1[2 {"id":"C1","text":"At h=1 the base-effect drift model has a lower RMSE than the naive forecast, with a relative improvement of about 45%, and the DM test (Bartlett LRV, normal) rejects equal accuracy at 5%.","results_keys":["rmse_model_h1","rmse_naive_h1","rel_improvement_rmse_h1","dm_stat_h1","dm_pvalue_h1","conclusion_code_h1"],"status":"supported","caveat":"Only 10 origins; with the HLN small-sample correction and t(9) values, p = 0.047, which is borderline."},3 {"id":"C2","text":"At h=3 the model has a lower RMSE than naive (about 20% improvement), but the difference is not statistically significant.","results_keys":["rmse_model_h3","rmse_naive_h3","rel_improvement_rmse_h3","dm_stat_h3","dm_pvalue_h3","conclusion_code_h3","boot_ci95_lo_h3","boot_ci95_hi_h3"],"status":"supported","caveat":"Overlapping 3-step errors and n = 10 make the DM long-run variance imprecise."},4 {"id":"C3","text":"Among the baselines, AR(1) improves only slightly on naive at both horizons; seasonal naive is the worst at h=1 but beats naive at h=3.","results_keys":["rmse_ar1_h1","rmse_ar1_h3","rmse_snaive_h1","rmse_snaive_h3","rmse_naive_h1","rmse_naive_h3"],"status":"supported","caveat":"No significance test was carried out among the baselines."},5 {"id":"C4","text":"The first-release YoY growth of the h=1 target months has since been revised by 0.548 pp on average in absolute value, and every one of these revisions is downward.","results_keys":["mean_abs_revision_h1_targets","mean_revision_h1_targets_current_minus_first","revisions_h1_targets"],"status":"supported","caveat":"Reflects the annual benchmark revision present in the 2026-10-30 current vintage; it may be revised again."},6 {"id":"C5","text":"The block bootstrap 95% interval for RMSE_model - RMSE_naive excludes zero at h=1 and includes zero at h=3.","results_keys":["boot_ci95_lo_h1","boot_ci95_hi_h1","boot_ci95_lo_h3","boot_ci95_hi_h3","random_seed"],"status":"supported","caveat":"A bootstrap on 10 observations is crude."}7]8