Merging waves of data from the British Household Panel Survey

[Harrison Smith]

Dear Forum Users,

I am currently performing a research in applied economics project concerning inter-generational social mobility in the UK. I would like to use the British Household Panel Survey to extend analysis performed by Paterson and Iannelli (2007) (http://www.socresonline.org.uk/12/6/15.html) to more recent waves of data to observe whether policy measures focused upon education successfully addressed the decline in social mobility that occurred during the 20th century.

I have retrieved the individual level data from the BHPS in Stata format from the UK Data Service (http://discover.ukdataservice.ac.uk/...#documentation) in separate files concerning the different waves, 1991-2009, from the survey, with the intention of creating birth cohorts for my analysis

I have familiarised myself with methods of merging and appending data through online tutorials. However, I am encountering difficulty in applying the techniques to the data in Stata and being confident in understanding the data that I dispose of subsequent to performing such techniques.

I would be extremely grateful if anyone were able to provide any insight.

Many thanks,

Harrison

[Marcos Almeida]

Hello Harrison,

I gather you've already typed "help merge" and "help append" in Stata.

If still in doubt, I suggest you present a visual an abridged version of the main variables to be merged.

Best,

Marcos

[Stephen Jenkins]

Harrison: welcome to the Forum. I think your formulation of your problems is too vague for posters to contribute useful assistance. Please read the FAQ for advice about how to solicit help most effectively. Amongst other things, you don't tell us what "online tutorials" you are referring to, and whether they are BHPS-specific.

Whatever, you also need to be very careful about combination of different files in the BHPS. You refer to "individual-level data", but there are 2 types of file for individuals -- "enumerated individuals" (wINDALL files, where "w" is the letter identifying waves) and "adult respondents" (wINDRESP). And some of the important data for individuals refers to their household context, which is held in household-level files (inter alia wHHRESP files). So, when you merge files from the same year, you need to be very careful that you get the data and samples you expect. Once these issues are sorted, then appending the data from different ways is relatively straightforward. Along the way, you also need to be very careful about whether are going to use only the original sample (began 1991, wave 1), or also use data from the later extension samples for Wales, Scotland, and Northern Ireland (which add complications if only because of the absolutely essential need to then use the correct sets of weights)

You might also want to look at a new book by researchers who work at the institute that produced the BHPS: https://www.iser.essex.ac.uk/2014/12...ing-panel-data. It has lots of Stata examples in it. (ISER also runs training courses.) Perhaps also look at John Haisken-DeNew's "PanelWhiz" program as a means of semi-automating the data combination process. (Google for it.)

[Harrison Smith]

Thank you Marcos and Stephen for your helpful comments.

I understand the importance in this instance in understanding the results of merging the data and hence am wary of exercising caution prior to performing the econometric analysis.

I am grateful for resources that you have suggested and hope that they will allow me to address the data appropriately.