Book excerpt: Cost and affordability of healthy diet (CoAHD) metrics developed in a handful of academic studies have quickly become mainstream food security indicators among major development institutions. The World Bank and FAO now report CoAHD statistics in their widely used databanks, and the UN’s State of Food Insecurity and Nutrition in the World (SOFI) reports CoAHD metrics on an annual basis, with the headline conclusion being that over 3 billion people worldwide cannot afford a healthy diet. While quantifying affordability constraints is indeed a vital addition to the suite of global food security indicators, there is a dearth of scientific analysis on the accuracy and sensitivity of CoAHD methods. Published global CoAHD estimates rely on three implicit assumptions: that demographic differences across countries have little effect on average diet costs; that non-food expenditure requirements have little systematic variation across countries; and that international food price data is representative in a population sense and product coverage sense. Testing these assumptions on the cost of the EAT-Lancet reference diet, we find sizable sensitivity of baseline methods to adjusting diet affordability estimates for systematic cross-country differences in demographic profiles and non-food expenditure requirements, smaller effects of adjusting for inadequate food product coverage in international price data, and inconclusive evidence on issues of urban bias in price surveys. Our proposed methodological improvements significantly change country, regional and global estimates of healthy diet affordability, though not the headline conclusion that several billion people cannot afford a healthy diet. Even so, the accuracy, rigor, and reliability of CoAHD statistics warrant closer investigation given their widespread adoption and utilization.