Typed Abstention in Administrative Disability Registries: Certification Intensity, Subregistration, and the Limits of Cantonal Normalization in Ecuador's National Disability Registry (2023)

Abstract

Administrative disability registries record certified cases, not disability as such: a person absent from the registry may lack certification because no certifiable impairment exists, because access to an evaluation board was unavailable, or because an evaluation is pending - three epistemically distinct situations that a single observed count cannot, by itself, distinguish. This paper reframes a nationwide descriptive analysis of Ecuador's National Disability Registry (Registro Nacional de Discapacidades, CONADIS/MSP; cut-off December 2023; N = 113,203 aggregated records representing 480,776 certified persons across 222 province-canton units) as an instance of typed abstention applied to administrative subregistration. We distinguish "not registered" (no consta: absence of a positive record, uninformative about the underlying condition) from a hypothetical "registered negative" (consta que no: an explicit denial after evaluation, a state that exists within the underlying certification process but is not exposed in the published open-data resource used here) and from internally conflicting evidence. Data quality auditing verified 100% structural completeness across all eight operational fields. The demographic and clinical profile shows a predominance of males (56.48%), adults aged 51 years and older (47.13%), physical disabilities (44.87%), and moderate impairment severity (45.78%). Guayaquil and Quito concentrate 30.83% of absolute registration volume, yet fall to the 134th and 144th of 221 population-normalized ranks, below the national median. Because the published registry cannot, by construction, separate certification capacity from underlying need, we add an illustrative synthetic simulation in which two data-generating mechanisms - pure certification-capacity heterogeneity and pure prevalence heterogeneity, each confounded with population size - produce observable correlations between the normalized rate and population that differ by only 0.021 (Pearson r = 0.949 vs. 0.928) despite sharing no causal mechanism, formalizing why the normalized rate cannot be read as need without an explicit, currently unavailable, capacity signal. We situate this identification problem within the literature on disability statistics (the Washington Group on Disability Statistics; the World Health Organization's International Classification of Functioning, Disability and Health), administrative and missing-data quality, and evidential reasoning frameworks for typed indeterminacy (Dempster-Shafer belief functions, three- and four-valued logics, AGM belief revision), none of which have previously been applied to a cantonal-level disability registry of this kind.