Abstract
Objective.
To examine whether housing instability - inclusive of eviction, homelessness, moving in with others, moving for cost reasons, or frequent moves - is associated with mental health among renters in the aftermath of the Great Recession of 2007–09.
Study Design.
A panel survey study.
Methods.
We used data from the Michigan Recession and Recovery Study (2009–10, 2011 and 2013), a population-representative sample of working-aged adults, and logistic regression with propensity score weights to examine the association between housing instability over a year and a half and anxiety attack or depression symptoms at follow up.
Results.
Respondents with any housing instability were 14 percentage points more likely to have had a recent anxiety attack, and those who had moved for cost reasons were 16 percentage points more likely. Respondents who experienced eviction were significantly more likely to meet criteria for major or minor depression at follow up, by 13 percentage points.
Conclusions.
Prior evidence of an association between housing instability and mental health is support by these findings, which are robust to potential confounders, including financial and life shocks, housing quality, and neighborhood poverty concentration.
Keywords: Housing instability, Involuntary moves, Anxiety, Depression, Propensity score weighting
INTRODUCTION
Adverse mental health implications of housing instability gained attention after the Great Recession of 2007–09.1,2 Involuntary relocation reduces feelings of personal control and increases stress,3,4 and moving may disrupt mental health-supportive social networks, instrumental supports, and access to institutions or resources.5,6 As the global COVID-19 pandemic and accompanying recession have amplified the threat of housing instability for renters across the U.S.,7 it is important to consider implications of the current recession for population mental health. This study addresses a gap in the literature by considering the various forms of housing instability collectively, evaluating other housing problems as alternative explanations, and considering confounding by sociodemographic position or shocks to finances or relationships.
Previous literature has shown poorer mental health among those with severe forms of housing loss, including homelessness,8 eviction,9 and frequent moves and moves made for cost reasons.10 By contrast, findings for the impact of doubling up on mental health have been mixed,11, 12 in part because intergenerational housing has benefits for older parents.13 However, very few studies have assessed the mental health impact of housing instability while considering its many forms. A comprehensive measure of housing instability is conceptually more appropriate than focusing on discrete events of a specific type (e.g., only eviction), as people may experience a complex, sequential pattern of multiple housing instability events over time14–16, or may be at risk of some, but not other types. Using a comprehensive measure of housing instability, we can more effectively distinguish the unstably housed from those who are not housing unstable in any way. One study that used a comprehensive measure of housing instability to capture recession-induced housing hardship for both homeowners and renters found a significant association between housing instability and mental health problems in the aftermath of the Great Recession.1 However it is unclear whether the previously demonstrated mental health consequences of foreclosure among homeowners drove this association,17,18 or if there are similar consequences for renters. U.S. housing policies treat homeowners and renters differently, making it critical to distinguish them in examining the mental health consequences of housing instability.
In addition, past studies rarely have accounted for other housing problems associated with poorer mental health, even though these are likely to be more common among those experiencing housing instability. Stressful and physically dangerous housing conditions include inadequate climate control, hazardous or degraded structural conditions, environmental pollutants, and vermin.8 A study of depression and generalized anxiety disorder among U.S. mothers in the Fragile Families and Well Being Study controlled for housing quality (i.e., peeling paint, holes in floor) and housing disarray (i.e., dark, crowded, noisy), but it solely focused on frequent moves, ignoring other types of housing instability.10 Studies on the impact of any form of housing instability on mental health have rarely controlled for community level conditions, such as neighborhood poverty, that could shape housing options and residents’ health, even net of their own characteristics and experiences. Involuntary moves may well correlate with both poor physical housing conditions and impoverished communities.19
Even after adjusting for housing quality problems or neighborhood disadvantage, housing instability may only appear to be linked to poorer mental health because of confounders. Housing instability could occur as a function of life or income shocks, including relationship disruption (e.g., partner loss), dramatic decline in income, or loss of housing assistance, and each of these shocks could also predict poorer mental health. The socio-demographic profiles of individuals could also predict both housing instability and poorer mental health. Many prior studies have adjusted for differences in educational attainment, income, or other dimensions of individual socioeconomic position (SEP).9,10 However, many have also used samples of persons who have all experienced housing problems or are at particularly high risk (e.g., studies of homeless individuals, studies using samples that over-represent socioeconomically less-advantaged people), restricting variation in individual SEP and limiting generalizability of findings.
We contribute to the evidence linking involuntary housing instability among renters to subsequent mental health in several ways. We use a measure that encompasses homelessness, eviction – including threatened, in progress, and completed eviction – moving in with others, frequent moves, and cost-related moves. We examine the link between housing instability and subsequent anxiety and depression in a population-based, representative sample of adult residents of the three counties surrounding Detroit who were renters at baseline in late 2009 or early 2010. Estimates adjust for substandard housing quality and neighborhood poverty at follow-up, and use regression-based and propensity score weighted models to address a range of potential confounders. This approach yields novel estimates of the connections between housing instability and changes in mental health over a period of 18 months in the wake of the Great Recession.
METHODS
Data
We use data from the Michigan Recession and Recovery Study (MRRS), a stratified random sample of adults 19–64 drawn from the general population of the three Detroit-area counties, Macomb, Oakland, and Wayne. Wave 1 interviews with 914 respondents, representing a response rate of 82.8%, were conducted between October 2009 and April 2010. We re-interviewed 847 respondents between April and August of 2011, a response rate of 94% of survivors. Among these, 751 took part in follow-up interviews between June and October 2013, a response rate of 90% of survivors from wave 2. This study draws on data from respondents who were renters across all three waves (N = 262). We excluded 7 respondents who were missing any of the variables of interest. To increase analytic power, we created up to two person-spells for each respondent, one including information from waves 1 and 2, and another including information from waves 2 and 3. Spells contain information on mental health at both the baseline and follow-up observations and on housing instability that occurred between them. This results in an analytic sample of 510 observations from 255 respondents.
Measures
Health Outcomes
At each wave, respondents were asked whether in the last 4 weeks they have had an anxiety attack—suddenly feeling fear or panic (yes = 1, no = 0). This item came from the PHQ-brief instrument, a validated scale.20 Depression (yes = 1, no = 0) in the last 2 weeks was assessed using the Patient Health Questionnaire, a validated nine-item scale based on the diagnostic criteria for major depressive disorder in the Diagnostic and Statistical Manual Fourth Edition.21
Housing-related variables
Our comprehensive measure of housing instability (yes = 1, no = 0) indicates reporting any of the following during the person-spell, as measured at the second observation and reported retrospectively: homelessness, eviction (threatened, in progress, or completed), having moved in with others to save on costs, having moved more than twice over the person-spell, or a cost-related move. Not all housing instability events in our comprehensive measure are equally severe. But each event in our comprehensive measure of housing instability is widely used as an indicator of housing instability.1,14
At the second observation of the person-spell, respondents reported on housing quality indicators (summed and categorized as none, 1–2 or 3+): leaky roof or ceiling; a toilet, hot-water heater, or other plumbing that does not work properly; rats, mice, roaches or other insects; broken windows; a malfunctioning heating system, stove, or refrigerator; exposed wires or other electrical problems; peeling paint; or other problem. We used 5-year estimates from the 2011 American Community Survey to measure whether respondents were residing in a high poverty census tract (>20 percent of families below the poverty line; 1 = yes, 0 = no) at the second observation of the spell.
Life or income shocks
We considered three types of shocks: loss of a partner, loss of housing assistance, or reduction in income. Partner loss includes those by death, divorce, or cessation of cohabitation over follow up. Housing assistance loss indicates respondents who reported receiving federal housing assistance at baseline but not receiving it at follow up. Income loss was measured by an income-to-needs ratio decline of more than 25% between observations in the person-spell. Income-to-needs ratio was calculated based on the federal poverty level by household size and the number of children.
Sociodemographic characteristics
All sociodemographic characteristics were measured at the second observation of the person-spell, including race (1 = Black, 0 = other), sex (1 = female), relationship status (1 = currently married or cohabiting), educational attainment (1 = some college or more, 0 = less), age (in years), presence of children in the home (1 = yes), and income-to-needs ratio.
Analytic Methods
We ran a logistic regression model for each mental health outcome with a propensity score weight, a method less sensitive to the misspecification of the functional forms of the association between a set of covariates and an outcome variable than a traditional regression model.22,23 Propensity score weighting rather than propensity score matching accommodated our relatively small sample size. Propensity score weighting achieves balance in the distribution of independent variables between the treatment group (experienced housing instability) and comparison group (housing-secure) by reweighting observations based on the probability of treatment assignment for each observation when considering a set of observed covariates. The inverse probability weighting approach applies the inverse of the probability of treatment assignment for those cases in the treatment group and the inverse of 1 minus the probability of treatment assignment for those in the comparison group.23 We calculated the probability of treatment assignment using the generalized boosted regression approach, rather than the more commonly used logistic regression approach, because the former is a nonparametric approach and robust to misspecification of the functional relationship between a set of covariates and the outcome variable.24,25
We also address a prominent alternative explanation for the association of housing instability and subsequent mental health, which suggests that those experiencing housing instability already had a higher rate of mental health problems at the start of the person-spell. Using a lagged dependent variable to capture earlier mental health in regression analysis could result in biased estimators because we include multiple observations for each individual.26,27 Thus, we have included mental health at the baseline of the person spell in the model that predicts the probability of treatment assignment.26,27 We addressed other potential confounders using the same approach.
We examined the characteristics of our analytic sample overall, then tested for differences by housing instability status over follow up, using t-tests. We conducted bivariate analyses of covariates and two mental health outcomes. Next, we estimated logistic regression models weighted by propensity scores for each health outcome using the comprehensive measure of housing instability. We then considered each type of housing instability individually to assess whether the comprehensive measure disguised heterogeneous associations. To examine the association for each individual type of housing instability, while ensuring that we were comparing to those with no experience of housing instability over follow up, we created a three-category variable for each type: experienced the focal type, experienced other type, or did not experience any type (reference group).
RESULTS
Table 1 presents the characteristics of our analytic sample, overall and stratified by experience of housing instability during the person-spell. Those who had experienced housing instability were more likely to report anxiety and depression at both the follow-up (31% versus 12% and 25% versus 16% respectively) and the baseline (35% versus 19% and 31% versus 14% respectively). In the sample overall, 34% reported housing instability over follow up, including 13% who experienced more than one form. These were distributed among moving for cost (43%), multiple moves (42%), doubling up (33%) or experiencing eviction (32%) and homelessness (11%). More than one in four respondents overall had a drop in income to needs ratio of more than 25% over follow up, but only about 3% had lost housing assistance. At follow-up, those who had experienced housing instability were significantly more likely to have experienced a partner loss (13% versus 2%) and one or two housing quality problem (38% versus 24%), but less likely to have three or more (11% versus 17%), and were more likely to be living in a high poverty census tract (42% versus 30%).
Table 1.
Characteristics of the Pooled Analytic Sample from the MRRS, overall and stratified by housing instability events over follow up
| Overall | Housing instability events over follow up |
||
|---|---|---|---|
| Yes | No | ||
|
| |||
| Number of observations | 510 | 191 | 319 |
|
| |||
| Mental health outcomes at follow up of spell, % (SE) | |||
| Anxiety | 18.2 (3.2) | 30.7 (2.8)** | 11.6 (2.2)** |
| Depression | 18.8 (2.6) | 25.1 (3.8)* | 15.5 (2.4)* |
| Mental health at baseline of spell, % (SE) | |||
| Anxiety | 24.1 (4.5) | 34.8 (3.2)** | 18.6 (4.4)** |
| Depression | 20.0 (2.3) | 30.6 (3.2)** | 14.4 (2.4)** |
| Housing instability event over follow up, % (SE) | 34.4 (3.7) | – | – |
| Eviction | 10.9 (2.0) | 31.6 (5.0) | – |
| Homelessness | 3.9 (1.1) | 11.4 (3.0) | – |
| Moving in with others | 11.2 (2.7) | 32.7 (5.7) | – |
| Moved for cost | 14.6 (2.6) | 42.5 (6.6) | – |
| Multiple moves | 14.4 (1.8) | 41.9 (4.2) | – |
| Life or income shocks over follow up, % (SE) | |||
| More than 25% decrease in income-to-needs ratio | 28.1 (3.3) | 30.9 (4.8) | 26.6 (3.1) |
| Lost housing assistance | 2.6 (0.9) | 2.8 (1.3) | 2.4 (1.0) |
| Partner loss | 5.7 (1.5) | 12.5 (3.4)** | 2.1 (0.9)** |
| Housing problems at follow up, % (SE) | |||
| Number of substandard housing problems | |||
| None | 56.6 (3.1) | 51.2 (5.4)* | 59.4 (4.0)* |
| One or two | 28.6 (3.2) | 37.8 (5.1)* | 23.8 (3.4)* |
| Three or more | 14.8 (2.0) | 11.0 (2.5)* | 16.8 (3.3)* |
| Living in a high poverty tract (>20%) | 33.7 (5.8) | 41.8 (7.0)* | 29.5 (5.9)* |
| Demographic & SES characteristics at follow up | |||
| Black, % (SE) | 51.7 (5.1) | 62.4 (8.9) | 46.1 (5.5) |
| Female, % (SE) | 58.4 (5.1) | 65.7 (6.1) | 54.6 (8.3) |
| Age in years, mean (SE) | 39.5 (1.2) | 37.2 (1.3) | 40.7 (1.3) |
| Married or cohabiting, % (SE) | 41.7 (4.8) | 39.0 (6.2) | 43.2 (5.8) |
| More than high school education, % (SE) | 57.1 (6.4) | 49.1 (7.1) | 61.2 (7.3) |
| Child presence, % (SE) | 52.3 (5.3) | 61.0 (7.2) | 47.8 (6.0) |
| Income-to-needs ratio, mean (SE) | 2.3 (0.3) | 2.1 (0.4) | 2.3 (0.3) |
Note 1. Housing instability events over follow up capture whether respondents had experienced any of the following between survey waves: eviction (including eviction threat, and eviction being in progress), homelessness, moving-in with others, frequent moves (more than twice over follow up), or cost-related move.
Note 2. Using t-tests (f-test for number of substandard housing problems), we examined whether the distribution of covariates significantly differs between respondents who experienced housing instability over follow-up and those who did not.
p<0.001
p<0.01
p<0.05
Table 2 presents results from the bivariate analyses of covariates and the two mental health outcomes. The first column shows coefficients and standard errors for bivariate models predicting anxiety attack. Respondents who reported baseline anxiety are more likely to report anxiety at follow up. Also, respondents who experienced at least one type of housing instability were more likely to report an anxiety attack at follow up. When considering specific types of housing instability, we find that respondents who experienced eviction, homelessness, or moved for cost reasons over follow up were more likely to report a recent anxiety attack at follow up. Respondents who experienced an income shock over follow up were more likely to report a recent anxiety attack at follow up, while those with a higher income-to-needs ratio were less likely than those with fewer income resources.
Table 2.
Bivariate Analysis of Covariates and Two Health Outcomes
| Anxiety attack | Depression | |||
|---|---|---|---|---|
|
| ||||
| Coef. | SE | Coef. | SE | |
|
| ||||
| Focal mental health at baseline of spell | 2.32*** | .36 | 1.11* | .49 |
| Housing instability over follow up | 1.21** | .38 | .60* | .26 |
| Housing instability event over follow up (ref = respondents w/o any housing instability events) | ||||
| Eviction | 1.64*** | .38 | 1.13** | .31 |
| Homelessness | 1.19* | .52 | 1.16+ | .56 |
| Moving in with others | 1.24+ | .65 | .21 | .42 |
| Moved for cost | 1.28*** | .32 | .63 | .39 |
| Multiple moves | .70 | .43 | .75* | .31 |
| Life or income shocks over follow up | ||||
| More than 25% decrease in income-to-needs ratio | .79* | .32 | .39 | .30 |
| Lost housing assistance | 1.15* | .50 | 1.28* | .58 |
| Partner loss | .86 | .59 | .01 | .39 |
| Housing problems at follow up | ||||
| Number of substandard housing problems (ref = None) | ||||
| One or two | .58 | .37 | −.16 | .42 |
| Three or more | .39 | .53 | .27 | .35 |
| Living in a high poverty tract (>20%) | .38 | .40 | 1.41** | .43 |
| Demographic & SES characteristics at follow up | ||||
| Black | .02 | .42 | .84* | .38 |
| Female | .86 | .51 | .47 | .35 |
| Age in years | .01 | .02 | .05*** | .01 |
| Married or cohabiting | −.29 | .36 | −.85+ | .42 |
| More than high school education | −.55 | .48 | −.97+ | .53 |
| Child presence | .28 | .37 | −.38 | .36 |
| Income-to-needs ratio | −.26** | .09 | −.55*** | .13 |
Note. ref = reference group; Coef. = Coefficient; SE = standard error.
p<0.001
p<0.01
p<0.05
p<.1
The second column of values in Table 2 shows coefficients and standard errors for bivariate models predicting depression. Similar to the results for anxiety attack, respondents who reported baseline depression were more likely to meet criteria for major or minor depression at follow up. Considering specific types of housing instability shows that respondents who experienced eviction or multiple moves over follow up were more likely meet criteria for depression at follow up. Additional predictors of depression at follow up include loss of housing assistance over follow up, living in a high poverty tract, Black race, older age, and a lower income-to-needs ratio.
Table 3 shows results from 12 separate logistic (for any housing instability) or multinomial logistic regression models (for individual types of housing instability) weighted by propensity scores. This weighting addresses the influence of differences in demographic and socioeconomic characteristics, life or income shocks over follow up, number of substandard housing problems and living in high poverty neighborhood at follow up, and focal mental health condition at baseline of person-spell. The first column of values in Table 3 shows coefficients, standard errors, and average marginal effects for models predicting anxiety attack. Respondents who experienced any housing instability over follow up were 14 percentage points more likely to have had an anxiety attack in the 4 weeks prior to the follow up interview. Moves for cost reasons were a significant predictor of anxiety attack, with all other individual types of housing instability except homelessness showing positive associations of varying strength. The second column of results shows that respondents who experienced eviction over follow up were more likely to meet criteria for depression at follow up by 13 percentage points. All other specific types of housing instability, and the overall measure, show positive but insignificant associations with depression.
Table 3.
Coefficients, Standard Errors, and Average Marginal Effect from propensity score weighted logistic regression models predicting mental health outcomes among renters
| Anxiety attack | Depression | |||||
|---|---|---|---|---|---|---|
|
|
|
|||||
| Coef. | SE | AME | Coef. | SE | AME | |
|
| ||||||
| Housing instability over follow up | .92* | .36 | .14 | .38 | .32 | .06 |
| Types of housing instability over follow up (ref = respondents w/o any housing instability events) | ||||||
| Eviction | .60+ | .33 | .09 | .65* | .30 | .13 |
| Homelessness | −.14 | .55 | −.02 | .30 | .55 | .05 |
| Moving in with others | .28 | .43 | .04 | .32 | .40 | .06 |
| Moved for cost | .93** | .32 | .16 | .38 | .33 | .07 |
| Multiple moves | .31 | .36 | .04 | .55+ | .32 | .11 |
| Person-spells | 510 | 510 | ||||
Note. Results represent odds ratio from 12 separate logistic regression models weighted by propensity scores to balance the distribution of covariates, including demographic and socioeconomic characteristics, life or income shocks over follow up, housing problems at follow up (number of substandard housing problems and living in high poverty neighborhood), focal mental health condition at baseline of person-spell, and an indicator of person spell. To examine the association for each type of housing instability, we created a three category variable: experienced the focal type of housing instability, experienced other type of housing instability (coefficients not shown), or did not experience any type of housing instability (reference group). ref = reference group; Coef. = Coefficient; SE = standard error; AME = average marginal effect.
p<0.001
p<0.01
p<0.05
p<.1
We conducted several checks for robustness of these results (not shown but available). In one, we substituted measures of competing housing problems (housing quality problems, high poverty census tract) measured at the start of the person-spell, rather than at the end. Adjusting across treatment and control groups for these other housing problems at follow up helps to isolate the focal housing instability experience from the conditions in the new residence. Adjusting instead for prior housing-related problems helps eliminate potential historical impacts of prior poor living conditions. We also conducted the analysis without any adjustment for housing quality or neighborhood levels of poverty. All these scenarios produced similar results.
DISCUSSION
This study explored the mental health consequences renters faced in the aftermath of the Great Recession. Using a population-representative sample of working-aged adults in a diverse metropolitan area in Michigan, we examined the association of housing instability and mental health among renters while addressing potential confounders, including financial and life shocks, housing quality, and neighborhood poverty concentration. Our findings raise concern as eviction moratoriums that were instituted to support social distancing and prevent COVID-19 are receding, even as economic recovery appears to lag.
About one in three renters in our sample experienced housing instability between 2009–2013. Our study does not measure the effect of the Great Recession on housing instability. But our findings complement past work finding that rent burden28 and doubling up28,29 increased in the aftermath of the Great Recession compared to the years immediately before it and that the number of extremely low-income U.S. households per affordable rental unit increased by 17 percent in 2007–2010.29
Our results from propensity score models suggest that respondents who experienced housing instability over follow up were more likely to report a recent anxiety attack at follow up. We addressed a limitation of prior studies of this association - most did not consider the influence of housing problems (other than instability itself)1,10 that have been shown to increase when people experience housing instability, for example, with involuntary moves.19 To address this alternative explanation, we adjusted for post-move housing quality, poverty concentration in the neighborhood, and life events that could predict both housing instability and subsequent mental health. When we disaggregated the comprehensive measure into specific types of housing instability over follow up, however, we only found a significant association with anxiety for cost-related moves. This might be primarily attributed to the limited statistical power of our small sample, especially for the relatively rare, extreme forms of housing instability. Future work also should assess whether the measurement of some types of housing instability is conceptually unrefined due to factors like heterogeneity in the types of doubled-up households.30
Our results show that the association between any housing instability and subsequent symptoms of depression weakened considerably after controlling for symptoms at the baseline wave. This difference from the most comparable, population-based study with a comprehensive measure of housing instability1 may reflect the fact that the earlier study only captured prior depression symptoms from more than four years prior to the measurement of housing instability. We may have been better able to capture cases of mental health predicting subsequent housing troubles, or our study may suffer from insufficient power due to the size of the sample. A larger data source could address this limitation, though such sources appear to be rare. For specific types of housing instability over follow up, we found significant associations with depression for eviction and also (marginally) for multiple moves. This suggests that the previously documented association of housing instability and depression in a more restricted sample of economically vulnerable people9,10 is more widely present.
Our analysis has limitations that should be noted. Even using the methods applied here, unmeasured community-level factors associated with both housing instability and mental health could influence the association between the two. Given the age and geographic parameters of our sample, our results are not broadly generalizable. Future studies could use broader regional or national data. Also, our relatively small sample size limits the power of statistical analyses, but few data sources that provide detailed information on a comprehensive range of housing instability types, as well as mental health, are available. Lastly, findings from the aftermath of the Great Recession may not be representative of those prevailing under all macroeconomic conditions. Nonetheless, they may be particularly relevant to the conditions in the wake of the COVID-19 economic downturn.
This study has several implications for public health, housing policy, and health services. First, it suggests that just addressing an income loss or the poor quality of housing stock do not necessarily remove the risks associated with housing instability. Future work could look at the mechanisms by which different types of housing instability influence mental health, as well as seeking to identify formal or informal supports that could buffer against these effects.
Second, findings call for the continuation of policies to reduce involuntary moves among renters to protect mental health.31 Even prior to the COVID-19 pandemic, housing needs had long outstripped the constraints limited funding imposes on federal housing programs and income support programs. Median income growth among renters has lagged behind the median rent increase for decades and more than half of low-income renters are severely housing cost burdened and at risk of involuntary moves.32 Past research shows housing assistance and unemployment insurance reduce housing instability.14,33 But only one in four income-eligible families receive federal low-income housing assistance32 while a large amount of federal housing assistance goes to the higher end of income distribution through the mortgage interest deduction.34,35 It is critical to consider ways to restructure the currently inequitable allocation of federal housing resources. Building on the public health tradition of examining the health implications of social policies within a social determinants of health framework,8,36 we have found that recession-induced housing instability could negatively affect population-level mental health. Just as public health scholars have shaped discussions on the eviction moratorium during this pandemic, future research on housing policy, housing instability, and health can provide a venue for public health scholars to support critical evidence-based policy making to improve population health.
Author Statements
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Data collection for this study was supported by funds provided to the National Poverty Center (NPC) by the Office of the Assistant Secretary for Planning and Evaluation at the U.S. Department of Health and Human Services, the Office of the Vice-President for Research at the University of Michigan, the John D. and Catherine T. MacArthur Foundation, and the Ford Foundation. This study was approved by the Institutional Review Board (IRB) of University of Michigan, Ann Arbor.
Table A1.
Covariate Balancing in Propensity Score Weighted Sample
| Anxiety attack | Depression | |||||
|---|---|---|---|---|---|---|
|
|
|
|||||
| Housing instability events over follow up | Standardized mean difference | Housing instability events over follow up |
Standardized mean difference | |||
|
|
|
|||||
| Yes | No | Yes | No | |||
|
| ||||||
| Focal Mental health at baseline of spell | .29 | .21 | .19 | .25 | .19 | .13 |
| Life or income shocks over follow up | ||||||
| More than 25% decrease in income-to-needs ratio | .29 | .27 | .05 | .30 | .28 | .04 |
| Lost housing assistance | .03 | .03 | .02 | .03 | .03 | −.00 |
| Partner loss | .08 | .03 | .18 | .07 | .04 | .11 |
| Housing problems at follow up | ||||||
| Number of substandard housing problems | .96 | 1.08 | −.07 | .96 | 1.08 | −.07 |
| Living in a high poverty tract (>20%) | .42 | .34 | .17 | .40 | .34 | .13 |
| Demographic & SES characteristics at follow up | ||||||
| Black | .63 | .52 | .21 | .60 | .52 | .18 |
| Female | .63 | .57 | .11 | .63 | .57 | .11 |
| Age in years | 39.02 | 40.16 | −.09 | 38.65 | 40.08 | −.11 |
| Married or cohabiting | .39 | .42 | −.06 | .41 | .42 | −.02 |
| More than high school education | .53 | .58 | −.11 | .53 | .58 | −.10 |
| Child presence | .56 | .51 | .10 | .57 | .51 | .11 |
| Income-to-needs ratio | 2.04 | 2.21 | −.07 | 2.13 | 2.22 | −.04 |
| Follow-up spell | .53 | .48 | .10 | .53 | .47 | .11 |
| Number of person-spells | 191 | 319 | 191 | 319 | ||
Note 1. The standardized mean difference was calculated by dividing the difference in the means of covariates between respondents having experienced housing instability and those who have not by the standard deviation of the covariates.
Note 2. Since focal mental health at baseline of spell is different for each outcome, we created propensity scores separately for each outcome. We also conducted separate covariate balancing for a series of analyses using each type of housing instability as a key predictor (not presented here, but available upon request). The distribution of a few covariates still remains unbalanced in the propensity score weighted sample, but including those remaining unbalanced covariates in the main analytic models predicting health outcomes did not change our substantive findings.
p<0.001
p<0.01
p<0.05
REFERENCES
- 1.Forbes MK, Krueger RF. The great recession and mental health in the United States. Clinical Psychological Science. 2019;7(5):900–913. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Burgard SA, Seefeldt KS, Zelner S. Housing instability and health: findings from the Michigan Recession and Recovery Study. Social science & medicine. 2012;75(12):2215–2224. [DOI] [PubMed] [Google Scholar]
- 3.Eriksen MD, Ross A. The impact of housing vouchers on mobility and neighborhood attributes. Real Estate Economics. 2013;41(2):255–277. [Google Scholar]
- 4.Nettleton S, Burrows R. Mortgage debt, insecure home ownership and health: an exploratory analysis. Sociology of health & Illness. 1998;20(5):731–753. [Google Scholar]
- 5.Greenbaum S, Hathaway W, Rodriguez C, Spalding A, Ward B. Deconcentration and social capital: Contradictions of a poverty alleviation policy. Journal of Poverty. 2008;12(2):201–228. [Google Scholar]
- 6.Keene DE, Geronimus AT. “Weathering” HOPE VI: the importance of evaluating the population health impact of public housing demolition and displacement. Journal of Urban Health. 2011;88(3):417–435. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Aspen Institute. The COVID-19 Eviction Crisis: An Estimated 30–40 Million People in America at Risk. 2020.
- 8.Shaw M Housing and public health. Annu Rev Public Health. 2004;25:397–418. [DOI] [PubMed] [Google Scholar]
- 9.Desmond M, Kimbro RT. Eviction’s Fallout: Housing, Hardship, and Health. Social Forces. 2015;94(1):295–324. [Google Scholar]
- 10.Suglia SF, Duarte CS, Sandel MT. Housing Quality, Housing Instability, and Maternal Mental Health. J Urban Health. 2011;88(6):1105–1116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Evans GW, Saegert S, Harris R. Residential Density and Psychological Health among Children in Low-Income Families. Environment and Behavior. 2001;33(2):165–180. [Google Scholar]
- 12.He Y, O’Flaherty B, Rosenheck RA. Is shared housing a way to reduce homelessness? The effect of household arrangements on formerly homeless people. Journal of Housing Economics. 2010;19(1):1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Aranda L Doubling up: A gift or a shame? Intergenerational households and parental depression of older Europeans. Social Science & Medicine. 2015;134:12–22. [DOI] [PubMed] [Google Scholar]
- 14.Kim H, Burgard SA, Seefeldt KS. Housing assistance and housing insecurity: A study of renters in southeastern Michigan in the wake of the great recession. Social Service Review. 2017;91(1):41–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Sosin M, Piliavin I, Westerfelt H. Toward a longitudinal analysis of homelessness. Journal of Social Issues. 1990;46(4):157–174. [Google Scholar]
- 16.Kleit RG, Kang S, Scally CP. Why do housing mobility programs fail in moving households to better neighborhoods? Housing Policy Debate. 2016;26(1):188–209. [Google Scholar]
- 17.McLaughlin K, Nandi A, Keyes K, et al. Home foreclosure and risk of psychiatric morbidity during the recent financial crisis. Psychological Medicine. 2012;42(7):1441. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Osypuk T, Caldwell C, Platt R, Misra D. The consequences of foreclosure for depressive symptomatology. Annals of Epidemiology. 2012;22(6):379–387. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Desmond M, Shollenberger T. Forced displacement from rental housing: Prevalence and neighborhood consequences. Demography. 2015;52(5):1751–1772. [DOI] [PubMed] [Google Scholar]
- 20.Löwe B, Gräfe K, Zipfel S, et al. Detecting panic disorder in medical and psychosomatic outpatients: comparative validation of the Hospital Anxiety and Depression Scale, the Patient Health Questionnaire, a screening question, and physicians’ diagnosis. Journal of psychosomatic research. 2003;55(6):515–519. [DOI] [PubMed] [Google Scholar]
- 21.Kroenke K, Spitzer R, Williams J. The PHQ-9: validity of a brief depression severity measure. The Journal of General Internal Medicine. 2001;16:606–613. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Li F, Morgan KL, Zaslavsky AM. Balancing covariates via propensity score weighting. Journal of the American Statistical Association. 2018;113(521):390–400. [Google Scholar]
- 23.Morgan SL, Winship C. Counterfactuals and causal inference. Cambridge University Press; 2015. [Google Scholar]
- 24.Lee BK, Lessler J, Stuart EA. Improving propensity score weighting using machine learning. Statistics in medicine. 2010;29(3):337–346. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.McCaffrey DF, Ridgeway G, Morral AR. Propensity score estimation with boosted regression for evaluating causal effects in observational studies. Psychological methods. 2004;9(4):403. [DOI] [PubMed] [Google Scholar]
- 26.Dafoe A Nonparametric identification of causal effects under temporal dependence. Sociological Methods & Research. 2018;47(2):136–168. [Google Scholar]
- 27.Bellemare MF, Masaki T, Pepinsky TB. Lagged explanatory variables and the estimation of causal effect. The Journal of Politics. 2017;79(3):949–963. [Google Scholar]
- 28.Colburn G, Allen R. Rent burden and the Great Recession in the USA. Urban Studies. 2018;55(1):226–243. [Google Scholar]
- 29.Lens MC. Extremely low-income households, housing affordability and the Great Recession. Urban Studies. 2018;55(8):1615–1635. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Harvey H Cumulative effects of doubling up in childhood on young adult outcomes. Demography. 2020:1–28. [DOI] [PubMed] [Google Scholar]
- 31.Benfer EA, Vlahov D, Long MY, et al. Eviction, health inequity, and the spread of covid-19: Housing policy as a primary pandemic mitigation strategy. Journal of Urban Health. 2021;98(1):1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Joint Center for Housing Studies. America’s Rental Housing 2020. 2020.
- 33.Berkowitz SA, Basu S. Unmet Social Needs And Worse Mental Health After Expiration Of COVID-19 Federal Pandemic Unemployment Compensation: Study examines unmet social needs and mental health after Federal Pandemic Unemployment Compensation expired. Health Affairs. 2021;40(3):426–434. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Infranca JJ. Housing resource bundles: Distributive justice and federal low-income housing policy. University of Richmond Law Review. 2014;49:1071–1137. [Google Scholar]
- 35.Dreier P Federal housing subsidies: Who benefits and why. In: Bratt R, Stone M, Hartman C, eds. A right to housing: Foundation for a new social agenda. Philadelphia: Temple University Press; 2006:105–138. [Google Scholar]
- 36.Krieger J, Higgins DL. Housing and health: time again for public health action. American journal of public health. 2002;92(5):758–768. [DOI] [PMC free article] [PubMed] [Google Scholar]