What’s up with the allocation item for earnings in the March CPS? It drops from a reasonable 22% to around 2% between 1987 and 1988. I’m using data from IPUMS-CPS (qincwage). This does not jibe with the Table 16-4 in the CPS Technical Documentation (pg 16-5).
This is a good data mystery. Here’s a link to the values on the allocation item for 10 years of data (1984-1993) for those who don’t have access to your underlying data.
First, the imputed values in the CPS Technical documentation are based on the outgoing rotation files, not the March CPS. And, the two files use different imputation methods [see Footnote 5 from this link]. But, the March values should still be reasonably close to the published figures based on the Outgoing Rotation files, not to mention the mystery of the drastic drop in the percent allocated between 1987 and 1988.
The explanation for this is that the CPS changed its data collection process in 1988 as well as its edits:
There are two wage allocation flags. One is hidden on the IPUMS website as it is on the second page of the income/wage choices (QINCLONG, which goes with the item INCLONGJ). And, then’s there’s an allocation flag that IPUMS does not make available at all – a whole-case imputation flag, (FL-665).
Here is some code that shows how to calculate imputation, which I got from the Census Bureau. The variable names are not IPUMS-harmonized names, but instead match the original March CPS names. This makes sense as for now, you have to go to another source to get the FL-665 item. I suspect IPUMS will include the new whole-case imputation flag in their next release of the March CPS data. They were probably not aware of it. Here are two choices for computing code: