Children and Youth Services Review 36 (2014) 124–132

Contents lists available at ScienceDirect

Children and Youth Services Review

j ourna l homepage: www.e lsev ie r .com/ locate /ch i ldyouth

Foster parenting together: Foster parent couples

John G. Orme ⁎, Terri Combs-OrmeCollege of Social Work, University of Tennessee, United States

⁎ Corresponding author at: College of Social Work, UnHall, Knoxville, TN 37996, United States.

E-mail address: jorme@utk.edu (J.G. Orme).

0190-7409/$ – see front matter © 2013 Elsevier Ltd. All rihttp://dx.doi.org/10.1016/j.childyouth.2013.11.017

a b s t r a c t

a r t i c l e i n f o

Article history:Received 20 July 2013Received in revised form 18 November 2013Accepted 19 November 2013Available online 28 November 2013

Keywords:Foster familiesFoster parentsFoster fathersCouplesFoster parent applicants

Almost two-thirds of children in non-relative foster care live in homes headed bymarried couples. However, theliterature includes little about either the role of foster fathers or the complexity and importance of the maritalcontext for fostering. This study examined family functioning, overt interparental hostility, depression, andparental acceptance in a sample of 111 heterosexual foster parent couples and the relationship between familycontext and potential to foster parent successfully. Latent class analysis revealed three distinct types of fosterparent couples, “Good Context” (33%), “Typical” (56%), and “Bad Context-Discordant” (11%) couples. “BadContext-Discordant” couples are notable for having much lower scores on all context measures, for differencesbetween mothers and fathers, and for the fact that most fathers had depression scores in the clinical range.Finally, results indicated that better family context was related to greater potential to foster successfully andthat on three of five dimensions measured mothers exhibited greater potential.

© 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Nearly half (47%) of the 394,833 children in foster care in the US atthe end of the 2011 federal fiscal year lived in non-relative fosterhomes. Of these 184,532 children, almost two-thirds (63%) lived infoster homes headed by married couples (National Data Archive onChild Abuse and Neglect, 2012). Yet little is known about marriedcouples who serve as foster parents in terms of marital and family func-tioning, parenting, and other contextual variables. We begin by brieflyreviewing the literature concerning foster fathers and foster parentcouples. This is followed by a discussion of the importance of both thequality of parenting and the role of the marital relationship itself inparenting. Finally, we discuss parental mental health as an importantaspect of the foster family context.

1.1. Foster parents as couples

The foster parenting literature includes little about either the role offoster fathers or the complexity and importance of the marital contextfor fostering. Mothers frequently provide the only data about fosterfamilies (Inch, 1999), in part because they reportedly handle most ofthe interaction with foster care agencies (Eastman, 1982; Wilson,Fyson, & Newstone, 2007) and are believed to be the primary caregivers(Gillian, 2000). Indeed, foster fathers sometimes have been regardedmerely as sources of support for foster mothers (Dando &Minty, 1987).

iversity of Tennessee, Henson

ghts reserved.

Available research about foster fathers is mainly based on qualitativestudies of small samples. This small body of knowledge, however, dem-onstrates that foster fathers may be important sources of care for fosterchildren. Foster fathers report playing active roles in parenting, servingas rolemodels for fathering andhaving positive influences on foster chil-dren (Denuwelaere & Bracke, 2007; Riggs, Delafabbro, & Augostinos,2010;Wilson et al., 2007). Foster fathers recognize that children's expe-riences with their foster fathers are likely to be significant and complex,because somany foster children have not experienced positive relation-ships with their birth fathers (Fernandez, 2007).

The marital relationship itself is also an important resource for chil-dren in foster couple homes. Lamb (2012) reviewed and summarizedfour decades of research on determinants of child adjustment andreached the conclusion that although family structure (in terms ofwhether parents are married or not) is not important, quality of parent-ing, quality of relationships between parents, and economic circum-stances are key determinants of successful child adjustment. Bothmarital harmony and conflict are known to have independent signifi-cant influences on children's adjustment (Lamb), possibly by influenc-ing children's emotional security (Buehler, Benson, & Gerard, 2006).Given histories of maltreatment, poor parenting, and difficult familycontexts, it is logical that foster children's needs exceed those of otherchildren in these areas.

1.2. Quality of parenting

Quality of parenting is inadequately studied in the foster care litera-ture, despite its clear importance to child outcomes. This may be in partbecause, as Berrick and Skivenes (2012) discuss, so much of good fosterparenting goes beyond “basic” good parenting to deal with foster chil-dren's special needs, backgrounds of maltreatment, and traumatization

125J.G. Orme, T. Combs-Orme / Children and Youth Services Review 36 (2014) 124–132

due to removal from their homes and placements with strangers, in thecontext of the foster care system.

Good parenting has been defined andmeasured in toomanyways torecount in this manuscript. We have previously defined context-basedparenting as “behaviors that, given a child's age, developmental needs,and special circumstances, are optimal for promoting the child's healthygrowth and development” (Combs-Orme, Wilson, Cain, Page, & Kirby,2003, p. 440). This definition seems particularly apt for foster parenting,as foster children's needs are likely to be many and unique. Certainly,marriage and family relationships and functioning are part of that con-text as well.

Parenting includes parental attitudes, knowledge, and behavior.Attitudes about parenting can be thought of as preconceptions of desir-able parenting behavior or generalized models of how parents shouldbehave while fulfilling childrearing duties. The question of how parentsare differentiated by their attitudes about parenting practices has been asubject of study throughout most of the 20th and 21st centuries,primarily using descriptive studies.

The unique nature of foster parenting requires dealing with childrenwho have serious behavioral and emotional issues, complicated familyrelationships, and other problems. Frequently, foster parents do noteven have complete knowledge of their foster children's previous livesand may need to adjust their parenting as they go to deal with childrenwho have been scarred by sexual and physical abuse, deprivation, andchronic insecurity.

The influence of attitudes on behavior is not consistently clear.Parents' attitudes toward corporal punishment, for example, are notperfect predictors of use of corporal punishment andmay bemoderatedby other variables (e.g., Crouch & Behl, 2001). However, parenting atti-tudes are generally believed to make important contributions to actualbehavior, even if this relationship is not absolute (Holden&Buck, 2002).

Limited research has examined parenting attitudes specific to fosterparenting. Harden, Meisch, Vick, and Pandohie-Johnson (2008) devel-oped a measure of parenting attitudes central to foster parenting,emphasizing attachment/commitment to the child as most important.Others included attitudes toward: child's relationship with biologicalfamily; previous experiences with caregiving and child emotional andbehavioral problems; role and relationship with the foster agency; andchild-centered motivation to foster. Generally low reliability of thesub-scales (only one demonstrated internal reliability greater than.70) may limit the usefulness of the measure, but their findings dosuggest that parents who have positive attitudes toward the fostercare experience also report more appropriate parental attitudes overall.

1.3. Quality of marital relationship

The quality of the marital relationship for married parents and thequality of parenting are consistently correlated in the parenting re-search (Grych, 2002), and it is easy to presume that these two attributesalways go together. However, the variety of global measures employedin this researchmakes it difficult to be more precise about how they in-teract to influence child outcomes. Belsky and Fearon (2004) assertedthat “…there are different types of families in which marital quality,parenting, and child development go together in rather differentways” (p. 503). Belsky and Fearon examined this issue with 829 two-parent intact families. They found that 15% of the married families inthis nationally representative sample demonstrated high quality inboth marriage and parenting; 43% showed moderate quality of both;and 16% showed poor quality on both. The rest (26%) exhibited amixed pattern.

Not surprisingly, Belsky and Fearon (2004) found that in first grade,children who lived in families with both good parenting and strongmarital relationships had the best outcomes, and children whose fami-lies lacked both types of resources showed the poorest outcomes. Thetwo groups also differed on a number of important antecedent factors,including income, maternal age and education. Consideration of these

pre-existing differences attenuated some, but not all, of the significantrelations.

Findings for the mixed groups revealed that outcomes for childrenwho grew up with only one set of resources (high-quality parentingor high-quality marital relationships) were not as good as childrenwho had both and were better than those of children who had neither.When only one advantage was present, high-quality parenting was themost advantageous. It is reasonable to assume that similar relationshipswould exist for foster children and parents, albeit probably in morecomplicated ways due to legal complexities of the foster care systemand children's histories. Moreover, parenting standards are higher forfoster parents, because the state has assumed legal authority over chil-dren and by doing so has assumed the obligation of providing the bestpossible environments for those children (Shlonsky & Berrick, 2001).

1.4. Foster parent mental health

Foster care agencies seek resources for children with many and var-ied problems. Foster children thus require a high-quality home environ-ment as all children do, but foster parents oftenmust demonstrate extraskills and knowledge to nurture them, given their unique histories ofmaltreatment and other disadvantages (Shlonsky & Berrick, 2001).

Foster care agencies' training materials recognize the impor-tance of foster parents' having good mental and emotional health(e.g., Dickerson, Allen, & Pollack, 2011), and parenting research isclear that mental illness and symptoms are detrimental to parentingand child outcomes, but few studies have focused on foster parents'mental health issues (for exceptions see Burgess & Borowsky, 2010;Cole & Eamon, 2007; Mennen & Trickett, 2011; Orme et al., 2004).Foster care workers are alert for signs of mental illness that mightimpair foster parenting, but research has not provided guidance onthe prevalence of mental illness in foster parents nor on its effects.

Extensive research demonstrates thatmaternal depression is damag-ing to infants and young children; depression in particular impairs par-ents' emotional availability and sensitivity to a child's needs (England& Sim, 2009). A recentmeta-analysis of 193 studies (not including fosterchildren) reported that the offspring of depressed mothers are at highrisk of depression and other psychiatric disorders (Goodman et al.,2011).

Although paternal depression has been much less studied, researchalso shows that paternal psychiatric disorders may be as damaging forchildren as maternal disorders (Ramchandani & Psychogiou, 2009).Depression has been the most often studied; the children and adultoff-spring of depressed fathers are at higher risk for behavioral andemotional problems.

Parental depression and possibly other disorders affect children'soutcomes through several avenues, including effects on parent–child in-teractions, the marital relationship, and even the financial resourcesavailable to the family and child. Parents with psychiatric disordersprovide less competent physical care and nurturing for their childrenand spend less time with their children (Ramchandani & Psychogiou,2009). Of course it is impossible to know how genetics and epigenetics(the influence of the environment on gene expression) influence theserisks, just as it is impossible to calculate the effects of foster children'sexperiences of previous poor parenting and maltreatment, but thesefindings do suggest that depression influences parenting behavior inways that are damaging to children.

1.5. Research questions

This study seeks to understand important aspects of the family con-text among married foster parent couples, examining both individual-level and dyadic data with both foster mothers and fathers. First, weexamine family context in a sample of foster parent couples. Second,we determinewhether this overall sample contains discrete types (sub-populations) of foster parent couples with different dyadic patterns of

126 J.G. Orme, T. Combs-Orme / Children and Youth Services Review 36 (2014) 124–132

family context. Third, we describe these family context patterns andcompare them to patterns found in the overall sample. Finally, weexamine whether these discrete dyadic patterns are meaningful in thesense that they are differentially associated with potential to providequality foster care.

2. Methods

2.1. Design and sample

This studywas approved by theUniversity of Tennessee InstitutionalReview Board, and signed informed consent forms were obtained fromall participants.

This study was part of a larger project designed to develop and testfoster parent assessment instruments. The first phase of the study wasconducted with foster mothers only. Using a cross-sectional design,we recruited a convenience sample of 304 foster mothers from a list ofstate and local foster parent associations obtained from the NationalFoster Parent Association (NFPA) website and sent recruitment infor-mation packets to those associations by mail and email. Recruitmentpackets included information about the study and tools for distributinginformation to foster parents, cover letters describing the study,endorsement letters from the president of the NFPA, flyers about thestudy for reproducing and distributing, and information about thestudy for placement in association newsletters or on websites. Weasked these associations to distribute flyers to members and to placeannouncements in newsletters and on their state association websitesand message boards.

All advertisements included an email address and a toll-free projecttelephone number, and we asked foster mothers to use one of thesemeans to contact us to participate. Thosewilling to participate providedtheir names, mailing addresses and, if available, email addresses. Addi-tional information about the study was sent upon request. Fostermothers who chose to participate received $50 gift certificates uponcompletion of the study measures.

Initially, we mailed recruitment packets only to state foster parentassociations. About three months into the project, we resent recruit-ment materials to state foster parent associations in states from whichwe had not received any completed questionnaires. Also at this time,we sent recruitment materials to local foster parent associations fromthose states.

Initially we did not have funding to study foster fathers, but wehoped to obtain such funding, so we asked participating marriedmothers to recruit their husbands for a future study of foster fathers.Funding to study foster fatherswas obtained, and data collection for fos-ter fathers started approximately one year after data collection startedfor mothers. Of the 238 participating mothers who were married, 148reported that their husbands would be interested in participating, and111 completed the measures. The focus of this study is on these 111couples.

Participating mothers and fathers completed the measures de-scribed below and provided other information as part of the broaderstudy (Orme, Cox, Rhodes, Coakley, Cuddeback, & Buehler, 2006;Orme, Cuddeback, Buehler, Cox, & Le Prohn, 2006). More specifically,we mailed the questionnaires and accompanying instructions to eachparticipant along with a consent form and a pre-paid self-addressedenvelope. We asked participants to sign and return informed consentforms with their completed questionnaires, and we included an extraconsent form that participants were asked to retain for their records.Also, we provided an email address and a toll-free project number onthe consent forms, and we asked participants to call the toll-freenumber or email if they had any questions about the study. In addition,we asked mothers and fathers to complete the questionnaires indepen-dently, although we have no way of knowing if they followed theseinstructions. Upon receipt of the completed questionnaire we mailedeach participant was a $50 Wal-Mart gift certificate.

2.2. Family context measures

It would not be possible, practical, or parsimonious to include all ofthe marital/family functioning, psychological, and parenting measureswe collected in one latent class model. Therefore we selected a set ofmeasures that have excellent reliability and are particularly pertinentfor the research questions. There are two types of family context mea-sures. In thefirst type (twomeasures) both fathers andmothers providetheir independent perceptions of family functioning and inter-parentalhostility. In the second set of family context measures, mothers andfathers provide self-reports of depression and parental acceptanceattitudes.

2.2.1. Family functioning style scale (FFSS)The FFSS is a 26-item scale designed to measure perceptions of

family strengths in terms of interactional patterns, family values, copingstrategies, family commitment, and resource mobilization (Deal,Trivette, &Dunst, 1988). These characteristics are particularly importantfor the current study because they provide a picture of eachrespondent's assessment of the functioning and interaction of the entirefamily, including the parents and other children in the home.

Each item of the FFSS is rated on a 5-point scale: (1) not at all like myfamily, (2) a little like my family, (3) sometimes like my family, (4) usuallylike my family, and (5) almost always like my family. The total scale scoreis the sum of the 26 item scores (potential range from 26 through 130).Higher scores indicate better family functioning.

2.2.2. Overt interparental hostility (OIH)The OIH is a 6-item scale designed to measure each respondent's

assessment of howoften overt verbal and physical expressions of hostil-ity occur between parents (Buehler et al., 1998). Each item is rated on a4-point scale: (1) never, (2) once in a while, (3) fairly often, and (4) veryoften. The total scale score is the sum of the six item scores (potentialrange from 6 through 24). Higher scores indicate greater interparentalhostility. We selected this instrument in part because of the generalliterature demonstrating the negative impact of interparental hostilityon child outcomes (Buehler et al., 2006) and also because such hostilitymight be particularly problematic for foster children.

2.2.3. Center for Epidemiological Studies—Depressed Mood (CES-D)The CES-D is a widely-used 20-item measure of depressive symp-

toms (Radloff, 1977; Radloff & Locke, 2008). We selected this measurebecause research has shown depression in parents to be related topoor parenting and poor child outcomes (Committee on Depression,Parenting Practices, and the Healthy Development of Children, 2009).Respondents rate the frequency of occurrence of each symptom inthe past week on a 4-point scale: (0) rarely or none of the time (lessthan 1 day), (1) some or little of the time (1–2 days), (2) occasionally ora moderate amount of time (3–4 days), and (3) most or all of the time(5–7 days) (3). Items 4, 8, 12, and 16 are reverse-scored before comput-ing the scale score. After reverse-scoring these items, all of the itemscores are summed to create the total scale score (potential rangefrom 0 through 60). Higher scores indicate greater severity of depres-sive symptoms, and a score of 16 or more typically is employed as acut-off that indicates clinical depression.

2.2.4. Parental acceptance scale (PAS)The PAS is a 10-item subscale constructed by Ruiz, Roosa, and

Gonzalez (2002) based on the Children's Report of Parental BehaviorInventory (CRPBI; Schaefer, 1965). It is designed to measure broadlyparents' behaviors and feelings indicating acceptance, warmth, andsupport for a child, and the construct is frequently labeled using theseterms. For example Barber, Stolz, Olsen, and Maughan (2005) use theterm parental support. Higher scores indicate greater parental accep-tance, warmth, or support.

Table 1Foster parents' demographic characteristics.

Characteristic Mothers Fathers

RaceEuropean-American 93.7 91.9African-American 4.5 5.4Hispanic .9 0American Indian 0 1.8

Highest degreebHS 1.8 5.4HS/GED 24.3 39.6College, no degree 33.3 22.5Two-year degree 14.4 9.0Bachelor's degree 17.1 12.6Advanced degree 9.0 10.8

Employment statusFull time 27.0 77.5Part-time 18.9 1.8Unemployed, looking for work .9 1.8Homemaker, not employed outside home 43.2 2.7Disabled or retired, not employed outside home 2.7 11.7Other 7.2 4.5

Yearly family incomeb10,000 .9 .910,000–19,999 3.7 2.720,000–29,999 7.3 7.230,000–39,999 21.1 18.940,000–49,999 9.2 12.650,000–59,999 15.6 18.960,000–69,999 11.0 9.070,000–79,999 12.8 11.780,000–89,999 6.4 5.490,000–99,999 4.6 2.7≥100,000 7.3 9.9

Note. Income data were missing for twomothers. Race percentages do not add up to 100%because parents were asked to select all that applied.

127J.G. Orme, T. Combs-Orme / Children and Youth Services Review 36 (2014) 124–132

Each item of the PAS (As a parent, I'm a person who…believes inshowingmy love for this child, gives this child a lot of care and attention,is easy to talk to, etc.) is rated on a 3-point scale: (1) not like me,(2) somewhat like me, and (3) a lot like me. The total scale score is themean of the 10 item scores (potential range from 1 through 3).

2.2.5. Casey Foster Applicant Inventory (CFAI)The Casey Foster Applicant Inventory (CFAI) is a standardized self-

report measure designed to assess the potential to foster parentsuccessfully (Cherry & Orme, 2011; Delgado & Pinto, 2011; Orme,Cuddeback, Buehler, Cox, & Le Prohn, 2007). Each item is rated using a4-point response format: strongly disagree (1), disagree (2), agree (3),and strongly agree (4). The CFAI was developed using a rigorous processthat involved input from foster parents, foster care workers, and formerfoster children, aswell as standards developed byprofessional organiza-tions and foster care research.

The CFAI has six subscales, eachwith a potential range of values from0 through 100. Higher subscale scores indicate greater potential to:promote foster child development (FCD: 60 items); foster challengingchildren (CC: 13 items); manage challenging relationships with fostercare workers and agencies (WAC: 9 items); co-parent foster children(CP: 10 items, applicable only in two-parent foster families); integratefoster children into a foster family with birth or adopted children (IFC:6 items, applicable only to foster families with birth or adoptedchildren); and provide care to children of a relative. [The Kinship Caresubscale was excluded from this study because it was only applicableto a small subsample of mothers (n = 28) and fathers (n = 30).]

2.3. Latent class analysis

Weused latent class analysis (LCA) to: (1) explorewhether there arediscrete types of couples in terms of reported family context; (2) deter-mine the size of the subgroups; and (3) assign couples to subgroups.

LCA is a statistical method used to identify discrete subtypes (sub-populations) of similar cases, in this case couples, within an overallpopulation (Collins & Lanza, 2010; Muthén, 2008). These subtypes arereferred to as “latent classes,” and the classes form the categories of adiscrete latent variable. The variable is latent in that it is not directlyobservable, but instead is inferred from a set of observed variables(indicators) (e.g., family context measures as completed by mothersand fathers).

We had no a priori hypotheses about the number of latent classes(i.e., number of different types or subpopulations of couples), so weestimated and compared models with one through three classes. Thisis analogous to determining the number of factors in an exploratoryfactor analysis and, as in exploratory factor analysis, ultimately it isimportant to validate the empirically derived variables. We usedMplus 7 (Muthén & Muthén, 1998–2012) (syntax available uponrequest) and the robust maximum likelihood ratio estimator (MLR)for these analyses, treating the variables as continuous.We transformedraw scale scores into standard scores and used these in the analyses tofacilitate interpretation of results. We fixed covariances among thelatent class indicators to zero (i.e., “local independence” assumed),and allowed means and variances of the indicators to vary acrossclasses.

There is general agreement that theory, interpretability, parsimony,and class size should drive selection of the number of latent classes. Inaddition, there are several statistical indices that can be used in modelselection. However, there is no definitive statistical test of the “true”number of classes so we considered three of the most promising statis-tical criteria used to select the most plausible number of classes(Henson, Reise, & Kim, 2007; Nylund, Asparoutiov, & Muthén, 2007):(1) the bootstrap likelihood ratio test (BLRT; McLachlan & Peel, 2000;Nylund et al., 2007); (2) the Bayesian information criterion (BIC:Schwarz, 1978); and (3) the sample-size adjusted Bayesian informationcriterion (SSABIC; Sclove, 1987). In addition, after identifying the most

plausible model we considered the degree of error associated with theassignment of couples to classes as indicated by entropy, an aggregatemeasure of classification uncertainty and, more specifically and impor-tantly, the mean probability of class membership for each class.

Models with the lowest BIC or SSABIC, and the model at which theBICs in nested models begin to level off, suggest the best model fit.The BLRT provides a p-value that can be used to determine statisticallysignificant improvement in fit when an additional class is added; a non-significant p-value indicates that the more parsimonious model withone fewer class is preferred. Finally, entropy values (Ramaswany,DeSarbo, & Robinson, 1993) and mean probabilities of class member-ship near 1 indicate a high degree of certainty in classifying cases toclasses. Entropy should not in itself be used to determine the optimalnumber of classes, but it provides an important summary of the extentto which a model generates classification errors.

3. Results

3.1. Sample

3.1.1. Non-participantsWe used participation as the dependent variable in a binary logistic

regression to examine whether participating married couples weredifferent in terms of mothers' reported age, family income, race(European-American), or education. Results indicated no statisticallysignificant differences. We conducted a similar analysis using mothers'scores from the FFSS, OIH, CES-D, and PAS and found no statisticallysignificant differences. (Results available upon request.)

3.1.2. Demographic characteristics of participating couplesThe 111 participating couples included in this study lived in 29

different states. Over 90% of mothers and fathers were European-American, and had a high school education or some college (Table 1).

128 J.G. Orme, T. Combs-Orme / Children and Youth Services Review 36 (2014) 124–132

The vast majority of fathers were employed full-time (77.5%), as wereabout one-fourth of mothers. Themedian yearly family income approx-imated $50,000. Mothers reported four children in their homes(M = 3.75, SD = 2.00), two of whom were birth or adopted children(M = 1.77, SD = 1.59), and one who was younger than six (M =1.18, SD = 1.25). (We did not collect this information from fathers.)Mothers were approximately 42 years (M = 42.28, SD = 9.51), andfathers 46 years old (M = 45.79, SD = 9.92). The median number ofyears fostered was three for mothers and five for fathers and the inter-quartile range (IQR) was two to 10 years for both. The median numberof children fostered was 12 for mothers (IQR = 4.0–30.0) and 14 forfathers (IQR = 6.0–32.5). (This discrepancy was likely due to the factthat data collection for foster fathers started approximately one yearafter data collection started for mothers.)

3.2. Reliability of family context scores and differences between mothersand fathers

Internal consistency reliability was .80 or greater, except formothers' marital hostility (α = .66) and mothers' parental acceptance(α = .68; Table 2).

Mothers' and fathers' scores were positively correlated (Table 2).Self-perceptions (depression and parental acceptance) were not ashighly correlated as appraisals of dyadic variables (family functioningand marital hostility).

Paired-sample t-tests (Table 2) indicated that mothers in the overallsample reported higher mean family functioning and parental accep-tance scores than fathers, although these differences were small asquantified by the g-index, a bias-corrected standardizedmeandifferencefor dependent groups (Borenstein, Hedges, Higgins, & Rothstein, 2009,pp. 28–30). In addition, t-tests for dependent group variances (Glass &Stanley, 1970) indicated greater variability among fathers in family func-tioning (t(109) = 3.27, p = .001), marital hostility (t(109) = 2.84,p = .005), and parental acceptance (t(109) = 6.97, p b .001), butthere was no statistically significant difference between mothers andfathers in variability of the severity of depressive symptoms (t(108) =.88, p = .381).

3.3. Reliability of CFAI scores and differences between mothers and fathers

Internal consistency reliabilitywas .70 or greater except for potentialto integrate foster children (α = .68) scores reported by fathers(Table 2).

Mothers and fathers' CFAI-A scores were positively correlated(Table 2). However, paired-sample t-tests indicated that mothers hadhigher mean potential to foster overall, to co-parent, and to integrate

Table 2Descriptive statistics and comparisons between mothers and fathers for family context and Ca

Scale Mothers Fathers

M SD α M SD α

Family contextFFSS 109.89 11.01 .92 105.62 14.55 .95OIH 7.58 1.50 .66 7.77 1.91 .77PAS 2.84 .19 .68 2.71 .35 .89CES-D 4.99 5.03 .80 5.05 5.45 .83

Casey Foster Applicant InventoryFCD 80.99 8.79 .95 75.52 8.96 .95CC 61.10 12.19 .83 59.68 9.78 .78WAC 58.13 15.10 .78 57.09 11.69 .72CP 84.32 12.85 .89 79.40 13.25 .88IFC 69.63 12.70 .70 67.06 10.50 .51

Note. For the CES-D n = 110, and for the IFC n = 98 (only applicable to those with children in(parental acceptance scale), CES-D (Center for Epidemiological Studies—Depressed Mood),(WAC), co-parenting (CP), and integrating foster children (IFC).

foster children into the home, but these differences were in the small-to-moderate range as indicated by the g-index (Table 2). In addition,there was greater variability among mothers in potential to fosterchallenging children (t(109) = 2.60, p = .011), meet worker andagency challenges (t(109) = 2.94, p = .004), and integrate fosterchildren into the home (t(96) = 2.13, p = .035), but there was nostatistically significant difference in variability in overall potential(t(109) = − .21, p = .832) or potential to co-parent (t(109) = − .36,p = .723).

3.4. Latent class analysis results

We selected the three-classmodel for several reasons. First, the BLRTsuggested that it fit better than the two-class model, the three classmodel fit better than the two-class model, and the four-class modelwould not converge. Second, the BIC decreased for the first threemodels, so theBLRT andBIC suggest that the three-classmodel providedthe best balance between model fit and parsimony. Fourth, there wasminimal error associated with assignment of couples to the three latentclasses, suggesting that the model adequately defines the data; meanprobability of class membership was .97 for two of the classes and .98for the remaining class, and entropy was .91. Fourth, class sizes wereadequate in the three-class model. Fifth, multiple sets of starting valuesclearly indicated a single maximum likelihood solution for the three-class model, suggesting that it was identified. Finally, characteristics ofthe three classes were interpretable and substantively meaningful(Table 3).

One class of couples accounted for 56% of the sample;wewill refer tothese as Typical couples for reasons described below. The second classaccounted for 33% of the sample; we will refer to these as Good Contextcouples. The third class accounted for 11% of the sample;wewill refer tothese as Bad Context-Discordant couples. It is important to note thatthese labels refer to the relative standing of these three types of couples,not to an absolute standard such as a clinical cut-off. As shown below,however, these classes are related to potential to foster in ways thatprovide support for these interpretations.

Fig. 1 shows the profile plot of standard scores for the three differenttypes of couples. Typical couples are characterized by scale scores closeto the mean; average absolute deviation from the mean equals .16standard deviations. Also, this class is characterized by small differencesbetween mothers and fathers (.09 to .23 SD).

Good Context couples are characterized by scores that average.50 standard deviations from the mean in the favorable direction(i.e., higher family functioning and parental acceptance and lowermar-ital hostility and depression scores). However, Good Context coupleshave marital hostility scores close to the mean, similar to Typical

sey Foster Applicant subscales (N = 111).

Mothers vs. fathers

t p g-index r 95% CI p

−3.24 .002 −.24 .44 .27, .58 b .0011.12 .264 .08 .48 .32, .61 b .001

−4.22 b .001 −.31 .33 .15, .48 b .001.10 .920 .01 .19 .00, .37 .044

−5.80 b .001 −.61 .37 .20, .52 b .001−1.29 .199 −.13 .46 .29, .59 b .001−0.73 .468 −.08 .39 .22, .54 b .001−3.75 b .001 −.37 .44 .27, .58 b .001−2.12 .037 −.18 .48 .31, .62 b .001

the home). FFSS (family functioning style scale), OIH (overt interparental hostility), PASfoster child development (FCD), challenging children (CC), worker-agency challenges

Table 3Latent class analysis models.

Number oflatent classes

Number ofparametersestimated

BIC SSABIC BLRT BLRTp-value

Entropy

1 16 2586.09 2535.53 N/A N/A N/A2 33 2419.68 2315.40 .893 50 2380.11 2222.10 119.64 b .001 .91

Note. BIC = Bayesian information criterion; SSABIC = sample-size adjusted Bayesianinformation criterion; BLRT = bootstrap likelihood ratio test.

129J.G. Orme, T. Combs-Orme / Children and Youth Services Review 36 (2014) 124–132

couples, and so marital hostility does not differentiate these two typesof couples. Overall, though, on average there is a .59 standard deviationdifference between these couples and Typical couples in the direction ofmore favorable scores, and this mean difference is larger (.77) if maritalhostility is excluded. Finally, as with Typical couples, this class is charac-terized by small differences between mothers and fathers (.00 to .16SD).

Bad Context-Discordant couples are characterized by scores that av-erage 1.14 standard deviations from themean in the unfavorable direc-tion. Differences between this class and the other two classes are alsolarge; on average there is a 1.63 standard deviation difference in com-parison toGood Context couples and a 1.07 standard deviation differencein comparison to Typical couples. Finally, BadContext-Discordant couplesare characterized by differences between mothers and fathers in self-perceptions of the severity of depressive symptoms and parentalacceptance and, to a lesser degree, their separate appraisals of familyfunctioning and marital hostility. That is, mothers reported higherparental acceptance (.72 SD difference) and family functioning (.45 SDdifference) scores than fathers; these results are consistent with thetotal sample, but larger. Mothers also reported much lower depressionscores (1.09 SD difference) and somewhat lowermarital hostility scores(.35 SD difference) than fathers; these results are not consistent withthe total sample in which these differences were .08 and .01 standarddeviations, respectively.

Very fewmothers (3.6%) or fathers (7.3%) were in the clinical rangeon the CES-D (i.e., ≥16), the only latent class indicator with a clinicalcutting score. However, of the eight fathers in the clinical range, allbut twowere in the Bad Context-Discordant group (50% of the 12 fathersin this group) and the remaining twowere in the Typical group. The fourmothers in the clinical range were all in the Typical group.

3.5. Relationship between potential to foster, family context, and gender

We used mixed-model ANOVA to examine differences in potentialto foster among latent classes (between-subjects) and betweenmothers and fathers (within-subjects) (Kenny, Kashy, & Cook, 2006,

FFSS-FF FFSS-FM ACPT-FF ACPT-FM OIH-FF OIH-FM CESD-FF CESD-FM

Bad (11%) -1.32 -0.88 -1.34 -0.62 1.53 1.18 1.65 0.56

Typical (56%) -0.03 -0.22 -0.11 -0.21 -0.25 -0.16 0.05 0.27

Good (33%) 0.52 0.67 0.65 0.57 -0.12 -0.14 -0.65 -0.65

-1.80

-1.40

-1.00

-0.60

-0.20

0.20

0.60

1.00

1.40

1.80

Mea

n Sc

ale

Scor

es (

z-sc

ores

)

Fig. 1. Latent class profile plot.

pp. 67–70). Given the disparity in sample sizes among groups weexamined Box's test of the null hypothesis that the observed covariancematrices of the dependent variables are equal across groups. We couldnot reject this null hypothesis in any case.

The significant main effect for family context indicates that thebetter the family context, the greater the potential to foster in all areasexcept potential to meet worker-agency challenges (Tables 4 and 5).More specifically, in each of these four areas, planned contrasts(Helmert) indicated that mean potential to foster was lower for thosein the Bad-Discordant group compared to the Typical and Good groups,and higher in the Good group compared to the Typical group, exceptfor potential to foster challenging children (results available uponrequest). The largest effect of context by far was on potential to co-parent (partial η2 = .22).

Gender also had amain effect. Mothers had greater overall potentialto foster, to co-parent, and to integrate foster children into the home.The largest effect of gender was on overall potential to foster (partialη2 = .22), but the effect on potential to co-parent was similar (partialη2 = .18).

Two interaction effects influence interpretation of the findings. First,mothers and fathers differed on the effect of family context on potentialto co-parent (Fig. 2). Mothers and fathers in the Typical group hadmoredesirable co-parenting scores than mothers and fathers in the Bad-Discordant group, but this difference is larger for fathers (17.67 vs.5.48). Mothers and fathers in the Good group had higher co-parentingscores than mothers and fathers in the Typical group, but the differenceis smaller for fathers (4.25 vs. 8.30).

Second, the effect of family context on potential to integrate fosterchildren into the home differed for mothers and fathers (Fig. 3).Mothers in the Bad-Discordant group had about the same potential tointegrate foster children as mothers in the Typical group, but mothersin the Good group demonstrated more potential. Conversely, fathers inthe Typical group scored higher on potential to integrate foster childrenthan fathers in the Bad-Discordant group, but they had about the samepotential as fathers in the Good group.

4. Discussion

This section highlights three important issues in the findings, butfirst we recognize the limitations of the study. The study employs aconvenience sample drawn from state and local foster parent associa-tions, so it is likely that respondents represent the most committedfoster parents, possibly those who have more leisure time to completethe instruments, or perhaps parents who were motivated by the $50incentive offered to participate. In addition, data from fathers were

Table 4Descriptive statistics for CFAI subscales by family context and gender.

Subscale Fathers Mothers

Bad Typical Good Total Bad Typical Good Total

FCD M 69.97 75.49 77.37 75.52 78.33 79.37 84.56 80.99SD 8.63 8.33 9.53 8.96 8.47 8.28 8.82 8.79n 12 62 37 111 12 62 37 111

CC M 56.98 59.42 60.98 59.68 54.06 60.01 65.22 61.10SD 8.20 8.95 11.47 9.78 15.63 11.19 11.45 12.19n 12 62 37 111 12 62 37 111

WAC M 53.70 57.11 58.16 57.09 55.86 57.24 60.36 58.13SD 8.28 10.18 14.68 11.69 15.51 14.00 16.81 15.10n 12 62 37 111 12 62 37 111

CP M 62.22 79.89 84.14 79.40 76.67 82.15 90.45 84.32SD 11.40 10.73 13.41 13.25 13.33 12.59 10.75 12.85n 12 62 37 111 12 62 37 111

IFC M 59.72 67.86 68.52 67.06 66.20 66.69 76.48 69.63SD 7.16 10.36 10.95 10.50 11.47 11.07 13.66 12.70n 12 56 30 98 12 56 30 98

Note. Foster child development (FCD), challenging children (CC), worker-agency chal-lenges (WAC), co-parenting (CP), and integrating foster children (IFC).

Table 5Mixed-model ANOVA: CFAI subscales by family context and gender.

Subscale Source df Mean square F p Partial EtaSquared

FCD Gender 1 1494.52 30.74 b .001 .22Gender × context 2 91.81 1.89 .156 .03Error 108 48.61Intercept 1 855,049.80 8511.76 b .001 .99Context 2 516.45 5.14 .007 .09Error 108 100.46

CC Gender 1 14.41 0.22 .642 .00Gender × context 2 140.23 2.11 .126 .04Error 108 66.38Intercept 1 502,857.49 2978.27 b .001 .97Context 2 584.83 3.46 .035 .06Error 108 168.84

WAC Gender 1 79.68 0.70 .406 .01Gender × context 2 29.18 0.25 .775 .00Error 108 114.45Intercept 1 463,512.61 1836.11 b .001 .94Context 2 208.36 0.83 .441 .02Error 108 252.44

CP Gender 1 2092.69 23.25 b .001 .18Gender × context 2 399.75 4.44 .014 .08Error 108 90.01Intercept 1 893,851.42 4599.29 b .001 .98Context 2 2987.22 15.37 b .001 .22Error 108 194.35

IFC Gender 1 654.88 10.29 .002 .10Gender × context 2 459.92 7.23 .001 .13Error 95 63.64Intercept 1 611,064.98 3316.19 b .001 .97Context 2 930.01 5.05 .008 .10Error 95 184.27

Note. Foster child development (FCD), challenging children (CC), worker-agencychallenges (WAC), co-parenting (CP), and integrating foster children (IFC).

Bad Typical Good

Fathers 59.72 67.86 68.52

Mothers 66.20 66.69 76.48

50556065707580859095

100

Inte

grat

ing

Fos

ter

Chi

ldre

n Sc

ores

Fig. 3. Effect of family context on potential to integrate children into the home asmoderated by gender.

130 J.G. Orme, T. Combs-Orme / Children and Youth Services Review 36 (2014) 124–132

collected a year after data frommothers, so it is possible that family cir-cumstances had changed during this interval, contributing to fathers'less optimal scores. Finally, all the instruments used are based on theparents' perceptions and self-reports and so results are limited bywhat these parents know and by what they are willing to discloseand, in particular, information given by foster parents may be affectedby their desire to give a good impression.

The relatively small sample of foster parent couples is another limi-tation. In particular, too small a sample can lead to choosing too fewclasses to adequately describe the underlying model (Dziak, Lanza, &Tan, in press). So, it may be that our classification of foster parentcouples may oversimplify the true underlying model, and this is a criti-cal question for future research in this area. Indeed Belsky and Fearon(2004), for example, used a much larger sample and identified fivedifferent types of families based onmultiplemeasures ofmarital qualityand parenting in a general population of parents.

Bad Typical Good

Fathers 62.22 79.89 84.14

Mothers 76.67 82.15 90.45

50556065707580859095

100

Co-

Par

enti

ng S

core

s

Fig. 2. Effect of family context on potential to co-parent as moderated by gender.

The major strength of the current study may lie in the examinationof married foster parents not just as individuals, but also as dyads (cou-ples). An attractive and unique aspect of co-parenting research is thatthe dyad (e.g., coparental relationship) or triad (if including the child)is the focus of study. However, this also means that the dyad is theunit of analysis and, therefore, analytical approaches traditionally usedin family research are not adequate. At the most fundamental level,members of a dyad are two individuals who are not independent.Rather, it is expected that they have something in common that willinfluence their scores. This commonality, or nonindependence, is truewhen the scores of members of a dyad score are more similar (or dis-similar) than those of people who are not members of a dyad (Kennyet al., 2006). The analytical method used must account for this lack ofdependence in order to avoid biased estimates. In the current study,Latent Class Analysis permitted us to identify patterns among fosterparent couples that were not evident when looking at the sampleoverall.

Second, these findings are similar to other studies of foster mothersand fathers in that mothers' scores on all measures generally indicatedgreater potential for foster parenting than fathers'. In addition, overalland as expected, mothers' and fathers' scores were significantly corre-lated. However, this was not true in the small Bad-Discrepant sub-group, where mothers' and fathers' scores were quite discrepant. Dothe discrepancies between husbands and wives constitute a problem,in and of themselves? We don't know, but it is clear that couples whodisagreed on these important issues in our sample also appeared tohave less potential for competent, nurturing foster parenting.

It has been suggested that discrepancies about marital and familyfunctioning between foster mothers and fathers might indicate thatfathers are more objective and better gauges of family well-being. Forexample, Orme et al. (2004) reported that husbands perceived familyfunctioning to be poorer than did their wives, and Cautley (1980)found that fathers weremore accurate about howwell the foster familywas doing. Cautley speculated that, becausemothers generally initiatedbecoming foster parents and were often the primary caretakers, theywere more emotionally invested. Consequently, foster mothers mightbe less objective about problems with foster child placements due totheir desire to succeed as foster parents. By contrast, foster fathersmight be more able and willing to acknowledge household conflictand strain.

Further examination of the Bad-Discordant class suggests anotherexplanation, however. Nearly half of the fathers in this small groupdemonstrated depression scores above the clinical cut-off; this is instriking contrast to very few clinically depressed mothers and fathersin the other classes. It is possible that clinical depression is the sourceof the poor scores and the discrepancies, in two different ways. First,rigorous research in many settings and using a variety of measureshas reported depressive symptoms to be correlatedwith poor parenting

131J.G. Orme, T. Combs-Orme / Children and Youth Services Review 36 (2014) 124–132

in both mothers and fathers (Engle & McElwane, 2013; Wilson &Durbin, 2010). Second, the “depression-distortion hypothesis” suggeststhat clinically depressed individuals generally may appraise both them-selves and others in a more negative light (Najman et al., 2000). Bothsituationsmay be operation; that is, depressed fathersmay be less com-petent parents, and they also may be negatively biased in their reports.

Finally, and in keeping with the above point, this study emphasizesthe importance of looking specifically at foster fathers, both in couplesand as individuals. Research has shown that one competent parentcan sometimes compensate for a less nurturing parent (Chang,Halpern, & Kaufman, 2007), although in the current case the mothersin the Bad-Discordant group may be less able to do so because theirscores were also relatively low. Moreover, foster fathers may be espe-cially important to foster children because somany have not had fatherswhowere positive presences in their lives. In addition, it is important toask whether in some cases non-nurturing foster fathers might do addi-tional harm to children who have already suffered. More investigationabout foster fathers and their specific influences on foster children iswarranted.

Acknowledgments

The authors thank Cheryl Buehler, Donna J. Cherry, Tanya M.Coakley, Mary Ellen Cox, Gary Cuddeback, and Katie Rhodes for theirextensive help in the development and implementation of the overallstudy of which the present study is a part. We thank the many fosterparents who made this study possible.

References

Barber, B. K., Stolz, H. E., Olsen, J. A., & Maughan, S. L. (2005). Parental support, psycholog-ical control, and behavioral control: Assessing relevance across time, culture, andmethod. Monographs of the Society for Research in Child Development, 70(4, SerialNo. 282).

Belsky, J., & Fearon, R. P. (2004). Exploring marriage-parenting typologies and their con-textual antecedents and developmental sequelae. Development and Psychopathology,16, 501–523. http://dx.doi.org/10.1017/S095457940400464X.

Berrick, J.D., & Skivenes, M. (2012). Dimensions of high quality foster care: Parenting Plus.Children and Youth Services Review, 34, 1956–1965. http://dx.doi.org/10.1016/j.childyouth.2012.05.026.

Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction tometa-analysis. Chichester, UK: Wiley.

Buehler, C., Benson, M. J., & Gerard, J. M. (2006). Interparental hostility and early adoles-cent problem behavior: The mediating role of specific aspects of parenting. Journal ofResearch on Adolescence, 16, 265–292. http://dx.doi.org/10.1111/j.1532-7795.2006.00132.x.

Buehler, C., Krishnakumar, A., Stone, G., Anthony, C., Pemberton, S., Gerard, J., et al. (1998).Interparental conflict styles and youth problem behavior: A two-sample replicationstudy. Journal of Marriage and the Family, 60(1), 119–132.

Burgess, A. L., & Borowsky, I. W. (2010). Health and home environments of caregiversof children investigated by child protective services. Pediatrics, 125, 273–281.http://dx.doi.org/10.1542/peds.2008-3814.

Cautley, P. W. (1980). New foster parents: The first experience. New York: Human SciencesPress.

Chang, J. J., Halpern, C. T., & Kaufman, J. A. (2007). Maternal depressive symptoms, father'sinvolvement, and the trajectories of child problem behaviors in a US national sample.Archives of Pediatrics and Adolescent Medicine, 161(7), 697–703.

Cherry, D. J., & Orme, J. G. (2011). Validation study of a co-parenting scale for foster cou-ples. Journal of Public Child Welfare, 5, 564–589. http://dx.doi.org/10.1080/15548732.2011.617285.

Cole, S. A., & Eamon, M. E. (2007). Predictors of depressive symptoms among foster care-givers. Child Abuse & Neglect, 31, 295–310. http://dx.doi.org/10.1016/j.chiabu.2006.06.010.

Collins, L. M., & Lanza, S. T. (2010). Latent class and latent transition analysis: With applica-tions in the social, behavioral, and health sciences. Hoboken, NJ: Wiley.

Combs-Orme, T., Wilson, E., Cain, D. S., Page, T., & Kirby, L. D. (2003). Context-based par-enting in infancy: Background and conceptual issues. Child and Adolescent SocialWork, 20, 437–472. http://dx.doi.org/10.1023/B:CASW.0000003138.32550.a2.

Committee on Depression, Parenting Practices, and the Healthy Development of Children(2009). Depression in parents, parenting, and children: Opportunities to improve identi-fication, treatment, and prevention. Washington, D.C.: National Academies Press.

Crouch, J. L., & Behl, L. E. (2001). Relationships among parental beliefs in corporal punish-ment, reported stress, and physical child abuse potential. Child Abuse & Neglect, 25(3),413–419.

Dando, I., & Minty, B. (1987). What makes good foster parents? British Journal of SocialWork, 17(4), 383–399.

Deal, A. G., Trivette, C. M., & Dunst, C. J. (1988). Family functioning style scale: An instrumentfor measuring strengths and resources. Asheville, NC: Winterberry Press.

Delgado, P., & Pinto, V. S. (2011). Criteria for the selection of foster families and monitor-ing of placements: Comparative study of the application of the Casey Foster ApplicantInventory-Applicant Version (CFAI-A). Children and Youth Services Review, 33,1031–1038. http://dx.doi.org/10.1016/j.childyouth.2011.01.005.

Denuwelaere, M., & Bracke, P. (2007). Support and conflict in the foster family andchildren's well-being: Comparison between foster and birth children. FamilyRelations, 56(1), 67–79. http://dx.doi.org/10.1111/j.1741-3729.2007.00440.x.

Dickerson, J. L., Allen, M., & Pollack, D. (2011). How to screen adoptive and foster parents.Washington, D.C.: National Association of Social Workers.

Dziak, J. J., Lanza, S. T., & Tan, X. (in press). Effect size, statistical power and sample sizerequirements for the bootstrap likelihood ratio test in latent class analysis.Structural Equation Modeling (in press).

Eastman, K. S. (1982). Foster parenthood. Marriage & Family Review, 5(2), 95–120.England, M. J., & Sim, L. J. (Eds.). (2009). Depression in parents, parenting, and children:

Opportunities to improve identification, treatment, and prevention. Washington, DC:The National Academies Press.

Engle, J. M., & McElwane, N. L. (2013). Parental depressive symptoms and marital intima-cy at 4.5 years: Joint contributions to mother-child and father-child interaction at6.5 years. Developmental Psychology. http://dx.doi.org/10.1037/a0032450 (Advanceonline publication).

Fernandez, E. (2007). How children experience fostering outcomes: participatory re-search with children. Child & Family Social Work, 12, 349–359. http://dx.doi.org/10.1111/j.1365-2206.2006.00454.x.

Gillian, R. (2000). Men as foster carers. A neglected resource? Adoption & Fostering, 24(2),63–69.

Glass, G. V., & Stanley, J. C. (1970). Statistical methods in education and psychology. Engle-wood Cliffs, NJ: Prentice-Hall.

Goodman, S. H., Rouse, M. H., Connell, A.M., Broth, M. R., Hall, C. M., & Heyward, D. (2011).Maternal depression and child psychopathology: A meta-analytic review. Clinical Childand Family Psychology Review, 14, 1–27. http://dx.doi.org/10.1007/s10567-010-0080-1.

Grych, J. H. (2002). Marital relationships and parenting. Handbook of Parenting, 4,203–225.

Harden, B. J., Meisch, D. A., Vick, J. E., & Pandohie-Johnson, L. (2008). Measuring parentingamong foster families: The development of the Foster Parent Attitudes Questionnaire(FPAQ). Children and Youth Services Review, 30, 879–892. http://dx.doi.org/10.1016/j.childyouth.2007.12.015.

Henson, J. M., Reise, S. P., & Kim, K. H. (2007). Detecting mixtures from structural modeldifferences using latent variable mixture modeling: A comparison of relative modelfit statistics. Structural Equation Modeling: A Multidisciplinary Journal, 14, 202–226.http://dx.doi.org/10.1080/10705510709336744.

Holden, G. W., & Buck, M. J. (2002). Parental attitudes toward childrearing. In M. H.Bornstein (Ed.), Handbook of parenting (2nd ed.)Being and becoming a parent, Vol. 3,Mahwah, N.J.: Lawrence Erlbaum Associates.

Inch, L. J. (1999). Aspects of foster fathering. Child and Adolescent Social Work Journal,16(5), 393–412.

Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic data analysis. New York: Guilford.Lamb, M. E. (2012). Mothers, fathers, families, and circumstances: Factors affecting

children's adjustment. Applied Development Science, 16, 98–111. http://dx.doi.org/10.1080/10888691.2012.667344.

McLachlan, G. J., & Peel, D. (2000). Finite mixture models. New York: Wiley.Mennen, F. E., & Trickett, P. K. (2011). Parenting attitudes, family environments, depres-

sion, and anxiety in caregivers of maltreated children. Family Relations, 60,259–271. http://dx.doi.org/10.1111/j.1741-3729.2011.00646.x.

Muthén, B. (2008). Latent variable hybrids: Overview of old and new models. In G. R.Hancock, & K. M. Samuelsen (Eds.), Advances in latent variable mixture models(pp. 1–24). Charlotte, NC: Information Age Publishing.

Muthén, L. K., & Muthén, B. O. (1998-2012). Mplus user's guide (7th ed.)Los Angeles, CA:Muthén & Muthén.

Najman, J. M., Williams, G. M., Nikles, J., Spence, S. U. E., Bor, W., O'Callaghan, Michael,et al. (2000). Mothers' mental illness and child behavior problems: Cause–effect asso-ciation or observation bias? Journal of the American Academy of Child & Adolescent Psy-chiatry, 39(5), 592–602.

National Data Archive on Child Abuse and Neglect (2012). Adoption and Foster Care Analysisand Reporting System (AFCARS) Adoption file annual supplement, 2011 version 1 [Data fileand code book]. (Available from NDACAN site: http://www.ndacan.cornell.edu)

Nylund, K. L., Asparoutiov, T., & Muthén, B. (2007). Deciding on the number of classes inlatent class analysis and growth mixture modeling: A Monte Carlo simulation study.Structural Equation Modeling: A Multidisciplinary Journal, 14, 535–569.

Orme, J. G., Buehler, C., McSurdy, M., Rhodes, K. W., Cox, M. E., & Patterson, D. A.(2004). Parental and familial characteristics of family foster care applicants.Children and Youth Services Review, 26, 307–329. http://dx.doi.org/10.1016/j.childyouth.2004.01.003.

Orme, J. G., Cox, M. E., Rhodes, K. W., Coakley, T., Cuddeback, G. S., & Buehler, C. (2006).Casey Home Assessment Protocol (CHAP): Technical manual (2nd ed.)Knoxville, TN:University of Tennessee, Children's Mental Health Services Research Center.

Orme, J. G., Cuddeback, G. S., Buehler, C., Cox, M. E., & Le Prohn, N. (2006). Casey FosterApplicant Inventory (CFAI) Technical manual (2nd ed.)Knoxville, TN: University ofTennessee, Children's Mental Health Services Research Center.

Orme, J. G., Cuddeback, G. S., Buehler, C., Cox, M. E., & Le Prohn, N. S. (2007). Measuring fosterparent potential: Casey Foster Applicant Inventory—Applicant version. Research onSocial Work Practice, 17, 77–92. http://dx.doi.org/10.1177/1049731506295084.

Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the gen-eral population. Applied Psychological Measurement, 1, 385–401. http://dx.doi.org/10.1177/014662167700100306.

132 J.G. Orme, T. Combs-Orme / Children and Youth Services Review 36 (2014) 124–132

Radloff, L. S., & Locke, B. Z. (2008). Center for Epidemiologic Studies Depression Scale(CES-D). In A. J. RushJr., M. B. First, & D. Blacker (Eds.), Handbook of psychiatricmeasures (pp. 506–508) (2nd ed.). Washington, DC: American PsychiatricAssociation.

Ramaswany, V., DeSarbo, W., & Robinson, W. (1993). An empirical pooling approachfor estimating marketing mix elasticities with PIMS data. Marketing Science, 12,103–124. http://dx.doi.org/10.1287/mksc.12.1.103.

Ramchandani, P., & Psychogiou, L. (2009). Paternal psychiatric disorders and children'spsychosocial development. The Lancet, 374, 646–653. http://dx.doi.org/10.1016/S0140-6736(09)60238-5.

Riggs, D. W., Delafabbro, P. H., & Augostinos, M. (2010). Foster fathers and carework: En-gaging alternate models of parenting. Fathering, 8, 24–36. http://dx.doi.org/10.3149/fth.0801.24.

Ruiz, S. Y., Roosa, M. W., & Gonzalez, N. A. (2002). Predictors of self-esteem for MexicanAmerican and European American youths: A reexamination of the influence of

parenting. Journal of Family Psychology, 16, 70–80. http://dx.doi.org/10.1037/0893-3200.16.1.70.

Schaefer, E. S. (1965). Children’s reports of parental behavior: An inventory. ChildDevelopment, 36(2), 413–424.

Schwarz, G. (1978). Estimating the dimension of a model. Annals of Statistics, 6, 461–464.Sclove, S. L. (1987). Application of model-selection criteria to some problems in multivar-

iate analysis. Psychometrika, 52, 333–343.Shlonsky, A.R., & Berrick, J.D. (2001). Assessing and promoting quality in kin and nonkin

foster care. Social Service Review, 75, 60–83. http://dx.doi.org/10.1086/591882.Wilson, K., Fyson, R., & Newstone, S. (2007). Foster fathers: Their experiences and contri-

butions to fostering. Child & Family Social Work, 12, 22–31. http://dx.doi.org/10.1111/j.1365-2206.2006.00443.x.

Wilson, S., &Durbin, C. E. (2010). Effects of paternal depression on fathers’parentingbehav-iors: Ameta-analytic review. Clinical Psychology Review, 30, 167–180. http://dx.doi.org/10.1016/j.cpr.2009.10.007.