Copying Swedish Free School Reforms

England’s schools are currently facing the prospect of the most radical reform since the dismantling of selective schooling four decades ago with the Conservatives looking to replicate Sweden’s free school reforms. The work of Swedish economists used to support the argument that choice and competition has improved academic performance, is however less unambiguous than the Conservative spokesman has claimed. Here I give a non-technical summary of the impact of the reforms on test scores, evaluating the relative merits of the papers and explaining why they disagree in their findings.

Background to reforms

In 1992, Sweden introduced a voucher scheme whereby privately-run (including for-profit) schools could receive public funding for each pupil they educated on the same terms as municipality schools. Like all market reforms of public services, the exact nature of the institutional structures, financing and regulation are critical to ensuring success.

However, the important context for the changes in Sweden was a backdrop of radical supply-side reforms, intended to facilitate innovation and more efficient resourcing decisions, including deregulating teacher pay and conditions, decentralising school financing and increasing school discretion of curriculum, goal setting and test regimes (see Björklund et al., 2005, for an overview of the reforms). Each reform was almost the exact opposite of the New Labour education reforms that were taking place at the same time in the UK.

Today about ten percent of lower secondary aged pupils in Sweden choose to attend free schools, with places strictly allocated on a first-come, first-served basis. There has been large regional variation in free school expansions: it has been greatest in urban, affluent and gentrifying areas and in those places with 2nd generation immigrant communities. Within these municipalities, more educated parents and 2nd generation immigrants are most likely to use the free schools, so the overall system is stratifying a little (Böhlmark and Lindahl, 2007). The largest group of free schools are for-profit providers of a general education, but special pedagogy, religious and special language/ethnic group schools are also prevalent.

Evidence of the impact on academic achievement

The ideal situation for identifying the causal impact of free school competition on pupil achievement is random assignment of the policy experiment across the 290 Swedish municipalities with no families moving across municipalities in response to the reforms. The actual Swedish situation does not reflect this ideal because free schools were not set-up at random across municipalities: they are more prevalent where the municipality is politically supportive and offers high per-pupil funding. Movement of pupils across municipality boundaries is also permissible, although not particularly common. Evaluation of the reforms is especially difficult because Sweden does not routinely collect administrative test score and demographic data on all pupils in the country, as England does. Externally marked test score data in maths, English and Swedish is available for around 30 municipalities, but for the remainder researchers are restricted to using grade levels that are not consistently standardised across the country.

The empirical papers described below adopt a variety of strategies to deal with the observation that areas with many free schools have demographic characteristics that make them systematically different. Two studies rely on cross-sectional data, hoping that the good quality of control variables (and the use of an instrument to predict free school supply) avoid confounding influences. The more recent two studies use a municipality panel to measure the extent to which changes in the share of free schools in the municipality are associated with changes in pupil test scores. This latter approach requires less arduous identification assumptions since it controls for time-invariant demographic factors in the municipality. There is still a problem establishing causality because trends in social demographics (such as an influx of immigrants) are associated with trends in free school growth, but this can be dealt with by collecting data on time-varying demographic characteristics and/or by accounting for pre-reform trends in test scores.

Böhlmark and Lindahl (2007, 2008)

The most recent paper is described first here since it uses the most robust data and methods, requiring relatively few identification assumptions. By describing this first, the relative drawbacks of the other papers’ approaches can be understood. The greatest advantages of their research come from the construction of a long-panel of data from 1988/89 to 2002/3. This gives three years of pre-intervention trend and over a decade of post-reform data to apply a difference-in-differences approach with municipality fixed effects to compare changes in areas with a large growth in free schools to changes in areas with smaller growth in free schools. The sample of data they are able to draw is also very large: a 20 percent sample from the population of pupils across all municipalities. The quality of the background control variables is good, including parental education, income, age and immigrant status, although it does not include any measures of the child’s prior attainment. They do rely on non-standardised attainment data, but have externally-marked test scores for a sample of pupils that is sufficient to confirm that biases in teacher-assessment are not correlated with the policy reform of interest.

They show a moderately positive impact of free school growth on municipality academic performance at the end of 9th grade (end of lower secondary school). This finding is convincing because it is consistently estimated across almost all subjects and model specifications; the greatest beneficiaries are found to be children from highly educated families (the impact on low educated families and immigrants is close to zero). By tracking siblings within family who differed in whether they attended municipality or free schools, they show that this superior performance of areas with private schools is due to both the greater effectiveness of private schools and also municipality schools making improvements in response to school competition, with the latter likely to be more important than the former.

However, they find that the advantages that children who were educated in areas with free schools have by age 16 do not translate into greater educational success in later life. Although there is some (weak) evidence that students in areas with many free schools are more likely to take an academic track in high school, they score no better in high school exit tests at the age of 18/19 and are no more likely to participate in higher education than those who were schooled in areas without free schools. They explore a variety of explanations for this, but conclude that the educational advantages of school competition are simply too small to persist into any long-term gains for young people.

Björklund et al (2005)

Björklund et al.’s short panel of data for municipalities between 1998 and 2001 is only able to analyse the relationship between growth in private school share in a municipality and changes in test scores over a short period of time, with correspondingly less variation in the parameter of interest (just a one percentage point change in free schooling share between these dates) and no pre-reform data to account for pre-existing social trends. They compare estimates between a sample of around 30 municipalities for which they have good quality data and all 290 municipalities where data quality is poor. Overall, they do not find a consistently positive impact of free school share on educational attainment: they identify a small positive impact on English and Swedish attainment, but a zero or even negative impact in maths. Their findings are not consistent across the sample and the population of municipalities, suggesting there may be selection problems in the municipality sample. This is a significant observation regarding data quality since the following two studies both rely on this sample of 30 municipalities.

Ahlin (2003)

Ahlin estimates the impact of the share of private schooling on grade nine test scores in a cross-section of 34 municipalities from 1997/8, hoping that the quality of her control variables are sufficient to avoid any confounding influences. This is the only study that includes the prior attainment of the pupil in grade six and further background controls, thus accounting for systematic differences in the levels of attainment across municipalities but not dealing with differences in expected rates of progress from grade six to grade nine that are due to home background factors. Her findings reverse those of Björklund et al. with quite large positive effects of private schools on overall municipality achievement in maths, but not in Swedish or English.

Sandström and Bergström (2005)

Sandström and Bergström were the first researchers to explore the impact of the free school reforms on overall academic standards in Sweden. Their finding of large positive gains to the reforms have been widely reported and is surprising given their data comes from quite early in the reform period (1997/8) before growth in the free school section became substantial. Their study relies on the largest number of identifying assumptions since they use a cross-section of only municipality schools in just 30 municipalities, using a parametric sample-selection correction to address composition changes caused by lack of data on pupils in free schools. They use a two-stage approach, with an instrument of political control predicting the municipality share of free schools. Critics argue the instrument may not meet the excludability criterion of predicting the growth of free schools but not directly determining education attainment because they are not able to control for most of the social factors in the municipality that explain household educational practices. Given this, it is hard to argue that their large positive finding should contribute to our current knowledge of the impact of the reforms.

Concluding remarks

The experience of Sweden is helpful, but necessarily limited, in the extent to which it can help us predict the impact of school reforms in England. One reason for this is that they also underwent a radical decentralisation of the education system, which would seem to be critical to promoting diversity and productivity gains through experimentation in free schools. They also have fewer reasons to be concerned that a free school system will produce greater school stratification since their lower levels of income and skill inequalities mean there is far less need for parents to choose schools based on social composition. It is also possible that their greater tradition of non-standard schooling (such as Steiner and Montessori) is leading to a greater diversity of provision than English parents would ever demand.

The econometric evidence on the impact of the reforms suggests that, so far, Swedish students do not appear to be harmed by the competition from private schools, but they have not yet transformed educational attainment in Sweden. Bunar (2009) argues that growth in free schools in the first decade was too slow to bring a great transformation due to unclear regulations and uncertainty as to whether the ruling Social Democratic party would further deteriorate the financial conditions for free schools. Also, the rising pupil population in Sweden during the 1990s meant that existing public schools did not lose students in great numbers as free schools opened and poorly performing schools did not need to close. In the past few years, overall demand for school places has fallen as the pupil population shrinks and the supply of free schools places has rapidly grown, so the prospect of a true competitive threat is now real and efficiency gains over the next decade could be larger.


Ahlin, Å. (2003) Does school competition matter? Effects of a large-scale school choice reform on student performance, Deparment of Economics, Uppsala University Working Paper, 2.

Björklund, A., Clark, M., Edin, P.-A., Fredriksson, P., and Krueger, A. (2005) The market comes to education in Sweden: an evaluation of Sweden’s surprising school reforms, New York: Russell Sage Foundation.

Björklund, A., Edin, P.-A., Fredriksson, P., and Krueger, A. (2004) Education, equality and efficiency – an analysis of Swedish school reforms during the 1990s. IFAU report, 1.

Böhlmark, A. and Lindahl, M. (2007) The impact of school choice on pupil achievement, segregation and costs: Swedish evidence, IZA discussion paper 2786.

Böhlmark, A. and Lindahl, M. (2008) Does school privatization improve educational achievement? Evidence from Sweden’s voucher reform, IZA discussion paper 3691.

Bunar, N. (2009) Can Multicultural Urban Schools in Sweden Survive the Freedom of Choice Policy? The Stockholm University Linnaeus Center for Integration Studies Working Paper 3.

Sandström, F. M. and Bergström, F. (2005) School vouchers in practice: Competition will not hurt you, Journal of Public Economics, 89:351–380.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s