Small Area Estimation (SAE) in Public Safety Statistics in Latin America

On Thursday, April 9, 2026, the virtual event “Application of Small Area Estimation (SAE) in the Field of Public Safety Statistics in Latin America” was held, organized within the framework of ECLAC’s Knowledge Transfer Network. The meeting brought together specialists from national and international organizations to discuss the use of SAE methodologies in the production of victimization and perceived insecurity statistics with greater territorial disaggregation.

During the event, it was highlighted that the growing demand for information at the local level—including municipalities, communes, and neighborhoods—poses significant challenges for National Statistical Offices, given that traditional surveys are designed primarily to produce national or regional estimates. In this context, SAE techniques emerge as a key tool for generating reliable estimates in small geographic areas without the need to significantly expand the sample size.

The session began with the presentation “SAE Methodology,” delivered by Andrés Gutiérrez (ECLAC), who outlined the conceptual and methodological foundations of small-area estimation, highlighting its role in generating disaggregated official statistics. He emphasized the importance of these methods for advancing the “leave no one behind” principle of the Sustainable Development Goals, as they enable the production of comparable indicators at different territorial levels by integrating surveys, censuses, and alternative sources such as satellite imagery. He also mentioned that the course “Disaggregation of Estimates in Small Areas using R” will be offered again in 2026.

Subsequently, the presentation “Victimization Survey” was delivered by Adriana Oropeza and Justo Rojas of the UNODC-INEGI Center of Excellence, addressing the role of these surveys in measuring public safety and their alignment with international standards such as the VICLAC Initiative. Their contribution to measuring key SDG 16 indicators—including victimization, perceptions of safety, and crime reporting—was highlighted, as was their potential for integration with SAE methodologies to expand the territorial disaggregation of data based on the integration of multiple sources and institutional capacities to consolidate a statistical ecosystem that links sociodemographic, territorial, and security information.

In the session on national experiences, the presentation “Public Safety and Justice” was delivered by José Antonio Gallegos, Director of Government Information Models at INEGI (Mexico). This presentation outlined the application of SAE methodologies to estimate perceptions of insecurity at the municipal level using the ENVIPE survey . Given that a significant proportion of municipalities lack sufficient sample size to generate accurate direct estimates, models based on the Fay-Herriot approach were implemented, including variants such as EBLUP and its spatial extension (SEBLUP). These models integrate auxiliary information from the census and administrative records—such as population density, employment, and crime rates—to improve the accuracy of the estimates. This approach enabled the generation of consistent and comparable statistical information at disaggregated territorial levels, strengthening the evidence base for the formulation of public safety policies.

Finally, from Chile, the presentation “Small Area Estimation (SAE) for Generating Estimates at the Communal Level of the Household Victimization by Violent Crimes (VHDV) Indicator in the 2024 National Survey on Crime and Safety (ENUSC)” was delivered by Jorge Arratia, a Methodological Analyst based at the ENUSC (INE Chile). This presentation demonstrated how the implementation of SAE enabled the generation of estimates at the municipal level, based on the results of the 2024 ENUSC, capturing the territorial heterogeneity of victimization and responding to the growing demand for local information to target public policies. Likewise, the use of predictive models and the incorporation of multiple information sources were highlighted to improve the accuracy of estimates within the framework of experimental statistics.

These estimates can be consulted on the website of the National Institute of Statistics of Chile, where the municipal results obtained through SAE for the VHDV of the 2024 ENUSC are presented. In this section, you will find the methodological summary, statistical tables, the methodological report, as well as the data visualizer.

Background, Coverage, and Methodology

Small Area Estimation (SAE) is a set of statistical techniques designed to generate reliable estimates for subpopulations or geographic areas that were not included in the original sample design of a survey. These methodologies combine survey data with auxiliary variables from censuses, administrative records, and other sources, thereby improving the accuracy of indicators at more disaggregated territorial levels.

In the field of public safety, their application is particularly relevant, as phenomena such as victimization and perceptions of insecurity exhibit high territorial heterogeneity that cannot be adequately captured by aggregate estimates. Experiences presented at the event show that the use of SAE enables the generation of local-level statistics, strengthens the evidence base for public policy, and improves the international comparability of indicators.

The development and application of these methodologies in Latin America are part of efforts to strengthen national statistical systems and advance the production of information aligned with international standards, including the VICLAC Initiative and SDG 16 indicators.

Relive the event—here you can find the recording and presentations: Videoconference on the Application of Small Area Estimation (SAE) in the Field of Public Safety Statistics in Latin America | Knowledge Transmission Network