How has subway ridership changed since the COVID-19 pandemic in NYC neighborhoods with varying language access needs?
In my previous post, I examined which neighborhoods in New York City have more Limited English Proficiency (LEP) residents as well as whether those neighborhoods tend to have more public transit commuters. To recap the findings: LEP residents are concentrated in relatively few areas, and those neighborhoods do commute by public transportation more than the median NYC neighborhood. There were two major limitations to that analysis, however. Commute method data from the Census served as a proxy for actual use of public transportation, and the data from 2016–2020 likely did not fully account for the effects of the COVID-19 pandemic on transit ridership.
This analysis attempts to fill in the gaps by examining actual subway ridership from recent weeks in 2022 and from 2019 to see how it has changed for stations in different parts of NYC. Given the reporting on working-class neighborhoods driving subway ridership recovery, I hypothesized that stations in neighborhoods with more (or greater proportions of) LEP residents have recovered more of their pre-pandemic ridership. If so, the data would further show the need for language access improvements in those neighborhoods’ stations.
Have subway stations in NYC neighborhoods with a higher proportion of LEP residents recovered more of their pre-COVID-19 ridership?
To answer this question, I first calculated average daily ridership (in terms of entries) for each station in the past four weeks from the MTA’s turnstile data. The steps I took in R to do so were adapted from Chris Whong’s approach. Then, I brought in 2019 ridership data for comparison. Lastly, I located each station within a census tract and a neighborhood tabulation area (NTA) and related it to data about LEP residents in the area. See the Methodology section for more information.
The data indeed shows a positive correlation between ridership recovery and the neighborhood proportion of LEP residents. Ridership recovery in this case means the average daily entries in my 2022 sampling period (10/08/22 to 11/05/22) as a percentage of average daily entries in 2019. In other words, stations in “LEP-dense” NTAs have recovered (or perhaps kept) more of their pre-COVID-19 ridership.
Notably, stations in the most “LEP-dense” neighborhoods (about 40% LEP or above) all sit above a certain “floor” of ridership recovery. That floor of about 37% ridership recovered could signal a baseline level of ridership in those neighborhoods because of reliance on public transportation. By contrast, for many stations in neighborhoods with lower proportions of LEP residents — including some large stations — ridership recovery remains below that floor.
Accounting for station size, how has ridership changed in neighborhoods with varying LEP population sizes?
Even if ridership recovery rates tend to be higher for stations in “LEP-dense” neighborhoods, stations differ greatly in terms of ridership size. A large station like Times Square-42nd Street that has lost a greater proportion of its riders could still see more daily entries than an outer-borough local station that has lost a smaller proportion. To address this, I examined total entries on an average day after grouping stations by their neighborhood’s LEP population size.
Unsurprisingly, stations in the lowest quintile of neighborhood LEP population account for the most riders on an average day because they include station complexes in central districts with few LEP residents. Examples include Grand Central-42nd Street and the Fulton Street complex. It’s worth noting that the riders at these stations likely include many non-NYC residents as well, whereas stations in more peripheral neighborhoods are likely to have mostly local residents as riders. This fact calls attention to a limitation of this analysis: riders entering a station can come from neighborhoods other than the one in which the station is located.
The decline in ridership is stark for all quintiles, but at a glance seems to be steeper for the lowest quintile and gentler for the highest quintile (the stations with the most neighborhood LEP residents). To confirm this difference, I looked at the share of total subway ridership for each quintile on an average day.
As suspected, the stations in neighborhoods with the most LEP residents (quintiles 4 and 5) have seen their share of total ridership on an average day increase since the pandemic. On the flip side, stations in neighborhoods with fewer LEP residents now represent a lower share of total ridership.
While a snapshot of recent weeks in 2022 does not guarantee future trends, the shift in ridership share from 2019 to the present should at least push the MTA to consider how much of their core subway ridership is coming from areas with many LEP residents. Riders in those neighborhoods likely have less ability to telecommute and are thus continuing to ride the subway, so language access improvements are even more important today than they were before the pandemic.
To illustrate the need and potential for cost-effective investments in subway language access, we can use a ridership comparison with the Long Island Rail Road system. The top quintile of subway stations in terms of neighborhood LEP population saw over 375,000 entries on an average day in the past 4 weeks. That’s despite representing “only” 20% of total subway ridership, as shown in the chart above. Those 375,000 entries are well above even the estimated ridership on an average weekday in October 2019 for the entire LIRR system (fewer than 300,000). This bears repeating: the recent daily ridership — accounting for both COVID-19’s impacts and weekend lulls—of just the one-fifth of stations serving the most LEP residents exceeds even the pre-COVID-19 average weekday ridership of the entire LIRR network.
The soon-to-be completed East Side Access project serving only LIRR riders will end up costing over $11 billion. How much benefit could be had from language access improvements in subway stations made for, say, $10 million?
Conclusion
The connections shown here between ridership recovery and neighborhoods with more (or greater proportions of) LEP residents help to confirm assumptions about the changing nature of who tends to be riding the subway. Both the sheer volume and the increased share of riders who would likely benefit from language access improvements should provide support for those kinds of investments.
That said, there are several limitations to my methodology. First, because of the lack of demographic information about subway riders, neighborhood LEP population and proportion for each station stand in as proxies, even though LEP residents could be taking the subway at lesser or greater rates than the general populace within any given neighborhood. Second, as mentioned earlier, using the surrounding neighborhood to characterize a station does not account for riders entering that station from other areas, such as those transferring from buses from other neighborhoods. Third, using entries to estimate ridership does not account for in-system transfers. Lastly, 12 stations were excluded from the analysis due to lack of data (see methodology), potentially affecting results slightly.
In addition, the four weeks’ data that I used to study recent ridership may reflect trends that are unique to this time of year, which may skew the 2022 data compared to the 2019 data taken from that entire year. Future analyses could study the entirety of 2022 to compare results.
Also worth studying would be bus ridership, which is a component of public transit important to study because bus riders are more likely to be low-income and foreign-born than the average New Yorker. I hypothesize that there is probably a correlation between bus ridership and LEP status as well.
Methodology and Sources
See comments in the R code for detailed methodology, data dictionary for descriptions of the data analyzed, and sources for the data sources.
