1[2 {"id":"C1","text":"The naive forecast gdp_yoy(t-2) has RMSE 0.814 pp at h=1 and 1.140 pp at h=3 over 48 origins (MAE 0.642 and 0.872).","results_keys":["rmse_naive_h1","rmse_naive_h3","mae_naive_h1","mae_naive_h3","n_origins"],"status":"supported","caveat":"Current-vintage data; timing simulation only."},3 {"id":"C2","text":"The employment-augmented direct ADL model has lower RMSE and MAE than the naive forecast at both horizons (RMSE 0.743 and 0.902 pp; relative improvement 0.087 at h=1 and 0.208 at h=3).","results_keys":["rmse_model_h1","rmse_model_h3","mae_model_h1","mae_model_h3","rel_improvement_rmse_h1","rel_improvement_rmse_h3"],"status":"supported","caveat":"Point estimates only; see C3."},4 {"id":"C3","text":"The improvements are not statistically significant: Diebold-Mariano (HLN-corrected) p-values are about 0.19 (h=1) and 0.18 (h=3).","results_keys":["dm_test"],"status":"supported","caveat":"Low power with 48 overlapping errors."},5 {"id":"C4","text":"How large the gain is depends on the COVID-period training exclusion; without it the relative RMSE improvement is 0.012 at h=1 and 0.304 at h=3, but the model still beats naive at both horizons.","results_keys":["sensitivity_no_covid_exclusion"],"status":"supported","caveat":"Sensitivity only, not the headline model."},6 {"id":"C5","text":"All forecasts use GDP only through t-2 and employment only through t-1; at the last origin these are 2026-02-01 and 2026-03-01.","results_keys":["last_target_date_available","last_predictor_date_available","last_origin"],"status":"supported","caveat":"Enforced in code and verified in out/state/checks.md."}7]8