FT data

This graph tells you lots about what is wrong with English state schools:

To create it, we gave every 16-year-old who took GCSEs at a state school in 2010 a point score for their exam performance: 8 points for an A* down to 1 point for a G. We standardised the lot, then divided them up by the poverty of their neighbourhoods and displayed them on the graph. Children in deprived postcodes are at the left of the graph and the richest are at right.

If the school system were able to overcome all of a child’s background, the line would be flat. As you can see, it is not.  If the government is serious about improving social mobility, flattening that curve must be a major priority.

The Department for Education has one high-profile tool which it can use to try and level things out: academy conversion. Bad schools tend to serve poorer children. If these schools can be handed over to new management who will improve results, this would help poorer children. This so-called “sponsor academy” conversion does tend to lead to improved results, as research by Steve Machin and James Vernoit from the LSE has shown.

But it’s important to ponder how much impact a conversion programme like that, which tries to crack the hard core of bad schools, can plausibly have on social mobility.

Here’s a quick and dirty way to gauge it: how much does the curve flatten when you strip out those failing schools? This is a simple way of showing how much impact those weak schools make on the distribution.  The rather startling truth is: not very much.

The government has a “floor target”: it expects that 35 per cent of children at every secondary school should get five A* to C including English and maths. Schools that fail to meet that target (and a few other criteria) are at risk of having their management replaced. That floor target captures about 100 schools out of 3,000.

If you imagine those schools away, the line changes a little like this:

So it moves a little at the bottom. But, that horrific gradient is still there. Okay, so maybe the bar isn’t high enough? What if you lift it to 40 per cent? This means you switch out about 200 schools.

What about 45 per cent? This drops almost 400 schools from the sample.

What about 50 per cent? This cuts out about 470 schools.

Consider that red line: it is, in effect, a guess at what the school system would look like if children at the bottom sixth of secondary schools were instantaneously dispersed into the rest of the school system in a way that did not damage the performance of those other schools.

How big a difference is that? We already have a measure for that. We can describe how important a factor poverty is under different scenarios using a simple measure. We look at the β in a regression that looks like this.

Academic percentile =  cons + β*(poverty percentile) + ε

If you drop schools below the 35 per cent line, the coefficient falls by 0.007. Up at the 50 per cent line, it falls by 0.021. By comparison, between 2005 and 2010, that coefficient linking poverty and academic performance declined by 0.036. (There’s more on what these numbers mean over here.)

That’s not nothing, but it’s not big. Indeed, given that in this calculation we are instantaneously erasing the worst sixth of all secondary schools, that’s a pretty weak effect.  So what does that all suggest?

The killer problem for social mobility is not that there are a few schools which have all the poor children in them (though that is a factor), it is that poorer children tend to do badly even when they go to good schools.

Sort out the bad schools, by all means, but social mobility is much tougher than that.

The previous graphs showed pupil-level performance, the next one shows school level performance. Lower-performing schools at left, through to stronger schools at right. The dark blue line shows the average point score for pupils in each percentile of schools. As you move from left to right, schools get ever-better results.

The pink line shows the same number, but only for poorer children within those schools (those living in the bottom fifth of households, as measured by the deprivation of the postcode). As the blue line climbs up rapidly, the pink one remains distressingly flat, only crossing the national average line at around the 85th percentile.

This is the graph that ought to haunt the dreams of every school reformer. The social mobility problem is not that there is a small number of weak schools serving a lot of poor kids. It is that poor children do badly in the majority of England’s schools.