Does immigration induce “native flight” from public schools?

Most studies available draw attention to the United States, where some have looked at private schools, while others have studied the so-called charter schools. The latter emerged in the beginning of the 1990s in some US states as an alternative to public schools. Charter schools are often started for the purpose of attracting certain minority groups “to keep students in school” (Tedin and Weiher, 2004 p. 1112).

See Rothstein (2006) and Urquiola (2005), raising the issue of children sorting into different schools on grounds other than school efficiency, which blunt the impact of school choice on educational outcomes. One motive in school choice could be to ensure one’s offspring to attend schools that are in line with the parents’ religious identity. See Cohen-Zada (2006) for a theoretical model and empirical work on such mechanism.

See also Gibbons et al. (2008) for a discussion and references on theoretical explanations on possible mechanisms involved. As they point out, there are two main arguments for efficiency gains due to school choice. The first holds that more competition will enforce schools to adopt/improve “teaching technologies.” The other argument holds that free choice will lead to better matches as pupils will sort into schools based on educational preferences and needs. Chubb and Moe (1988) put forward another argument for efficiency gains by stressing the aspect of the institutional-level decisions concerning schools financing and standards. They argue that private schools have more scope to set their own standards and methods.

About 85 % of those enrolled in private schools in their sample attend religious (i.e., catholic) schools.

Betts and Fairlie address endogeneity issues by using IV-estimations, where their instrument for demographic changes from 1980 to 1990 is the share of immigrants in the year 1980. Due to the rather small sample size of 132 metropolitan areas, these IV-estimates are not very precise and therefore rather unreliable.

Their results point to sorting along socioeconomic lines but they do not find improvements in various outcome variables on school achievem ents in their sample of Chilean municipalities (sorting along ethnic lines is not addressed). Their results are consistent with Cullen et al. (2005), who look at the outcome of “open enrolment rules” into Chicago public schools. In a study by Böhlmark and Lindahl (2007), the authors look at the effect of the introduction of free schools in Sweden in 1992. Their main focus is on educational outcomes, but they also report results on sorting. They find a positive correlation between the fraction of immigrants in a municipality and enrollment into free schools, that is, pupils with immigrant background show a larger propensity to enroll into such schools than native Swedes.

A study by Gould et al. (2009) states a negative effect as to the presence of immigrants in elementary schools for the chances of passing matriculation exams in Israel. Other studies that have looked at the interplay of ethnic heterogeneity and school performance did not find similar negative correlations; see, e.g., Angrist and Lang (2004) and Guryan (2004).

See Epple and Romano (2008) for a discussion on voucher systems that can avoid cream-skimming effects.

This option has been used by about 58 % from (1992) to 37 % (in 2004) of the students leaving grade 9. A small number of schools also offer an 11th grade option. In addition, pupils aged 14–18 can attend the so-called continuation schools (efterskole), that is, schools that aim to build a bridge from elementary schools to secondary education, and which “enjoy complete freedom of curriculum,” see Johnson (1997). These schools are not included in the estimations. For a more detailed description of the educational system in Denmark, see Danish Ministry of Foreign Affairs (2008).

According to Andersen (2008), some municipalities apply a formal rule of funding public schools according to the number of pupils enrolled, but in praxis this does not seem to be a very strictly administered policy.

There are a small number of private schools that mainly rely on tuition fees, but they only make a minor feature of the Danish educational system, so the focus in this study is on free schools.

For more details and an account of studies on charter schools, see Hanushek et al. (2007). They describe charter schools as “...hybrids of public and private institutions that allow independent development and decision-making along with public financing and state accountability for performance” (p. 824). In contrast to private schools, minority groups, especially Afro-Americans, are over-represented in charter schools.

In 2006, the operational costs paid by the state government amounted to 41,000 dkr (per year), and the fee paid by the parents was 9,000 dkr on average. The local governments reimbursed the state by paying 31,000 dkr for each pupil. See “Private Schools in Denmark,” Danish Ministry of Education, http://​www.​eng.​uvm.​dk/​.

Other studies on the effect of private school on educational achievements in Denmark draw similar conclusions; see Nannestad (2004), and Schindler Rangvid (2008).

Betts and Fairlie (2003) make a distinction on pupils attending high schools and those in elementary schools. They argue that the impact of immigration should be larger in high schools, as elementary schools are covering smaller residential districts, that is, parents can easier ensure their preferred peer groups due to residential segregation. Differences in the number of school districts at the primary and secondary level have been utilized by Urquiola (2005), for the purpose of disentangling mechanisms of socioeconomic and/or racial stratification.

For 2005, the private school attendance rates for children with immigrant background is .054 in smaller municipalities, .052 in medium-sized municipalities, and .104 in larger municipalities. There is considerable variation across municipalities. For example, for the city of Copenhagen, the average private school attendance rate is about 0.28 and 0.18 for native Danes and those of foreign descendent, respectively.

The Danish governmental system grants considerably leeway to local governments to decide on local tax rates and the way tax money is spent. Education accounts for a main part in the local budgets; see e.g. Gerdes (2011).

See also Black and Machin (2011) for a discussion.

The effect of immigration on housing prices is still open to debate. For references on some studies done for the US, see Card (2007). A study by Bayer et al. (2007) concludes that race is not capitalized in housing prices as sorting based on race would be sufficient to satisfy “self-segregating preferences.”

As an additional control (not shown here) I also include a control variable for Total taxes on real property, which did not change the results. Estimations are available on request by the author.

For more information regarding the way the placement policy was accomplished, see Gerdes (2011) and references given there.

Access to private schools across municipalities has not been considered as a factor in placement of refugees.

One main reason to weight the results by population size within municipalities is to take in to account of the fact that some municipalities only host few immigrant children, which makes the according percentage shares less reliable.

This is about similar to the age group used in Betts and Fairlie (2003), that is, they use as their measure of immigration pupils aged 7–16. They motivate this choice by the fact that redistribution of school resources due to immigration will affect both lower and upper grades.

The reason for using a one year time lag is to secure that the sequence of events is a choice of private schools after an influx of immigrant children to the municipality has taken place.

All standard errors are robust to arbitrary within-municipality serial correlation and heteroskedasticity. Included covariates are weakly dependent, i.e. correlation diminishes over time. In such situation fixed effects are preferred to first difference estimations. See Wooldridge (2002), p. 302 for a discussion.

The considerably smaller R-squared values shown in the first-difference estimations and the IV-estimations are caused by the fact that these estimations are based on within transformations rather than using municipality dummies as in the previous regression estimations in Table 2. For both the first-difference estimations and the IV-estimations, the “xtivreg2” command has been used; see Schaffer (2010). The first stage F-test for significance of the instrument in the IV-estimations lies in the range of 25–51; see last row in Table 5.

A similar argument is made in Betts and Fairlie (2003) and Fairlie and Resch (2002).

This is calculated by multiplying the ratio of a standard deviation and mean change in the share of immigrant children aged 6–16 with parents from countries outside the EU and OECD for all municipalities (see column (1) in Table 7) and the coefficient estimate from Table 2, column (4), i.e., \([.029/.048]*.0429\).

For their sample of 132 municipality areas, Betts and Fairlie (2003) estimate a 0.34 percentage point increase in private secondary schools attendance rate for natives following an increase of 1.06 percentage points in immigrant shares in the age group 7–16 for the period 1980–1990.

According to Andersen and Serritzlew (2007), in 2001, about one-third of the municipalities in Denmark did not have a private school within their boundaries, while about one-third had just one. The average numbers of public and free schools in each municipality were 6 and 2, respectively.

This interpretation is in line with the main finding in Urquiola (2005), saying that an increase in school district availability reduces enrollment into private schools at the same time as it affects children’s peer groups “with respect to both race and parental education,” p. 1310.

Sudden (unexpected) demographic changes due to refugee immigration have been used in numerous studies to establish causal relations, most prominently perhaps in Card (1990). Angrist and Kugler (2003) use the influx of refugee migrants from former Yugoslavia to different areas in Europe to evaluate the importance of labor market institutions for employment outcomes of natives and immigrants.