Before You Use Our GPS Travel Data To Formulate Coronavirus Policy, Make Sure You Understand the Data
Confusing travel distance with actual human mingling is no way to create smart policy.
Companies are using aggregated GPS location data to track trends on how well citizens are changing their habits in response to COVID-19. While this information might be helpful, it has also prompted some unsettling and maybe misguided power plays.
It started in mid-March, when Unacast, a company that tracks and analyzes location data from people's phones, put together what a "Social Distancing Scoreboard" to attempt to calculate how successfully the citizens of each state are changing their travel habits.
Calling it a "Social Distancing Scoreboard" is itself a mistake. Social distancing is supposed to refer to the amount of physical space between individuals. But this scoreboard initially graded states on the basis of the distance people were traveling. Ostensibly this was an attempt to see if people were making fewer non-essential trips. That may be valuable data, but it's not what "social distancing" means.
It got worse when they started grading states. The only state to get an F was (and still is) Wyoming, which currently has 153 confirmed cases of COVID-19 and no deaths. New York state, now the world's epicenter in coronavirus infections and deaths, gets an A-. (The death toll in New York City alone has topped 1,500.) The reason for the difference should be obvious. Wyoming is a profoundly rural state. Of course its residents won't reduce their travel distances as much as someone in a big city. Sweetwater County, the physically largest county in Wyoming, has a population density of 4.2 people per square mile. New York City has a population density of 27,000 people per square mile.
And why should it matter that someone in Wyoming travels further to buy necessities? The goal is supposed to be reducing exposure within the population. People who live in high-density areas are going to have to take harsher measures than those who do not. I live in Los Angeles, and I have not traveled more than a mile from my apartment since the second week of March. But within that distance is a grocery store, and dozens of nearby restaurants are begging me to get their food delivered. I suspect the dynamics are different in Rock Springs, Wyoming, population 23,000.
Unacast subsequently upgraded its methodology to factor in whether a trip is "essential" or "non-essential." Its definition of "non-essential" venues includes consumer electronics stores, office supply stores, toy stores, and movie theaters. But it also includes restaurants and hotels, which in some contexts might be very essential.
Unacast seems aware that rural environments are going to be different, and the company's explanation of its methodology makes it clear that these data map people's behavior, not the path of COVID-19's spread. They hope, they explain, to "provide direct aggregated feedback to policy makers and community leaders on how well their social distancing measures are being adopted by the general public, and if more severe restrictions do lead to a reduction in the number of reported cases of COVID-19."
Because we are in the grip of "We have to do something, anything, to stop the spread of coronavirus," leaders are turning to this aggregated data to justify "more severe restrictions." Tennessee Gov. Bill Lee just used aggregated travel data to justify stricter "stay at home" orders, even though Unacast's data showed about a 45 percent decline in travel distance and "non-essential visits" even before his order.
In Vermont, the state has ordered big-box stores to stop selling items they deem "non-essential" within their stores and to cordon them off from shoppers. (Delivery and curbside purchases are still permitted.) Of course, just because the government says these goods and products are not essential, doesn't make it true. Garden goods have been deemed non-essential, which doesn't seem the right message for a time when people are supposed to be holed up at home with plenty of food.
Now Google is offering aggregated reports showing travel data from phone users who have turned on their location history settings. The information is being aggregated anonymously so that nobody's privacy will be violated. But nevertheless, these data can be abused. Wyoming is showing a decline in travel to most places but an increase in visits to parks. That tells you nothing about whether people are social distancing in those parks. But as we've already seen in the United Kingdom, police are quite capable of confusing "traveling to parks to get exercise" with "not engaging in social distancing." On Thursday, Los Angeles County sheriff's deputies arrested a paddleboarder off the coast of Malibu for violating a stay-at-home order. Bringing the man to the sheriff's station in Calabasas for processing before releasing him exposed him to a much greater risk of COVID-19 infection than if they had just left him alone.
Google warns against using its data to compare regions that are different from each other. Whether officials will pay attention to that warning remains to be seen.
Aggregated trend data can be useful, but that usefulness depends on reasonable, responsible decision-making. Unfortunately, when leaders haven't always grasped even how COVID-19 is transmitted—neither Georgia Gov. Brian Kemp nor New York Mayor Bill de Blasio seem to have understood until this week that asymptomatic people can spread the virus—there's reason to doubt that we'll see smart decisions.