Background The Education Maintenance Allowance (EMA) provides a stipend to young people (aged 16+) in low-income households if they participate in education or training. It was introduced UK-wide in 2004 and increased educational participation. However, its relationship with mental health is unclear. Austerity led to the EMA being phased out at the end of the 2010 school year in England, but it continues in the Rest of the UK (RUK). These policy differences created a natural experiment: we therefore investigated if EMA removal was associated with poorer mental health in England.

Methods We utilise an intention to treat difference-in-differences design using Understanding Society data for the 2009-2018 academics years. The analytic sample consists of people aged 16 or 17 on September 1^st of each year and in the bottom 10% of household incomes (meeting EMA age and income criteria). The exposure to EMA policy regime was identified using an interaction between two variables: UK area (RUK vs England) and EMA policy period. Academic years starting in 2009/2010 indicated the EMA Period, 2011 a transition period, and 2012-2018 as post-EMA. Mental health was assessed using the General Health Questionnaire (GHQ-12) score (0-36 points), the SF12 Mental Health Subscale and life satisfaction. SF12 Physical Health Subscale was considered a falsification outcome. Linear regression using robust standard errors, adjusting for sex and month of interview, was conducted. Weights were applied to account for non-response and analysis by academic year.

Results The analytic sample consisted of 1,328 observations from young people in the bottom 10% of household income. In England, relative to RUK and the EMA period, the transition period (OR4.2 95%CI 1.1 to 7.3) and the post EMA period (2.9 95%CI 0.7 to 5.1) were associated with a worse GHQ-12 score. Similar, but weaker estimates were found for the other mental health outcomes. There was little evidence of associations with the falsification outcome in the transition period (2.0 95%CI -1.8 to 5.9) or post EMA period (-2.1 95%CI -5.1 to 0.8). In sensitivity analyses widening the sample to the bottom 15% of incomes resulted in weaker associations.

Conclusion Young people living in low-income households in England appeared to have worse mental health following the removal of the EMA, compared to RUK. However, a slightly stronger effect in the transition rather than EMA period, suggests cointerventions could confound estimates. Supporting continuing education might provide short-term mental health benefits, in addition to longer-term from improved education.