How much do the poor spend on food when their income increases? We estimate a key economic parameter—the income elasticity of food expenditures—using data from the randomized evaluations of five conditional cash transfer programs in Mexico, Nicaragua, the Philippines, and Uganda. The transfers provided routine, exogenous increases of 12 to 23 percent of baseline income for at least a year to recipients at or below the global poverty line. Using pooled ordinary least squares and Bayesian hierarchical models, we first show that expenditures on all food categories increase with income. But even among some of the poorest people in the world, all of whom are experiencing high hunger levels, our estimated income elasticity for food is 0.03, i.e., much smaller than many published estimates that either rely on cross-sectional variation or study responses to large income shocks. Next, we run the first credible test of Bennett’s Law—the empirical regularity whereby poor households respond to income increases by (i) shifting spending from coarse to fine staples, or (ii) spending more on protein than staples—and find partial support for it. While income increases lead consumers to substitute fine grains for coarse grains and protein for staples, again the estimated shifts are smaller than previous estimates. Quantifying how small and routine income changes affect food demand in low- and middle-income countries can inform the policy discourse on poverty reduction, nutrition, and social protection, as well as the debate on the impact of economic growth on global carbon emission patterns.
