LOCKED DOWN
A story of human movement in the US during 2020
In January 2020, COVID-19 began its march across the world, leaving a trail of devastation.
The United States was safe at first. But soon, the virus came knocking on its doors.
Eventually, most states in the US would issue lockdown orders. Some areas instituted stricter lockdowns than others, and some acted later or not at all. But...
What were the actual effects of these public policies on human behavior?
How closely did people adhere to lockdown guidelines, and if such guidelines were not present, did people change their behavior anyways?
If only there was some way to measure the movement of people over time...
Couture et al. compile 2 sets of smartphone movement data, showing both intra-state and inter-state mobility during the pandemic.
DEX
(Device Exposure Index)
A measure of intra-state mobility
On average, how many devices does one device “meet” over the course of a day?
LEX
(Location Exposure Index)
A measure of inter-state mobility
What proportion of devices in Location A recently arrived from Location B in the past 14 days?
Some caveats...
DEX is limited to “commercial venues”, i.e. no hospitals, schools, churches, outdoor parks, etc. are represented
Device mobility ≠ human mobility
(people can move around without their phones!)
For more information, see full dataset here
Nevertheless, DEX and LEX are a good approximation of intra− and inter-state mobility.
Let’s take a look at the data!
We’ll start with DEX.
Drag the slider to see how intra-state mobility changed throughout 2020.
Evidently, people did change their behavior at the start of the pandemic.
Within every state, movement dropped dramatically in mid-March.
But what about interstate travel?
Did that also change throughout the year?
Let’s now take a look at mobility between states, or LEX.
Drag the slider to see how inter-state mobility changed throughout 2020.
Click on a state to see its top 10 connections (either arrivals or departures).
It is clear that interstate movement also dropped during the start of the pandemic.
But were these movement patterns due to lockdowns?
Or did other factors influence people’s behavior?
Let’s look at DEX in the context of lockdowns.
Pay attention to the temporal order of lockdown versus mobility drop. Which comes first?
Did lockdowns precede mobility drops, or vice versa?
Do you see any differences between states which implemented lockdowns and those which did not?
In states like California, DEX hits bottom just at the start of lockdown.
But in other states such as Alabama, DEX hits bottom before the lockdown even starts. In fact, you’ll notice this pattern for most of the states in the South.
Moreover, for states like Utah, DEX drops with no lockdown at all.
Go back to the previous figure and check out the magnitude distribution of the March mobility drop.
The distributions between Lockdown and No Lockdown states (if you ignore outliers such as Nevada) are pretty similar!
Let’s take a closer look at Nevada, the state with the largest mobility drop.
Here, we’ve added a timeline with more information about the state’s response to COVID.
It turns out that, prior to the official lockdown, there were many events which hinted at the severity of the pandemic.
In fact, this holds true for all states.
Lockdowns were often preceded by emergency declarations, school and business closures, and public health announcements.
So did lockdowns make a difference?
Lockdowns were just one of many factors which influenced people’s behavior during the pandemic.
Were they the most important factor? Probably not.
You may have noticed that mobility started increasing for most states right after the initial drop.
By December, mobility in many states (regardless of lockdown status) were back at their February levels.
It seems fair to conclude that depending on lockdowns alone is not enough.
Effectively controlling the pandemic requires a combination of factors, such as public health messaging, targeted legislative intervention, rapid testing, and vaccination.
Our Team
Angel Hsu
Harvard SEAS
Data Science
Jimmy Young
Harvard College
Biomedical Engineering
Andrew Zhang
Harvard College
Biology & CS
Team Lead
