HRHHID in the CPS

CPS
1

Hello, I have a few questions about the HRHHID variable :

  1. Is the HRHHID of the household in the March Basic sample identical to the HRHHID of the same household in the ASEC sample?
  2. Is it fair to assume that any values of HRHHID in the ASEC sample which are not in the March Basic sample correspond to the oversampled households?
  3. Is it fair to assume that households in the March Basic sample should also exist in the ASEC sample, and therefore all values of HRHHID in the March Basic sample can be matched to the same value of HRHHID in the ASEC sample?

Thank you!

2

I will address each question one at a time.

(1) Yes, the HRHHID variable in the March Basic sample should be identical to the HRHHID of the same household in the ASEC sample.

(2) Yes, households in the ASEC sample but not in the March Basic sample are part of the ASEC oversample. These “oversample” households can be identified with the ASECOVERH variable.

(3) Yes, in principle this is correct. However, depending on the sample year, this can be complicated. You may find the discussion in this paper we have on linking the March basic sample to the ASEC sample useful.

3

Hi Jeff,

Thanks for your response. I tried matching households in the CPS Basic Monthly (March) to the CPS ASEC in 1983-88 using HRHHID and HUHHNUM (since MARBASECID is not available before 1989). I selected one observation from each household and tried to merge using these two variables. I had a few issues in the process:

  1. A number of observations in the basic monthly March sample do not have a counterpart in the ASEC sample. This doesn’t make sense because every household in the basic monthly March sample should be in the ASEC sample as well.

  2. For the households which did match (based on HRHHID and HUHHNUM), I tried to compare the variable FAMINC which is recorded in the basic monthly sample and the ASEC sample. Between 1983-87 there was a large share of cases where the value of the variable for the same household was quite different.

I’m not sure what the issue is, so I’d appreciate any advice you have to offer. Thanks!

4

I’ll try to address each question one at a time.

(1) Although you are correct that, in theory, all observations in the March basic monthly sample should appear in the ASEC sample, in practice, there are a number of complications that limit this perfect 1:1 match. These complications are some of the reasons why the IPUMS CPS Team has yet to release the MARBASECIDP variable for pre-1989 samples. From previous investigations we know you’ll need to eliminate some duplicates and use demographic variables to ensure the records are being correctly matched. The attached article by Madrigan and Lefgren will probably be a helpful resource.

(2) Note that although the variable FAMINC is included in both the ASEC and basic monthly samples, the response categories are often different when this variable is included in the ASEC compared to when it is included in the basic monthly samples. This inconsistency is likely one major source of the discrepancies you are seeing.