Data: We aggregate the total number of trips originating from each census tract within the following ranges of dates:

04-01-2019 to 10-18-2019

04-01-2020 to 10-18-2020

We refer to them as 2019 and 2020 data and consider them pre and post Covid lockdowns respectively.

The reasons for the choice of these date ranges are:

  • The starting date of 2020 date range is the first of the month following March 22, the first day when the stay at home order went into effect in NYC. And the latest date Replica has data for 2020 is 10-18.
  • To accurately compare 2019 travel behavior with 2020’s, we use the same exact start and end days and months for 2019.

We can aggregate trip counts per census tract based on either the location of origin or destination of trips, and we choose to use the data aggregated based on the trip origins. Our reasoning is that when we look at both the origin and destination-based aggregations, we see that they were very similar; presumably because Replica counts a return trip as two separate trips if a non-travel activity takes place at the destination: “If a person walks from home to a cafe, buys a coffee, and then walks to work, two trips have occurred.” ( In this case both the person’s home and the cafe are each counted once as origins and once as destinations.


Map 0: Change in the total number of trips & Outliers

First we look at the change in the overall number of trips recorded by Replica and visualize the difference in this number between 2019 and 2020. The dark blue census tracts show a positive change, however, these are tracts with a very small population and most are open spaces so we consider these outliers and remove them from our data set.

From here on, we do not visualize the census tracts where total trip numbers have increased from 2019 to 2020.


Maps 1 & 2: Change in the total number of trips and % of change in the total trips.
Assuming Replica collected data from a similar number of people for both date ranges, we first compare the total trip numbers originating from each census tract. Lighter colored census tracts that are concentrated in Manhattan show more than 40% decrease in the number of trips recorded and census tracts in dark red, scattered in Brooklyn, Bronx and Queens show that there has been little decrease in the number of trips recorded before and after the pandemic.

(data aggregated per NTA)

Maps 3-6: Change in the total number of bike trips and % of change in the total bike trips.
We are interested in the cycling behavior, therefore, again assuming that Replica tracked a similar number of people during the two date ranges, we look at the change in the absolute number of bike trips for each census tract, as well as the percentage of change in the absolute number of bike trips. Here we see that, North Brooklyn, Long Island City, Harlem and Upper West Side neighborhoods saw a considerable increase in bike use, despite the overall decrease in the trip numbers seen in the previous map.

(data aggregated per NTA)

The percentage of change in the number of bike trips

(data aggregated per NTA)

Maps 7 & 8: Change in the mode share for cycling and public transit.  
What is the change in the percentage of trips made on a bike from 2019 to 2020? Due to the initial modal share of bike use being low, the change in modal share rarely went over 1, however, the increase in modal share from 2019 to 2020 was over 50% and often 100% in many Brooklyn census tracts.

What is the change in the percentage of trips made using public transit from 2019 to 2020? What is interesting for us in this map is the areas where public transit dropped less than 50%. We anticipate that the population living in these areas, shown in reds had a sustained demand to use public transit even during the pandemic months, and  a majority of them are likely essential workers.


Maps 9 & 10: Percentage of essential workers to the population +16 years of age per census tract*
To double check our assumption, we look at the % of working population who are essential workers per census tract. We are interested in seeing if there are any overlaps between areas with a high population of essential workers and sustained transit ridership. We expect to see transit ridership not going down in areas where a high % of essential workers live.

(data aggregated per NTA)

* Based on the list of occupations considered as essential in the below Boston study and Census 2019 data (Table 2401)
(NYC essential occupations list is very detailed to filter census data with):

Charts 1 & 2: Which census tracts had the highest gains in bike trips while not losing a lot of transit riders? Where do the most number of essential workers live?
We assume that if the public transit ridership did not go down significantly after the start of the lock down in a region, it is likely that a higher number of essential workers live in that area, compared to the others where transit ridership decreased significantly.

(data aggregated per NTA)

Map 11: Map of census tracts that have gained bike trips with bike lanes overlay


Map 12: Crash data with bike trip gains and bike lanes
Crash data is from within the date range of 04-01-2020 and 10-18-2020. Small dots represent crashes with cyclist injuries and large circles, crashes with cyclist deaths.

Map 13: Which census tracks have bike lanes, and which need them?
The census tracts overlaid with shades of blue have either a protected bike lane, standard bike lane or sharrows passing though it, those that are not, do not have any. Underlying colors show the percentage of change in cycling from 2019 to 2020. Where we see no blues and bright-dark reds are tracts where people used bikes significantly more in 2020 than they did in 2019, but there are no bike lanes on their streets.

Maps 14 & 15: Census tracts where: (1) Increase in cycling from 2019 to 2020 is above the median, (2) There are no protected bike lanes and (3) There has been at least one crash where a cyclist was injured or killed between 04-01-2020 and 10-18-2020. Hover over each tract to see more detail.

For this map, bike lanes were intersected with the census tract polygons, and the facility class (I,II,III or Link) of the longest intersecting lane is assigned to the census tract. We exclude census tracts intersecting with protected bike lanes (Facility Class I). We keep those that have negligible intersections.

The same census tracts are rendered based on the percentage of essential workers to the population +16 years of age.

Extra visuals
What is the change in public transit use?