The aim of the project was to develop a regional approach to modelled demographic data and its characteristics. This would not only provide consistent, timely and inclusive data that could underpin development policy, but it would also provide a framework through which countries could request support for modelled population estimates and build sustainable capacity to be able to replicate the outputs.

Deliverables

Country-level support: The original plans for country-level support were to provide bi-lateral co-development to 2-3 countries based around their census requirements and the available input data. This type of co-development requires a significant investment of time to develop bespoke methods therefore the time constraints on the project necessitated a transition to a more regional approach. Countries could still be supported, however by running a regional workshop, it would allow a number of countries to be trained at once.

This regional technical workshop was delivered for the census and GIS leads of 10 Caribbean countries from the 9-14 March 2026 to introduce them to the methodological concepts of population modelling, and to train them in the use of the PyPopRF tool that can disaggregate statistical data using geospatial covariates. The workshop was organised jointly with UNFPA as part of their regional workshop on degree of urbanization. This approach was taken to improve the efficiency of delivery, reduce the burden on countries to participate and provide better value for money to the project. This regional approach also allowed WorldPop to start discussing country level requirements with countries.

Across the region, participating countries demonstrated strong engagement with population modelling tools and gridded data methods, with the majority successfully running the pypopRF model during training exercises — in several cases using national datasets to assess real-world applicability. Whilst census data remains the preferred primary source for most national statistical offices, there is broad recognition of the value of gridded population estimates for addressing persistent data gaps, particularly in hard-to-enumerate communities such as gated developments, and for producing inter-censal population projections. Technical capacity and data availability were identified as common constraints to fuller implementation, and a clear appetite exists across the region for continued capacity development, including in-country training and the establishment of a community of practice. In preparation for the 2030 census round, interest in shorter questionnaire designs was widely expressed, alongside emerging discussion of weighting methodologies and the integration of modelling approaches to complement short-form data collection. Where modelling has already been employed during previous census exercises — including to address under-coverage and low response rates — countries highlighted the likelihood that such methods will again be required, underscoring the need for sustained technical support and regional knowledge-sharing in the years ahead

Census Strategy: The Terms of Reference for a census strategy consultant were provided to WorldPop in January 2026 for review. WorldPop were able to review these through the lens of modelled population data, but were also able to forward the ToRs to ONS who could provide a more comprehensive set of comments given their experience with census delivery. The census consultant was appointed in January and is expected to deliver the inception report to CARICOM by the end of February 2026. The draft report was circulated by the end of March and this will be reviewed and finalised by June 2026.

COD-PS: An initial modelled output has now been produced for the region . The datasets provided by UNFPA were explored, cleaned, prepared, and processed. A total of 49 geospatial covariates were used and building footprints metrics were ranked by the model as the most important predictors. Each grid cell contains between 0-698 people per pixel and there is no indication of extreme values. The initial model achieved an explained variance of approximately 94%. Age–sex disaggregated raster datasets were produced for commonly reported demographic groups (5-year age groups) as well as an additional layer for females aged 15–49 (women of childbearing age). These should help to support decision-making in the region.

 Work is still needed to finalise the disaggregated data and the processing script will also be shared to ensure that these outputs are open and repeatable by UNFPA, CARICOM or other agencies.