An evaluation of the Fiji Fertility Survey based on the postenumeration survey.

Type Working Paper - World Fertility Survey Occasional Papers No. 2
Title An evaluation of the Fiji Fertility Survey based on the postenumeration survey.
Publication (Day/Month/Year) 1979
The reliability of the responses obtained in the Fiji World Fertility Survey of 1974 was evaluated by conducting a post-enumeration survey, 7 weeks after the initial survey, among a 10% subsample of the original surveyed population. This analytic report sought to assess the value of post-enumeration surveys as an evaluation technique. The post-enumeration survey proved to be an effective evaluation tool. The technique permitted the investigator to identify sources of sampling and non-sampling error and to compare the magnitude of these errors. These identified sources of error provided suggestions for improving survey procedures in the future. The degree to which the subsample was representative of the original survey sample was determined. Although there were differences, the subsample, in general, was representative of the original surveyed population. The responses obtained in the original and in the post-enumeration survey were compared with each other by calculating net and gross errors for a large number of characteristics investigated in the survey. The analysis compared the responses given by the same respondent at 2 points in time. The magnitude of errors attributable to events which occurred between interviews were examined and found to be negligible except in regard to items dealing with contraceptive knowledge. The questions asked in the initial survey affected interviewee knowledge about contraception in the later survey. After adjusting for these extraneous sources of error, the magnitude of sampling and non-sampling errors was determined. The magnitude of the sampling error was in most cases small. The magnitude of the non-sampling errors was greatest in regard to knowledge about contraception. Responses concerning ever use of contraceptives had a greater magnitude of gross error than responses concerning current usage. In general, responses which required factual information were less subject to error than responses dealing with the past or with the future. Included were 1) the methods and formulas used to calculate gross and net errors and 2) tables which provided gross and net error measures for a large number of data elements. Only analytic results were provided. Descriptive results were reported elsewhere.

Related studies