Home Foreclosure and Child Protective Services Involvement

Home Foreclosure and Child ProtectiveServices InvolvementLawrence M. Berger, PhDa, J. Michael Collins, PhDa, Sarah A. Font, PhDb, Leah Gjertson, JDa, Kristen S. Slack, PhDa,Timothy Smeeding, PhDa

abstractOBJECTIVE: We estimated associations between experiencing a home foreclosure filing andexperiencing a child protective services (CPS) investigation or substantiation.

METHODS: We linked a large sample drawn from administrative data on foreclosure filings, CPSinvolvement, and participation in a host of other public programs for .60 000 Wisconsinhouseholds over a 4-year period from 2008 to 2011. Our empirical analyses used piecewiseexponential survival models to estimate the risk of CPS involvement (investigation orsubstantiation) as a function of a home foreclosure filing and a set of individual and householdcharacteristics. We fitted these models with and without the inclusion of propensity scoreweights.

RESULTS: Households that experienced a foreclosure filing had a much higher probability of CPSinvolvement. This was true in the year before the filing as well as the year after the foreclosurefiling. However, these associations were generally largest in the period before or shortlyafterward.

CONCLUSIONS: Experiencing a foreclosure filing is associated with increased CPS involvement.However, it is not clear that this association is driven by the foreclosure filing action itself.Rather, increased risk of CPS involvement is apparent during the process of moving toward thefiling as well as the year or so after the filing, both of which are likely characterized by limitedeconomic resources as well as by financial and other stress.

WHAT’S KNOWN ON THIS SUBJECT: Prior studieshave found a positive relationship betweenmacro-level indicators of home foreclosure andchild maltreatment rates. The extent to whichhome foreclosure may be associated with childprotective services involvement at the microlevel is largely unknown.

WHAT THIS STUDY ADDS: Foreclosure filings arepositively associated with child protectiveservices involvement. However, this is true of theperiods before and after a filing, which arecharacterized by economic and other stress,which may drive this association more than thefiling itself.

aInstitute for Research on Poverty, University of Wisconsin-Madison, Madison, Wisconsin; and bPopulationResearch Center, University of Texas at Austin, Austin, Texas

Drs Berger, Collins, Slack, and Smeeding conceptualized and designed the study; Dr Bergersupervised data management and analysis and drafted the initial manuscript; Dr Font and MsGjertson carried out the initial analyses and reviewed and revised the manuscript; Drs Collins,Slack, and Smeeding critically reviewed the manuscript; and all authors approved the finalmanuscript as submitted and agree to be held accountable for all aspects of the work.

www.pediatrics.org/cgi/doi/10.1542/peds.2014-2832

DOI: 10.1542/peds.2014-2832

Accepted for publication May 1, 2015

Address correspondence to Lawrence M. Berger, PhD, Institute for Research on Poverty, Universityof Wisconsin-Madison, 3420 William H. Sewell Social Sciences Bldg, 1180 Observatory Dr, Madison,WI 53706. E-mail: [email protected]

PEDIATRICS (ISSN Numbers: Print, 0031-4005; Online, 1098-4275).

Copyright © 2015 by the American Academy of Pediatrics

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant tothis article to disclose.

FUNDING: This research was generously funded by the John D. and Catherine T. MacArthurFoundation How Housing Matters to Families and Communities Initiative.

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts ofinterest to disclose.

PEDIATRICS Volume 136, number 2, August 2015 ARTICLEby guest on December 10, 2015Downloaded from

Socioeconomic disadvantage has beendescribed as “the most consistent andstrongest” predictor of childprotective services (CPS)involvement.1 An extensive literaturedocuments that economic stress andhardship, low income, poverty status,and related factors are associatedwith increased risk of both childmaltreatment and CPSinvolvement.2–23 Economic factorsmay directly and indirectly affect riskof abuse or neglect.24 Limitedeconomic resources, or a declinetherein, may directly increasea family’s likelihood of engaging inmaltreatment by preventing theability to fully meet a child’s basicmaterial needs.25–27 A chronic lack ofresources may indirectly increase theprobability of maltreatment throughits association with increasedparental stress and depression.28,29

Economic instability may haveparticularly important effects;a sudden decline in resources(negative shock) may lead toa concomitant increase in parentalstress and a deterioration of the homeenvironment beyond that associatedwith limited but stable economicresources.30 Finally, limited economicresources, or a drop therein, mayincrease the likelihood that a familybecomes known to CPS, even holdingparenting behaviors constant.

The recent housing crisis and GreatRecession were characterized byrampant economic stress as largenumbers of families experienced jobloss and home foreclosure.31 Whereasmany studies have examined linksbetween unemployment and variousmeasures of child maltreatment, weare aware of only 2 studies to directlyexamine associations betweenforeclosure and maltreatment,despite research indicating that homeforeclosure is inversely associatedwith well-being for both adults andchildren.32–35 Wood et al examinedassociations of metropolitanstatistical area-level unemployment,mortgage delinquency, and

foreclosure rates with hospitaladmissions for physical abuse andhigh-risk traumatic brain injury. Theyfound positive associations ofmortgage delinquency andforeclosure rates with both types ofadmissions.36 Frioux et al estimatedassociations of county-levelunemployment and mortgageforeclosure rates with childmaltreatment investigations andsubstantiations. Their results indicatethat both factors are positivelyassociated with CPS investigationsand substantiations.37 These studiesprovide evidence suggesting a linkbetween home foreclosure and childmaltreatment. However, one cannotassume that these macro-levelassociations hold at the individual orhousehold level. We are aware of nostudy that examines associationsbetween foreclosure and childmaltreatment at the micro level.

We use extensive administrative datafrom Wisconsin to examineassociations between experiencinga home foreclosure filing andexperiencing a CPS investigation orsubstantiation. We conceptualizehome foreclosure as a dynamicprocess, rather than a static event.That is, homeowners are likely to facea period of missed mortgagepayments and mortgage delinquency,which may be accompanied byincreasingly aggressivecommunications from lenders. Thisprocess may unfold over an extendedperiod of time and is likely to beaccompanied by considerablefinancial stress. The next step in theforeclosure process in a judicialprocedure state such as Wisconsinoccurs when the mortgage holderfiles a legal motion to foreclose. Sucha filing marks the beginning ofa formalized legal process throughwhich the lender seeks to evict thehomeowner and reclaim the property.During the subsequent months(sometimes years), the homeownermay make adequate payments toretain the home, may voluntarilyleave the home, or may enter into

a court process that results in theforeclosure being upheld or denied. Inaddition to the period precedinga foreclosure filing, the period inwhich the filing occurs is also likely tobe characterized by added stress,uncertainty, and perhaps instability.Furthermore, if a family chooses toleave the home, or is forced to do so,and is unable to find another suitablehome in a reasonable amount of time,it may be at risk for child neglect byfailing to provide adequate housingfor the children. Even if the familyfinds a new, adequate home, theprocess of doing so is likely to bechaotic. Evidence from the few priorstudies that have considered thetiming of foreclosure-related eventssuggests that the period surroundinga foreclosure filing is associated withgreater risk of adverse outcomes thanis the actual loss of the home,32,35

which generally occurs much later, ifat all. We expect the period leading toa foreclosure filing (during whicha family is likely to be experiencingfinancial hardship), the filing eventitself, and the period after the filing toeach be positively associated withCPS involvement. It is unclear,a priori, which period should be moststrongly associated with CPSinvolvement. Our empirical workexamines the timing of household-level foreclosure filings in relation tothe timing of household-level CPSinvolvement over a 4-year period,focusing on quarterly time intervals.

METHODS

Data

We used linked administrative datafrom the Wisconsin ConsolidatedCourt Automation Programs (WICCAP) and the Multi-Sample PersonFile (MSPF) database housed at theInstitute for Research on Poverty.38,39

WI CCAP data include caseinformation from the Wisconsincircuit courts case managementsystem. Court record summaries arepublic records under Wisconsin law.The MSPF database includes linked

300 BERGER et alby guest on December 10, 2015Downloaded from

individual-level administrative datafrom a host of public social welfareprograms. Inclusion in the MSPFnecessitates that an individual or hisor her family or household membersparticipated in $1 of these programsat some point. Any individual who isobserved as being part of a socialwelfare benefit case is included,regardless of being the actual benefitrecipient. Once an individual entersthe MSPF, he or she is followedbackward and forward in time,regardless of prior or ongoing benefitreceipt. For example, once anindividual enters the MSPF, we haveaccess to and can track his or herearnings (through unemploymentinsurance contributions) in allsubsequent and prior periodsregardless of program participation inthose periods.

To construct our data set, we firstextracted the entire universe of57 661 person-level foreclosure filingrecords in the WI CCAP from 2008 to2011. These records could includemultiple individuals in the samehousehold if listed in the filing. Wethen linked these files to MSPFrecords by using a deterministicmatching strategy based onidentifying information available inboth data sources. We were able tomatch 55 880 of the foreclosure cases(97%) to MSPF records. Because theWI CCAP and MSPF data were linkedat the individual rather thanhousehold level, households wereduplicated if multiple adults wereaffiliated with a foreclosure filing. Toensure that our analyses did notinclude duplicate households, werandomly selected 1 adult from eachhousehold and excluded 24 995duplicate household records. Thisresulted in a sample of 30 885households, observed each quarterfrom 2008 to 2011 (constituting494160 household-quarter observations).

We constructed a comparison groupby randomly selecting 40 000individuals in the MSPF. Because theWI CCAP data include the full

universe of WI foreclosures, thecomparison group consisted only ofhouseholds that did not experiencea foreclosure filing. We then excluded10 424 individuals (26%) who were,18 years old in the first period inwhich we could observe foreclosurefilings (quarter 1 of 2008). Thisresulted in a total sample of 60 461households (30 885 foreclosure and29 576 comparison group), observedeach quarter from 2008 to 2011(constituting 967 376 household-quarter observations).

Measures

We focused on 2 measures of CPSinvolvement including whethera household had been (1)investigated for child maltreatmentand (2) substantiated formaltreatment. Our key predictor waswhether a foreclosure had been filedregarding the address associated witha given MSPF record in a particularquarter. With the exception of thefinal set of “consolidated” models, forwhich we coded foreclosure asa discrete event constrained to thequarter of the filing, we codedforeclosure as 1 in the quarter inwhich it was filed and all subsequentquarters. In different modelspecifications, we assessedassociations between foreclosure andCPS involvement over a 2-yearperiod, accounting for various lag andlead periods. Our models controlledfor gender, race, and age of the adultassociated with the MSPFobservation; dummies for wagecategory (zero reported wages andwage quintile for those with positivewages, as well as an indicator that wewere unable to link to unemploymentinsurance wage data because therespondent’s social security numberwas not available in theadministrative data); and indicatorsfor Temporary Assistance for NeedyFamilies (TANF), the SupplementalNutrition Assistance Program (SNAP),Medicaid, child support received,child support paid, andunemployment in the current quarter.

Finally, we controlled for the year inwhich the observation occurred.

We used piecewise exponentialsurvival models to estimate risk ofCPS involvement as a function ofa foreclosure filing and the controlvariables. Piecewise exponentialmodels are particularly suited to ouranalyses because they do not requirethat the hazard rate be proportionalacross groups or over time. In otherwords, our models allow the hazardof CPS involvement to benonproportional across householdsthat did and did not experienceforeclosure. Additionally, the model isnot constrained by the assumptionthat the baseline hazard is flat(constant) or changing at a constantrate. Rather, it may vary across timeintervals identified in the model, solong as it meets the constancyassumption within a given interval.We allowed the baseline hazard tovary by year of observation. (We alsoestimated models allowing thebaseline hazard to vary by quarter ofobservation; however, this resulted ina considerably poorer model fit.) Thisstrategy enabled us to modelforeclosure filings using a variety oflag and lead periods over a 2-yearspan, without having to meet theconstancy assumption for thebaseline hazard. The reference groupin all models is households that didnot experience foreclosure.

We cannot identify homeownershipin the MSPF data. Thus, by necessity,the comparison group includes bothhomeowners and renters. This shouldbias our results toward zero, asrenters tend to be less advantagedthan homeowners and should,therefore, have a greater underlyingprobability of CPS involvement. Tobetter adjust for such differences, wealso implemented a propensity scoreweighting strategy. We generatedpropensity scores representing thelikelihood of foreclosure based on allcovariate values in quarter 1 of 2008.After computing a propensity scorefor each household, we conducted

PEDIATRICS Volume 136, number 2, August 2015 301by guest on December 10, 2015Downloaded from

weighted regression analyses inwhich the weights were equal to theinverse probability (propensity) ofexperiencing a foreclosure. Thisserved to more heavily weight themost similar households.

RESULTS

Table 1 presents descriptive statisticsand results from t tests for meandifferences in the characteristics ofhouseholds that experienceda foreclosure filing and those that didnot, as well as in the characteristics ofthose households that experienceda foreclosure when observed in pre-and postforeclosure quarters.

Households that experienceda foreclosure were substantially morelikely to have been involved with CPS.These households weredisproportionately white and hadhigher wages and less welfare receipt.They were also more likely to receiveSNAP and Medicaid, to pay andreceive child support, and to haveexperienced unemployment. Amonghouseholds experiencinga foreclosure filing, the probability ofCPS involvement was slightly, thoughnot substantially, less common in theperiod after the filing than before it.Households were substantially moredisadvantaged in the postfiling periodthan prefiling; they reported lower

wages, more unemployment, andgreater receipt of SNAP, Medicaid,and TANF in the postforeclosureperiod.

Our first set of survival analysisestimates is shown in Table 2. Theupper panel presents results withoutthe propensity score weights; thelower panel presents propensityscore weighted results. The estimatespresented in each column representthe risk of experiencing a CPSinvestigation for foreclosedhouseholds relative to nonforeclosedhouseholds, all else being equal. Eachcolumn shows results from a separatemodel focusing on a particular periodin which a filing may have occurred,

TABLE 1 Descriptive Statistics by Foreclosure Filing Status

Measure Full Sample Households Experiencing Foreclosure Filing

No Foreclosure Filing Foreclosure Filing Before Foreclosure Filing After Foreclosure Filing

CPS involvementCPS investigation (ever) 3.61 11.85a

CPS investigation in current quarter 0.24 0.78a 0.81 0.76b

Substantiated investigation (ever) 0.74 8.56a

Substantiated investigation in current quarter 0.03 0.11a 0.12 0.09c

Time constant covariatesCPS involvement before 2008 2.30 7.07a

Female 49.05 44.46a

Missing gender 3.31 0.60a

White 47.35 67.04a

Black 9.90 9.12a

Hispanic 5.25 6.16a

Asian 2.23 2.79a

American Indian 1.71 1.98a

Race unknown 33.55 12.91a

Missing age 10.29 4.32a

Mean age in years in 2008 (SD) 45.43 (20.27) 40.82 (10.30)a

Time varying covariatesNo reported wages 50.95 35.98a 32.48 38.92a

Unable to link to wage data 20.63 10.23a 10.78 9.78a

0th–20th percentile wages 17.98 11.96a 10.69 13.03a





TANF receipt this quarter 0.36 0.21a 0.10 0.30a

SNAP receipt this quarter 8.16 10.77a 7.43 13.59a

Medicaid receipt this quarter 11.71 22.03a 18.61 24.90a

Received child support this quarter 4.20 11.36a 9.63 12.82a

Paid child support this quarter 4.39 9.01a 7.07 10.65a

Unemployed this quarter 5.47 12.73a 12.18 13.19a

Household-quarter observations 473 216 494 160 225 672 268 488

Values are expressed as % unless noted otherwise. Table shows 967 376 household-quarter observations of 60 461 households observed from Quarter 1, 2008, through Quarter 4, 2011(analyses are based on 1 observation of each household in each of these quarters, for a total of 16 observations per household). The “unable to link to wage data” category is a subset ofthe “no reported wages” category, such that these categories are not mutually exclusive. t tests were used to assess mean differences between groups.a P , .001.b P , .05.c P , .01.


including subsequent (impending)filings, concurrent filing, and prior(lagged) filings. Once a householdexperienced an initial filing, allsubsequent quarters were codedaffirmatively for having experienceda filing.

The results indicate that householdsthat will go on to experiencea foreclosure filing in the next 6 to 12months are at 70% greater risk ofa CPS investigation than householdsthat will not subsequently experiencea foreclosure. Indeed, the entireperiod from 12 months before to 6months after the filing is associatedwith elevated risk of CPSinvestigation. Interestingly, thepattern of the coefficients suggeststhat such risk is greatest during the 6to 12 months before the foreclosureand decreases in magnitude in eachconsecutive quarter. Approximately 6months after a filing, households areno longer at elevated risk. Thepropensity score weighted resultsreveal a consistent pattern, thoughthe estimates are somewhat larger.Additionally, these results indicatethat households are also at increasedrisk of CPS involvement during the 12months after a filing. As noted above,foreclosure group households tendedto be more socioeconomicallyadvantaged than comparison grouphouseholds (all of the formerachieved homeownership, whereasthe comparison group includes bothrenters and homeowners), whichshould likely bias our estimatestoward zero. Indeed, these resultsindicate that, once the data areweighted to account for suchdifferences, we find even greater riskof CPS involvement in the periodsbefore, during, and shortly aftera foreclosure filing. Table 3 presentsresults for substantiations. Thepattern and magnitude of thecoefficients is quite consistent withthose for investigations.

Table 4 shows results when all of thelag and lead periods are included inthe same “consolidated” model. Here,TA

BLE2

Piecew

iseExponentialS

urvivalM

odel

Results

forCPSInvestigation,

Separate

Modelsby

Timingof

ForeclosureFiling

Timing

ImpendingForeclosureFiling

Concurrent

ForeclosureFiling

Lagged

ForeclosureFiling

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Withoutpropensity

scoreweights

Foreclosurefiling3–4quarters

after

1.701(0.055)a

Foreclosurefiling2quarters

after

1.471(0.047)a

Foreclosurefilingin

nextquarter

1.398(0.045)a

Foreclosurefilingin

currentquarter

1.316(0.043)a

Foreclosurefilingin

priorquarter

1.266(0.043)a


before

1.167(0.041)a


before

1.046(0.041)

Foreclosurefiling$5quarters

before

0.782(0.068)b

With

propensity

scoreweights


1.739(0.080)a

after Foreclosurefiling2quarters

after

1.539(0.068)a

Foreclosurefilingin

nextquarter

1.481(0.065)a

Foreclosurefilingin

currentquarter

1.410(0.062)a

Foreclosurefilingin

priorquarter

1.359(0.060)a


before

1.271(0.058)a


before

1.154(0.057)b


before

0.867(0.083)

Values

areexpressedas

hazard

ratios(SEs).Foreach

setof

results

(withoutandwith

propensity

scoreweights),each

column(1

through8)

presents

results

from

aseparate

piecew

iseexponentialsurvival

model

exam

iningwhether

ahousehold

experiencedaforeclosurefilinginthespecified

period.Tableincludes

967376household-quarterobservations

of60

461households

observed

from

Quarter1,2008,through

Quarter4,2011

(analysesarebasedon

1observationofeach

householdineach

ofthesequarters,for

atotalof16observations

perhousehold).Observations

arecensored

afterthefirstfailure

(foreclosurefiling).Allmodelscontrolfor

thecovariates

show

ninthelower

panelofTable1as

wellasafixedeffectforyear

ofobservation.

aP,

.001.

bP,

.01.


we are concerned only with the initialquarter in which a foreclosure filingoccurred; we thus coded theforeclosure variable as 1 in thatquarter only and not in subsequentquarters, allowing for multiple CPSevents (failures). These resultsindicate that experiencinga foreclosure in any of the periodsconsidered is associated with anincreased likelihood of a CPSinvestigation and substantiation, with2 exceptions. We find no associationsfor experiencing a substantiation inthe same quarter as the filing, nor forexperiencing a substantiation $5quarters after the filing. Consistentwith our prior estimates, themagnitude of the association betweenforeclosure and CPS involvementtended to decrease steadily over time,such that it was highest ∼12 monthsbefore a foreclosure and lowest ∼12months after the foreclosure.However, these results also reveala notable temporary dip in the hazardfor each outcome in the quarter of thefiling. Finally, they suggest thathouseholds are at substantiallyincreased risk of CPS involvement inthe 6 to 12 months after a filing, aswell as during the 12 months before.

DISCUSSION

Experiencing a foreclosure filing isassociated with increased risk ofa CPS involvement. This pattern,which is apparent in our raw data andthe estimates from our survivalanalyses, supports findings from priorresearch linking macro-levelforeclosure rates with childmaltreatment prevalence.36,37 Anexamination of the timing of theseevents, however, reveals thatincreased risk of CPS involvement isapparent as much as 12 monthsbefore a foreclosure filing, althoughhouseholds continue to be atheightened risk of CPS involvementthrough the 6 to 12 months after thefiling.

On the whole, this leads us toconclude that factors such asTA

BLE3

Piecew

iseExponentialS

urvivalM

odel

Results

forCPSSubstantiation,

Separate

Modelsby

Timingof

ForeclosureFiling

Timing

ImpendingForeclosureFiling

Concurrent

ForeclosureFiling

Lagged

ForeclosureFiling

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Withoutpropensity

scoreweights


after

1.696(0.144)a


after

1.448(0.123)a

Foreclosurefilingin

nextquarter

1.330(0.115)a

Foreclosurefilingin

currentquarter

1.200(0.107)b

Foreclosurefilingin

priorquarter

1.308(0.120)c


before

1.169(0.114)


before

1.058(0.117)


before

1.054(0.234)

With

propensity

scoreweights


after

1.769(0.173)a


after

1.563(0.151)a

Foreclosurefilingin

nextquarter

1.429(0.142)a

Foreclosurefilingin

currentquarter

1.332(0.135)c

Foreclosurefilingin

priorquarter

1.406(0.148)c


before

1.314(0.145)b


before

1.200(0.149)


before

1.221(0.292)

Values

areexpressedas

hazard

ratios(SEs).Foreach

seto

fresults(withouta

ndwith

propensityscoreweights),each

column(1

through8)

presentsresults

from

aseparatepiecew

iseexponentialsurvivalm

odelexam

iningwhether

ahousehold

experiencedaforeclosurefilinginthespecified

period.Tableincludes

967376household-quarterobservations

of60

461households

observed

from

Quarter1,2008,through

Quarter4,2011

(analysesarebasedon

1observationofeach

householdin

each

ofthesequarters,for

atotalof16observations

perhousehold).Observations

arecensored

afterthefirstfailure

(foreclosure

filing).Allmodelscontrolfor

thecovariates

show

ninthelower

panelofTable1as

wellasafixedeffectforyear

ofobservation.

aP,

.001.

bP,

.05.

cP,

.01.


decreased economic resources,increased economic instability, andhigh levels of economic and otherstress that characterize the periodsurrounding a foreclosure are likelydriving the association betweenforeclosure and CPS involvement.That is, whereas our resultsdemonstrate an association betweenforeclosure and CPS involvement,they do not necessarily indicate thatthere is a causal link between thefiling itself and increased CPSinvolvement. As such, policies andprograms aimed at stabilizingeconomic resources and preventingor compensating for adverseeconomic shocks may be moresuccessful at preventingmaltreatment than those solelyfocused on helping households avoidforeclosure without addressing otherunderlying economic needs andstressors.

Our findings are particularly notablegiven that those households thatexperienced foreclosure were, onaverage, more advantaged than thosethat did not. To begin with, theforeclosure group consistedpredominantly of homeowners,whereas the comparison group

included a combination of renters andowners. In addition, the foreclosuregroup experienced higher wages andgreater child support payment andreceipt than the comparison group.These factors all suggest greatereconomic advantage. At the sametime, households that experienceda foreclosure were likely to have alsoexperienced an adverse economicshock. This is reflected by the largedecrease in wages and increase inunemployment, as well as TANF,SNAP, and Medicaid receipt, inpostforeclosure compared withpreforeclosure quarters (Table 1). Onone hand, we would expect moreadvantaged households, such as thosethat were able to purchase a home, tohave lower risk of CPS involvement.On the other hand, prior evidencesuggests that adverse economicshocks are associated with increasedmaltreatment.29 Our findings clearlyprovide support for the latterconjecture. After adjusting for otherdifferences in socioeconomicadvantage, the 24-month periodsurrounding a foreclosure filing isassociated with roughly a 40% to95% greater likelihood of a CPSinvestigation or substantiation; such

risk is highest in the 6 to 12 monthsbefore or after a foreclosure filing.

CONCLUSIONS

Homeownership has long been linkedto a variety of social and economicbenefits.40 However, recent researchsuggests that that these benefitsprimarily reflect the stabilityassociated with owning a home, ratherthan homeownership itself.41,42

Owners move less often than renters,and this stability allows families andchildren to establish a foundation ina community, remain in a consistentschool, and use a known set ofcommunity institutions. Whereas ourresults imply that the economicinstability surrounding foreclosure isassociated with increased CPSinvolvement, it is important to notethat we focused only on foreclosurefilings, not on foreclosure finalizationor eviction. It is possible that beingevicted may compound the adverseeffects of preexisting economic stressand lead to even greater risk of CPSinvolvement. This is an area that isripe for future research.

Overall, we find consistent evidencethat households that experience

TABLE 4 Piecewise Exponential Survival Model Results for CPS Investigation and Substantiation, Consolidated Model

Timing Investigation Substantiation

Without PropensityScore Weights

With PropensityScore Weights

Without PropensityScore Weights

With PropensityScore Weights

Foreclosure filing 3–4 quarters after 1.804 (0.089)a 1.950 (0.112)a 1.832 (0.235)a 1.985 (0.274)a

Foreclosure filing 2 quarters after 1.653 (0.107)a 1.693 (0.119)a 1.922 (0.307)a 2.202 (0.387)a

Foreclosure filing in next quarter 1.619 (0.101)a 1.671 (0.116)a 1.894 (0.293)a 1.863 (0.300)a

Foreclosure filing in current quarter 1.468 (0.093)a,b 1.549 (0.111)a,b 0.812 (0.178)b,c,d 0.957 (0.227)b,c,d

Foreclosure filing in prior quarter 1.682 (0.103)a 1.696 (0.114)a 1.857 (0.290)a,e 1.775 (0.291)a,e

Foreclosure filing 2 quarters before 1.607 (0.103)a 1.667 (0.120)a 1.705 (0.288)e,f 1.760 (0.321)e,f

Foreclosure filing 3–4 quarters before 1.459 (0.078)a,b 1.497 (0.094)a,b 1.403 (0.208)e,g 1.473 (0.231)g

Foreclosure filing $5 quarters before 1.281 (0.058)a,b,c,d,h,i,j 1.338 (0.076)a,b,c,d,h,i 1.148 (0.148)b,c,d,h,i 1.281 (0.185)b,c

Values are expressed as hazard ratios (SEs). Table includes 967 376 household-quarter observations of 60 461 households observed from Quarter 1, 2008, through Quarter 4, 2011(analyses are based on 1 observation of each household in each of these quarters, for a total of 16 observations per household). Observations are not censored after the first failure; thatis, these models allow for multiple failures. All models control for the covariates shown in the lower panel of Table 1 as well as a fixed effect for year of observation.a P , .001.b P , .05 versus “Foreclosure 3–4 quarters after.”c P , .05 versus “Foreclosure 2 quarters after.”d P , .05 versus “Foreclosure in next quarter.”e P , .05 versus “Foreclosure in current quarter.”f P , .01.g P , .05.h P , .05 versus “Foreclosure in prior quarter.”i P , .05 versus “Foreclosure 2 quarters before.”j P , .05 versus “Foreclosure 3–4 quarters before.”


a foreclosure filing are at increasedrisk of CPS involvement in both theyear before and the year after thefiling. However, this does notnecessarily suggest that there isa causal link between the filing itselfand becoming involved with CPS.Indeed, we do not observe animmediate jump in CPS involvementin the period during or directly afterthe filing. Rather, households are atequal or greater risk of CPSinvolvement in the year before thefiling and shortly thereafter. In plainterms, foreclosure appears to bea symptom of larger economicproblems that create stress within thefamily, especially for relativelyadvantaged families that are notusually thought to face suchpressures. As such, childmaltreatment prevention practicesand policy would be well served bysupporting all households thatexperience an adverse economicshock rather than focusingspecifically on those householdsfacing foreclosure.

ACKNOWLEDGMENTS

We are grateful to Jane Smith for datamerging assistance and to ShainiKothari and Laura Cuesta for dataextraction assistance. We assumeresponsibility for all errors ofcommission and omission.

ABBREVIATIONS

CPS: child protective servicesMSPF: Multi-Sample Person fileSNAP: Supplemental Nutrition

Assistance ProgramTANF: Temporary Assistance for

Needy FamiliesWI CCAP: Wisconsin Consolidated

Court AutomationPrograms

REFERENCES

1. Cameron G, Freymond N. Understandinginternational comparisons of childprotection, family service, and

community caring systems of child andfamily welfare. In: Cameron G, FreymondN, eds. Towards Positive Systems of Childand Family Welfare: InternationalComparisons of Child Protection, FamilyService, and Community Caring Systems.Toronto, Canada: University of TorontoPress; 2006:3–25

2. Aber L. Poverty, violence, and childdevelopment: untangling family andcommunity level effects. In: Nelson CA,ed. Threats to Optimal Development:Integrating Biological, Psychological,and Social Risk Factors. The MinnesotaSymposia on Child Psychology. Vol 27.Hillsdale, NJ: Lawrence Erlbaum; 1994:229–272

3. Berger LM. Income, family structure, andchild maltreatment risk. Child Youth ServRev. 2004;26(8):725–748

4. Coulton CJ, Crampton DS, Irwin M,Spilsbury JC, Korbin JE. Howneighborhoods influence childmaltreatment: a review of the literatureand alternative pathways. Child AbuseNegl. 2007;31(11–12):1117–1142

5. Coulton CJ, Korbin JE, Su M, Chow J.Community level factors and childmaltreatment rates. Child Dev. 1995;66(5):1262–1276

6. Courtney ME, Dworsky A, Piliavin I, ZinnA. Involvement of TANF applicant familieswith child welfare services. Soc Serv Rev.2005;79(1):119–157

7. Cowal K, Shinn M, Weitzman BC,Stojanovic D, Labay L. Mother-childseparations among homeless andhoused families receiving publicassistance in New York City. Am JCommunity Psychol. 2002;30(5):711–730

8. Culhane J, Webb D, Grim S, et al.Prevalence of child welfare servicesinvolvement among homeless and low-income mothers: a five year birth cohortstudy. J Sociol Soc Welf. 2003;30(3):79–95

9. Drake B, Pandey S. Understanding therelationship between neighborhoodpoverty and specific types of childmaltreatment. Child Abuse Negl. 1996;20(11):1003–1018

10. Eckenrode, J, Smith, EG, McCarthy, ME,Dineen, M. Income inequality and childmaltreatment in the United States.Pediatrics. 2014;133(3):454–461

11. Fein D, Lee W. The impacts of welfarereform on child maltreatment in

Delaware. Child Youth Serv Rev. 2003;25(1):83–111

12. Font S, Warren E. Inadequate housingand the child protection systemresponse. Child Youth Serv Rev. 2013;35(11):1809–1815

13. Gelles R. Poverty and violence towardchildren. Am Behav Sci. 1992;35(3):258–274

14. Jones ED, McCurdy K. The links betweentypes of maltreatment and demographiccharacteristics of children. Child AbuseNegl. 1992;16(2):201–215

15. McDaniel M, Slack KS, Lewis D. Major lifeevents and the risk of a childmaltreatment investigation. Child YouthServ Rev. 2005;27(2):171–195

16. Paxson C, Waldfogel J. Work, welfare, andchild maltreatment. J Labor Econ. 2002;20(3):435–474

17. Sedlak A, Broadhurst D. Third NationalIncidence Study of Child Abuse andNeglect. Washington, DC: US Departmentof Health and Human Services; 1996

18. Sedlack AJ, Mettenburg J, Basena M,et al. Fourth National Incidence Study ofChild Abuse and Neglect (NIS-4): Reportto Congress. Washington, DC: USDepartment of Health and HumanServices, Administration for Childrenand Families; 2010

19. Sidebotham P, Heron J, Golding J; ALSPACstudy team. Child maltreatment in the“Children of the Nineties”: deprivation,class, and social networks in a UKsample. Child Abuse Negl. 2002;26(12):1243–1259

20. Slack KS, Holl JL, Lee BJ, McDaniel M,Altenbernd L, Stevens AB. Childprotective intervention in the context ofwelfare reform: the effects of work andwelfare on maltreatment reports. JPolicy Anal Manage. 2003;22(4):517–536

21. Slack KS, Lee BJ, Berger LM. Do welfaresanctions increase child protectionsystem involvement? A cautious answer.Soc Serv Rev. 2007;81(2):207–228

22. Jonson-Reid M, Drake B, Kohl PL. Is theoverrepresentation of the poor in childwelfare caseloads due to bias or need?Child Youth Serv Rev. 2009;31(3):422–427

23. Institute of Medicine and NationalResearch Council of the NationalAcademies. New Directions in ChildAbuse and Neglect Research.


Washington, DC: The National AcademiesPress; 2013

24. Berger LM, Waldfogel J. EconomicDeterminants and Consequences of ChildMaltreatment. OECD Social, Employmentand Migration Working Paper No. 111.Paris, France: Organization for EconomicCooperation and DevelopmentPublishing; 2011

25. Berger LM. Socioeconomic factors andsubstandard parenting. Soc Serv Rev.2007;81(3):485–522

26. Pelton LH. The role of material factors inchild abuse and neglect. In: Melton GB,Barry FD, eds. Protecting Children FromAbuse and Neglect. New York, NY:Guilford Press; 1994:131–181

27. Shook K. Does the loss of welfare incomeincrease the risk of involvement with thechild welfare system? Child Youth ServRev. 1999;21(9–10):781–814

28. Conger RD, Ge X, Elder GH Jr, Lorenz FO,Simons RL. Economic stress, coercivefamily process, and developmentalproblems of adolescents. Child Dev. 1994;65(2 Spec No):541–561

29. Conger RD, Wallace LE, Sun Y, Simons RL,McLoyd VC, Brody GH. Economic pressurein African American families: a replicationand extension of the family stress model.Dev Psychol. 2002;38(2):179–193

30. Votruba-Drzal E. Income changes andcognitive stimulation in young children’s

home learning environments. J MarriageFam. 2003;65(2):341–355

31. Danziger S. Evaluating the effects of theGreat Recession. Ann Am Acad Pol SocSci. 2013;650(1):6–24

32. Fowler KA, Gladden RM, Vagi KJ, BarnesJ, Frazier L. Increase in suicidesassociated with home eviction andforeclosure during the US housing crisis:findings from 16 National Violent DeathReporting System States, 2005-2010. AmJ Public Health. 2015;105(2):311–316

33. McLaughlin KA, Nandi A, Keyes KM, et al.Home foreclosure and risk of psychiatricmorbidity during the recent financialcrisis. Psychol Med. 2012;42(7):1441–1448

34. Pollack CE, Lynch J. Health status ofpeople undergoing foreclosure in thePhiladelphia region. Am J Public Health.2009;99(10):1833–1839

35. Bradbury K, Burke MA, Triest RK. Theeffect of foreclosure on Boston PublicSchool student academic performance.Federal Reserve Bank of Boston WorkingPaper (13-12); 2013. Available at: www.bostonfed.org/economic/wp/wp2013/wp1312.pdf. Accessed January 5, 2015

36. Wood JN, Medina SP, Feudtner C, et al.Local macroeconomic trends andhospital admissions for child abuse,2000-2009. Pediatrics. 2012;130(2).Available at: www.pediatrics.org/cgi/content/full/130/2/e358

37. Frioux S, Wood JN, Fakeye O, Luan X,Localio R, Rubin DM. Longitudinalassociation of county-level economicindicators and child maltreatmentincidents. Matern Child Health J. 2014;18(9):2202–2208

38. Brown P, Ross D, Smith JA, et al.Technical report on lessons learned inthe development of the Institute forResearch on Poverty’s Multi-SamplePerson File (MSPF) data system.Madison, WI: Institute for Research onPoverty; 2011

39. Wisconsin Consolidated CourtAutomation Programs (WI CCAP)[electronic database]. Wisconsin CircuitCourt Access (WCCA), Wisconsin CourtSystem, Director of State Courts.Available at: https://www.wicourts.gov/courts/offices/ccap.htm

40. Aaronson D. A note on the benefits ofhomeownership. J Urban Econ. 2000;47(3):356–369

41. Di ZX, Belsky E, Liu X. Do homeownersachieve more household wealth in thelong run? J Hous Econ. 2007;16(3):274–290

42. Herbert CE, Belsky ES. Initial housingchoices made by low-income andminority homebuyers. Cityscape. 2008;10(2). Available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1341190.Accessed May 8, 2015


DOI: 10.1542/peds.2014-2832; originally published online July 20, 2015; 2015;136;299Pediatrics

Slack and Timothy SmeedingLawrence M. Berger, J. Michael Collins, Sarah A. Font, Leah Gjertson, Kristen S.

Home Foreclosure and Child Protective Services Involvement

ServicesUpdated Information &

/content/136/2/299.full.htmlincluding high resolution figures, can be found at:

References

/content/136/2/299.full.html#ref-list-1at:This article cites 31 articles, 3 of which can be accessed free

Subspecialty Collections

/cgi/collection/child_abuse_neglect_subChild Abuse and Neglectthe following collection(s):This article, along with others on similar topics, appears in

Permissions & Licensing

/site/misc/Permissions.xhtmltables) or in its entirety can be found online at: Information about reproducing this article in parts (figures,

Reprints /site/misc/reprints.xhtml

Information about ordering reprints can be found online:

rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.Grove Village, Illinois, 60007. Copyright © 2015 by the American Academy of Pediatrics. All and trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elkpublication, it has been published continuously since 1948. PEDIATRICS is owned, published, PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly

by guest on December 10, 2015Downloaded from

DOI: 10.1542/peds.2014-2832; originally published online July 20, 2015; 2015;136;299Pediatrics

Slack and Timothy SmeedingLawrence M. Berger, J. Michael Collins, Sarah A. Font, Leah Gjertson, Kristen S.

Home Foreclosure and Child Protective Services Involvement

/content/136/2/299.full.html

located on the World Wide Web at: The online version of this article, along with updated information and services, is

of Pediatrics. All rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.Boulevard, Elk Grove Village, Illinois, 60007. Copyright © 2015 by the American Academy published, and trademarked by the American Academy of Pediatrics, 141 Northwest Pointpublication, it has been published continuously since 1948. PEDIATRICS is owned, PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly

by guest on December 10, 2015Downloaded from

Documents

Home Foreclosure and Child Protective Services Involvement