Selective systems are precisely what “harm opportunities for other people’s children” – a further comment rejecting the HEPI report on selective schooling

Posted in: Data, politics and policy, Education, Evidence and policymaking, Welfare and social security

Dr Lindsey Macmillan is an Associate Professor (Reader) of Economics at UCL Institute of Education, Dr Matt Dickson is a Reader in Public Policy at Institute for Policy Research, University of Bath, and Professor Simon Burgess is a Professor of Economics at University of Bristol.

We appreciate the response by Iain Mansfield to the widespread criticism of his paper on selective schooling. However, the points we made about the dataset used and the methods employed remain.

A major critique that has yet to be answered is the inappropriate comparisons made when analysing progression to Higher Education (HE). The key part of the argument about the effectiveness of selective schools is hinged on analysis that is far too simple to support the strong statements made. Mansfield returns to the 39% vs 23% rates of progression from selective compared to non-selective areas in his response. The fact that he again attributes these large differences in progression rates directly to the schooling systems, rather than other factors involved that muddy the waters, is a basic stats mistake. The comparison group of all non-selective areas is wrong. If instead the non-selective group of areas were chosen with similar characteristics to the selective areas, the progression rate would be much closer to the 39% rate for selective areas (we will soon be publishing a working paper that shows this is the case). The point is that the Higher Education Policy Institute (HEPI) report attributes all of the difference in selective HE participation rates to the schooling system, while ignoring completely the very significant basic differences between people living in different areas (which also drive different participation rates). Correlation is not causation.

When you make the statistically-right comparison, the progression rates are very similar between the two types of schooling system. There are non-selective areas that are just as good as selective areas at getting this type of progression. Yes, they have particular characteristics, but that is very much the point. It’s those characteristics, rather than the systems per se driving this high participation rate.

The point on the data is that the sample is far from representative – it is missing a lot of high earners. This is illustrated by 70% of the sample falling below the national median figure. Even though median income is defined within sample for the analysis, the fact that the dataset is missing a whole chunk of self-employment data makes any stat about the income profile misleading. Consider the case where one parent runs a company and the other makes a small sum of money in a part-time job. The family will be defined as low income in this data based on the earnings of the parent who makes a small sum of money, rather than reflecting the full sum of income available from the CEO parent. The report uses the figure that 45% of grammar school pupils come from households with below median income to conclude that grammars increase social mobility – it’s mentioned a number of times in that way – but this is simply misleading and unreliable given the significant issues with the dataset.

Finally, the emotional argument about “not holding other children back” shows an astonishing lack of perspective on a selective system. It completely ignores the 80% of children that don't go to grammars, many of whom are dramatically “held back”. What if the child of the warehouse supervisor, the immigrant parent or the shop assistant is one of the majority who miss out on a grammar school (as is far more likely to be the case based on more reliable analysis of the backgrounds of grammar school pupils)? As has been shown before, those who live in selective areas but don't get into the grammar school suffer significant penalties in terms of education and wages.

Emotional arguments might appeal to the masses but they are not a good basis for making policy decisions. That should be based on evidence from sound analysis of high-quality data. The real reason why academics are united against grammar schools is that the evidence that they damage social mobility is robust and conclusive.

Taking all these points together, this discussion continues to be a distraction from genuinely improving the life chances of children from disadvantaged backgrounds. Anyone who seriously wants to do that knows that grammars are not the answer.

Posted in: Data, politics and policy, Education, Evidence and policymaking, Welfare and social security


  • (we won't publish this)

Write a response

  • An even bigger issue is that he is using DfE KS5 Destinations Data for school sixth forms only.

    He therefore ignores 35% of selective university entrants from non-selective LAs who go to a comprehensive age 11-16 and then an FE College or 6th Form College 16-18.

    All in all, this means his data excludes 2/3 of 18-year-olds.

    He has admitted/claimed that this was a deliberate methodological choice on the blogpost you link to.

    And the of course you’ve got the fact 25% of Grammar students come from non-selective LAs, 12% are privately educated up to age 11 and so on.

    I’m shocked that HEPI published this with such shoddy peer review to critique this policy-based evidence-making

  • They wouldn't ever be able to keep an eye on the performances of all the so-called teams
    playing different disciplines. This could be the fourth
    stop on the UKIPT season and there has become
    a lots of action already. However the strength of one's hand - specifically your two hole cards - determine to a large extent what
    to play, things to fold since the hand progresses and when you try to
    create your opponents think you have better cards than you really do.