# Contract (P0)

Question: At the end of each month t (48 origins, 2022-05..2026-04), how accurately can Canadian monthly real-GDP y/y growth be forecast for t+1 and t+3 using only data published by then, and does a model that adds LFS employment y/y growth beat the naive forecast? Estimands: RMSE and MAE (percentage points) over the 48 origins for h=1,3 of (a) naive = gdp_yoy[t-2], (b) one model; relative improvement 1 - RMSE_model/RMSE_naive. Meaningful under: fixed origin set, current-vintage data (simulates publication timing only, not real-time vintages), squared/absolute loss. Outputs: out/forecasts.csv (192 rows; origin_date,target_date,horizon,forecast,actual,model,last_target_date_used,last_predictor_date_used); out/results.json (keys per TASK.md table); out/report.md (Question; Data and exclusions; Method; Results; Uncertainty; Limitations and open issues); out/analysis.py; out/claims.json; plus protocol files (state/, exclusions.csv). Information rules: at origin t, GDP columns only for months <= t-2, employment only for months <= t-1; estimation, selection, transformations and plugged-in values obey this; training targets must be observed (<= t-2). Invalidity: any use of GDP > t-2 or employment > t-1; missing actuals for targets (latest target 2026-07 is available, so feasible); claims of real-time validity.