Korean smart farm subsidies have expanded rapidly since 2018, with the seedling industry — the upstream supplier of seedlings for downstream crop production — a strategic policy target. Conventional wisdom holds that smart farm adoption produces catch-up effects benefiting low-revenue farms most, justifying targeted subsidies for low performers. Using census-level microdata on all 1,567 registered Korean seedling firms, we initially find an apparent catch-up pattern: estimated effects of facility adoption decline monotonically from 9.7% at q10 to 3.8% at q90 of sales-per-㎡. However, a diagnostic re-estimation reveals this pattern is largely a composition artifact. Once we control for business type (Q7: specialized vs. diversified seedling firms), the catch-up gradient vanishes (test q10 = q90: p = 0.640). Sub-sample analysis shows that diversified firms gain 5.8× more from adoption than specialized firms (β = 0.085 vs. 0.015, interaction p = 0.059), with the two groups exhibiting opposite distributional patterns: diversified firms show middle-quantile concentration (peak at q25, β = 0.148), while specialized firms show modest frontiershift (q90 effect, p = 0.090). The true structure of heterogeneity is therefore two-dimensional — by business model and within-model distribution — rather than catch-up along the revenue distribution. We construct a Smart Seedling Production Index (SSPI) following Porter and Heppelmann's (2014) three-tier capability framework, and validate identification through peer-effect IV placebos, Durbin-Wu-Hausman tests, and wild cluster bootstrap inference. The findings imply that targeting low-revenue farms based on apparent catch-up is misguided; effective policy should adopt two-dimensional targeting that channels subsidies to diversified firms in the mid-revenue range while supporting Tier-3 advanced technologies for specialized firms operating near the technological frontier.
