DfE research recommends weighting output metrics in policy appraisal

It would be awfully embarrassing if research commissioned by the DfE argued for weighted rather than absolute outcome measures, wouldn't it? Well, you'll never guess what David Kernohan has just read.

David Kernohan is an Associate Editor of Wonkhe

So there is, it turns out, a salary premium for individual GCSE grade points.

According to the Department for Education, getting a B rather than a C in Maths at 16 could yield men an extra £32k in salary over their working life. And if you thought LEO for graduates was riven with issues and entirely unsuitable for serious education policy making, this effort is not going to change your mind.

However, a companion release rejoicing in the unpromising title of Schools Policy Appraisal Handbook represents a serious look DfE at policy appraisal using outcomes measures, and as such there’s some lovely advice our stellar cast of current ministers can take into higher education policy. So let’s dive into that, and if you read to the end you can look at some nice charts about GCSEs and salaries.

What’s salary got to do with it?

The policy appraisal handbook is best seen as a supplement to the Treasury Green Book – the manual that defines common ways that costs, benefits, and trade-offs are calculated in policy development, selection, and appraisal. These approaches turn up in things like Bill Impact Assessments – you might recall the excitement caused by the one in the Higher Education (Free Speech) documentation.

Simply put, cost benefit analysis (CBA) takes the costs of a particular policy approach and weighs them up against the “welfare benefits” – how much better it makes people’s lives over time. This is not limited to purely financial benefits, although you will often see a financial value assigned to something that is arguably not linked (or linked very loosely) to a financial benefit.

Education has a huge impact on a person’s welfare and life course – the research on this is clear. But it is not the only thing that has an impact, and we are encouraged in policy appraisal to account for these other determinants. The handbook glosses research that suggests that

the strongest childhood predictor of a satisfying adult life is emotional health in childhood, followed by behavioural outcomes, and only then by cognitive attainment

But it is also noted that robust data is generally only available for some impacts on attainment. Toby Young would be delighted to read a nod at the literature on the impact of genetics on wellbeing, I would think that the impact of background and childhood experience is probably the key one – our authors suggest that it is also important that account for the interdependence between individuals and – indeed – the jealousy between individuals differently performing groups that formed the basis of recent tedious arguments about the white working class.

Weighting tables

Imagine the embarrassment caused by writing the right of a regulator to use “absolute” outcomes measures into actual proposed legislation in the face of advice like:

Weighting could prove transformational in the appraisal of education policy. Given that both intergenerational social and income mobility is relatively low in the UK, it is likely to make a considerable difference to the outcome of any CBA, leaning toward targeted policies for disadvantaged pupils and deprived areas. This is particularly important because we find that the earnings returns to attainment are, on average, smaller for more disadvantaged groups

If a policy (the example given is schools policy, but it is equally applicable to higher education – indeed, getting more people into HE is arguably a schools policy itself) has a uniform effect on attainment using this simple absolute value would suggest it should be applied to all students, which would perhaps be prohibitively expensive. But if we looked at lifetime earnings we could see that the impact of increased attainment on the earnings of students who had been eligible for free school meals is greater than for other groups, making a targeted policy intervention look more attractive.

It’s an economics-heavy way of arriving at a conclusion that makes instinctual sense – disadvantaged groups are disadvantaged, so focusing on these groups aids rather than prevents equality. One for the Higher Education team at DfE to highlight, I think.

The signal and the noise

There’s been some debate – in higher education in particular – as to how much of the individual benefit of a qualification is in the development of new skills, and how much is due to a signalling effect where a qualification indicates non-skill attributes of a person.

Though the general consensus is that this is a distinction that matters little to the individual learner if we assume a salary-maximising Homo Economicus, it is a big deal for policy makers. If the benefit is entirely a signalling one the actual content of the qualification in terms of skills taught is irrelevant. After all, though I very much doubt that Conservative candidate selection boards know what an MA at Oxford entails, the label seems to keep impressing them.

If you are making education policy you need to have a position as to whether attainment indicates the amount of productivity-enhancing impact that a given course has, or whether attainment just tells you how great a person a given student already was. The research-backed example given is an interesting one: many employers and providers require an English GCSE at grades A to C, even though it is clear that the difference in attainment and thus likely productivity between a low C and a high D is very, very small.

Though that’s a pretty clear signalling effect, it’s only one piece of research. The current consensus points in the other direction, meaning that we currently assume a signalling effect so weak that we don’t account for it. That said, the advice is:

This can only be treated as a general rule and concern for signalling effects ought to be addressed in each policy appraisal

So if you are contemplating, say, a minimum grade to access higher education fee loans, you’d probably have some really good research that suggests you aren’t just perpetuating a signalling effect.

Wouldn’t you?

This is where it all went wrong for you

Unless you’ve read the rest of the article, you’re not allowed to play with these charts.

[Full screen]

My example cleaves to a theme that I do tend to push in these salary articles – we really do pay artists badly. Here we see a man getting an A* rather than an A in GCSE music is waving goodbye to just over £20k in lifetime earnings. We’re not given an official interpretation of this phenomenon – but it seems reasonable to expect that outstanding early success may be linked to further study in music followed by a richly rewarding yet poorly remunerated career in the music industry.

To be clear, the standard errors are pretty large and there’s a whole lot of assumptions underpinning the calculations – diligently documented in the research paper, but there clearly wasn’t space in the press release. The model is very similar to the influential IFS examination of lifetime earnings I wrote about last year. We get a look at earnings by qualification level – and I am delighted to see some hourly earnings estimates in here too. I do wish that would that we could see hourly earnings in the main LEO release, as it would be an elegant solution to the part time problem.

[Full screen]

There’s also an estimated impact on lifetime earnings from the number of GCSE you take, that has a fascinating interaction with how you actually did – even though free school meals eligibility has a typically profound impact.

[Full screen]

Leave a Reply