Testing Objective Criteria Weights Against Equal Weights Under Serial Dependence: Size, Power, Decision Regret and a Saudi Arabia–China Risk-Factor Panel

Entropy and CRITIC weighting can produce unequal weights through sampling variation alone, so departures from equal weights call for a test. We propose stability-shrinkage entropy–CRITIC (SSEC), a fitted-null abstention test that keeps equal weights unless data-driven weights differ beyond bootstrap noise. It targets the weight functional, not equal importance. Cell-fitted size lies in the practical band (0.02–0.10) at all 27 points, with its interval wholly inside at 24 and wholly outside at none. Self-fitted size point estimates are in band at 25 of 27 points; among their exact intervals, 7 lie wholly inside and 20 cross a band limit. Strong-signal power exceeds 0.8 in the point estimate at all 15 assessed points, and its lower confidence bound exceeds 0.8 at 14. Outside the calibration family, SSEC is in band at 5 of 10 exact-null cells; criterion-specific persistence marks the calibration boundary. Among evaluated implementations, with the B-type tests in their registered whole-panel form, SSEC alone qualifies at every dependence level; fitted-null Brown–Forsythe qualifies at ϕ≤0.5. The gated policy has lower maximum rank-loss regret than data-driven weights. On a Saudi Arabia–China panel outside the demonstrated calibration domain, SSEC abstains in every sensitivity row, illustrating the decision rule.