Information Provision and Public Preferences over Education Finance Rules in Korea

The intervention consists of a randomized information module embedded in a nationally representative online survey of Korean adults (ages 19–65). After completing identical pre-treatment modules covering demographics, baseline fiscal knowledge, and neutral primers on the LEFG, respondents are randomized at survey entry into one of four conditions. All modules use the same visual format, neutral tone, closed-ended response scales, and comparable length.

1. Placebo (control): Non-directional fiscal context. Displays the education budget by function and program category, introduces the Ministry of Education's fiscal-assessment framework, and elicits beliefs about the LEFG's relationship to public education quality. Contains no information that could directionally shift spending or rule preferences.

2. Efficiency treatment (expenditure efficiency under demographic change): Presents factual content on the mechanical relationship between cohort decline and per-student resources under the fixed LEFG formula, including recent trends and long-run projections of per-student spending, student-teacher ratios, and class sizes.

3. Reference-point treatment (international benchmarks): Provides cross-country comparisons anchoring evaluations of Korea's education spending to OECD peer-economy levels, covering student-teacher ratios, class sizes, and spending shares by education level.

4. Trade-off treatment (cross-sector budget trade-offs): Embeds education finance in the economy-wide spending basket by presenting long-term fiscal projections of mandatory vs. discretionary spending, breaking out social welfare spending by category, and overlaying LEFG projections with demographic structure.

2024-10-18

2024-10-24

We plan to capture data on the following:
- Perceived adequacy of current education spending
- Preferred education budget share
- Long-run spending priority rankings across sectors
- Perceived adequacy of the LEFG formula (20.79% linkage)
- Preferred LEFG formula share
- Support for changing the current formula
- Preferred form of formula reform (e.g., population-based vs. project-based)
- Preferred allocation of surplus education funds across competing uses

- Perceived adequacy of the LEFG share within local education budgets
- Preferred LEFG share within local education budgets
- Perceived need for budget expansion across welfare sub-sectors

An individually randomized survey experiment with approximately 3,000 Korean adults assigned to four arms in roughly equal proportions. The sampling frame is the Hankook Research master panel, with proportional quotas by region, sex, and age benchmarked to recent population statistics. The survey is fielded via web mode.

The three information treatments are designed to isolate distinct channels: (1) demographic-efficiency, (2) reference-point, and (3) opportunity-cost. Across arms, we hold constant visual format, length, neutral tone, and the absence of partisan or elite cues. The placebo arm provides non-directional fiscal context to separate information effects from mere survey exposure.

Quality controls include soft prompts for skipped items, internal consistency checks, and time-on-page constraints on data-heavy screens.
Pre-treatment items include a short review quiz measuring baseline comprehension. We plan to explore heterogeneous treatment effects along dimensions such as prior awareness of the LEFG, parental status, education background, and survey attentiveness.

All respondents complete identical pre-treatment modules (demographics, labor-income covariates, neutral primers on education outlays and the LEFG) before randomization into one of four information modules. Immediately after the assigned module, all respondents answer identical post-treatment outcome items, followed by comprehension checks.

A supplementary survey of public-finance experts will also be conducted to provide an external benchmark.

Computer randomization at survey entry, implemented within the web survey platform.

Individual respondent

No

3,000 individual respondents

3,000 individual respondents

Approximately 750 respondents per arm (four arms of roughly equal size)

With approximately 750 respondents per arm and individual-level randomization, the minimum detectable effect size for a two-arm comparison at 80% power and 5% significance is approximately 0.14 standard deviations.