January 2023 update:

Beginning with the week of June 13th, 2022, the source of mobile device data changed from Safegraph to Spectus.

To protect user privacy, Spectus does not report information on 'sensitive' points of interest- a subset of which are present in Safegraph data. Furthermore, there is no way of reconciling whether the two sources use the same panel dataset. The delivery method of Spectus device information necessitated the use of polygons to make spatial queries over the study period. Rather than collect information on each city's full boundary, only downtown areas were queried during the initial phase of this transition to account for the size and computational complexity constraints of repeating this operation for 66 cities over 1434 days.

Reconciling the reported device counts from the two sources was performed like so:

Spectus data was first subset on the records that were most likely to have been collected in a manner similar to Safegraph. This was done using a t-test for each week for all cities that compared the distribution of device counts between the two sources. There was insufficient evidence to reject the null hypothesis that the paired samples came from the same distribution over the entire study period for all but 3 cities: Albuquerque, Quebec, and Honolulu. Halifax was excluded from this test due to there being no recorded devices collected according to the method closest to Safegraph's prior to May 2021.

The Spectus values were scaled by the average of the observed Safegraph values over the most recent 3-month interval then appended to the existing time series. The weekly downtown RQs from June 2022 through November 2022 and seasonal averages were then computed according to the methodology used previously.

As the team continues this work with Spectus, we aim to incorporate city recovery quotients and relative location quotients with the updated data.

End of updated methodology


Point of Interest. From Safegraph:

A point of interest is a specific physical location which someone may find interesting. Restaurants, retail stores, and grocery stores are all examples of points of interest.


In the United States, the smallest unit of observation in this study is the ZCTA. For Canada, it is a census dissemination area.


The set of observations in a city with the greatest employment density.


Measures the number of raw POI visits (scaled by state sampling rate) in post-COVID period vs pre-COVID period per geography.


Recovery Quotient:

\[ RQ_{2022} = \frac{\text{Downtown POI visits in 2022}}{\text{Downtown POI visits in 2019}} \]


Measures the proportion of POI visits in downtown vs. the rest of the city in post-COVID period vs. pre-COVID period.


Location Quotient:

\[ LQ_{2022} = \frac{\frac{\text{Downtown POI visits in 2022}}{\text{Entire City POI visits in 2022}}}{\frac{\text{Downtown POI visits in 2019}}{\text{Entire City POI visits in 2019}}} \]

What is an observation?

In the United States, the smallest unit of observation in this study is the ZCTA. For Canada, it is a census dissemination area.

Definition of downtown

Observations with the greatest employment density. See to this document for full list of downtown areas by zip code and map

Time periods

March 2020 - May 2020

June 2020 - August 2020

September 2020 - November 2020

December 2020 - February 2021

March 2021 - May 2021

June 2021 - August 2021

September 2021 - November 2021

December 2021 - February 2022

March 2022 - May 2022

June 2022 - August 2022

September 2022 - November 2022

Comparative Map

Displays 2D maps of all metrics and all periods for all cities at the national level. A value greater than 100% indicates an increase in activity from the comparison period. A value less than 100% indicates the opposite. A value equal to 100% means the activity level has not changed between the two periods.

The purpose of this map is to provide an overview of all metrics over all periods for all cities in the study.

Relevant libraries

dplyr, tidyverse, leaflet, geojson

Back to top

Recovery Rankings

Displays a ranked bar chart wth city names and metrics superimposed on the bars. The bars are filled according to which region they belong to (Canada, US Midwest, US Northeast, US Pacific, US Southeast, US Southwest).

A value greater than 100% means that for the selected inputs, the mobile device activity improved from the comparison period. A value less than 100% means the opposite, and a value equal to 100% means the activity did not change. Since all cities in the study were included because they were the largest cities in Canada and the United States, ‘large’ in this case refers to the largest 20 cities in the study. All other cities are considered ‘medium’.

Relevant libraries

dplyr, tidyverse, ggplot

Back to top

Recovery Patterns

This chart displays a rolling weekly average of the selected metric. The mobile device count for a given week during the pandemic is divided by the mobile device count for the corresponding week pre-pandemic. Safegraph reports weekly data every Monday. Each reporting date is converted to the ISO week number that it belongs to. Weeks are considered to correspond with one another if their ISO week number matches.

Relevant libraries

dplyr, tidyverse, ggplot

Back to top

Explanatory Variables

Plots the relationship of a selected socioeconomic factor and selected recovery metric. This plot takes all cities into consideration, but a select few can be highlight for comparison’s sake. You can click on any point in the plot to see the exact value of the selected independent and dependent variables, rounded to two decimal points. The points are colored according to region.

Relevant libraries

dplyr, tidyverse, ggplot

Back to top