paternal incarceration and the housing security of urban mothers

17
AMANDA GELLER Columbia University ALLYSON WALKER FRANKLIN John Jay College of Criminal Justice Paternal Incarceration and the Housing Security of Urban Mothers High rates of imprisonment among American fathers have motivated an ongoing examination of incarceration’s role in family life. A growing literature suggests that incarceration creates material and socioemotional challenges not only for prisoners and former prisoners but also for their families and communities. The authors examined the relationship between fathers’ incarceration and one such challenge: the housing insecurity of the mothers of their children. Using data from the Fragile Families and Child Wellbeing Study ( N = 4,125) and a series of longitudinal regression models, they found that mothers’ housing security was compromised following their partners’ incarceration, an association likely driven in part, but not entirely, by financial challenges following his time in prison or jail. Given the importance of stable housing for the continuity of adult employment, children’s schooling, and other inputs to healthy child development, the findings suggest a grave threat to the well-being of children with incarcerated fathers. Housing security has long been recognized as an integral component of the economic, Columbia University Mailman School of Public Health, 722 West 168th St., New York, NY 10032 ([email protected]). John Jay College of Criminal Justice, North Hall, Room 2414, 524 W. 59th St., New York, NY 10019. Key Words: Fragile Families and Child Wellbeing Study, housing, incarcerated parents, mothers. physical, and emotional health and well-being of individuals and families (Bradley, Oliver, Richardson, & Slayter, 2001; Lee, Tyler, & Wright, 2010; Postmus, Severson, Berry, & Yoo, 2009). The lack of safe and stable housing is often viewed as an indicator of severe social exclusion, particularly for individuals vulnerable in other aspects of their lives (Lee et al., 2010). Children’s schooling, the receipt of social services, treatment for medical conditions, and the search for employment are each facilitated by a stable home address (Bradley et al., 2001; Buckner, 2008; Rafferty, Shinn, & Weitzman, 2004). Even in the absence of homelessness, hous- ing insecurity threatens grave consequences for health and well-being. Although federal guide- lines (42 U.S.C. § 11302) define homelessness as the lack of a “fixed regular, and adequate nighttime residence,” researchers have noted that eviction, frequent moves, difficulty paying rent, doubling up, and living in overcrowded conditions represent “a manifestation of the same underlying relationship between hous- ing costs and household resources” (Honig & Filer, 1993) and threaten individual and fam- ily functioning (Gilman, Kawachi, Fitzmaurice, & Buka, 2003; Kushel, Gupta, Gee, & Haas, 2005; Ma, Gee, & Kushel, 2008; Reid, Vit- tinghoff, & Kushel, 2008; U.S. Department of Housing and Urban Development, 2003). Crowded conditions and excessive residential mobility also have the potential to disrupt children’s school attendance and performance (Cunningham, Harwood, & Hall, 2010; Goux Journal of Marriage and Family 76 (April 2014): 411–427 411 DOI:10.1111/jomf.12098

Upload: allyson-walker

Post on 06-Apr-2017

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Paternal Incarceration and the Housing Security of Urban Mothers

AMANDA GELLER Columbia University

ALLYSON WALKER FRANKLIN John Jay College of Criminal Justice∗

Paternal Incarceration and the Housing Security

of Urban Mothers

High rates of imprisonment among Americanfathers have motivated an ongoing examinationof incarceration’s role in family life. A growingliterature suggests that incarceration createsmaterial and socioemotional challenges notonly for prisoners and former prisoners butalso for their families and communities. Theauthors examined the relationship betweenfathers’ incarceration and one such challenge:the housing insecurity of the mothers of theirchildren. Using data from the Fragile Familiesand Child Wellbeing Study (N = 4,125) anda series of longitudinal regression models,they found that mothers’ housing securitywas compromised following their partners’incarceration, an association likely driven inpart, but not entirely, by financial challengesfollowing his time in prison or jail. Given theimportance of stable housing for the continuityof adult employment, children’s schooling, andother inputs to healthy child development, thefindings suggest a grave threat to the well-beingof children with incarcerated fathers.

Housing security has long been recognizedas an integral component of the economic,

Columbia University Mailman School of Public Health,722 West 168th St., New York, NY 10032([email protected]).∗John Jay College of Criminal Justice, North Hall, Room2414, 524 W. 59th St., New York, NY 10019.

Key Words: Fragile Families and Child Wellbeing Study,housing, incarcerated parents, mothers.

physical, and emotional health and well-beingof individuals and families (Bradley, Oliver,Richardson, & Slayter, 2001; Lee, Tyler, &Wright, 2010; Postmus, Severson, Berry, &Yoo, 2009). The lack of safe and stable housingis often viewed as an indicator of severe socialexclusion, particularly for individuals vulnerablein other aspects of their lives (Lee et al.,2010). Children’s schooling, the receipt of socialservices, treatment for medical conditions, andthe search for employment are each facilitatedby a stable home address (Bradley et al., 2001;Buckner, 2008; Rafferty, Shinn, & Weitzman,2004).

Even in the absence of homelessness, hous-ing insecurity threatens grave consequences forhealth and well-being. Although federal guide-lines (42 U.S.C. § 11302) define homelessnessas the lack of a “fixed regular, and adequatenighttime residence,” researchers have notedthat eviction, frequent moves, difficulty payingrent, doubling up, and living in overcrowdedconditions represent “a manifestation of thesame underlying relationship between hous-ing costs and household resources” (Honig &Filer, 1993) and threaten individual and fam-ily functioning (Gilman, Kawachi, Fitzmaurice,& Buka, 2003; Kushel, Gupta, Gee, & Haas,2005; Ma, Gee, & Kushel, 2008; Reid, Vit-tinghoff, & Kushel, 2008; U.S. Departmentof Housing and Urban Development, 2003).Crowded conditions and excessive residentialmobility also have the potential to disruptchildren’s school attendance and performance(Cunningham, Harwood, & Hall, 2010; Goux

Journal of Marriage and Family 76 (April 2014): 411–427 411DOI:10.1111/jomf.12098

Page 2: Paternal Incarceration and the Housing Security of Urban Mothers

412 Journal of Marriage and Family

& Maurin, 2005) and other aspects of familyfunctioning.

We examined the risk of housing insecurityamong a vulnerable population of growing inter-est to researchers and policymakers: familiesexperiencing a father’s incarceration. The sharpand unprecedented expansion of the correctionalpopulation in the past 40 years, combined withconsistently high rates of fatherhood amongincarcerated men, has led to an increasing num-ber of families with fathers in prison and jail.A growing literature has suggested that fathers’incarceration adversely affects their partners andchildren, including their economic and materialwell-being (Comfort, 2007; Geller, Garfinkel,& Western, 2011; Schwartz-Soicher, Geller, &Garfinkel, 2011). Although safe and stable hous-ing has the potential to enable family resilienceto these challenges, these economic challengesconversely have the potential to underminehousing security, further destabilizing familylife (Wildeman, 2013). Using data from alarge longitudinal survey of urban families,we estimated the relationship between fathers’incarceration and mothers’ housing insecurity,considering challenges ranging from relativelycommon occurrences, such as skipping a rentor mortgage payment, to more disruptive hard-ships, such as eviction or homelessness. Tothe extent that fathers’ incarceration increasessuch instability, their partners and children mayrequire specialized attention by social serviceproviders.

HYPOTHESIZED RELATIONSHIPS

There are a number of reasons to expect thatfathers’ incarceration might compromise thehousing security of their families. When fathersreside with their children, incarceration removesthem from the household and incapacitates themfrom the labor market, depriving their families ofa potential source of income. Even fathers whodo not live with their children often contributefinancially in the form of child support (Gelleret al., 2011; Nepomnyaschy & Garfinkel, 2007),visit with their children (Geller, 2013), andmaintain involvement in their lives and day-to-day routines (Swisher & Waller, 2008; Tach,Mincy, & Edin, 2010). Travis, McBride, andSolomon (2005) noted that 68% of incarceratedfathers had provided the primary source ofincome to their families. Incarceration not onlylimits these contributions but also threatens the

earning power of remaining family members,who may sacrifice work time to perform taskspreviously done by the incarcerated father(Lynch & Sabol, 2004) or struggle to coverexpenses associated with his incarceration, suchas legal representation or maintaining contactthrough phone calls and visits (Comfort, 2008).

Furthermore, the family financial instabil-ity associated with a father’s incarceration islikely to persist past the time that he spendsin prison or jail. A substantial body of researchhas documented the labor market challenges thatex-prisoners face upon reentry (Bushway, Stoll,& Weiman, 2007; Holzer, 2007), and womenwith formerly incarcerated partners receive lessin either shared earnings or child support than dosimilarly situated women whose partners werenot incarcerated (Geller et al., 2011). The earn-ings challenges of formerly incarcerated men arecompounded by the tendency of their roman-tic relationships to dissolve (Western, 2006),reducing the likelihood that even their limitedearnings are shared with their former house-holds (Geller et al., 2011). On the basis of theseand other strains, a father’s incarceration mayalso elevate maternal stress levels (Wildeman,Schnittker, & Turney, 2012) and diminish moth-ers’ mental health (Fishman, 1990), reducingtheir ability to manage family resources. Accord-ingly, Schwartz-Soicher et al. (2011) foundthat families with incarcerated fathers experi-enced significantly more material hardship thanfamilies with no paternal incarceration.

In addition to causing financial and emotionalstrain, incarceration may limit the quantity andquality of housing available to prisoners andtheir families. Certain classes of offenders,most notably sex offenders, face restrictions onwhere they are allowed to live (Levenson &Cotter, 2005; Zandbergen & Hart, 2006; Zgoba,Levenson, & McKee, 2009). In addition, certaintypes of housing assistance (most notably, publichousing) may be unavailable to the families ofincarcerated individuals, due to a “One strike andyou’re out” style of regulation that authorizespublic housing authorities to evict and excludefrom the application process for a “reasonableamount of time” any household containing aperson with a background of criminal activitythat the public housing authority believeswould endanger the health or safety of thecommunity. Although the total number ofindividuals excluded from public housing due toone-strike policies is unknown, Human Rights

Page 3: Paternal Incarceration and the Housing Security of Urban Mothers

Paternal Incarceration and Maternal Housing 413

Watch (Carey, 2004) estimated that more than3.5 million people with criminal convictionswould be denied access to housing assistance asthe result of “one strike” policies over a 5-yearperiod. These individuals may have partnersand children at risk of losing access to publichousing, even if these family members haveno criminal histories of their own (Venkatesh,2002).

CONFOUNDING FACTORS

Despite the challenges faced by families ofincarcerated fathers, the extent of incarceration’scausal effect on partner housing insecurity—ifone exists at all—is unknown. The effect offathers’ incarceration on their partners and chil-dren is dependent on the relationship that fathershad with their families before going to jail orprison (Geller, Cooper, Garfinkel, Schwartz-Soicher, & Mincy, 2012; Western & Wildeman,2009). Criminal behavior is more prevalentamong unmarried men (Sampson & Laub, 1990),and although many unmarried and nonresidentfathers maintain involvement in the lives of theirchildren (Swisher & Waller, 2008; Tach et al.,2010), we know little about prior involvementamong fathers who become incarcerated, inparticular whether these fathers were involvedenough for their families to be affected by theirabsence (Sampson, 2011, though see Geller,2013). Moreover, many incarcerated fathershad complicated families of origin (Rumbaut,Gonzales, Komaie, Morgan, & Tafoya-Estrada,2006; Sampson & Laub, 1993); such com-plexity has also been associated with unstablerelationships in adulthood (Doucet & Aseltine,2003) and may attenuate effects of incarcerationon their partners or former partners.

Furthermore, as a population, incarceratedmen face a number of educational, cognitive,mental health, and socioeconomic challengeseven before their contact with the criminaljustice system (Western, 2006). They aredisproportionately young, are predominantlymembers of racial/ethnic minority groups,have low levels of education (Pew Centeron the States, 2008), and would likely facechallenges in the labor market even in theabsence of their incarceration, limiting theirfinancial contributions to their families. Manymen who become incarcerated also facechallenges related to substance use (Drucker,2011), mental health (Drucker, 2011), physical

health (Curtis, 2011; Patterson, 2010), andimpulse control (Farrington, 1998; Gottfredson& Hirschi, 1990), which have the potentialto both limit their earnings and destabilizetheir family relationships (Western, 2006).Furthermore, to the extent that women partnerwith men who are demographically and socio-ecomically similar (Vanyukov, Neale, Moss, &Tarter, 1996), those with incarcerated spousesand partners are likely to be vulnerable inthe housing market as well. Women withincarcerated partners have disproportionatelylow incomes, with few economic resources andlittle social capital (Sugie, 2012), and they facehigh rates of stress even before their partner’sincarceration (Wildeman et al., 2012). Womenwith incarcerated partners are also more likelyto themselves have histories of incarceration(Geller, Garfinkel, Cooper, & Mincy, 2009).It is unclear whether partner incarceration hasan independent effect that exacerbates thesechallenges or whether the observed disadvantageis driven solely by pre-incarceration conditions.In addition, to the extent that mothers withincarcerated partners receive greater levels ofpublic assistance (Sugie, 2012), any hardshipscreated by their partners’ incarceration may beattenuated.

EMPIRICAL EVIDENCE

Although a growing literature has documenteda nexus among incarceration, homelessness,and the housing insecurity of former prisoners(Geller & Curtis, 2011; Gowan, 2002; Greenberg& Rosenheck, 2008; Herbert, 2005; La Vigne& Parthasarathy, 2005; Lee et al., 2010;Metraux & Culhane, 2004; Metraux, Roman,& Cho, 2008; Travis, Solomon, & Waul,2001), far less is known about the effectof incarceration on the housing security ofthe family members of incarcerated men.However, recent research has suggested thatthese family members are at serious risk.Wildeman (2013) found strong associationsbetween recent paternal incarceration and therisk of homelessness among young urbanchildren. Foster and Hagan (2007) also identifiedpaternal incarceration as a key predictor of socialexclusion during the transition to adulthood and,notably, as predictive of having experiencedhomelessness by this point in the life course.Although these studies identified importantrisks facing the children of incarcerated men,

Page 4: Paternal Incarceration and the Housing Security of Urban Mothers

414 Journal of Marriage and Family

their focus on homelessness, the most severeform of insecurity, is likely to have missedsubstantial disruption resulting from otherchallenging housing situations (Geller & Curtis,2011). Other studies examined less severeindicators of housing insecurity, such as skippingrent payments (Schwartz-Soicher et al., 2011),eviction (Schwartz-Soicher et al., 2011), orresidential mobility (Phillips, Erkanli, Keeler,Costello, & Angold, 2006), but only ascomponent pieces of broader indicators ofmaterial hardship, rather than with a specificfocus on housing needs.

CONTRIBUTIONS OF THE CURRENT ANALYSIS

We advance the literature on incarcerationand housing by estimating the extent towhich mothers’ housing insecurity might becompromised by the incarceration of theirromantic partners. We focused on a broadindicator of insecurity, based on circumstancesrelatively common among low-income familiesas well as those that are rare and indicativeof extreme social exclusion. We controlled fordetailed measures of women’s socioeconomicdisadvantage as well as past housing insecurity toreduce the likelihood that observed relationshipsare confounded by factors other than theirpartners’ incarceration. In so doing, we sought toisolate the effects of paternal incarceration fromother factors that might compromise maternalhousing security. We tested five specifichypotheses:

Hypothesis 1: Mothers’ housing security is nega-tively associated with their partners’ incarceration,an effect felt most strongly for women living withtheir partners prior to the incarceration.

Hypothesis 2: The relationship between paternalincarceration and maternal housing security ismediated by fathers’ financial contributions, whichare undermined following an incarceration.

Hypothesis 3: The relationship between paternalincarceration and maternal housing security ismediated by maternal stress, which increasesfollowing fathers’ incarceration.

Hypothesis 4: Mothers in public housing are atincreased insecurity risk when their partners areincarcerated.

Hypothesis 5: Mothers receiving housing assis-tance (but not living in public housing) areprotected from the housing insecurity that mayfollow a father’s incarceration.

METHOD

Data Source, Analysis Sample, and MissingData

Data were drawn from the Fragile Families andChild Wellbeing study (hereafter Fragile Fami-lies), a population-based survey of urban coupleswith children. The Fragile Families study fol-lows a stratified random sample of nearly 5,000couples in large U.S. cities with children bornbetween 1998 and 2000 (see Reichman, Teitler,Garfinkel, & McLanahan, 2001, for a descrip-tion of the research design). The study wasdeveloped to allow researchers to understandthe capacities of and challenges facing unwedurban parents and contains detailed questions onthe role of fathers in family life and the socialand material well-being of fathers, mothers, andchildren.

The study oversamples unmarried parents,and the sample is highly socioeconomicallydisadvantaged, with a high prevalence ofincarceration among the fathers. More than40% of the fathers, including approximatelyhalf those unmarried at their child’s birth, havespent time in prison or jail. The fathers withno history of incarceration are also of relativelylow income, with low levels of education, andprovide a valuable comparison sample for theassessment of incarceration’s unique risks. Thestudy surveys both men and their partners at thetime of their child’s birth, with follow-up surveysconducted when the children are 1, 3, and 5 yearsold. Our analysis sample consisted of the 4,125mothers reporting on their housing security atYear 5. (When weighted to represent the 20Fragile Families cities, the analysis sample dropsto 4,041.)

Although our analysis sample consisted ofmothers who reported on their housing securityat the time of their child’s fifth birthday, ouranalysis may be vulnerable to selection bias ifmothers were missing data on key predictorsof Year 5 housing insecurity and the propen-sity for missing data was unevenly distributed.To the extent that mothers facing housing hard-ships are more difficult to retain in surveys,the role of incarceration as a determinant of

Page 5: Paternal Incarceration and the Housing Security of Urban Mothers

Paternal Incarceration and Maternal Housing 415

housing insecurity may be misstated in our esti-mates. We therefore assessed the sensitivity ofour findings to missing data using two methodsof analysis. First, we estimated a set of modelsthat used complete case analysis (also knownas listwise deletion), which dropped familiesfrom a regression model if they were missingdata on any variables in the model. Althoughcomplete case analysis has the potential to pro-duce unbiased coefficient estimates, this requiresthat data be missing “completely at random”(Allison, 2002). This is unlikely to be the casein a longitudinal survey where retention mighthave been affected by factors also related tofamily stability. We therefore also used an impu-tation procedure (specifically, multiple imputa-tion through chained equations—see Royston,2004, and Van Buuren, Boshuizen, & Knook,1999—and the ice and micombine commands inStata) to estimate missing values of incarcera-tion and father involvement indicators as well aspotential confounders. We examined the sensi-tivity of findings to our choice of a missing datastrategy.

Variables

Housing insecurity. We measured mothers’housing insecurity using indicators identifiedby Geller and Curtis (2011) that identifiedinsecurity at several different degrees of severitybased on mothers’ living conditions at the timeof each follow-up survey and on hardshipsshe reported experiencing in the year leadingup to her survey. Mothers were consideredinsecure if they indicated having skipped arent or mortgage payment due to a lackof funds, moving in with others due tofinancial constraints (also known as “doublingup”), moving residences more than once peryear in the past wave (Gilman, Kawachi,Fitzmaurice, & Buka, 2003), having beenevicted, or having been homeless (per 42 U.S.C.§ 11302). Our primary outcome of interestwas a constructed binary indicator of whetherrespondents indicated any of these insecuritiesover the year preceding their Year 5 survey. Wealso examined whether our findings were robustto our choice of outcome measure or if somecomponents of the measure (e.g., skipping a rentpayment, doubling up, etc.) were more closelyassociated with fathers’ incarceration thanothers.

Incarceration. Our measure of paternal incar-ceration was based on fathers’ self-reports,supplemented with additional indicators (e.g.,mother reports, subcontractor reports that fatherswere incarcerated when contacted for follow-up)to reduce the risk of measurement error associ-ated with underreporting (Groves, 2004). Ateach follow-up wave, fathers were asked to self-report whether they had been charged with acrime in the years leading up to the interview;if they had, they were asked if they had beenconvicted and, if they had, they were asked ifthey had been incarcerated.

We constructed measures of both “recentincarceration” (i.e., whether fathers had beenincarcerated in the period following the Year 1survey, leading up to the Year 5 survey) and“distal incarceration” (i.e., in the period leadingup to the Year 1 survey).

Covariates

As noted, families of incarcerated men arelikely to differ from other families in waysthat influence their housing circumstances. Ouranalyses therefore controlled for a series ofcovariates we expected to be correlated withboth fathers’ incarceration and mothers’ housingsecurity. We focused predominantly on maternalcharacteristics, which we posit are highlycorrelated with those of her partner (Vanyukovet al., 1996).

Covariates included mothers’ race, nativity,baseline age and education, and family history(i.e., family mental health history and whethershe was living with both her biological par-ents at age 15) as well as time-stable traitssuch as cognitive ability and impulsivity. Cog-nitive ability was measured using the WechslerAdult Intelligence Scale (Weschler, 2008), andimpulsivity was measured using the Dickman(1990) scale of dysfunctional impulsivity. Wemeasured mental health history using moth-ers’ self-reports of whether their own motherssuffered from depression in order to avoid con-founding with contemporaneous conditions. Wealso controlled for a rich set of family, behav-ioral, and economic characteristics, includingthe couple’s relationship status (married, cohab-iting, or nonresident) at the Year 1 follow-upsurvey, and several indicators of mothers’ phys-ical and mental health as well as substance use atthe time of their baseline and Year 1 surveys. Inaddition to measuring mothers’ family history of

Page 6: Paternal Incarceration and the Housing Security of Urban Mothers

416 Journal of Marriage and Family

mental health, we also constructed indicators ofself-reported health at baseline (scored 1 if theyreported “excellent” or “very good” health, and 0if they reported “good,” “fair,” or “poor” health),parenting stress, and perceived social support.Parenting stress was measured as an additivescale (α = .61 at Year 1) constructed from theextent to which mothers agreed or disagreed withfour statements: (a) “Being a parent is harderthan I thought [it would be],” (b) “I feel trappedby my parental responsibilities,” (c) “Takingcare of the children is more work than pleasure,”and (d) “I often feel tired and worn out from rais-ing my family.” Each individual item was codedon a scale from 1 to 4 (1 = strongly disagree,2 = disagree, 3 = agree, 4 = strongly agree), andresults were summed. Perceived social supportwas measured using a principal-components fac-tor (α = .74) that explained 58% of the variancein whether mothers reported having someonethey could count on to (a) loan them $200 in thenext year, (b) provide them a place to live in thenext year, (c) help with emergency child care,and (d) co-sign for a loan for $1000.

We constructed and controlled for severalindicators of mothers’ socioeconomic status aswell as the number of children in her household,to indicate how many people must be supportedat a given level of income. Socioeconomicstatus indicators included mothers’ employmentand earnings at Year 1 as well as the amountof money her partner had contributed to thehousehold in the first year of the study. Financialcontributions were computed in terms of sharedearnings for resident fathers and child supportfor nonresident fathers (Geller et al., 2011). Wealso controlled for whether mothers reportedowning their home at Year 1 or reported receiptof public assistance, housing-specific assistancein particular. Mothers were classified into oneof three mutually exclusive categories: (a) thoseliving in public housing, (b) those receivingvouchers or other housing assistance, and(c) mothers not receiving assistance with theirhousing costs. Finally, we controlled for whethermothers reported any history of incarceration bythe Year 5 survey.

Potential Mechanisms

As noted above and described in greater detailbelow, we assessed the plausibility of severalpotential mechanisms that may govern therelationship between incarceration and housing.

In addition to the Year 1 measures indicatedabove, we also constructed Year 5 measuresof fathers’ financial contributions and maternalstress, which we expected to mediate anyestimated effects.

Modeling Strategy

We assessed the relationship between fathers’incarceration and mothers’ housing securityusing a series of logistic regression modelsthat leveraged the longitudinal structure of theFragile Families data to examine how mothers’housing circumstances change following herpartner’s incarceration. A sensitivity analysisrunning comparable linear probability models,with substantively similar results, is availableon request. We focused our interpretation onwhat we refer to as “recent” incarceration,between the Year 1 and Year 5 survey waves,denoted as INC15, and the associated regressioncoefficient β1. To isolate the predictive role ofrecent partner incarceration from other factorsthat might influence insecurity, we controlled forthe covariates described above (X) as well as forpartner incarceration experiences that precededthe Year 1 survey (INC1):

Logit(INSECURE5) = β0 + β1INC15

+ β2INC1 + β3X + ε

To further isolate the effects of incarcerationfrom other factors that might influence mothers’housing insecurity, we estimated Model 2, whichcontrolled not only for covariate vectors Xbut also for a vector of Year 1 insecurityindicators, INSECURE1. In this model, β1identified changes in insecurity associated withincarceration incidents between Years 1 and 5:

Logit(INSECURE5) = β0 + β1INC15

+ β2INC1 + β3X + β4 INSECURE1 + ε

If unobserved changes in family or commu-nity circumstances lead to both a new incar-ceration experience and subsequent housinginsecurity, the estimate of β1 in Model 2 wouldreflect a spurious relationship in addition to anycausal one, overstating the effects of fathers’incarceration. We therefore took the coefficientβ1 in Model 3 as an upper bound, reduced-form estimate of the extent to which mothers’

Page 7: Paternal Incarceration and the Housing Security of Urban Mothers

Paternal Incarceration and Maternal Housing 417

housing insecurity may have been affected bytheir partner’s incarceration.

To the extent that Model 2 suggested aneffect of fathers’ incarceration on mothers’housing insecurity, we hypothesized, on thebasis of research conducted by Western andWildeman (2009) and Geller et al. (2012), thatestimated effects would be more pronouncedamong families in which the father wasmore involved before his incarceration. Wetested the plausibility of our Model 2 findingsusing a third model that estimated howthe association between paternal incarcerationand maternal housing might be moderatedby fathers’ pre-incarceration residence status.Model 3 controlled not only for Year 1relationship status but also for the interactionbetween Year 1 coresidence (RES1) andsubsequent incarceration (INC15). In this model,β1 represents the effect of a nonresidentpartner’s incarceration on housing insecurity,and the effect of incarceration of a residentpartner is denoted by the sum β1 + β5. Tothe extent that the coefficient β5 suggestedstronger associations for partners that werepreviously coresident, Model 3 would increaseour confidence that fathers’ incarceration had acausal effect on mothers’ housing security:

Logit(INSECURE5) = β0 + β1INC15

+ β2INC1 + β3X + β4INSECURE1

+ β5INC15 × RES1 + ε

We hypothesized that our estimates in Models1 through 3 would suggest a robust associationbetween fathers’ incarceration and mothers’subsequent housing insecurity. Accordingly, wenext estimated a series of models examiningpotential mediators and moderators of therelationship between incarceration and ouraggregate measure of housing insecurity. Ourtests for mediation and moderation werevariations of Model 2, which estimated theaverage effect of paternal incarceration onmothers’ housing outcomes. We assessed theplausibility of each potential mediator using amodification of a Sobel–Goodman mediationtest, focusing specifically on linear probabilitymodels and the complete case sample. (TheStata sgmediation test precluded the test ofmediation in nonlinear models or modelsbased on multiply imputed data sets.) TheSobel–Goodman test identified the extent to

which (a) fathers’ incarceration was associatedwith a given potential mediator, (b) fathers’incarceration was associated with mothers’housing insecurity when the mediator was notconsidered, (c) the unique association betweenthe potential mediator and mothers’ housinginsecurity, and (d) the extent to which theassociation between incarceration and housinginsecurity was reduced when the potentialmediator is considered. Although controls forposttreatment circumstances must be interpretedwith caution and cannot be assumed to representa causal mechanism (Gelman & Hill, 2007), thistest helped to assess the plausibility of potentialmediators whose causal effects can be exploredin future research.

We first tested the hypothesis that fathers’incarceration might undermine their partners’housing security through a reduction in house-hold income. Model 4 included an additionalcontrol for fathers’ financial contributions atYear 5:

Logit(INSECURE5) = β0 + β1INC15 + β2INC1

+ β3X + β4INSECURE1

+ β5CONTRIB5 + ε

We next tested the extent to which fathers’incarceration might undermine their partners’housing insecurity through mental health strains.Model 5 added a control for STRESS5, maternalstress at Year 5, and we tested its plausibility asa mediator using the sgmediation test:

Logit(INSECURE5) = β0 + β1INC15 + β2INC1

+ β3X + β4INSECURE1

+ β5STRESS5 + ε

Finally, we tested the extent to whichincarceration’s effects might be tied to housingpolicy by testing the moderating role of publichousing and other forms of housing assistance.Model 6 returned to the imputation sample andlogit functional form and included an interactionbetween fathers’ incarceration between Years 1and 5 and mothers’ residence in public housingat Year 1 (PH1) and the receipt of otherhousing assistance at Year 1 (VOUCHER1). Thecoefficient β1 represents the estimated effects ofpartner incarceration on the insecurity of womennot receiving housing assistance at Year 1, the

Page 8: Paternal Incarceration and the Housing Security of Urban Mothers

418 Journal of Marriage and Family

estimated effect of partner incarceration onwomen in public housing is denoted by β1 + β5,and the effect on women receiving other housingassistance is estimated by β1 + β6. As noted,we hypothesized that β5 would be positive,indicating increased insecurity among motherssubject to the scrutiny of public housing, but thatβ6 would be negative, reflecting a protective roleof other housing assistance against the insecurityassociated with incarceration:

Logit(INSECURE5) = β0 + β1INC15

+ β2INC1 + β3X

+ β4INSECURE1 + β5INC15 × PH1

+ β6INC15 ∗ VOUCHER1 + ε

Sensitivity Analyses

As noted above, our primary estimates ofincarceration’s effect on housing insecurity werefocused on an aggregate measure of insecurityand a sample constructed through multipleimputation. We assessed the sensitivity of ourfindings to our choice of outcome measureand missing-data approach. After estimatingthe associations between fathers’ incarcerationand our aggregate measure of mothers’ housinginsecurity, we estimated the extent to whichincarceration was associated with each of thecomponent indicators of insecurity, replicatingeach model to predict, in turn, each of thefive indicators of insecurity. We also assessed

the sensitivity of findings to our missing dataapproach by reestimating models on a completecase sample.

RESULTS

Sample Description

A summary of maternal housing insecurityamong our analysis sample is provided inTable 1, underscoring the prevalence of housinginsecurity facing urban mothers. Our data,weighted to be representative of the 20 FragileFamilies cities, showed that nearly one fifth(18%) of mothers in these cities reported someform of insecurity around the time of theirchild’s fifth birthday. However, the prevalenceof housing insecurity varied substantially bydomain; the most common type of insecurity,having skipped a rent or mortgage payment,was experienced by 11% of urban mothers,and the most rare type of insecurity, eviction,was experienced by only 2%. For eachtype of housing insecurity, Table 1 identifiessignificant differences in maternal housinginsecurity by partner incarceration history, withmothers whose partners had been incarceratedsignificantly more likely to experience eachform of housing insecurity than mothers whosepartners had no history of incarceration.

Table 1 also suggests that the differencesbetween our two analysis samples were onlyslight. Because both the complete-case andimputed samples focused on mothers reportingon all components of housing insecurity, the

Table 1. Mothers’ Year 5 Housing Insecurity by Partner Incarceration Status

Partner incarceration status

Complete cases Imputed

Insecurity typeFull sample(N = 4,041)

Partner everincarcerated(n = 1,808)

Partner neverincarcerated(n = 1,817)

Partnerincarceration

unknown (n = 416)Partner everincarcerated

Partner neverincarcerated

Any insecurity 18 28 13 22 28 15Skipped payment 11 15 9 12 15 9Doubled up 6 11 4 7 11 4Moved > 1× per year 4 7 2 5 7 2Homelessness 3 5 1 6 6 2Evicted 2 3 1 4 3 2

Note: All table values are percentages. All differences between ever-incarcerated and never-incarcerated cases arestatistically significant at p ≤ .001. In weighted and imputed data, the number of incarcerated fathers varies from 1,103 to1,129; the number of never-incarcerated fathers varies from 2,912 to 2,938 across the five imputed data sets. Data are weightedto represent 20 Fragile Families cities.

Page 9: Paternal Incarceration and the Housing Security of Urban Mothers

Paternal Incarceration and Maternal Housing 419

Table 2. Mothers’ Demographic and Socioeconomic Background, by Partner Incarceration Status

Ever incarcerated Never incarceratedVariable M (SD) or % M (SD) or %

Mother raceWhite 17% 34%Black 53% 27%Hispanic 27% 30%Other 3% 9%

Mother foreign born 11% 32%Mother age 24.0 (5.8) 28.3 (6.0)Mother lived with both parents 33% 62%Mother cognitive ability (range: 0–15) 6.5 (2.6) 7.1 (2.9)Mother impulsivity (range: 0–6) 0.91 (0.74) 0.74 (0.72)Family history of depression 31% 22%Married (Year 1 [Y1]) 20% 69%Cohabiting (Y1) 27% 16%Nonresident (Y1) 53% 16%Mother completed less than high school (baseline [BL]) 43% 24%Mother high school graduate, not college graduate (BL) 55% 49%Mother college graduate (BL) 2% 27%Father’s financial contributions (Y1) $3,158 ($4,536) $10,856 ( $12,411)Maternal employment (Y1) 41% 52%Maternal earnings (Y1) $8,219 ($10,387) $16,072 ($24,494)Baseline maternal health (“excellent” or “very good”) 60% 72%Mother’s substance use (BL) 5% 1%Father’s substance use (BL) 16% 3%Maternal benefit receipt (Y1) 55% 17%Number of children (Y1) 2.3 (1.6) 2.0 (1.2)Maternal home ownership (Y1) 10% 32%Mother in public housing (Y1) 19% 7%Other maternal housing assistance (Y1) 10% 4%Maternal social support −0.27 (1.17) 0.10 (0.70)Any maternal incarceration by Year 5 12% 4%

Note: Percentage sums (race, education, and relationship status) may not total 100% due to rounding. All differencesare statistically significant at p < .001 with the exception of Hispanic race. Mothers with and without partner histories ofincarceration were equally likely to be Hispanic. In weighted and imputed data, the number of incarcerated fathers varies from1,103 to 1,129; the number of never-incarcerated fathers varies from 2,912 to 2,938 across the five imputed data sets. Data areweighted to represent 20 Fragile Families cities.

full-sample rates of housing insecurity wereequal across samples; the difference between the“partner ever incarcerated” and “partner neverincarcerated” rates was driven by the allocationof mothers whose partner incarceration historieswere unknown in the complete-case sample.Imputing the incarceration histories of thesefathers suggested virtually no changes inrates of insecurity among women with eitherincarcerated or never-incarcerated partners.Subsequent tables therefore present results basedon the imputation sample, with complete case

results discussed only in terms of deviationsfrom the imputation findings.

Although rates of housing insecurity weresignificantly higher among mothers whosepartners had histories of incarceration, thesehardships were likely driven not only bythe incarceration itself but also by othersocioeconomic challenges faced by incarceratedfathers and their families. As shown inTable 2, women whose partners had historiesof incarceration faced significant disadvantagesin addition to their increased housing insecurity;they were younger when the focal child was

Page 10: Paternal Incarceration and the Housing Security of Urban Mothers

420 Journal of Marriage and Family

born, were less likely to have been marriedat the time of the birth, and were morelikely to live apart from their child’s father.They faced higher rates of depression amongtheir own mothers, displayed higher levels ofimpulsivity, scored lower on cognitive tests,and had lower levels of educational attainmentand employment. They reported higher rates ofsubstance use and worse health. They had morechildren, were less likely to be homeowners,reported lower levels of social support, and weremore likely to have histories of incarceration.Each of these factors, significant at p < .05,might have compromised mothers’ housingsecurity even in the absence of their partners’incarceration. On the other hand, mothers withincarcerated partners were more likely to live inpublic housing at Year 1 and reported higherrates of both housing assistance and publicassistance more generally, suggesting greaterdisadvantage, but also that welfare benefitsmight have helped to mitigate socioeconomichardships. Table 2 underscores the need tocontrol for the numerous socioeconomic factorsthat might confound estimates of the causaleffect of fathers’ incarceration on mothers’insecurity.

Insecurity Risk

Table 3 presents odds ratios (ORs) indicatingthe estimated increase in the odds of housinginsecurity among women with histories ofpartner incarceration, suggesting that womenexperiencing their partners’ incarceration facedsignificantly elevated risks. Model 1 suggestedthat women whose partners were recentlyincarcerated faced odds of insecurity nearly50% higher (OR = 1.49) than women whosepartners were not recently incarcerated. Thisdifference was of smaller magnitude than thatnoted in Table 1 (due to covariate adjustment),but substantial as well as highly statisticallysignificant. The estimated difference associatedwith recent partner incarceration was above andbeyond that associated with her own lifetimehistory of incarceration and of approximatelyequal magnitude. (Note that mothers’ ownincarceration histories were independentlyassociated with increased housing insecurity.)

Model 1 also suggested that several othersocioeconomic factors were associated withmothers’ housing insecurity. Women displayinghigher levels of impulsivity were significantly

more likely to report housing insecurity, as werewomen with family histories of depression andthose reporting higher levels of parenting stressat Year 1. On the other hand, women who hadgraduated college reported significantly lowerrates of housing insecurity, as did those reportinghigher rates of social support. Year 1 homeown-ership was associated with a marginal reductionin mothers’ odds of housing insecurity at Year 5,as was mothers’ baseline health. It is importantto note that although these associations neednot reflect causal relationships (e.g., womenwith greater stability in their personal lives mayhave been better positioned both to completetheir education and to maintain their housingsecurity), controlling for other aspects of stabil-ity reduced the likelihood that the relationshipbetween incarceration and housing insecuritywas driven by unobserved heterogeneity.

Model 2 suggested that the increased risk ofinsecurity among women with incarcerated part-ners was not driven by housing insecurity thatpreceded the Year 1 survey or any subsequentincarceration of their partner. The increased oddsof insecurity associated with recent incarcera-tion did not substantially change when earlierhousing insecurity was considered (OR = 1.49);covariate associations were slightly altered(most often, reduced) in both magnitude andsignificance, but qualitative results remainedsimilar. Using repeated measures of housinginsecurity allowed the observation of changes,increasing our confidence that the observed riskof mothers’ insecurity associated with their part-ners’ recent incarceration represented a causaleffect.

Model 3 further tested the plausibility ofour estimated relationships by examining theextent to which the association between fathers’incarceration and mothers’ experience ofinsecurity was moderated by pre-incarcerationcoresidence. As noted earlier, we hypothesizedthat paternal incarceration was most disruptiveto families who had been living together beforethe father’s time in prison or jail. This hypothe-sis was supported by Model 3, which suggesteda significantly stronger incarceration–insecurityrelationship for mothers whose partners werecoresident prior to their incarceration (interac-tion OR = 1.51). In fact, Model 3 suggested thatthe association between incarceration and hous-ing insecurity was only statistically significantfor couples who had been coresident prior to theincarceration.

Page 11: Paternal Incarceration and the Housing Security of Urban Mothers

Paternal Incarceration and Maternal Housing 421

Table 3. Summary of Logistic Regression Analysis of Mothers’ Year 5 Housing Insecurity (Any): Effect Estimation

Model 1 Model 2 Model 3

Socioeconomiccontrols only

Socioeconomic controlsand Y1 housing

insecurityIncluding interaction

by Y1 residence

Predictor b SE OR b SE OR b SE OR

Recent incarceration (Y1–Y5) 0.40∗∗∗ 0.11 1.49 0.40∗∗∗ 0.11 1.49 0.22 0.14 1.25Distal incarceration (by Y1) −0.03 0.12 0.97 −0.09 0.12 0.91 −0.08 0.12 0.92Incarceration (Y1–Y5) × resident (Y1) 0.41∗ 0.19 1.51Early life covariates

Black 0.01 0.13 1.01 0.09 0.13 1.09 0.08 0.13 1.08Hispanic −0.13 0.15 0.88 −0.10 0.15 0.90 −0.10 0.15 0.90Other race 0.27 0.25 1.31 0.30 0.25 1.35 0.28 0.26 1.32Mother foreign born 0.05 0.15 1.05 0.05 0.15 1.05 0.07 0.15 1.07Mother age −0.01 0.01 0.99 −0.01 0.01 0.99 −0.01 0.01 0.99Mother lived with both parents −0.15 0.09 0.86 −0.13 0.09 0.88 −0.13 0.09 0.88Mother cognitive ability (range: 0–15) 0.04 0.02 1.04 0.03 0.02 1.03 0.03† 0.02 1.03Mother impulsivity (range: 0–6) 0.16∗∗ 0.06 1.17 0.18∗∗ 0.06 1.20 0.19∗∗∗ 0.06 1.21Family history of depression 0.61∗∗∗ 0.09 1.84 0.54∗∗∗ 0.10 1.72 0.54∗∗∗ 0.10 1.72

Contemporaneous covariatesCohabiting 0.04 0.13 1.04 0.04 0.13 1.04 0.01 0.13 1.01Nonresident 0.01 0.15 1.01 −0.02 0.15 0.98 0.12 0.16 1.13Less than high school −0.01 0.09 0.99 −0.01 0.10 0.99 −0.02 0.10 0.98College graduate −0.43∗ 0.21 0.65 −0.39† 0.21 0.68 −0.28† 0.21 0.76Father’s contributions (Y1) −0.01 0.01 0.99 −0.01 0.01 0.99 −0.01 0.01 0.99Maternal employment (Y1) 0.05 0.10 1.05 0.04 0.10 1.04 0.04 0.10 1.04Maternal earnings (Y1) 0.00 0.00 1.00 0.00 0.00 1.00 0.00 0.00 1.00Baseline maternal health −0.14† 0.09 0.87 −0.14 0.09 0.87 −0.15† 0.09 0.86Maternal stress (Y1) 0.16∗ 0.07 1.17 0.10 0.08 1.11 0.10 0.08 1.11Mother’s substance use (BL) 0.07 0.17 1.07 −0.03 0.17 0.97 −0.04 0.17 0.96Father’s substance use (BL) 0.14 0.13 1.15 0.09 0.13 1.09 0.08 0.13 1.08Maternal benefit receipt (Y1) 0.03 0.10 1.03 −0.06 0.11 0.94 −0.07 0.11 0.93Number of children (Y1) 0.04 0.03 1.04 0.03 0.03 1.03 0.03 0.03 1.03Maternal home ownership (Y1) −0.28 0.16 0.76 −0.16 0.16 0.85 −0.15 0.16 0.86Mother in public housing (Y1) −0.16 0.14 0.85 0.00 0.14 1.00 0.00 0.14 1.00Other housing assistance (Y1) 0.22 0.17 1.25 −0.07 0.18 0.93 −0.07 0.18 0.93Maternal social support −0.19∗∗∗ 0.05 0.83 −0.14∗∗ 0.05 0.87 −0.14∗∗ 0.05 0.87Any maternal incarceration (Year 5) 0.40∗∗ 0.14 1.49 0.35∗ 0.15 1.42 0.34∗ 0.15 1.40Skipped rent (Y1) 0.81∗∗∗ 0.12 2.25 0.81∗∗∗ 0.12 2.25Evicted (Y1) 0.38 0.24 1.46 0.38 0.23 1.46Doubled up (Y1) 0.64∗∗∗ 0.13 1.90 0.64∗∗∗ 0.13 1.90Homeless (Y1) −0.11 0.29 0.90 −0.12 0.29 0.89Moved > 1× per year (Y1) 0.38∗∗ 0.13 1.46 0.38∗∗ 0.13 1.46

Pseudo R2 range 0.07–0.08 0.10–0.11 0.11

Note: N = 4,125 in each imputed data set. OR = odds ratio; Y1 = Year 1; Y5 = Year 5; BL = baseline.†p < .10. ∗p < .05. ∗∗p < .01. ∗∗∗p < .001.

Potential Mechanisms

Having estimated significant associationsbetween fathers’ incarceration and mothers’subsequent experience of housing insecurity,

we examined potential mechanisms thatmight have governed these relationships. InTable 4 we present tests of potential mediatorsand moderators of the relationship between

Page 12: Paternal Incarceration and the Housing Security of Urban Mothers

422 Journal of Marriage and Family

Table 4. Results of Regression Analyses Examining Potential Mediators and Moderators of the Relationship BetweenFathers’ Incarceration (Year 1–Year 5 [Y1–Y5]) and Mothers’ Housing Insecurity and Results of Sobel–Goodman Tests for

Mediation

Model 2(restated)

Model 4(+ Y5 contributions)

Model 5(+ Y5 stress)

Model 6(housing assistance)

Predictor b SE OR b SE OR b SE OR b SE OR

Logit model results(based on multiple imputation samples, N = 4,125 per sample)

Recent incarceration(Y1–Y5)

0.40∗∗∗ 0.11 1.49 0.29∗∗ 0.11 1.33 0.38∗∗∗ 0.11 1.49 0.46∗∗∗ 0.12 1.58

Fathers’ Y5 contributions −0.07∗∗∗ 0.02 0.93Maternal stress (Y5) 0.34∗∗∗ 0.07 1.41Incarceration (Y1–Y5)

× PH (Y1)−0.23 0.23 0.79

Incarceration (Y1–Y5)× other assistance (Y1)

−.021 0.39 0.81

Results of Sobel–Goodman mediation tests(based on linear probability model with complete case sample, N = 1,849)

Unmediated estimate 0.06∗ 0.03 0.50† 0.02Estimated incarceration

effect on mediator−1.82∗∗∗ 0.37 0.11∗∗ 0.04

Estimated effect ofmediator on insecurity

−0.01∗∗∗ 0.00 0.06∗∗∗ 0.02

Estimated indirect effect 0.01∗ 0.00 0.01∗ 0.00Estimated direct effect 0.05∗ 0.03 0.04† 0.02Proportion of total effect

that is mediated0.21 0.13

Note: Models 4, 5, and 6 include all covariates included in Model 2 and displayed in Table 3. Results are based onunweighted data.

†p < .10. ∗p < .05. ∗∗p < .01. ∗∗∗p < .001.

fathers’ incarceration and mothers’ housinginsecurity.

In the upper panel of Table 4, the first set ofcolumns restates the results of Model 2, indi-cating that mothers with recently incarceratedpartners faced odds of housing insecurity 49%higher than those of mothers whose partnerswere not incarcerated. The second set of columnspresents the results of Model 4, which added anadditional control for fathers’ Year 5 contribu-tions. Model 4 suggested that fathers’ financialcontributions mediated a portion of incarcera-tion’s effects on mothers’ housing insecurity:Fathers’ financial contributions were associatedwith a significant reduction in mothers’ odds ofinsecurity, and the coefficient on recent incarcer-ation was reduced from .40 to .29 when fathers’post-incarceration contributions were consid-ered. This mediating role was further supportedby the Sobel–Goodman test (in the lower panelof the table), which found reduced financial

contributions among formerly incarceratedfathers and an estimated protective effect offinancial contributions on mothers’ housing inse-curity. However, even when financial contribu-tions were considered, incarceration maintaineda significant independent association with subse-quent housing insecurity. The Sobel–Goodmantest suggested that 21% of incarceration’s totalassociation with mothers’ housing insecuritywas explained by reductions in fathers’ financialcontributions and that other mechanisms werealso likely at play.

The third set of columns in Table 4 presentsresults from Model 5, which examined Year5 maternal stress as a potential mediator ofincarceration’s effects on housing insecurityand found its mediating role to be limited.Although maternal stress was a significantpredictor of mothers’ housing insecurity,the Sobel–Goodman mediation test foundthat maternal stress explained only 13% of

Page 13: Paternal Incarceration and the Housing Security of Urban Mothers

Paternal Incarceration and Maternal Housing 423

incarceration’s total association with housinginsecurity, again suggesting that other mecha-nisms may have been at play.

Finally, the last columns of Table 4 presentresults from Model 6, which tested the plau-sibility of housing policy as a determinantof mothers’ housing security following theirpartners’ incarceration. Specifically, we hypoth-esized that the risks associated with fathers’incarceration would be exacerbated for mothersin public housing, who were subject to regula-tions such as one-strike enforcement but that theeffects of incarceration would be mitigated formothers receiving other forms of housing assis-tance. Model 6 provided little support for thesehypotheses, as neither public housing nor otherhousing assistance was a significant moderatorof incarceration’s estimated effect. Moreover,the interaction between incarceration and pub-lic housing was associated with diminishedinsecurity rather than more.

DISCUSSION

Summary of Findings

The goal of our study was to assess the effects offathers’ incarceration on the housing security oftheir families and specifically of the women withwhom they have children. As demonstrated inTable 3, we found strong and robust associationssuggestive of adverse effects: Mothers withrecently incarcerated partners, on average, facedapproximately 50% greater odds of housinginsecurity than other mothers. The negativeassociations between fathers’ incarceration andmothers’ housing security were concentrated infamilies in which the father had been residentprior to his incarceration, which supportedthe first of our entering hypotheses and wasconsistent with prior research suggesting afather’s removal from the household as a driverof incarceration’s adverse effects on families(Geller et al., 2012).

Our examination of other potential mech-anisms suggested only limited support for ourmediation and moderation hypotheses (Hypothe-ses 2–5). As noted above, fathers’ financialcontributions were a significant predictor ofmothers’ housing insecurity and explained 21%of the association between mothers’ insecu-rity and fathers’ incarceration. However, theirmediating role was limited: The majorityof the association was left unexplained whencontributions were considered. Maternal stress

was even less closely associated with theincarceration–insecurity relationship, and wefound little evidence of moderation of theincarceration–housing relationship by eithermothers’ residence in public housing or otherreceipt of housing assistance. However, it bearsnoting that the interaction coefficients representan average level of moderation and may maska heterogeneous relationship whereby housingassistance was protective in some circumstancesbut risky in others.

Sensitivity to Missing-Data Approach

Although the majority of results we havepresented were based on a multiple-imputationsample, our key findings were largely similarwhen models were estimated for a completecase sample. (Detailed results are available onrequest.) In our complete case analysis, Models 1and 2 suggested relationships that were reducedin magnitude and only marginally significant(ORs reduced from approximately 1.49 to 1.35,p = .057); this reduced significance was duein part to the dropped cases in the completecase sample (N = 1,852). Model 3 suggested astatistically significant association (OR = 1.81,p < .01) among mothers whose partners hadpreviously been coresident, but not those wholived separately from their partners, suggestingthat our key substantive finding of an adverserelationship concentrated in formerly coresidentfamilies was observed in both the complete caseand imputed data sets.

Sensitivity to Outcome Choice

As noted, our outcome of primary interest was anaggregate indicator of whether mothers experi-enced “any insecurity” (among five components)in the year leading up to their Year 5 interviews.Examining each type of insecurity (results avail-able on request), we found that our generalobservation of compromised housing securityamong mothers with recently incarcerated part-ners is robust across outcomes. However, themagnitude and significance of these associationsvaried. Using the aggregate measure, we foundthat mothers with recently incarcerated partnersfaced odds of insecurity that were nearly 50%greater than mothers whose partners were neverincarcerated (p < .001). This association wasslightly diminished for skipping a rent payment,the most common type of insecurity (OR = 1.33,

Page 14: Paternal Incarceration and the Housing Security of Urban Mothers

424 Journal of Marriage and Family

p = .029). Eviction, moving more than onceper year, doubling up, and homelessness wereeach more closely associated with partner incar-ceration than the aggregate measure of inse-curity (adjusted ORs = 1.71, 1.75, 1.77, and2.36, respectively). The statistical significance ofthese relationships also varied: Frequent moves,doubling up, and homelessness were each signif-icantly associated with incarceration at α = .05or less, whereas eviction was only marginallyassociated with incarceration (p = .059).

When we examined differences in theincarceration–insecurity relationship by fathers’pre-incarceration residence, we found that, likeour aggregate indicator, skipping a rent paymentand moving more than once per year were eachsignificantly associated with incarceration onlyfor families in which the incarcerated fatherhad been coresident prior to his incarceration.Eviction, only marginally associated withincarceration overall, was limited to a marginalassociation for formerly coresident motherand was not significantly associated withincarceration for mothers who had not beenliving with their partners in the wave beforetheir incarceration. Each of these findingsunderscores the potentially destabilizing effectof an incarcerated father’s removal from thefamily household. On the other hand, doublingup was associated with incarceration for familiesliving separately as well as those previouslycoresident, and mothers’ experiences of home-lessness, the most serious form of housinginsecurity, were associated with incarcerationonly for women who had not been living withtheir partner before his incarceration.

Limitations and Directions for Future Research

Although we observed significant associationsbetween fathers’ incarceration and mothers’subsequent housing insecurity that were of con-siderable magnitude and largely robust, ourfindings must be interpreted with caution. Ascer-taining causal effects from observational datapresents challenges because observed increasesin mothers’ housing insecurity may reflect aspurious relationship driven by unobserved dif-ference between families with and withoutpaternal incarceration histories. This concernis particularly relevant to our homelessnessfindings. We had hypothesized that the effectsof paternal incarceration on maternal housingwould be greatest for couples that had previ-ously been coresident. Although this was the

case for several domains of housing insecurity,we found that the increased risk of homelessnesswas greatest for nonresident partners, suggestingthat the experience of homelessness may havebeen tied to other aspects of family disadvantagerather than the shock of incarceration.

This counterintuitive finding points to twodirections for future research. The first is ananalysis more focused on causal inference. Weconsider our current estimates to be an upperbound on incarceration’s causal effects, butan analysis examining exogenous shocks to afamily’s risk of paternal incarceration is likelyto more closely identify incarceration’s uniqueeffect. The second direction we propose forfuture research is a detailed analysis of mothers’experiences of different domains of housinginsecurity. A growing literature suggests thatincarceration undermines family income (Gelleret al., 2011), which may in turn affect the abilityto pay rent (Phillips et al., 2006; Schwartz-Soicher et al., 2011). However, the links betweenincome loss and more severe experiences ofhousing insecurity, such as homelessness, maybe mediated by social support and other factors.Although the current analysis drew on a richset of observable control variables (includingsocial support) and presents preliminary testsof mediation and moderation, a large portionof the relationship between incarceration andhousing insecurity remains unexplained. Thesecomplex relationships would benefit fromfurther analysis, potentially using a structuralequation modeling framework.

Finally, our analysis was limited by alack of detailed information about the fathers’experience in the criminal justice system. Weknew little about the crime for which fatherswere incarcerated, the length of time they spentincarcerated, whether they were incarceratedin prison or jail, or how recently they werereleased. The effects of incarceration are likelyto vary widely on the basis of these factors(Comfort, 2007; Metraux et al., 2008). Althoughour estimates represented average effects acrossa range of circumstances, the exploration ofeffect heterogeneity is an important direction forfuture research.

Conclusions and Policy Implications

Despite these limitations, our findings add tothe growing literature documenting the extremedisadvantage facing the millions of families

Page 15: Paternal Incarceration and the Housing Security of Urban Mothers

Paternal Incarceration and Maternal Housing 425

with currently or formerly incarcerated fathers,and they provide guidance for policymakers,criminal justice practitioners, and social serviceproviders who wish to strengthen familiesfollowing a father’s incarceration. Specifically,the partially mediating role of fathers’ financialcontributions suggests that efforts to increasethe consistency of monetary support couldhave substantial payoff for family stabilityand housing security in particular. Previouslycoresident couples may be helped by short-termrent subsidies that enable them to maintaintheir ability to pay rent and stay in theirhomes following the loss of a breadwinner toincarceration. Increasing attention is also beingpaid to the child support system, one of the fewsystems to administratively connect incarceratedfathers to their partners and children, as a modeof stabilizing the income of families in whichnonresident fathers are incarcerated (CDCRToday, 2011). Housing-focused subsidies andconsistent child support payments have thepotential to help vulnerable families, includingthe partners of incarcerated men, maintainconnections to their local communities and betterhandle the other challenges they face.

NOTE

Research for this article was supported by a grant from theRobert Wood Johnson Foundation. The Fragile Familiesand Child Wellbeing Study was supported by GrantR01HD36916 from the Eunice Kennedy Shriver NationalInstitute of Child Health and Human Development (NICHD),and the project described in this article was supported byAward Number R24HD058486 from NICHD. The contentis solely the responsibility of the authors and does notnecessarily represent the official views of the NICHD orthe National Institutes of Health. For a complete list offunders of the Fragile Families and Child Wellbeing Study,see http://www.fragilefamilies.princeton.edu/funders.asp.

We are grateful to Marah Curtis, Christopher Wildeman,and colloquium participants at New York University forvaluable feedback on preliminary analyses. Anahid Modrekand Chelsea Davis provided valuable research assistance.

REFERENCES

Allison, P. D. (Ed.). (2002). Missing data. ThousandOaks, CA: Sage.

Bradley, K. H., Oliver, R. B., Richardson, N. C., &Slayter, E. M. (2001). No place like home: Housingand the ex-prisoner. Issue brief, CommunityResources for Justice, Boston.

Buckner, J. C. (2008). Understanding the impact ofhomelessness on children: Challenges and futureresearch directions. American Behavioral Scien-tist, 51, 721–736. doi:10.1177/0002764207311984

Bushway, S., Stoll, M. A., & Weiman, D. F. (Eds.).(2007). Barriers to reentry? The labor marketfor released prisoners in post-industrial America.New York: Russell Sage Foundation.

Carey, C. A. (2004). No second chance: Peoplewith criminal records denied access to pub-lic housing. Human Rights Watch. Retrievedfrom http://www.hrw.org/reports/2004/usa1104/usa1104.pdf

CDCR Today. (2011, March 21). Child supportobligations education begins for Californiainmates [Press release]. Retrieved from http://cdcrtoday.blogspot.com/2011/03/child-support-obligations-education.html

Comfort, M. (2007). Doing time together: Love andfamily in the shadow of the prison. Chicago:University of Chicago Press.

Cunningham, M., Harwood, R., & Hall, S. (2010).Residential instability and the McKinney–VentoHomeless Children and Education Program: Whatwe know, plus gaps in research. Research brief,The Urban Institute. Washington, DC.

Curtis, M. A. (2011). The effect of incarcerationon urban fathers’ health. American Journalof Men’s Health, 5, 341–350. doi:10.1177/1557988310385104

Dickman, S. J. (1990). Functional and dysfunctionalimpulsivity: Personality and cognitive correlates.Journal of Personality and Social Psychology, 58,95–102. doi:10.1037/0022-3514.58.1.95

Doucet, J., & Aseltine, R. H. (2003). Childhood familyadversity and the quality of marital relationshipsin young adulthood. Journal of Social andPersonal Relationships, 20, 818–842. doi:10.1177/0265407503206006

Drucker, E. (2011). A plague of prisons. New York:New Press.

Farrington, D. P. (1998). Predictors, causes, andcorrelates of male youth violence. Crime andJustice, 24, 421–475.

Fishman, L. T. (1990). Women at the wall. Albany:State University of New York Press.

Foster, H., & Hagan, J. (2007). Incarceration and inter-generational social exclusion. Social Problems, 54,399-433. doi:10.1525/sp.2007.54.4.399

Geller, A. (2013). Paternal incarceration andfather–child contact in fragile families. Jour-nal of Marriage and Family, 75, 1288–1303.doi:10.1111/jomf.12056

Geller, A., Cooper, C. E., Garfinkel, I., Schwartz-Soicher, O., & Mincy, R. B. (2012). Beyondabsenteeism: Father incarceration and child devel-opment. Demography, 49, 49–76. doi:10.1007/s13524-011-0081-9

Geller, A., & Curtis, M. A. (2011). A sort ofhomecoming: Incarceration and the housingsecurity of urban men. Social ScienceResearch, 40, 1196–1213. doi:10.1016/j.ssresearch.2011.03.008

Page 16: Paternal Incarceration and the Housing Security of Urban Mothers

426 Journal of Marriage and Family

Geller, A., Garfinkel, I., Cooper, C. E., & Mincy, R. B.(2009). Parental incarceration and child wellbeing:Implications for urban families. Social ScienceQuarterly, 90, 1186–1202. doi:10.1111/j.1540-6237.2009.00653.x

Geller, A., Garfinkel, I., & Western, B. (2011).Paternal incarceration and support for childrenin fragile families. Demography, 48, 25–47.doi:10.1007/s13524-010-0009-9

Gelman, A., & Hill, J. (2007). Data analysis usingregression and multilevel/hierarchical models.New York: Cambridge University Press.

Gilman, S. E., Kawachi, I., Fitzmaurice, G. M., &Buka, L. (2003). Socio-economic status, familydisruption, and residential stability in childhood:Relation to onset, recurrence, and remission ofmajor depression. Psychological Medicine, 33,1341–1355. doi:10.1093/ije/31.2.359

Gottfredson, M., & Hirschi, T. (1990). A generaltheory of crime. Stanford, CA: Stanford UniversityPress.

Goux, D., & Maurin, E. (2005). The effect ofovercrowded housing on children’s performance atschool. Journal of Public Economics, 89, 797–819.doi:10.1016/j.jpubeco.2004.06.005

Gowan, T. (2002). The nexus: Homelessness andincarceration in two American cities. Ethnography,3, 500–534. doi:10.1177/1466138102003004007

Greenberg, G. A., & Rosenheck, R. A. (2008). Jailincarceration, homelessness, and mental health: Anational study. Psychiatric Services, 59, 170–177.doi:10.1176/appi.ps.59.2.170

Groves, R. M. (2004). Survey errors and survey costs.Hoboken, NJ: Wiley.

Herbert, T. (2005). The invisible tenant: Living infederally assisted housing after prison. New York:Family Justice.

Holzer, H. J. (2007). Collateral costs: The effects ofincarceration on the employment and earnings ofyoung workers. IZA Discussion Paper No. 3118,Institute for the Study of Labor, Bonn, Germany.

Honig, M., & Filer, R. K. (1993). Causes of intercityvariation in homelessness. American EconomicReview, 83, 248–255.

Kushel, M. B., Gupta, R., Gee, L., & Haas, J. S.(2005). Housing instability and food insecurityas barriers to health care among low-incomeAmericans. Journal of General Internal Medicine,21, 71–77. doi:10.1111/j.1525-1497.2005.00278.x

La Vigne, N. G., & Parthasarathy, B. (2005). Pris-oner reentry and residential mobility. ReturningHome Illinois Policy Brief, The Urban Institute,Washington, DC. Retrieved from http://www.urban.org/UploadedPDF/311213_residential_mobility.pdf

Lee, B. A., Tyler, K. A., & Wright, J. D. (2010).The new homelessness revisited. Annual Reviewof Sociology, 36, 501–521. doi:10.1146/annurev-soc-070308-115940

Levenson, J. S., & Cotter, L. P. (2005). Theimpact of sex offender residence restrictions:1,000 feet from danger or one step fromabsurd? International Journal of Offender Therapyand Comparative Criminology, 49, 168–178.doi:10.1177/0306624X04271304

Lynch, J. P., & Sabol, W. J. (2004). Assessingthe effects of mass incarceration on informalsocial control in communities. Criminology &Public Policy, 3, 267–294. doi:10.1111/j.1745-9133.2004.tb00042.x

Ma, C. T., Gee, L., & Kushel, M. B. (2008).Associations between housing instability andfood insecurity with health care access in low-income children. Ambulatory Pediatrics, 8, 50–57.doi:10.1016/j.ambp.2007.08.004

Metraux, S., & Culhane, D. P. (2004). Homelessshelter use and reincarceration following prisonrelease: Assessing the risk. Criminology &Public Policy, 3, 201–222. doi:10.1111/j.1745-9133.2004.tb00031.x

Metraux, S., Roman, C. G., & Cho, R. (2008).Incarceration and homelessness. In D. Dennis, G.Locke, & J. Khadduri (Eds.), Toward understand-ing homelessness: The 2007 National Symposiumon Homelessness Research. Washington, DC: U.S.Department of Housing and Urban Development.

Nepomnyaschy, L., & Garfinkel, I. (2007). Childsupport, fatherhood, and marriage: Findings fromthe first five years of the Fragile Familiesand Child Wellbeing Study. Asian Social Workand Policy Review, 1, 1–20. doi:10.1111/j.1753-1411.2007.00002.x

Patterson, J. (2010). Incarcerating death: Mortalityin U.S. state correctional facilities, 1985–1998.Demography, 47, 587–607. doi:10.1353/dem.0.0123

Pew Center on the States. (2008). One in 100: Behindbars in America 2008. Washington, DC: Author.

Phillips, S. D., Erkanli, A., Keeler, G. P., Costello,E. J., & Angold, A. (2006). Disentangling therisks: Parent criminal justice involvement andchildren’s exposure to family risks. Criminology& Public Policy, 5, 677–702. doi:10.1111/j.1745-9133.2006.00404.x

Postmus, J. L., Severson, M., Berry, M. & Yoo,J. A. (2009). Women’s experiences of violenceand seeking help. Violence Against Women, 15,852–868. doi:10.1177/1077801209334445

Rafferty, Y., Shinn, M., & Weitzman, B. C. (2004).Academic achievement among formerly homelessadolescents and their continuously housed peers.Journal of School Psychology, 42, 179–199.doi:10.1016/j.jsp.2004.02.002

Reichman, N., Teitler, J., Garfinkel, I., & McLanahan,S. (2001). Fragile Families: Sample and design.Children and Youth Services Review, 23, 303–326.doi:10.1016/S0190-7409(01)001414

Page 17: Paternal Incarceration and the Housing Security of Urban Mothers

Paternal Incarceration and Maternal Housing 427

Reid, K. W., Vittinghoff, E., & Kushel, M. B.(2008). Association between the level of housinginstability, economic standing, and health careaccess: A meta-regression. Journal of Health Carefor the Poor and Underserved, 19, 1212–1228.doi:10.1353/hpu.0.0068

Royston, P. (2004). Multiple imputation of missingvalues. Stata Journal, 4, 227–241.

Rumbaut, R. G., Gonzales, R. G., Komaie, G.,Morgan, C. V., & Tafoya-Estrada, R. (2006).Immigration and incarceration: Patterns andpredictors of imprisonment among first- andsecond-generation young adults. In M. RamiroJr. & V. Abel Jr. (Eds.), Immigration and crime:Race, ethnicity, and violence (pp. 64–89). NewYork: New York University Press.

Sampson, R. J. (2011). The incarceration ledger:Toward a new era in assessing societal con-sequences. Criminology & Public Policy, 10,819–828. doi:10.1111/j.1745-9133.2011.00756.x

Sampson, R. J., & Laub, J. H. (1990). Crime anddeviance over the life course: The salience of adultsocial bonds. American Sociological Review, 55,609–627.

Sampson, R. J., & Laub, J. H. (1993). Crime in themaking: Pathways and turning points through life.Cambridge, MA: Harvard University Press.

Schwartz-Soicher, O., Geller, A., & Garfinkel, I.(2011). The effects of paternal incarceration onmaterial hardship. Social Service Review, 85,447–473. doi:10.1086/661925

Sugie, N. (2012). Punishment and welfare: Pater-nal incarceration and families’ receipt of pub-lic assistance. Social Forces, 90, 1403–1427.doi:10.1093/sf/sos055

Swisher, R. R., & Waller, M. R. (2008). Confiningfatherhood: Incarceration and paternal involve-ment among nonresident White, African Ameri-can, and Latino fathers. Journal of Family Issues,29, 1067–1088. doi:10.1177/0192513X08316273

Tach, L., Mincy, R. B., & Edin, K. (2010).Parenting as a “package deal”: Relationships,fertility, and nonresident father involvementamong unmarried parents. Demography, 47,181–204. doi:10.1353/dem.0.0096

Travis, J., McBride, E. C., & Solomon, A. L.(2005). Families left behind: The hidden costsof incarceration and reentry. Washington, DC:The Urban Institute.

Travis, J., Solomon, A. L., & Waul, M. (2001). Fromprison to home: The dimensions and consequences

of prisoner reentry. Washington, DC: The UrbanInstitute.

U.S. Department of Housing and Urban Development.(2003). Affordable housing needs: A report toCongress on the significant need for housing.Washington, DC: Office of Policy Developmentand Research.

Van Buuren, S., Boshuizen, H. C., & Knook, D. L.(1999). Multiple imputation of missing blood pres-sure covariates in survival analysis. Statistics inMedicine, 18, 681–694. doi:10.1002/(SICI)1097-0258(19990330)18:63.0.CO;2-R

Vanyukov, M. M., Neale, M. C., Moss, H. B.,& Tarter, R. E. (1996). Mating assortmentand the liability to substance abuse. Drug andAlcohol Dependence, 42, 1–10. doi:10.1016/0376-8716(96)01255-0

Venkatesh, S. A. (2002). The Robert Taylor HomesRelocation Study. New York: Columbia UniversityCenter for Urban Research and Policy.

Weschler, D. (2008). Wechsler Adult IntelligenceScale, Fourth Edition (WAIS–IV). San Antonio,TX: Harcourt Assessment.

Western, B. (2006). Punishment and inequality inAmerica. New York: Russell Sage Foundation.

Western, B., & Wildeman, C. (2009). The Blackfamily and mass incarceration. Annals of theAmerican Academy of Political and SocialScience, 621, 221–242. doi:10.1177/0002716208324850

Wildeman, C. (2013). Parental incarceration, childhomelessness, and the invisible consequencesof mass imprisonment. Annals of the AmericanAcademy of Political and Social Science, 651,74–96. doi:10.1177/0002716213502921

Wildeman, C., Schnittker, J., & Turney, K. (2012).Despair by association? The mental health of moth-ers with children by recently incarcerated fathers.American Sociological Review, 77, 216–243.doi:10.1177/0003122411436234

Zandbergen, P. A., & Hart, T. C. (2006). Reducinghousing options for convicted sex offenders:Investigating the impact of residency restrictionlaws using GIS. Justice Research and Policy, 8(2),1–24. doi:10.3818/JRP.8.2.2006.1

Zgoba, K. M., Levenson, J., & McKee, T.(2009). Examining the impact of sex offenderresidence restrictions on housing availability.Criminal Justice Policy Review, 20, 91–110.doi:10.1177/0887403408322119