1[2 {"id":"C1","text":"In real time over 8 origins (2024-01-31 to 2025-12-31), the AR(1) had lower RMSE than the naive forecast at h=1 (1.49 vs 2.01 pp, 26% improvement).","results_keys":["rmse_model_h1","rmse_naive_h1","rel_improvement_rmse_h1","n_origins","first_origin","last_origin"],"status":"supported","caveat":"Point estimate on 8 forecasts; not statistically significant (HLN-DM p=0.18)."},3 {"id":"C2","text":"At h=2 the AR(1) had lower RMSE than naive (1.40 vs 2.27 pp, 38% improvement) and the smaller absolute error at all 8 origins.","results_keys":["rmse_model_h2","rmse_naive_h2","rel_improvement_rmse_h2","n_model_smaller_abs_error_h2"],"status":"supported","caveat":"Not significant at 5% (HLN-DM p=0.11); overlapping errors."},4 {"id":"C3","text":"MAE also favours the AR(1) at both horizons.","results_keys":["mae_model_h1","mae_naive_h1","mae_model_h2","mae_naive_h2"],"status":"supported","caveat":"Same small sample."},5 {"id":"C4","text":"The advantage of the AR(1) over naive is not statistically significant at the 5% level at either horizon.","results_keys":["dm_stat_h1","dm_hln_pvalue_two_sided_h1","dm_stat_h2","dm_hln_pvalue_two_sided_h2"],"status":"supported","caveat":"DM test is itself unreliable with n=8."},6 {"id":"C5","text":"First releases of the h=1 target quarters have since been revised by 0.56 pp on average in absolute value.","results_keys":["mean_abs_revision_h1_targets"],"status":"supported","caveat":"Relative to the 2026-10-30 vintage; future revisions may change it."},7 {"id":"C6","text":"The AR(1) gain comes mainly from shrinking toward the post-1985 mean growth rate (low persistence, phi about 0.34).","results_keys":["last_origin_ar1_intercept","last_origin_ar1_phi","last_origin_ar1_implied_mean"],"status":"supported","caveat":"Interpretive; parameters shown for the last origin only."}8]9