Heterogeneidad socioeconómica y protección económica de los adultos mayores en Ecuador: segmentación y clasificación mediante aprendizaje automático a partir de la ENEMDU 2025
Year: 2026
Type: article
Study design: Observational Study
Abstract: This study examined socioeconomic heterogeneity and the labor-income and pension dimensions of economic protection among older adults in Ecuador using the 2025 annual National Survey of Employment, Unemployment, and Underemployment (ENEMDU). The analytical dataset comprised 45,959 records for adults aged 65 years or older. Applying the annual survey weight fexp yielded a population estimate of 1,855,076 (approximately 1.86 million) for January–December 2025; this is not a count of distinct respondents, and some person–household panel positions were observed more than once. K-means identified four profiles: retirees with limited employment (26.0% weighted), higher-income workers (1.8%), lower-income workers (35.7%), and inactive adults without pensions or employment (36.5%). Clustering stability was high (mean adjusted Rand index = 0.9905). Supervised models used only sociodemographic predictors, excluding the economic variables used to form the clusters. On the row-level test set, Random Forest achieved macro-F1 = 0.4709 and accuracy = 0.5504, compared with 0.4676 and 0.5424 for XGBoost. For XGBoost, raw macro ROC-AUC was 0.788 and macro average precision 0.505; training-selected isotonic calibration reduced the Brier score from 0.585 to 0.524. An additional panel-key-grouped evaluation yielded macro-F1 = 0.4708 for Random Forest but did not externally validate cluster formation. The typology offers an integrated descriptive view of work and pension protection without supporting causal or individual eligibility claims.
Institution: Universidad Estatal de Milagro
Language: English
Subfield: Life-span and Life-course Studies
Field: Social Sciences
Domain: Social Sciences
Open Access status: diamond
OpenAlex ID: W7225983125
