Data for Social Good

Insights from Mobility Data for COVID-19

A curated list of impactful Global Use Cases

Parvathy Krishnan
6 min readApr 28, 2022

This work has been done entirely using publicly available data and was co-authored with Mahdi Fayazbakhsh and Kai Kaiser. All errors and omissions are those of the author(s).

Photo by Martin Sanchez on Unsplash

Quantifying human mobility patterns gives insights into many aspects of life from the regional and global spread of infectious diseases to economic fluctuations. With over two-thirds of the world population having access to mobile phones in 2020, there has been an influx of applications trying to model users, places, and trajectories of movements through innovative ways of collecting mobility data. The mobility data can be used in many applications ranging from epidemic modeling for assessing the recovery of communities after natural disasters.

In the wake of COVID-19, there was a great demand for such mobility data to be made public, with epidemiologists, researchers, and humanitarian agencies corroborating that aggregated mobility data could help fight COVID-19. Such aggregated datasets had the potential to examine the impact of social distancing messaging or policies, and also help in optimizing policy relaxations without causing a major resurgence, given that the personal privacy protection of people is given the utmost importance while collecting and aggregating this information.

This resulted in the availability of multiple mobility datasets being made available in the public domain with different methodologies, granularities, and spatial coverage.

Figure 1. Publicly available mobility indicators and mobility change measures (Source — Change of human mobility during COVID-19: A United States case study)

This blog outlines two such initiatives to make mobility trends (anonymised and adhering to the data protection and privacy rules) publicly available during the COVID-19 pandemic and have resulted in creating a huge impact in driving data-informed and data-driven decision making across sectors globally. Globally these datasets coupled with other structured and unstructured data sources have been instrumental in driving decision- and policy- making in the wake of COVID-19 curves. We also present a curated list of these use cases to demonstrate the values of these datasets.

First is the COVID-19 Community Mobility Report by Google which provides location-specific mobility trends by anonymizing location data obtained by Google users (who have turned on the location history share feature). The second initiative that we will look at is the Facebook Data for Good platform which has made available two metrics — Change in Movement and Stay Put — as an indicator of human mobility change.

Google Community Mobility Reports

Google collects anonymized data from its users regarding real-time locations to help optimize its applications such as routing in traffic or recommending the nearest restaurant. This data when analyzed at scale can be used to look at movement trends across different countries and regions. Google realized the importance of making this data available without including personally identifiable information such as individual movements, locations, or contact. This resulted in the Community Mobility Reports which provide the percentage change in visits to given categories of location (retail and recreation, parks, grocery and pharmacy, transit stations, residential, and workplaces) compared to a baseline. The baseline day is the median value from the 5-week period between Jan 3, and Feb 6, 2020, for that day of the week. This also accounts for intra-week changes because the baseline for weekends may be significantly different from the baseline for the week.

This aggregated data for six categories of location can provide insights into how far movement has dropped, for example, the proportion of people staying at home after the lockdown was announced. Initially launched for a coverage of 131 countries, currently, it has data for 135 countries.

The mobility data is made available in forms ready for end-users (e.g., WebApps or reports) or data access resources (APIs). The latter may offer more granular data access, and also allow data presentations to be customized in particular ways (e.g., by intersecting with other temporal or spatial data series).

The following code shows how to access this data in Python and explore the availability for different countries.

Figure 2. Data columns fetched from the Google Mobility Reports (Image by Authors)

The data is also available at Level 1 and Level 2 administrative units for many countries. Level 1 units are provinces and municipalities whereas Level 2 includes districts, provincial cities, and district-level towns. Out of the 135 countries, 95 of them have data available at Level 1 and 42 countries at Level 2 administrative levels.

Figure 3. List of countries where Google Mobility Data is available at Level 1 and Level 2 administrative levels (Image by Authors)

Example Use Cases

  1. Mining Google and Apple mobility data: temporal anatomy for COVID-19 social distancing | Scientific Reports
  2. Nowcasting Economic Activity in Times of COVID-19: An Approximation from the Google Community Mobility Report
  3. Linking excess mortality to Google mobility data during the COVID-19 pandemic in England and Wales
  4. Effects of COVID-19 on geographic mobility and working habits | Implications of Remote Working Adoption on Place-Based Policies: A Focus on G7 Countries
  5. Exploring the Utility of Google Mobility Data During the COVID-19 Pandemic in India: Digital Epidemiological Analysis

Facebook Movement Range Trends

Facebook, through the Data for Good published two metrics for quantifying human mobility changes — Change in Movement and Stay Put. Change in Movement looks at how much people are moving around and compares it with a baseline period that predates most social distancing measures, while Stay Put looks at the fraction of the population that appears to stay within a small area during an entire day.

People who use Facebook on a mobile device have the option of providing their precise location in order to enable products like Nearby Friends and Find Wi-Fi and to get local content and ads. Movement Range Trends are produced by aggregating and de-identifying this data. Only people who opt into Location History and background location collection are included. People with very few location pings in a day are not informative for these trends, and, therefore, only those people whose location is observed for a meaningful period of the day are included in the trend metric.

Each metric in this data set is produced for a given administrative region once per day. Beyond U.S. counties, the disaggregation includes level 3 statistical regions from the Nomenclature of Territorial Units for Statistics (NUTS) for European countries and level 2 divisions from the Database of Global Administrative Areas (GADM) for other countries around the world. Each data point corresponds to a full day and night, from 8:00 p.m. one day to 7:59 p.m. the next day in local time.

To generate a data point for a given region, the locations of users who spend evenings in that particular region are aggregated. To protect the privacy of users and to make sure there is a significant number of data points to estimate trends, any region with fewer than 300 qualifying people is omitted from the data sets.

Our team has written a very detailed blog and published a notebook on how to access, analyze and derive insights from the Facebook Movement Range Maps which is available on the following medium blog.

Example Use Cases

  1. Mobility data to aid assessment of human responses to extreme environmental conditions
  2. Socioeconomic status determines COVID-19 incidence and related mortality in Santiago, Chile
  3. Genomics, social media, and mobile phone data enable mapping of SARS-CoV-2 lineages to inform health policy in Bangladesh
  4. Spatial Heterogeneity of COVID-19 Impacts on Urban Household Incomes
  5. Temporary reduction in daily global CO2 emissions during the COVID-19 forced confinement

As demonstrated by the use cases, the changes in mobility — be it the data from Google or Facebook — need to be contextualized with additional data sources like the demographics, socio-economic indicators, infrastructure, etc. to better model and understand the underlying reasons for the variations observed in the datasets. For example, regions with better infrastructure might show lower mobility for categories such as grocery and pharmacy because of the shorter trips and easier access to such services. Also, it is important to note that the data is collected only from people who have enabled location and history sharing settings and hence might be overrepresenting certain populations while underrepresenting others.



Parvathy Krishnan

Lead Data Scientist | CTO at Analytics for a Better World | Public Sector Consultant