COVID-19 Mobility Data Analysis: Observations on using de-identified and aggregated data for social impact analysis in India

PCQ Bureau
New Update
COVID-19 Mobility Data Analysis

The COVID-19 pandemic has introduced a range of public health, safety, and economic challenges in India Researchers from the Indian Institute of Technology – Tirupati (IIT-T) partnered with Facebook to create data-driven tools to help policymakers and public health experts to make informed decisions and interventions.


How it Works: Facebook provides de-identified and aggregated datasets on population density, movement, and social connectedness to academic researchers and nonprofits, including those at IIT-T. These institutions and organizations use these insights to generate analysis for governments, healthcare workers, first responders, and civil society organizations.

Globally government COVID-19 response units and public health officials use these visual insights to inform interventions. This includes healthcare delivery and food aid distribution as well as policymaking decisions such as the lifting of and compliance with lockdown orders. Meanwhile, researchers use these datasets to conduct trend analysis to better understand the economic and social impact of the pandemic on communities. Critically, partnering with academic and civil society stakeholders like IIT-T helps ensure independent subject matter expertise, stringent privacy protections and promotes global scalability.

Project Conceptualization and Approach: Researchers at IIT-T use Facebook movement maps and population density maps to analyse population movement and disease containment in the states of Andhra Pradesh and Odisha. They did this by calculating the estimated number of people moving through COVID-19 hotspots at the ward- or block-level at different times of the day. IIT-T then shared daily, visual reports with the state governments of Andhra Pradesh and Odisha on high-mobility regions, geographies with high outflows of people, and areas with high population density to help governments make decisions about interventions in red zones to help slow the spread of disease.


Implementation: IIT-T developed a real-time dashboard to visualise and understand the effectiveness of the lockdown imposed due to COVID-19. The analysis was created by combining various public datasets along with de-identified datasets provided by Facebook. In addition to Facebook, IIT-T also directly worked with the Andhra Pradesh and Odisha state governments to construct the COVID-19 response dashboard.

Datasets Used:

Privacy by Design: Facebook datasets are available to academics and nonprofits who sign a datasharing agreement that limits the purpose and sharing of the datasets. The aggregated, deidentified datasets can be downloaded in an open format (csv) from a portal, and are regularly updated to allow for real-time response reports. In order to protect user privacy, Facebook follows a three-tiered framework.


User consent: Only Facebook app users who opt-in to location history and background location collection are included in the datasets

Data aggregation: Each data point is generated by aggregating data of a number of users and any region with fewer than 300 qualifying people is omitted from the data sets.

Differential privacy (DP): Facebook applies a differential privacy (DP) framework. Differential privacy minimizes the risk of re-identification of individual data with the help of possible additional information — even information that cannot be anticipated now. Applying a DP framework takes into account the sensitivity of the datasets and adds noise proportionally to ensure with high probability that no one can reidentify users.

Figure 1: Odisha COVID-19 Daily Dashboard

Figure 1: Odisha COVID-19 Daily Dashboard

Figure 1: Odisha COVID-19 Daily Dashboard

Figure 2: Andhra Pradesh Movement Map

Generating Insights: IIT Tirupati

Different states required different types of analysis, initially, most states requested analysis of average population movement over 24 hours at inter-state, inter-district and intra-district levels. As the COVID-19 pandemic continued, state governments started seeking more customised insights, such as mobility analysis for specific periods during the day, mobility in quarantine zones, and population movement in major cities, such as Puri in Odisha, to better pinpoint outbreaks.


These highly-localised insights helped authorities identify potential hotspots and make informed decisions about where to ease lockdown orders and where to retain or increase restrictive measures in order to protect public safety.


“We have partnered with Facebook and through them with IIT Tirupati to get research-based analysis for insights to understand the degree to which communities in red zones are adopting physical distancing and whether more stringent enforcement needs to be deployed. We are reading these reports daily to have a real-time view of important correlates of disease transmission and thank Facebook and IIT Tirupati for supporting us with such important research work,” said Manoj Mishra, Secretary of Electronics and IT in the Government of Odisha, Odisha.” Odisha used Tirupati IIT mobility analysis to track infection, Hemant Kumar Rout, The New Indian Express, 26th May 2020 “Technology is playing a vital role in helping us track and limit the spread of COVID-19. The mobility analysis reports are very helpful as these are providing insights to understand the degree to which people in the hot spots and containment zones are adopting physical distancing and whether more stringent enforcement action is needed. We are reading these reports daily to have a real-time view of important correlates of disease transmission and mobility. IIT Tirupati is supporting us with the analysis from Facebook's aggregated and deidentified datasets and we appreciate them for engaging with us on such critical research work,” said Pratap Bhimireddy, Special Representative, Investment Promotion & Infrastructure Development, Govt of Andhra Pradesh.” Andhra Pradesh ropes in Facebook, IIT-Tirupati for better management of COVID-19, Appaji Reddem, The Hindu, 11th July 2020.


Key Lessons Learnt:

Encourage Organizations to Voluntarily Develop Data Tools for Social Impact

Value of Purpose-Specific Datasets: IIT-T has been using Facebook Data for Good<3> datasets since 2018, initially to support research on crisis response by analysing population displacement after natural disasters. In late 2019, IIT-T participated in Facebook’s external stakeholder consultation to recommend how the datasets could be used to provide support for health emergencies. This consultation led to the development of Facebook’s Disease Prevention maps launched in 2019<4>. Though Disease Prevention maps were created specifically for health emergencies, they did not fully address the needs of health officials battling the COVID-19 crisis, and had to be adapted significantly during the early months of COVID19 response. This chronology illustrates the amount of time and successive consultations required to create relevant datasets. In the absence of a clearly defined end use case aggregated datasets’ relevance maybe limited.

Limitation of Reliability and Representativeness: A number of technology firms including Facebook, Google, Camber systems among others have shared mobility datasets to support COVID-19 response efforts. Third-party researchers play a vital role in this process – particularly in understanding the reliability of various datasets and determining the extent to which they are representative of a population as a whole. This is important because private datasets provided by technology companies have limitations and may not represent the situation on the ground with 100 percent accuracy. Researchers can help verify that accuracy over time and can also identify where datasets may fall short of being fully representative of demographics in a particular geography. For example, vulnerable segments of society such as women and lower income groups may not have access to broadband or devices that would allow them to be captured in a particular dataset.

Promote the Free Flow of Data and Expertise

Collaborate to Shorten Response Time: Ensuring the free flow of data and ideas across borders helped reduce response time. Professor Satchit Balsari and Professor Caroline Buckee, and their teams, from the COVID-19 Mobility Data Network<8> had been using the Facebook datasets to generate mobility reports for counties in the US. They conducted several online sessions to share their findings regarding the utility of the maps in the US and possible applications in India with the IIT-T team. The COVID-19 Mobility Data Network produced the initial mobility reports for India and shared insights on the overall structure of the reports and dashboards. By leveraging their expertise and experience, the IIT-T team was able to provide useful, actionable data maps and dashboards to state governments more quickly.

Enable Access to Global Expertise and Experience: In addition to enabling more rapid analysis, crossborder research linkages and data sharing can help produce more sophisticated insights. The COVID-19 Mobility Data Network is a global, multi-disciplinary team with expertise across health and data science.

Such diverse expertise is difficult to aggregate at a country level – especially on such a short notice. The COVID-19 Mobility Data Network shared their health expertise across a range of subjects including the ethical guidelines around health research. This complemented IIT-T capabilities in data science, allowing for more sophisticated analysis.

Strengthen Academic Networks to Contribute at Scale:

Grow the Independent Expert Ecosystem in India: It is critical to expand academic partnerships and networks globally and in India in order to promote scalability. Data scientists at research institutions bring not just expertise, but the freedom to iterate, explore, collaborate, and share insights with a wide community. This can often be more effective than partnering with another private sector company or a government agency. IIT-T’s role as an expert partner illustrates the importance and efficacy of this type of private sector-academic linkage. IIT-T contributed in two significant ways, providing:

Subject matter experts: As data scientists, IIT-T developed statistical models to create a visual map or graphic from Facebook datasets, sometimes overlaying it with other public datasets in response to specific policy development requirements.

Local context: A familiarity with the Indian administrative functioning helped IIT-T develop relevant statistical models, such as aggregating insights from a ward-level to match administrative units.

Create a Knowledge Hub: In order to expand the contribution of data science for social impact, a more structured approach to academia and private sector collaboration would be beneficial. Since the launch of Data for Good program in 2017, Facebook has expanded its network of academics and researchers in India. In 2019, Facebook convened leading academics to discuss and debate the various applications for data science for social impact and introduce the Facebook Data for Good program to the larger audience. Looking ahead, it will be crucial to build on this progress by creating a consortium or knowledge hub of local data science partners that can not only help the private sector conduct analysis but develop data-driven solutions to social and development challenges.

Route to Scale: Robust, in-country networks help rapidly scale up partnerships. IIT-T leveraged its strong network of PhD students and faculty to scale operations. This eventually included supporting two state governments and developing a real-time dashboard covering mobility analysis in multiple states.

Engaging IIT-T’s network allowed Facebook to scale up the Data for Good program, as well as provided students with hands-on data science experience in real-life scenarios. The success of this project illustrates the critical importance of academic institutions in the innovation ecosystem. In partnerships with the private sector, academic institutions are well-positioned to ensure the long-term sustainability and success of projects of this type.


Dr. Kalidas Yeturu Indian Institute of Technology Tirupati,

Charu Chadha, Facebook

Krishna Prapoorna, Ph.D. Indian Institute of Technology Tirupati,

Rohit Singh, Facebook