# Amendment 02 — price forecast stability (development-only observation)

After introducing funded exploration (Amendment 01), an unseen good entered with essentially zero past trading price. A one-step naive last-price forecaster produced exceptionally large price-and-output feedback and extreme regrets in seeds 11–13. These are genuine outcomes of the unbounded forecast rule and remain reproducible by setting `forecast_cap_multiplier=0`.

For *new subsequent trials* we adopt a common ex-ante, static prior-price band `[prior/4, prior*2.5]` for all institutions' production forecasts, with `prior` computed only from parameters known at time zero. This guards an ill-conditioned observed-price initialization for previously unavailable goods, and not a treatment-label-specific rescue. It does, however, alter the economic environment after pilot inspection; results must be called amended exploration and compared against an uncapped sensitivity arm. No claim of confirmatory preregistration may use the pilot data that motivated this modification.