Education policy has recently focused on improving accountability and incentives of public providers for actual learning outcomes, often with school-based reward programmes for high performers. The Learning Guarantee Programme in Karnataka, India, is prominent among such efforts, providing cash transfers to government schools that achieve learning at specified high levels. This study examines whether schools that self-selected into the incentive programme are different than those that did not. The answer has important implications for how to evaluate the impact of such a programme. Although we find no significant differences in resources and characteristics, we do find significant and substantial differences in test scores prior to selection into the programme, with better performing schools more likely to opt-in. These findings also provide insight into how incentive-based programmes that focus on levels of (rather than changes in) achievement can exacerbate inequality in education. Failing schools, since they are more likely to opt-out of incentive programmes, are likely to require other targeted programmes in order to improve. In addition, our findings reinforce the value of randomised controlled trials to evaluate incentive programmes since evaluations that rely on matching schools based on resources (if, for instance, pre-programme test scores are unavailable) will be biased if resources poorly predict test scores.