1[2 {"id":"C1","text":"The employment-augmented direct regression has lower out-of-sample RMSE than the naive forecast at h=1 (0.741 vs 0.814 pp; relative improvement ~8.9%).","results_keys":["rmse_model_h1","rmse_naive_h1","rel_improvement_rmse_h1"],"status":"supported","caveat":"Point estimate only; DM test p=0.19, not significant."},3 {"id":"C2","text":"At h=3 the model's RMSE is about 17% lower than the naive forecast's (0.947 vs 1.140 pp).","results_keys":["rmse_model_h3","rmse_naive_h3","rel_improvement_rmse_h3"],"status":"supported","caveat":"DM test p=0.22, not significant; 48 overlapping origins in a single episode."},4 {"id":"C3","text":"The model's MAE is also lower than the naive forecast's at both horizons.","results_keys":["mae_model_h1","mae_naive_h1","mae_model_h3","mae_naive_h3"],"status":"supported","caveat":"Same caveats as for RMSE."},5 {"id":"C4","text":"The improvement over the naive forecast is not statistically significant at conventional levels at either horizon.","results_keys":["dm_pvalue_h1","dm_pvalue_h3"],"status":"supported","caveat":"HAC-based DM test with HLN correction; approximate in small samples."},6 {"id":"C5","text":"The gain is attributable specifically to LFS employment information.","results_keys":[],"status":"not_evaluated","caveat":"No ablation without employment was run; the gain may partly reflect mean reversion captured by the regression."},7 {"id":"C6","text":"All forecasts respect the publication-lag rules (GDP <= t-2, employment <= t-1, training targets <= t-2).","results_keys":["last_target_date_available","last_predictor_date_available"],"status":"supported","caveat":"Timing only; current-vintage data, not real-time first releases."}8]9