I am trying to determine the characteristics of lower income households who transition from renting their dwelling to owning, over time, using the panel aspect of the CPS data.
Would a reasonable approach include downloading several samples over time, using HRHHID 1 and HRHHID2 to identify each household uniquely, isolating which of these made the renting to owning transition, while filtering only for lower income households?
While the combination of HRHHID and HRHHID2 uniquely identifies households within each sample month, they do not link households across their appearances in the CPS panel. For linking households across months, IPUMS provides the linking variable CPSID (CPSIDP and CPSIDV for linking individuals). It is important to note, however, that the CPS samples dwellings, and then interviews members of households within the sampled dwellings. If a family or household moves dwellings, they are not followed by the CPS; the new inhabitants of the sampled dwelling will take their place. The dwelling retains the new household-level linking key, CPSID, linking the new inhabitants to those who resided in the dwelling previously even if they are different individuals. This means that simply looking for changes in tenure at the person or household level over time will not only capture people who purchase the dwellings they live in and formerly rented, but people who purchased and moved into dwellings that another household was previously renting.
To identify individuals in the CPS who transitioned from renting their dwelling to owning that same dwelling during their CPS observation period, you should use CPSIDV in combination with OWNERSHP. CPSIDV is a unique person identifier that links individual respondents over their CPS panel. It is a validated identifier, meaning it only makes links between those records whose SEX and RACE values do not change and whose AGE values change in expected ways over time. Using CPSIDV, you can be sure that you are looking at the same individuals over time. Materials from our 2021 CPS Virtual Summer Data Workshop on linking the CPS may be of use to you. See, for example, this presentation on CPS linking and validation.
OWNERSHP reports whether the household rented or owned its housing unit. Regardless of who within the household owned or rented the housing unit, all members of the household will share the same value of OWNERSHP. Therefore, it is okay to study OWNERSHP at the person level using CPSIDV to link individuals, rather than at the household level.
You cannot capture individuals who transitioned from renting their dwelling to owning and living in another dwelling. I expect the number of individuals who make the transition you are interested in (the former) to be very small. With very small sample sizes, your standard errors will be large, and it may be impossible to infer meaningful information. See this IPUMS Forum thread for a quick discussion of small sample sizes.
Thanks Isabel, as always this is extremely useful. I have a few follow ups…
First, to be more clear: I am interested in studying the characteristics of households who transitioned from renting to owning any dwelling, not necessarily the same dwelling.
Second, I want to confirm I understand what you wrote. Is this correct:
–CPSID does not follow households over time but rather follows dwellings. We may therefore find that the traits of two households with the same CPSID (downloaded using “Longitudinal, 1 year apart” samples) are in fact different individuals.
–By contrast, CPSIDV uses traits like SEX, RACE, and AGE to ensure that individuals from samples 1 year apart with the same CPSIDV are in fact the same people.
Is that correct?
Finally, you wrote, “You cannot capture individuals who transitioned from renting their dwelling to owning and living in another dwelling.” This is concerning because that is the primary group I am interested in (I am interested in all individuals transitioning from renting to owning, and most of these will be renting in one dwelling at T-0 and owning in another dwelling at T-1).
I am a bit confused about why we can’t use CPSIDV to track individuals transitioning from renting to owning (in the same or different dwelling). Is this because the census drops individuals from the rotation group when they move?
To be specific: when downloading linked samples (“Longitudinal, 1 year apart”), shouldn’t an individual, identified with CPSIDV, with OWNERSHP 22 (cash rent) in year 1, who shows up as the same CPSIDV in year 2 with OWNERSHP 10 (owning or buying) represent someone who went from renting to buying, and not necessarily in the same dwelling (but rather, renting in one dwelling and buying/moving into another)?
This is correct: “CPSID does not follow households over time but rather follows dwellings. We may therefore find that the traits of two households with the same CPSID (downloaded using “Longitudinal, 1 year apart” samples) are in fact different individuals. By contrast, CPSIDV uses traits like SEX, RACE, and AGE to ensure that individuals from samples 1 year apart with the same CPSIDV are in fact the same people.”
Because of how the CPS samples dwellings, and not the people living within them, anyone who moves dwellings during their CPS observation period is dropped from the CPS. The people who move into the dwelling after them will be surveyed in their place (if they choose to participate in the survey).
Simply put, you cannot track someone who moves during the CPS to their new dwelling. So, the vast majority of people you are interested in are not identifiable in the CPS, and are in fact not even present in the CPS by nature of the sampling methods. The only renters-to-movers you can identify are those who bought the dwelling they previously rented (and made the purchase during their CPS observation period, which is a pretty narrow window).
Got it! Thanks for the precise and concise reply! Super helpful Isabel. Really appreciate all your help on this topic.
John Voorheis (at the Census Bureau), in this Twitter thread, differentiates between the homeownership rate for household heads (67%) vs all adults. When I run an SDA IPUMS analysis (output attached), I see the “official” rate of ownership of 67%—ie, the rate that John suggests only looks at household heads. But I haven’t specified pernum=1 here, and the columns total 333 mn, ie the entire population.
Is Census treating all non-household heads as having the ownership status of the head of their household? That seems to be the only way to reconcile what I am finding with John’s analysis.
The ACS variable OWNERSHP is a household level variable. The variable indicates if the dwelling the household lives in is owned by anyone in the household. The text from the 2022 ACS questionnaire is:
Is this house, apartment, or mobile home
-Owned by you or someone in this household with a mortgage or loan?
-Owned by you or someone in this household free and clear (without a mortgage or loan)?
-Occupied without payment of rent?
Each household member has the same value of OWNERSHP which refers to the same dwelling. The variable does not indicate which household member(s) are the legal owners of the dwelling. Using OWNERSHP, you can calculate the share of households that own the dwelling they reside in, but you cannot calculate the share of individual people who own a home. Filtering to PERNUM=1 ensures that only one member of each household is counted (in order to not count a dwelling more than once).
Excellent thanks. And so if I want to calculate the ownership rate by person (adult) rather than dwelling, is it reasonable to assume all adult pernum of 3 and higher do not own their own home?
That is not really possible with ACS data, because you cannot identify which member(s) of a household own(s) the home. In the ACS, the householder is, in most cases, the person or one of the people in whose name the home is owned, being bought, or rented. However, if no such person is present in the household, the householder can be any person age 15 and older.
It is not possible to determine, within a household that owns the dwelling they live in, whether only the householder or the householder in addition to another member or members of the household own the home. To provide a few examples, you can imagine how this would cause issues with married and unmarried couples who live together in a home owned solely by the householder and not the other member of the couple, or with an adult living in a home owned by their parents who live elsewhere.
You may want to look into data from the American Housing Survey if the ACS cannot meet your needs in this case.
Voorheis assigned “homeowners” status to the householder and their spouse, and non-homeowner status to any other adult in the household. For you to replicate that assignment, you would need to use the RELATE variable (available in CPS and ACS) to determine who in the household are the householder and, if existing, the spouse of the householder. They would be classified as “homeowners”, and any other adult in the household would be coded as non-homeowner.