1[2 {"id":"C1","text":"At the 1-month horizon the AR(2)-on-differences model has higher RMSE than the naive forecast (relative improvement about -0.024) and higher MAE.","results_keys":["rmse_naive_h1","rmse_model_h1","mae_naive_h1","mae_model_h1","rel_improvement_rmse_h1","n_evaluated_h1"],"status":"supported","caveat":"35 evaluated pairs; final-vintage data; one fixed model specification."},3 {"id":"C2","text":"At the 3-month horizon the AR model has slightly higher RMSE than naive (relative improvement about -0.008) and higher MAE.","results_keys":["rmse_naive_h3","rmse_model_h3","mae_naive_h3","mae_model_h3","rel_improvement_rmse_h3","n_evaluated_h3"],"status":"supported","caveat":"Overlapping errors; difference not statistically significant (DM p=0.18)."},4 {"id":"C3","text":"The simple AR model did not beat the naive forecast at either horizon over the 36 origins 2023-07 to 2026-06.","results_keys":["rel_improvement_rmse_h1","rel_improvement_rmse_h3","n_origins","first_origin","last_origin"],"status":"supported","caveat":"Pseudo-out-of-sample on revised data; short, calm window."},5 {"id":"C4","text":"The h=1 loss difference is nominally significant (HLN-corrected DM p about 0.016) while h=3 is not (p about 0.18).","results_keys":["diag_dm_stat_h1","diag_dm_pvalue_two_sided_h1","diag_dm_stat_h3","diag_dm_pvalue_two_sided_h3"],"status":"supported","caveat":"Forecast differences are hundredths of a point vs 0.1 pp data rounding; small sample; interpret cautiously."},6 {"id":"C5","text":"The estimated AR coefficients are small, so the model forecasts are close to a random walk.","results_keys":["diag_ar_coef_last_origin","diag_ar_nobs_last_origin"],"status":"supported","caveat":"Likely influenced by the 2020 COVID outliers in the estimation sample (not tested)."},7 {"id":"C6","text":"One observation (2025-10-01) is missing, leaving 35 evaluated pairs per horizon out of 36 origins; 941 observations are in the last training window.","results_keys":["data_issues","n_evaluated_h1","n_evaluated_h3","n_analysed","n_input_rows"],"status":"supported","caveat":""}8]9