Post: 19 March 2018
Over the weekend I made a shapefile of Jobcentre locations in Britain.
It’s fine for my own purposes. But I have some questions about the source data.
As a starting point I used the following dataset, released last week, which we will call Register 1:
DWP jobcentre register as of February 2018
The dataset is a table containing addresses of “offices open as of 28 February 2018”, with the office names and a NOMIS code for each.
According to DWP:
This document is a register of the Department for Work and Pensions (DWP) jobcentres.
It has been created as part of the Government Transformation Strategy to build a national data infrastructure of registers (authoritative lists that are held once across government).
(Held once.)
But then we have Register 2:
This is a linked dataset, managed by DWP and available in various file formats under the auspices of the Government Digital Service’s Registers programme. According to GDS, these registers “represent the government’s approved version of that data.”
(Approved version.)
You can probably see where I’m going with this.
Even allowing for the difference in cutoff dates, these are different datasets. Register 2 contains additional fields (some of which I’ve matched and appended into my shapefile). More significantly Register 2 contains more than a dozen records of additional Jobcentre locations, without closure dates, that don’t appear in Register 1. There’s also at least one Jobcentre that appears in Register 1 but not Register 2 (Bordon in Hampshire).
These inconsistencies suggest Register 2 is not being kept up to date. (“Registers are live lists …”)
DWP also maintains an online tool that jobseekers can use to find contact details for local Jobcentres:
Search by postcode and you get back the address of the nearest Jobcentre.
The dataset behind this tool doesn’t seem to conform to either Register 1 or Register 2. For example at the time of writing the search results include a Jobcentre at 25 Station Street, Nottingham NG2 3BZ – a location that appears in neither register.
What was that about a “canonical source of truth”?