According to the 2015 report of NYC Bureau Policy, approximately a quarter of NYC households and 32% of unemployed people lack the broadband internet at their home (NYC Bureau Policy Report, 2015). As a practical solution to this issue, the Mayor of New York in 2014 launched an initiative by partnering with CityBridge. Five years later, CityBridge had launched around 1.774 of 7500 planned wifi-kiosks, more popularly known as LinkNYC. The main idea behind this project was to create an open, equal and connected city (Huber, 2016 in vice.com). Several sources also emphasize the goal of this project to bridge the ‘digital gap’ and cater to the need of those people really in need of free public wifi services. However, some findings denote that the wifi kiosks are mostly located in Manhattan where there are fewer people particularly in need for this kind of services (Correal, 2019 in Newyorktimes.com). In addition, less than 10% LinkNYC is planned to be installed in Bronx where there is the highest percentage of households without internet access (NYC Bureau Policy Report, 2015 and Huber, 2016 in vice.com). Hence, it is interesting to analyze just how central the ‘digital gap’ is to the mission and real implementation of this project. Finally, in this essay, we would like to identify how this project has been catering to particular people in need of this service. After defining our aim of this project, we break down our aim to several sub-questions which can help us sequence analyzing the process. The main questions for this project comprise of:
What is the main borough where most public wifi services are located?
Between the five boroughs in NYC, where is the pocket of poverties in NYC mostly concentrated?
Is there any statistical or spatial correlation between poverty/income-level and public wifi location?
To succinctly summarize this project, we would like to analyze the relationship between public wifi services and poverty rates in NYC. We mainly worked with two main datasets: NYC Wifi hotspot locations and 2010 census tract data, which derived from NYC Open data portals. A similar data with regard to census tract also available in Kaggle which provide a cleaner version, but we used the official source to ensure the quality of the data. Both datasets were downloaded in its CSV form and we did the data processing using Numerical Python. We distinguished the step of this project to data cleaning, analysis, and visualization. The detailed descriptions of each step will be explained in the next section.
Source 1.NYC Wifi Hotspot Location [https://data.cityofnewyork.us/Social-Services/NYC-Wi-Fi-Hotspot-Locations/a9we-mtpn?fbclid=IwAR1YLzOXETv-pxHqtK5d8zwe1JZGI_mow4h67ot9s0SBfe4abRiULXQlpwI]
The NYC Wifi Hotspot location contains information of all the publically accessible Wifi. Information includes the location of the Wifi kiosks, providers, types (free/limited free), and activated time.
Source 2. U.S. Census [https://factfinder.census.gov]
The census data include information regarding population, poverty level, race, income level, and census tract is used as the geographical unit, and geometry point.
The NYC Wifi Hotspot location has a column pertain to borough code and census tract code. This code is a sub-part of census tract code which also available in the Census Tract dataset. Therefore, our first step is to eliminate the unnecessary column in NYC Wifi dataset such as latitude, longitude, geometry point, type of kiosk, etc. Another thing to consider is that the NYC Wifi dataset contains multiple census-tract codes. For example, in “BOROCT2010” column there are four rows of ‘ 3118400’ values which translate that there are four public wifi-kiosk in the ‘3118400’ tract. We address these issue by making two separate dataset. First, we make a new dataset which contains a new column. the new column will represent the new census tract code that similar with the id in census tract dataset. We run some code in python to make the new column representing a set of number which composed of borough code and the last five number of each value in ‘‘BOROCT2010’. From the US Census website, we were informed that the first 5 number of the census tract is the code for each borough. for example, Manhattan code is 36061. Therefore, the new ID in NYC Hotspot dataset which has a value of Manhattan in ‘BORONAME’ column and 3118400 in ‘BOROCT2010’ column is 36061118400. Using this method, later we can join the NYC Wifi dataset and the census tract dataset. Second, we created the new dataset which contains a new column containing a total of public wifi for each census tract. We use group.by method in Pandas to make the new dataset. The census data already contains an essential element for our analysis. There is no significant issue such as missing value. After creating a new dictionary which uses the census tract ID (similar to the new ID in Public Wifi dataset), the dictionary then exported to a data frame to make the analysis process easier. The following step is to merge the new data frame of NYC Public wifi which contains the new ID and the census tract data frame, we rename the key column to ‘index’ to merge both data frames. To make it more comprehensive for our analysis, we merge the newly created data frames with the second data frame in Step 1 which contains the number of public wifi for each census tract. The final dataset from this data cleaning process is shown in picture 1, there are approximately 756 rows and 16 columns in this new dataset.
Table 1.The final dataset 
Graph 1.Public Wifi per Borough 
Map 1.Public Wifi Hotspot Location 
To make this graph, we use python and seaborn package to compute the total wifi per each borough and populate a new dictionary which contains the sum of public wifi spot for each borough. We also added spatial mapping using ArcGIS to complement the results.
Manhattan has the highest number of public wifi services, more than double the number of public wifi in Brooklyn (the second one). Ranking from the highest to the lowest one, there are Manhattan, Brooklyn, Queens, Bronx, and lastly Staten Island The possible cause= ad revenue for CityBridge so it is profitable to put LinkNYC in Manhattan where most tourist visit, and a high presence of business district
Graph 2.Percentage of Low-income Households per Borough 
As the graph shows the percentage of low-income people out of the total population for each respective neighbourhood, Brooklyn has the highest percentage of people living under the poverty rate. Following Brooklyn, from the second highest to the lowest one there are Manhattan, Bronx, Queens, and Staten Island at the last. The percentage of poor people in Manhattan, Queens, Brooklyn, and Bronx results in a relatively similar range (between 35% to 42%). However, it must be noted that the total population for each borough is different
Graph 3.Correlation Matrix 
Graph 4.Scatter Plot
Map 2. Wifi Location and Median Household Income
Link to Wifi locaiton and Income-level by Census Tract Map https://arcg.is/199muG0
Using heatmap and scatterplot function in Seaborn, we try to map the correlation between both variables, the poverty and total public wifi. We also consider the percentage of racial class in the first correlation heatmap to broaden our findings with regards to poverty in NYC. To simplify the data frame, we also drop some irrelevant column such as provider, location, and borough code.
According to the Correlation Matrix below, the correlation between poverty and public wifi location is relatively low, suggesting the possibilities of other factors have greater influences over the location choice of public wifi. Furthermore, according to the scatterplot there is no clear pattern of neither negative or positive correlation. Looking at the dataset, the correlation figure is around -0,07 and signifies a very weak correlation. The map demonstrates a clear spatial correlation between the income level and the location of public wifi: higher-income neighbourhoods enjoy more access to public wifi while free public wifi is commonly absent in low-income neighbourhoods. The trend is clearest in Manhattan borough and least obvious in Staten island as only few wifi hotspots exist and income gap between census tract is not as substantial.
The goal of this paper is to examine the digital gap in New York City- whether there is an equity issue regarding public wifi access. It is rational to place LinkNYC (public wifi kiosks) in higher-density neighbourhoods, and as a matter of fact, Manhattan does outnumber all other boroughs. However, our findings raise the issue of equity and question whether the program has managed to serve its target population- people who lack access to the internet. We find a clear spatial pattern that Link NYC are concentrated in higher-income/lower-poverty neighbourhoods and vice versa. We urge the City to address this digital gap and do more to provide public wifi access for low-income communities. Campaigns and other education programs can also be initiated to encourage more equitable uptake of free public wifi.
Is poverty a good indicator to represent people who lack access to broadband internet at home? As a general rule of thumb, it might show some correlation but further research can complement this assumption. For example, in this project we find less correlation between poverty rate and total wifi installed. The compliance of time attribute between two datasets. For a metropolis city such as New York, it is highly possible that the demographic composition can significantly change in a relatively short year. However, we also bumped into the fact that the comprehensive demographic data like census is only conducted once every 5/10 years. Other factors may also have a high correlation with access to Wifi. For example, population density and tourists volume. Being an open, equal, connected city can include various objectives, and due to limited resources, some areas may be prioritized due to other concerns.