This study reports evidence from an unusual policy intervention – The Reaching Out of School Children (ROSC) project – in Bangladesh where school grants and education allowances are offered to attract hard-to-reach children to schools comprised of a single teacher and a classroom. The operating unit cost of these schools is a fraction of that of formal primary schools. We use panel data to investigate whether ROSC schools are effective in raising enrolment and learning outcomes. Our findings suggest that there is a modest impact on school participation: ROSC schools increase enrolment probability between 9 and 18% for children in the two age cohorts 6-8 and 6-10. They perform as well as non-ROSC schools in terms of raising test scores, and even have positive impacts on academically stronger students. There is also strong evidence of positive externalities on non-ROSC schools in program areas. These results point to the effectiveness of a new model of non-formal primary schools that can be replicated in similar settings.