This study examines the causal effects of the 2017 school dinner discontinuation policy implemented by the Gyeonggi Provincial Office of Education on academic achievement and private education in general high schools, employing a quasi-experimental de...
This study examines the causal effects of the 2017 school dinner discontinuation policy implemented by the Gyeonggi Provincial Office of Education on academic achievement and private education in general high schools, employing a quasi-experimental design. The Gyeonggi Provincial Office of Education initially pursued a complete abolition of evening self-study sessions but, facing opposition from schools, employed an indirect policy instrument—discontinuing school dinner services—to achieve its policy objective. Consequently, the proportion of general high schools providing dinner in Gyeonggi Province dropped from 80.2% in 2016 to 57.7% in 2017 (a 22.5 percentage point decline), further decreasing to 41.0% by 2019. No comparable changes were observed in Seoul or other regions during the same period. This study leverages this policy change as a natural experiment to empirically identify the unintended consequences of indirect policy instruments.
The analysis covers the period from 2013 to 2019 to exclude confounding effects of COVID-19. Gyeonggi Province serves as the treatment group, while Seoul (primary comparison) and four metropolitan cities and provinces in the Yeongnam region (secondary comparison) constitute the control groups. The identification strategy employs difference-in-differences (DiD) with 2017 as the treatment year, supplemented by triple difference (DDD) to examine heterogeneous effects by school ownership (public vs. private) and school type (non-selective vs. selective). The primary outcome variables are the mean standardized scores and grade distributions in the mathematics section of the College Scholastic Ability Test (CSAT), a national university entrance examination. For private education, the analysis examines participation rates, expenditures, and the number of private academies (hagwon) and instructors. The models include school and year fixed effects.
The main findings are as follows. First, the dinner discontinuation policy negatively affected mathematics achievement in the natural sciences track, one of two curricular tracks in Korean high schools (Hypothesis 1). The mean standardized mathematics score in Gyeonggi Province declined by 0.05 SD (approximately 1 point, p<0.01) relative to Seoul, while no such effect was observed in the humanities track. Furthermore, structural shifts in the achievement distribution were identified: the proportion of top-tier students (top 23%) decreased by approximately 1.7 percentage points, while the proportion of bottom-tier students (bottom 23%) increased by approximately 1.4 percentage points, indicating a pattern of downward leveling. Second, no buffering effects based on institutional characteristics were confirmed (Hypothesis 2). The DDD analysis revealed that private schools did not absorb the policy shock better than public schools, and selective high schools exhibited even greater achievement declines than non-selective schools. Third, the balloon effect toward private education was limited (Hypothesis 3). While private education participation rates increased slightly compared to Seoul, expenditures and the number of academies and instructors showed lower growth rates than in Seoul. This suggests that the reduction in public evening learning opportunities did not fully transfer to formal private education but may have dispersed to informal learning spaces (e.g., study cafés, reading rooms) not captured in official statistics.
These findings suggest that the indirect policy instrument of discontinuing school dinners may have released the “lock-in effect” that dinner services provided for keeping students at school, while increasing transaction costs for maintaining evening study routines—thereby producing unintended consequences including declining academic achievement and deteriorating achievement distributions. This study demonstrates the need for careful examination of the ripple effects of indirect policy instruments and provides empirical evidence for evidence-based education policy design.