{ "task_id": "T1-forecast", "model_name": "BED (base-effect drift)", "model_description": "Base-effect drift model on the RSAFS vintage dated V only. Log level forecast ln L_{last+h} = ln L_last + h*mu, with mu = mean monthly log growth over the last 12 published months = g_last/12 (percent). Hence forecast g_T = 100*ln(L_last / L_{T-12}) + h*g_last/12 for T = last(V)+h, where L_last and L_{T-12} are read from vintage V (T-12 <= last(V) always). It exploits that 12-h of the 12 monthly growth terms in the target YoY are already published at the origin (known base effects), replacing only the h unknown months by their trailing 12-month average. No parameters are estimated; estimation window = last 13 levels of the vintage. PAYEMS and UMCSENT were not used. The specification was fixed before any out-of-sample error was computed; no tuning on the evaluation origins.", "n_origins": 10, "first_origin": "2025-10-31", "last_origin": "2026-07-31", "rmse_naive_h1": 0.996874361153601, "mae_naive_h1": 0.8487221999999999, "rmse_naive_h3": 1.4200403019751233, "mae_naive_h3": 1.2004486, "rmse_snaive_h1": 1.674542814799132, "mae_snaive_h1": 1.3066698, "rmse_snaive_h3": 1.3160788608225193, "mae_snaive_h3": 1.0658687, "rmse_ar1_h1": 0.940524354787796, "mae_ar1_h1": 0.7976509999999999, "rmse_ar1_h3": 1.1871176370926342, "mae_ar1_h3": 0.9526893999999999, "rmse_model_h1": 0.5502976390310066, "mae_model_h1": 0.48633190000000004, "rmse_model_h3": 1.1422196608203694, "mae_model_h3": 0.9339774000000001, "rel_improvement_rmse_h1": 0.4479769362367869, "rel_improvement_rmse_h3": 0.19564278617186803, "dm_stat_h1": 2.4289778933188826, "dm_pvalue_h1": 0.015141455540092164, "dm_stat_h3": 1.546577321548468, "dm_pvalue_h3": 0.12196520179653336, "conclusion_code_h1": "model_better_significant", "conclusion_code_h3": "model_better_not_significant", "mean_abs_revision_h1_targets": 0.5482198028388884, "feasibility": "complete", "feasibility_reason": null, "data_issues": [ { "issue": "Irregular publication timing: last(V) is two months before the origin for 2025-10-31 (last 2025-08), 2025-11-30 (last 2025-09), 2025-12-31 (last 2025-10), 2026-01-31 (last 2025-11), 2026-02-28 (last 2025-12), 2026-03-31 (last 2026-01); vintage 2026-04-30 adds 2026-02 and 2026-03 at once, so 2026-03 is never an h=1 target (it is the h=1 target of no origin) and 2026-02 is reached from the 2026-03-31 vintage.", "handling": "Targets defined as last(V)+h from the vintage content, per the task rule, not from the calendar." }, { "issue": "Benchmark/annual revision of RSAFS levels: the current vintage is ~4% lower in level and its YoY growth is systematically lower than first release; the base level used in first_release jumps between targets 2026-07 and 2026-08 (rebasing between the 2026-08-31 and 2026-09-30 vintages), affecting h=3 targets 2026-08 and 2026-09.", "handling": "No adjustment: actuals are first-release values by definition; the revision is quantified in mean_abs_revision_h1_targets and discussed." }, { "issue": "UMCSENT has 210 empty values (all before 1978, quarterly survey era) and is a current (not real-time) vintage.", "handling": "Not used by the model; would have been usable under the stated no-revision rule." }, { "issue": "PAYEMS vintages sometimes extend beyond last(V) of RSAFS (e.g. 2025-12-31 vintage ends 2025-11 while RSAFS ends 2025-10).", "handling": "Not used by the model." }, { "issue": "first_release.csv dates the first release of 2025-08 (last month of the 2025-10-31 vintage) to the unshipped 2025-09-30 vintage; its level in the 2025-10-31 vintage still equals level_first_release.", "handling": "None needed; checked for consistency." } ], "n_input_rows": { "first_release.csv": 28, "vintages/PAYEMS_vintage_2025-10-31.csv": 1040, "vintages/PAYEMS_vintage_2025-11-30.csv": 1041, "vintages/PAYEMS_vintage_2025-12-31.csv": 1043, "vintages/PAYEMS_vintage_2026-01-31.csv": 1044, "vintages/PAYEMS_vintage_2026-02-28.csv": 1045, "vintages/PAYEMS_vintage_2026-03-31.csv": 1046, "vintages/PAYEMS_vintage_2026-04-30.csv": 1047, "vintages/PAYEMS_vintage_2026-05-31.csv": 1048, "vintages/PAYEMS_vintage_2026-06-30.csv": 1049, "vintages/PAYEMS_vintage_2026-07-31.csv": 1050, "UMCSENT.csv": 887, "vintages/RSAFS_vintage_2025-10-31.csv": 404, "vintages/RSAFS_vintage_2025-11-30.csv": 405, "vintages/RSAFS_vintage_2025-12-31.csv": 406, "vintages/RSAFS_vintage_2026-01-31.csv": 407, "vintages/RSAFS_vintage_2026-02-28.csv": 408, "vintages/RSAFS_vintage_2026-03-31.csv": 409, "vintages/RSAFS_vintage_2026-04-30.csv": 411, "vintages/RSAFS_vintage_2026-05-31.csv": 412, "vintages/RSAFS_vintage_2026-06-30.csv": 413, "vintages/RSAFS_vintage_2026-07-31.csv": 414, "RSAFS_current.csv": 417 }, "n_analysed": 20, "random_seed": 20261030, "supplementary": { "bootstrap_rmse_diff": { "h1": { "rmse_diff_model_minus_naive": -0.44657672212259436, "ci95": [ -0.6982530105268047, -0.2612591327392867 ], "share_draws_model_better": 1.0, "block_length": 2, "n_boot": 9999, "dm_lrv": 0.8091373211601599, "n_origins_model_smaller_sq_error": 8 }, "h3": { "rmse_diff_model_minus_naive": -0.2778206411547539, "ci95": [ -0.6376736715848398, 0.14490466807245417 ], "share_draws_model_better": 0.8764876487648765, "block_length": 3, "n_boot": 9999, "dm_lrv": 2.1185178340561874, "n_origins_model_smaller_sq_error": 6 } }, "origins": [ { "origin": "2025-10-31", "last": "2025-08", "lag_months": 2 }, { "origin": "2025-11-30", "last": "2025-09", "lag_months": 2 }, { "origin": "2025-12-31", "last": "2025-10", "lag_months": 2 }, { "origin": "2026-01-31", "last": "2025-11", "lag_months": 2 }, { "origin": "2026-02-28", "last": "2025-12", "lag_months": 2 }, { "origin": "2026-03-31", "last": "2026-01", "lag_months": 2 }, { "origin": "2026-04-30", "last": "2026-03", "lag_months": 1 }, { "origin": "2026-05-31", "last": "2026-04", "lag_months": 1 }, { "origin": "2026-06-30", "last": "2026-05", "lag_months": 1 }, { "origin": "2026-07-31", "last": "2026-06", "lag_months": 1 } ], "revisions_h1_targets": { "2025-09": -0.5502167925401702, "2025-10": -0.6641763167214578, "2025-11": -0.4389199822781502, "2025-12": -0.3286268165544084, "2026-01": -0.9756759757612237, "2026-02": -0.40995563089387543, "2026-04": -0.7481728291014038, "2026-05": -0.16533992600871716, "2026-06": -0.7440888135516674, "2026-07": -0.45702494497780943 }, "ar1_note": "AR(1) re-estimated by OLS at each origin on the vintage, 1993-02..last(V)" } }