Does Social Distancing Work? (Yes)
When the Covid-19 outbreak took the United States by storm in late March 2020, I began collecting data on confirmed cases and deaths from the virus and uploading it to my company GitHub page. The numbers have been staggering —1.3 million confirmed cases and around 75 thousand deaths in the US as of May 11. In about two months, Covid-19 has fundamentally altered human life in the US and around the world.
The scariest thing about this virus is how little we know about it. It has been widely reported by now that the United States was late to the game when it comes to testing. Serological studies suggest actual Covid-19 infection totals likely exceed the number of confirmed cases in many areas by several orders of magnitude.
The other thing about the virus, however, is that it affects people. And we know a lot about people. By focusing on what we already know, we can develop a better understanding of the virus, how it spread through the country, and how to stop it.
So to answer the question that titles this article, we can take a look at what we know well about people: where they are. Data-savvy corporations know far more than many may realize about where we are, thanks to the abundance of location data we generate with our smart phones. A practical use of this data comes from the company Descartes Labs, who has generated detailed aggregate statistics on social mobility. From their site:
“Our methodology looks at a collection of mobile devices reporting consistently throughout the day. We calculate the maximum distance moved in kilometers (excluding outliers) from the first reported location. Using this value, we calculate the median across all devices in the sample to generate a mobility metric for selected [regions].”
Using DL’s social mobility index data and the date of each state’s stay-at-home order, we are able to quantify at scale the effect of social distancing guidelines on human behavior. By matching this information to Covid-19 data from Johns Hopkins University, we can analyze the effectiveness of stay-at-home orders as it varies from city to city.
To illustrate these effects, I created the map below. The circles represent the 107 most populous metropolitan counties in the US. Circle color represents the level of adherence social distancing guidelines (i.e., the reduction in overall mobility after their lock down order). Circle radius represents the number of new Covid-19 cases (per million people) in the 10-day period immediately following local stay-at-home orders.
From the map we can see a few discernible trends, such as regional variation in adherence to stay-at-home orders. Compare, for example, the small red bubbles on the West coast to the big yellow bubbles throughout the South. However, the most important pattern to notice is that the smallest circles tend to be red. This means that the counties which saw the highest reduction in mobility also saw the smallest increase in confirmed cases (kudos to California for all those little red bubbles).
In fact, I found that among the 107 largest US counties, there is a significant negative correlation between our measure of adherence to stay-at-home orders and the number of new cases per million people. Thus, it seems as if social distancing, in isolation (no pun intended), has been a very effective model for containing the spread of the virus. It also seems as if certain states are either taking stay-at-home orders more seriously or are simply more capable of maintaining social distance than others.
Now that we have the data to show demonstrate this, maps like the one in this article (or like this one, or this one) can act as a sort of social score card. While our location data is very useful for analyzing the policies and interventions that keep us safe, its use by private companies or government agencies has profound implications for society.
Thanks for reading.
