Comprehensives that control their own admissions are less likely to admit poor, low-achieving or special needs pupils, a study of London secondaries has found.
Authors Professor Anne West and Audrey Hind from the London School of Economics are now calling for a centrally-controlled admissions system.
"There is a strong case for decisions about who is allocated to which school to be taken away from individual schools, given that they have a vested interest in terms of who is admitted," they say.
Their findings, presented to the American Educational Research Association conference, will be seized on by opponents of the education Bill which encourages schools to become their own admissions authorities.
They follow research for the Sutton Trust which showed that most top-performing comprehensives control their admissions and do not take a proportionate number of disadvantaged children.
The LSE study compared the proportions of pupils in 367 London comprehensives eligible for free meals - widely used as an indicator of poverty.
Among the 197 community and voluntary-controlled schools which followed local authority admissions criteria a third of pupils qualified for free meals.
This compared to only 20 per cent in the 170 comprehensives with foundation and voluntary-aided status that controlled their own admissions.
There were also differences between comprehensives using local authority admissions systems. The 32 which adopted practices such as favouring children of staff, had a quarter of pupils eligible for free meals compared to 35 percent of the 165 that did not.
An analysis of 24 London comprehensives with fewer than 5 per cent of deprived pupils on free meals showed 22 controlled their own admissions, 14 used religious criteria to select students and 14 used some potentially selective practice. Their percentages of deprived pupils were between two and 11 times lower than the averages for their authority.
The researchers also compared schools by placing them in nine bands - with nine the highest - according to the average point score in national tests taken by 11-year-olds in 2001.
The average band for the community and voluntary-controlled comprehensives was 4.47 rising to 5.84 for those with foundation or voluntary aided status. Community and voluntary-controlled comprehensives also had higher proportions of pupils with special needs - 3 per cent with statements and 20.9 per cent without - compared to 2.3 and 14.2 per cent in foundation and voluntary-aided schools.
The report compared the composition of London's 19 grammars with 89 comprehensives in the same local authorities and found significant differences in ethnic make-up.
In grammars 4 per cent of pupils were black compared to 12 per cent in comprehensives, while Indian children made up 11 per cent of grammar pupils but only 6 per cent of comprehensives'.
Pupils with Chinese or Asian backgrounds outside India, Pakistan and Bangladesh made up 10 per cent of grammar pupils but only 3 per cent of comprehensives. The report calls for monitoring of which students apply to which schools and which are admitted.
"State-maintained schools in London are publicly-funded, yet access to a significant number is restricted for various reasons; on account of academic selection, selection by religiosity and selection via other admissions criteria and practices; all of these privilege some students over others."