The skills problem may be harder and more complex than we thought

While the rest of you were reading the Sue Grey report, a long-awaited set of government publications slipped by unnoticed.

If you haven’t clocked a new government seriousness about skills provision, you must have been at quite some party. But the rhetoric about higher technical skills qualifications, local skills improvement plans, and the impacts of upskilling on levelling up that has become so familiar has seldom been backed up by anything like what we might call research evidence.

The reasons for this are manifold – we’ve never really collected data on skills as opposed to employment, or skills as opposed to qualifications. And that’s just the supply (what skills do people have) data – on the demand side, things are even worse. When you mix in local issues at the kind of granularity you need to really understand where areas of deprivation are – not all the North West is like Manchester, not all of Manchester is like Bury, not all of Bury is like Ramsbottom – and here we see a clear data and evidence gap.

A work event

Between 2020 and 2022 the Skills and Productivity Board worked on this issue – addressing three key questions posed by none other than Gavin Williamson:

• Which areas of the economy face the most significant skills mismatches or present growing areas of skills need?
• Can the board identify the changing skills needs of several priority areas within the economy over the next 5-10 years?
• How can skills and the skills system promote productivity growth in areas of the country that are poorer performing economically?

That Board has now published all but one of a set of reports (work on changing skills needs within a sample of job roles will follow in June), before handing the baton to the new and shiny Unit for Future Skills – which also chose yesterday to drop some reports and data dashboard. This new one has a role closer to innovation than the Board’s purer research focus, and is working on:

• building up the data infrastructure to create new links between datasets to identify how skills are used in the economy
• conducting analysis and present data in new ways to address data gaps and improve our evidence-base
• becoming a centre of expertise on future skills, developing robust methodology and insights on current and future skills needs

So let’s start with that one.

Would you look at the size of this absolute unit

We had a peep last week at a new LEO (Longitudinal Educational Outcomes) release that included both subject of study and the industry in which graduates work. It was pretty awesome, but they’ve only gone and improved it – covering every exit qualification held by people working in a given industrial sub-sector aged between 25 and 30 (in 2018-19). It’s a genuinely astonishing level of detail (though without personal characteristics or institution of study it isn’t complete – although there is regional data (at NUTS3 level) available), and even though usual LEO rules apply (part time is not poorly paid full time) there is a lot to learn.

Here’s median salary by final qualification, showing the number of employees in each main industrial sector by final qualification.

[Full screen]

It’s a lot to take in (I’ve not added the broad region component as I don’t think it adds a lot here – this is England-wide) so this is a version that lets you drill into particular industrial subsectors using the filters at the top.

[Full screen]

You’ll see in both that higher level qualifications generally, but not always, lead to higher pay within a given industry – there are exceptions but these tend to be in very small groups of employees.

There’s even data letting you track qualification journeys – choose a qualification and see what qualification people tended to do next before working in a particular industrial sector!

[Full screen]

Skills demands and horizon scanning

The Unit for Future Skills has also released a report conducted in partnership with Rand Europe, examining skills needs in a range of future scenarios. A series of qualitative research interventions generated five scenarios that feel a bit more like trends:

• digital greening
• living locally
• protectionist slowdown
• continued disparity
• generating generalists

The policy prescriptions are fairly vanilla – clearly, we need more STEM (but when you drill in this is mainly ICT/digital skills), there’s work to be done on flexible pathways for learners and micro-credentials, the role of local stakeholders, the need for foundational skills (a theme as we will see), and – most notably – a renewed emphasis on the need for employee training.

That’s the general sector stuff – but it is recognised that:

there may be a need for additional specific policies within the above broad framework to address skills shortages due to migration in health care and construction, for example, or because of the way in which technology is implemented or netzero ambitions develop in the transport and energy sectors. This could also have implications for the skills needs and policies in the higher education sector both in terms of research and delivery

That’s both a useful pointer to future research and a reminder that – despite some of the more outre DfE messaging – universities can be and need to be sitting squarely in the middle of this revitalised skills agenda.

Board meeting

The Unit for Future Skills explicitly builds on the work of the Skills and Productivity Board, and we get a whole range of delights from that source as well.

We learn locality (in the fullest and most precise sense) is a huge deal in describing variations in productivity, and skills interventions alone are not enough to close these gaps.

The key message is that all this stuff needs to be joined up: investment in skills, infrastructure, and leadership capacity need to go hand in hand to have any chance to make a difference – and even then nothing may change overnight, or even on usual political timescales.

But first up, there’s some stuff about data.

Improving labour market information

Data is the guts of this problem – we are very short of high-quality data on skills. The supply-side stuff is very limited – we do get various releases (LEO) and collections (the Census, Annual Survey of Hours and Earnings) – and although moving industrial sectors into LEO has helped link areas of work to specific education experiences (and surely occupations will follow) has helped we’re still looking at actual skills through a prism of what jobs people are doing rather than what skills they are using.

Demand-side data is even worse – we have very little systematic data on what employers are looking for (there is an Employer Skills Survey but it doesn’t really get into a lot of detail). In terms of recommendations, there’s not much hope here – a combination of expert-led predictions within priority occupations, and wider trends-based (economic, demographic, even policy-driven trends) projections are about the best we will get.

The onset of Local Skills Improvement Plans means that this data will need to be available at a local (even hyper-local) resolution, and be easy enough to make sense of for non-specialists to make reasonable sense of. There’s some talk of new sources of data (web scraping of job listings, no less) but this is very heavily caveated – obviously there are selection, omission, and skills inflation risks so we’d need to calibrate it with other more sensible measures.

One of the key recommendations for this paper is on the need for a skills taxonomy – a standardised list of skills and definitions that we can use to plot the skills needed for emerging jobs (or new skills within existing jobs) and be sure we are all talking about the same thing. It’s all going to be qualitative to start with, and I suspect very expensive.

Skills taxonomies

Happily DfE has paid Frontier Economics to get this sorted – and their recommendation is fairly simple: there’s a US taxonomy called O*NET and we should use that for most of the things we want to do. O*NET is developed by the US Department of Labor and covers skills in broad, digestible chunks. Where we need finer detail we can use the European Skills Competences Qualifications and Occupations (ESCO) taxonomy or there’s one Nesta knocked up.

The big job that needs to be done here is mapping all of these to the data we do have (on occupations) and to UK qualifications frameworks – this mapping will need monitoring and updating too. In parallel, ONS is extending its Standard Occupational Classification (SOCs) which will result in some additional fine detail (Sub Unit Groups, SUGs, named purely to allow me to make jokes about it being “madness”) to map as well.

The thing that always bugs me about the graduate job data collections is that things are graduate jobs because lots of graduates did them last time we updated the coding – there is a parallel risk that jobs that “need” particular skills may just end up being done by some people that have those skills, or that required qualifications for professional roles turn out not to teach some of the “required” skills. Again vacancy scraping rears its head as another potential source of data – again it is dismissed because of the flaws enumerated above.

As the report notes:

A sensible starting point would be to expand gradually on some of the very high-level skills groupings that have been tested to date (e.g. cognitive, physical, interpersonal) and establish whether these give plausible results

We are – in other words – at the very beginnings of even being able to define the things we are talking about.

Current and future skills needs

What would that SOCs/O*NET mapping look like right now, and what would it tell us about skills needs now and in the future?

Well:

The analysis identifies a set of ‘core transferable skills’ that are currently in high demand across many occupations, including in the priority areas, and are likely to continue being in high demand in the future. These include communication skills, digital and data skills, application of knowledge skills, people skills, and mental processes.

This does feel a lot like the kind of things universities talk about for right now, but what about the future:

Skills that are growing in importance and used across many occupations in the economy include people skills, mental processes and application of knowledge skills, and skills associated with being able to teach others and be a good learner. Skills that are growing in importance, even though they are used in relatively fewer occupations, include STEM knowledge (particularly relevant for Health and Science and Technology occupations, and already likely to be in shortage now), care skills, important for Health occupations, and a range of management skills.

The methodology here is simply mapping O*NET and SOCs, and then using demand for particular occupations as a proxy for the skills needed intensively in those roles. It’s inferences, in other words, all the way down – the future perspective comes from another projection exercise performed by the University of Warwick for DfE that extends six whole years into the future on the basis of current trends. The last one started from 2017 data, so you can all make your own sarcastic comments about pandemics here.

Even assuming all this works, we don’t know if the skills needs identified are linked to undersupply (not many people have them) or underutilisation (not many people want to do those particular jobs). For this reason, there is a lot of hedging about what we can do with these findings.

And here are the findings – a current ranking of skills along the Y axis, and a future ranking along the X axis. The size of the dots shows the likelihood of future increased demand – you’ll note mental health/wellbeing related skills lower down the ranking are projected to grow as work becomes even more terrible in future.

[Full screen]

Skills, places, and productivity

Three research reports tackle the issues more closely related to the “levelling up” agenda.

Left behind/levelling up

There is no single way to fix the problems an area has, says report author Kenneth Mayhew – each deprived local area (to burlesque Tolstoy) is deprived in its own way, and we have to look at very small areas to see what is going on.

We are introduced to three general types of “lagged locality”:

• Long term disadvantage (places that have not had a healthy supply of good jobs in living memory… the example given is parts of Cornwall)
• Secular decline of former dominant sectors (parts of the North West)
• Macro-economic shock (places where an “anchor” employer has recently closed)

In contrast, successful areas tend to have excellent transport links, to already attract the brightest and best, and to have good regional agglomeration of supply chains and industry processes.

Again, just addressing skills won’t fix everything – there needs to be steady work on scaling up and improving local businesses (particularly in terms of the “good job” measures around job quality, design, and organisation of work), interventions to improve teaching (especially early years provision) and better careers support.

Productivity in local areas

Skills demand is derived from business needs, so even if we do invest in skills provision local businesses need to be able to absorb a better-qualified workforce. We need, in other words, to be thinking about business improvement needs as well. Some areas have what report author Ewart Keep calls “low skills equilibrium” – few industries locally need advanced skills, few of the local workforce have those skills, and so the cycle continues.

If you are thinking “high tech clusters” here the story is a mixed one – clusters like this can be useful, but tend to generate a few high-skilled jobs and many more low paid ones that do nothing to level and area enough. Skilled workers will arrive, and will need childcare, retail, and hospitality to meet their needs – which is often what the locals end up having to do.

On a higher education note this report recognises that there is a lot of great regional partnerships led or fostered by universities, and that DfE needs to get a lot better at understanding where these are and how they work.

LEO analysis

We all know LEO – but this is a tantalising glimpse of micro-local data for graduates. We learn that the wage differential for higher levels of qualifications tends to be similar in magnitude all over the country (graduates will always earn more on average than those with L3 qualifications, who will earn more than those with L2 qualifications. However, higher-level qualifications make you more mobile:

Individuals from across the education spectrum – at all qualification levels – are similarly likely to move across local neighbourhoods (MSOAs), but those with higher education degrees are far more likely than those with lower level qualifications to move across labour markets (TTWAs). This may be because, even by the age of 27, there is relatively little economic benefit to moving for those without a higher education degree

Non-graduates, it seems, would also move across labour markets if there was any point in doing so. Which is a shame, because this study suggests that a third of cross area variation in earnings is explained by the characteristics of the area, rather than particular places.

What we have learned

It turns out that levelling up is hard. It takes more than investment in a local FE college (as we saw yesterday via an Open University intervention) to turn an area round – and even when the analysis is focused on skills it becomes clear that there are numerous interlocking parts to the plight of a struggling area. We can take heart that we know how to do some of this – even though it is hard. Upskilling businesses feels very much in the university sector wheelhouse, while local infrastructure and local education feels like something local and national government could get stuck into.

The question really is political will. Why stick all this stuff calling for systematic, long term, (multi-parliament) action with an (albeit very welcome) tiny contribution to local skills provision. The hard stuff is getting the data flowing, and using data to construct meaning. Let us hope against hope for at least one Johnson administration legacy that is not a picture of some bottles on a table in No 10.