This small study has looked at only one aspect of the reliability of survey data–their repeatability at an interval of 2–6 months with a different interviewer. Even for basic factual data such as date of birth, household composition and number of children there were some errors. Some were coding errors, some punching, some appeared to be reporting errors, but the numbers were small and no discernible patterns emerged.

For other factual data, such as ability to have more children or the use of different methods of contraception, the repeatability seemed to vary with the hardness of the categories. Sterilization was reported consistently, but more loosely defined characteristics, such as difficulty in conceiving or carrying a pregnancy to term, proved relatively unreliable; birth control methods which involved use of an appliance were more consistently reported than withdrawal and the safe period.

If ‘hard’ facts were reported more consistently than ‘soft’ ones it might be thought that attitudes would be reported less consistently than facts. But the dimension of hard versus soft cuts across both factual and opinion data. Fortunately for us, the question about whether they wanted more children appeared relatively hard; only 2% appeared to have changed their minds radically about this. Their memory of their intentions about becoming pregnant on the last occasion was more consistent than their reports about use of contraception around the time of conception.