Using Data Science To Solve US Homelessness
Rebellion's goal is to shed some insight on homelessness and suggest some ways to reduce the homelessness crisis from a data science perspective by analyzing the distributions of homeless people and shelters in the United States.
The data set used in this model consists of two parts: the 2007 to 2018 Housing Inventory Count by Continuums of Care (CoC) and the 2007 to 2018 Point in Estimate by CoC. The data came from Hud Exchange. (https://www.hudexchange.info/resource/5783/2018-ahar-part-1-pit-estimates-of-homelessness-in-the-us/)
Through data analysis and visualization, several patterns on homeless distributions and shelter distributions are revealed.
What's the trend for the total number of homeless population and its sub-division in the US?
First, let’s take a look at the total number of homeless population changes over time.
From the chart above, the Overall Homeless sum has a decreasing trend from over 600,000 to about 550,000. This trend also applies to most sub-divisions of the homeless population, such as the total unsheltered population, total homeless people in families, and total chronically homeless individuals.
As shown in the graph, individuals make up the largest proportion of the overall homeless sum, dropping below 400,000 over the past ten years.
The overall homeless problem seems to be improving based on what the numbers tell. Homelessness might be decreasing as a result of the improvements in homeless assistance and increasing funding by the federal government.
What are the trends for the US homeless population in different geographical regions?
We first compute the total number of homeless populations in different regions in 2018.
According to the charts below, most homeless people are distributed in coastal regions, while there are comparatively less homeless people in the Midwest region. Next, as shown in the second bar chart, the number of homeless people in the North and South are similar.
Furthermore, we compute how the yearly number of the homeless population has changed over the years. According to the table below, we observe that the yearly percentage decrease of the homeless population in the Midwest region is much larger than that of in the coastal region.
Especially more recently, the Midwest is improving at a rate several times above the coastal region.
However, we find something more intricate when we zoom in the data. In the Midwest, from 2007 to 2018, the homeless percentage decreased by 8.3%, but from 2013 to 2018, the homeless percentage decreased by 21.8%.
This discrepancy implies a surge in the homeless population from 2007 to 2013 in the Midwest. Thus, it is only from 2014 to 2018 that the homeless problem in the Midwest actually improved.
What are the trends for US homeless shelters?
There are three types of beds considered in the data set: emergency shelter (ES), transitional housing (TH), and supportive housing (SH).
ES is usually where people turn for support after they experience sudden economic shock and staying in this type of housing is usually temporary.
TH is short-term (up to 24 months) housing for people to transit from homeless to permanent housing.
SH is more stable than ES and TH, offering residents both housing and services that help them treat diseases, alcoholic abuse, and support them with training life skills.
Shelters have a significant correlation with the homeless problem. I observed through data analysis that the homeless population far exceeded the bed availability before 2013, so there still existed a great deal of unsheltered homeless people.
Recall that the overall homeless population is around 600,000, whereas from the graph below the overall available bed amounted to less than 400,000 before 2013.
Only afterwards, the surge in emergency shelters has dragged the total number of beds upward to fill the gap.
It is also worth noting that transitional housing has seen an overall decrease except between 2012 and 2013, and moreover, supportive housing is very limited––minimum compared to the other two types of housing.
Thus, most of the funding has been invested to build emergency shelters since 2013. Despite the different trends of different beds, we discovered a more important pattern: The surge of beds from 2013 has occurred at the same time with the decrease of homeless population especially in the Midwest, which implies shelter turned out to be a significant factor to help alleviate the homeless problem.
If we could gather more regional data of the bed availability, we might be able to quantify the correlation between bed surge and decrease in homeless
What are some recommendations to be taken from the data?
We think it’s important to look into the policies during 2013 which have succeeded in bringing down the homeless population in the Midwest.
We might be able to draw some information on what was successful from it to tackle the homeless issue in other regions.
The surge in emergency shelter and transitional housing were the main stimulus, so it might be worth it to continue investing in both of them and examine the effects on the homeless.
In addition, supportive housing seems to be very limited, so if construction is affordable, we should explore whether investing in supportive housing also helps to fight the homeless problem by first experimenting at the local level.
Written by Fang Du
Edited by Alexander Fleiss, Gihyen Eom, Michael Ding Rohan Mehta & Jared Nussbaum