📈 economics

Impact of Housing Choice Voucher on Crime: A Spatial Econometric Analysis

Using a spatial Durbin model with fixed effects on St. Louis County data, this paper finds that Housing Choice Vouchers significantly increase local crime rates through both direct effects and spatial spillovers, resulting in a total impact of approximately a 2.5% rise in crime for every additional voucher per 1,000 housing units.

Original authors: Muritala Ogunsiji

Published 2026-09-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Muritala Ogunsiji

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the United States, the cost of housing has climbed steadily, leaving millions of families struggling to find a place to live that does not consume a dangerous portion of their income. To help, the federal government runs a massive program called the Housing Choice Voucher, often known as Section 8. This program gives low-income families a subsidy that allows them to rent apartments in the private market, with the government paying a portion of the rent directly to the landlord. The goal is to help families move out of poverty and into safer, better neighborhoods. However, a persistent worry has shadowed this policy for decades: does moving a large number of subsidized families into a neighborhood change the safety of that area? Specifically, does the arrival of these vouchers lead to more crime? This question sits at the intersection of economics and public safety, where researchers try to untangle whether a housing policy is helping families or inadvertently harming the communities they enter. The challenge is that crime and housing are deeply connected to the physical layout of a city; what happens in one neighborhood often ripples into the next, making it difficult to measure the true effect of a policy without looking at the whole map.

A researcher at Northern Illinois University decided to investigate this question by looking closely at St. Louis County, a sprawling suburban area with over ninety different towns and a sharp divide between wealthy and poor communities. Instead of just counting crimes and vouchers in a single neighborhood, the study used a sophisticated method that treats the county like a connected web. This approach recognizes that a change in one census tract—a small, defined geographic area—does not happen in a vacuum. If a neighborhood next door sees a change in its housing mix, that change can influence the crime rate in the original neighborhood, just as a ripple in a pond spreads outward from a single stone. By using a model that accounts for these connections, the researcher could separate the direct effect of vouchers in a specific neighborhood from the indirect effect of vouchers in the neighborhoods surrounding it.

The study analyzed eight years of data, from 2015 to 2022, covering 123 census tracts in St. Louis County. The researcher focused on property crimes, such as theft and burglary, because these are the types of offenses most likely to be influenced by economic conditions and opportunity. The data showed that on average, there were about 24 property crimes for every 1,000 residents in these areas. When the researcher looked at the relationship between the number of housing vouchers and the crime rate, a clear pattern emerged. In a neighborhood where the number of vouchers increased by one for every 1,000 housing units, the local crime rate rose by approximately 1.7 to 1.8 percent. This was the direct effect, the immediate impact felt within the tract where the new voucher holder moved in.

However, the story did not end at the neighborhood border. The analysis revealed that the impact of vouchers spilled over into neighboring areas. When a nearby tract added one voucher per 1,000 units, it contributed an additional 0.6 to 0.7 percent increase in the crime rate of the original neighborhood. When the researcher combined the direct effect and this spillover effect, the total impact of adding a single voucher unit was an increase of about 2.5 percent in the crime rate. To put this in concrete terms, in an average neighborhood with 4,300 residents, this increase translates to roughly 2.6 additional property crimes per year. This finding suggests that the standard way of looking at these programs, which often ignores the connections between neighbors, underestimates the full scope of the impact.

The study also tested whether the wealth of a neighborhood changed how vouchers affected crime. Some theories suggested that in wealthier areas, the presence of more valuable targets might increase crime, while others argued that wealthier communities have better resources to prevent it. While the primary model used in the study did not find a statistically significant interaction, a secondary analysis using a different spatial model (the Spatial Autoregressive model) did find evidence supporting the "buffer effect" hypothesis. This suggests that in higher-income neighborhoods, the negative impact of vouchers on crime may be moderated by better neighborhood resources and social controls. Consequently, the increase in crime is likely unevenly distributed, driven more by spikes in lower-income tracts that lack these protective buffers, rather than being a uniform effect across all areas. The researcher also checked if the results held up when using different ways of measuring the data, such as looking at the number of vouchers relative to occupied homes versus total homes, or by checking if the effect happened a year later. In every case, the connection between vouchers and a rise in property crime remained steady and significant.

These findings challenge the notion that housing vouchers are neutral regarding neighborhood safety. The study suggests that when vouchers are concentrated in specific areas, they can create a feedback loop that increases crime not just in that specific location, but in the surrounding communities as well. The researcher argues that to address this, policymakers should focus on spreading out the vouchers more evenly, rather than allowing them to cluster in distressed areas. By encouraging families to move into higher-income neighborhoods where the social environment might better absorb the change, the negative effects could be reduced. The study concludes that while the vouchers are essential for helping families, the way they are distributed matters deeply for the safety of the entire region. The evidence points to a need for a strategy that considers the whole map, ensuring that the solution to housing insecurity does not inadvertently create new problems for the communities it aims to help.

Technical Summary: Impact of Housing Choice Voucher on Crime: A Spatial Econometric Analysis

Problem Statement
The Housing Choice Voucher (HCV) program is the primary federal housing subsidy in the United States, designed to assist low-income households in accessing affordable private market housing. However, a persistent controversy exists regarding whether the concentration of voucher holders in specific neighborhoods generates negative externalities, specifically an increase in local crime rates. While numerous studies have investigated the relationship between subsidized housing and crime, existing literature suffers from two critical limitations:

Mixed Empirical Results: Previous studies utilizing data from various jurisdictions (e.g., Chicago, Dallas, Charlotte) have yielded inconsistent findings, ranging from no significant effect to positive associations with property crime.
Neglect of Spatial Dependence: Most prior research relies on non-spatial models that assume independent observations. This assumption is frequently violated in spatial data where crime in one census tract is inherently dependent on and influences crime in contiguous tracts. Ignoring these spatial spillovers can lead to biased estimates and a failure to quantify the broader, indirect consequences of the HCV program.

Methodology
To address these gaps, this study employs a spatial econometric approach using panel data from St. Louis County, covering 123 census tracts from 2015 to 2022. The analysis focuses on property crime, utilizing data from the National Incident-Based Reporting System (NIBRS) and the FBI Uniform Crime Report, aggregated to the census tract level.

Data Construction: The study constructs a balanced panel of 984 observations. Census tracts from the 2010 and 2020 decennial censuses were harmonized using IPUMS National Historical GIS (NHGIS) crosswalk files with interpolation weights. Missing data for crime and HCV counts (up to three years) were addressed via linear interpolation.
Variables: The dependent variable is the natural logarithm of the property crime rate per 1,000 residents. The primary independent variable is the HCV rate (occupied HCV units per 1,000 housing units). Control variables include median household income, poverty rate, minority share, male population percentage, under-18 population percentage, and homeownership rate.
Estimation Strategy:
1. Baseline Models: The study first estimates standard Fixed Effects (FE) and dynamic Arellano-Bond Generalized Method of Moments (GMM) models to establish a baseline and compare results with prior literature.
2. Spatial Durbin Model with Fixed Effects (SDM-FE): The core analysis utilizes the SDM-FE, specified as:
  $\ln(Crime_{it}) = \rho(W \times \ln(Crime_{it})) + \beta_1 HCV_{it} + \beta_2 (W \times HCV)_{it} + X_{it}\phi + WX_{it}\vartheta + \mu_i + \varepsilon_{it}$
  This model accounts for spatial dependence in both the response variable (via the spatial lag of the dependent variable, $\rho$ ) and the explanatory variables (via the spatial lag of the independent variables, $W \times HCV$ ).
3. Spatial Weight Matrix: A $k$ -nearest neighbors (KNN) matrix with $k=4$ is used to define spatial relationships.
4. Decomposition: Following LeSage and Pace (2009), the coefficients are decomposed into direct effects (impact within the tract) and indirect/spillover effects (impact on neighboring tracts).

Key Results
The spatial econometric analysis yields statistically significant findings regarding the impact of HCV concentration on crime:

Direct Effects: A one-unit increase in the HCV rate (defined as one additional voucher per 1,000 housing units) within a census tract is associated with a 1.7% to 1.8% increase in the local property crime rate.
Indirect (Spillover) Effects: An additional voucher unit in neighboring tracts contributes a further 0.6% to 0.7% increase to the local crime rate.
Total Impact: The aggregate effect of an additional voucher unit is approximately a 2.5% increase in the crime rate. Evaluated at the sample mean of 24 crimes per 1,000 residents, this translates to approximately 0.6 additional property crimes per 1,000 people, or roughly 2.6 additional annual crime incidents in an average census tract of 4,300 residents.
Robustness: These findings remain robust across alternative HCV rate measures (e.g., per occupied housing units), functional forms (log-log models), and lagged specifications. The results also hold when using the Spatial Autoregressive (SAR) model, though the SDM is preferred based on Likelihood Ratio (LR) tests and Akaike Information Criterion (AIC) scores.
Income Interaction: The study tested whether neighborhood income moderates the HCV-crime relationship (the "buffer effect" vs. "high-target opportunity" hypotheses). While the interaction term was statistically insignificant in the preferred SDM specification, the SAR model suggested a potential buffer effect where higher income mitigates the adverse impact. The author notes that the SDM's inclusion of neighbor-specific controls may obscure this specific interaction.

Significance and Claims
The paper claims to make three primary contributions to the literature:

Quantification of Spillovers: It provides the first estimates of the spillover effects of housing vouchers on crime using the Spatial Durbin Model with fixed effects (SDM-FE). By demonstrating that the total impact (2.5%) is significantly larger than the direct impact alone, the study argues that non-spatial models systematically underestimate the full consequences of the HCV program.
Methodological Rigor: The study advances the field by employing the Spatial Durbin Model with fixed effects, which accounts for unobserved heterogeneity and spatial dependence in both dependent and independent variables. This approach addresses the bias inherent in models that ignore spatial feedback loops.
Suburban Context: Unlike previous studies focused on large urban centers with centralized governance, this research examines a highly fragmented suburban context (St. Louis County) characterized by over 90 municipalities and sharp socio-economic boundaries (the "Delmar Divide").

Policy Implications
Based on the findings, the author suggests that to mitigate the crime effects associated with subsidized housing, policymakers should focus on deconcentrating occupied HCV units. The study advocates for:

Limiting voucher concentration in distressed tracts.
Encouraging mobility to higher-income areas where a "buffer effect" may exist.
Adopting Small Area Fair Market Rents (SAFMRs) to broaden the range of neighborhoods accessible to voucher holders.
Pursuing deconcentration alongside place-based investments in distressed tracts to address underlying socioeconomic drivers.

Limitations
The author modestly acknowledges that the findings are specific to the socio-spatial context of St. Louis County and may not be fully generalizable to other jurisdictions. Furthermore, while the study controls for spatial dependence, it does not fully resolve potential endogeneity arising from reverse causality (i.e., crime rates influencing voucher allocation), suggesting that future research should employ quasi-experimental designs to more credibly identify causal relationships.

Technical Summary: Impact of Housing Choice Voucher on Crime: A Spatial Econometric Analysis

More like this