Primary teachers have been giving seven-year-olds artificially low assessment scores to put their schools' overall performance in a better light, research suggests.
The tactic, uncovered by an analysis of government data, allows schools to improve their key stage 2 value-added scores by underplaying pupils' achievements at the end of KS1, the report argues.
Researchers compared assessments in primaries with those conducted by infant schools, which have no KS2 pupils and therefore no incentive to record low KS1 scores.
Seven-year-olds in infant schools have consistently scored higher than their primary counterparts since at least 1998, according to Education Datalab, a new research organisation set up by charity FFT. Meanwhile, pupils leaving junior schools are consistently behind their primary peers in terms of progress since KS1.
"It seems highly unlikely that infant schools are systematically effective institutions while junior schools are systematically ineffective institutions," notes the report (available at www.educationdatalab.org.uk).
But although junior schools have often claimed that the phenomenon is owing to infant schools inflating the KS1 results they are judged on, the data suggests another cause.
The researchers find that the gap between primary and infant KS1 results widened suddenly after 2003, when there was a switch from externally marked tests to internal teacher assessment. Primary scores dropped dramatically while infant school scores remained stable.
"The pattern of this divergence is very clear," the report says. "There is little evidence that infant schools are taking advantage of teacher assessment to inflate the scores they give pupils.
"Instead, teacher assessment in primary schools produces lower judgements of KS1 attainment, thus lowering their bar to show impressive pupil value-added at KS2."
Rebecca Allen, director of Education Datalab, said: "This isn't cheating as we have seen in the US, where people get a rubber and they erase the things that children have written.
"It is asking people to apply subjective judgements knowing that the decisions they make are going to be critical to the success of the school in a few years' time - in the way that they are judged by Ofsted and by everybody else. That's a really tough ask."
But Dr Allen said the situation was complicated because not every primary school was likely to be depressing teacher assessment results, and those that were could be doing so to varying degrees.
The research also finds that primary teachers seem particularly cautious about assessing seven-year-olds as performing above expectations by attaining level 3.
But Dr Allen does not think the problem will disappear with the end of national curriculum levels, because schools will be able to make baseline assessments that could be subject to the same pressure.
The phenomenon was probably explained by primary teachers acting in the interests of their schools, rather than being instructed to lower results by their headteachers, she said.
But Ruth Whymark, headteacher of Cranmer Primary School in Merton, South London, had doubts about the theory.
"I do not understand why primary schools would intentionally deflate their KS1 results," she said. "When Ofsted come in they are looking for progress throughout the school. If primary schools deflated KS1 results, they would flag that up as a significant issue for progress throughout the early years and KS1. Good schools will always assess their children accurately."
Amanda Spielman, chair of exams watchdog Ofqual and education adviser to the Ark academies chain, said: "This analysis shows how difficult it is for someone making a judgement not to be influenced by their knowledge of the use to which that judgement will be put.
"As assessment data is put to more uses, it becomes even more important to make sure that the context is recognised when the data is interpreted."
Why physics teachers don't need a degree
Having a physics degree makes no difference to an individual's effectiveness in teaching the subject, an Education Datalab report published this week suggests.
Researchers used a contextual, value-added measure of GCSE physics results to compare the performance of 1,128 secondaries, accounting for prior attainment and pupils' backgrounds.
The study groups the schools into three categories depending on the proportion of science teachers with physics degrees, but finds no link to physics GCSE success. Schools' average science GCSE points score also fails to rise in line with the number of teachers who have physics degrees.
Sam Freedman, research director at Teach First, said: "It's not really that surprising. The type of content a physics undergraduate deals with is very different from the GCSE syllabus."