The effect of pandemic relief efforts on nyc shelter occupancy

Final Projects


In New York City with an estimated population of 8.3 million people, roughly one of every 106 people is unhoused.  Every night thousands of people sleep on the street, in public spaces or shelters throughout the City.  One of the primary causes of homelessness is a lack of affordable housing, particularly among families. Other factors that contribute to homelessness include mental illness, substance abuse, untreated medical issues, traumatic events and difficulties sustaining employment.  Job loss, eviction, overcrowding, domestic violence or hazardous living conditions can result in families or individuals losing their homes and seeking residences in shelters.

Homelessness in NYC has been on the rise over the past decade, yet a recent report from the Coalition for the Homeless indicates that, while the numbers of single, unhoused adults has continued to increase, there has been a sudden, significant decline in the numbers of families who are unhoused.  They have attributed this downward trend to the effects of Covid-19 pandemic relief efforts such as the moratorium on evictions and extended, augmented unemployment benefits.  Another factor in the decline of shelter rates was the relocation of many unhoused people from shelters to hotels. These hotels were taken over by the City in an attempt to stem the spread of the highly contagious virus within overcrowded shelters. Ultimately, while the numbers have gone down temporarily, this decline is probably unsustainable and unlikely to continue once the short-term relief efforts are exhausted or halted.  

Exploring the Data

It can be challenging to track exactly how many people are homeless within the City at any given time, but shelter headcounts can provide some insight.  Through analyzing the shelter data over time it is possible to see how the occupancy rates change, and if any trends exist that demonstrate the reported declines.  For the first part of this project, I wanted to chart the occupancy numbers from the past few years and create visualizations to see if they reflected the impact of Covid-19 relief efforts on homeless families. 

To begin my exploration I used data collected by the NYC Department of Homeless Services (DHS) that I found online using the NYC Open Data website.  This dataset includes information regarding the daily number of families and individuals residing in the (DHS) shelter system and the daily number of families applying to the DHS shelter system from 2013 to present (2021).  While this dataset does not account for any unhoused people who are not living in a DHS shelter, the fluctuation of its numbers can be indicative of the rates of homelessness in general.  The data set is broken down into different demographic groups including single individuals and those in families.  The single adults are further divided into male and female, while the families are broken down into adults and children.

Sample of the DHS dataset pertaining to shelter occupancy rates from the NYC Open Data website

To assist in my exploration and visualization of the data I utilized Tableau Public.  Tableau Public is a free software that allows you to import data sets and create charts, graphs that can be interactive and easily shared.  Fairly intuitive to use, with no knowledge of coding required, Tableau is great for beginners and experts alike.  

To start, I uploaded a CSV file of the DHS data set to Tableau.  First I looked at the total number of individuals — both singles and the people in families.  I used Tableau’s measure tool to get the average number of people for each year.  It was important to use the average number in order to fairly compare previous full years with 2021, which is currently only half way through at the time of this investigation.  The resulting line graph showed that while the decline in rates had already begun prior to the Pandemic in 2019, with a drop of about 500 people, the past two years of 2020 and 2021 saw a much steeper descent (~4-6K per year).  After charting the data, I edited the “Number of Individuals” axis to start at a higher number (45,000) so that there was less white space on the graph.  I also felt it helped to better visualize the decrease in numbers, although I was a little concerned that the steeper line might dramatize the decline and skew the impact of the results for viewers.

Line graph charting total rates of shelter occupancy from 2013-2021

After looking at the total numbers, I wanted to compare the rates between single adults versus those in families.  To accomplish this I created bar charts showcasing the yearly averages, but this time, instead of showing the total numbers, I used Tableau’s quick table calculation to chart the difference in totals from the year prior.  Rather than having users of the charts do the mental math of subtracting one year’s totals from the year prior, I found that charting the differences was an easier way of showing how drastically the rates were changing.  Using this process I was able to create several bar charts to investigate the data and observe which groups demonstrated the decline in shelter occupancy that has been attributed to the Covid-19 Pandemic relief efforts.  

First I created a chart for total singles with the resulting differences all being positive — indicating that overall numbers were on the rise. This seemed to be in line with the Coalition’s reporting that single adults have not seen diminished homelessness during Pandemic. However, the past year (2021) did see significantly less of an increase than prior years (372 versus ~1000).  This could be a result of the Pandemic relief efforts, but it is not a conclusive conclusion.  I next looked at the difference in average totals for individuals in families.  The resulting bar chart showed a decline starting in 2017, but with much more significant decreases occurring in 2020 and 2021 (~2-3x less individuals in families in shelters).  Comparing these rates to the prior yearly differences, it is clear that shelter occupancy among families has seen a drastic decline during the Pandemic.     

Bar charts demonstrating that while average occupancy rates for singles has been on the rise, families have seen a steep decline attributed to pandemic relief efforts

Looking at the bar charts for children and adults who are part of families, there was also a significant decline in numbers as compared to earlier differences, beginning in 2020.  These results, with their very steep declines over the past two years of the Pandemic, more strongly corroborate the supposition that the temporary relief efforts have had a great effect on decreasing the rates of shelter occupancy and homelessness among people in families.  

Both adults and children in family have seen a steep decline in average occupancy rate during the past two years of the Covid-19 Pandemic

Formatting the Charts

When formatting all of my visualizations I sought external advice to determine the final layout and stylistic choices.  Initially I had different colors for each bar graph, but after review by some peers, it was determined that using one color provided consistency throughout.  I chose a neutral yellow color for the initial line graph of the total numbers, as well as for the bars of the subsequent charts.  A dark blue border was added to the bars in order to make them stand out a bit more.  I also added labels to each graph that marked the numbers of individuals per year.  I found this to be extremely helpful for comparing exactly how steeply the rates were changing from one year to the next.  Additionally the label’s inclusion of a plus or minus sign served to reinforce if the rates were going up or down.  I chose to use the same dark blue for those labels as the borders of the bar charts in order to provide more cohesiveness throughout my visualizations.  After consulting with a peer who works in design and production, I added a grey background to the numbers axis to highlight it. From that consultation, I also made some edits to the font sizes, making the titles bigger and bolder so that they were easier to read by users.

After further discussion during my class review, I opted to combine some of the bar charts, rather than have them stand alone, in order to better visualize the comparison in occupancy rates that I was trying to highlight. Within Tableau I was able to add these visualizations onto a story board with titles and captions so that users of the graphs would have some background on what the charts were trying to convey. Click this link to view the full story.

The Effect of Covid-19 Relief Efforts on NYC Shelter Occupancy
Screenshot of the story board created in Tableau with titles and captions

Predictions for the Future

Unless outside measures are taken, in the next few years I would guess that the steep rates of decline among families that we’ve seen in the past two years will drop off, especially after the moratorium on evictions and foreclosures ends in August 2021.  To continue this research, I would look to compare the numbers over the next several years as things return to “normal”, pre-pandemic life. It will be interesting to see if the general decline that started around 2019 would continue or if, with the halting of the pandemic related benefits, more people would find themselves unable to afford housing and forced to seek residence in shelters.  Unfortunately that may be the case, with renters who are facing eviction in the city owing on average $8150 in unpaid rent.  Thinking about this potential influx of people into the shelter system, for the second part of this project I sought to create an interactive map of resources available for people in need of housing assistance or who are looking to help someone in need. This map could be used directly by unhoused people or people facing the loss of their home, as well as by those who work in social services or know someone who is struggling with their housing situation.

Building the Map

To create my map I utilized Carto, an online software that allows users to create geospatial visualizations.  Carto offers a free year trial that I was able to take advantage of for creating my map.  I again looked to NYC Open Data’s repository of datasets to search for resources that would be useful for assisting unhoused people or people in danger of losing their homes.  Using the free version of Carto you are limited to only four layers, or datasets, for each map that you create.  The first layer I added was using the dataset for Homebase locations where services are available for unhoused people.  I then followed that with adding locations for homeless drop in centers, Medicaid offices, and job centers.  Once I had added all of my layers, I then turned to formatting the visualization.  

I wanted to make sure each type of resource center was easily visible, so I chose a different icon for each distinct category.  Carto offers several standard icons such as a circle or square, as well as various images to choose from.  I tried to select icons that represented each location. I opted for a house for the Homebase locations, a bed for the drop in centers, a medical cross for Medicaid, and a briefcase for the job centers.  Next I gave each one a distinct, bright color to differentiate them from each other.  Making each resource’s color unique was important since some of them overlapped on the map, and I wanted to make sure users could see there was more than one resource available at those locations.  I changed the point size as well to make them larger and thus more visible on the map.  

Map of Resources for Unhoused People in NYC

To assist users with understanding what each icon was referencing I added a legend that showed each icon and the facility it represented.  The final step was to add pop-ups to each layer so that users could click on a given facility and find its name, address and any contact information that’s available.  While I was limited by the four layer restraint using Carto’s free version, the goal would be to identify other facilities that offer services to unhoused people, and add those locations to the map as well. I asked a peer who does not have a background in design to review the map and engage with it in oder to determine its usability. She was able to move around the map and successfully identify what the various resources were, as well as the corresponding contact information after clicking on the icons.


The rise of homelessness in NYC is nothing new.  However, with the temporary relief efforts related to the Covid-19 Pandemic, there has been a decline in families requiring access to shelters as evidenced in the visualizations above.  With the looming expiration of both the extended unemployment benefits and the moratorium on evictions, as well as the on-going removal of people from the Pandemic hotels while rents continue to rise, the future looks bleak for many people who are already behind in rent due to loss of income and work opportunities.  The steep declines in family shelter occupancy that we’ve been seeing over the past two years will likely cease, and in fact will probably skyrocket after the benefits and temporary relief efforts halt.  Hopefully there will be some outside intervention, perhaps in the form of additional governmental relief to curtail the potential surge in shelter occupancy.  Regardless if any such measures are taken, resources such as the map that I created will be valuable for people requiring housing assistance and for those who are looking to assist them.