State-Based Eviction Moratoria and Child Maltreatment During the COVID-19 Pandemic

Child maltreatment is a pressing public health issue with detrimental consequences.¹ About 3.1 million children received Child Protective Services (CPS) investigations in 2022.² Eviction and housing insecurity elevate the risk of child maltreatment and CPS investigations.^3–5 A large proportion of maltreatment allegations, particularly for neglect, involve housing and material hardship, which may in part reflect structural adversities beyond a family’s control.^6,7 Nonetheless, most states’ legal statutes consider a family’s failure to provide secure housing to be child neglect.⁸ In 2022, inadequate housing was associated with about 8% of substantiated cases² and 11% of child removals in the child welfare system.⁹ Eviction and housing instability may also increase maltreatment risk by affecting caregivers’ stress, mental health, and parenting behaviors.⁵ 

Although not all child maltreatment is preventable, promoting housing security through policy interventions may effectively reduce maltreatment risk. Existing housing and renter support programs, such as the Keeping Families Together program in New York City and Connecticut’s Supportive Housing for Families, have demonstrated promise in improving child and family well-being by ensuring secure housing.^10–13 While experimental studies offer robust causal inference, they are often costly and challenging to scale. Natural experiments, when combined with appropriate statistical methodologies, offer a cost-effective means of examining causal inference using secondary data.¹⁴ However, research on the impact of housing policies on child maltreatment has been limited due to challenges in obtaining comprehensive data on policy variation and child maltreatment.

The COVID-19 pandemic compounded risks of child maltreatment by exacerbating employment and housing insecurity, particularly for low-income families and communities of color who already experience high rates of housing insecurity and CPS involvement prior to the pandemic.^15,16 In response to housing security concerns, policymakers issued eviction moratoria and renter-supportive measures at a national scale for the first time to help prevent the spread of COVID-19 while allowing families impacted by the economic crisis to retain housing. The federal Coronavirus Aid, Relief, and Economic Security (CARES) Act and the Centers for Disease Control and Prevention (CDC)’s eviction moratoria order temporarily halted evictions nationwide and authorized states to implement their own eviction moratoria and housing-related support programs, leading to state variations in implementation.¹⁷ Early evidence shows that county- and state-based eviction moratoria succeeded in slowing COVID-19 transmission, reduced eviction filing, and reduced mental distress,^18–21 suggesting that such supports may have also prevented child maltreatment by safeguarding families’ material needs and reducing stress. Leveraging variation in state-based eviction moratoria, we examine their potential effects on CPS report rates with focus on maltreatment subtypes, including physical abuse, sexual abuse, and neglect, using quasi-experimental difference-in-differences (DiD) methods.

We linked nationwide data from multiple sources aggregated by state, county, and/or time, spanning January 2019 through August 2021. State eviction moratoria went into effect as early as March 2020. We thus included child maltreatment data prior to moratoria implementation to assess the influence of such moratoria on changes over time in CPS report rates across counties that implemented moratoria relative to those that did not. In addition, for many states (eg, Kansas), eviction moratorium status fluctuated over time, as shown in Figure S1 in the Supplemental Material. To ensure the examination of consistent policy exposure, we restricted our treatment group to counties in states that continuously had eviction moratoria in effect (“always treated”) from mid-March 2020 to mid-August 2021 (the treatment period). These states included California, Hawaii, Illinois, Minnesota, Maryland, New Jersey, New Mexico, New York, and Washington, along with Washington DC. Similarly, the control group comprised counties in states that never implemented eviction moratoria (“never treated”) during this period, including Arkansas, Georgia, Missouri, Ohio, Oklahoma, South Dakota, and Wyoming. This approach allowed for an equal amount of treatment time for all counties in the treatment group.

Our child maltreatment data were drawn from the National Child Abuse and Neglect Data System (NCANDS), a national administrative data system encompassing case-level data on all child maltreatment investigations and child welfare services in the United States. NCANDS provides unique state and county identifiers—federal information processing standard codes—for all reports, allowing the calculation of the total number of county-level maltreatment reports as well as linkage to other data sources at the state- and county-level. NCANDS suppresses identifiers for counties with fewer than 1000 reports for confidentiality protection. This restriction implies that our results, like those from all county-level studies using NCANDS data, may not generalize to small counties.

Our eviction moratoria data are drawn from the COVID-19 Eviction Moratoria and Housing Policy Database (EMHPD) housed at the Inter-university Consortium for Political and Social Research.²² EMHPD is a policy surveillance database that includes information on state housing-related responses to the pandemic, such as eviction moratoria and renter-supportive measures in all 50 states between March 13, 2020 and March 1, 2022. Following the EMHPD’s suggestion, we used measures describing critical aspects of eviction moratoria uniformly across states. The CARES Act eviction moratorium, in effect from March 27 to August 23, 2020, was followed by the CDC-issued federal moratorium from September 4, 2020 to August 26, 2021. Combined, the federal-level eviction moratorium, lasting from March 2020 to August 26, 2021, overlapped with our study period for state-level moratoria. The state-based moratorium had broader protections and stronger enforcement while the federal-level moratorium provided a standardized baseline to prevent evictions. Because the federal moratorium did not vary by state, we excluded it from our analyses. Further, we could not find comprehensive county-level moratoria data, and the available city-level data covered only large cities,²³ making them incompatible with our county-level child maltreatment data. We used Johns Hopkins Center for Systems Science and Engineering COVID-19 time series data to address state variation in COVID-19 contexts.

We calculated biweekly counts of all screened-in CPS reports between January 2019 and mid-August 2021 at the county level. NCANDS rounded the month and day of the maltreatment report to either the eighth or the 23rd of the month for confidentiality. We thus calculated maltreatment report rates per 1000 children in the county on a biweekly basis. We further calculated rates by maltreatment type including physical abuse, sexual abuse, and neglect. The subtypes were not mutually exclusive because reports often involve multiple maltreatment types. We computed the individual type as “any inclusion” of that type.

our primary “treatment” indicates whether a state-level moratorium was in place in each county during a biweekly period.

In addition to eviction moratoria, some states also implemented utility moratoria that banned utility shutoffs during the pandemic, which we included as a time-varying control variable. Also, we adjusted for ongoing utility disconnection bans from a prepandemic federal program that allows states to halt utility disconnection under certain weather conditions and timeframes.²⁴ State-level rates of accumulated COVID-19 cases and deaths were included for each biweekly period to account for the potential impact of the pandemic on eviction and child maltreatment rates as well as to serve as proxies for health care system disruptions. Additionally, we controlled for state-level emergency rental assistance by including the number of recipient households divided by the total renting households per state and the rental assistance amount relative to the total renting households from January to August 2021.

We employed a DiD strategy to analyze the effects of eviction moratoria on child maltreatment rates by comparing average biweekly maltreatment rate changes over time between counties with and without moratoria.²⁵ We employed a county-level DiD approach comparing pre- and postpolicy maltreatment rates in counties with eviction moratorium vs those without. Although examining state-level policy effects on both state- and county-level child maltreatment rates may lead to qualitatively similar results at both levels, using county-level data allows us to account for the extent to which the policy impact on child maltreatment may have varied across counties within states that adopted the policy. Using county-level data also allows for sufficient sample sizes to achieve the statistical power for subgroup analyses. Because the study focused on rates of child maltreatment rather than case counts, linear regression was chosen over Poisson regression to provide more straightforward interpretations of the findings.

We first tested the DiD assumption of common maltreatment rate trends between counties with and without moratoria by estimating an event study model focusing on the interaction between the treatment group (counties with moratoria) and time compared with a reference prepolicy time point. Because the policy was implemented in the second biweekly period of March 2020, the reference period was the first biweekly period of March 2020, the period before policy implementation. The assumption that prepolicy trends in the treatment and control groups were parallel is supported if the prepolicy event study coefficients are not significantly different from 0.²⁶ 

Of the 318 counties in 17 states analyzed (including Washington, DC), 185 counties in 10 states consistently instituted state-level eviction moratoria over the study period from mid-March 2020 to mid-August 2022, as shown in Figure S1 in the Supplemental Material.

Figure 1 illustrates biweekly regression-adjusted differences in total child maltreatment, physical abuse, sexual abuse, and neglect report rates in counties with and without eviction moratoria compared with the biweekly period before policy implementation. We present point estimates with 95% CIs from the event study analysis. Overall, the prepolicy event study coefficients are not significantly different from 0 for total child maltreatment and all types of report rates, supporting the assumption that prepolicy trends in the treatment and control groups are parallel.

Table 1 displays the estimated effects of eviction moratoria on child maltreatment rates from our primary DiD models. The analysis compares changes in maltreatment rates between counties that implemented eviction moratoria and those that did not during the periods before and after moratoria were enacted. The first row presents the coefficients for eviction moratoria, which show that eviction moratoria were significantly associated with reduced biweekly reports of physical abuse (b [coefficient estimate] = −0.073; 95% CI, −0.119 to −0.027), sexual abuse (b = −0.034; 95% CI, −0.051 to −0.018), and neglect (b = −0.217; 95% CI, −0.346 to −0.088). Eviction moratoria were negatively associated with overall child maltreatment report rates, but the coefficient was not statistically significant. The second row presents the average county-level pre-eviction moratoria rate, which we use to compute the effect size (percent change) associated with each coefficient by dividing the coefficient by the estimated prepolicy maltreatment rate in counties with eviction moratoria. Effect sizes are presented in the third row. Results indicate that the effect sizes of eviction moratoria on maltreatment report rates were relatively large, reducing physical abuse rates by 16.04%, sexual abuse rates by 21.12%, and neglect rates by 12.17%.

Leveraging state variation in the implementation and expiration of eviction moratoria as a natural experiment, we found that the implementation of eviction moratoria was associated with reduced rates of child neglect and physical and sexual abuse reports. As the first study to investigate the impact of eviction moratoria on CPS reports in the United States, our findings contribute to a growing body of evidence on public health implications of such policies. Prior research has demonstrated moratoria policies’ efficacy in reducing COVID-19 infections and deaths, eviction filings, and mental distress among renters.^18–21 Our study builds upon this foundation by examining their impact on preventing child maltreatment as measured by CPS involvement.

More broadly, our findings align with evidence of the positive effects of housing support programs in mitigating child maltreatment risk.^10–13 Previous studies have highlighted the link between housing insecurity—manifested in home foreclosure and eviction filings—and increased child maltreatment reports.^3–5 Collectively, these findings suggest that eviction moratoria may be effective in reducing CPS involvement.

This study has several limitations that should be considered in contextualizing its findings. First, we relied on child maltreatment reports to state agencies, which may not fully capture actual child maltreatment because many instances of maltreatment go unreported, potentially due to lack of visibility to mandatory reporters. As such, our results may not necessarily translate to true reductions in maltreatment. Nonetheless, CPS involvement is consequential for families and highly policy relevant. Second, eviction prevention policies during the pandemic were (and are) complex and vary by state and county, and many states implemented additional renter-supportive measures such as providing grace periods for rent payments. Future research should explore potential interactions among eviction moratoria and other factors to understand their collective impact on child maltreatment. Third, this study did not account for other COVID-19-related federal and state policies, such as expanded unemployment insurance and child tax credit, which may also impact child maltreatment risk; however, we have no reason to suspect that the effects of such policies would have varied in states with and without eviction moratoria. Additionally, schools’ instructional modes (in person, virtual, or hybrid) had substantial impacts on maltreatment reporting. Because school modes were operationalized at the school district level rather than at the county level, it was challenging to aggregate data at consistent geographic levels, and our study is silent on the potential effects of school mode. Finally, although we used a natural experiment approach to produce plausibly causal estimates, threats to causal inference remain. Future research should incorporate these policy elements to provide a more comprehensive assessment of their impact on child welfare system involvement.

Economic support programs, including the Supplemental Nutrition Assistance Program; the Special Supplemental Nutrition Program for Women, Infants, and Children program; minimum wage; the Temporary Assistance for Needy Families program; Medicaid expansion; and housing support programs, have shown promising results in reducing neglect, physical abuse, and sexual abuse.^27–31 Our results are generally consistent with those from these studies. Given the growing body of research suggesting a causal link between income and both child maltreatment and CPS involvement,^32,33 programs offering cash or rental assistance have the potential to effectively reduce eviction threats and, thereby, child maltreatment. Also, it is important to recognize that, while eviction moratoria provided temporary relief to economically challenged families, they may have only delayed rather than stopped evictions. As such, it is unclear that maltreatment reports were fully avoided rather than delayed. As federal eviction moratoria expired in August 2021, eviction cases surged in the first 2 months of the year,³⁴ as well as the CPS reports comparing fiscal year 2022 with 2021,² highlighting the vulnerability of low-income families to both eviction and CPS involvement. Additionally, eviction moratoria faced criticism regarding the financial burden placed on landlords who may have struggled with missed rent payments. Although stimulus packages allocated $46.5 billion for rental assistance, the disbursement of these funds was slow and inadequate.³⁵ Given the impacts of eviction moratoria, policymakers should consider sustainable housing assistance to support financially struggling families while not penalizing landlords, both immediately following a public health crisis and over the long run.

Our findings are encouraging. Eviction protection policies may help prevent CPS involvement by ensuring secure housing for families. Future research would benefit from examining whether housing stability serves as the core mechanism that links eviction moratoria to child maltreatment. Given the varying findings across maltreatment types, examining potential mechanisms through which such policies may impact particular forms of maltreatment warrants further investigation.

b

coefficient estimate

CARES

Coronavirus Aid, Relief, and Economic Security

CPS

Child Protective Services

CDC

Centers for Disease Control and Prevention

NCANDS

National Child Abuse and Neglect Data System

EMHPD

Eviction Moratoria and Housing Policy Database

DiD

difference-in-differences