Chapter 2


Box 2.1 Reliability of Methods of Food Consumption Measurement

A number of studies have shown that estimates of food consumption figures taken from expenditure surveys show
much greater variation with income level than estimates taken from 24-hour-recall surveys. A study of 496
households in Kenya undertook both food-expenditure and 24-hour-recall surveys and the results are shown in the
table below.

Family calorie availability and calorie intake per capital, by food group and expenditure quartile

 Data Source & Food Expenditure Quartile Quartile 4
 Group minus Quartile

 All I 2 3 4
 Calorie Intake'
 Maize 1130 1098 1149 1140 1130 +32
 Other staples 230 185 222 248 264 +79
 Meat/Fish 145 112 120 161 187 +75
 Vegetables 43 39 39 48 46 +7
 Fruits 21 19 16 II 37 +18
 Others 293 197 292 303 381 +184
 Total 1862 1651 .1839 1912 2045 +394
 Calorie Availability2
 Maize 926 799 878 990 1037 +238
 Other Staples 134 259 374 471 540 +281
 Meat/Fish 138 90 123 148 192 +102
 Vegetables 50 37 44 57 63 +26
 Fruits 129 100 125 125 165 +65
 Others 230 155 216 251 296 +141
 Total 1884 1441 1759 2043 2293 +852

 Source: Bouis, Haddad & Kennedy, "Does It Matter How We Survey Demand for Food? Food Policy, 1992.
 Calories computed from 24-hour-recall survey.
 Calories computed from food-expenditure survey.

 There is a much greater increase in calorie availability than intake as between the top and bottom deciles, and most
 of this difference arises from higher estimates of staple consumption in the top quartile and lower estimates of
 maize consumption in the lowest quartile. The authors suggest that high income groups underestimate the number
 of meals served to non-household members and therefore food expenditures overstate family food consumption.
 Low income groups underestimate the number of meals eaten outside the house and therefore food expenditures
 understate family food consumption.

 The differences shown are great enough to be potentially misleading if used in isolation to estimate the impact of
 income increases on low-income food consumption and nutrition. Though the food expenditure data would
 probably give reasonable information on the way that market demand for food would change with changing
 income, the 24-hour recall data would be more appropriate to examine the relationship between nutritional status
 and income. In both cases, the figures should be used with caution.