The Characteristics of Active Airbnb Listings in Manhattan
Using 10 years’ record (2009 – 2019) of short term rental activity scraped from the Airbnb website, we investigate how the activity varies spatially and temporally in Manhattan. The number of listings can be used as an indicator of short term rental activity and how it changes over time. Having information about the most dynamic neighborhoods can help service sector targeting tourists by knowing their preferences (price, privacy, proximity to main attractions... etc.) and identify clusters (if any). In addition, this data can provide some insights into the recent banned on short-term rental in NYC and how it affects the spatial distribution of these listings.
The focus of the analysis presented here is to test two hypotheses. First, the majority of listings are concentrated in Midtown Manhattan regardless of time. Second, the lower homeownership percentage in one neighborhood the higher the number of listings.
The spatial distribution of Airbnb listings in New York City.
1.Data Preparation
The data was scraped from the original Airbnb website in September 2019. As part of the cleaning process, listings without reviews were removed as well as any hosts with more than 4 listings. This filtering allows focusing on short-term rentals posted by individuals rather than companies. Find more about the cleaning process here.
The number of records in New York City by borough.
2. Hypothesis A:
The majority of listings are concentrated in Midtown Manhattan regardless of time.
2.1 Density Mapping
To asses this hypothesis visually, the count of listings in Manhattan zipcodes was calculated and visualized as a choropleth map.
Overall, Lower East Side, East Village, and Hills Kitchen can be identified as neighborhoods with the highest short term rental activity. On the other hand, the Financial District and far uptown (i.e. Fort George and Inwood) are the least active neighborhoods. The east side of Manhattan starting from 34th street and above seems to be less active compared to the west side of the island despite the fact that it is still relatively close to Times Square and might seem attractive to tourists.
2.2 Centrographic Statistics
Although many listings found to be concentrated in East Village and Hills Kitchen, this pattern might be influenced by the room type. The website offers four types of rentals: Entire apartment, Private room, Shared room, and Hotel room. The spatial distributions of each type are shown in Fig.2.
Figure 1. The concentration of Airbnb listings in Manhattan.
Figure 2. Spatial distribution of listings by room type.
Figure 3. The cerographic statistics of (a) Entire home/apt, (b) Private room, and (c) All listings in Manhattan compared to (a) and (b).
The majority of listings are either an Entire home/apt or Private room. For this reason, the spatial means and the standard deviational ellipses are used next to compare these two types of rentals in Manhattan.
As expected, the standard deviation ellipses are stretched vertically and tilted towards the west. This makes sense because there is a cluster of active rentals in the range (480-720) around Morningside Heights with slight variation from one type of rental to another. It should not be a surprise that the overall canter of mass falls closer to the Entire home/Apt center of mass since it accounts for 59.8% of the data while the Private room is only 37.6% of total rentals.
3. Change over time
Were there more rentals last year than this year?
The most challenging part of this exploratory analysis is to capture the change of activity over time. This information is not available explicitly in the raw data. In order build this data, the year of the first review, as well as the year of last review, were extracted for each rental regardless of their type. New variables representing each year are added to the data frame to capture the activity of listing for the entire time frame (2009 – 2019). Take a look at the code here.
This would be useful in building a prediction model of business growth. A good approach would be to identify neighborhoods with demographic characteristics similar to the ones with already existed active rentals. This information is also useful in exploring any changes in activity since the New York City Council voted to restrict online short-term rentals in 2018. Although it seems like there is an increase in rental 2019, the spatial distribution might have change over the years and the number of minimum nights should be investigated further before driving any conclusions.
4. Hypothesis B:
The lower homeownership percentage in one neighborhood the higher the number of listings
Is there a relationship between the number of Airbnb listings in one neighborhood and the rate of homeownership? How about nonfamily households? We would expect neighborhoods with family and high homeownership rates to be less active as they would be reluctant to share their homes or expose their families to strangers. For this part of the analysis the data was prepared such that
Y = Number of listings per neighborhood
X = [ Percent of homeownership, Nonfamily households]
Interested? contact me for the full report with more details on Hypothesis A and the results of Hypothesis B.
Figure 4. The change of Airbnb activity in Manhattan over time.