1# Contract (P0) — F3-conf-0423**Question.** Over 40 expanding-window origins t = 2015Q4 … 2025Q3, does a chosen time-series model forecast Canadian quarterly CPI inflation (400·Δln of the quarterly-average NSA CPI, % annual rate) better than the seasonal-naive forecast (value of quarter t+h−4) at h = 1, 2, 4; and at h = 4 is the mean squared-error loss differential different from zero once overlap-induced serial correlation is accounted for?45**Estimands.** Per horizon and model: out-of-sample RMSE and MAE over the 40 origins (pp, annualised); relative improvement 1 − RMSE_model/RMSE_snaive. At h=4: d̄ = mean(e²_snaive − e²_model), HAC (long-run variance) SE with truncation lag ≥ 3, DM statistic d̄/SE and two-sided p-value. Meaningful under: covariance-stationarity of d_t (DM assumption), the data as given (current vintage, not real-time), a fixed evaluation sample of 40 points (small; asymptotic normal approximation is rough).67**Outputs.** out/forecasts.csv (240 rows: origin_date, target_date, horizon, forecast, actual, model); out/results.json with keys task_id, model_name, model_description, n_origins, first_origin, last_origin, rmse/mae_{snaive,model}_h{1,2,4}, rel_improvement_rmse_h{1,2,4}, loss_diff_mean_h4, loss_diff_se_h4, hac_lags_h4, hac_kernel_h4, dm_stat_h4, dm_pvalue_h4, feasibility, feasibility_reason, data_issues, n_input_rows, n_analysed, random_seed; 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).89**Information rules.** Forecast at origin t uses only quarters ≤ t (estimation, selection, transformations). For direct h-step regressions, the training pairs (x_s, y_{s+h}) must satisfy s+h ≤ t. Snaive uses quarter t+h−4 ≤ t. No real-time claim.1011**Invalidating conditions.** Missing target quarters (t+h up to 2026Q3 needed — present), look-ahead in estimation, HAC lag < 3, data hash mismatch with README.12