![Sample our Education journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5932/TOC-Subject-Sample-2021-Education.jpg"})
Abstract
The voucher system in Denmark combines unrestricted generous subsidies with substantial autonomy of private schools as to schedule and teaching methods. This has produced a private school sector with a wide variety of school types. This paper uses data on eight cohorts of students (over 510,000 individuals) to compare educational attainment in public and private voucher schools, including religious schools (Catholic and Protestant) and various types of non‐religious schools. The findings suggest that, after controlling for individual and peer characteristics, the average public student would attain moderately higher levels of education if he/she attended grammar or Catholic school, relative to the public alternative. Attainment of students at Protestant, international and German minority schools is not different from public schools. However, attending free, boarding and, particularly, little and Waldorf schools is associated with substantially lower completion rates at the upper secondary level, which is probably at least partly due to the clustering of special education students in these school types, which cannot be controlled for. At the tertiary level, differences between private and public schools generally vanish.
Notes
1. McEwan (Citation2001, Citation2002) provides evidence on Chile and Argentina, Somers et al. (Citation2004) on several Latin‐American countries, Angrist et al. (Citation2002) on Columbia, Vandenberghe and Robin (Citation2004) on OECD countries, Dronkers and Robert (Citation2003) on 19 OECD countries and Dronkers (Citation2004) reviews evidence on public versus religious private schools' effectiveness in seven European countries.
2. Vandenberghe and Robin (Citation2004) present results for Denmark but, due to data limitations, they cannot (unlike in the present study) divide private schools into different categories.
3. Strictly speaking, the voucher is not given to individual families. Rather, it is a direct subsidy to the school. Though the physical mechanism of payment is different, the two policies are similar in directly linking the school budget to enrolment.
4. Throughout this article, I use the term “educational attainment” indicating length of completed education, while “achievement” is used for test scores or grades.
5. Rangvid (Citation2003) offers only tentative evidence by private school types due to data limitations.
6. A range of previous research shows that there are differences between the effectiveness of different types of private schools (Dronkers, Citation2004; Dronkers & Robert, Citation2003; McEwan, Citation2001).
7. Test score data are not available for these student generations. However, a recently released dataset for younger generations of Danish students includes grades of school‐leaving exams. I hope to examine these data in future work. Vandenberghe (Citation2003) uses data from the international PISA study to assess test score differences in Denmark and other countries, yet his data do not allow identification of different types of private schools and can therefore not be used to study such effects.
8. The Danish education system combines the primary and lower secondary level in one school (preschool to 9/10th grade).
9. In the following, I will use the word “effect” to describe (statistical) associations. However, this does not necessarily imply a causal relationship.
10. Yet some countries offer even more generous vouchers, e.g. the Dutch system provides a voucher that is of the same amount for public and private students.
11. When I use the term private schools, I mean government‐dependent private schools. They are private in the sense that they are run, quite autonomously, by a board of parents of the school. The public influence is mainly limited to financial support of these schools, and some few requirements by law. Government‐independent private schools are almost non‐existent in Denmark.
12. Setting up a private school is in principle easy: in order to establish a school and receive public vouchers, a parent or teacher must only gather a few students (a total of at least 28 pupils in the first to seventh grade) and establish a board of governors.
13. Yet this has changed during the last seven years. As a consequence of the not so favourable results from the first PISA‐wave in 2000, the Ministry of Education now requires schools to provide detailed results from national school‐leaving exams, which are disseminated via the Ministry's website. Also, today schools are required to maintain a school website providing information on the educational and pedagogical orientation of the school.
14. Yet some indirect channels of quality control exist: (1) Danish private schools are indirectly accountable to parents for good results on the national examinations that pupils typically take at the end of lower‐secondary school; (2) the government requires that private schools charge some (modest) minimum tuition (500 Euro per year) to all parents except those for whom it would cause undue financial hardship. This is to ensure that parents have a financial stake in the school they choose for their children and thus a high degree of commitment to the school. Parental commitment to and involvement with private schools are viewed as a cornerstone of their success; (3) additionally, as a means of assuring direct parental control of private schools, each one must be governed by a board elected by parents.
15. Due to the low enrolment in Muslim schools for these generations, they are not included in the subsequent analysis, although Muslim schools have become increasingly popular during the last decade.
16. Waldorf schools provide both conventional forms of special needs education and others that are unique to Waldorf education. Waldorf‐specific methods of special needs education would include curative eurhythmy, which is claimed to be especially therapeutic in its effects and is a development of the art of movement (eurhythmy) developed by Rudolf Steiner.
17. It is important to notice that a number of academically oriented (partly high‐prestige) boarding schools is in the category of “academically oriented private schools” rather than in the boarding school category. Thus, the group of boarding schools in this paper contains only boarding schools with special education programmes.
18. In our sample, average enrolment of eighth‐graders in little and free schools is only eight students. By‐grade enrolment in the other types of schools varies between 50 in public and grammar schools, 37 in Catholic schools, 27 in boarding schools and between 14 and 19 in the remaining school types (note: the number of students in Table is to be divided by eight to yield by‐grade enrolment, as there are eight cohorts of students in the sample.)
19. This percentage is estimated to 20–30% of ADHD‐students.
20. Binary outcomes are estimated by OLS as linear probability models, as done frequently in the recent literature (e.g. Cullen, Jacob, & Levitt, Citation2005).
21. In this paper, the term students' socio‐economic background covers a broad set of students' individual characteristics, and characteristics of students' family background that in the empirical educational literature have been suggested to be related to students' educational achievement and/or attainment.
22. The standard error for the estimate of the difference is calculated using the formula given in McEwan (Citation2001): , where V Pu
S is the variance–covariance matrix from the public school estimation, and
the matrix from the estimation of private school type j. X− is a vector with the background characteristics of the average public school student.
23. Students who have left the country permanently less than 10 years after completing eighth grade are deleted from the sample (about 8,000 individuals or 1.5% of the sample) because these observations are censored, since we cannot be certain that they have completed their schooling. These students are unequally distributed across school types: for most types of schools, only 1–3% of students leave the country within the first ten years after completing eighth grade, while this is the case for 15% of students at international schools and 8% of German minority schools' students. Due to the nature of our data, namely that we do not observe the completed education level for students after they leave the Danish school system, keeping these observations would probably induce a downward bias in the educational attainment measure for such schools.
24. We have data on the highest educational degree attained. Most of the coding is straightforward: seventh, eighth, ninth and tenth grade are coded to 7, 8, 9 and 10 years of schooling. Further, a vocational or academic degree at the upper secondary level is coded to 12 years, short college to 14 years, a bachelor degree to 15 years, a master degree to 17 years and a PhD to 20 years of education.
25. Data on parents' education are missing in a significant number of cases (for 4% of students' mothers and 7% of fathers—either because there is no information on the parent in the register altogether or because specifically the education variable is missing).
26. Also, 14% of all schools did not fit into any of the nine categories. These schools are gathered into a category named “other private schools”. Since they do not represent a distinct type of school, they are not included in the further analysis.
27. This is not only due to the fact that schools are located exclusively in an area of Denmark (Southern Jutland) with low education levels Compared with the country mean: even Compared with their public school counterparts in the region, the level of education of parents in German minority schools is considerably lower (0.28 standard deviation).
28. The results from these regressions are available from the author on request.
29. The results from these regressions are available from the author on request.
30. These results were obtained by a simple OLS regression (without controls) including a set of dummies for private school types with public schools being the omitted category.
31. Therefore, the sample is slightly smaller than for the preceding outcomes.
32. From hearsay it seems that little schools are popular among parents whose children are difficult to accommodate in larger student groups and who might be helped by the small and individualised environment that small classes and small schools can offer. Also, this probably introduces negative selection on (for the researcher) unobserved characteristics into little schools and most certainly causes downward bias in the little school effect. Even though we have not found evidence of this, free schools might also attract similar students, since average by grade enrolment in free schools is similarly low as in little schools.
33. A formal analysis of wages is postponed to a later study.
34. And perhaps in grammar schools, although the size of the difference to the high‐SES estimate is small.
35. If low‐SES students are over‐represented among those who are referred to special boarding school programmes due to behaviour problems and social problems in their families, there is a higher degree of negative selection of low‐SES students, which might bias the estimate for low‐SES students downward relative to that of high‐SES students.
36. This was done separately for each private school type to approximate the relative supply of private schooling.
37. Using origin in English‐, French‐ and German‐speaking countries as instruments for international schools and only German‐speaking countries (Germany, Austria, Switzerland) for German minority schools.
38. Problems encountered with the potential instruments ranged from weak predictive power in the school choice regressions to producing abnormally huge (and sometimes significant) private school coefficients in the outcome regressions.
39. However, only 13% say that this is a prerequisite or has high priority. For an overview of the PISA 2000, see OECD (2001).
40. However, there are two caveats: first, the existence of differential admission policies between different types of private schools cannot be investigated within the PISA‐framework (since we have only a public/private dummy available, but not type of private school); and second, the PISA data are from 2000, while the student generations investigated in our estimation sample are eighth graders in 1985–1993. If the pattern of selective admission policies has changed since then, the data from the PISA study may provide imprecise or even incorrect information.
41. To assess school effectiveness properly for this group, one would also need information on the severity of the disability, as the least able probably attend public special education schools, which are not included in the data.
42. Corten and Dronkers (Citation2006) consider private government dependent schools and argue that most of them are religious (mainly Catholic) schools in most of the countries they consider in their study.
