Auditing Scoring Architecture in Financial Capability Surveys: Monte Carlo Evidence and a Peruvian Workplace Case on Perceived Financial Freedom

Abstract

Self-report financial surveys require auditable scoring as well as evidence that their items measure the intended constructs. This study combines a Peruvian workplace archive of 50 records with an independently reconstructed Monte Carlo experiment. The audit distinguishes archived categories, direct sums, a historical partial key and correction of all four explicit constraint items. Full correction changes the overall correlation from −0.081 to −0.331 (record-level permutation p = 0.01930), while perceived-financial-freedom alpha falls from 0.708 to 0.255. The archive contains only 20 distinct response patterns; a pattern-cluster sensitivity interval includes zero. Archived categories are not deterministic functions of the stored sums. The simulation evaluates 82,500 synthetic datasets across a 27-cell baseline and 28 targeted extension cells. It separates measurement-related displacement, protocol distortion and finite-sample error. At N = 150 and a latent correlation of 0.40, the empirical key-map scenario yields a joint correlation change of −0.158 and a paired interaction of −0.041 (MCSE 0.0020). Mean observed-target baseline coverage is 0.949 for corrected sums and 0.948 for direct categories; latent-target inclusion is a different diagnostic. The contribution is an operational audit combining software invariants, item and provenance checks, and conditional robustness assessment. Scoring repair and synthetic evidence do not validate the source instrument or establish a substantive financial-capability effect.