Geography and Covid-19 Part 2 – Cellphone location data provide insights on physical distancing

Many Android and Apple apps record people’s location data unless this is disabled. While it can raise privacy concerns, the data – when anonymized and aggregated – can provide information about physical distancing during the pandemic. (© Richard McGuire Photo)

As B.C. and other jurisdictions around the world begin to lift physical distancing measures, how will we know how well it’s working?

Human behaviour never exactly mirrors the measures implemented by governments. Some people defy guidelines to stay at least at least two metres apart in public. On the other hand, just because a nail salon or gym opens again, it doesn’t mean people will return in pre-Covid-19 numbers. The public makes its own judgments about what is safe.

In a previous article, I talked about how geography can be used to understand the spread of the novel coronavirus. In this article, I’ll talk about how cellphone data can be used to estimate people’s movements and adherence to social distancing guidelines.

Most people are somewhat aware that cellphones track their moves unless location data is disabled. Tracking can be done either through triangulation from different cell towers, from GPS (Global Positioning System) satellite data, or both.

If I look in my Google Maps Timeline, for example, I can see red dots all over a map of the world indicating the places I’ve been and when I was there. This very personal information is normally only available to me, but if I were being investigated for a crime, law enforcement agencies could access to it.

Many phone apps from third parties also track users’ location data, even when the apps aren’t being used. Typically, companies only receive the data in anonymized form, often aggregated with millions of other users. This means they can see where people go without identifying the individual.

Privacy advocates have concerns, and there is great potential for abuse. But such data in anonymized and aggregated form provides a valuable tool in understanding where people are going during the Covid-19 pandemic.

Google regularly publishes a Covid-19 Community Mobility Report showing over time how visits to certain types of businesses or facilities have increased or decreased during the pandemic.

The most recent one for Canada, with data up to May 9, shows that use of transit stations in B.C. is 45% lower than the baseline. At the same time, use of parks is up 84% from the baseline. Not surprisingly, attendance at workplaces is down 35%, while time at people’s residences is up 10%. Retail and recreation is down 39%, but visits to grocery stores and pharmacies are up 3% from the baseline.

These numbers vary greatly from province to province depending on the severity of the outbreak and physical distancing measures imposed. In Ontario, where there’s a far higher infection rate, park use is down 32% and time at grocery stores and pharmacies is down 14%.

The numbers also change over time. In most jurisdictions, the peak time for physical distancing was late March and early April. Since then, it has been returning in fits and starts toward normal.

B.C. park use, for example, was well below the baseline in late March when many people took the order to “stay home” too literally, thinking that meant stay inside. When B.C. Provincial Health Officer Bonnie Henry started emphasizing the need to get outdoors and exercise, park use started increasing.

Provincial parks were closed between April 8 and May 13. The May 9 data doesn’t reflect the reopening, but subsequent reports should.

As usual, much more detailed data and maps are available for the U.S.

A company called SafeGraph is using data from third-party apps to track county-level data of Americans’ movements over the course of the pandemic. This includes analyses of foot traffic to various businesses as well as mapping of the geographic response to shelter in place.

The company uses a “Shelter In Place Index” to show in percentage terms how much people are staying home in every U.S. county as compared to a baseline of the seven days ending Feb. 12.

Not surprisingly, the highest compliance is in hard-hit urban counties and the lowest compliance is in low-density rural counties in states that saw few or only mild restrictions.

Topping the list is Queens, NY, with 26.83. Queens has one of the highest rates of Covid infections and deaths in the country. At the bottom of the list is McCone, MT, with -4.82 – suggesting even more people are out and about than in the February baseline period.

There are some surprises. Trousdale, TN currently has the highest infection rate on a per capita basis in the entire U.S. due to a major prison outbreak. Yet its score of -2.65 suggests fewer people are staying home – although the prisoners may have little choice.

Another measure, percentage staying home, also shows Queens at the top with 54.1%. At the bottom of the list is Lamar, AL, with 17.5%.

The date when the highest number of Americans stayed home was April 7 at 43.80%. The latest figure is from May 16 when 36.70% stayed home nationally. This suggests that while more people are getting out, there is still a lot of caution and many remain off work.

Georgia is the American state that was one of the last to close and first to open, even prompting a rebuke from President Donald Trump. Yet the peak of 41% staying home on April 9 in that state has only declined to 31.2% as of May 16 – suggesting that many Georgians are more cautious than their governor, Brian Kemp.

The SafeGraph data also looks at foot traffic to such businesses as movie theatres (way down) and supermarkets (close to normal except for a spike in late March when there was some panic buying).

Mississippi is that state with the closest-to-normal foot traffic. Not surprisingly, New York has seen the second biggest decline in foot traffic. The state with the highest decline is Hawaii, though the reason isn’t explained.

Foot traffic to bars is closest to normal in Montana, while Nevada’s bars have seen the biggest impact. Bars and casinos were closed in that state.

The Washington Post has used SafeGraph’s data to map the time people spent at home on various dates in every U.S. county. Four maps show the seven-day periods ending on March 1, April 1, April 7, and April 30.

The changes between the maps are dramatic. On March 1, the country was mostly wide open. A month later, the change was significant except in a few less populated states. April 7 reflects the peak of stay-at-home. The April 30 map more closely resembles the April 1 map, suggesting fewer staying at home, but still far from normal.

Researchers in the U.S. have also used the SafeGraph data, as well as data from Veraset, to determine which types of businesses are safest to visit. Their analysis looks at such factors as visitors per square foot, how long they stay, when they turn up and how much physical contact there is.

Golf courses are relatively safe because people are outdoors and can easily keep a distance from each other.

Gyms and fitness centres are much riskier because there are many visitors who stay for extended periods, often touching many surfaces and using the same equipment as others.

Businesses where people make a quick in-and-out tend to be safer than those where people stay longer. And of course businesses with tightly packed customers or where there is physical contact have added risk.

The SafeGraph tracking even looked at visits to specific brand stores in the U.S. Not surprisingly, as home-bound people took on reno projects, visits to Home Depot were considerably higher than the other stores measured. At the bottom of the list was Starbucks, where many states don’t allow sit-down service. Target and Walmart did better than Costco. McDonalds was around the middle of the pack most recently.

Interestingly, Starbucks and McDonalds topped the list in late February and early March. In mid-March, as people grasped the reality of the pandemic and decided to stock up and hunker down, Costco saw a surge.

Cellphone data also provides insight into what happens when one jurisdiction opens up more aggressively than its neighbours.

A week after Georgia opened sit-down restaurants, hair salons and other businesses on April 24, visitors flooded in from neighbouring states at a rate of around 62,000 additional trips per day.

Travel between different states and provinces can spread the virus, so this result concerned researchers.

These types of uses of cellphone data that deal with anonymous aggregate information can provide important insights on public movement during the pandemic.

But privacy concerns are likely to increase as various options are discussed for contact tracing of individuals infected or exposed to the virus. That’s a topic for a future article.

Author: Richard McGuire

Richard McGuire is an Osoyoos photographer who worked at the Osoyoos Times between 2012 and 2018, first as reporter and then as editor. He has a long career in journalism as well as research, communication and management at the House of Commons in Ottawa and in the federal government.

Leave a Reply