If you’re still at home, you’re not alone.
After a peak week of sheltering in place in early April, U.S. residents began to inch out of their homes, according to new cellphone data. But even as states begin to “open up,” more Americans appear to be staying put than sprinting out the door.
Share of time spent at home, for the seven-day period ending on:
Pinch to zoom. Tap to explore county details.Hover to explore county details
Note: Counties with fewer than 100 devices in the dataset are shown in gray. Source: Washington Post analysis of SafeGraph data. See full methodology at bottom of article.
According to a Washington Post analysis of data provided by SafeGraph, a company that aggregates cellphone location information, the peak period of our collective, coronavirus-induced lockdown was the seven-day period ending April 7. (There was also a one-day spike on Easter Sunday.) During that time, U.S. residents spent a whopping 93 percent of their time at home, up from the early March averages of roughly 70 percent.
Some governors began relaxing restrictions meant to curb the spread of covid-19 at the end of April, after they were prodded by President Trump and cautioned by public health officials. But as of April 30, people were still at home an average of 89 percent of the time.
The numbers vary greatly by geography. People in some counties appear to barely leave their homes, especially in large metro areas, including hard-hit New York, Detroit and Chicago. People in areas that are less populated tend to leave a bit more, and residents in at least one very rural county in Arizona are approaching the amount of free-roaming time they enjoyed in early March.
To determine when people are home, SafeGraph obtains GPS data through regular pings from smartphones that are running one or more apps from an undisclosed list. (This type of tracking, which requires users to allow location data when they use certain apps, has rattled privacy experts. App users are often not aware that their data is being used by third-party companies.)
The company defines “home” as a common location from which a phone pings between 6 p.m. and 7 a.m. When the phone pings from elsewhere, SafeGraph assumes the phone is away from home.
It’s not a perfect measure, but the data, aggregated from an average of 18 million phones per day, reveals broad insights.
Here’s where people are going
Just because people are leaving their homes more doesn’t mean they’re flouting stay-at-home orders or that they’ve abandoned social distancing. They simply may be enjoying warmer weather outdoors, especially in places such as the Upper Midwest, which had an unseasonable April cold snap.
Pings from phones within businesses and recreational areas show where people are — and aren’t — showing up. They’re going to parks and grocery stores, according to Google data, which tracks users via their Google Accounts and compares their locations to a baseline from before stay-at-home orders were in place.
Park visits have trended upward recently.
Workplace visits, however, have not. The percentage of phones (and presumably, their owners) in workplaces dropped steeply in mid-March and has barely risen. More than 30 million people have filed for unemployment benefits in the past six weeks.
SafeGraph’s more granular data shows people are again hitting general merchandise stores such as Walmart and Target at near-March levels. Some are satisfying fast-food cravings at restaurants with counter service.
The data lines up with responses in a Washington Post-University of Maryland poll released May 5 in which more than half of people said they were comfortable visiting grocery stores, but 78 percent said they would not be comfortable eating at a sit-down restaurant, and 67 percent said they wouldn’t be comfortable shopping in a clothing store. Only 18 percent of respondents were in favor of reopening movie theaters.
State orders don’t necessarily dictate behavior
While governors have issued guidelines for reopening their states, businesses and customers have been slow to return in many locales.
One of the earliest states to relax restrictions was Georgia, where a group of southwestern counties are suffering through some of the highest per capita covid-19 death rates in the country. Gov. Brian Kemp (R) allowed spas, hair salons, tattoo parlors, bowling alleys and gyms to reopen on April 24. As of April 30, it did not appear that most Georgians immediately flocked out to get haircuts and hit weight rooms. On average, they were still staying home 86 percent of the time, compared with 92 percent at their early-April peak.
Some states are starting to leave home more than others
Share of time spent at home, for the seven-day period ending on each date
Select a state:
In New Jersey, second only to New York in covid-19 deaths, people are spending 96 percent of their time at home, just 1.7 percent less than at the peak. That is both the highest stay-home percentage and the smallest change of any state. Gov. Phil Murphy (D) reopened state parks and golf courses on May 2, but his stay-at-home order is still in effect.
Montanans, by contrast, are going out 10 percent more compared with the peak, the largest change in any state. The state has reported 456 covid-19 cases and 16 deaths as of May 5. Gov. Steve Bullock (D) has allowed retail and churches to reopen and will allow restaurants and schools to open May 7.
Alaskans and South Dakotans are spending the least time at home, 80 percent on average. South Dakota Gov. Kristi L. Noem was one of seven governors to never issue a stay-at-home order. The state has one of the largest coronavirus clusters in the country, at a Smithfield pork-producing plant in Sioux Falls.
Variations on how people move around within states are common and sometimes large, and they often — but not always — follow the pattern of virus outbreaks.
In Florida, people in the populous southern counties of Miami-Dade, Broward and Palm Beach, where most of the state’s covid-19 cases have been concentrated and where beaches remain closed, are spending about 4 percent more time away from home than in early April. But people in the adjacent counties of Collier and Monroe, which includes the Florida Keys, were out of their houses 13 percent more as of April 30 than at the peak.
Areas of Georgia around Atlanta showed almost no decrease, led by Forsyth County, which barely budged from its peak of 97 percent at home. By contrast, people in rural Dooly County in the central part of the state were going out 12 percent more than they had at their stay-home peak, despite having a covid-19 death rate that is 19 times higher than Forsyth’s.
In Arizona’s La Paz County, residents are away from home 25 percent more than they were at the peak — the largest change of any county in the country — and they’re home just 57 percent of the time, roughly the same rate as in early March. In Maricopa County, which includes Phoenix, the difference is just 4 percent off the peak, and as of April 30, people were still home 90 percent of the time.
In general, people in rural areas are away more than people in urban and suburban areas, in large part because stores, services and essential workplaces are much farther away and delivery services are rare or nonexistent. But the trend lines are very similar across all types of places.
The data clearly shows that when Americans were told it was time to stay home, most people did. It also indicates that they are deciding for themselves when to go back out.
About this story
Data for this story was provided by SafeGraph, a company that aggregates location data from tens of millions of devices and compares it with building footprints. The Post analyzed its social distancing metrics data set to estimate the share of time spent at home for each county.
The raw numbers fluctuate from day-to-day and county-to-county, and some counties have far fewer devices than others. To account for these challenges, we adjusted the device counts based on population and performed a multilevel regression using a random effects model on the county, state and national levels. Because people tend to leave their homes far more on weekdays than weekends, we chose to calculate a seven-day rolling average, which results in each “day” of data including five week and two weekend days.
This analysis relies on defining a “home” location for every device. SafeGraph does this by finding the usual nighttime location of each device over the course of six weeks. It then buckets that location into a grid of roughly 150-meter-wide boxes across the country. When the device remains in that box, it is considered as being home. Of course this is not perfect, but in aggregate and at scale, SafeGraph’s at-home numbers track with other data sources, including location-data research published by other companies and public polling.
Urbanization figures are based on county designations from the National Center for Health Statistics’s urban-rural classification scheme.