No abstract

BACKGROUND In recognition of the empowering potential of digital technologies, gender equality in internet access and digital skills is an important target in the United Nations (UN) Sustainable Development Goals (SDGs). Gender-disaggregated data on internet use are limited, particularly in less developed countries. OBJECTIVE We leverage anonymous, aggregate data on the online populations of Google and Facebook users available from their advertising platforms to fill existing data gaps and measure global digital gender inequality. METHODS We generate indicators of country-level gender gaps on Google and Facebook. Using these online indicators independently and in combination with offline development indicators, we build regression models to predict gender gaps in internet use and digital skills computed using available survey data from the International Telecommunications Union (ITU). RESULTS We find that women are significantly underrepresented in the online populations of Google and Facebook in South Asia and sub-Saharan Africa. These platform-specific gender gaps are a strong predictor that women lack internet access and basic digital skills in these populations. Comparing platforms, we find Facebook gender gap indicators perform better than Google indicators at predicting ITU internet use and low

Abstract. Mapping the distribution of poverty in developing countries is essential for humanitarian organizations and policymakers to formulate targeted programs and aid. However, traditional methods for obtaining socioeconomic data can be time-consuming, expensive, and labor-intensive. Recent studies have demonstrated the effectiveness of combining machine learning and satellite images to estimate wealth in sub-Saharan African countries (Xie et al., 2016, Jean et al., 2016). In this study, we investigate the extent to which this method can be applied in the context of the Philippine archipelago to predict four different socioeconomic indicators: wealth level, years of education, access to electricity, and access to water. We also propose an alternative, cost-effective approach that leverages a combination of volunteered geographic information from OpenStreetMap and nighttime lights satellite imagery for estimating socioeconomic indicators. The best models, which incorporate regional indicators as predictors, explain approximately 63% of the variation in asset-based wealth. Our findings also indicate that models trained on publicly available, volunteer-curated geographic data achieve the same predictive performance as that of models trained using proprietary satellite images.

The United Nations Sustainable Development Goals (SDGs) are a global consensus on the world's most pressing challenges. They come with a set of 232 indicators against which countries should regularly monitor their progress, ensuring that everyone is represented in up-to-date data that can be used to make decisions to improve people's lives. However, existing data sources to measure progress on the SDGs are often outdated or lacking appropriate disaggregation. We evaluate the value that anonymous, publicly accessible advertising data from Facebook can provide in mapping socioeconomic development in two low and middle income countries, the Philippines and India. Concretely, we show that audience estimates of how many Facebook users in a given location use particular device types, such as Android vs. iOS devices, or particular connection types, such as 2G vs. 4G, provide strong signals for modeling regional variation in the Wealth Index (WI), derived from the Demographic and Health Survey (DHS). We further show that, surprisingly, the predictive power of these digital connectivity features is roughly equal at both the high and low ends of the WI spectrum. Finally we show how such data can be used to create gender-disaggregated predictions, but that these predictions only appear plausible in contexts with gender equal Facebook usage, such as the Philippines, but not in contexts with large gender Facebook gaps, such as India.

In times of crisis, real-time data mapping population displacements are invaluable for targeted humanitarian response. The Russian invasion of Ukraine on February 24, 2022 forcibly displaced millions of people from their homes including nearly 6m refugees flowing across the Nowcasting population displacement in Ukraine using social media border in just a few weeks, but information was scarce regarding displaced and vulnerable populations who remained inside Ukraine. We leveraged near real-time social media marketing data to estimate subnational population sizes every day disaggregated by age and sex. Our metric of internal displacement estimated that 5.3m people had been internally displaced away from their baseline administrative region by March 14. Results revealed four distinct displacement patterns: large scale evacuations, refugee staging areas, internal areas of refuge, and irregular dynamics. While this innovative approach provided one of the only quantitative estimates of internal displacement in virtual real-time, we conclude by acknowledging risks and challenges for the future.

Much research has examined how crime rates vary across urban neighborhoods, focusing particularly on community-level demographic and social characteristics. A parallel line of work has treated crime at the individual level as an expression of certain behavioral patterns (e.g., impulsivity). Little work has considered, however, whether the prevalence of such behavioral patterns in a neighborhood might be predictive of local crime, in large part because such measures are hard to come by and often subjective. The Facebook Advertising API offers a special opportunity to examine this question as it provides an extensive list of “interests” that can be tabulated at various geographic scales. Here we conduct an analysis of the association between the prevalence of interests among the Facebook population of a ZIP code and the local rate of assaults, burglaries, and robberies across 9 highly populated cities in the US. We fit various regression models to predict crime rates as a function of the Facebook and census demographic variables. In general, models using the variables for the interests of the whole adult population on Facebook perform better than those using data on specific demographic groups (such as Males 18-34). In terms of predictive performance, models combining Facebook data with demographic data generally have lower error rates than models using only demographic data. We find that interests associated with media consumption and mating competition are predictive of crime rates above and beyond demographic factors. We discuss how this might integrate with existing criminological theory.

In times of volatility and crisis, it is essential to have real-time data mapping population movements to facilitate a rapid and effective humanitarian response. Considerable attention has been placed on the 5.8 million Ukrainian refugees crossing the border as of early May 2022, but information is scarce to quantify and locate over 38 million people who remain in the country, many of whom have been displaced from their homes or remain vulnerable to conflict. This report and its supplementary data leverage near real-time digital trace data from social media users, along with demographic and geo-spatial methods to produce daily estimates of current population sizes and changes sub-nationally disaggregated by age and sex. Using our estimation methods, we quantify large reductions in populations from conflict areas (e.g. Kyiv city), particularly women and children, and population increases in western Ukraine (e.g. Lviv Oblast). Examining additional Oblasts (administrative regions) such as Cherkasy, we show evolving and mixed demographic changes as populations transit through different areas. Mapping population dynamics through time, we illustrate the net changes in population sizes in Oblasts from the start of the war until present, providing a daily metric of total negative population drops. As of May 06, this metric suggested that 6,529,949 people were displaced away from their baseline Oblast. While this data and approach is innovative and is one of the few estimates available to quantify and map internal displacement in virtual real-time during a crisis, we conclude by acknowledging deficiencies and future extensions.

Having reliable and up-to-date poverty data is a prerequisite for monitoring the United Nations Sustainable Development Goals (SDGs) and for planning effective poverty reduction interventions. Unfortunately, traditional data sources are often outdated or lacking appropriate disaggregation. As a remedy, satellite imagery has recently become prominent in obtaining geographically-fine-grained and up-to-date poverty estimates. Satellite data can pick up signals of economic activity by detecting light at night, it can pick up development status by detecting infrastructure such as roads, and it can pick up signals for individual household wealth by detecting different building footprints and roof types. It can, however, not look inside the households and pick up signals from individuals. On the other hand, alternative data sources such as audience estimates from Facebook's advertising platform provide insights into the devices and internet connection types used by individuals in different locations. Previous work has shown the value of such anonymous, publicly-accessible advertising data from Facebook for studying migration, gender gaps, crime rates, and health, among others. In this work, we evaluate the added value of using Facebook data over satellite data for mapping socioeconomic development in two low and middle income countries – the Philippines and India. We show that Facebook features perform roughly similar to satellite data in the Philippines with value added for urban locations. In India, however, where Facebook penetration is lower, satellite data perform better.