Bee Colony-Loss Forecasting Benchmark: a pre-registered out-of-sample test on USDA NASS data

A pre-registered out-of-sample forecasting benchmark for U.S. honey-bee colony loss built on public USDA NASS Honey Bee Colonies data. Reported stressor prevalence (Varroa, disease, pesticides) adds no statistically detectable marginal forecasting value over a model of past loss; the only out-of-sample skill is recalibration of a seasonal-naive baseline. Releases a cleaned state-by-quarter panel (2015–2025) and a frozen forecasting benchmark. Code and pre-registration: https://github.com/BenMalaga/bee-colony-loss-forecast