Twenty-four-hour recall and observed-weighed food record surveys are generally regarded as the gold
standard of food consumption surveys, but they are expensive, conducted infrequently, and not available
in most countries. As a result, public health nutritionists and food security analysts and policymakers have
increasingly turned to alternative data sources, including that already being collected in household
consumption and expenditure surveys (HCES). HCESs are nationally representative, multipurpose
surveys that are routinely conducted in more than 120 countries. While they are readily available, using
them to estimate food consumption is plagued by several data limitations, perhaps foremost among which
is how data on consumption of food prepared away from home (FAFH) and composite foods (CF) are
collected. Generally, no information is provided about the quantity FAFH or about the ingredients used
to make many CF resulting in underestimation of food consumption and nutrient intakes, and
overestimation of the prevalence of inadequate nutrient intake, including food insecurity, and the size of
the nutrient intake gap.
To address the FAFH/CF information gap, many analysts employ an imputation method first
developed by Subramanian and Deaton (S–D). The method consists of imputing the food’s caloric content
as the product of expenditures on the food and the reciprocal of the average national cost per calorie,
where the latter is calculated using all other foods for which quantity and caloric content data are
available. This paper assesses the S–D approach for imputing the caloric intake of households from FAFH
and CF by juxtaposing it with the imputations of alternative approaches, and extends these approaches to
four additional nutrients—vitamin A, iron, zinc, and calcium.