This is an extract from one of the links I posted about the 2012 trial and rolling it out nationally:
Broadening the benefits?
The pilot found that Year 6 students in areas of universal provision made an additional two months’ progress over the course of two years relative to similar students in other areas. These are significant effects, roughly the same size as the benefits from national programmes such as the “literacy hour”. However, it’s far from certain that universal free school meals would be as effective if rolled out nationally.
One reason for caution is the difference between the pilot areas and the average English local authority. Both Newham and Durham are relatively disadvantaged. If pupils in better-off areas are more likely to pay for school meals or to have healthier packed lunches, the gains from making school lunches free to all students are likely to be smaller.
Further, while the pilot study found that universal free school meals improved test scores, it wasn’t able to pinpoint how these improvements came about. Evidence from other countries suggests that Without understanding what’s driving the headline academic gains, it’s difficult to know whether the relationship between free school meals and test scores would be the same in other areas with a different local context.
Rather than providing free school lunches for all children, policymakers wanting to tackle student hunger could support school breakfasts instead. IFS research has found that support for a one-year breakfast programme in disadvantaged schools delivered similar academic benefits to universal free school meal provision (though the gains were higher in Year 2 than Year 6). The breakfast clubs also significantly improved behaviour and concentration, and reduced absences – and did so at around one-tenth of the cost per pupil of universal free school meals.
Extending free school meals to all primary school children would cost around £950 million each year. It would not directly benefit the poorest children, who are already entitled to free lunches. While there is some evidence it might raise attainment overall, we don't understand how or why, and so the effect of extending this nationwide is uncertain. In the context of constrained public resources it is important to be much clearer about effectiveness before spending a large amount of money on a new universal entitlement.
www.ifs.org.uk/publications/9202