{"subscriber":false,"subscribedOffers":{}} Can Medicaid Expansion Prevent Housing Evictions? | Health Affairs

### Research Article

Culture Of Health

# Can Medicaid Expansion Prevent Housing Evictions?

Affiliations
1. Heidi L. Allen ([email protected]) is an associate professor in the School of Social Work, Columbia University, in New York City.
2. Erica Eliason is a doctoral candidate in the School of Social Work, Columbia University.
3. Naomi Zewde is a postdoctoral research scientist at the Center on Poverty and Social Policy, School of Social Work, Columbia University.
4. Tal Gross is an associate professor in the Department of Markets, Public Policy, and Law, Questrom School of Business, Boston University, in Massachusetts.
PUBLISHED:Open Accesshttps://doi.org/10.1377/hlthaff.2018.05071

## Abstract

Evictions are increasingly recognized as a serious concern facing low-income households. This study evaluated whether expansions of Medicaid can prevent evictions from occurring. We examined data from a privately licensed database of eviction records in fourteen states (286 counties) and used a difference-in-differences research design to compare rates of eviction before and after California’s early Medicaid expansion (51 counties). Early Medicaid expansion in California was associated with a reduction in the number of evictions, with 24.5 fewer evictions per month in each county from a pre-expansion average of 224.7. These results imply that for every thousand new Medicaid enrollees in California, Medicaid expansion was associated with roughly twenty-two fewer evictions per year. Additionally, we found a 2.9-percentage-point reduction in evictions per capita associated with early expansion. The effects were concentrated among counties with the highest pre-expansion rates of uninsurance. We conclude that health insurance coverage is associated with improved housing stability.

Medicaid expansion has been shown to reduce medical debt and poverty by alleviating the financial burden of medical care on low-income beneficiaries and their families. Another of the challenges facing this population is home eviction. The growing rate of home eviction in the United States is drawing the attention of researchers and policy makers. We investigated whether Medicaid coverage plays a role in mitigating the devastating phenomenon of home eviction.

Eviction can have long-term damaging consequences for the families affected. Prior research suggests that experiencing a housing eviction may lead to poorer physical and mental health outcomes and a considerable worsening of health-related behaviors.1 There are several potential mechanisms for this decline. First, with a legal record of eviction, renters may be forced into a higher-risk rental market in higher-crime, higher-poverty neighborhoods with substandard housing conditions.2 Second, eviction precipitates acute risk of homelessness, which imparts a well-documented toll on health.3 Third, for those receiving health care, forced displacement can interrupt treatment continuity and disrupt patient-provider relationships,4 thus increasing the likelihood of worsened health outcomes. The causes of eviction are understudied and poorly understood, and these factors are of growing importance as evictions become increasingly common.

In the most immediate sense, eviction is initiated by a court judgment in favor of a landlord, usually in a dispute over unpaid rent. The tenant then either voluntarily vacates the premises in two weeks or is forcibly removed with their belongings by local police. The rising incidence of urban eviction may result from narrowing margins between rising rents and stagnant incomes.5 For example, in Los Angeles County, California, which has seen a sharp increase in homelessness,6 working full time at minimum wage and renting a one-bedroom apartment leaves \$148 a month for all other expenses. Nationwide, most low-income families spend over half of their income on housing.2 Poor health can be a contributor to eviction when people with limited resources cannot afford both housing and medical care. Poor health can cause absenteeism or job loss, further constraining resources through income reduction, becoming uninsured, or both. As a result, poor health can increase one’s exposure to health care costs, which can contribute to financial decline and increase one’s risk for housing instability.

Medicaid may mitigate the risk of eviction directly by reducing the cost of medical care and indirectly by protecting earning potential through better health. The strength of the latter effect is moderated in this population by concurrently weak ties to the labor force and a high burden of work-limiting disease and disability.7 Within this context, medical debt or urgent medical needs, acute or chronic (for example, filling prescriptions), may compete with housing obligations when finances are tight. If so, providing otherwise uninsured low-income adults with affordable health care coverage may help prevent eviction.

In this study we used a quasi-experimental difference-in-differences design to examine whether the expansion of eligibility for Medicaid, which provided medical care at low or no cost for low-income adults, could prevent eviction. Our hypothesis that Medicaid expansion might reduce evictions was supported by prior evidence of the program’s effects on social and economic aspects of the lives of low-income patients. Evidence from a randomized trial of Medicaid expansion in Oregon has shown that the coverage nearly eliminated catastrophic medical debt (defined as out-of-pocket medical spending totaling more than 30 percent of income),8 decreased the probability of having debt in collections, lowered the amount owed in existing collections, and lowered financial strain in general (including borrowing money from friends and family members and skipping some bills to pay others).9 In California, Medicaid expansion was also associated with a reduction in the use of payday loans.10 Thus, it is possible that improved finances related to Medicaid could help families avoid being sent to court for an eviction or being evicted by allowing tenants to come to a financial agreement with their landlords.

We analyzed data from California because it is a large state that initiated early Medicaid expansions through the Affordable Care Act (ACA) county by county over a period of months in late 2011 and throughout 2012. The California Low Income Health Program (LIHP), also called the Bridge to Reform, was established through a Medicaid Section 1115 waiver and built upon a demonstration program in ten counties, the Health Care Coverage Initiative, in 2007–10. Most counties opted to participate in the LIHP expansions. Eligibility for expanded coverage was established by counties and varied considerably. For example, Sacramento County expanded eligibility to people with incomes that were 67 percent of the federal poverty level, while Alameda County used 200 percent of poverty. An estimated 680,000 people gained coverage through the LIHP (based on unduplicated cumulative enrollment in December 2012), and most of them were transferred to Medi-Cal or the state Marketplace (Covered California) in January 2014.

Our data set of fourteen states allowed us to compare 51 California counties that expanded Medicaid early to 235 counties in states that did not expand Medicaid during the same time period, as well as to counties in California that opted out of the LIHP. As noted above, we used a difference-in-differences research design. We were unable to evaluate the 2014 ACA Medicaid expansions using this data source and methodological approach, as most states in our database adopted the 2014 expansions. As a result, we included data only through 2013 in our analysis.

## Study Data And Methods

### Data

This study used a commercial evictions database that was originally designed to help landlords screen tenants, from American Information Research Services (AIRS). The database was licensed to us for the purpose of this research. It has been used and validated in previous eviction research by the Eviction Lab at Princeton University.11 AIRS retrieves public eviction records from counties in periodic batches, through automated retrieval, or both, depending on the contractual arrangement with the county.

For our main analyses we used evictions in 286 counties in fourteen states (including California) for each month in the period January 2008–December 2013—roughly including thirty-six months before and twenty-four months after the early California Medicaid expansions in 2011 and 2012. For counties in several states (Arizona, California, Nevada, New Jersey, Utah, Washington, and Wisconsin) we analyzed the universe of evictions in the time period above. For counties in several other states (Delaware, Florida, Hawaii, Massachusetts, Ohio, and Pennsylvania) some time periods were not covered in the data.

We conducted a number of analyses that excluded particular states from the analysis to ensure that the results did not depend on the inclusion or exclusion of particular states, but our findings were not sensitive to the choices. In our final analyses, two states (Massachusetts and Hawaii) that had achieved near-universal coverage before the start of the California Medicaid expansions were labeled “always treated” states. A description of our sample, showing the share of counties with Medicaid expansion over our study period, is in the online appendix (also see appendix exhibit A.1. and table A.2. for more details on the data set).12

### Analyses

We evaluated the effect of California’s early Medicaid expansion on evictions by running a difference-in-differences regression, comparing 51 early-expanding “treated” counties to 235 “control” counties without early expansion. The regression compared the average monthly change in evictions for expansion counties before and after they expanded Medicaid to the average monthly change in nonexpansion counties over the same time period observed in our study.

We aggregated information from the evictions database to the county-month level. We included county-specific and year-month fixed effects in all main models, in addition to adjusting for county-specific linear time trends. We controlled for yearly county unemployment using data from the Local Area Unemployment Statistics of the Bureau of Labor Statistics. We controlled for annual county-specific poverty, using American Community Survey data published by the Census Bureau and aggregated by the Princeton Eviction Lab.11

Our outcomes of interest were the number of evictions in each county-month, the number of evictions per capita, the natural logarithm of the number of evictions, and the number of evictions per rental unit. Per capita estimates adjust for total population, but evictions are at the household level, which vary in size. Estimates per rental unit are better in this regard, but we had only one measure of rental units per year, compared to precise monthly estimates of evictions.

To calculate evictions per capita, we used estimates of county population size for each county from the 2000 census. To calculate evictions per rental unit, we used the number of rental-occupied housing units in each county for each study year from the American Community Survey.

To examine whether our results were driven by populations most likely to obtain coverage under the Medicaid expansion, we stratified our estimation model of monthly numbers of evictions between counties, using a higher versus a lower share of uninsured residents before the expansion. We stratified counties by the share of residents with incomes below the threshold for Medicaid eligibility (138 percent of the federal poverty level) who were uninsured in 2010. We obtained estimates for the share uninsured by income from the Small Area Health Insurance Estimates published by the Census Bureau. We present results from the difference-in-differences regression of monthly number of evictions, estimated separately for counties with above- and below-median shares of uninsured residents.

### Supplemental Analyses

In our appendix12 we present the results of several exercises that we conducted to visually inspect our results and assess the soundness of our empirical approach: We show the results of our main specification without adjusting for county-specific linear trends. We present a “raw count” graph of the percentage changes in evictions from January 2008 levels, comparing California counties that expanded Medicaid in January 2012 to all other counties until 2014. We display “event studies” for all of our outcome measures to probe the parallel trends assumption of the difference-in-differences design. In each exercise, but particularly in the last, we were trying to evaluate the counterfactual assumption that in the absence of Medicaid expansion, trends in evictions would not have differed between treated and control counties during the study period.

Our visual inspection of the event studies suggested that treated California counties had higher eviction rates than control counties did before Medicaid expansion. Pre-expansion differences in trend could partially account for the magnitude of change observed in difference-in-differences models, which is why we adjusted for county-specific linear trends in our main specifications. Additionally, we explored the robustness of our main results using a synthetic control approach as an alternative to difference-in-differences regression.13,14 Synthetic control methodology is not sensitive to an assumption of parallel trends. For this approach, we created a synthetic control group of counties for each California county using an algorithm created by Alberto Abadie and colleagues, the economists who developed this specific methodology for comparative case studies.13 The algorithm assigned a weight of 0 to 1 to each synthetic control county so as to closely match the pre-expansion trend in evictions in the treated county across the pre-expansion period. We then calculated the average difference between the actual evictions each month in the treated county and the evictions in the synthetic control group after expansion. Details on the synthetic control approach are in the appendix.12

### Limitations

The primary limitation of our study was the aggregate nature of the data. We were unable to directly link individual-level data on insurance status to people’s rental histories or eviction records. The data we had on evictions were administrative in nature and included very few details about people’s demographic characteristics such as insurance status, income level, or age. To examine the link between insurance status and eviction, we performed stratified analyses by uninsurance rates, expecting that the main effect of the expansion would be most prominent in counties with a greater share of uninsured residents before expansion if the result was tied to the change in insurance status.

There were also issues related to using a commercial evictions database instead of undergoing primary data collection (which would have been prohibitively costly and time-consuming). One issue was that we could not verify that every eviction was included in our data set. However, the Eviction Lab at Princeton University examined undercoverage in the AIRS data by comparing internally aggregated counts of evictions in AIRS with publicly reported aggregated counts of evictions by states. Fewer than 5 percent of counties in AIRS states were suspected to have counts below the externally verified ones.11

An additional limitation involved the assumptions underlying the empirical approach. The difference-in-differences estimation strategy assumed that evictions in California and other states were evolving along parallel trends. That assumption might be violated, and in fact exhibit 1 suggests a slight divergence in trend before California expanded Medicaid. For that reason, we explored other empirical approaches, such as the synthetic control methodology, and adjusted for county-specific linear trends in our main specification.

We also could not completely rule out changes in the housing market that may have affected rates of eviction during our study period, although we had no reason to believe that such changes would have coincided specifically with Medicaid expansion. The nation faced a housing crisis during the Great Recession (2007–09), which led to increased regulation and oversight of the mortgage industry. New federal regulations were not specific to California, although California experienced a disproportionately high increase in evictions during the Great Recession. California did implement a law in 2017 that allowed some evictions to be sealed from disclosure, but this did not affect our study.

## Study Results

We found that the Medicaid expansion in California led to a significant reduction in the number of evictions occurring in each county. Exhibit 1 illustrates this relationship as an event study of the monthly number of evictions. Before the expansion, California expansion counties had slightly higher rates of evictions (particularly following the Great Recession) than comparison counties did. Following the expansion, we saw a change in sign, with California expansion counties having significantly fewer total numbers of evictions. Point estimates of the reduction in evictions following expansion grew as more time elapsed following expansion, reaching significance more than twelve months post-expansion. The event studies of the estimated effect of Medicaid on eviction rates per capita and per rental unit are presented as appendix exhibits D.3.–D.5.12

The difference-in-differences regression estimate, adjusted for county-specific linear trends, suggests that expansion was associated with 24.517 fewer evictions per month in each county, compared to a pre-expansion mean of 224.718 evictions per month (exhibit 2). Based on publicly available estimates of the number of Californians who gained coverage during Medicaid expansions, these estimates imply that for every thousand new enrollees, there were approximately twenty-two fewer evictions per year15 (see appendix B.2. for an explanation of calculating eviction reductions per new enrollee).12 Relative to the number of residents in each county, we found that the expansion reduced evictions per capita by 0.029 from a pre-expansion mean of 0.261 (exhibit 2). When we examined evictions per rental unit, we saw a nonsignificant reduction of 0.172 evictions from a pre-expansion mean of 4.037.

 Dependent variables Independent variables Evictions per month Evictions per capita Log of evictions Evictions per rental unit California, after expansion −24.517** −0.029** −0.117** −0.172* Unemployment rate 0.617 0.005 0.019 0.045 Poverty rate −3.853 −0.006 −0.018 −0.069** Pre-expansion mean 224.718 0.261 4.037 4.037 R2 0.949 0.921 0.946 0.918 Number of county-month observations 17,925 17,925 16,974 17,925

SOURCE Authors’ analysis of data from American Information Research Services (AIRS). NOTES The exhibit shows the results of difference-in-differences regressions that were adjusted for county-specific fixed effects, year-month-specific fixed effects, and county-specific linear time trends (not shown). The sample consists of county-month observations. More information on the composition of the sample and the dates of Medicaid expansion is in appendix exhibit A.1., table A.2., and table B.1., and appendix table C.1. is a version of this exhibit with standard errors and detailed p values (see note 12 in text).

*$p<0.10$

**$p<0.05$

As expected, we found a more pronounced effect of Medicaid expansion on evictions in counties with above-median shares of uninsured residents before expansion, as measured in 2010, compared to those with below-median shares: 51.505 versus 3.270 fewer evictions after expansion (exhibit 3).

 Dependent variable: number of evictions Independent variable Share uninsured above median Share uninsured below median California, after expansion −51.505** 3.270 Unemployment rate 1.341 2.150** Poverty rate −6.384 −1.427 Pre-expansion mean 318.858 117.568 R2 0.946 0.954 Number of county-month observations 8,419 9,506

SOURCE Authors’ analysis of data from American Information Research Services (AIRS). NOTES The exhibit shows the results of difference-in-differences regressions in which counties were stratified by the share of the population that was uninsured in 2010. The median share uninsured across counties in the sample was 0.28. Regressions were adjusted for county-specific fixed effects, year-month-specific fixed effects, and county-specific linear time trends (not shown). The sample consists of county-month observations. More information on the composition of the sample and the dates of the Medicaid expansion is available, as explained in the notes to exhibit 2. Appendix table C.2. is a version of this exhibit with standard errors and detailed p values (see note 12 in text).

**$p<0.05$

When we examined our main results without the adjustment for county-specific linear trends, we observed both much greater reductions in evictions and an increase in statistical significance (see appendix table D.1. for more detail).12 A visual inspection of the event studies suggested to us that there were differences in trends following the Great Recession that would support an additional examination of the data using synthetic control methods. Reassuringly, we found that the results were robust to using a synthetic control approach of matching expansion counties to nonexpansion counties by their pre-expansion eviction trends. Using this approach, we also saw a significant divergence following Medicaid expansion, with expansion counties experiencing fewer evictions than their nonexpansion counterparts did (see appendix table E.1. for differences between actual evictions and evictions in a synthetic control group).12 These estimates suggest an effect of roughly seventy fewer evictions, on average, which is quite similar to the results in the difference-in-differences models that were unadjusted for county-specific linear trends. All of the associated p values in the post-expansion period were significant below conventional levels.

We explored the effect of Medicaid expansion on evictions in four ways, using difference-in-differences regressions, event study regressions, a synthetic control approach, and a visual comparison of the “raw count” percentage change from January 2008 to January 2014 (see appendix exhibit D.2.).12 Some patterns in the results suggested that we did not have a perfect comparison group. That said, all four approaches pointed in the same direction and implied that Medicaid expansion was associated with a decrease in evictions.

## Discussion

This study found that Medicaid expansion reduced the rate and number of home evictions, using data from the counties in California that expanded Medicaid early. The estimated magnitude of this relationship falls within the range expected from related results in the existing literature on the economic impacts of Medicaid expansion. Examining public records that covered credit reports, evictions, bankruptcies, and wage garnishments, Sarah Miller and colleagues found a reduction of 11–16 percent in adverse financial outcomes in the first year following Medicaid expansion in Michigan.16 Similar to our findings, that study showed that the estimated effect increased over time, with the emergence of significant effects six to ten months following expansion.

Furthermore, Naomi Zewde and Christopher Wimer found that Medicaid reduced annual out-of-pocket medical spending by approximately \$100 per beneficiary, on average.17 People who became eligible for Medicaid under the expansion necessarily live on small margins between income and expenses. Thus, the share of the population with Medicaid-related spending reductions near or greater than the \$100 mean could plausibly better meet housing obligations and avoid eviction.

This study contributes to an existing body of literature suggesting that one of the principal benefits of Medicaid expansion is related to protection from encumbering medical debt, leading to improved financial well-being. Research has shown that California’s early Medicaid expansion was also associated with a reduction in payday borrowing, which provides further evidence of the financial protection that the expansion provided.10 In addition, the Oregon Health Insurance Experiment’s randomized trial findings demonstrated an improvement in financial stability and a reduction in catastrophic medical debt among Medicaid recipients.8,9 Health care coverage may be keeping households from “falling over the brink,” helping them meet their living expenses as the growing cost of medical care constrains household budgets.

Health care coverage may be keeping households from “falling over the brink.”

## Conclusion

Our findings suggest that Medicaid expansion may play an important role in preventing eviction and its devastating long-term consequences for the physical and mental health, housing quality, and financial well-being of already burdened low-income families. The findings add to a body of evidence suggesting that the economic consequences for recipient households will be felt beyond the immediate coverage implications. While voters in Idaho, Nebraska, and Utah recently approved Medicaid expansion through ballot referendums,18 other states are experimenting with programmatic retractions that include instituting work requirements.19 This body of literature helps inform policy makers of the ramifications of expanding or retracting public coverage on the economic well-being of the population. Our findings indicate that Medicaid not only is an important part of the health care safety net but also may be considered a key strategy for addressing poverty-related housing instability.

## ACKNOWLEDGMENTS

A draft of this article was presented as part of the Michael M. Davis Lecture, Center for Health Administration Studies, University of Chicago, on October 9, 2018. An earlier version of the article was presented at the AcademyHealth Annual Research Meeting in Washington, D.C., June 2–4, 2019. The authors thank Katherine Baicker for her thoughtful feedback on an earlier draft of the article and the Russell Sage Foundation for funding the study (Award No. 94-16-02). This is an open access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt, and build upon this work, for commercial use, provided the original work is properly cited. See https://creativecommons.org/licenses/by/4.0/.

## NOTES

• 1 Vásquez-Vera H, Palència L, Magna I, Mena C, Neira J, Borrell C. The threat of home eviction and its effects on health through the equity lens: a systematic review. Soc Sci Med. 2017;175:199–208. Crossref, Medline
• 2 Desmond M, Shollenberger T. Forced displacement from rental housing: prevalence and neighborhood consequences. Demography. 2015;52(5):1751–72. Crossref, Medline
• 3 Fazel S, Geddes JR, Kushel M. The health of homeless people in high-income countries: descriptive epidemiology, health consequences, and clinical and policy recommendations. Lancet. 2014;384(9953):1529–40. Crossref, Medline
• 4 Jego M, Grassineau D, Balique H, Loundou A, Sambuc R, Daguzan Aet al. Improving access and continuity of care for homeless people: how could general practitioners effectively contribute? Results from a mixed study. BMJ Open. 2016;6(11):e013610. Crossref, Medline
• 5 Marr T. Millions of renters face eviction—why today’s housing market is partially to blame. Redfin Real-Time [blog on the Internet]. 2016 Dec 12 [cited 2019 Jul 9]. Available from: https://www.redfin.com/blog/2016/12/millions-of-renters-face-eviction-why-todays-housing-market-is-partially-to-blame.html Google Scholar
• 6 Katz MH. Homelessness—challenges and progress. JAMA. 2017;318(23):2293–4. Crossref, Medline
• 7 Tipirneni R, Goold SD, Ayanian JZ. Employment status and health characteristics of adults with expanded Medicaid coverage in Michigan. JAMA Intern Med. 2018;178(4):564–7. Crossref, Medline
• 8 Baicker K, Taubman SL, Allen HL, Bernstein M, Gruber JH, Newhouse JPet al. The Oregon experiment—effects of Medicaid on clinical outcomes. N Engl J Med. 2013;368(18):1713–22. Crossref, Medline
• 9 Finkelstein A, Taubman S, Wright B, Bernstein M, Gruber J, Newhouse JPet al. The Oregon Health Insurance Experiment: evidence from the first year. Q J Econ. 2012;127(3):1057–106. Crossref, Medline
• 10 Allen H, Swanson A, Wang J, Gross T. Early Medicaid expansion associated with reduced payday borrowing in California. Health Aff (Millwood). 2017;36(10):1769–76. Go to the article
• 11 Eviction Lab. Methodology report: v.1.1.0 [Internet]. Princeton (NJ): Princeton University; 2018 May 7 [cited 2019 Jul 9]. Available from: https://evictionlab.org/docs/Eviction%20Lab%20Methodology%20Report.pdf Google Scholar
• 12 To access the appendix, click on the Details tab of the article online.
• 13 Abadie A, Diamond A, Hainmueller J. Synthetic control methods for comparative case studies: estimating the effect of California’s Tobacco Control Program. J Am Stat Assoc. 2010;105(490):493–505. Crossref
• 14 Hu L, Kaestner R, Mazumder B, Miller S, Wong A. The effect of the Patient Protection and Affordable Care Act Medicaid expansions on financial wellbeing [Internet]. Cambridge (MA): National Bureau of Economic Research; [revised 2018 Feb; cited 2019 Jul 9]. (NBER Working Paper No. 22170). Available from: https://www.nber.org/papers/w22170.pdf Google Scholar
• 15 California Department of Health Care Services. Medi-Cal’s historical period of growth: a 24-month examination of how the program has changed since December 2012. Sacramento (CA): DHCS; 2015 Aug. Google Scholar
• 16 Miller S, Hu L, Kaestner R, Mazumder B, Wong A. The ACA Medicaid expansion in Michigan and financial health [Internet]. Cambridge (MA): National Bureau of Economic Research; 2018 Sep [cited 2019 Jul 9]. (NBER Working Paper No. 25053). Available from: https://www.nber.org/papers/w25053.pdf Google Scholar
• 17 Zewde N, Wimer C. Antipoverty impact of Medicaid growing with state expansions over time. Health Aff (Millwood). 2019;38(1):132–8. Go to the article
• 18 Kliff S. Idaho, Nebraska, and Utah vote to expand Medicaid. Vox [serial on the Internet]. 2018 Nov 7 [cited 2019 Jul 9]. Available from: https://www.vox.com/2018/11/7/18055848/medicaid-expansion-idaho-nebraska-utah Google Scholar
• 19 Wilson JC, Thompson J. Nation’s first Medicaid work requirement sheds thousands from rolls in Arkansas. Health Affairs Blog [blog on the Internet]. 2018 Oct 2 [cited 2019 Jul 9]. Available from: https://www.healthaffairs.org/do/10.1377/hblog20181001.233969/full/ Google Scholar