Average attainment figures showing that pupils with English as an additional language (EAL) outperform native speakers are "profoundly misleading" because they mask "enormous variation" between children with different first languages, a thinktank has said.
In a report published today, the Education Policy Institute (EPI) also argues that statistics on the performance of EAL students are plagued by missing or misleading prior attainment records, and that "better official statistics" are required to prevent the "urgent needs of some sub-groups" being missed.
According to performance statistics published by the Department for Education last month, in 2017 GCSE pupils with EAL outperformed native speakers on all DfE measures for the first time.
The EPI analysed earlier data from 2016, in which EAL pupils had an identical Attainment 8 score to the national average, made greater than average progress during school, and were more likely to achieve the English Baccalaureate than those with English as a first language.
EAL students' attainment
However, while some commentators have hailed the performance of EAL students, the EPI said that average attainment figures for these students are "profoundly misleading" because they encompass a group which is "extremely heterogenous".
The thinktank points out that some pupils reported as EAL speakers were born in this country and speak English fluently - while also happening to speak another language at home – whereas others are recent arrivals in the country with no English at all.
According to the report, attainment varies significantly based on a child's first language and how late they enter the school system.
At key stage 2, pupils from six language groups – Pashto, Panjabi, Turkish, Portuguese, Czech and Slovak – have attainment below the national expected standard, even for children who had arrived in English state-funded schools as infants.
At the other end of the spectrum, three groups – Tamil, Chinese and Hindi – have attainment that is at least one point above the expected standard for children who arrived as late as Year 5.
Likewise, at secondary level, pupils with Chinese as their first language have above-average Attainment 8 scores at age 16 even if they have arrived as late as Year 10, whereas pupils with Portuguese as their first language have below-average attainment even if they have attended English schools since the age of five.
'False sense of security'
The report argues that "current statistical outputs mask this variation", creating a "false sense of security around the ability of current policies to deliver fair outcomes, because smaller groups of children with high and unmet needs are hidden within a larger group".
The EPI also highlights a number of methodological issues that could distort the DfE's data in a way which inflates the average attainment and progress of EAL pupils.
"Measurements of attainment by children with EAL are misleading because assessments undertaken before English proficiency is reached will under-estimate academic attainment. Attainment is mediated by the child’s English proficiency at the time of the test," the report states.
Many children with EAL will also have missing attainment data because they arrived after the time of the assessment.
The EPI estimates that around three in 10 children with EAL fall into this category in primary schools – and around one in 10 children with EAL in secondary schools.
The report calls for "better official statistics", which acknowledge the "wide spread" of EAL outcomes – such as attainment breakdowns by first language, statistical benchmarking by time of arrival in English schools and analysis of new English proficiency assessments that the DfE began collecting in 2016.
The report states: "We believe it is important to work towards intelligent benchmarks for EAL attainment to avoid complacency about outcomes for high-ability groups – and to avoid the average masking the urgent needs of some sub-groups, but also because there can be no progress towards accountability for the funding allocated in respect of EAL without a reasonable benchmark with which to compare performance."