The effects of school choice on students and public education: evidence from South Korea

ABSTRACT

This study examines the effects of school choice on students and public education using the case of the application-based random assignment system prevailing in Seoul, South Korea. I find little evidence that winning this educational lottery improves students’ academic performance, class attitude, or overall manners, which are quality-related indices of public education. While students’ school satisfaction increases significantly when they are assigned to their preferred school, this positive effect does not persist over time. I also find no robust evidence that the quality of public education, as measured by examining the behaviours of principals and teachers, is enhanced by South Korea’s school choice policy. These results do not align with advocates’ arguments that the school choice programme can normalise public education by stimulating competition among schools. One possible interpretation is that students may exploit the school choice program to facilitate their access to private tutoring and that this circumstance does not ultimately raise the quality of public education. Overall, my results suggest that the positive effect of school choice is limited and that a school choice system should be designed carefully.

KEYWORDS:

• School choice
• application-based random assignment
• policy evaluation
• south Korea

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. Parag Pathak won the John Bates Clark Medal in 2018 for his contributions to examining the impacts of school choice systems on education.

2. It is impossible to carry out a controlled experiment on students since introducing a beneficial educational policy to some pupils while forbidding it for others would be unethical. While many educational experiments can be both unethical and unfeasible, Seoul’s quasi-natural experimental circumstances from the random assignment system provide a close approximation to such an experiment.

3. In 2009, the government introduced a new type of school called an autonomous private high school. This type of school has a higher level of autonomy in school management and receives financial support from the government. Following the introduction of autonomous private high schools, Park (2016) investigates whether such autonomy and financial incentives can improve the quality of education.

4. My study complements Choi and Hwang (2017) work since it does not specify the individual’s school choice. They also use cross-sectional data from 2009 to 2011 and therefore cannot check individual-level self-selection bias directly.

5. According to Statistics Korea, the percentage of students who advance to high school is 99.6% (2009), 99.7% (2010) 99.7% (2011), 99.7% (2012), 99.7% (2013), 99.7% (2014), and 99.7% (2015).

6. The exact assignment mechanism is strictly confidential, however, one crucial point is that academic performance and family background are not used to allocate students to high schools. According to Dustmann, Ku, and Kwak (2018), the Seoul Metropolitan Office of Education first determines the total quota for high school entrance, which varies year by year. Second, students participate in the assignment procedure and a computer program randomly matches them to a high school.

7. The general school district can be a part of several combined school districts.

8. Unfortunately, the survey item “My high school is the one I wanted to attend before the lottery” is not applicable to all the respondents in the dataset; therefore, the sample size is rather small. However, the small sample size does not hamper the examination of this question. To support my argument, I compare the demographic characteristics of the surveyed group with those of the not surveyed group. The value of the F-test across the variables reveals that the null hypothesis is not rejected at the 5% significance level, suggesting little evidence of any systematic differences between them. Moreover, if there is any meaningful effect of the school choice program on students within the small sample size, this would further support the hypothesis that the school choice policy influences the actual educational traits of pupils.

9. While individual characteristics related to how much effort a student makes in school appear to be relatively high for students who answered “yes” to the survey’s question each year, as shown in Table 4, the combined effect (2010–2012) on their baseline academic performance shows no such systematic disparity. Thus, the risk of self-selection caused by over-reporting (by students putting more efforts into studying) is slight.

10. The exams were universally carried out for the surveyed students every year. Each test comprised 30 to 35 questions to examine students’ achievement based on the curriculum.

11. The validity and comparability of these subjective measures have been criticised. However, these measures do allow us to focus on the more subjective determinants of students’ well-being so that they can directly reflect students’ welfare (Kahneman and Krueger 2006).

12. The selection bias issue may also remain if more able students tend to select schools that are competitive and vice versa, even if they are from the same school district. At the very least, the school district fixed effect relieves these concerns.

13. I control for sex, number of siblings, birth order, free food program, private tutoring, family income, parents’ age, and parents’ schooling level. Explanations of the variables are given in the Appendix.

14. I control for school gender type (single-sex or co-ed), school foundation type (public or private), and average class size. Explanations of the variables are given in the Appendix.

15. Here, concerns over the different effects of school characteristics are rare since regular high schools are equalised and homogenised. First, teachers in public schools rotate to a new school every few years. Hence, little selection for teacher quality exists among high schools. Second, the Ministry of Education centrally regulates the curriculum and tuition in regular high schools. Consequently, class numbers, lesson hours, and academic materials are similar across teachers and schools. Teachers and schools have to follow this course of study (i.e., self-selection on the basis of coursework is unlikely). Third, school facilities are equalised since the central government allocates resources across schools based on equity and fairness. These equalised and homogenised public education environments enable us to distinguish the policy effects from other confounding factors (e.g., school, teacher, and peer effects).

16. If the effect is similar with and without control variables, it suggests that selection based on observable characteristics is not a significant concern in my model specification (Altonji, Elder, and Taber 2005b).

17. I average the test results of the three subjects (Korean, maths, and English) in the 3rd grade of middle school to calculate the exam score.

18. (1) Interesting: The level of a student’s interest in the lecture; (2) Helpful: The level of class helpfulness to a student’s learning; (3) Concentration: The level of a student’s concentration on the lecture; (4) Participation: The level of a student’s participation in the classroom; and (5) Diligence: The level of a student’s preparation for the class.

19. (1) Understanding: The level of a student’s understanding of the lecture through class materials and the teacher’s teaching skills; (2) Fun: The level of a student’s sense of fun with their studies; (3) Purpose: The level of a student’s proclivity for setting a specific purpose for how he or she studies and actively engages in class; (4) Self-Preference: The level of a student’s self-perception regarding his or her preference based on school attendance; and (5) Self-Advantage: The level of a student’s self-perception of his or her comparative advantage based on school attendance.

20. I apply the simple model above the ordered logit/probit regression to interpret the result as the degree of change. The qualitative results in the two settings are unchanged. The results also do not change when I extend the model to include the covariates.

21. It would be somewhat speculative to hypothesise why the effect is temporary. First, students may become accustomed to their school environments, which mitigates the positive effect on school satisfaction. Second, they could experience gaps between their expectation and the actual circumstances. Third, students are likely to be stressed while preparing for college exams, leading to lower school satisfaction in later grades. I leave this question to future research.

22. I apply the following self-evaluated variables to measure the behaviours of teachers and principals on a five-point scale from one (strongly disagree) to five (strongly agree). (1) Teachers: The level of activity towards class lecture improvement, level of participation in the counselling program, level of participation in speciality classes, and level of participation in a curriculum improvement program. (2) Principals: The level of support in teacher training programs, level of budget support, level of support in teacher evaluation systems, and level of support in class lecture improvement.

23. Recent studies seek to examine the underlying reasons for school choice. For example, Abdulkadiroğlu et al. (2020) show that parents in New York choose schools on the basis of causal effectiveness rather than on peer characteristics.

24. A total of 256 of the 328 students (78%) students answered the question. The attrition from the survey may cause bias. However, this is unlikely considering that the percentages of each category in the survey are balanced between the winners and losers of the lottery.

25. Private tutoring is pervasive in Seoul. See “Students not free from study during vacation.” The Korean Times, 16 August 2018.

26. However, this low R-squared value does not imply that the model is not properly established. The fact that the R-squared value is the coefficient of determination and that the regression line is close to the data indicates that various factors other than school choice can explain the education system.

Additional information

Notes on contributors

Hoyong Jung

Hoyong Jung is an assistant professor in the department of economics, Kookmin University, South Korea. He worked as a deputy director in the Ministry of Economy and Finance at South Korea during 2010-2019 and as an assistant professor in the department of economics, Chonnam National University during 2019-2020. He holds Ph. D. in economics at the University of Wisconsin-Madison.