A report from Martin Delap of the AEB’s research department found that about half of candidates did worse than their teachers expected. He collected 7,000 predicted grades from about 450 schools in 11 subjects, and says teachers “were not inclined to provide estimates of low grades (ie, N or U) for their candidates; approximately 650 candidates were estimated to obtain grades N or U whereas over 1,700 of these grades were awarded. Similarly, more candidates obtained grade A than were estimated to obtain grade A.”

Tony Higgins, chief executive of the Universities and Colleges Admissions Service (UCAS), said many students were basing their university applications on wrong information, since so many teachers overestimated, which could lead to disappointment if they lost their conditional places.

The findings, published in the journal Assessment in Education, will provide fresh evidence for those who want the universities admissions system changed so that candidates are not admitted until their A-level grades are known. Teacher associations favour such a change, and the Committee of Vice Chancellors and Principals is looking at a range of changes to university admissions. But a spokesman said major change was unlikely before the Dearing review on 16 to 19 qualifications was complete.

Martin Delap’s study shows a great deal of variation among subjects and grades. For instance, in physics, the proportion of accurate estimates of grade A was high (84 per cent), but for grade C it was very low (18 per cent). Chemistry, meanwhile, proved very hard to predict at grade A (28 per cent), but easier at grade U (65 per cent). English was hard to predict altogether, with accurate estimates in 40 per cent of cases or less at every grade.

The study concludes: “The analyses have shown that the teachers’ estimates were not very accurate; about half of the estimates were optimistic and estimated grades of C, D or E were accurate on about one in four occasions.”

There was some evidence that teachers are more generous when estimating girls’ achievements, particularly in biology, geography and maths. And independent schools were likely to make lower predictions of English grades than comprehensives.

Delap says that while a change in admissions policy to wait for actual grades “may be desirable for other reasons, the accuracy of the estimated grades themselves should not necessarily lead to such a conclusion.” His report says that although estimates appear unreliable on an individual level, “this may not be of paramount concern to higher education institutions”, which can work on the basis of prediction trends. He says a 1976 study found teachers’ estimates of performances provided a better indication of final degree success than the actual performances.

Copies of “Teachers’ Estimates of Candidates’ Performances in Public Examinations” can be purchased for Pounds 10 from Carfax Publishing, P O Box 25, Abingdon, Oxfordshire OX14 3UE

Percentage of accurate teacher-estimated grades for each subject Grade Physics Maths French Biology Chemistry English History Geography

A 84 75 67 54 28 37 41 53 B 29 50 48 40 34 27 35 39 C 18 29 24 29 27 22 19 23 D 22 25 27 27 27 30 26 26 E 22 23 29 33 32 31 24 30 N 15 25 15 30 27 40 26 17 U 54 78 77 43 65 40 78 69