It is a rare thing for an international study to provide conclusive evidence on what education systems work best. So we should all sit up and listen when such a study shows unequivocally that comprehensive school systems have produced narrower social differences than selective systems.

A recent analysis of the Organisation for Economic Co-operation and Development’s (OECD) PISA survey of knowledge and skills of 15-year-olds across industrialised countries has come to this conclusion. More precisely, it looked at how secondary schools “differentiate” students, such as the age at which they are first divided into separate schools or programmes, the number of such “tracks” and the extent to which weaker students repeat grades. It found a pronounced association between the extent of such differentiation and the average gap in reading performance between students from more and less privileged backgrounds, on an internationally comparable scale.

The results show that the gap between the least and most socially advantaged students at the end of schooling tends to be wider than average in the countries that still divide pupils early in secondary education. Nor is greater equality in comprehensive schools achieved by levelling down. Countries that divide students also have, on average, lower student performance. Countries such as Finland, Japan and Korea all have comprehensives: they combine high achievement with small social differences. Others, such as Germany and Hungary, have below-average performance and wide social differences. They separate pupils at the start of secondary education.

Our own comprehensives have produced students who, overall, have higher reading literacy than average for OECD countries. However, social differences in performance in the UK remain relatively high, as they do in the US, Australia and New Zealand.

This does not mean that our comprehensives have high inequalities. Ignoring the effect of class, the amount of variation in student results is about average for the OECD - pretty impressive progress when you consider our history of dividing students.

However, those differences that persist are more closely linked to social inequalities than in other comprehensive systems such as Scandinavia’s. British students from less advantaged backgrounds are more likely to go to worse-performing schools, and also underperform within each school. Yet this should not lead us to conclude that our comprehensives have failed. The PISA results show just how wrong it is to assume that working-class children in countries such as Germany are more likely to achieve their potential because, if bright, they go to superior academic track-schools or, if less academic, they thrive in a different kind of school adapted to their needs.

PISA assessed whether, near the end of compulsory education, students had the basic reading skills needed to do well in adult life. The quarter of German students with the least favourable social backgrounds were on average only just proficient at level 2 on a five-level literacy scale: this was the third worst result in 27 OECD countries - just below Portugal. In the UK, the least-advantaged quarter had a very respectable average score, just within level 3, which was the average proficiency of all students in Germany.

In the face of such evidence, it would be hard to argue that dividing students at an early age into programmes suited to their needs is the way to reinvigorate our secondaries. When the Government talks about specialisation and diversity, might it be leading us back in this direction?

he idea that different schools should be able to develop different specialisms and characters is intricately tied up with the idea of choice. Otherwise, every family would end up with the particular character of school that happened to exist in their catchment area, even if they preferred something quite different.

While choice may not necessarily increase social segregation, the possibility is there. Here, there are further lessons to be learned from abroad. In the Netherlands, for example, many immigrant families opt for the least ambitious of the four secondary-school types, because they are convinced that they will feel more at home there.

UK parents have proved quite adept at this kind of social clustering. In recent years this has been related to geographic location and league-table performance rather than specialism or ability-testing. The tables place schools in a hierarchy, and prompt the cross-town school runs and moves to different catchment areas more common among the middle classes.

An important objective of giving schools different characters should be to flatten this hierarchy, providing equally esteemed choices, rather than to provide new signals of which are the plum schools. This makes it important to move quickly towards a system where each school is recognised as having something special, rather than just those with special status. Otherwise, the present policy risks providing opportunities to some pupils, at the expense of those who most need them.

Donald Hirsch is an international consultant on education policy.