Place matters – the Office for Students says so and I agree.
This new briefing accompanies the release of a further cut of the geography, environment, and earnings data (GEE) that I wrote about back in June. If you can remember that far back, you might recall that I had three substantive complaints about the initial release: why are we mucking about with ten year old census data? why use LEO and not Graduate Outcomes data? and why use Travel To Work Areas (TTWAs) rather than smaller areas? I’m gratified to see that the OfS data team has taken steps to address all these issues, but there is a lot of other data of interest here too. So let’s start with that.
For all the talk of rootless cosmopolitan graduates, universities are very closely linked to the space they occupy. It feels fatuous to say it, but as well as the documented contributions to civic society and the local economy most universities literally bear the name of the place they are rooted in. For the majority of providers, the area in which students study is the area in which a plurality of graduates remain to work and live.
Provider data
The improved GEE data includes details of the number of graduates of each provider residing in each travel to work area. This is derived from LEO (combining three years of data for three years after graduation), which has previously provided this data only at (NUTS3) regional level.
This dashboard lets you look at the graduates of an individual provider, using the filter at the top. The map is self explanatory, the chart ranks TTWAs by the number of that providers graduates and you can click on an area on the map or use the highlighter to search for a particular area.
This isn’t exact. As OfS caveats:
For each university or college on the OfS register this table shows the number of graduates, and the proportion living in each TTWA according to the LEO data. Universities and colleges are only included in this table if there were at least 25 graduates in the relevant population. To preserve confidentiality, areas are only shown if at least five graduates from across the three years of data were living there, otherwise those graduates are counted in the ‘Other areas’ section. For two providers, all or nearly all of the graduates were in a single area. To avoid revealing anything about these individuals we modify this data as though some graduates were in ‘Other areas’.
For such a novel dataset, the English regulator makes very little noise about how useful this will be. And as we shall see, it is very useful.
Area data
First up, we get details of the proportion of graduates at each provider that live in each quintile on two measures. The first one is LEO based – the proportion that earn more than a threshold salary (£24,000 – the national median earnings for those aged 25-29 in 2018-19) or are in any further higher education study (not just at a higher level). Because this is LEO we suffer from the usual issues regarding part-time work.
The other area measure relates to the proportion of graduates resident in an area who reported being in highly skilled employment (SOC2020 groups 1-3), further study, or other “positive outcomes” (travelling, retired, or caring for someone) in the Graduate Outcomes (GO) survey. That’s the same definition as is used in Proceed.
You can choose a provider of interest and view this information as a bar chart on this dashboard:
Do not be put off by how simple this graph looks. If – in future – your provider is hauled over the regulatory coals on a B3 (Outputs) related infraction you can use this data to argue that:
- Many or most of your graduates live in areas where graduates get poor salaries or struggle to get high skilled jobs
- And this is good for the local area because levelling up
It may not, of course be good for those individual graduates (the OfS does have a point here), but the new government policy direction around the idea of place means – by implication – that individual outcomes are not the only game in town.
It would be possible – and I’m sure OfS are on this (if not have this one on me Richard) to use this dataset to develop a salary and skilled employment benchmark – given where graduates from a particular provider tend to live, how much would we expect them to earn/how highly skilled should their jobs be. Of course, this would be for a parallel universe where OfS was not mystifyingly committed to absolute outcomes measures.
It gets better
We can see this effect in more detail by cross plotting data on each individual TTWA with the number of graduates from a given provider in that area.
The size of each blob shows the number of graduates from each provider in a given TTWA, the two axes show the proportion of graduates in each TTWA (not the proportion of graduates from each provider within the TTWA) above the earnings threshold and in highly skilled work. The dots, of course, do not move. But flicking between providers shows how the centre of gravity that underpins the bars in the chart above shifts.
I’ve also done two reworks of the top dashboard showing the LEO and the Graduate Outcomes quintiles and underlying data with the number of graduates of each provider living in each TTWA.
What else has changed since last time?
You’ve probably spotted that Graduate Outcomes data on skilled employment has replaced the 2011 census data, which now sits alongside LEO data on salary. I’m glad that the skilled employment measure includes the expanded list of other “positive” outcomes we saw in Proceed – just as 2020 saw government and the regulator move away from salary as a proxy for teaching quality, 2021 is seeing both shift away from the deeply problematic concept of the “highly skilled job” – eventually we’ll all use something similar to the IFS approach to graduate mobility alongside the HESA measures around “fair work”.
I had concerns about the use of travel to work areas as the geographic unit of analysis and an Annex B attempts to address these. The analysis there doesn’t convince – table B1 suggests, for instance, that more than 40 per cent of MSOAs in LEO qunitile 1 are not in TTWA level 1 quintiles. It’s frustrating, because we’ve been round this loop before with POLAR resolution – small areas (as HESA has shown recently) can highlight pockets of deprivation glossed over in lower resolution analysis.
I’ll grudgingly accept that using TTWAs allows us to look at provider level data, which generated some genuinely useful and useable findings. But the arguments in annex B don’t quite stack up.
This bit is really abstract, sorry
Here I’ve plotted the LEO rate against the number of graduates for each MSOA, and coloured using the TTWA quintile.
Fundamentally, these LEO quintiles are only useful in that they characterise an area in which graduates live. The question always is (with this, and with other area-based measures that use quintiles) how well the assignment of one of these five labels reflects the actual experience of people in a given area – too large and you start to elide the existence of pockets of prosperity in deprived areas and of pockets of deprivation in well-off areas, too small and you run into problems of low population sizes.
For instance – the quintile 5 Castlefield and Deansgate area is in Manchester, a travel to work area in quintile 1. Brimsdown Avenue (quintile 1) sits within Enfield (quintile 5). There are many other examples like this with a reasonably sized graduate population – and to providers looking to understand and improve graduate outcomes these distinctions are important.
Other OfS area based measures (POLAR, TUNDRA, adult participation) use the MSOA or LSOA as a unit of analysis. There is clearly not a problem with publishing data at this resolution as it has been done. And OfS made the opposite argument regarding TUNDRA when it published the MSOA and LSOA variants of that measure – it was clear that these pockets of difference in small areas are important and notable.