Error patterns in children's age reports in retrospective sample surveys.

S. Mahmud, S. Becker

Research output: Contribution to journalArticlepeer-review

Abstract

Age misreporting distorts age distributions as well as affects fertility and mortality estimates in a population. Reported ages in Bangladesh are often inaccurate because respondents do not know their exact ages; ages must be estimated by the interviewer. This paper attempts to validate reports of children's ages in retrospective sample surveys by comparing them to the actual ages recorded in a vital registration system in Matlab Thana, Bangladesh. In 1 of 2 field surveys in 1980, 2076 women aged 15-50 reported on their 3859 living children. 19% of the children's ages were reported within 1 month of their true age. The ages of 34% were overreported, and the ages of 46% were underreported. Incorrect ages were, on average, off by 14 months. Women less than 30 years old correctly reported the ages of 28% of their living children, twice the percentage of women over30 years old. Women with 3 or fewer live births were more likely to report their children's ages correctly than women with more than 3 live births. In general, as a woman grew older and had more children, not only did she report incorrect ages for a higher proportion of her living children, but the errors were largerin magnitude and mostly negative in sign. Correct ages were reported equally often for male and female children. Overstatement of age was more common for children under 5 years old. It was also found that the quality of age data deteriorated significantly with the progress of field work.

Original languageEnglish (US)
Pages (from-to)123-135
Number of pages13
JournalBangladesh Development Studies
Volume12
Issue number4
StatePublished - Dec 1 1984
Externally publishedYes

ASJC Scopus subject areas

  • Medicine(all)

Fingerprint

Dive into the research topics of 'Error patterns in children's age reports in retrospective sample surveys.'. Together they form a unique fingerprint.

Cite this