Why do grammar school systems increase inequality?

Lindsey Macmillan (IOE), Matt Dickson (University of Bath), Simon Burgess (CMPO)
The role of grammar schools is still a hotly contested topic in education policy in England. We contribute to this debate by showing that earnings inequality is higher under a selective system in which pupils are allocated to secondary schools based on their performance in tests at age 11. While selective systems have declined since their heyday in the mid-1960s, a number of areas retain a selective system and some believe that this system should again be expanded.
In our recent paper, we moved away from typical questions around grammar schools such as whether access to them is fair (it isn’t) and what the impact of grammar schools is for the marginal student (debatable), to ask about the longer term impacts of these types of systems on earnings inequality.
Using a nationally representative panel data source, Understanding Society, we considered the adult earnings distributions of over 2500 individuals born between 1961 and 1983, comparing those who grew up in an area operating a selective schooling system to those who grew up in very similar areas operating a comprehensive schooling system.
We ensure that the areas we are comparing are very similar by matching areas that are comprehensive to selective areas based on the average hourly wage, unemployment rate and proportion of private schools in both areas. The rich data source also allows us to control for things that may be driving the choice of area and the later earnings distributions, such as parental education and occupation when the individual was 14, gender, age, ethnicity and current area of residence.
We therefore compare the adult earnings of people who have very similar characteristics, live as adults in very similar areas and grew up in very similar areas: the main difference being that one area operated a selective system and the other a comprehensive system.
When we consider these two groups, then, we see that earnings inequality is greater for those who grew up in areas operating a selective system compared to those who grew up in comprehensive areas.
Comparing individuals of similar characteristics, the variance of earnings (2009-2012) for those who grew up in selective areas is £29.22 compared to £23.10 in non-selective areas. Put another way, the difference in pay between those at the 90th percentile of the wage distribution and those at the 10th percentile for those who grew up in a selective system is £13.14 an hour compared to £10.93 an hour in comprehensive systems.
On a personal level, if you grow up in a selective system and end up with earnings at the 90th percentile, you earn £1.31 more an hour (statistically significant) than the similar individual who grew up in a comprehensive system. At the other end of the scale, if you grow up in a selective system and don’t do so well – earning at the 10th percentile, you earn 90p less an hour (statistically significant) than the similar individual who grew up in a comprehensive system.
We can also compare the 90-10 wage gap between selective and non-selective areas to the overall 90-10 wage gap in the sample. As noted, in selective areas the 90-10 wage gap is £2.21 an hour higher than in comprehensive areas. This accounts for 18% of the overall 90-10 wage gap in our sample. So selective systems account for a large proportion of inequality in earnings. The message is clear. Grammar systems create winners and losers.
There are also interesting differences by gender. If we look separately at males and females, we see that males in selective systems at the top of the earnings distribution do significantly better than their non-selective counterparts (£2.25 an hour) while there is no difference for those at the bottom of the earnings distribution.
For females, the picture is the opposite. Females growing up in selective systems who do well look very similar to successful females from non-selective systems but those who do badly earn significantly less (87p an hour) than their comprehensive system counterparts. We think this could be because males were outperforming girls at school for the cohorts we consider and so more males attended grammars and more females attended secondary moderns within selective systems, although we cannot observe this directly.
What lies behind these differences? Inequality in earnings comes from inequality in qualifications and these in turn might derive from differences in peer effects and teacher effectiveness between the systems. We speculate that in the 1970s and 1980s more able teachers might have been more effectively sorted in a selective system into schools with high attaining pupils. The evidence on peer effects in the UK is mixed but the evidence on teacher effectiveness points to this as a possible key mechanism.
Whatever might be driving this phenomenon, our research shows that inequality is increased by selective schooling systems. If this is combined with evidence that sorting within selective systems is actually more about where you are from rather than your ability, then selective systems may not be the drivers of social mobility that some claim. The pros and cons of a system which creates greater inequality will doubtless continue to be passionately debated. What we cannot ignore is that there are losers as well as winners in this story.