Beyond LEO information at a (NUTS1) regional level, we’ve never really had a good handle on graduate opportunities and work in local areas.
I had a look at some elements of this a while ago, focusing on the industries and locations in which graduates find themselves working and drawing on some institutional level Graduate Outcomes data. I, too, struggled to find a way to get a full understanding of what was going on.
A new experimental release from the Office for Students fills in another corner of this largely blank map – looking particularly at the skill level of jobs in a small area, and the difference between graduate salaries and a (national) median salary. “A Geography of Employment and Earnings” (let’s call it “GEE”, for headline pun purposes) examines salary data from LEO against skilled employment data from the 2011 Census, for Travel to Work Areas (TTWAs).
Head, meet desk
Now, some of these data design choices are a little puzzling. There are, to be clear, good reasons in the documentation covering why these decisions have been made – and I’m all in favour of expediency and data reuse. But here are my concerns:
- Employment data from LEO? We know that ethnicity, socio-economic background, gender (particularly), subject of study, and provider of study have an impact on graduate destinations alongside region of domicile, and that any robust use of LEO needs to control for this. We are also painfully aware that LEO utterly collapses when it comes to salary data for part-time employment – to the extent that LEO salary data is largely for entertainment purposes only. I would hope in future that Graduate Outcomes data can fill this role.
- The 2011 Census was a decade ago. It’s good data in terms of being, literally, census (rather than survey) data but it does give an outdated picture of regional employment and it doesn’t separate out graduates from non-graduates. In future (possibly starting this summer) new iterations of GEE will use regional graduate employment data from (you guessed it) Graduate Outcomes.
- Travel to Work Areas were chosen to reflect likely employment location based on residence, which makes a little bit of sense – but there is a distinct paucity of data available at this particular resolution, and as these areas are built based on Lower Super Output Areas there is not a lot of interesting data from constituent areas (which tends to be available at MSOA level or LTLA/UTLA level) to bring in to gain a better understanding. This choice probably also determined the use of 2011 Census data, rather than – say – any of the ONS’ labour surveys.
What we can see
As we already broadly knew that things were better for graduates in London and the home counties, and not so good in Wales (I’ll admit I didn’t see Aberdeen coming) we would need to ad a new level of data to derive new insights. Though I can’t bring institution level data directly into the picture, I have added the physical location of campus delivering higher education (from the Unistats) to the basic plot provided by the OfS. The area colours show the “LEO quintile”, which just splits the proportion of graduates earning more than the national median salary or studying at a higher level into 5 buckets; the provider colours show mission groups.
In doing this, we can start to get an appreciation of what impact having a higher education campus in a travel to work area might be having on graduate earnings and activity, or the proportion of the wider population in skilled work. As you might expect, there is almost no correlation between these two measures, but TTWAs with no campuses in (coloured red here) tend to do less well on both. Or maybe have more graduates working part-time, because LEO.
If we’re prepared to accept that many graduates are likely to be in the TTWA that contains the campus they studied at three years after graduation, we can start looking at provider-level impacts of a sort (if you ignore London, which is a good rule for life if I’m honest). This plot shows each campus against the two measures – you’ll see (even though there is a scatter applied so the figure shown isn’t completely correct) a much greater correlation between the two – suggesting to me that the region in which a provider has a campus has a large (and significant) impact on graduate outcomes. Not a great advertisement for the use of absolute outcomes data in regulation.
What a TTWA
This is where, with a better area definition, things could get very interesting. There is a great deal of good data (say by local authority area, or – banter intensifies – parliamentary constituency) that we could look at in combination with the fine-grained Graduate Outcomes data to understand what is going on. We could compare with (recent) median salaries in each area, occupations, industries, as well as with referendum and general election voting patterns.
So, although the ostensible rationale for releasing this data is that people are interested in local area graduate outcomes, this release doesn’t exactly sate that interest. We are left with many more questions that we would love to answer. And we’d love to factor in part-time employment patterns while doing so.
Fundamentally, stuff like this starts getting really useful when we have some idea of the location of graduates from particular providers (perhaps in particular subjects). That’s not a dataset we have at this resolution, but if this release points the way to something like that I’m all in favour.