Tuesday, March 23, 2010

The trouble with single respondents in household surveys

Much of the development literature is based on the administration of household surveys, with typically one member of the household answering all question on the behalf of all members of the household. Of crucial importance here is whether is actually matters who answers.

Monica Fisher, Jeffrey Reimer and Edward R. Carr use survey data from Malawi and verify whether the estimates of the spouse's income by the head of household were accurate. It turns out not: they are statistically not reliable. While interviewing only one person may save time and cost, it makes a substantial part on a survey absolutely useless. Worse, they make important analysis unreliable. In their example, one cannot establish the determinants of poverty. I wonder how many results in the development literature need to be reexamined.

No comments: