Using IPUMS USA 2000 5% sample I am trying to identify whether each worker’s place of work is Los Angeles city of the rest of the LA County. I found that PWPUMA00 variable is very imprecise and aggregated to the extent that it is impossible to work with. For example, PWPUMA00 variable with the value of ‘4890’ belongs to almost the whole LA County. Also, there are 0 observations in the whole 2000 5% sample that would have PWCITY = “3730”, which is the city of Los Angeles. Does that mean that there is absolutely no way to identify the place of work of a person?
Similar to PUMAs, Place of Work PUMAs are defined based on population limits and geographic boundaries. In many cases Place of Work PUMAs map directly onto standard PUMAs of residence. Some Place of Work PUMAs, however, are construct from a group of PUMAs of residence, which is what you are observing in the 2000 5% sample of Los Angeles city. The PWCITY is actually constructed from PWPUMA00 where possible. This means that because Los Angeles city cannot be identified using PWPUMA00, Los Angeles city will not be available in PWCITY.
Is there any particular reason why this is the case only for Los Angeles? There are plenty of other cities which are much smaller, yet identified by PUMAs. But in the case of Los Angeles PUMAs are merged into 1 PWPUMA that covers almost the whole County. According to the definition, PUMAs are the areas with 100,000+ residents, so this is done not because of the confidentiality issue, but due to some other reasons.
Is it only for the year 2000? Is there a way to correctly identify the PWCITY for Los Angeles for any samples other than 2000 5% sample?
Unfortunantely, the Census Bureau does not have very detailed documentation regarding the criteria used to identify Place of Work PUMAs. In the 1990 samples Place of Work PUMAs were nearly identical to PUMAs of residence. However, starting with the 2000 samples, Place of Work PUMAs started representing more aggregations of PUMAs. When the Census Bureau redefined the PUMA boundaries again for the 2010 Decennial (these new 2010 PUMA boundaries are used for ACS samples from 2012 onword, see PUMA note regarding Multi-year files) they more clearly set the requirements for Place of Work PUMAs as being primarily county based. You can see both the 2000 and 2010 Place of Work PUMA boundaries here.
However, Los Angeles city is an identifiable PWCITY in the 1990 1% and 5% samples as well as 1980 samples.