1. FAO assessment of global undernourishment. Current practice and possible improvements Carlo Cafiero, ESS Rome, 12-13 September 2011 1 CFS Round Table on Monitoring Food Security

2. Outline Definitions Definitions Criticisms Criticisms ◦ Appropriateness ◦ Usefulness ◦ Validity ◦ Precision Conclusions and moving forward Conclusions and moving forward Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 2

3. FAO’s Undernourishment indicator It is used as the MDG indicator 1.9 Basic elements habitual caloric intake representative individual ◦ A distribution for habitual caloric intake is defined for a representative individual in the population Food Balance Sheets AND Household Surveys ◦ Parameters of such a distribution are estimated based on data from Food Balance Sheets AND Household Surveys Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 3

4. FAO’s Undernourishment indicator Basic elements (cont’d) representative individual falls below its minimum Prevalence of Undernourishment ◦ The probability that intake for the representative individual falls below its minimum level of calorie requirement compatible with a healthy and active life is taken as an estimate of the Prevalence of Undernourishment (PoU) in the population Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 4

5. Criticisms criticisms The FAO indicator has been subject to criticisms as a global indicator of “hunger” Criticisms should be distinguished depending on whether they refer to the: – appropriateness – appropriateness of the underlying operational definition of food insecurity – usefulness – usefulness for policy guidance and program monitoring – validity – validity as a measure of the accepted definition of food insecurity – precision – precision of the estimates produced, given the quality of the available data Rome, 12-13 September 2011 5 CFS Round Table on Monitoring Food Security

6. Appropriateness continued (chronically) insufficient caloric intake Undernourishment is defined as a continued (chronically) insufficient caloric intake chronic food deprivation population level The embedded operational definition of food insecurity is of chronic food deprivation, at the population level Criticisms: ◦ Quality of the diet ◦ Quality of the diet, as linked to an appropriate balance of macronutrients (protein/fat/carbohydrates) and essential micronutrients, may be equally important ◦ Temporary ◦ Temporary food insecurity may also be important and occur even more frequently than chronic hunger individual household ◦ Monitoring should be conducted at the individual or household level to better target interventions Rome, 12-13 September 2011 6 CFS Round Table on Monitoring Food Security

7. Appropriateness other perspectives Criticism is valid as it points to the need for other perspectives to be added to capture more fully the dimensions of food insecurity ◦ see discussion in the previous session still great value in assessing the extent of chronic hunger However, there is still great value in assessing the extent of chronic hunger, especially in recognition of the too limited progress achieved so far globally Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 7

8. Usefulness cannot be produced in a timely manner no real-time monitoring of food crises The major criticism on the usefulness of the indicator for policy purposes is that it cannot be produced in a timely manner, as the needed data are available only with a significant delay, so that no real-time monitoring of food crises is possible ◦ Steps are being taken to ensure that needed data inputs are made available sooner (tomorrow’s focus) projections of estimated PoUprojected economic contingency ◦ Research is ongoing to allow forming projections of estimated PoU based on projected economic contingency Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 8

9. Validity Two major criticisms on the validity effective use of household survey data 1. The FAO does not make effective use of household survey data (Smith 1999) single threshold level for the dietary energy requirement 2. The choice of a single threshold level for the dietary energy requirement is bound to generate errors (Svedberg 2000, 2002) Rome, 12-13 September 2011 9 CFS Round Table on Monitoring Food Security

10. Validity: use of surveys Why a statistical model? Why a statistical model? defining a minimum dietary energy requirement at the individual level ◦ Controversies exist on the possibility of defining a minimum dietary energy requirement at the individual level habitual household food consumption significant noise ◦ Information on habitual household food consumption obtained from household surveys data is affected by significant noise Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 10

11. Validity: the energy requirement threshold The argument in Svedberg (2000) stands on existence of a continuous joint distribution of intakes (x) and requirements (r) in the population. no one being considered as “adequately nourished” A continuous joint distribution would assign zero probability to the event (x = r), therefore any population would be partitioned in two groups: the under and the over-nourished, with no one being considered as “adequately nourished” ◦ See Sukhatme’s criticism of Dandekar in the ’70s. representative individual Use of the marginal distribution of intakes only, defined as pertaining to a representative individual of a broad group, avoids reference to a joint distribution for intake and requirement ◦ See Naiken (2007) Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 11

12. Precision precision of the FAO estimate at individual country level is perforce rather limited (i.e., large confidence intervals) Though never explicitly reported, precision of the FAO estimate at individual country level is perforce rather limited (i.e., large confidence intervals) ◦ Failure to report standard errors has likely contributed to misinterpretations of the actual value of country level estimates Precision of the estimate depends on: ◦ Quality of the data ◦ Appropriateness of the ancillary modeling assumptions, i.e. the assumption of Log Normal distribution for food consumption in the population Rome, 12-13 September 2011 12 CFS Round Table on Monitoring Food Security

13. Precision: quality of the data both data sources Quality of the data is a problem for both data sources ◦ Data informing compilation of FBS ◦ Data collected through household surveys Integration and comparison of the two sources Integration and comparison of the two sources of data when available for the same country and the same year promises to improve the precision of the estimate Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 13

14. Precision: modeling assumptions shape of the distribution Of the modeling assumptions, the most critical one is on the shape of the distribution and on the way its parameters are updated distribution’s parameters for many countries have not been updated ◦ Due to lack of data, the distribution’s parameters for many countries have not been updated over many years the log normal model may no longer be appropriate ◦ If over time relative shares of food consumption in the population change, not only the coefficient of variation of food consumption may have changed, but also the log normal model may no longer be appropriate Rome, 12-13 September 2011 14 CFS Round Table on Monitoring Food Security

15. Precision: modeling assumptions alternative distribution models Testing alternative distribution models ◦ Problem: we need a model because we do not “trust” household level data, but we need precise household level observations to test the model… Ongoing research in ESS is tackling this issue skewed-t distribution ◦ Preliminary results point to the skewed-t distribution as a promising alternative to the log- normal. Rome, 12-13 September 2011 15 CFS Round Table on Monitoring Food Security

16. Conclusions and moving forward is a valid measure The FAO indicator is a valid measure of the share of population having insufficient access to food should not be interpreted as a comprehensive measure of food insecurity It should not be interpreted as a comprehensive measure of food insecurity, but rather used in combination with other indicators focusing on different dimensions of food insecurity extensions of the basic framework Its usefulness will be raised thanks to extensions of the basic framework model in two directions: ◦ Developing projection models for undernourishment ◦ Calculating more indicators, e.g. estimates of excessive caloric intake which is expected to be highly correlated with nutrition related problems such as obesity Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 16

17. Conclusions and moving forward increased quality of the basic data Its precision at individual country levels could be greatly improved through increased quality of the basic data, both from Food Balance Sheets and Household Surveys the collaboration with countries For that, it is imperative to improve the collaboration with countries in collecting better and more timely data Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 17

18. Questions Improving the data (both FBD and HS): ◦ What can be done? ◦ How can be it achieved? Improving the model: ◦ Directions for added flexibility? ◦ Modeling/projecting over years with no underlying data? ◦ Expansion of the scope of the analysis? Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 18

19. Thanks ! Rome, 12-13 September 2011 CFS Round Table on Monitoring Food Security 19