investigating a more complicated family of statistical models to take account of the new focus on frequency issues. While taking account of frequency of occurrence is a great idea, I believe this latter recommendation would reduce the transparency of the survey results and hinder their use for important communication and public policy purposes. For example, with these models, it will no longer be possible to classify households as "food secure" or "food insecure with hunger" by simply counting up the number of affirmative responses to 18 survey questions. Also, the proposed family of models relies on more severe statistical assumptions than those that are required for the current approach. Nobody needs an "Item Response Theory (IRT)" model to measure unemployment or poverty, and it is not clear to me that ever more arcane members of this family of models will contribute to clear policy-relevant information about food security and hunger either.

Speaking of policy-relevance, the report recommended that USDA stop using the rate of food security and hunger as way of measuring the success of anti-hunger programs, under the provisions of a law known as the "Government Performance and Results Act (GPRA)."

The panel concludes that an overall national estimate of food insecurity is not appropriate as a measure for meeting the requirements of the GPRA. Even an appropriate measure of food insecurity or hunger using appropriate samples would not be a useful performance indicator of food assistance programs, because their performance is only one of many factors that result in food insecurity or hunger. Consequently, changes in food insecurity and hunger could be due to many factors other than the performance of the food safety net.

This recommendation also entails a loss. The GPRA act specifically asks federal agencies to evaluate the success of their programs using genuine outcome measures (like reducing hunger) rather than process measures (like delivering a certain number of free meals). Furthermore, I am not sure this recommendation is based on clear reasoning. It is quite correct that factors other than government programs also influence food insecurity and hunger, and yet rates of food insecurity and hunger may still provide useful information about the success or failure of major national anti-hunger and anti-poverty programs over time.

I suppose my conclusions run quite contrary to the broad themes of this report. To provide good information about the hunger problem in America, and to facilitate continued improvements in anti-hunger programs, future changes to the food security and hunger measure should aim for ever-greater simplicity, transparency, and policy-relevance.