Human Factors of XR: Using Human Factors to Design XR Systems
Was the Global Food Crisis Really a Crisis? Self-reported Data from the Gallup World Poll
1. ETHIOPIAN DEVELOPMENT RESEARCH INSTITUTE WAS THE GLOBAL FOOD CRISIS REALLY A CRISIS? Simulations versus Self-reporting Derek Headey IFPRI ESSP-II Ethiopian Economic Association Conference July 21, 2011 Addis Ababa 1
3. 7/20/2011 1. Introduction From 2006 to 2008 the international prices of wheat and maize roughly doubled, rice prices tripled Widely assumed that the poor would suffer: The World Bank estimated that as many as 100 million people would be thrown into poverty, subsequently raised to 160 million FAO claimed 75 million were thrown into hunger No IFPRI model, but we also emphasized that the poor would suffer
4. 7/20/2011 1. Introduction But how good is the evidence base for these claims? All of the global evidence thus far is based on simulations of poverty or hunger incidence Part of what comes out of these models is what goes in; other flaws too This paper takes a different track by assessing self-reported food insecurity from the Gallup World Poll (GWP)
5. 7/20/2011 2. Brief critique of simulation studies 2 types: FAO/USDA hunger and WB poverty estimates Hunger estimates deeply flawed Extremely difficult to measure calorie intake Extremely difficult to measure calorie requirements No means of gauging access shocks, because there are no prices or incomes (e.g. Asian financial crisis) In fact, FAO has relied on USDA estimates, not its own Conceptually flawed: with rising prices, people may cut back on quality not quantity; and micronutrients matter
6. 7/20/2011 2. Overview of simulation studies USDA model estimates impacts of higher prices on calorie availability based on reduced imports This has problems but 2 main problems are: USDA omits China, Mexico, Brazil – big countries USDA estimates are contradicted by their own historical data: cereal consumption did not go down in any region over 2007/08, except wheat consumption in Africa So calorie approach is very flawed for looking at impact of shocks; widely reported FAO numbers not credible
7. 2. Overview of simulation studies World Bank poverty simulations better Measure net benefit ratios (whether HHS are net food consumers or producers) so price changes easy to model But sample sizes are weak (China always excluded) and there are significant doubts about the shock imposed on the model: only food prices rise, not nonfood commodities, not fuel, and not incomes. The latter is a big issue: incomes in developing world have been growing strongly (cited as cause of crisis!)
8. 3. Gallup World Poll (in brief) Since 2005/06 GWP has interviewed HHs in around 150 countries using randomly selected, nationally representative samples Surveys are smaller than LSMS, but margins of error around 3.3 percentage points (though possibly more for specific questions) No major sampling biases reported, though some oversampling in 3 Chinese provinces, which Gallup claims to have made adjustments for
9. 7/20/2011 3. Gallup World Poll (in brief) “Have there had 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/No?” [food insecurity] Potential problems: No common definition of food Standard biases: fear, pride, media coverage Sensitivity to question ordering. In China in 2006 food affordability question followed income question. Did this prime respondents to answer “yes” more often?
10. 3. Are trends in self-reported indicators plausible? In the paper I look at cross-section patterns, and I show strong correlations with other indicators, as well as some potential problem countries Yet my main interest is not levels, but trends. Indicators could be biased in levels and still be unbiased in trends. To test validity of self-reported food insecurity, I regress changes in this measure against changes in food CPI and percentage changes in GDP per capita (i.e. disposable income) using a panel of countries
11. 4. Are trends in self-reported indicators plausible? Note: I use monthly food CPI data, which matches recorded month of GWP survey. I only have yearly GDP data, so match is imperfect I interact growth and food inflation with levels of income for obvious reasons Panel dataset is unbalanced cross-sectionally and survey timings are also different
12. Table 5.2—Are changes in self-reported food insecurity explained by economic growth and food inflation?
13. 7/20/2011 4. Are trends in self-reported indicators plausible? So both economic growth and food inflation have expected signs & strong marginal impacts Strong impact of growth is particularly important – elasticity is as strong here as it is in poverty & growth literature – but impacts conditional on income levels Recall that rising incomes don’t feature in any simulation analysis scenarios Caveat is that with fixed effects excluded, R-squared is quite low (0.10), so other things are driving self-reported food insecurity, and measurement error may be sizeable
14. 5. Trends in self-reported food insecurity Given some validation from the regression results, it seems that trends in GWP data may impart some useful information I focus on 70 LDCs - 70% of developing world population – and look at changes between a pre-crisis (2005/06), a food crisis (2007/08), and an early financial crisis wave (2008/09). Note that 2007/2008 refers to surveys conducted in 2008, but asking questions about previous 12 months (etc.) I also run a range of sensitivity analyses relating to China/India, omitted countries and using predicted changes rather than actual changes (like instrumental variables)
15. Table 14. Alternative estimates of the global food insecurity trends (millions of people): 2005/06 to 2007/08
16. 7/20/2011 5. Trends in self-reported food insecurity So self-reported food insecurity went down by a huge number under almost any assumption about errors What explains this? In a nutshell, it appears to be because of strong economic growth and very limited food inflation in big countries: China, India, Indonesia, etc Note that econometric results suggest that food inflation did increase poverty, but only if incomes are held constant (same as WB poverty simulations) Also note that GWP food insecurity did rise in Africa
17. Figure 8. Rapid economic growth and limited food inflation in most large countries: 2005/06 to 2007/08
18. Table 13. Estimating changes in self-reported food insecurity by backcasting and forecasting
19. 7/20/2011 7. Conclusions Self-reported data come with caveats so we need to be very careful: we still don’t know the true impacts of the “crisis” But GWP certainly shed light on the potential importance of rising incomes, and certainly cast doubt on simulation results FAO should consider abandoning calorie approach WB should consider modeling income changes as well as food price changes Both institutions and others should think about how we can improve measurement of food insecurity Food consumption scores and anthropometric measures arguably get closer to what we really mean by food insecurity
20. Table 9. Trends in self-reported food insecurity prevalence in The developing world: weighted and unweighted means Table 10. Estimated trends in the numbers of food insecure people (millions) in 58 developing countries