Photo: Colourbox

Corona Crisis: Behavioural data reveals movement patterns

Monday 22 Jun 20


Sune Lehmann Jørgensen
DTU Compute
+45 45 25 39 04

Explore the researchers’ interactive visualizations yourself

The research project ‘HOPE – How Democracies Cope with COVID19 – A Data-Driven Approach’ is also a project aimed at disseminating the research to a broad target group. It is therefore possible to explore the interactive visualizations which are continually updated. Here you can also read the researchers’ reflections on the latest data
Researchers are using behavioural data from Facebook to understand how the Danish population is reacting to the government’s corona interventions and the gradual opening up of the country.

In March, DTU, the University of Copenhagen, and Aarhus University received a DKK 25 million donation from the Carlsberg Foundation to implement the Semper Ardens research project ‘HOPE’ – How Democracies Cope with COVID19 – A Data-Driven Approach’.

The project, which combines big data and machine learning with insights from social sciences, aims to help authorities understand the link between the government’s handling of the crisis, media coverage, the population’s behaviour, and the spread of the virus.

On a new website, researchers have now compiled several types of anonymized and aggregated behavioural data. The data comes from Facebook, which for the past two years has made datasets available to universities and NGOs via the product ‘Facebook Data for Good’ that aims to improve well-being and save lives.

“The data are collected through users’ mobile apps and are unique in that they say a lot about our movement patterns and where we are staying—not about the things we do on Facebook,” says Sune Lehmann, a professor at DTU Compute who is responsible for the handling and technical analysis of the large volumes of data.

This is not the first time that researchers have used these data sets for data visualization, but DTU Compute has now developed a number of new analyses of the data sets which have made it possible, for example, to estimate where Danes are spending their time during the corona crisis, how they move around the country during lockdown, and how their behaviour has changed over time. The researchers are also able to compare different municipalities and cities with one another.

The observations show, for example, how the opening of childcare affected the behaviour of Danes and how the number of residents has changed over time in the Danish municipalities. While many Danes left the major cities when Denmark went into lockdown, Fanø, Odsherred, Dragør, Gribskov, Læsø, and Halsnæs recorded population growth. 

 Photo: DTU

Danes’ movement data on Monday, 6 April 2020. Green indicates an increase in people and red indicates a decrease relative to the beginning of the corona crisis. The figure is interactive and can be explored here.

“It’s been exciting to see how we can use movement data to understand the public’s response to government intervention—and subsequently the easing of lockdown conditions,” says Sune Lehmann.

“We also want to understand how we can use mobility data to better model the disease’s pattern of spread. However, it’s not entirely straightforward, as people’s collective patterns of movement don’t say anything about who is meeting who—which is what determines the spread of infection. At the same time, the two data sources aren’t independent, so how we use the dataset is an exciting research challenge.”

The next step in the project is to help other countries gather data on population behaviour. It is also hoped that DTU students can take part in similar projects in developing countries where it is harder to obtain reliable public information about Covid-19.

Currently, Brazil, Iceland, Italy, Norway, New Zealand, Spain, Sweden, and Germany can also be explored on the website. Comparisons between countries can help the researchers understand how countries’ multiplicity of interventions have affected movement patterns differently.

News and filters

Get updated on news that match your filter.