The research aims to capture the vibrancy and expansiveness of NYC's nightlife landscape by interrogating how people access it: with Citibike, subways, or for-hire vehicles (FHV). Nightlife is a major economic and cultural driver for NYC, and our initial research goal was to use multi-modal time series transportation data as an advocacy tool to help local stakeholders in the nightlife industry (dance clubs, music venues, bars, and more) demonstrate their contribution to NYC’s citywide and hyper-local night-time economies. With the arrival of the COVID-19 pandemic, we’ve expanded our research goal to examine how the night economy has been uniquely impacted by the crisis in the hopes of enabling policymakers to develop a nuanced understanding of the role of nightlife in the City and develop a uniquely targeted package of policies and aid that ensure its continued vitality. The resulting drop in public transportation usage presents urban scientists with an incredible research opportunity– we’ve built a baseline model of the transportation landscape as it looked in comparable months from 2019 and will explain exactly how each mode (bike, subway, and taxi/for hire vehicle) were affected at each step of the onset of the pandemic.


Problem Statement

1. Where does NYC's nightlife occur? Is it centralized or diffuse? Are there specific (or surprising) clusters?
2. How do people access nightlife in NYC and are there any predictors of difference in mode choice?
3. What is the citywide and borough-wide decrease in transportation utilization due to Covid-19?







Spatial Analysis

  1. A key component of our work involved figuring out how to compare data aggregated at multiple geographic levels. The taxi pickups/drop offs, for example, were aggregated to Taxi Zones. This leaves analysis performed on that data susceptible to the Modifiable Areal Unit Problem (MAUP) a classic form of statistical bias in spatial analytics, where individual spatial phenomena (in this case, taxi origin/destination data) are aggregated into geometries of arbitrary shape and size and the resulting aggregates are subsequently misleading or obfuscating. We attempt to minimize this bias by decomposing the data through areal interpolation (Fig. 4.2).

  2. The first map in Figure depicts the citywide hexes, collecting locations of nightlife venues, with a z-score (number of standard deviations from the mean) of 2 or greater. Given that mu = 3 nightlife venues and that sigma = 11 venues, that means each of the hexes depicted contains at least 25 nightlife venues. That's quite a few venues, considering that the hexbins represent just a few square blocks, and it's unsurprising that that kind of density doesn't exist outside of Manhattan or the parts of Brooklyn closest to it along the L train. In an attempt to de-bias our nightlife clusters, we standardize the data by the share of businesses in a given hexbin that are venues rather than the absolute count. We filter the hexes to only include those with a minimum of 150 total businesses and having a venue share z-score of at least .5 (or ~2.5% of the businesses categorized as nightlife). The application of those filters produces the second map in Figure, which we can see captures a much more geographically diverse spread of nightlife (Fig. 4.1).






We use D3.js and the MapboxGL Javascript API to construct a data visualization interface. Users can monitor daily or nightly changes of inflow and outflow for different mobility usage in hexagons or music venues’ hotspots by clicking on the map. Time series analysis can help to get insights for anomaly identification and trend summary for selected cases. On the other hand, the interface allows users to compare the observation in 2020 with the baseline in 2019 to show differences in mobility demands potentially caused by the outbreaks of COVID-19. The changes in demands can be correlated to the milestones of government actions.



To further analyze the hidden factors over the changes in daily usage of transportation tools during night-time, we initiated a multi-variable regression. In the regression model, we separated three main transportation mode during night-time and research for the correlation with demographic features, built environment features (retail area, commercial area, residential area), weather features (temperature, precipitation), temporal factors (day of the week, month, covid-19), and traffic-related factors (station counts in the area).



The first period, between Feb 12 and Feb 28, marks the normal condition. The second period, between Mar 1 and Mar 17, is when people started to be aware of the outbreaks while business and venues were operating. The third period, between Mar 18 and Apr 3, is when all nightlife had been closed and the city started to lock down.

MTA usage presents a slight loss in citywide and borough-wide statistics in period 1, however it tends to remain in the normal fluctuation (Figure 5.1, 5.2). In period 2, it shows an increased reduction around 10-15% in general. Losses at night are greater than day-time in every sector. Also, Manhattan holds the biggest loss for both day and night transit, followed by the nightlife clusters which indicates that people rely on taking subways to access those venue places. Citibike performance shows a reversed trend. There is a boost in usage around 30% in both period 1 and 2 (Figure 5.3, 5.4). The outbreak of COVID-19 doesn’t seem to interfere with bike riding in the first place. After venues were closed and all business have been shut down, MTA and Citibike show the significant loss in around 80% and 60% respectively.


Result of Regression

During night-time, New Yorkers prefer using taxis as their transportation mode. If the destination is Manhattan, FHV will have more night-time inflows as well as Citibike. But not the subway. The borough that people love taking MTA at night is Staten Island. Once COVID-19 hits on NYC, fewer people are visiting Brooklyn and Manhattan at night. The population and built environment factors have a minor impact on the transportation inflows.

Manhattan has a strong influence on venue locations. Citibike as the last mile transit solution has a higher positive coefficient on venue counts. Different land use has less impact on venue counts. Overall, venue locations have strong relationships with boroughs, transportation station locations and little relationship with populations.

New Yorkers prefer to use FHV during weekends and the subway during the weekdays. Citibike does not have a dramatic difference on weekends and weekdays. Once it is nighttime, all transportation usages decrease, especially the subway. Rainy days will have a bigger impact on subway ridership but minor impact on Citibike. For FHV, we do not have significant evidence to indicate the relationship with weather.





Spatial Analyst/ Data Engineer



Data Engineer/ Data Visualization



Data Analyst/ Data Modeling



Data Analyst/ Data Modeling



unnamed (1).png



VibeLab is bringing the nightlife movement around the world
Many cities are creating new nightlife policy. The time is now for your city to grow as a global nightlife capital 
We are having active conversations with Bogota, London, Los Angeles, Mumbai, Sydney, Tokyo, Vancouver, and Amsterdam, and several more international nightlife capitals about bringing the tools and experiences to their cities




New York University’s Center for Urban Science And Progress (CUSP) is an interdisciplinary research center dedicated to the application of science, technology, engineering, and mathematics in the service of urban communities across the globe.



Smart Cities Postdoctoral Associate/ Mentor

Kim Mahler is a Smart Cities Postdoc affiliated to both NYU CUSP and NYU WIRELESS. His research interests involve the development of viable 5G applications, drone/vehicular communications, localization using 5G millimeter wave and user-centric innovation developments. He received the M.Sc. degree with honors and the Dr.-Ing. degree (magna cum laude)