COVID-19 has introduced a new era for the world as a whole, with infection numbers continuously rising, and movement and interaction restrictions implemented far and wide. As governments issued shelter in place orders, schools closed, and non-essential businesses shut down, a recurring question is whether or not people are actually following these restrictions. At Skyhook, we work with mobile location data in various forms. This data is incorporated into our Mobility Index, which can help answer some of those questions about current population behaviors and community movement. Below is a synopsis of some of the things we have seen in March and most of April.
These days most people carry a cell phone or another GPS or WiFi-enabled device that uses location services to provide latitude, longitude, and timestamp records corresponding to real world positions of the device over time. The Skyhook Mobility Index is a powerful new tool that looks at mobile device movements in the aggregate as a way of modeling population behaviors. The Mobility Index can measure the amount of time a device spends in a location, the distance a device traveled, and the number of locations visited in each day. This can be measured and aggregated at global, national, state, and even county levels. By understanding device movement patterns and how they change over time, we can present a powerful (and often complementary) way to better understand changes in public health policies and shelter-in-place orders.
Measuring Social Distancing
We wanted to look at the changes in mobility since COVID-19 started to spread, specifically looking at the restrictions on movement in U.S. cities and how those were followed by residents. To analyze these, we looked at the degrees of mobility per mobile device from their home location.
To analyze mobility data, Skyhook first divides the world into millions of “tiles,” which are individual areas of 100 meters wide by 100 meters long. From there, the Mobility Index measures the number of tiles a device enters during a given day. As devices visit more tiles, the degree of mobility increases. Skyhook takes individual device movement statistics and aggregate them across various regional levels like counties, states, and countries, allowing us to make general observations regarding movement patterns within those different levels.
As tiles are bound to specific counties, states, and countries, and all devices in a tile are aggregated, it’s possible to observe and analyze movement patterns at different levels. The diagram to the right illustrates how a device travels through five individual tiles, relative to its home location, which is denoted with an enlarged circle.
The Mobility Index pairs an anonymized device ID with timestamps across different locations. As the device enters each additional tile, the tile count associated with that device is increased by a value of one.
City and State Analysis
Skyhook leveraged Mobility Index data to show where there were changes in mobility in U.S. cities related to the virus restrictions. We looked at some cities and states throughout the United States to get a feeling for how school and non-essential business closures affected population movement. (Date of interventions are from http://covid19.healthdata.org/.) By comparing these timelines with movement patterns after these dates, we can see the changes in lifestyle behaviors and motion based on restrictions.
All of the cities we discuss below are observed between March 1 - April 24, 2020, providing almost two months of data. While we provide some speculation in this post, we encourage readers to look at the data themselves and come to their own conclusions about influences on population behaviors and movement.
New York City:
In New York City we see the mobility index change over time, with drastic drops after schools closed in-person operations. The heat map below shows changes in activity over time. Blue indicates a decrease in relative activity, while red indicates an increase. Something interesting to note is that the Verrazano bridge and Staten Island ferry path are very blue, which qualitatively shows people in Staten island are leaving the island less. We can also see that lower parts of Manhattan have decreased activity which makes sense as they are primarily commercial buildings.
Trends in Population Movement in the US
Looking at these graphs - we can see that closing schools had a large impact on population movement and behaviors. There are dips on weekends which we assume to be essential workers who aren’t working on the weekends. Across all of these cities, there was a sharp drop in mobility on March 8. On this date the CDC released updated guidance for evaluating and testing for COVID-19, and released an avoid non-essential travel notice. Other cities show a dip on March 21, when Italy reported record-high single-day death toll from coronavirus, one day after previous record-high, and New York state surpassed 10,000 confirmed cases of coronavirus, with more than 6,000 in New York City. On March 29 President Donald Trump extended social distancing orders in a press briefing, which may contribute to the additional observed drop.
Stockholm County, Sweden
We’ve included Stockholm County in Sweden in this round-up due to Sweden’s unique response to the COVID-19 pandemic. Sweden, unlike the majority of the world, has mostly forgone forced social distancing and stay at home orders, preferring to depend on individual responsibility, requesting citizens stay at home if sick and monitor their own health. This approach is different, and you can see that average movement is higher than the United States.
Over time as travel restrictions ease, Skyhook will continue to monitor how people approach reintegration into society as we observe devices visiting additional tiles in the Mobility Index. At the current time, one of the more interesting questions is around whether mobility is now increasing even in areas that haven’t lifted bans, given “fatigue” with staying at home. Subscribe to the Skyhook blog to see a follow up post as we continue to follow movement trends. This is a sampling of the data we can pull, and a basic example of what you can learn about social distancing behaviors from foot traffic analysis - reach out to us if there are specific cuts of data you are interested in observing.