Data material and study population

This study takes advantage of data from a new Swedish database covering all judicial processes (e.g. all stages in summary proceedings and the eviction process) registered by the Swedish Enforcement Authority [30]. The Enforcement authority is the only authority in Sweden sanctioned to execute home evictions. After a court order or a decision after a summary proceeding, the creditor can make an application at the Enforcement Authority for the judgment to be executed. The most common reason for eviction is rent arrears; other common reasons are repeated late payments and nuisance [30, 31].

In addition to information on individuals exposed to eviction, the database also includes a 10% random sample of individuals (n ≈ 770 000) not exposed to eviction, representative of the Swedish population (aged 16 years or older on December 31, 2012). Through use of the personal identification number assigned to all Swedish individuals at the time of birth or at the time of immigration, individual-level, longitudinal data from other national registers were linked to the information on the evicted and non-evicted individuals. The personal identification numbers are replaced by random reference numbers before data is made available to researchers, and all data are anonymous. The researchers did not have access to any information that could identify individuals in the dataset. The project was approved by the ethics committee in the Stockholm region before any records were linked (2014/24-31/5).

For the purpose of the current study, we used information on all applications for evictions and all executed evictions during 2009–2011 retrieved from the database described above. A pooled dataset was created, containing record-linked national register data (see Table 1 for a list of registers) for the evicted and the non-evicted (i.e. the 10% random sample) individuals [32]. Among the evicted individuals, families with children were identified through The Multi-generational register. The exposed group consisted of all children born in 1995–2008 who had a parent who was evicted in 2009–2011 (i.e. when the children were aged 1–16 years) (n = 2224). The Multi-Generational register was also used to identify families with children among the 10% representative sample. Children of individuals included in the 10% representative sample, who were born in 1995–2008, constituted the non-evicted group (n = 250 146). Children with previous experiences of placement in out-of-home care were not included in the study population (n = 4919).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. List of national registers that are combined in the study.

https://doi.org/10.1371/journal.pone.0195295.t001

In Sweden, after a notice to quit a lease has been served and there has been a decision from a court procedure, the landlord can apply for the eviction to be executed. At this stage of the eviction process, about 30% of the evictions are executed [33]. Thus, a large number of individuals are served with an application for an execution of an eviction but the eviction is never executed. One explanation is that many tenants move without being formally evicted since an eviction reduces their chances of a new lease [30]. In additional analyses we analyzed information from children of individuals who received a notice of eviction in 2009–2011, but who were not formally evicted (n = 7847). We refer to this sample as children from families under threat of eviction.

Placement in out-of-home care

The outcome measure was defined as a first (ever) placement in out-of-home care (foster family or residential care) in 2012–2013, according to the Swedish Register of Children and Young Persons Subjected to Child Welfare Measures. Information on placements was only available until 2013 and due to the short-term follow-up period, the current study does not address duration of placement.

Covariates

We use a large number of observed background variables to control for confounding. To the extent unobserved confounding is related to observed confounding, our comprehensive controls indirectly control for unobservables [34]. Evictions are more common in resource-poor households, and previous studies have shown an overrepresentation of individuals with foreign origin among evicted individuals [35]. Low educational levels and poor economic resources are more common in parents of children placed in out-of-home care and previous research also suggests that the overrepresentation of children with a foreign background in out-of-home care can be attributed to socioeconomic differences [28]. The potential importance of socioeconomic factors was considered through adjustment for parental educational level and recipiency of social assistance. Social assistance recipiency was analyzed as a dichotomous indicator of whether the childhood household had received means-tested social assistance of any amount in 2008. A dichotomous indicator of whether the parents lived together in 2008 was also included in the analyses. Information on year of birth and sex of the child, country/region of birth of child and parent, and geographic residency/municipality was also included in the analyses. Covering the entire territory of the country, municipalities are the lower-level local government entity. Information on geographic residency was used in the analyses in two ways: included in the regression models as dummy variables for the 290 Swedish municipalities, (i.e. fixed effects), and as a categorical variable where the 290 municipalities were categorized as “city”, “town” and “rural” (see S1, S2 and S3 Tables describing ORs and CIs of all covariates).

A number of parental psychosocial factors may co-occur with both evictions and placement in out-of-home care [13, 14, 28, 29]. Parental psychiatric disorders were defined by at least one hospitalization with a diagnosis indicating psychiatric disorders and/or self-inflicted injuries during 1995–2008, according to the national Hospital Discharge Register. Information from this register was also used to identify parents with at least one hospitalization with an ICD diagnosis indicating alcohol or illicit drug use during 1995–2008. Parental criminality was defined as at least once having been convicted of a crime during 1995–2008, according to the Register of Court Convictions.

Statistical analyses

In studies with the aim of assessing effects of exposures, several approaches could be employed. Often baseline characteristics of the individuals are imbalanced between exposed and non-exposed groups and adjustments need to be made. This can be accomplished either through appropriate regression modeling or by various forms of matching estimators. Regression modeling can be described as a particular sort of weighted matching estimator [36], and therefore the differences between regression and matching estimates are unlikely to be of major empirical importance. It has also been shown that propensity score methods gave similar results to traditional regression modeling in observational studies [37]. In the current study, we use regression analysis to account for observed imbalances between evicted and non-evicted children.

Logistic regression models based on the Karlson/Holm/Breen method were used to estimate odds ratios (OR) and corresponding 95% confidence intervals (CI) [38]. The KHB method ensures that the crude and adjusted coefficients presented are measured on the same scale [39]. Since the prevalence of first placement in out-of-home care was low in our population (less than 10%), estimated ORs may be interpreted as risk ratios [40].

We examined the risk of placement in out-of-home care for a) evicted children in comparison with non-evicted children, b) children from families under threat of eviction in comparison with non-evicted children, and c) evicted children in comparison with children under threat of eviction, in three separate regression models. The latter analysis may to some extent approximate the true effect of eviction beyond the selective processes that lead to eviction since it is reasonable to assume that this contrast should substantially mitigate the usual omitted variable bias.

Potential differences with regard to gender of the child was investigated by means of including an interaction term between gender and exposure. Evicted families often include younger children [41]. Since younger children are more likely to be placed in out-of-home care because of reasons related to parenting problems and maltreatment, and since this group of children tend to be a more socially selected [28], we were also interested in investigating whether age moderated the association. We analyzed age of the child at the time of the eviction categorized as 1–6 years, 7–12 years, and 13–17 years.

Our data have a three-level structure: children (level 1, n = 260 217) who are nested in families/households (level 2, n = 158 832) which in turn are nested in municipalities (level 3, n = 290). Viewing the nested data structure as nuisance that needs correction, we use cluster-robust standard errors to account for the within-family error correlation [42]. To account for the clustering within municipalities, we use fixed-effects (dummy variables) which absorb all of the variation that occurs between these administrative units. When including all but one municipality-dummy in the model then there cannot be any between municipality variation explained by municipality-level variables such as variation in child welfare practices and preventive measures regarding evictions.

Stata version 14.2 was used for all analyses, the KHB-command was utilized for the logistic regression. The vce(cluster) option was used to account for family clustering [39].

Strengths and limitations

The strengths of this study include the large cohort sample (including a sizeable number of children who had experienced an eviction), the use of national register data with low attrition, comprehensive controls of robust socioeconomic and psychosocial confounders related to the parents, and temporal ordering of the variables of primary interest.

Although our comprehensive set of controls of observed confounders may have bought some protection against confounding by unobservables [34], data that could be vital for our understanding of the links between family evictions and subsequent placement in out-of-home care are often outside the scope of register-based studies. For example, the Swedish child welfare register does not contain data on reasons for placement in out-of-home care (e.g. abuse or neglect). We also do not have information on homelessness, e.g. usage of shelters, or repeated residential changes. We further do not have data on the hypothesized link between exposure to deprived residential neighborhoods and child maltreatment/parenting styles [51, 52], although a number of studies based on individual-level data suggest that this link is probably non-causal [53, 54]. Another limitation is that children with separated or divorced parents are recorded as living with just one of the parents, and we cannot know to what extent children from non-intact families were living with only one parent or alternately with both parents. However, the proportion of children in joint physical custody, i.e. where children alter their residency between the parents’ homes, is high in Sweden; 35–40% of all children from non-intact families spend more or less equal time with both parents [55], and previous studies have also shown that a majority of Swedish children have regular contact with both parents even if the parents do not live together [56]. Furthermore, a previous Swedish study showed that in 75% of the families under threat of eviction the child lived permanently in the dwelling from which the tenants are evicted [57]. According to Swedish law, the landlord is obliged to inform the local social welfare board when the tenant has been served with a notice to quit and the lease has been terminated, and the social welfare board must also be informed by the Enforcement Authority after the formal decision for the eviction to be executed. Thus, another potential limitation is the possibility that children from evicted families who are abused or neglected to a greater extent are detected by social services, compared to children in the comparison group. We have, however, excluded children with previous placements in out-of-home care, and the strength of the associations are therefore more likely to be underestimated, rather than the opposite.

Implications

The findings of the present study indicate that children who experience an eviction constitute a disadvantaged group in the national population and that these children are at significant risk of being placed in out-of-home care. Out-of-home placements were consistently more common in evicted children compared to children from non-evicted families with similar characteristics, indicating that eviction adds to the risk in already vulnerable families. Recently published studies emphasize the vulnerability of children placed in foster care, both as a result of early life circumstances but also since foster care placement in itself may be a risk factor for health problems in childhood [58–60]. Furthermore, even though the aim of an out-of-home placement is to provide children from adverse birth homes with better opportunities, these individuals constitute a group with a particular high risk of negative long term outcomes [19, 20]. Former foster youth also experience increased risks of homelessness, low educational attainment and unemployment [61]. Our findings emphasize the need of support for evicted children and their parents and the importance of addressing underlying economic and psychosocial problems in families at risk of eviction. A recent report from the European Commission [21] highlights a number of prevention measures, both general macro-level housing policies but also secondary prevention measures directed at individuals with a potentially high risk of eviction and homelessness, including counselling, support through social services and public assistance with housing costs and rent arrears. Avoiding households with children from being evicted, e.g. through such preventive policies, seems to be an important and viable intervention path.

Conclusion

Our results demonstrated that evicted children had a twofold elevated risk of being placed in out-of-home care, compared with children from the general population net of socioeconomic and psychosocial circumstances of the parents. These findings emphasize the need of support for these children and their parents, the importance of addressing underlying economic and psychosocial problems in families at risk of eviction, and of developing strategies focusing on avoiding households with children from being evicted.