Where did New Yorkers hide from the pandemic and how do they return?
Intro
To explore how New York City (NYC) community changed over the pandemic, we took data from GPS aggregators and looked at the relocation patterns of New Yorkers choosing to leave their homes until the pandemic is over, or even longer.
In the previous article, we explored people’s outflow from NYC after the first pandemic news, COVID cases and lockdown measures (winter-spring 2020–2021). The mobility between NYC and its hinterland continued in the following months. New Yorkers were not only leaving the city but also started to return when work and education activities resumed.
To explore this further relocation and return to the city, we selected two additional periods for analysis, between June-September (summer) and October-December (autumn). The study is comparing pre-pandemic mobility in 2019 with the following two years, 2020 and 2021, to see whether a recovery can be observed in the current relocation patterns.
The results of this continued study are presented below.
It is clearly seen that the mobility in the last months of 2021 signals return to normal, with outflows from NYC having similar geography to pre-pandemic period, and inflows getting stronger.
Data
Habidatum has a long history of working with mobility data — both from telco and GPS aggregators. Besides the basic features like the number of user signals, we identify origin-destination points for all users.
Origin and destination points are often used for transportation planning. We tested it in many countries, USA inclusive. One of the latest examples are our studies on the night bus network plan for Florida Department of Transportation (FDOT), and the analysis of the impact of the expansion of Heathrow (London) airport on passenger and local population mobility patterns.
In this article, we look at the change of people’s consistent nighttime and daytime activity locations — presumably home and work. We used aggregated anonymized GPS data coming from mobile phones apps, like e-commerce, transportation, weather, sport, quantified self and the like.
Apart from basic signals, our data providers offer such attributes as the home area of a user, updated weekly, and calculated via the most frequent location of the user’s long stays (dwell time) during the night hours.
Methodology
As the home area of a user can be updated weekly, we can understand the change of the home location of all users over time with a high level of granularity. So we looked at the home locations of all available users at the start and at the end of the chosen time intervals. The intervals are shown below:
- June 2019 — September 2019 (~pre-pandemic summer)
- June 2020 — September 2020 (~post-first-lockdown summer)
- June 2021 — September 2021 (most recent summer)
- October 2019 — December 2019 (~pre-pandemic autumn)
- October 2020 — December 2020 (~post-first-lockdown autumn)
- October 2021 — December 2021 (most recent autumn)
We took all users identified as residents of NYC counties / boroughs in either one of the periods mentioned above. Then we explored where they lived at the start and the end of each interval and looked if they relocated (changed home area) or not. Those who relocated were grouped by origin and destination county (see tables and maps below).
All data is aggregated on a county and block group level. It can also be seen by more granular locations such as streets and blocks or small grid cells with a side of 50 meters or lower, being still anonymized, aggregated and fully compliant with the regulations and rules on private and sensible data.
1. Comparing inflows and outflows
1.1. Summer 2019–2020–2021
In summer, more people are leaving the city than returning.
However, in 2020 the negative difference between the outflows and inflows was the smallest as probably fewer people were leaving the city for vacation purposes (they either left before, during the first lockdown or stayed in the city).
The number of relocations was overall x1.5 higher in summer 2019 compared to 2020 (and x 1.3 higher in 2019 vs. 2021, correspondingly). The number of vacation trips shrunk in pandemic period and is not fully recovered.
1.2. Autumn 2019–2020–2021
Unlike the summer patterns, autumn signals a recovery.
The number of relocations in autumn has an overall less significant difference between 2019-2020-2021 than in summer.
Also, in autumn 2020–2021, the inflow to NYC exceeded the outflow (unlike 2019). People who relocated in the first lockdown started returning to NYC more actively (when the school year started and gradual reopening continued).
2. Geography of relocations
2.1. From NYC
The closest NYC surroundings are the key relocation areas. Among them, the leaders are Long Island counties Nassau and Suffolk (east of NYC borough Queens), as well as the second-wealthiest county in NY, Westchester (north of NYC borough Bronx, Hudson Valley).
In summer, Monmouth, NJ, the northern part of Jersey Shore, the famous place for boating and fishing, is also among the leaders. The share of relocations to Monmouth is twice larger in summer than in autumn. The same pattern is observed in Suffolk, the county locating famous vineyards (Long Island American Viticultural Area), beaches and high-end estates. Among the farther locations, Florida (Broward, Miami-Dade, Palm Beach) and California (Los Angeles, Orange) are the top destinations.
Overall, the high-end areas are the most attractive for relocation, which is seen on the zoom-in relocation maps by census block groups below.
In autumn, Hudson (Jersey City) in 2019 and Bergen (the most populous county in New Jersey) in 2021 were among the leaders, and there’s also a larger share of relocations between NYC boroughs (including Staten Island) which signals the shift from vacation areas to offices and commercial hubs. The same can be mentioned about more distant relocations: in autumn, Chicago is among the leaders (Cook County, the second-largest by population in the US after Los Angeles), but only in 2019 and 2021 (it is not in the top-list of 2020, see table below).
The general geographic patterns of relocation in 2019 and 2021 are similar, while there were more differences in the peak pandemic year of 2020. This can most clearly be seen on the zoom-in relocation maps.
2.2. To NYC
The closest NYC surroundings are both attracting and originating relocations. Stable leaders are Nassau, Suffolk (Long Island), Westchester (Hudson Valley).
In autumn, Monmouth, NJ, is also among the leaders, like the more distant Orange County, CA: more inflows in summer, more outflows in autumn.
Among other distant origins of returning trips to NYC, the leaders are Broward, Miami-Dade, Palm Beach + Osceola in summer 2020 in Florida, Los Angeles in California and Chicago (Cook County, only in 2019 and 2021, similarly to outflows).
The geography of returns to NYC differs a lot between summer and autumn:
- 2019 and 2021 are more similar in summer (for outflows (section above), this is both summer and autumn),
- autumn relocations on the map look more similar for 2020 and 2021, especially within the Greater New York area (these are also two periods with inflows to NYC exceeding outflows).
End of 2021 signals NYC recovery:
- more people return than leave,
- returns are from the areas that accepted more relocations in the first COVID months,
- the geography of outflows from the city is similar,
- more business areas are involved in relocations.
Terms
Data: GPS mobile phone data
Relocation: users get qualified as relocated if their home locations at the beginning of the observed time period are different from their home locations at the end of the period.
Home location: location where user lives (down to block group level), defined by a ratio calculated on the following four variables:
- Number of days user spent in that location in the last month
- Daily average number of hours spent in that location
- Time of the day spent in the location (mostly nighttime)
- User records have been present in the dataset for at least 15 days prior to the observation point
By New York City or NYC in this text we mean 5 counties = 5 boroughs of the city.
Authors: Daniel Gorokhov, spatial data scientist at Habidatum, Katya Letunovsky, VP at Habidatum
