Erasure Skew is a proposed provenance-fairness measure by Nobel Glas.
Where Provenance Erasure Rate measures how much attribution is lost, Erasure Skew measures whether the loss is associated with source power. It is estimated as the slope of per-source provenance retention on a declared power coordinate such as a Retrieval Capital proxy.
A positive value means higher-power sources retain more attribution than lower-power sources. A value near zero means retention is not associated with power. The paper recommends permutation testing and reports the measure as undefined when there are too few sources or no variation in power.
The deposit demonstrates the estimator on synthetic data and pre-registers tests on real retrieval or summarization output. It does not report those field results and does not infer intent from a positive score.