Key Points
Question
Is Medicaid expansion associated with decreased home evictions, and is this association heterogenous across states and counties?
Findings
In this cohort study of 25 398 county-year observations (across 40 US states) for the years 2002 through 2018, Medicaid expansion was significantly associated with a decrease in mean number of county eviction judgments and a reduction in the rate of eviction judgments. Associations varied across both states and counties, with 29% of the variation being explained by across-state differences, and 9% being explained by county-level demographic and uninsurance differences.
Meaning
These results suggest that Medicaid expansion is associated with reductions in eviction judgments and eviction judgment rates; these associations may vary considerably both across as well as within states (across counties).
This cohort study compares county-level rates of home evictions among states that expanded Medicaid under the Affordable Care Act vs those that did not.
Abstract
Importance
Prior research has identified associations between housing insecurity and poor health outcomes.
Objective
To evaluate the association between US state Medicaid expansions and reductions in eviction; to examine the persistence of these associations and how they vary across US states and counties.
Design, Setting, and Participants
This cohort study of 25 398 county-year observations (across 40 states) used US eviction and census data for the years 2002 through 2018 (ie, 17 years). County-level associations were estimated using interactive fixed effects counterfactual estimators, and models were selected using cross validation. Across-county treatment association heterogeneities were assessed using multivariable regression methods. Analyses were performed in July of 2022.
Exposure
State-level Medicaid expansion under the Patient Protection and Affordable Care Act.
Main Outcomes and Measures
Eviction judgments; eviction judgments per 100 renter-occupied households.
Results
Among a total of 774 treated counties (with Medicaid expansion) and 720 control counties (untreated, without Medicaid expansion), mean (SD) eviction judgments for treated counties were 534.78 (1945.84) eviction judgments in the pre-2014 period (mean [SD] eviction rate, 2.25 [2.18] per 100 households), which decreased to 463.67 (1499.39) eviction judgments in the post-2014 period (mean [SD] eviction judgment rate, 2.02 [1.81] per 100 households). Control group mean (SD) county eviction judgments were 477.22 (1592.18) eviction judgments (mean [SD] eviction judgment rate, 1.91 per 100 households) pre-2014, and 490.22 (1575.19) eviction judgments (mean [SD] eviction judgment rate, 1.89 per 100 households) post-2014. Model estimates indicate that Medicaid expansion was associated with reductions in county eviction judgments by −66.49 (95% CI, −132.50 to −0.48; P = .047) and reductions of the eviction judgment rate by −0.25 (95% CI, −0.35 to −0.14; P < .001). Associations remained broadly consistent between 2014 and 2018, although some diminishment of associations occurred in 2018. Approximately 29% of the across-county treatment association variation was explained by across-state differences, while 9% was explained by county-level demographic and uninsurance differences.
Conclusions and Relevance
In this cohort study, Medicaid expansion was associated with reductions in eviction judgments and eviction judgment rates; however, these associations were found to vary considerably both across as well as within states (across counties). These findings suggest that the channel between Medicaid expansion and evictions is sensitive to state environments as well as county specific population demographics and uninsurance levels.
Introduction
In the US, estimates indicate that about 1 in 5 households were housing insecure in 2020, that is, had no or only slight confidence in being able to make their next rent or mortgage payment.1 Housing insecurity rates are furthermore higher among minority populations.1 For example, Ong et al2 report that in the state of California, renters from racial or ethnic minority groups are more likely than their non-Hispanic White counterparts to struggle to keep up with rent payments; and when it comes to receipt of rental relief, Asian and Latino renters are underrepresented relative to White renters. Racial and ethnic disparities within housing security is an ongoing concern, as a growing body of work has helped identify linkages between individuals’ housing security, health care access,3 and individual health outcomes.4,5,6 Recent trends of rising inflation, and in particular rental prices, are further anticipated to exacerbate this problem via increased rates of eviction filings and subsequent eviction judgments forcing tenants out of their residence.7 While policies such as eviction moratoria have provided short-term solutions to the threat of eviction facing many families, alternative policy solutions are desirable.8 To this end, recent studies have indicated important associations between policies that can help alleviate financial distress among low-income families and in turn reduce home evictions. Pertaining to minimum-wage legislation, Agarwal et al9 show that state minimum-wage increases reduced the risk of renters defaulting on their lease contracts by 1.7 percentage points, representing 10.6% fewer rental defaults. Additional work has shown similar results pertain to state-level adoption of Medicaid expansions under the Affordable Care Act. In particular, Zewde et al10 show that in the first 3 years following Medicaid expansion, states that adopted these expansions experienced significant reductions in evictions.
Given the policy importance of the Medicaid expansion channel in being able to ameliorate eviction problems within US nonexpansion states, this study sought to build on, and add to, prior work in this area. This was done in 2 ways. First, we present results using a longer post-Medicaid expansion adoption time-period. While prior work has used 3 years of posttreatment data,10 we included 5 years of data to provide updated results. Additionally, our analysis of a longer posttreatment period allowed us to examine whether associations remain persistent across longer periods of time. Second, while prior work has focused upon the estimation of the overall US average treatment effect, we also focused on the estimation of potentially heterogenous treatment effects that may exist across states, as well as within states across different counties. This study examined whether such heterogeneities exist, and whether they vary systematically across state differences pertaining to their timing of expansion as well as whether they adopted 1115 waivers as part of their expansions. As 1115 waivers may allow states greater flexibility in their design and implementation of their Medicaid programs, all while remaining budget neutral to the federal government,11 we hypothesized that added program flexibility may translate into increased program efficacy. We also examined within-state county variation based on population demographics, uninsurance, and pretreatment levels of evictions. As such, we sought to provide updated results as well as details on important nuances that may exist pertaining to Medicaid adoption in terms of its potential association with home evictions. Given these objectives, this study also contributes to a broader field of research examining the wide implications of Medicaid expansions.12,13
Methods
Conceptual Model
The net association of Medicaid expansion with county level eviction rates is theoretically ambiguous, as 2 separate (and opposite acting) pathways may affect individuals’ likelihood of eviction (Figure 1). First, the expansion of Medicaid eligibility, and with that the access to public insurance, has been hypothesized to alleviate individual budgetary constraints (for covered individuals) as these individuals no longer need to purchase private coverage and face lower medical bills when they receive medical care.12,13 What could be expected to result would be a reduction in covered individuals’ risk of eviction. Additionally, we note that insurance coverage, by means of lowering the price of medical care, may induce individuals to increase their investment into their own health capital, which (within the framework of the Grossman model of health consumption) would act to decrease the time that individuals spend sick and increasing the amount of time that they are able to allocate toward working.14 Therefore, reduced cost of medical care can act to change individual behavior toward increased health investment and increased hours of work, which may also translate into reducing covered individuals’ risk of eviction. Second, since Medicaid coverage is tied to income, it is possible that it may act to reduce certain individual’s incentive of pursuing (or retaining) employment (this pertains to individuals whose labor market participation decision directly affects their Medicaid eligibility).15 This unintended consequence can in turn reduce individual employment and increase the risk of eviction. While these channels are described as 2 separate pathways, they are not mutually exclusive. While results from prior work suggests that the first of these 2 channels is the (on average) most dominant,10,16 both of these channels because their relative strengths may vary across regions, and as such, this variation may help explain heterogeneities within resulting Medicaid expansion treatment effect sizes when assessed across US county regions.
Figure 1. Conceptual Model of Pathways Between Medicaid Expansion and Home Evictions.
Upper pathway (path 1) illustrates how Medicaid expansion may yield increased disposable income, and in turn, reduce the risk of evictions. The lower pathway (path 2) illustrates how Medicaid expansion may have the unintended consequence of disincentivizing work, which may reduce individual disposable income, and with that lead to an increased risk of eviction. It should be noted that pathways 1 and 2 are not mutually exclusive.
This study was based on public and deidentified data and did not constitute human subjects research as defined by 45 CFR §46.102. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.
Research Design and Study Sample
For our primary analysis we used county-year–level eviction judgment data for the 2002 to 2018 period from the Princeton Eviction Lab.17 These data are based on eviction judgment information from LexisNexis Risk Solutions and are nationally representative with the exception of 2 states (North Dakota and South Dakota).17 County-level population data were added for each county-period that were sourced from the Area Health Resource Files,18 which obtained these data from the US Census Bureau. State-level information on Medicaid expansion status, 1115 waiver status and year of expansion were all sourced from the Kaiser Family Foundation.19 Based on these data, we restricted our sample to (1) counties with continuous eviction data reporting for the full 2002 to 2018 period and (2) states with Medicaid expansion adoption in the years of 2014, 2015, and 2016. Our resulting analysis sample consisted of 25 398 county-year observations. Additional study sample creation details are available in eMethods and eTable 1 in the Supplement.
Lastly, for the purpose of decomposing and explaining estimated heterogenous treatment effect sizes, we additionally used demographic data for the purpose of constructing a secondary analysis sample. This sample combined our estimated county-level average treatment effect upon the treated estimates with Census Bureau data from the 2013 AHRQ county database.20 This resulted in a secondary analysis sample of 774 (treated) county observations.
Study Variables
Outcome Measures
Our primary measure of household evictions was based on the county eviction judgment rates per 100 renter occupied households. This measure captures the rate of cases that result in an eviction judgment against the defendant, forcing them to vacate a premise by a specific date. Our secondary measure was the county-level raw eviction judgment count.
Exposure Measure
The Medicaid expansion status of a county within our data was captured by a treatment vector that took the value of zero in periods of nonexpansion and the value of 1 once a county was treated with the Medicaid expansion. As such, the language of treatment refers to the notion of state-level Medicaid expansion adoption, while nontreatment (or control) refers to the absence of such a policy adoption.
Decomposition Measures
To decompose estimated average treatment effects upon the treated, we used additional data. First, we used data on whether a given state had enacted a 1115 Waiver and the year of expansion adoption. Second, we used state indicators to capture across-state differences within the average treatment effects upon the treated, which allowed us to capture overall latent legal and structural differences across states. Two examples of latent (state-level) structural factors that may factor into the average treatment effect upon the treated are differences in the supply of subsidize housing that may lessen the risk of eviction of individuals, and across-state differences in the cost of housing. Third, we also used county-level (pre-expansion) information to examine whether the average treatment effects upon the treated counties varies systematically within states across readily available county demographics. Specifically, we included county-level information on: population share below 18 years of age; population share ages 65 years and older; the proportion of non-Hispanic Black, Hispanic, Asian, and female individuals, and population count. Additionally, we also used information on the share of the population under the age of 65 years that was uninsured and reported an income at or below 138% of the federal poverty line, as this population may be economically vulnerable and sensitive to the expansion of Medicaid eligibility.
Statistical Analysis
To estimate average treatment effects associated with counties treated with an expansion of Medicaid eligibility, we drew upon counterfactual models that included recently developed fixed effects and interactive fixed effects models.21,22 These methods utilized pretreatment period data on untreated control counties to estimate counterfactual (synthetic) outcomes for the treated units (posttreatment). These methods have the important advantages of producing more reliable estimates than traditional 2-way fixed-effects models when treatment effects are heterogenous (as our conceptual model would have us believe they are within our setting) and when there may exist unobserved time-varying confounders.21,22 Additionally, these methods allow researchers to reduce their reliance on the strong parallel trends assumption of traditional difference-in-difference methods.22
In terms of our implementation, we used a fixed-effects counterfactual model for the analysis of our eviction rate outcome measure, and an interactive fixed-effects counterfactual model for the analysis of our eviction count measure. Both models were selected based on cross-validation results across multiple models (see eMethods and eTable 2 in the Supplement). Each model controlled for the county-level population count. The appropriateness of each model was further assessed using a placebo test (ie, a test for significant treatment effects in the pretreatment periods when no such effects are expected) and tests for no pretreatment trends—both tests indicated good model fit (see eMethods and eTable 3 in the Supplement).21,22,23 Robustness was additionally assessed using 3 alternative estimation designs, consisting of: (1) traditional 2-way fixed-effects estimation; (2) the interaction weighted estimator approach; and (3) the inverse probability tilting approach.24,25 Results from these robustness checks are provided in eMethods and eTables 4 through 6 in the Supplement.
Lastly, the source of variation within the estimated county-specific average treatment effects were examined. This secondary decomposition analysis was achieved using standard multivariable regression analyses with and without adjustment for state fixed effects and county covariates.
Analyses were performed using R version 4.2.1 (R Foundation) and Stata MP version 17 (StataCorp LLC). Extension packages consisted of fect, csdid, and eventstudyinteract.21,24,26 Statistical significance was assigned at the 95% CI level, and hypothesis tests were 2-sided.
Results
Our analysis sample consisted of a balanced panel of 25 398 county-year observations (1494 unique counties across 40 states and 17 years). This corresponded to 774 counties treated with Medicaid expansion across 26 states, and 720 untreated control counties across 14 states (Figure 2; eMethods in the Supplement).
Figure 2. Treated and Control States and Counties.
Treated geographic areas were counties that were part of a state-level Medicaid expansion between the years 2014 and 2016. Control counties did not participate in expansion, while gray states or counties indicate an absence of data.
Eviction judgment rates by county in the pretreatment period of 2013 and in the final year of our data (2018) indicated considerable variation across state boundaries as well as across-counties within several states in our sample (Figure 3). Sample descriptives further indicated that mean (SD) county evictions for treated counties were 534.78 (1945.84) evictions in the pre-2014 period (mean [SD] eviction rate, 2.25 [2.18] per 100 renter-occupied households); evictions in treated counties fell to 463.67 (1499.39) evictions in the post-2014 period (eviction rate, 2.02 [1.81] per 100 renter-occupied households). Control group mean county evictions were 477.22 (1592.18) evictions pre-2014 (eviction rate, 1.91 [1.94] per 100 renter-occupied households) and 490.22 (1575.19) evictions post-2014 (1.89 [1.61] per 100 renter-occupied households). This descriptive evidence supported a negative treatment association between Medicaid expansions and evictions (and eviction rates) within our sample.
Figure 3. Eviction Rates by County.
The model estimates for the eviction judgment rate and eviction judgments during posttreatment years indicated that the expansion of Medicaid eligibility was associated with reductions in county eviction judgments by 66.49 (95% CI, −132.50 to −0.48; P = .047) and reductions in the eviction judgment rates by 0.25 (95% CI, −0.35 to −0.14; P < .001) (Figure 4).
Figure 4. Estimated Average Treatment Effect Over Time (ATT) for Counties Participating in Medicaid Expansion.
Regression results that utilized these eviction rate effect estimates as their outcome measures indicated that treated states with 1115 waiver adoption had on average a −0.31 (95% CI, −0.49 to −0.14; P < .001) larger reduction in home eviction rates than did non 1115 waiver states (Table). The timing of Medicaid expansion also appeared to factor into eviction rates, as later expansion states (with expansions initiated in 2015 and 2016) experienced less benefit associated with Medicaid expansion than did the early expansion states (with expansion in 2014).
Table. Regression of Average Treatment Effect on Treated (ATT) Estimates on State and County Characteristics.
| Characteristics | ATT estimate (95% CI) | |||
|---|---|---|---|---|
| 1 a | 2 b | 3 c | 4 d | |
| State with 1115 waiver | −0.31 (−0.49 to −0.14) | NA | NA | NA |
| Year of expansion | ||||
| 2015 | 0.69 (0.32 to 1.06) | NA | NA | NA |
| 2016 | 0.81 (0.59 to 1.02) | NA | NA | NA |
| Judgment rate | NA | NA | −0.16 (−0.25 to −0.07) | −0.15 (−0.24 to −0.06) |
| Age | ||||
| <18 y | NA | NA | NA | 1.16 (−1.69 to 4.01) |
| ≥65 y | NA | NA | NA | 6.16 (3.97 to 8.35) |
| Race or ethnicity | ||||
| Asian | NA | NA | NA | 1.96 (−1.02 to 4.95) |
| Hispanic | NA | NA | NA | 0.16 (−1.03 to 1.36) |
| Non-Hispanic Black | NA | NA | NA | 0.62 (−0.43 to 1.67) |
| Women | NA | NA | NA | −3.85 (−7.78 to 0.08) |
| Population | NA | NA | NA | 0 (−0.03 to 0.02) |
| Uninsured population, %e | NA | NA | NA | −2.71 (−4.84 to −0.58) |
| Constant | −0.18 (−0.26 to −0.11) | 0.62 (0.62 to 0.62) | 0.69 (0.65 to 0.73) | 2.49 (0.42 to 4.56) |
| Counties observed, No. | 774 | 774 | 774 | 774 |
| R2 | 0.05 | 0.29 | 0.34 | 0.38 |
| State fixed effects | No | Yes | Yes | Yes |
Abbreviations: FPL, federal poverty level; NA, not applicable.
Model 1 examines the association between ATT and state 1115 waiver and year of expansion.
Model 2 examines the association between ATT and state fixed effects (ie, state indicators).
Model 3 examines the association between ATT and preexpansion judgment rates in 2013 and state fixed effects.
Model 4 examines the association between ATT and preexpansion county covariates in 2013 and state fixed.
Defined as having income ≤138% of FPL and age ≤65 years.
Several additional results were significant when considering variations across states and counties (Table). Approximately 29% of the variation in the county-level associations was explained by across state differences. The addition of the pretreatment county judgment rates in 2013 increased the explanatory power by an additional 5%. The results here suggest that counties with high rates of evictions in the pretreatment period of 2013 had disproportionately larger reductions in eviction rates following Medicaid expansion. Third, when additional county-level demographics were added, areas with a larger share of individuals aged 65 years and above experienced on average less beneficial effects from Medicaid expansion (coefficient estimate, 6.16; 95% CI, 3.97 to 8.35; P < .001), and areas with a larger share of economically vulnerable uninsured individuals experienced more benefits associated with Medicaid expansion (coefficient estimate, −2.71; 95% CI, −4.84 to −0.58; P = .013).
Discussion
Results from our study provide further support of Medicaid expansions having had an overall beneficial associated with lowering county eviction rates within expansion states. As such, our study builds upon important prior work that has previously documented this association either using a shorter posttreatment period10 or within the context of a single state.16 Second, results from our study highlight that treatment effects upon the treated are heterogenous both across states as well as within states across counties. This is an important observation as it highlights that the overall association of a Medicaid expansion decision on eviction rates within a given county will depend on the state environment—for example, in terms of its legal structures and protections of renters, as well as county-specific population demographics and uninsurance rates. Pertaining to state-level differences, estimated treatment effects were found to vary across states with and without 1115 waivers, as well as across the year of Medicaid expansion. Here, we note that the larger reduction effects within 1115 waiver states may stem from the greater flexibility offered by these waivers in terms of how states design and implement their Medicaid programs.11 Related to within-state county-level differences, we observed that Medicaid expansion was associated with the greatest reduction in eviction rates within counties with higher eviction rates prior to Medicaid expansion adoption, as well as in counties that had a larger share of potentially eligible individuals (ie, individuals under the age of 65 years). This observation seems reasonable as individuals ages 65 years and older are eligible for Medicare, and thus less likely to directly benefit from Medicaid expansions. We also observed that counties with a larger share of economically vulnerable uninsured individuals experienced greater benefit associated with Medicaid expansion. This seems sensible as low-income uninsured individuals may be more likely to have foregone insurance due to budgetary constraints, as opposed to other reasons.
Something that we did not observe was any systematic variation within the average treatment effect upon the treated in terms of the racial and ethnic makeup of counties. This is important to highlight as non-Hispanic Black and Hispanic populations are overrepresented within the segment of the US population that lives in poverty,27 is uninsured,28 and has a higher risk for eviction.1 Data from a 2019 survey29 and a 2022 study30 further indicate that non-Hispanic Black and Hispanic families’ wealth is on average less than 15% that of non-Hispanic White families. While our decomposition analyses adjust for some of these channels, we were not, for example, able to adjust for wealth differences across regions. As such, the null finding regarding the population share that is non-Hispanic Black, Hispanic, and Asian suggests potential Medicaid enrollment disparities across minority populations, as we would otherwise have expected some potential heterogeneity across this dimension of county population characteristics. This observation appears further supported by data showing that non-Hispanic Black and Hispanic individuals are overrepresented within the population of individuals that are eligible for Medicaid but who remain unenrolled.31 We also note that alternative mechanisms may be at play here, and that an examination of such mechanisms represents an opportunity for future research.
Our findings appear to suggest 2 important policy implications. First, our results indicate that the expansion of Medicaid income eligibility, on average, may help reduce households’ risk of eviction. As such, this is an important benefit that state policy makers need to consider in weighing the costs and benefits associated with Medicaid expansion. Second, we find evidence that the adoption of Medicaid expansion alone might not be enough to ensure qualified individuals gain coverage, and therefore are able to benefit from associated eviction risk reduction. The observations that counties with higher racial and ethnic minority prevalence do not experience significantly larger gains in eviction risk reduction, and that these populations are otherwise overrepresented among those who are Medicaid eligible but uninsured, suggests that barriers may exist to prevent these populations from obtaining Medicaid coverage. As such, additional efforts are needed to ensure that all qualified populations gain access to Medicaid coverage as part of expansion efforts. This recommendation seems particularly potent given that many remaining nonexpansion states (particularly those located in the south of the US) have high populations of racial and ethnic minorities.
Limitations
This study had several limitations. The validity of our identification strategy—and hence of our estimated results—hinges upon the satisfaction of the identifying assumptions of our approach. While the use of a synthetic counterfactual control approach helps ameliorate the strong assumption of parallel pretreatment trends of treated and control units, the satisfaction of this condition is not guaranteed. Our model showed an appropriate fit, as the average treatment effects upon the treated appear to be aligned with a null effect within the pretreatment period of our sample (Figure 4); however, we provide 2 sets of additional results in support of the goodness of fit of our model, and hence of the appropriateness of our approach (eTable 3 in the Supplement). First, we show that placebo test results based on projected average treatment effects for 5 pretreatment periods do not reject the null of no average treatment effect during this period. Second, equivalence tests applied to the placebo period yielded the same conclusion. Third, pretrend tests of the full pretreatment period yielded further supporting evidence of there not existing any pretreatment trends. While these tests provided support of the goodness of fit of our model, we also provided additional robustness results based on 3 alternative estimation designs (2-way fixed effects estimation; the interaction weighted estimator approach24; and the inverse probability tilting approach25). Results from these robustness checks indicated average treatment effects upon the treated that were qualitatively similar to our main results, and as such, these additional results indicated that our main findings remained robust across these alternative designs (see eTable 4, eTable 5, and eTable 6 in the Supplement). Lastly, we note that while our data allowed us to include counties across 40 geographically diverse US states, this study did not include a complete set of US states; additionally, we also note that in cases where Medicaid expansion adopting states may have simultaneously adopted other policies that support low-income individuals (in a disproportionate fashion to nonadoption states), this may act to confound and potentially overstate our estimates.
Conclusions
In this cohort study, Medicaid expansion was associated with reductions in eviction judgments and eviction judgment rates; however, these associations were found to vary considerably both across as well as within states (across counties). These findings suggest that the channel between Medicaid expansion and evictions is sensitive to state environments as well as county-specific population demographics and uninsurance levels.
eMethods.
eTable 1. Analysis Sample Creation
eTable 2. Cross Validation Results Based on Mean Squared Prediction Errors
eTable 3. Placebo Tests and No Pretrend Tests
eTable 4. Two-Way Fixed Effects Regression Average Treatment Effect on Treated Estimates
eTable 5. Inverse Probability Tilting Average Treatment Effect on Treated Estimates (Based on Aggregate Effects).
eTable 6. Interaction Weighted Estimator Average Treatment Effect on Treated Estimates (Based on Aggregate Effects).
eReferences.
Data Sharing Statement
References
- 1.Centers for Economic and Policy Research . Housing Insecurity by Race and Place During the Pandemic. CEPR press release. April 5, 2021. Accessed August 24, 2022. https://cepr.net/report/housing-insecurity-by-race-and-place-during-the-pandemic/.
- 2.Ong P, Gonzalez S, Nazario P. Housing Insecurity Persists for Renters of Color amid the COVID-19 Pandemic. UCLA Center for Neighborhood Knowledge . April 2022. Accessed on November 1, 2022. https://escholarship.org/uc/item/1h62d5wv#main
- 3.Martin P, Liaw W, Bazemore A, Jetty A, Petterson S, Kushel M. Adults with housing insecurity have worse access to primary and preventive care. J Am Board Fam Med. 2019;32(4):521-530. doi: 10.3122/jabfm.2019.04.180374 [DOI] [PubMed] [Google Scholar]
- 4.Tsai AC. Home foreclosure, health, and mental health: a systematic review of individual, aggregate, and contextual associations. PLoS One. 2015;10(4):e0123182. doi: 10.1371/journal.pone.0123182 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Taylor, LA. Housing and health: an overview of the literature. Health policy brief. HealthAffairs. 2018;10. doi: 10.1377/hpb20180313.396577 [DOI] [Google Scholar]
- 6.Nyamathi AM, Salem BE. The impact of unstable housing on health. Lancet Public Health. 2021;6(5):e265-e266. doi: 10.1016/S2468-2667(21)00035-9 [DOI] [PubMed] [Google Scholar]
- 7.US Government Accountability Office . (2020). How Covid-19 Could Aggravate the Homelessness Crisis? GAO WatchBlog. August 25, 2020. Accessed August 24, 2022. https://www.gao.gov/blog/how-covid-19-could-aggravate-homelessness-crisis
- 8.US Centers for Disease Control and Prevention . (2022). CDC Issues Eviction Moratorium Order in Areas of Substantial and High Transmission. CDC press release. August 3, 2021. Accessed August 24, 2022. https://www.cdc.gov/media/releases/2021/s0803-cdc-eviction-order.html
- 9.Agarwal S, Ambrose BW, Diop M. Minimum wage increases and eviction risk. J Urban Econ. 2022;129:103421. doi: 10.1016/j.jue.2021.103421 [DOI] [Google Scholar]
- 10.Zewde N, Eliason E, Allen H, Gross T. The effects of the ACA Medicaid expansion on nationwide home evictions and eviction-court initiations: United States, 2000–2016. Am J Public Health. 2019;109(10):1379-1383. doi: 10.2105/AJPH.2019.305230 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.National Conference of State Legislatures . Understanding Medicaid Section 1115 Waivers: A Primer for State Legislators. 2017. Accessed November 1, 2022. https://www.ncsl.org/Portals/1/Documents/Health/Medicaid_Waivers_State_31797.pdf
- 12.Antonisse L, Garfield R, Rudowitz R, Artiga S. The effects of Medicaid expansion under the ACA: updated findings from a literature review. June 2016. Accessed August 24, 2022. http://www.nationaldisabilitynavigator.org/wp-content/uploads/news-items/KFF_Effects-of-Medicaid-Expansion_June-2016.pdf
- 13.Antonisse L, Garfield R, Rudowitz R, Artiga S. The effects of Medicaid expansion under the ACA: updated findings from a literature review. February 2017. Accessed August 24, 2022. https://nationaldisabilitynavigator.org/wp-content/uploads/news-items/KFF_Effects-of-Medicaid-Expansion-Lit-Review_Feb-2017-chart.pdf
- 14.Grossman M. On the concept of health and the demand for health capital. J Polit Econ. 1972;80(2):223-255. doi: 10.1086/259880 [DOI] [Google Scholar]
- 15.Peng L, Guo X, Meyerhoefer CD. The effects of Medicaid expansion on labor market outcomes: evidence from border counties. Health Econ. 2020;29(3):245-260. doi: 10.1002/hec.3976 [DOI] [PubMed] [Google Scholar]
- 16.Allen HL, Eliason E, Zewde N, Gross T. Can Medicaid expansion prevent housing evictions? Health Aff (Millwood). 2019;38(9):1451-1457. doi: 10.1377/hlthaff.2018.05071 [DOI] [PubMed] [Google Scholar]
- 17.Gromis A, Fellows I, Hendrickson JR, et al. Estimating eviction prevalence across the United States. PNAS. 2022;119(21):e2116169119. doi: 10.1073/pnas.2116169119 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Health Resources & Services Administration . Area Health Resource Files. Updated July 31, 2021. Accessed August 24, 2022. https://data.hrsa.gov/topics/health-workforce/ahrf
- 19.Kaiser Family Foundation . Status of State Medicaid Expansion Decisions: Interactive Map. November 9, 2022. Accessed August 24, 2022. https://www.kff.org/medicaid/issue-brief/status-of-state-medicaid-expansion-decisions-interactive-map/
- 20.Agency for Healthcare Research and Quality . AHRQ social determinants of health county database. Published December 2020. Accessed November 1, 2021. https://www.ahrq.gov/sdoh/data-analytics/sdoh-data.html
- 21.Liu L, Wang Y, Xu Y. A practical guide to counterfactual estimators for causal inference with time-series cross-sectional data. Am J Pol Sci. Published online May 2022. 1-17. doi: 10.1111/ajps.12723 [DOI] [Google Scholar]
- 22.Xu Y. Generalized synthetic control method: causal inference with interactive fixed effects models. Polit Anal. 2017;25(1):57-76. doi: 10.1017/pan.2016.2 [DOI] [Google Scholar]
- 23.Hartman E. Equivalence testing for regression discontinuity designs. Polit Anal. 2021;29(4):505-521. doi: 10.1017/pan.2020.43 [DOI] [Google Scholar]
- 24.Sun L, Abraham S. Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. J Econom. 2021;225(2):175-199. doi: 10.1016/j.jeconom.2020.09.006 [DOI] [Google Scholar]
- 25.Callaway B, Sant’Anna PH. Difference-in-differences with multiple time periods. J Econom. 2021;225(2):200-230. doi: 10.1016/j.jeconom.2020.12.001 [DOI] [Google Scholar]
- 26.Sun L. eventstudyinteract: interaction weighted estimator for event study. 2021. Accessed November 1, 2022. https://github.com/lsun20/eventstudyinteract
- 27.Kaiser Family Foundation . State Health Facts—Poverty Rate by Race/Ethnicity. 2022. Accessed November 7, 2022. https://www.kff.org/other/state-indicator/poverty-rate-by-raceethnicity/
- 28.Artiga S, Hill L, Orgera K, Damico A. Health coverage by race and ethnicity, 2010–2019. Kaiser Family Foundation . July 16, 2021. Accessed August 24, 2022. https://www.kff.org/racial-equity-and-health-policy/issue-brief/health-coverage-by-race-and-ethnicity/
- 29.Bhutta N, Chang AC, Dettling LJ, Hsu JW. Disparities in wealth by race and ethnicity in the 2019 Survey of Consumer Finances. FEDS Notes. Published online September 28, 2020. Accessed November 7, 2022. doi: 10.17016/2380-7172.2797 [DOI] [Google Scholar]
- 30.South E, Venkataramani A, Dalembert G. Building Black wealth—the role of health systems in closing the gap. N Engl J Med. 2022;387(9):844-849. doi: 10.1056/NEJMms2209521 [DOI] [PubMed] [Google Scholar]
- 31.Orgera K, Rudowitz R, Damico A. A Closer Look at the Remaining Uninsured Population Eligible for Medicaid and CHIP. Kaiser Family Foundation . November 18, 2021. Accessed August 24, 2022. https://www.kff.org/uninsured/issue-brief/a-closer-look-at-the-remaining-uninsured-population-eligible-for-medicaid-and-chip/
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods.
eTable 1. Analysis Sample Creation
eTable 2. Cross Validation Results Based on Mean Squared Prediction Errors
eTable 3. Placebo Tests and No Pretrend Tests
eTable 4. Two-Way Fixed Effects Regression Average Treatment Effect on Treated Estimates
eTable 5. Inverse Probability Tilting Average Treatment Effect on Treated Estimates (Based on Aggregate Effects).
eTable 6. Interaction Weighted Estimator Average Treatment Effect on Treated Estimates (Based on Aggregate Effects).
eReferences.
Data Sharing Statement