Will South Korea GDP growth in Q2 2026 be between 3.0% and 3.4%?

Note: only 6/8 ensemble models completed on this run.

Analysis

AGREEMENT: All six agents agree on the core facts — Q1 2026 YoY was 3.8% (strongest since 2020), Q2 2026 export data is exceptionally strong (June +70.9% YoY, semiconductors ~+200%), full-year institutional forecasts (BOK/KDI/OECD 2.5-2.6%) imply deceleration, and Trading Economics points to 3.5% for Q2. All six also land BELOW the market's 49.5%, and 5 of 6 said they would bet. DISAGREEMENT: The 1% estimates from Agents 1 and 5 are miscalibrated outliers — Agent 5's own evidence list (strong momentum) supports a directional view that growth exceeds 3.4%, but even so, 1% is absurdly low for a bucket adjacent to every plausible central estimate; a QoQ outcome of 0.0-0.3% is entirely plausible given payback dynamics. These drag the ensemble mean (26.8%) below what the evidence supports; the median (36.5%) is far more sensible. EVIDENCE QUALITY: Agents 3 and 6 have the strongest work — they did the explicit base arithmetic (Q2 YoY ≈ 3.1% + Q2 QoQ), identified that the bucket requires near-zero sequential growth, and noted the price-vs-volume caveat on semiconductor exports (which cuts both ways). Agent 4's 47% is largely market-anchored with little independent nowcast. ENSEMBLE CHECK: The mean is an artifact of two broken 1% estimates; the credible cluster is 35-47%. ANTI-CONFIRMATION: For the market's 49.5% to be right, the true Q2 QoQ distribution must be tightly centered at ~+0.1-0.2% — possible via payback, but the available nowcasts (TE 3.5%, ING 'well above 3.0%', record June exports through quarter-end) suggest the center is nearer 3.4-3.6%, splitting mass between this bucket and 3.5-3.9%. A 0.5pp band rarely deserves ~50% unless the nowcast is dead-center, and the visible evidence says it isn't. FINAL: 36%, deviating from the polluted ensemble mean toward the credible-agent median. Edge vs market is ~13.5% on the NO side with specific, verifiable evidence (export data, TE/ING nowcasts, base arithmetic). Recommendation: TRADE_NO, with moderate confidence — the main risk is a sharp Q1 payback that the market's Korean-source traders may be pricing better than my agents.

View on Polymarket