Bermudez 2026 — Pre-Registered Out-of-Sample Test Battery for the Axiom of Containment

This registration locks a pre-registered out-of-sample test battery specifying three empirical tests of novel predictions generated by the Axiom of Containment (Bermudez, 2026a) and the SysMoD computational model (Bermudez, 2026b). The predictions were derived from the formal architecture of the theory before any data specified in the protocol were accessed.

Test 1 (Trust-Gap Inversion) predicts that federal employee engagement scores (OPM FEVS) and costly-to-fake externalities (EEOC complaint rates, retaliation rates, attrition) will diverge within agencies over time, with stronger divergence under greater accountability pressure (GAO High Risk List). Test 2 (Capability Threshold) predicts that agencies with severe historical reporting suppression will show weaker responsiveness to transparency reforms than agencies without such histories. Test 3 (Diagnostic Window) predicts that the graduation-proficiency gap observed in early years of a superintendent’s tenure in New York State school districts predicts the settled-state gap at tenure’s end, even after controlling for poverty, district size, and baseline conditions.

The protocol specifies all predictions, data sources, variable operationalizations, statistical tests, decision rules, measurement contingencies, and competing explanations before data access. Protocol v2 incorporates revisions from two independent adversarial methodological audit panels. The full protocol document is attached as a PDF.

Data sources: OPM FEVS (2010–2024), EEOC Form 462 (2010–2021), FedScope (2010–2023), GAO High Risk List, OSC Annual Reports, NYSED Administrator Salary Disclosure (2005–2026), NYSED Report Card and Graduation Rate Databases (2010–2024). No raw microdata from these sources were accessed before this registration.

Upload: Attach the PDF file (Bermudez_2026_PreRegistered_Test_Battery_v2_SSRN.pdf)

Tags/Keywords: pre-registration, Axiom of Containment, Nash-Moral Paradox, institutional concealment, out-of-sample prediction, FEVS, EEOC, education finance, Goodhart Gap, trust-gap inversion, capability threshold, diagnostic window