Not on Record, Recorded as Negative, or in Conflict: Typed Abstention in Ecuador's Administrative Register of Missing Persons (2026)

Resumen

Administrative registers of missing persons routinely produce three types of outcomes that conventional descriptive statistics treat as if they were one: cases with no outcome yet (genuine indeterminacy, characteristic of an ongoing search process), cases with an explicit negative outcome (for example, a report that was retracted or dismissed), and cases in which two extracts or two operational definitions of the same register yield incompatible figures. This article takes as its case study the missing-persons register of Ecuador's Ministry of the Interior, already documented and corrected in a previous round of factual auditing, and reframes it as a textbook example of what we call typed abstention: the formal distinction between "not on record" (no outcome has been recorded yet), "recorded as negative" (explicit denial), and "evidence conflict" (sources or definitions that do not match). Using the real, already-verified data - 4,892 records as of the August 19, 2026 extract, with 3,485 resolved cases (71.2%) and 1,407 under investigation (28.8%), against 4,507 records and the 742/3,528/237 breakdown (under investigation/located alive/deceased) of the July extract - we show that the 385-record discrepancy between extracts and the disparity among the 78.28%, 83.54%, and 71.2% resolution rates are not errors but instances of exactly the three types of indeterminacy that the evidential-reasoning literature (Dempster-Shafer theory, AGM belief revision, Belnap-Dunn four-valued logic) has been formalizing for decades with tools native to that tradition. We complement the analysis of the real data with an illustrative simulation study - explicitly synthetic, not derived from the real register - that quantifies, through 20,000 synthetic cases and 100 Monte Carlo replicates per scenario, how much of the aggregate gap between two extracts of the same register can be explained by asymmetric underreporting without any real contradiction between the sources, and how much information is lost when a binary "resolved/unresolved" scheme collapses genuine indeterminacy and negative closure into a single ambiguous category. The simulation results show that an underreporting asymmetry of just 10 to 20 percentage points between two extracts can produce aggregate gaps of 6.8 to 13.6 percentage points in the observed resolution rate - of the same order of magnitude as the real 12.3-percentage-point gap in the resolution rate between the July and August 2026 extracts - without either source being objectively wrong. We conclude that explicitly typing these three states, rather than collapsing them, improves the interpretability of public indicators without requiring complex inferential statistics or assumptions about the causes of the phenomenon.