Game, set and match: new evidence from the Swedish voucher reform
The results, which show the impact of free school competition on pupils in both free and municipal schools, indicate that a 10 percentage-point increase in the share of 9th-grade pupils attending free schools in the municipality generates (1) about a 2 percentile rank point better test score in mathematics and English; (2) a 2 percentile rank point better performance in mathematics and English in the first year of upper-secondary school; (3) a 2 percentage point increase in the share who attend university; and (4) four more weeks of schooling on average. This is the first paper that finds long-term positive effects from a national voucher programme, while also presenting results indicating that grade inflation does not bias their findings.
Comparing the results in ninth grade with the authors’ earlier results, we see that the impact is about twice as strong. It should also be noted that they only estimate the impact of ninth-grade shares of free schools pupils. In prior papers, they found that using the overall share of free schools as a measure of competition made the coefficient double in size (to about the same effect as in this paper). While it is questionable whether the overall share of free school pupils induces competitive incentives – at least immediately – this indicates that the above-cited estimates are lower bound.
This is also the first paper ever – to my knowledge – that separates the general-equilibrium impact of for-profit and non-profit competition on achievement (briefly noting my IEA-study when doing so). They find no significant differences. This is strong evidence that both for-profit and non-profit competition is equally good at raising achievement. The key difference is the ability of for-profit actors to mobilise capital as well as scale up – thus providing more competition across the board. Michael Gove should pay attention.
Indeed, it is important to note that about 70-80% of the positive impact does not stem from the fact that free schools are better than municipal schools – but rather that competition forces municipal schools to improve. It also turns out that the effects are not significant until after about one decade after the reform. The authors argue that this is because free schools were a relatively marginal phenomenon until the early 2000s. However, it is also important to point out that the pupil population kept growing until 2003, so whatever competitive incentives increased for municipal schools remained marginal. The pupil population began decreasing in 2003, which coincides with the stronger increase in the free school share and also achievement. It is thus not surprising that it took about ten years before competition effects kicked in.
This clearly demonstrates that it takes time before competitive incentives have an impact – which has implications for how we evaluate other school choice programmes. Moreover, it indicates that competition is the key mechanism behind producing better outcomes, not the fact that some schools are better than others. It also shows that for-profit schools are just as good as non-profit schools when it comes to raising the overall attainment levels in a voucher system.
Finally, the authors find an insignificant negative effect on costs, rather than the significant positive impact they found in a previous paper. This indicates that all achievement gains can be interpreted as productivity gains. In other words, allowing schools to compete makes them more efficient.
This should settle the debate for now. There is no, I repeat, no evidence of any negative effects on achievement from school choice in Sweden. Indeed, the authors explicitly state that they ‘do not find any support for the belief that an increase in the share of independent school students provides an explanation for Sweden’s relative decline [in international comparisons]’.
While the Swedish voucher programme suffers from significant problems, which I discuss in more detail in a forthcoming study for the Institute of Directors, the evidence indicates that it has been somewhat positive despite them. This, of course, begs the question: how much would achievement increase if we fixed the problems?