The Daily Mail has published a rather startling story: from 2016, children will sit something akin to the old O-levels. Some parts of the story are relatively uncontroversial: the idea that there should be one exam board in each subject has many friends.

The newspaper also discusses abolishing the National Curriculum for secondary schools. However, if you have a single GCSE available in each subject, that sets a national curriculum in all but name. So these are less interesting than the sum of their parts.

But, if the Mail is correct, there is one proposal which stands out: splitting the GCSE. According to the report, under this new scheme some children would get the new O-level, and the bottom 25 per cent would take “CSEs”. This strikes me as a high-risk policy.

The GCSE’s strength is that it is a full-spectrum exam, measuring low to high ability. It includes questions designed to distinguish candidates that should get a G from candidates than deserve an F, as well as questions to filter A* candidates from those getting an A.

This is also its greatest PR weakness: it gets attacked by people citing the low-level questions. The Mail approvingly notes “questions like ‘Would you look at the Moon with a microscope or a telescope?’ from science GCSEs will be a thing of the past.”

The benefit of this system is that you get comparable qualifications, and there is no need for schools to attempt to sift children, guessing who will finish up with less than a C. The GCSE exams themselves do that work for them. But, according to the Mail:

Mr Gove believes those teenagers have been encouraged to think that a D, E, F or G grade at GCSE is a ‘pass’ when the real world treats those grades as a ‘fail’.

I confess that I do not see how it logically follows that the lower end of the GCSE should therefore be replaced with a CSE. The government would replace a D at GCSE with a certificate where the top grade is capped at a D. Maybe something got lost in the briefing.

The change would, however, have significant practical effects.

The most significant issues around this idea are related to social mobility: the CSE will tend to be an exam for poorer children. Consider who would take the CSE if schools could select the quarter of pupils with the lowest average grades with perfect foresight.

For this graph, I have ranked children by their average grade across all GCSEs (which, I think, is the fairest way to simulate the government’s thinking). I have then worked out the probability of children finishing in the bottom quarter on this measure nationally, split out by the poverty of the neighbourhoods in which they live. I have used the so-called IDACI measure to define poverty – the children are ranked by what proportion of the households near to where they live are in receipt of benefit payments.

At the left hand side, we have the poorest children. You can see they are much more likely to fall into this category than the richest children – at right. The line dips from just under 40 per cent among the very poorest to under 10 per cent for the very richest.

There will be a geographical effect, too, with some areas switching heavily to it. I have marked this map showing what proportion of children in each neighbourhood will finish in the bottom quarter on the same measure. The CSE will be a northern qualification, too. Take a look at the belt from Liverpool to Hull – the CSE towns of tomorrow.

Red areas are places where children have higher-than-average chances of getting results in the bottom quarter, and blue areas are places where they have lower-than-average chances. Dark blue areas do much better, dark red do much worse. If you click on the shapes on the map, you can see each area’s precise statistics.

This matters: one would expect it to lower aspirations among children put onto the CSE track – which would presumably happen at the age of 14. If a child gets moved onto the scheme who otherwise would be aiming for a C at GCSE, they may just stop trying.

This could happen, to an extent, when pupils only sit the so-called “foundation papers” for GCSEs now. But children taking such papers can still get a C at GCSE, and they do not have a qualification on their CV which suggests to employers that teachers thought they were low-ability.

Furthermore, schools do not have perfect foresight. Schools may struggle to place children with accuracy. This is not to belittle teachers: the lines are indistinct. Every child from the 29th to 48th percentile gets an average grade of “D”.

If that were not hard enough, children are moving targets – particularly in the early teenage years. There is a lot of movement between standardised tests. One third of the bottom quarter of children at the age of 11 break out of that grouping by the age of 16.

This uncertainty would, again, weigh on poorer children disproportionately. More than one in five poorer children get results that put them in the 25th to 40th percentiles at GCSE by average grade. They could more easily be dragged into the CSE net by accident than their richer peers.

I look forward to getting more detail on this proposal.

UPDATE 19th August 2012: The map went offline for a spell. Now back.