Social Media Data for Population Mapping: A Bayesian Approach to Address Representativeness and Privacy Challenges

Accurate and timely population data are essential for disaster response and humanitarian planning, but traditional censuses often cannot capture rapid demographic changes. Social media data offer a promising alternative for dynamic population monitoring, but their representativeness remains poorly understood and stringent privacy requirements limit their reliability. Here, we address these limitations in the context of the Philippines by calibrating Facebook user counts with the country’s 2020 census figures. First, we find that differential privacy techniques commonly applied to social media-based population datasets disproportionately mask low-population areas. To address this, we propose a Bayesian imputation approach to recover missing values, restoring data coverage for $5.5%$ of rural areas. Further, using the imputed social media data and leveraging predictors such as urbanisation level, demographic composition, and socio-economic status, we develop a statistical model for the proportion of Facebook users in each municipality, which links observed Facebook user numbers to the true population levels. Out-of-sample validation demonstrates strong result generalisability, with errors as low as ${\approx}18%$ and ${\approx}24%$ for urban and rural Facebook user proportions, respectively. We further demonstrate that accounting for overdispersion and spatial correlations in the data is crucial to obtain accurate estimates and appropriate credible intervals. Crucially, as predictors change over time, the models can be used to regularly update the population predictions, providing a dynamic complement to census-based estimates. These results have direct implications for humanitarian response in disaster-prone regions and offer a general framework for using biased social media signals to generate reliable and timely population data.

💡 Research Summary

This paper presents a novel Bayesian framework to overcome the dual challenges of representativeness and privacy in using social media data for dynamic population mapping. Focusing on the Philippines as a case study, the research calibrates spatially and temporally resolved Facebook user counts with official 2020 census data to produce reliable small-area population estimates.

The work addresses two major sequential problems. First, it identifies a critical bias introduced by the differential privacy mechanisms employed by social media platforms. These mechanisms suppress or mask data in tiles where user counts fall below a certain threshold, which disproportionately affects low-population, often rural areas, exacerbating existing urban biases in digital datasets. To recover these missing values, the authors develop a Bayesian multiple imputation model. This model leverages the observed data and spatial correlation structures between geographical tiles to impute the censored counts, successfully restoring data coverage for 5.5% of rural areas and providing a principled way to quantify the associated uncertainty.

Second, using the imputed Facebook data aggregated to the municipality (Bayan) level, the study constructs a statistical model to estimate the proportion of Facebook users in each area—the key scaling factor needed to convert social media signals into population estimates. This proportion is modeled as a function of several openly available predictor variables that serve as proxies for socio-economic and technological context: Degree of Urbanisation Classification, working-age population proportion, nighttime light intensity, and network usage density (from Ookla Speedtest data). The core methodological advancement lies in the specification of a Bayesian hierarchical model that explicitly accounts for overdispersion (where data variance exceeds the mean) and spatial autocorrelation (the similarity between neighboring areas). Incorporating these features is shown to be crucial for obtaining accurate point estimates and appropriate credible intervals that reflect true uncertainty.

Out-of-sample validation demonstrates the model’s strong generalizability, with mean absolute percentage errors as low as approximately 18% for urban and 24% for rural Facebook user proportions. A significant practical advantage of the framework is its potential for dynamic updates. Since predictors like nighttime lights and network data can be updated frequently (e.g., monthly or quarterly), the fitted model can be used to regularly refresh population predictions between traditional census cycles. This enables near-real-time population monitoring, which is invaluable for disaster response and humanitarian planning in crisis-prone regions like the Philippines.

In conclusion, this research provides a robust, generalizable framework for transforming biased, privacy-protected social media data into reliable and timely population estimates. By systematically addressing data missingness and building a context-aware conversion model, it offers a powerful complementary tool to static census data, with direct applications for improving the efficiency and targeting of humanitarian interventions.