Mobility trends during the pandemic
Written by: Chloe Lubin
We know that social distancing is key to “flattening the curve,” though the ripple effects of immobility are disastrous to people around the world. The impact is felt on a global scale as countries prepare for a recession. The impact is felt by companies that have suffered from a loss of clientele and are struggling to keep cash flows steady. The impact is felt by people who have lost their jobs.
As the world comes to a standstill amidst the pandemic, how have shelter-in-place and lockdown measures affected global mobility?
To answer that question, I will use Google’s community mobility dataset, publicly available here. It tracks data for 132 distinct countries between February 15 and May 2, 2020. Each datapoint represents a percent change (determined based on the median value computed for the period of 01/03–02/06/2020, for each day of the week) in visits and length of stay at the following location types:
Grocery and pharmacy (grocery markets, food warehouses, farmers markets, specialty food shops, drug stores, and pharmacies, etc.)
Parks (local parks, national parks, public beaches, marinas, dog parks, plazas, and public gardens)
Transit stations (subway, bus, and train stations, etc.)
Retail and recreation (restaurants, cafes, shopping centers, theme parks, museums, libraries, movie theaters, etc.)
Residential areas
Workplaces
Google developed the dataset to remediate the impact of COVID-19. By tracking people’s movements to areas that are generally crowded, its goal is to inform public health officials of potential movements that may contribute to the spread of the virus.
Using Tableau, I computed a 7-day moving average of the percent change to filter out the noise from random percent changes over time. I also added a trendline to visualize patterns in human mobility over time. A polynomial one was the best fit without over- or under-fitting the model. Feel free to check out the interactive visual here.
Based on the community mobility dataset produced by Google, my top observations are the following:
After a number of countries enforced lockdown measures mid-March, mobility to high-traffic areas (parks, transit stations, retail and recreation areas, and workplaces) dropped substantially.
At the same time, visits and length of stay at residential areas increased.
We observe increased mobility at the end of April, which coincides with the time when a number of countries started relaxing confinement measures.
A few comments about the dataset:
1. The data is aggregated and anonymized based on users who have turned their Location History setting on. Users who don't have that setting enabled are therefore excluded from the dataset.
2. A datapoint is only included if it meets Google's privacy threshold. If a location isn't busy enough to ensure anonymity of the person visiting the location, Google does not record a change for the day, which explains the presence of null values in the dataset.
3. Google explains that the data used to compute the percent change in mobility “depends on [a user's] connectivity,” among other things. For some areas with poor connectivity, it's fair to assume that the data may not be captured in the dataset.
For the reasons above, we should be careful about using Google’s community mobility data to estimate larger population parameters. The sample nevertheless presents interesting patterns, visualized above.
The graph shows that most high-traffic locations (groceries & pharmacies, workplaces, transit stations, etc.) have seen a considerable decrease in visits. Retail and recreation locations were the hardest hit: visits dropped by 61.33% on average between March 13, 2020, and April 26, 2020. The decrease happened abruptly mid-March, when a majority of countries announced shelter-in-place orders or lockdown measures to combat the spread of coronavirus.
The only location type that has seen an increase in mobility is residential areas (+20.59% between 03/13 and 04/26/2020). At the same time that people weren't visiting other locations, they increasingly remained in residential areas. I was surprised to see that the percent change of visits to residential areas didn't vary as much as other location types. It may be because people already visited residential areas or stayed at home a significant amount of time to begin with, explaining the low variance of visits to residential areas compared to the baseline.
Although mobility to locations central to economic activity has considerably decreased, the trendline suggests that activity is picking up again while residential mobility (or immobility) is slowly decreasing. This pattern coincides with new lockdown relaxation measures announced by a number of countries at the end of April. Increased mobility to high-traffic areas will facilitate the spread of the virus and public health officials will have to develop creative policies to combat the virus as people slowly come out of their homes.
Note: The data was last accessed on May 4, 2020. Google refreshes the data periodically, so some of the newer mobility data points collected aren’t reflected in this analysis.
Sources
Google LLC “Google COVID-19 Community Mobility Reports.” https://www.google.com/covid19/mobility/ Accessed: May 4, 2020
https://www.blog.google/technology/health/covid-19-community-mobility-reports?hl=en
This article originally appeared on The Medium