# Contract — F3-conf-02

Question. How accurately could US quarterly real GDP growth (A191RL1Q225SBEA, % q/q SAAR) have been forecast 1 and 2 quarters ahead in real time, using only the vintage published at each origin; does a simple time-series model beat the naive (last value) forecast?

Estimands. Over 8 origins (vintage dates 2024-01-31 … 2025-12-31) and h ∈ {1,2} (quarters after the last quarter in the origin's vintage): RMSE and MAE of naive and of one model, vs first-release values of the targets; rel improvement 1 − RMSE_model/RMSE_naive per h; mean |current − first release| over the 8 h=1 target quarters. Meaningful under: first release is the evaluation target (not "truth"); 8 errors per horizon → very large sampling uncertainty; h=2 errors overlap with h=1 errors of the next origin (serially correlated).

Outputs. out/forecasts.csv (32 rows; columns origin_date,target_date,horizon,forecast,actual,model,vintage_date,train_last_date,train_last_value,n_train); out/results.json (task_id, model_name, model_description, n_origins, first_origin, last_origin, rmse/mae naive & model h1/h2, rel_improvement_rmse_h1/h2, mean_abs_revision_h1_targets, 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 out/exclusions.csv and out/state/*.

Information rules. At origin d only the vintage file dated d may be used (all values as in that file). current_vintage.csv only for revision statistic/discussion. Actuals = first_release.csv. Naive = last value of vintage, both horizons.

Invalidity / impossibility. Using current vintage or later vintages in training; a first-release vintage date not after the origin (would mean actual known at origin — check); missing first releases for any target; vintage last quarter mismatched with stated.