Do Artificial Masks Represent Natural Missingness? External-Reference Evaluation in County-Level Power-Outage Records

Artificial masking is widely used to evaluate methods for missing records, but masked observations may not resemble records that are naturally absent. This study examines county-level EAGLE-I power-outage data for nine Florida events (2018-2025), using Florida Public Service Commission reports as a separately compiled external reference measurement rather than ground truth. It preserves four record states, compares artificial masks and natural absences against the same reference, and reports coverage and common-support diagnostics. Natural absences had higher zero shares and lower upper-tail reference outages than positive records. Full-coverage zero-fill contrasts were negative in all nine events, meaning larger reference error on the artificial masks. Selective comparisons were constrained by unequal issuance, weak overlap, and limited common support. No learned estimator was justified or reported. The evidence is Florida-only, event-specific, and descriptive.