AGEKID2 in Finland 2009 sample

TIME USE
1

Hi all,

I just wanted to flag a potential issue I found with “AGEKID2” in the Finland 2009 sample. In some households, AGEKID2 appears to be taking the age of the younger adult (who is clearly a spouse/partner), rather than the age of the youngest child. For example, in HLDID = “10675”, the household contains only two adults (ages 54 and 59) and no children, but AGEKID2 is coded as 54 (for both individuals). I’m seeing many similar cases where AGEKID2 is non-missing and corresponds to the younger adult in a two-adult household.

Thanks for looking into this!!

2

Another potential issue I’ve found with agekid2 comes up in the France 1999 and Finland 2009 samples. In these data, there are cases where adults are assigned an agekid2 value of –7, even though there are other household members under age 18 present in the household.

This seems incompatible with the IPUMS documentation for agekid2 which states: “If no household member is under 18 and no household member is the child of another household member, this variable takes the value –7.”

For example:

  • France 1999: household ID “2106055270”

  • Finland 2009: household ID “30908”

3

Thanks for your patience in awaiting a reply.

I looked into the issues you described and am sharing information about each of them below.

Universe of NCHILD and AGEKID2 in Finland 2009 sample
You did not ask about this but I noticed it while looking into your questions, and it is useful to know.

The stated universe of NCHILD in all MTUS samples is “all persons.” This variable reports the number of children under 18 living in each person’s household, and does not indicate who the child belongs to. The stated universe of AGEKID2 is all persons with child(ren) in the household. This variable reports the actual age of the youngest child in the household. The youngest child in this case can be anyone under age 18, or an individual age 18+ who is likely the child of another household member. Note that the Finland 2009 sample only includes information on persons age 10 and older. This means that NCHILD may not match the number of persons you observe in the data under age 18 in each household (e.g., if there is a 9-year-old in the household, they will be included in NCHILD, but will not have an associated person record).

It looks like in the Finland 2009 sample, these variables are only available for persons age 20 and older (AGE=22, see AGE comparability section for age codes for this sample). There are a small number of diary records for people under age 20 (22 records) with AGEKID2 not equal to -7. This is also the case for the AGEKID variable.
We will consult our partners at CTUR about whether the universe for this variable in Finland 2009 should be updated or whether there is a correction to be made to the variable.

Unexpected -7 codes in AGEKID2
In the France 1999 sample, I see 249 person records where AGEKID2 = -7 and NCHILD>0.
In the Finland 2009 sample, I see 1,258 diary records where there is at least one person under 18 in the household and AGEKID2 = -7. In 86 percent (1,094) of these cases, the diarist’s age is under 20 (AGE<22), and since those persons do not have values for AGEKID2, this makes sense. There are remaining 164 diary records in which the diarist is at least age 20 (so should be in universe for AGEKID2), and there is at least one person under 18 in the household, and AGEKID2 = -7. Note that in all of these cases, NCHILD=0, but I manually calculated the number of persons under 18 in the household. I suspect that because NCHILD was coded as 0, AGEKID2 was automatically coded as -7.

This is unexpected, but it affects a relatively small number of cases. While we provide some of the harmonization documentation from our partners at CTUR, I cannot say with certainty where or how these errors could have arisen. While these variables should be in agreement with one another (e.g., NCHILD and AGEKID should not contradict each other), I am not surprised to see a small number of cases like these.

Unexpected values of AGEKID2
I do see some instances where AGEKID2 does not make sense, as it points to a person who should not be identified as the “likely child” of someone else in the household based on their age and the other household members’ ages. In the France sample, 0.25 percent of individuals live in a household where the age gap between the oldest household member and the youngest child in the household is less than 15 years. In the Finland sample, there are no cases where AGEKID2 is less than 18 years younger than the oldest household member. I was not able to locate the HLDID 10675 in either sample.

We need to consult our partners at CTUR to identify the source of these issues.