Differences between fihhmngrs_dv over different data downloads
I was previously working with an older download of US that contained waves 1-7. I recently re-downloaded the data including waves 1-9 and comparing the fihhmngrs_dv variable across the two datasets there are appears to be differences. I tried to find some guidance on the website regarding this, but it mentions that no changes have been applied to the raw data.
Is there any reason why this variable might have changed (i.e. comparing the same household in the same wave across the different downloads of the data).
Apologies if this has been mentioned or asked previously, I have spent some time trying to understand this.
Thank you in advance!
#1 Updated by Rebecca Parsons 5 months ago
- Private changed from Yes to No
- Assignee changed from Stephanie Auty to Rebecca Parsons
- Status changed from New to In Progress
- File 6614_wave1_to_7_revisions_2018.pdf 6614_wave1_to_7_revisions_2018.pdf added
Looking back through the revision documents, changes were made to the income variables in the Wave 8 release, including fihhmngrs_dv. I've attached the pdf. Does this help?
On behalf of User Support
#2 Updated by Cameron Harries 5 months ago
Thank you for the prompt reply.
I had a look through the revision document and it appears the change came in the top coding of the variable. However, even when I use the top coding flag variable (fihhmngrs_tc) there are still differences between incomes in the same household and wave for the fihhmngrs_dv variable.
Thank you again!
#3 Updated by Rebecca Parsons 5 months ago
- % Done changed from 0 to 80
- Status changed from In Progress to Resolved
I've asked our income specialist, and his response is:
The reason is that imputed values get updated at each annual data release and income data contains a lot of imputed values. As the imputes get updated, the values of household income gets updated.
The reason the imputations get updated is because UKHLS uses a longitudinal imputation method that uses information from surrounding waves when filling in missing data. For example, when we get the wave 10 data it will be used to update the wave 9 imputes and so on. So as we get information from new waves, we improve the earlier imputes. Further details of UKHLS income data and imputation can be found here: https://www.understandingsociety.ac.uk/sites/default/files/downloads/working-papers/2019-08.pdf
I hope this helps,