Grade prediction system means the brightest, poorest students can miss out on top university places

By Gill Wyness

With UK tuition fees now among the highest in the world, but benefits from having a degree remaining substantial, choosing the right university has never been more important for young people. The government has tried to make this easier by offering more and more information not just on the university experience but on the quality of the institution and even the potential wage return students could reap.

Despite all these efforts to make the decision about where to apply as informed as possible, one issue remains: students still apply to university based on their predicted rather than actual qualifications. And these predictions are not always accurate.

Using information on university applicants’ actual and predicted grades and their university attended, obtained from the Universities and Colleges Admissions Service (UCAS), I find only 16% of applicants achieved the A-level grades that they were predicted to achieve, based on their best 3 A-levels.

Whilst the majority of predicted grades were within 1-2 points (where 1 point equates to 1 grade, i.e. the difference between AAA=15 points and AAB=14 points, or DDD and DDE), some 30% of grades were more than 2 points out (AAA vs ABB) either way, and 6% of applicants were out to the tune of 5 points – equivalent to a whole A-level at grade A.

This is perhaps not surprising. By this measure, teachers would have to be able to predict all three of a student’s A-level grades correctly. However it is interesting that the vast majority of applicants – 75% – are ‘over-predicted’ – they receive predictions that are higher than the results they actually go on to achieve. Among these, students from state schools and low SES (socioeconomic status) backgrounds are more likely to be over-predicted than those from independent or grammar schools and students from better-off backgrounds. To some extent this reflects ceiling effects. Its statistically less probable to under-predict someone with DDD (where the low SES state school pupils are more likely to be). and likewise there’s a greater chance of being under-predicted at AAA (where the independent school / better off pupils are more likely to be found).

Whilst we can argue that being over-predicted is not an issue, since this gives students more leeway to apply to good universities, this could eventually lead to trouble if they end up receiving offers from and attending universities that they are under-qualified for, and struggle academically.

However, perhaps more worrying is the group of students who are under-predicted. Being under-predicted is a real problem when it comes to applying for university, since students would be discouraged from applying to a course where they would have been accepted, instead applying to lower tariff institutions (and ultimately being penalised in the labour market). Here I find worrying evidence that ‘high ability’ (AAB or more) but disadvantaged students are significantly more likely to have their grades under-predicted than ‘high ability’ students from the most advantaged backgrounds, even after controlling for school type, gender, ethnicity and year. Although this is a small effect, it is nevertheless a worrying finding – it implies that some of our best students may be being misinformed about their likely potential.

What are the consequences of being under-predicted on students’ university choices? My research shows that under-predicted applicants are 10 percentage points more likely than applicants whose grades were accurate or over-predicted to have applied to a university that they are over-qualified for (defined here as where their own A-level achieved score exceeds the average for that university), and are also significantly more likely to be accepted at such a university. This is also true for the most high attaining students. So, if bright but poor students are more likely to be under-predicted, they are also going to be more likely to end up in universities that they are over-qualified for. And that is what my research finds; among high attaining students, it is the most disadvantaged that are more likely to be over-qualified for their university than the most advantaged.

And even if a prediction is only out by 1 grade – AAB vs AAA – this could still make a big difference to students’ university application decisions. This is particularly true if the under-predicted grade is in a subject required by the course. For example, many courses ask for an A in maths. If you are predicted to get a B you probably wont apply.

Somewhat in response to this, UCAS recently devised an amendment to the system whereby students who achieve higher A-levels than they were predicted to, can go through “adjustment” – i.e. choose to take up an alternative offer. My results show that a tiny proportion of qualified students actually availed themselves of this opportunity, suggesting that knowledge of this particular aspect of the UCAS process is still rather limited – or perhaps that the most desirable places are already full by the time students reach this stage.

So, what can we do to remedy this issue and make a fairer system for all students? One way would be to help teachers to make more accurate predictions, perhaps by increasing student testing throughout the A-level period. But surely a better way would be to simply end the system of predicted grades and let students apply to university based on their actual A-level attainment.

This blog was first published on the UCL Institute of Education blog and can be read here

The full report for UCU can be found here

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: