Crossing the capability threshold: ecosystem capacity and the transmission of productivity growth to manufacturing

Reproduction package for the paper "Crossing the capability threshold: ecosystem capacity and the transmission of productivity growth to manufacturing" (S. Kim and Y.-W. Sawng). It contains the country-year panel, an R script that reproduces every number reported in the paper, and documentation. Research question. The catch-up literature holds that technology becomes growth only once an economy has accumulated sufficient complementary capability, and it has been tested mainly where that threshold visibly binds. We ask whether the mechanism switches off once an economy is rich: does the transmission from productivity growth to manufacturing growth still rise with accumulated ecosystem capability among OECD members, and is any moderation specific to the productivity path? Data. One country-year panel of 24 OECD economies, 2002-2019 (432 rows; 374 used after first-differencing, lagging, and one incomplete series). Variables: manufacturing value-added growth (OECD STAN); TFP growth (Conference Board TED); lagged changes in a patent-based technology index and an export-sophistication index as controls; and three capability measures entered as country period means - business R&D share, institutional quality (first principal component of the six World Bank Worldwide Governance Indicators), and researchers per million (OECD MSTI, UNESCO UIS). Total R&D intensity is included for reference. Standardisation, index construction, terciles, lags and interactions are performed inside the script, so every derived variable is transparent from the raw columns. Findings. With country and year fixed effects and country-clustered standard errors, productivity growth is the dominant proximate driver of manufacturing growth (coefficient 1.33). Its interaction with the composite capability index is 0.404 per standard deviation, significant at one percent. By capability tercile the implied transmission coefficient is 1.14, 1.40 and 2.09: the bottom two thirds are statistically indistinguishable and the top third stands apart, the shape a threshold implies rather than a smooth gradient. The moderation is specific to the productivity path; replacing TFP growth with the lagged technology change yields 0.009 (SE 0.026). Results hold under principal-component weighting, for each dimension separately, for total R&D intensity, under alternative groupings and specifications, in all 24 leave-one-out regressions, and under a wild cluster bootstrap (p = 0.025). Interpretation and use. Results are conditional associations under a two-way fixed-effects design, not causal parameters: identification comes from interacting a time-varying regressor with a fixed country characteristic, so it recovers differences in the productivity-growth slope across countries, not the effect of accumulating capability within one. Running P3_verify_EINT.R in R (4.0+) with fixest and dplyr reproduces every number in the paper; each result prints next to the published value, flagged OK or DIFFERS.