27 March 2014

Correlation is not causation

Articles in the media commenting on education tend not to make clear the distinction between correlation and causation.

For example, a recent Daily Mail article reports on a study (by France’s National Institute for Demographic Studies) which was published in the European Sociological Review. The article states:
Comprehensive schools prevent pupils from poor backgrounds achieving their potential, a study has claimed.
Researchers compared reading standards in countries which have retained grammar schools with those which have phased them out, such as the UK.
They found that family wealth played next to no part in a child’s achievements when they were taught according to ability. But a disadvantaged background was more likely to count against youngsters in countries that shun selective education.
Presumably one is meant to think it is a good thing if a child’s achievement shows no correlation with family wealth. However, this would only reflect meritocracy if there were no correlation between parental success and offspring ability, which seems doubtful.

The fact that the presence of selection in a state educational system tends to go with low correlation between parental wealth and offspring achievement does not necessarily mean that selection generates more meritocracy. There may be other reasons, not to do with the absence of grammar schools, why in the UK the clever children of poorer parents do worse than expected.

* * *

William Alfred Green,
father of Celia Green
My father took the grammar school scholarship exam in the early 1900s. The population of Newham, where he lived, was about 400,000 then, which suggests a population of several thousand in his age group (by year). Of those taking the scholarship exam for Newham in the same year as him, twenty passed, and he came top. (My mother, also living in Newham, won a grammar school scholarship too.)

My father’s home circumstances were not propitious. His ostensible parents neglected him, there were few or no books in the house, and he appears not to have been a native English speaker, having come to England from Poland at about the age of eight. In spite of these bad circumstances, he gained the top scholarship.

In his case, parental neglect, lack of books in the home, and attendance at a low-grade primary school (from which he played truant) were associated with success in the scholarship. There were other factors, such as his very high intelligence and drive, but these factors are genetic and thus unlikely to be taken into account when modern ‘experts’ study school and exam performance. Academic studies tend to focus on the family and school environments, presumably because these factors are more amenable to social engineering.

My father’s success at achieving a grammar school place, in a fiercely selective system, was not sufficient to prevent his being handicapped by his unfavourable background. His ambitions were frustrated, and he ended up in the relatively lowly position of state primary school headmaster.

His deprived background and/or his exceptional abiltity were always against him. Very high ability can be enough to arouse hostility and opposition in other people and make life very difficult for the possessor of it.


Another example, from today’s Daily Mail (28 March).

A study (by the Higher Education Funding Council) claims to have found a link between type of school attended and class of degree awarded, with state school students doing better at university than those from private schools with the same A-level grades. Allegedly, this implies that the ability of private school pupils achieving a given level of A-level success is lower, and hence that private schools must be better at ‘pushing’ their pupils – supposedly justifying a ‘contextual’ admission policy, i.e. having a lower entrance requirement for state school students.

But interpreting the correlation in this way assumes that degrees are somehow more reflective of ability than A-levels.

I do not myself see that success in a modern degree course need have much correlation with ability at all. A study carried out on Oxford psychology students some decades ago ostensibly showed that class of degree was negatively correlated with IQ.

My unfunded independent university, which could be publishing analyses of the complex issues involved in the area of educational policy, has been effectively censored and suppressed for decades. Meanwhile, misleading and tendentious material on the topic continues to pour out from socially recognised sources.