Nine independent scoring periods. The Q5–Q1 governance gap is positive in every one. This page summarizes the key findings. The full statistical methodology is in Research Note #1.
The concentration at the bottom of the distribution is itself a finding: it quantifies the scale of governance deficits across the Korean market.
No firm achieves S-grade. This is not a calibration artifact — it reflects the structural reality that even Korea's best-governed firms have meaningful governance gaps in at least one axis. This distribution explains why the Korea Discount persists: the base rate of governance risk is high.
The governance–performance relationship holds across COVID recovery, post-vaccine rally, rate-hiking cycle, and normalization. The pre-COVID average (+12.8pp) and post-COVID average (+12.1pp) differ by only 0.7pp.
| Period | n | Q5–Q1 Gap | Regime |
|---|---|---|---|
| 2020 → 2021 | 489 | +14.1pp | COVID recovery |
| 2020 → 2022 | 495 | +8.5pp | COVID recovery |
| 2021 → 2022 | 563 | +17.6pp | Post-vaccine |
| 2021 → 2023 | 542 | +10.9pp | |
| 2022 → 2023 | 616 | +17.0pp | Rate-hiking cycle |
| 2022 → 2024 | 613 | +12.2pp | |
| 2023 → 2024 | 714 | +10.1pp | Normalization |
| 2023 → 2025 | 714 | +12.0pp | |
| 2024 → 2025 | 723 | +9.4pp | Most recent |
| Average | +12.4pp | Range: +8.5 ~ +17.6pp |
A governance signal that works only in specific market conditions has limited investment utility. Temporal stability across regimes is critical for institutional adoption.
| Archetype | n | % Univ. | 2024 ROE | 2-Yr Cum. | Debt | Loss Rate |
|---|---|---|---|---|---|---|
| Celestial | 95 | 11.3% | +7.7% | +15.1% | 123% | 7.4% |
| Poison Apple | 28 | 3.3% | +7.5% | +13.5% | 92% | — |
| Hidden Gem | 172 | 20.4% | +4.5% | +10.2% | — | — |
| Chameleon | 494 | 58.5% | +2.4% | +4.6% | 108% | — |
| Time Bomb | 55 | 6.5% | −5.6% | −12.4% | 155% | 32.7% |
| C–TB Gap | +13.3pp | +27.5pp |
Time Bomb loss rate (32.7%) is 4.4× the Celestial rate (7.4%). Revenue growth diverges significantly: Celestial +8.7% vs Time Bomb −5.0% (p < 0.001).
Poison Apple firms have nearly identical ROE to Celestials and the lowest debt ratio of any archetype. By every conventional financial screen, they appear healthy. Yet their governance structure — high transparency masking concentrated power — creates a fragility invisible to current-period financials. This is the failure mode that single-score ratings cannot detect: a firm that tells you everything but lets you decide nothing. Their executive pay ratio (7.5×), the highest of any archetype, corroborates the structural concentration of economic benefit in management.
KOSDAQ firms score higher on Transparency but dramatically lower on Balance and Risk. The pattern is counterintuitive — and consequential.
KOSDAQ has zero Celestial firms and 20.7% Poison Apples. For investors allocating to KOSDAQ, governance screening is not a refinement of analysis — it is a prerequisite. The base rate of governance risk is so high that deploying capital without governance due diligence is equivalent to accepting unscreened exposure to the D-grade majority.
Kill Switch firms do not look distressed in current financials. The governance data captures these risks before the balance sheet makes them obvious. Debt ratio difference: p = 0.03. ROE difference: p = 0.39 (n.s.), likely due to small sample.
The G-Score is not a trading signal. It is a pre-analysis filter — designed to help investors decide where to allocate analytical resources, and where not to.
What the evidence shows is that governance, when measured with sufficient granularity, carries economically meaningful and temporally stable predictive information about financial outcomes. The signal has held across nine independent periods spanning COVID recovery, rate hikes, and normalization.
For the Korean market specifically — where 54% of KOSPI firms and 80.5% of KOSDAQ firms receive D-grade scores — governance screening is not an optional refinement. It is a baseline requirement for informed capital allocation.
Universe: KOSPI 844 firms (2020–2025); KOSDAQ 1,255 firms (2023–2025). Backtest: 489–723 per period. 213 small-cap excluded (all D-grade).
Data: All inputs from DART. No surveys, self-assessments, or management interviews. Every data point independently verifiable.
Statistics: Bivariate correlations (KOSPI, n=630). *** p<0.001, ** p<0.01. Independent scoring per period — no forward-looking bias.
Limitations: Correlation ≠ causation. Survivorship bias. KOSDAQ insufficient for multi-period backtesting. Kill Switch n=17. Static annual scoring. Full discussion in Research Note #1.
Want the full methodology?
Research Note #1 contains all exhibits, statistical tests, and the complete limitations discussion.