Pre-registration
China's state-directed solar and wind manufacturing scale-up 2005-2020 delivered cost reductions on learning curves faster than any market-led OECD programme, demonstrating planning-led industrial policy's ecological potential.
Falsification criterion — what would disprove this
This hypothesis is considered falsified if:
The hypothesis is considered falsified if the pre-registered empirical test shows the opposite direction of the claim at conventional significance (p > 0.10), or if the primary outcome measure moves less than 10% in the claimed direction across the sample. Exact thresholds will be pinned in the variables and estimator blocks when this stub is promoted from draft.
formal test & threshold
test: Compare estimated solar-PV and wind-turbine learning rates (cost decline per cumulative-doubling) between CHN and OECD cohorts 2005-2020 (BNEF/IRENA series). Refute if CHN learning rate not statistically faster than OECD median at p<0.10.
Method
- Template
descriptive- Sample
- 10 countries · 2005 – 2020
- Evidence type
- associational
Two-cohort learning-rate comparison. Estimate solar-PV and wind-turbine learning rates (log cost on log cumulative capacity) separately for CHN and the OECD donor pool 2005-2020 using IRENA cost-and-capacity panels. Compare slope coefficients with bootstrap CIs. Descriptive — does NOT identify ownership-vs-demand-pull mechanism.
Data
| Variable | Source | Transform |
|---|---|---|
solar_pv_lcoe_usd_per_mwh outcome | irena:lcoe_solar_pvtier 2 | log |
wind_turbine_lcoe_usd_per_mwh outcome | irena:lcoe_wind_onshoretier 2 | log |
china_renewables_industrial_policy_indicator treatment | constructed:indicator = 1 for CHN from 2005 onwards (Renewable Energy Law) covering solar+wind manufacturing scale-uptier 5 | indicator |
cumulative_installed_capacity_log treatment | irena:installed_capacity_renewabletier 2 | log |
log_gdp_per_capita control | world_bank_wdi:NY.GDP.PCAP.KDtier 2 | log |
● ready · ● pending · ● reconstruct-needed
Detailed result card
Result card - china_renewables_industrial_policy_learning_curve
Verdict: INCONCLUSIVE_DATA_PENDING - global IRENA LCOE vintages support a transparent learning-curve diagnostic, but the original CHN-vs-OECD mechanism test remains blocked.
What Was Revived
The stale blocker is cleared for IRENA LCOE availability: local solar-PV and onshore-wind LCOE vintages now load. Because both LCOE files contain only country = World, this run does not grade the original China-vs-OECD claim. It records the narrower safe diagnostic: global LCOE learning curves against global installed capacity.
Results
- Solar PV: slope -0.663 (HC1 SE 0.027, p=0.0000); implied learning rate 36.8% per capacity doubling; n=15, 2010-2024.
- Onshore wind: slope -0.779 (HC1 SE 0.051, p=0.0000); implied learning rate 41.7% per capacity doubling; n=15, 2010-2024.
Specification
log(lcoe_usd_per_mwh) ~ log(global_installed_capacity_mw), estimated separately for solar PV and onshore wind with HC1 standard errors.
Remaining Blocker
The original pre-registered falsification test requires CHN and OECD cohort-specific LCOE. The exact local IRENA LCOE vintages are world-only, so the China industrial-policy mechanism still needs country/cohort LCOE from BNEF, IEA, or another source before it can be graded directly.
Local Data
- Solar PV LCOE:
data/vintages/irena/lcoe_solar_pv@2026-05-12T125721Z.parquet - Solar PV capacity:
data/vintages/irena/installed_capacity_solar_pv@2026-05-05T212314Z.parquet - Onshore wind LCOE:
data/vintages/irena/lcoe_wind_onshore@2026-05-12T125721Z.parquet - Onshore wind capacity:
data/vintages/irena/installed_capacity_wind@2026-05-05T212316Z.parquet
Generated by replication.py at 2026-05-15T18:06:42+00:00
Notes
Maps the eco-socialist school's China-renewables-learning-curve claim to a CHN vs OECD-cohort comparative learning-rate analysis. Estimator and prior set; full pre-registration awaits steelman + human sign-off.