# Contract — F3-conf-02

Question: How well could US quarterly real GDP growth (q/q % SAAR, A191RL1Q225SBEA) have been forecast 1 and 2 quarters ahead in real time at 8 origins (2024-01-31 … 2025-12-31), and does a simple time-series model beat the naive forecast?

Estimands: RMSE and MAE across the 8 origins, per horizon h∈{1,2}, of (a) the naive forecast (last value in the origin vintage) and (b) one model (chosen: AR(1), OLS, iterated); the forecast errors are measured against the first-release value of the target quarter. Also rel. improvement 1−RMSE_model/RMSE_naive per h, and the mean |current − first release| over the 8 h=1 targets. Meaningful only as descriptive real-time accuracy over 8 origins (very small sample; no significance claims).

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; out/analysis.py; out/claims.json.

Information rules: at origin d use ONLY vintage file d (all values as in that file). current_vintage.csv is used only for the revision statistic. Targets = last quarter in file + h. Actuals = first_release.csv.

Invalidity/impossibility: missing vintage or first release for a target; using current vintage or later vintages in estimation; selecting the model specification on the basis of out-of-sample results.