Multi-year process-drift model with year terms

The current drift estimate is a single-transition variance decomposition (FY2024→FY2025: Var(Δ)=2·mean(SD²)+τ²; τ 1.07–1.62pp). Published state PERs exist for many prior years with known sampling designs — fit a proper longitudinal model (state effects + year effects + AR(1)/random-walk process term, sampling variances known from simulated SDs at each year's sample sizes) to estimate drift and persistence from the full panel instead of one transition. Output feeds the simulator's optional drift band. Pre-register the estimator before touching the realized series.

Start by locating the simulator's optional drift-band input and the available multi-year state PER data, including each year's simulated sampling variances. Define and pre-register the longitudinal estimator before using the realized series; done means estimating state effects, year effects, persistence, and drift from the full panel for the simulator.