Is age distribution of mothers of children of certain age invalid as result of recoding for privacy?

USA
1

I want to use maternal age distribution by race and by year to predict number of children with Down syndrome born (absent terminations), as there is a direct link between maternal age and chance on a child with DS. I would think that the age distribution of mothers with children of 0 year of age in 1970, gives an indication of maternal age distribution at birth in 1970; age distribution of mother of children of 1 year of age in 1970 would then yield maternal age distribution (at birth) in 1969; 2 yrs of age for 1968, etc. However, maternal age distribution in the USA census IPUM-data constructed in this way does not match the information in the USA natality files. Well it does more or less for 1970 and 1969 (but seems to be shifted 1.5 years), but start to be reaaly different if you go down to 1968, 1967,…1961. Why? Is it a result of recoding ages for privacy reasons? Cause if that is so, I just should not use the IPUM data for this topic.

So, the question to IPUM is: has there been such a recoding of age that maternal age distribution are not very precise?

2

To be more precise, a mother in the 1970 Census with a child of less than 1 year old could have given birth in either 1970 or 1969. Similarly, a mother with a child of 1 year of age could have given birth in either 1969 or 1968. You would presumably get a more accurate measure of the maternal age distribution by directly using the child’s birth year (BIRTHYR) and attaching the mother’s age to the child’s observation. Unfortunately, BIRTHYR will not be accurate for samples after 1980 or before 1950 (see here for more details).

Age was self-reported by the survey respondents, and I am unaware of any recoding. It is quite possible that some respondents rounded or estimated their age. This could lead to some discrepancy when comparing to a different data source.

Hope this helps.

3

Staff, thank you for your answer. Very helpful. I should have used momrule and attached age to mother, as you suggest. My error was to use age of youngest child ( in stead of any child). That introduced a systematic bias.