Children from disadvantaged backgrounds are less likely to go to grammar schools than their more affluent peers even if they are just as clever.
Poor children in selective areas are half as likely to attend a grammar school as other pupils of similar ability, according to the study by academics at Bristol university.
In the 19 areas where significant selection remains, just two per cent of pupils attending grammar schools are entitled to free school meals.
This compares with 12 per cent eligible for free meals in other secondaries in those areas.
"Even among the very able poorer children, only a small minority make it," said the report, Selective Education: Who Benefits from Grammar Schools?.
Adele Atkinson and Paul Gregg, of Bristol university's Centre for Market and Public Organisation, suggest that the small number of disadvantaged pupils who attend grammars may be a result of the complexity of admissions systems which allow grammar schools to operate their own admission policies and in some cases set their own tests.
"It is possible that this approach is leading to a gulf in access between affluent and poor children," they said.
High-ability pupils with special needs or who speak English as a second language are also under-represented in grammar schools, they said.
Their findings are based on information about children's test and exam results between the ages of 11-16 taken from the national pupil database.
The report accused both supporters and critics of grammars of exaggerating their case.
"On average, there is little difference in achievement between pupils in local education authorities that still have selection and similar pupils in comparable non-selective LEAs," it said.
But it found selective systems make big differences to individual pupils.
Supporters of the 11-plus are correct to claim that the few bright pupils from poor backgrounds who succeed in gaining a grammar school place benefit from selection.
Their overall GCSE results are an average of eight grades higher than their peers in non-selective authorities.
Overall, grammar school pupils "do very well" compared with similar children in non-selective areas.
But this advantage for the 25 per cent of pupils who attend grammars is balanced by disadvantage for the 75 per cent who do not.
Selection creates large concentrations of poor children in individual schools and this reduces overall attainment, the study found.
Brian Willis-Pope, chairman of the National Grammar Schools Association, admitted poor parents might be put off by the complexity of the admissions system. He said they would be more likely to choose grammar schools if they were allowed to see the result of their child's 11-plus test before applying.
David Chaytor, MP for Bury North and chair of Comprehensive Future, a Labour group campaigning to end selection, said: "It is not surprising that working-class kids do better in schools with greater resources where they are surrounded by other clever children.
"Rejecting children at age 11 stores all kinds of problems for the future."
Selective Education: Who Benefits from Grammar Schools? by Adele Atkinson and Paul Gregg is available at www.bris.ac.uk DeptsCMPObulletin11j.pdf
Grammar schools Friday The Issue 11