ALBANY, N.Y. (Feb. 6, 2025) — Cheng Ren is an assistant professor at the University at Albany’s . His research centers on housing justice, particularly in the areas of tenant rights and eviction. Using open-access data and AI techniques, Ren is working to understand these issues in the United States, in order to improve access to stable housing and related services among vulnerable populations.
In January of this year, Ren received the 2025 Outstanding Social Work Doctoral Dissertation Award from the . In this work, Ren used advanced data science to analyze eviction data and get a clearer picture of the individual, community and macro-level factors that influence eviction decisions, drawing connections between those factors and eviction determinations.
Here, Ren sheds light on his research and the importance of studying evictions to advance housing justice.
What is the scale of the eviction problem in the U.S. and who does it impact most?
More than 2.3 million eviction cases are filed annually nationwide. However, it is difficult to know the true scale of the problem because little has been done to quantify how many of these eviction filings actually result in members of a household being evicted. We also, historically, have not had good data on surrounding factors that shape these outcomes.
We see especially high eviction rates in cities with rising rents and limited affordable housing. Marginalized populations are disproportionately affected, including low-income families and renters, particularly in Black and Hispanic communities, as well as the elderly and people with disabilities.
Evictions drive cycles of housing instability, which can worsen financial insecurity, degrade mental and physical health, and limit employment and educational opportunities. Understanding the scale of the eviction problem, and identifying factors that could influence whether or not someone keeps their home, are essential to advancing housing justice.
What are you exploring in your research currently?
The United States lacks a comprehensive dataset to accurately assess the severity of the eviction crisis. Typically, data is provided by cities, counties or states. Some states have established such infrastructure, but many have not.
To address the fundamental question of how many households are evicted, in the formal legal way, we can examine court case records. At this stage, we have used AI-driven methods to (1) extract information, such as the property’s location, from scanned court file documents, and (2) “read” and understand the narratives presented across multiple case files. The U.S. Department of Housing and Urban Development has invited my team to present to policy- and research-related staff who may be able to contribute historical eviction data and help complete a national dataset.
How can this information help individuals and advance housing justice?
To address the eviction problem, policymakers need guidance on where interventions—things like emergency housing assistance, tenants’ rights education, mental health services and employment programs—are needed, and how they could be applied most effectively. This requires an understanding of the details surrounding eviction cases—information that must be pulled from multiple sources, including lengthy case files, and organized into a comprehensive format.
To help develop this necessary data science infrastructure, my research team built digital eviction records datasets for states that did not have them. These efforts include building the most comprehensive eviction records database in California and contributing to a more efficient and accurate process for building eviction datasets in Washington state.
What are some examples of ways datasets like these can be put to use?
Using the California data, we looked at municipalities with or without a “crime-free” property policy and found that implementing such a policy on the county level does not appear to reduce the crime rate, but instead, increases the eviction rate. Some municipalities in California have begun to reconsider or even abandon this policy as a result of this finding.
In Washington, we obtained information from court files which included the case processing time after the eviction notice was filed, and whether tenants who received a summons responded to the court. Using machine learning models, we were able to identify where to allocate housing preservation interventions, such as tenant rights education, which could help ensure that renters understand their rights and options and can make an informed decision on how best to proceed.
In your recently awarded dissertation, you analyzed eviction case files to better understand factors that shape eviction outcomes. What did you find?
In this work, we identified trends in factors surrounding eviction cases that led to either eviction or case dismissal. Even a basic descriptive analysis revealed intriguing insights. For example, we found that simply replying to a court summons will lengthen the case processing time. This prevents a default judgment and provides the renter with more time to prepare their case.
At the individual-level, factors such as race, property sale records, legal representation, taxable property value and response to summons were all found to influence eviction filing outcomes. At the community level, poverty rates and the proportion of rent-burdened households emerged as strong predictors. The interaction between an individual’s race and the proportion of white people in the census tract reveals that people of color experience higher eviction filing outcomes compared to white individuals in the same community.
We also found that if tenants have access to emergency funding to pay for the late rent, the case is likely to be dismissed. However, the welfare department usually takes much longer than the court to process eviction cases. This means that even qualified individuals who received emergency eviction funding may already have been evicted. Infusing AI into the data collection process allowed us to detect this trend.
What are some of the biggest opportunities for data science and AI to benefit the welfare system?
One of the major opportunities presented by AI lies in assisting scholars in exploring data that is challenging to quantify— things like extensive textual data, including case files, and image data like property images, maps and videos. New innovative approaches can also empower general users to analyze these data more efficiently.
Of course, it is crucial to acknowledge concerns related to using AI in this context, particularly regarding privacy and safety. The social welfare system serves real people, and it would be risky to solely rely on AI in decision-making processes within the welfare system, at least at this stage. However, these new possibilities have strong potential to support evidence-based interventions and prevention strategies that could have wide-ranging benefits. Honing safe, reliable data processing approaches is part of what we are working on now.
RSS