Which Factors Explain the Recent Surge in Homelessness?

Leifheit et al. (2026) examined which state-level factors were associated with rising homelessness across U.S. states between 2019 and 2024. They analyzed annual state-level data on homelessness counts, rents, unemployment, eviction moratoria, emergency rental assistance, overdose deaths, immigration, and climate-related property damage. Their primary finding was that eviction moratoria and climate-related disasters were the factors most consistently associated with year-over-year changes in homelessness. Each additional 1% of person-time covered by an eviction moratorium was associated with a 0.36 percentage-point smaller increase in homelessness, while each additional home-equivalent lost to climate disasters per 10,000 residents was associated with a 1.0 percentage-point larger increase in homelessness.    

Why This Article Was Selected for The Policy Scientist

Homelessness is one of the most consequential public policy challenges because it reflects the interaction of housing markets, economic conditions, public health, disaster resilience, and government institutions. The recent acceleration in homelessness across many jurisdictions has elevated the issue from a longstanding social concern to a central policy question with implications for health systems, local governments, and economic stability. The topic is especially timely given the sharp post-pandemic increases documented across the United States. Leifheit and colleagues have published extensively on housing insecurity, eviction policy, and homelessness, making this study a logical extension of their broader research agenda. The article makes a valuable contribution by focusing on year-to-year changes in homelessness rather than the more commonly studied differences in homelessness prevalence across jurisdictions. In doing so, it builds on influential work emphasizing housing-market conditions as a driver of homelessness while examining the factors associated with recent increases. The dataset is comprehensive, covering all 50 states and Washington, D.C., over multiple years, although the well-known limitations of HUD point-in-time counts remain. The findings may be relevant to other developed countries confronting similar housing and climate-related pressures, though institutional differences could affect external validity. Methodologically, the study employs a two-way fixed-effects framework, a commonly used causal inference approach that controls for both state-specific and year-specific factors. This design is substantially stronger than conventional cross-sectional regression and provides a more credible basis for causal interpretation.

Full Citation and Link to Article

Leifheit, K. M., Robinson, L., Nkansah, M., Szilagyi, P. G., & Pollack, C. E. (2026). Factors associated with rising homelessness within US states, 2019 to 2024. JAMA Network Open, 9(4), e265187. https://doi.org/10.1001/jamanetworkopen.2026.5187

Central Research Question

The central research question is straightforward but highly consequential: What factors were associated with the sharp rise in homelessness observed across U.S. states between 2019 and 2024? Previous research has largely focused on explaining why some places consistently experience higher rates of homelessness than others. This study instead examines why homelessness increased rapidly in some states over a relatively short period while remaining stable or declining in others. The authors evaluate several competing explanations that have received substantial public and policy attention, including rising rents, unemployment, pandemic-era housing assistance, eviction moratoria, substance use, immigration, and climate-related disasters. By focusing on year-to-year changes rather than levels of homelessness, the study seeks to identify the factors most closely associated with recent increases in homelessness.

Previous Literature

The study builds upon an extensive body of research demonstrating that housing market conditions are strongly associated with homelessness prevalence. Prior work, particularly research summarized by Colburn and Aldern and by HUD-sponsored analyses, has found that high housing costs, constrained housing supply, and local economic conditions are among the strongest predictors of homelessness rates across jurisdictions. These studies generally focus on explaining differences in homelessness levels between states, metropolitan areas, or Continuums of Care.

The authors argue that far less attention has been devoted to understanding changes in homelessness over time. This distinction is important because the factors associated with long-run homelessness prevalence may differ from those associated with short-term increases or decreases. Structural conditions such as housing costs may explain why one jurisdiction consistently experiences higher homelessness rates than another, while acute shocks or policy changes may explain sudden increases within a jurisdiction. The article therefore extends the literature by examining the determinants of year-over-year change rather than cross-sectional variation.

The study also engages with emerging research on eviction policy, pandemic-era housing supports, substance use, migration, and climate change. The authors themselves have contributed substantially to the literature on eviction policy and housing insecurity, including studies examining the effects of eviction moratoria on health outcomes during the COVID-19 pandemic. This article broadens that line of inquiry by investigating whether such policies were associated with changes in homelessness itself.

Data

The analysis uses state-level panel data covering all 50 states and Washington, D.C. The primary outcome measure comes from the U.S. Department of Housing and Urban Development’s annual point-in-time (PIT) homelessness counts. These counts estimate the number of individuals experiencing sheltered and unsheltered homelessness on a single night in January each year. Because pandemic restrictions substantially disrupted data collection in 2021, that year was excluded from the analysis.

The final dataset contains 255 state-year observations spanning 2019 through 2024, excluding 2021. The authors examine overall homelessness as well as several important subgroups, including sheltered homelessness, unsheltered homelessness, adult homelessness, and child homelessness.

Multiple publicly available sources were used to construct explanatory variables. Housing market conditions were measured using population-weighted state averages of HUD fair-market rents. Labor market conditions were measured using state unemployment rates from the Bureau of Labor Statistics. Pandemic-era housing supports included both emergency rental assistance distributions and eviction moratoria. Substance use was proxied using state overdose mortality rates. Immigration was measured using Census Bureau migration data identifying individuals who had lived abroad one year earlier. Climate-related shocks were measured using estimates of property losses from natural disasters, converted into home-equivalent losses and normalized by population.

The dataset is unusually comprehensive because it combines homelessness, housing, labor market, policy, demographic, public health, and climate measures within a single longitudinal framework. However, the authors acknowledge limitations associated with PIT counts, which are generally believed to undercount homelessness. While this limitation affects measurement accuracy, the counts remain widely used for tracking trends over time.

Methods

The authors employ a two-way fixed-effects panel design. This approach incorporates both state fixed effects and year fixed effects. State fixed effects control for characteristics that remain relatively constant within states over time, while year fixed effects account for national trends affecting all states simultaneously. The models also use clustered standard errors at the state level.

The primary dependent variable is the percentage change in homelessness from one year to the next. The authors estimate both bivariate and multivariable regression models. The multivariable models include all explanatory variables simultaneously, allowing the analysis to assess the independent association of each factor while controlling for the others. The models also include a time-varying control for COVID-19 mortality to account for differences in pandemic severity across states.

Additional analyses examine specific homelessness subgroups, alternative outcome measures, and baseline homelessness prevalence. Several sensitivity analyses test the robustness of the findings, including alternative variable specifications, exclusion of extreme observations, restriction of the sample to post-pandemic years, and leave-one-state-out analyses.

Methodologically, the study is substantially stronger than a conventional cross-sectional regression because it leverages longitudinal variation within states and controls for both state-specific and national factors. The fixed-effects framework places the analysis within the broader family of causal inference approaches commonly used in policy research. However, because the study does not rely on a clearly exogenous natural experiment or other quasi-experimental identification strategy, causal interpretations should remain somewhat cautious.

Findings/Size Effects

The study identifies two factors that were consistently associated with year-over-year changes in homelessness: eviction moratoria and climate-related property damage.

The average state experienced a 7.0 percent annual increase in homelessness during the study period. Homelessness increased by an average of 16.0 percent annually among individuals experiencing unsheltered homelessness and by 8.4 percent among children experiencing homelessness.

Eviction moratoria were associated with smaller increases in homelessness. Specifically, each additional 1 percent of person-time covered by an eviction moratorium was associated with a 0.36 percentage-point reduction in the annual increase in homelessness. The magnitude of this relationship was substantial. The authors estimate that homelessness increased by approximately 11.0 percent in the average state during 2022. In the absence of eviction moratoria, that increase would have been approximately 19.9 percent.

Climate-related property damage was associated with larger increases in homelessness. Each additional home-equivalent lost to climate-related disasters per 10,000 residents was associated with a 1.0 percentage-point increase in homelessness growth. The authors estimate that absent climate-related property losses, the average increase in homelessness during 2022 would have been approximately 7.8 percent rather than 11.0 percent.

Several factors commonly cited in public discussions did not demonstrate consistent associations with year-over-year homelessness changes. Average rents, unemployment, emergency rental assistance, immigration, and overdose mortality generally failed to reach statistical significance in the primary multivariable models. Importantly, however, rents and unemployment remained strongly associated with overall homelessness prevalence, consistent with earlier literature. This distinction suggests that factors driving short-term changes may differ from those driving long-run levels of homelessness.

Conclusion

The study concludes that eviction moratoria and climate-related property damage were the factors most consistently associated with year-over-year changes in homelessness across U.S. states between 2019 and 2024. In contrast, variables frequently highlighted in public debate—including immigration, substance use, unemployment, and emergency rental assistance—were not consistently associated with recent increases in homelessness.

The article’s principal contribution lies in shifting attention from explaining homelessness prevalence to explaining homelessness growth. By distinguishing between these two outcomes, the authors demonstrate that factors associated with long-run homelessness levels may differ from those associated with short-term increases. Their findings suggest that policy interventions affecting housing stability and exposure to climate-related disruptions may play an important role in understanding recent homelessness trends.

More broadly, the study contributes to a growing literature examining homelessness as a dynamic phenomenon influenced by both structural conditions and short-term shocks. Its use of nationwide longitudinal data and a fixed-effects framework provides one of the most comprehensive analyses to date of the factors associated with recent changes in homelessness across the United States.