Evictions In Philadelphia

Analysis and prediction of the Philadelphia eviction

Zile Wu

December 13, 2022

Part 1 Intro

- Evicted: Poverty and Profit in the American City

"A single mother, after paying rent for a dilapidated apartment, has only $20 a month left to support herself and her two sons. A handicapped man who has lost both legs is tasked with taking care of the boys in the community while working for his landlord to pay off his debts. A good-hearted male nurse whose drug addiction has cost him his job and a place to stay ......"

Such stories may sound far removed from our lives. Yet they are real-life examples from Matthew Desmond's documentary novel, Evicted: Poverty and Profit in the American City. The book won the 2017 Pulitzer Prize in the nonfiction category. The Pulitzer Prize committee selected the book for its "thoroughly researched exposé of the aftermath of the 2008 U.S. economic collapse, in which mass evictions were the cause of poverty rather than the result of it.

What is the situation of evictions in the city of Philadelphia after the 2008 financial crisis, and can I foresee these potential populations in advance by means of spatial indicators of prediction? Can I anticipate these potential populations through spatial indicators and give them the necessary help before they are evicted? In this analysis, I will give the answers.

Part 2 Evictions

- Chart showing eviction in Philadelphia in 2010

Using 2010 data from evictionlab, we can clearly put Philadelphia on the map

Eviction Map （Move to see the number of Evictions）

The graph shows the two main types of data on eviction, the EvictionFilingRate and the number of Evictions per census tract within Philadelphia, An eviction filing is the result of a landlord filing a case in court to have a tenant removed from a property. Eviction represents the actual eviction that occurred.

As we can see from the graph, the number of EvictionFilingRate is significantly higher in West Philadelphia and North Philadelphia, and correspondingly, the number of Evictions actually occurring is also higher. In contrast, Northwest Philadelphia, South Philadelphia have lower EvictionFilingRate and lower actual Eviction.

Part 3 Analysis

- Relationship between between the eviction and other data

In this section we will look for some data related to Eviction, are they spatially consistent with Eviction? This step will prepare us for the next section to predict the potential Eviction.

01 / 05

Data Base

First I used the centroids of all census tracts as the spatial starting point for calculating the various types of data. Using IDW, I set the radius to 5000feet and the output cell size to 100feet to calculate the following four data.

PovRate

The first parameter is the poverty rate. From the distribution of IDW results, the poverty rate shows a similar trend to Eviction: higher in North and West Philadelphia and lower in Northwest and South Philadelphia.

RentBurden

RentBurden represents those households where rent accounts for more than 30% of household income. In terms of overall trends, RentBurden is only more concentrated in North Philadelphia compared to the more concentrated occurrence of poverty rates. However, the overall trend is still consistent with Eviction.

GrossRent

This data represents the amount of rent paid by households, and we know from Matthew Desmond's book that households with lower rents tend to live with poorer infrastructure, while the households themselves are less resilient to risk. In terms of spatial distribution, the distribution of rents also aligns with Eviction.

PctAfAmer

This graph reflects the distribution of African Americans in Philadelphia, and in Matthew Desmond's book, black households are at much greater risk of Eviction than white households. Therefore, we chose to use it as our Predictor despite the fact that it does not correspond exactly to Eviction in terms of spatial distribution.

Part 4 Prediction

- ISO-based forecasting

With the previous selection of data, the next step is to predict Eviction using Iso Cluster Unsupervised Classification. the Predictors are IDW_PovRate, IDW_RentBurden, IDW_GrossRent and IDW_PctAfAmer. a comparison of the prediction results with the actual situation is shown in the following figure.

The results are divided into five categories, with 5 representing the highest risk of eviction and 1 representing the lowest.

From the results, we can see that the model makes better predictions for eviction occurring in both North and West Philadelphia. The lower risk of Eviction in Northwest and South Philadelphia is also visually evident in the prediction results. We can use the prediction results to implement more relevant, e.g. Housing choice voucher, policies in the corresponding areas to help people avoid Eviction and transfer into life's dilemma.

Part 5 Summary

In this storymap, we start from the book and introduce Eviction, a major threat among low- and middle-income people. And using the tool of arcgis, we show the situation of Eviction in Philadelphia. Finally, using multiple predictors, we predict the high-risk areas where Eviction is likely to occur.

In fact, due to data and algorithm limitations, our model is still inaccurate in some areas. However, in general, our model can still be used as an aid to policy implementation, helping decision makers to better identify potential risk groups and provide timely support.