1{2 "task_id": "T1-forecast",3 "model_name": "BE-RWD (base-effect random walk with drift in log level)",4 "model_description": "Univariate real-time model on the RSAFS vintage dated V only. The log level is forecast as a random walk with drift: ln L_hat(last+h) = ln L(last) + h*d, with d = g_last/1200, i.e. the average monthly log growth over the 12 months ending at last(V) in that vintage. The YoY forecast is 100*(ln L_hat(T) - ln L(T-12)), where L(T-12) is read from the same vintage, so the forecast equals the naive forecast plus a known base-effect correction (the change in the year-earlier comparison level) and drift. No parameters are estimated; PAYEMS and UMCSENT are not used. Specification fixed before looking at any evaluation outcome; no alternative models were tried on the evaluation sample.",5 "n_origins": 10,6 "first_origin": "2025-10-31",7 "last_origin": "2026-07-31",8 "rmse_naive_h1": 0.9968743518607954,9 "mae_naive_h1": 0.848722188000456,10 "rmse_naive_h3": 1.4200403737520462,11 "mae_naive_h3": 1.200448633787668,12 "rmse_snaive_h1": 1.6745428991342943,13 "mae_snaive_h1": 1.3066698762543274,14 "rmse_snaive_h3": 1.3160788989552372,15 "mae_snaive_h3": 1.0658686914791946,16 "rmse_ar1_h1": 0.9405244111185227,17 "mae_ar1_h1": 0.7976510573253066,18 "rmse_ar1_h3": 1.1871178147422958,19 "mae_ar1_h3": 0.952689596144294,20 "rmse_model_h1": 0.5502977033274435,21 "mae_model_h1": 0.4863319392755308,22 "rmse_model_h3": 1.1422197516272885,23 "mae_model_h3": 0.9339774440853781,24 "rel_improvement_rmse_h1": 0.44797686659282443,25 "dm_stat_h1": 2.4289778515909086,26 "dm_pvalue_h1": 0.015141457282753734,27 "conclusion_code_h1": "model_better_significant",28 "boot_rmse_diff_ci95_lo_h1": 0.2615951068646396,29 "boot_rmse_diff_ci95_hi_h1": 0.6987722437362641,30 "share_origins_model_smaller_sq_error_h1": 0.8,31 "rel_improvement_rmse_h3": 0.19564276288194327,32 "dm_stat_h3": 1.5465768014315027,33 "dm_pvalue_h3": 0.12196532729732468,34 "conclusion_code_h3": "model_better_not_significant",35 "boot_rmse_diff_ci95_lo_h3": -0.14490476470017666,36 "boot_rmse_diff_ci95_hi_h3": 0.6265360274461692,37 "share_origins_model_smaller_sq_error_h3": 0.6,38 "mean_abs_revision_h1_targets": 0.5482198028389015,39 "feasibility": "complete",40 "feasibility_reason": null,41 "data_issues": [42 {43 "issue": "Irregular RSAFS publication timing: last(V)=2025-08 at V=2025-10-31 and 2026-01 at V=2026-03-31 (two-month lag); V=2026-04-30 adds two months at once.",44 "handling": "Targets defined as last(V)+h per task rule; the 2025-10-31 and 2026-03-31 origins therefore forecast months that are 2 and 4 months past the last calendar month."45 },46 {47 "issue": "h=3 forecast errors of consecutive origins overlap (and some h=1 and h=3 target months coincide), inducing serial correlation in the loss differentials; only 10 origins are available.",48 "handling": "DM long-run variance uses Bartlett kernel with truncation lag h-1 as specified."49 },50 {51 "issue": "Large level revision of RSAFS after the 2026-07-31 vintage (current levels ~4-5% below vintage levels); first_release.csv shows the break between targets 2026-07 and 2026-08.",52 "handling": "YoY always computed within one vintage; h=3 actuals for targets 2026-08/2026-09 are post-revision first releases, used as given; estimand (d) reflects this revision."53 },54 {55 "issue": "PAYEMS benchmark revision visible in the 2026-02-28 vintage; PAYEMS 2025-10 first appears in the 2025-12-31 vintage.",56 "handling": "PAYEMS not used in any forecast; documented only."57 },58 {59 "issue": "UMCSENT has 210 empty values before 1978-01.",60 "handling": "UMCSENT not used in any forecast; documented only."61 }62 ],63 "n_input_rows": {64 "RSAFS_current.csv": 417,65 "UMCSENT.csv": 887,66 "first_release.csv": 28,67 "vintages/PAYEMS_vintage_2025-10-31.csv": 1040,68 "vintages/PAYEMS_vintage_2025-11-30.csv": 1041,69 "vintages/PAYEMS_vintage_2025-12-31.csv": 1043,70 "vintages/PAYEMS_vintage_2026-01-31.csv": 1044,71 "vintages/PAYEMS_vintage_2026-02-28.csv": 1045,72 "vintages/PAYEMS_vintage_2026-03-31.csv": 1046,73 "vintages/PAYEMS_vintage_2026-04-30.csv": 1047,74 "vintages/PAYEMS_vintage_2026-05-31.csv": 1048,75 "vintages/PAYEMS_vintage_2026-06-30.csv": 1049,76 "vintages/PAYEMS_vintage_2026-07-31.csv": 1050,77 "vintages/RSAFS_vintage_2025-10-31.csv": 404,78 "vintages/RSAFS_vintage_2025-11-30.csv": 405,79 "vintages/RSAFS_vintage_2025-12-31.csv": 406,80 "vintages/RSAFS_vintage_2026-01-31.csv": 407,81 "vintages/RSAFS_vintage_2026-02-28.csv": 408,82 "vintages/RSAFS_vintage_2026-03-31.csv": 409,83 "vintages/RSAFS_vintage_2026-04-30.csv": 411,84 "vintages/RSAFS_vintage_2026-05-31.csv": 412,85 "vintages/RSAFS_vintage_2026-06-30.csv": 413,86 "vintages/RSAFS_vintage_2026-07-31.csv": 41487 },88 "n_analysed": 20,89 "random_seed": 2026103090}