Parenting stress and marital relationship as determinants of mothers' and fathers' parenting

Personal Relationships, 20 (2013), 259–276. Printed in the United States of America.Copyright © 2012 IARR; DOI: 10.1111/j.1475-6811.2012.01404.x

INVITED REVIEW

Parenting stress and marital relationship asdeterminants of mothers’ and fathers’ parenting

KOEN PONNET,a DIMITRI MORTELMANS,a EDWIN WOUTERS,a KARLA VANLEEUWEN,b KIM BASTAITS,a AND INGE PASTEELSa

aUniversity of Antwerp and bKatholieke Universiteit Leuven

AbstractUsing the actor–partner interdependence model, we explore how various sources of stress and support experiencedby fathers and mothers influence their own parenting styles and the parenting styles of the partner. Data from 227couples were analyzed, with mothers and fathers rating their parenting stress and marital relationship and childrenrating the parenting styles of both mothers and fathers. Structural equation models revealed actor effects of parentingstress on demanding and responsive parenting styles, as well as partner effects between positive aspects of maritalrelationship and responsive parenting style. The results further indicate that the strength of these pathways is similarfor both mothers and fathers and do not support the hypothesis that the parenting of fathers is more vulnerable thanthe parenting of mothers.

Ever since the advent of Belsky’s (1984) eco-logical model of parenting, it has been widelyaccepted that contextual sources of stress andsupport have an influence on parenting, whichin turn affects child outcomes (Crnic & Low,2002). The ecological approach also empha-sizes that all parts of the family system areinterconnected and that parents and children

Koen Ponnet, Research Centre for Longitudinal and LifeCourse Studies, and Media & ICT in Organizations andSociety, University of Antwerp, Antwerp, Belgium; Dim-itri Mortelmans, Research Centre for Longitudinal andLife Course Studies, University of Antwerp, Antwerp,Belgium; Edwin Wouters, Research Centre for Longitu-dinal and Life Course Studies, University of Antwerp,Antwerp, Belgium; Karla Van Leeuwen, Parenting andSpecial Education Research Group, Katholieke Univer-siteit Leuven, Leuven, Belgium; Kim Bastaits, ResearchCentre for Longitudinal and Life Course Studies, Uni-versity of Antwerp, Antwerp, Belgium; Inge Pasteels,Research Centre for Longitudinal and Life Course Stud-ies, University of Antwerp, Antwerp, Belgium.

Correspondence should be addressed to Koen Pon-net, Research Centre for Longitudinal and Life CourseStudies, and Media & ICT in Organizations andSociety, University of Antwerp, Sint-Jacob Street 2,2000 Antwerp, Belgium, e-mail: [email protected],[email protected].

are best studied within this network (Bel-sky, 1984; Bronfenbrenner, 1989). To date,research on the determinants of parenting hasfocused largely on the role of mothers (Pleck,2010). Fathers, however, are currently moreinvolved in rearing their children than wasthe case in the past. Although most fathersdo not take as active a role in the parentingprocess as most mothers do, the gap betweenmen’s and women’s participation in childrea-ring appears to be shrinking (Amato, Meyers,& Emery, 2009; Lamb, 2010; Woodworth,Belsky, & Crnic, 1996). Despite this generaltrend, most studies on the determinants ofparenting that do include fathers have typi-cally analyzed data from mothers and fathersseparately or used composite or aggregatedscores, thereby ignoring the interdependenceand mutual influence between the two parents.

Mothers and fathers parent within the samefamilies. In addition to the effects of hisor her own level of stress, the parentingof one partner also is likely to be affectedby the other partner’s level of stress aswell. Moreover, the nature of the relation

259

260 K. Ponnet et al.

between the contextual sources of stress andparenting may vary for mothers and fathers(Barnett, Deng, Mills-Koonce, Willoughby,& Cox, 2008). According to the fathering-vulnerability hypothesis (Belsky, Gilstrap,& Rovine, 1984; Cummings, Goeke-Morey,& Raymond, 2004; Cummings & O’Reilly,1997), fathering and father–child relation-ships might be more vulnerable to the stressof marital discord than is the case with moth-ering and mother–child relationships. In otherwords, negative marital relations might havestronger effects on father–child relationshipsthan they do on mother–child relationships(Cummings, Merrilees, & George, 2010). Onepossible explanation for the increased vul-nerability of fathers is that, for men, theroles of father and husband may be lessdistinct than the roles of mother and wifeare for women, thus making fathering moresensitive to marital stress or other exter-nal influences (Belsky, Youngblade, Rovine,& Volling, 1991; Coiro & Emery, 1998).Although it has not been the subject of exten-sive research, evidence is mixed regardingthe fathering-vulnerability hypothesis. On onehand, results of several studies suggest thatthe association between marital relationshipand parenting is not moderated by the genderof the parent. For example, in their meta-analysis of marital conflict, Erel and Bur-man (1995) found no evidence to supportthe assumption that marital quality in intactfamilies is differentially related to the qual-ity of mothering and fathering. On the otherhand, in a review on mothering and fathering,Coiro and Emery (1998) concluded that thereis tentative support for the hypothesis thatfather–child relationships are more sensitiveto marital conflict than mother–child relation-ships. Similarly, results from Krishnakumarand Buehler’s (2000) meta-analytic review ofrelations between interparental conflict andparenting provided at least some support forthe fathering-vulnerability hypothesis. Bothsupportive and nonsupportive results shouldbe interpreted with caution, however, giventhat analyses of mother–father differencesare not always based on within-couple com-parisons (Coiro & Emery, 1998; Cummingset al., 2004; Cummings et al., 2010). As such,

researchers focusing on the individual levelof analysis ignore nonindependence in dyadmembers’ scores on standard significance test-ing, which result in biased variances (Kenny,Kashy, & Cook, 2006).

In this study, we move beyond these lim-itations by analyzing data from both par-ents within the same family. By focusingon effects within and between partners, wecan differentiate between a personal and rela-tional component, because a parent’s parent-ing style depends on feelings of stress andsupport of his or her partner, besides hisor her own feelings of stress and support(Kenny et al., 2006). Our research makes sev-eral important contributions to the literatureon the contextual sources of stress and sup-port as determinants of parenting. First, wefocus on parenting stress (childrearing stressand feelings of role restriction), positive (mar-ital quality and partner support) and negative(ineffective arguing) aspects of marital rela-tionship as determinants of parenting. Whileparenting stress is situated at the parent–childlevel, the marital relationship is situated atthe parent–parent level. Second, consideringthe dyadic nature of parenting, we investigatehow sources of stress and support experiencedby fathers and mothers influence their respec-tive parenting styles, and we examine whetherthe strengths of the effects are similar forboth parents. Third, it can be assumed thatan individual’s parenting style is influencednot only by his or her own feelings of stressand support but also by the stress and sup-port experienced by the partner. We thereforeexamine possible pathways between the twoparents and test for gender differences. Fourth,we contribute to the literature using rigor-ous methodological techniques. All dependentand independent variables are treated as latentconstructs, using confirmatory factor analyses(CFA). The use of CFA offers the advan-tage of estimating relations among variables,while adjusting for measurement error. Tomodel the interdependence of dyad membersand the mutual influence between the twoparents, our analyses are grounded on theactor–partner interdependence model (APIM;Kashy & Kenny, 2000; Kenny et al., 2006),a family-system approach that proposes that

Stress, marital relationship, and parenting styles 261

the predictor variables of both the respon-dent (actor effects) and the respondent’s part-ner (partner effects) influence the respondent’soutcome variable (Yucel & Gassanov, 2010).The APIM allows the testing of both actorand partner effects. It also allows comparisonof actor and partner effects for both mothersand fathers and to test specific combinationsof actor and partner effects. Although the rela-tion between contextual sources of stress andsupport on one hand and parenting styles onthe other has frequently been studied, dyadicanalyses using such rigorous statistical con-trols are scarce (Nelson, O’Brien, Blankson,Calkins, & Keane, 2009).

Theoretical framework

The family system approach highlights theidea that the family is a complex, inte-grated whole, in which individual familymembers are necessarily interdependent (Cox& Paley, 1997; Minuchin, 1974). Severalprocesses have been proposed to explainhow family members influence one another(Erel & Burman, 1995; Repetti, 1987). Inthis study, we focus on two of these pro-cesses: spillover and crossover. Spilloveroccurs when an individual brings experiencesor feelings from one domain (e.g., the par-ent domain) into another domain (e.g., theparent–child domain). Transfer occurs in thesame valence. For example, father’s stressmight be linked to a less responsive parent-ing style. Crossover refers to the transfer ofexperiences or affect between people. Oneexample of crossover is when the stress expe-rienced by one parent is detrimental to theparenting of the other partner.

The APIM is a specific multisource family-system approach that uses the parent dyad asthe unit of analysis (Fincham & Beach, 2010).The model can be used to assess spilloverand crossover effects. The APIM approach tothe examination of family functioning is quiterecent. It was designed to estimate the impactof the independent variables of individuals ontheir own dependent variables (actor effects),as well as on the dependent variables of theirpartners (partner effect). It implies that thetwo members of the dyad influence each other

in the form of partner effects, which createinterdependence between members (Leder-mann & Macho, 2009). Because most studieson parenting use the individual as the unit ofanalyses, they focus exclusively on the waysin which the feelings of parents (e.g., mother’sfeelings of stress) are associated with theirparenting (i.e., actor effects), and they ignorethe mutual interdependence of family relation-ships. The use of the APIM provides a healthycorrective measure for past practices, in whichinterdependency in dyadic data was unrec-ognized, ignored, or addressed in suboptimalways (Fincham & Beach, 2010). In the APIMmodel, actor effects are indicative of spillovereffects, while partner effects are indicative ofcrossover effects. Applying the APIM notionto this study, we explored various pathways ofinfluence among fathers and mothers. Morespecifically, we examined actor and partnereffects between parenting stress and parentingstyles, as well as between marital relation-ship and parenting styles, and tested specificcombinations of actor and partner effects todiscern four specific patterns in the APIM:the actor only, the partner only, the couple,and the contrast pattern (Kenny & Ledermann,2010). The actor-only pattern is indicated if aperson’s parenting styles are a function of thatperson’s feelings of stress and support only,while the partner’s feelings of stress and sup-port have no impact. The partner-only patterntakes place if a person is affected by the part-ner’s feelings of stress and support, but notby his own feelings. In a couple-oriented pat-tern, the actor and partner effects are equalsuch that a parent is affected as much by hisor her own feelings of stress and support asby those of the partner. The contrast patternoccurs if actor and partner effects are equalin size but have opposite signs. For example,a person’s demanding parenting style is pos-itively affected by his or her parenting stressand negatively affected by the partner’s par-enting stress (Kenny et al., 2006; Kenny &Ledermann, 2010).

Parenting styles and contextual sourcesof stress and support

We conceptualize parenting styles in line withDarling and Steinberg’s (1993) definition of


styles as a reflection of the emotional climatein which socialization occurs. Although par-enting styles and parenting practices areoften used as interchangeable concepts, par-enting practices are distinct from parentingstyles. While parenting practices are directedtoward particular goals, parenting styles canbe regarded as the general context or climatein which the more specific parenting practicesare expressed (Darling & Steinberg, 1993).

Baumrind (1991) outlines two independentdimensions of parenting. The first dimen-sion, responsiveness, refers to the degreeof parental warmth, emotional expressive-ness, and positive reinforcement of the child.The second dimension, demandingness, refersto parental discipline, control, and level ofdemands. Following Baumrind’s perspective,we used a multidimensional measure that cap-tures parental responsiveness and demanding-ness (see the Method section).

Several contextual sources of stress andsupport have been associated with both dimen-sions of parenting style, two of which areexamined in this study: parenting stress andmarital relationship. In the following section,we briefly outline the two sources of stressand support and how they may affect parent-ing style.

Parenting stress and parenting styles

Stress is a multidimensional construct that canbe operationalized in various ways, depend-ing on the source or context of the stres-sor (Crnic, Arbona, Baker, & Blacher, 2009;Crnic & Low, 2002). Parenting stress is aspecific kind of stress, which can be distin-guished from stress in other domains of life(e.g., work stress and marital relationship).It refers to a condition or feeling experi-enced when parents perceive that the demandsassociated with parenting exceed the personaland social resources that are available tothem to meet those demands (Abidin, 1990;Cooper, McLanahan, Meadows, & Brooks-Gunn, 2009; Deater-Deckard, Smith, Ivy, &Petrill, 2005). Although parents may varyin the ways in which they handle stress(Bronte-Tinkew, Horowitz, & Carrano, 2010),all parents experience parenting stress to some

degree (Crnic & Greenberg, 1990; Hakvoort,Bos, Van Balen, & Hermanns, 2010). Thegender of parents, however, has been foundto be only modestly associated with levelsof parenting stress (Deater-Deckard, Dodge,Bates, & Pettit, 1998; Deater-Deckard &Scarr, 1996) Moreover, studies measuringparenting stress in both parents suggest thatmothers and fathers within the same familymight be more similar than different in theirlevels of parenting stress (Deater-Deckardet al., 1998; Deater-Deckard & Scarr, 1996).

A substantial body of literature documentsthe relation between parenting stress andparenting. More specifically, research indi-cates that parents reporting greater levelsof parenting stress are more demanding andless responsive in their parenting styles, andthey are less involved with their children(Belsky, Woodworth, & Crnic, 1996; Crnic,Greenberg, Ragozin, Robinson, & Basham,1983; Deater-Deckard, 1998; Deater-Deckard& Scarr, 1996). Despite this well-documentedrelation, the degree to which parenting stressaffects the parenting styles and involvementof mothers and fathers is not yet clearlyunderstood (Crnic, Gaze, & Hoffman, 2005;Fagan, Bernd, & Whiteman, 2007; Nelsonet al., 2009). Most studies on the relationbetween parenting stress and parenting focuson the role of mothers (Crnic et al., 2005;Crnic et al., 2009; Hutch-Bocks & Hughes,2008; Rodgers, 1998), and the few studiesthat have involved fathers analyzed data sep-arately from mothers and fathers (Bronte-Tinkew et al., 2010; Isacco, Garfield, &Rogers, 2010). Because the father is nowexpected to be an equal coparent with themother (Crnic et al., 2009; Pleck, 2010), itis important to determine the degree to whichthe paths from parenting stress to parentingstyles are uniform for mothers and fathers.

Marital relationship and parenting styles

Marital relationship is a broad construct,encompassing positive aspects such as partnersupport and marital quality and negativeaspects such as hostility or disagreement, andstudies investigating the association betweenmarital relationship and parenting differ in


the way they define and measure maritalrelationship (Kjobli & Hagen, 2009). Overthe past decades, considerable research has,however, demonstrated links between mari-tal relationship and parenting (Erel & Bur-man, 1995; Kaczynski, Lindahl, Mailik, &Laurenceau, 2006; Krishnakumar & Buehler,2000), although the direction of the effectsis not always clear (Grych, 2002). Consis-tent with the spillover hypothesis, a meta-analytic study by Erel and Burman (1995)reveals modest but consistent significant asso-ciations between marital quality and high par-ent–child relationship quality. Likewise, amore recent meta-analysis by Krishnakumarand Buehler (2000) demonstrates a signifi-cant negative relation between interparentalconflict and positive parenting. Little atten-tion, however, has been paid to crossover orpartner effects. In fact, most studies on theinfluence of marital relationship on parentingmostly focus on mothers (Benson, Buehler, &Gerard, 2008; Buehler & Gerard, 2002), andstudies that do involve both parents typicallyconsider marital relationship as a unit, with-out studying fathers separately (Buehler, Ben-son, & Gerard, 2006; Kaczynski et al., 2006;Kjobli & Hagen, 2009). There is nonethelesssome support for the crossover contention.For example, a recent article by Nelson andcolleagues (2009) reports crossover or part-ner effects between marital dissatisfaction andparenting behavior, although the effects are atthe trend level. The focus of Nelson and col-leagues’ study, however, was on supportiveand nonsupportive techniques that parents useto teach children about emotions rather thanon various dimensions of parenting style.

Research questions and hypotheses

Focusing on the APIM, the main aim ofthis study is to assess actor and partnereffects of parenting stress or marital relation-ship on parenting styles. More specifically,we address the following research questionsand/or hypotheses.

First, we examine actor effects of parent-ing stress, positive and negative aspects ofmarital relationship on the parenting stylesof mothers and fathers. On the basis of the

literature on parenting stress (Crnic & Low,2002), we anticipate significant actor effectsfor mothers and fathers, with higher levels ofchildrearing stress and role restriction associ-ated with a more demanding and less respon-sive parenting style. Furthermore, consistentwith other studies on marital relationship (Erel& Burman, 1995; Krishnakumar & Buehler,2000), we expect high marital quality andpartner support and low ineffective arguingto be associated with less demanding andmore responsive parenting styles. Second, weexamine whether the actor effects are equalfor mothers and fathers. On the basis of thefathering-vulnerability hypothesis, we expectthe parenting styles of fathers to be more sus-ceptible to deterioration in the face of stressderived from the marital relationship, which isineffective arguing. We expect no gender dif-ferences in the pathways from parenting stressto parenting styles, given that parenting stressis situated at the parent–child level, such thatfathers do not have to differentiate betweentheir roles of husband and father nor do moth-ers have to distinguish between their roles asmothers and their roles as wives. Third, wetest for partner effects, expecting that parent-ing stress, positive and negative aspects ofmarital relationship from one parent crossesover to the parenting style of the other part-ner, in the same ways as the actor effects.Fourth, we test for gender differences in part-ner effects. We make no specific hypotheses inthis regard, however, due to the general lackof previous literature on this topic. Finally,we test the relative size of the actor and part-ner effects to discern four specific patterns:the actor only, the partner only, the cou-ple, and the contrast pattern. On the basis ofthe findings of Nelson and colleagues (2009),we expect actor-oriented models for parentingstress as well as positive and negative aspectsof marital relationship; that is, actor effectswill be more prominent than partner effects.

Method

Procedure

The sample for this study was selected fromsubjects participating in the interuniversity


Relations in Flanders (RiF) project sponsoredby the agency for Innovation by Science andTechnology. The RiF project is based on aunique multiactor design, in which the childand both parents are interviewed. The researchpopulation was restricted to people in theDutch-speaking part of Belgium, who wereeither still in their first marriage or who hadexperienced one divorce. The sample wasdrawn from the Belgian National Register. Ifthe marriage was intact, both partners wereinterviewed face-to-face (Computer-AssistedPersonal Interview) in their current house-holds. If a divorce or separation had takenplace, both partners were interviewed face-to-face in their new residences. In additionto the parents, one resident child was inter-viewed face-to-face, provided that the childhad reached the age of 10. In case of multiplechildren above 10 years of age in the family,only the child whose birthday was closest tothe date of the interview was selected. Eventhough the same questions were asked of eachchild, the questionnaire was adapted to spe-cific characteristics of the child’s age (10–13,14–17, and 18+ years old).

Participants

For this study, we used a subsample of theRiF data: only nondivorced families (n =227) with a child ranging in age from 10to 18 years were included. We used mothers’(n = 223) and fathers’ (n = 200) reports ofthe various sources of stress, whereas we usedchild reports (n = 227) of the two dimensionsof parenting style. This approach is based onthe assumption that mothers and fathers aremore qualified than their children are to eval-uate their own levels of stress, while childrenare in a better position than their parents toreport on parenting styles. Self-reports fromchildren may be the most valid way of mea-suring parenting styles, as the feeling of beingcontrolled or criticized is very much a sub-jective experience (Aunola, Stattin, & Nurmi,2000; Barber, 1996; Litovsky & Dusek, 1985;Wentzel, 1994). Furthermore, parents’ reportsof their own parenting styles might be sub-ject to self-serving biases, as prior researchshows that parents rate their own styles and

skills more favorably than their children do(Purdie, Carroll, & Roche, 2004).

The average age of the children whowere interviewed was 14.12 years (SD =2.58), with 47% boys (n = 106) and 53%girls (n = 121). Univariate analysis of vari-ance revealed no between-group differencesfor age, F(1, 226) < 1. The average ageof the fathers who were interviewed was44.29 years (SD = 3.84, range = 28), and theaverage age of the mothers was 42.88 years(SD = 42.88, range = 21). A paired t testrevealed a significant difference between themean ages of the fathers and the mean ages ofthe mothers, t (195) = −6.98, p < .001. Edu-cation was measured as the highest level ofeducation achieved. The educational level offathers was significantly different from thatof mothers, χ2(4) = 38.09, p < .001. Withinour sample, 6.7% of the mothers and 12%of the fathers had completed no education oronly primary education, 39% of the mothersand 37.5% of the fathers had completed sec-ondary education, and 54.3% of the mothersand 50.5% of the fathers had completed atleast 3 years of higher education.

Analyses

To test our research questions and hypotheses,we used structural equation modeling (SEM),following the procedures outlined by Kennyand colleagues (2006). Raw data were struc-tured as triadic data. In other words, each linerepresented a triad, with variables reflectingscores from the mother, the father, and thechild. Statistical analyses were conducted inMplus 5.2 (Muthen & Muthen, 2008) withmaximum likelihood estimation. With SEM,it is possible to model several variables simul-taneously and to compare the relative magni-tudes of various regression paths.

We tested the fit of several successivemodels. First, we conducted CFA on all multi-item scales to identify whether the constructsare adequately measured by the indicators.Second, we built measurement models forthe relations between each of the five latentpredictors and parents’ demandingness andresponsiveness. Finally, we conducted severalAPIM structural equation models, for each


predictor variable (i.e., parenting stress ormarital relationship) and for each parentingstyle (i.e., demandingness or responsiveness),in which we evaluated actor and partnereffects among the latent variables. We alsotested whether actor and partner effects weresignificantly different for mothers and fathersand tested whether the models were moreactor, partner, contrast, or couple oriented.

The model fits of the CFA’s measurementand path models were evaluated according toseveral fit indices. Given that the χ2 is almostalways significant and not an adequate testof the model fit (Brown, 2006; Kline, 2005),we therefore report a χ2/df ratio as well. Aχ2/df ratio of 2:1 to 5:1 is required, andit indicates an acceptable fit, although val-ues of < 3 are considered favorable (Kline,2005). In addition, we examined the com-parative fit index (CFI; Bentler, 1990), theTucker–Lewis index (TLI; Tucker & Lewis,1973), root mean square error of approxima-tion (RMSEA; Steiger, 1990), and the stan-dardized root mean square residual (SRMR;Kline, 2005). The CFI and TLI range from 0to 1.00, with a cutoff of 0.95 or higher indicat-ing that the model provides a good fit and 0.90indicating that the model provides an ade-quate fit (Byrne, 2001; Hu & Bentler, 1999).RMSEA values below 0.05 indicate a goodmodel fit, and values between 0.06 and 0.08indicate an adequate fit (Brown, 2006; Raykov& Marcoulides, 2006). The SRMR is a stan-dardized summary of the average covarianceresiduals (Kline, 2005). A relatively goodmodel fit is indicated when the SRMR issmaller than 0.08 (Hu & Bentler, 1999).

Measures

With the exception of the control vari-ables (e.g., age of the parents), all measuresdescribed below were treated as unidimen-sional latent variables, which we constructedusing CFA with Mplus 5.2 (Muthen &Muthen, 2008). We investigated how thevarious indicators are related to the latentfactors, in addition to the relations amongindicator errors. Thereafter, we assessed thescale-composite reliability (ρ) for each latentconstruct, thus providing an appropriate and

desirable estimate of their reliability (Raykov,2009).

Parenting stress

Parenting stress was measured using thechildrearing stress scale and the role-restrictionscale (Van den Troost, 2005).

The childrearing stress of the parents wasmeasured according to three items, adaptedfrom a study by Van den Troost (2005).Each of the items is intended to reflect thedegree to which parents report experienc-ing childrearing as burdensome and problem-atic. Both mothers and fathers were asked torate the items along a 7-point Likert scaleranging from 1 (definitely disagree) to 7 (def-initely agree). Items are: “Raising my daugh-ter/son brings about a lot more problemsthan I expected”; “Raising my daughter/sonis harder than I thought it would be”; and“Raising my daughter/son frequently causesproblems.” These three items served as indi-cators of the latent construct representing thechildrearing stress of mothers and fathers.The initial model provided a good fit for thedata, χ2(8) = 10.81, p = .21; χ2/df = 1.35,CFI = 0.99, TLI = 0.99, RMSEA = 0.039,SRMR = 0.017. Factor loadings ranged from0.80 to 0.83 for mother reports and from0.77 to 0.86 for father reports. Both con-structs were interdependent (r = .53, p <

.001). Scale-composite reliability (ρ) was 0.86for the mothers as well as for the fathers.

Role restrictions were measured accord-ing to four items, adapted from a study byVan den Troost (2005). Both mothers andfathers rated the degree to which they feelrestricted by their parenting and childrearingroles in arranging their personal lives and ful-filling their personal interests (e.g., “Raisingmy children prevents me from doing thingsthat are important to me” and “Because ofyour children, you cannot plan your life asyou want it”). All the items were scored alonga 7-point Likert scale ranging from 1 (stronglydisagree) to 5 (strongly agree). The initialmodel showed a good fit, χ2(19) = 35.08,p < .05; χ2/df = 1.85, CFI = 0.97, TLI =0.96, RMSEA = 0.060, SRMR = 0.036. Fac-tor loadings ranged from 0.57 to 0.81 for


mother reports and from 0.59 to 0.85 forfather reports. Both constructs were interde-pendent (r = .33, p < .001). Scale-compositereliability (ρ) was 0.82 for mothers and 0.83for fathers.

Marital relationship

Positive marital relationship was measuredusing the Quality of Marriage Index (QMI;Norton, 1983) and a five-item Partner-SupportScale (Dykstra, van Tilburg, & Gierveld,2005). Negative marital relationship was mea-sured using the Ineffective Arguing Inventory(IAI; Kurdek, 1994).

The QMI (Norton, 1983) consists of sixitems measuring global perceptions of mar-ital satisfaction. The six items assess thenature of the marriage and quality of therelationship (e.g., “My relationship with mypartner makes me happy”). Both mothersand fathers rated the extent to which theyagreed with evaluative statements about theirmarriages, with five items anchored with 1(very strong disagreement) and 7 (very strongagreement) and the sixth anchored with 1(very unhappy) and 10 (perfectly happy).High scores on all items indicate a qual-ity marriage. All fit indices of the initialmodel were acceptable, with the exceptionof one score (RMSEA = 0.096). A modi-fication index suggested to freely estimatethe error covariance between two similarlyworded mother-reported items: “My relation-ship with my partner is very stable” and “Myrelationship with my partner is strong”. Byallowing this covariance, the model providedan adequate fit for the data, χ2(52) = 112.68,p < .001; χ2/df = 2.17, CFI = 0.98, TLI =0.97, RMSEA = 0.072, SRMR = 0.030. Fac-tor loadings ranged from 0.79 to 0.94 formother reports and from 0.74 to 0.96 forfather reports. Both constructs were interde-pendent (r = .71, p < .001). Scale-compositereliability (ρ) was 0.94 for both mothers andfathers.

Partner support was measured accord-ing to five items adapted from the Nether-lands Kinship Panel Study (Dykstra et al.,2005). Mothers and fathers rated the extentto which they receive support from their

partners regarding various life domains (e.g.,use of leisure and decisions about work). Allthe items were scored along a 4-point Lik-ert scale ranging from 1 (no support fromthe partner) to 5 (strong support from thepartner). The initial model showed a goodfit, χ2(34) = 39.79, p = .23; χ2/df = 1.17,CFI = 0.99, TLI = 0.99, RMSEA = 0.027,SRMR = 0.033. Factor loadings ranged from0.71 to 0.85 for mother reports and from 0.54to 0.79 for father reports. Neither constructwas independent (r = .50, p < .001). Scale-composite reliability (ρ) was 0.88 for mothersand 0.83 for fathers.

The eight-item IAI (Kurdek, 1994) wasused to assess the extent to which respondentsand their spouses engaged in couple-levelpatterns of ineffective arguing. Mothers andfathers were instructed to indicate the extentto which they agreed (1 = strongly disagree to5 = strongly agree) that each statement (e.g.,“Overall, our arguments are brief and quicklyforgotten”) fits their relationship. Confirma-tory factor analysis revealed that all indicatorsloaded significantly with the latent construct,although one father-reported indicator loadedonly 0.34. We decided to omit this item.The model showed a relatively acceptablefit, but it could be improved by allowing anerror covariance between two closely relatedfather-reported items (“Overall, I’d say we’repretty good at solving our problems” and“Overall, our arguments are brief and quicklyforgotten”). This resulted in an adequate-to-good fit of the model, χ2(102) = 199.39,p < .001; χ2/df = 1.95, CFI = 0.93, TLI =0.92, RMSEA = 0.065, SRMR = 0.057. Fac-tor loadings ranged from 0.45 to 0.89 formother reports and from 0.47 to 0.86 forfather reports. Both constructs were interde-pendent (r = .60, p < .001). Scale-compositereliability (ρ) was 0.86 for mothers and 0.85for fathers.

Parenting style

The Parenting Style Inventory II (Darling,Cumsille, & Pena-Alampay, 2005; Darling &Toyokawa, 1997) was administered to chil-dren to assess the parenting styles of mothersand fathers. The Parenting Style Inventory


was designed to assess the construct of parent-ing style independently of parenting practice(Darling & Toyokawa, 1997). Responsivenessassesses the extent to which parents showaffective warmth, acceptance, and involve-ment (e.g., “I can count on my mother to helpme out if I have a problem”). Demandingnessrefers to the extent to which parents show con-trol and supervision in their parenting (e.g.,“My mother really expects me to follow fam-ily rules”). All the items were scored along a5-point Likert scale ranging from 1 (stronglydisagree) to 5 (strongly agree).

Confirmatory factor analysis on the itemsof the demandingness subscale revealed thatone indicator (“If I don’t behave myself,my mother/father will punish me”) did notload significantly with the latent constructof mothers and that of fathers. After omit-ting this item, all indicators loaded signifi-cantly on the latent construct, although oneindicator’s loading was still low (0.18 formothers and 0.17 for fathers). We thereforeomitted this indicator as well, resulting in amodel with three indicators for demanding-ness on the part of the mother and on thepart of the father. To improve the model,we freed up an error covariance between thesame items in the latent construct of moth-ers and that of fathers (“My mother/fatherreally lets me get away with things”). Themodel then showed a good fit, χ2(7) = 10.93,p = .14; χ2/df = 1.56, CFI = 0.99, TLI =0.97, RMSEA = 0.050, SRMR = 0.040. Fac-tor loadings ranged from 0.38 to 0.92 formothers’ demandingness and from 0.52 to0.95 for fathers’ demandingness. Both con-structs were interdependent (r = .55, p <

.001). Scale-composite reliability (ρ) was 0.63for mothers’ demandingness and 0.74 forfathers’ demandingness.

Confirmatory factor analysis on the itemsof the responsiveness subscale revealed thatall indicators loaded significantly with thelatent construct, although one indicator withfathers’ responsiveness loaded only 0.22.Because the responsiveness of both moth-ers and fathers was rated by the same per-son, we decided to omit this item in thelatent constructs of both fathers and moth-ers. After omitting this item, factor loadings

ranged from 0.46 to 0.79 for mothers’ respon-siveness and from 0.44 to 0.78 for fathers’responsiveness. The model showed an accept-able fit to the data, χ2(19) = 46.94, p <

.01; χ2/df = 2.47, CFI = 0.94, TLI = 0.91,RMSEA = 0.080, SRMR = 0.053. Both con-structs were interdependent (r = .45, p <

.001). Scale-composite reliability (ρ) was 0.71for mothers’ responsiveness and 0.76 forfathers’ responsiveness.

Control variables

Researchers who have investigated parentingindicate that sociodemographic characteristics(e.g., educational level of the parents) areoften associated with parenting styles (Frias-Armenta & McCloskey, 1998; Simons, Whit-beck, Conger, & Melby, 1990). To determinewhether sociodemographic variables shouldbe included as covariates in the analyses,we examined the relation between the ageand gender of the child, the age and educa-tional levels of the mother and the father, andthe outcome variables. The analysis revealeda number of significant associations amongthe sociodemographic variables considered.Mother’s age was significantly associatedwith mother’s demandingness (β = −.173,SE = 0.075, p = .02), and father’s age wassignificantly associated with father’s demand-ingness (β = −.155, SE = 0.076, p = .041).Mother’s age and father’s age were there-fore included as covariates in the analysespredicting mother’s and father’s demanding-ness, respectively. Child age was significantlyassociated with father’s responsiveness (β =−.154, SE = 0.080, p = .054), and it wastherefore included as a covariate in the anal-yses predicting father’s responsiveness.

Results

Measurement models

We evaluated the model fit of all measure-ment models. As shown in Tables 1 and 2,all measurement models fitted the data well.Because the objectives of our research requirethe inclusion of separated scores for moth-ers’ and fathers’ latent constructs, we con-ducted for each construct a test to specify

268 K. Ponnet et al.Ta

ble

1.P

aren

ting

stre

ss,m

arit

alre

lati

onsh

ip,a

ndde

man

ding

ness

Test

ing

for

acto

rpa

rtne

ref

fect

sTe

stin

gfo

rge

nder

diff

eren

ces

Mot

her

Fath

erM

odel

fita

Dem

andi

ngne

ssβ

SEB

SER

2bR

2cχ

2df

pχ

2/d

fC

FIT

LI

RM

SEA

SRM

R

Chi

ldre

arin

gst

ress

Act

orM

othe

r0.

070.

100.

060.

02∗

5.7

7.5

66.6

630.

353

1.06

1.00

1.00

0.01

60.

040

Fath

er0.

220.

10∗

67.7

650.

384

1.04

1.00

1.00

0.01

40.

039

Part

ner

Mot

her

0.14

0.10

0.03

0.03

Fath

er0.

030.

10R

ole

rest

rict

ion

Act

orM

othe

r0.

030.

090.

010.

022.

42.

411

1.3

900.

064

1.24

0.98

0.97

0.03

20.

048

Fath

er0.

020.

0811

2.3

920.

074

1.22

0.98

0.97

0.03

10.

048

Part

ner

Mot

her

−0.0

60.

090.

000.

02Fa

ther

0.04

0.09

QM

I4.

62.

9A

ctor

Mot

her

0.23

0.11

∗0.

040.

0324

4.7

155

0.00

01.

580.

970.

970.

050

0.04

3Fa

ther

−0.1

00.

1124

6.4

157

0.00

01.

570.

970.

970.

050

0.04

3Pa

rtne

rM

othe

r−0

.18

0.11

−0.0

20.

03Fa

ther

0.09

0.11

Part

ner

supp

ort

3.9

2.3

Act

orM

othe

r−0

.04

0.09

−0.0

40.

0514

3.6

121

0.07

91.

190.

980.

980.

029

0.04

6Fa

ther

−0.0

30.

1014

4.6

123

0.08

81.

180.

980.

980.

028

0.04

6Pa

rtne

rM

othe

r−0

.11

0.10

−0.0

60.

05Fa

ther

−0.0

30.

09IA

I2.

63.

5A

ctor

Mot

her

−0.1

00.

110.

010.

0830

0.4

215

0.00

01.

400.

950.

950.

042

0.05

3Fa

ther

0.14

0.11

301.

821

70.

000

1.39

0.95

0.95

0.04

10.

053

Part

ner

Mot

her

0.05

0.11

−0.0

20.

07Fa

ther

−0.0

80.

10

Not

e.C

FI=

com

para

tive

fitin

dex;

TL

I=

Tuc

ker-

Lew

isin

dex;

RM

SEA

=ro

otm

ean

squa

reer

ror

ofap

prox

imat

ion;

SRM

R=

stan

dard

ized

root

mea

nsq

uare

resi

dual

;Q

MI

=Q

ualit

yof

Mar

riag

eIn

dex;

IAI

=In

effe

ctiv

eA

rgui

ngIn

vent

ory.

a Fir

stlin

esre

fer

toth

em

odel

fitof

mea

sure

men

tm

odel

,se

cond

lines

refe

rto

the

fitof

stru

ctur

alm

odel

s.bIn

clud

ing

mot

her

age.

c Inc

ludi

ngfa

ther

age.

∗ p≤

.05.


Tabl

e2.

Par

enti

ngst

ress

,mar

ital

rela

tion

ship

,and

resp

onsi

vene

ss

Test

ing

for

acto

rpa

rtne

ref

fect

sTe

stin

gfo

rge

nder

diff

eren

ces

Mot

her

Fath

erM

odel

fit

Res

pons

iven

ess

BSE

BSE

R2a

R2b

χ2

dfp

χ2/d

fC

FIT

LI

RM

SEA

SRM

R

Chi

ldre

arin

gst

ress

Act

orM

othe

r−0

.10

0.11

−0.0

30.

02<

16.

311

6.6

810.

006

1.44

0.97

0.96

0.04

40.

046

Fath

er−0

.06

0.11

117.

282

0.00

71.

430.

970.

960.

043

0.04

7Pa

rtne

rM

othe

rd

0.04

0.11

−0.0

20.

02Fa

ther

−0.1

40.

10R

ole

rest

rict

ion

Act

orM

othe

r−0

.13

0.09

−0.0

60.

02∗∗

0.03

7.3

156.

911

00.

002

1.43

0.96

0.95

0.04

30.

054

Fath

er−0

.21

0.09

∗15

8.8

111

0.00

21.

430.

960.

950.

044

0.05

5Pa

rtne

rM

othe

r−0

.07

0.10

0.02

0.02

Fath

er0.

160.

09†

QM

IA

ctor

Mot

her

0.03

0.12

−0.0

30.

03<

19.

526

3.3

179

0.00

01.

470.

980.

970.

046

0.04

4Fa

ther

−0.1

60.

1226

3.9

180

0.00

01.

470.

980.

970.

045

0.04

4Pa

rtne

rM

othe

r−0

.01

0.13

0.09

0.03

∗∗Fa

ther

0.34

0.11

∗∗Pa

rtne

rSu

ppor

tA

ctor

Mot

her

0.03

0.10

−0.0

10.

05<

16.

919

6.1

143

0.00

21.

370.

960.

960.

040

0.05

1Fa

ther

−0.0

60.

1019

7.0

144

0.00

21.

270.

960.

960.

040

0.05

0Pa

rtne

rM

othe

r0.

050.

110.

120.

05∗

Fath

er0.

230.

09∗

IAI

Act

orM

othe

r−0

.12

0.11

−0.0

20.

081.

16

372.

924

20.

000

1.54

0.93

0.92

0.04

90.

057

Fath

er0.

060.

1137

3.7

243

0.00

01.

540.

930.

920.

049

0.05

2Pa

rtne

rM

othe

r0.

060.

12−0

.08

0.09

Fath

er−0

.21

0.11

∗

Not

e.C

FI=

com

para

tive

fitin

dex;

TL

I=

Tuc

ker-

Lew

isin

dex;

RM

SEA

=ro

otm

ean

squa

reer

ror

ofap

prox

imat

ion;

SRM

R=

stan

dard

ized

root

mea

nsq

uare

resi

dual

;Q

MI

=Q

ualit

yof

Mar

riag

eIn

dex;

IAI

=In

effe

ctiv

eA

rgui

ngIn

vent

ory.

a Inc

ludi

ngch

ildag

e.bFi

rst

lines

refe

rto

the

mod

elfit

ofm

easu

rem

ent

mod

el,

seco

ndlin

esre

fer

toth

efit

ofst

ruct

ural

mod

els.

†p

<.1

0.∗ p

≤.0

5.∗∗

p≤

.01.


whether such a distinction was warranted. Fol-lowing a procedure used in a study by Yuceland Gassanov (2010), we compared models inwhich each maternal and paternal constructsare modeled separately to produce a modelin which both constructs are combined intoa single latent construct. The χ2 differencetests indicated that combining the parent-reported determinants decreased the fit sig-nificantly, with χ2(1) = 177.61, p < .001 forchildrearing stress, χ2(1) = 246.02, p < .001for role restriction, χ2(1) = 650.1, p < .001for marital quality, χ2(1) = 222.50, p < .001for partner support, and χ2(1) = 239.12, p <

.001 for ineffective arguing. Furthermore, χ2

difference tests indicated that combining thechild-reported outcome constructs decreasedthe fit significantly, with χ2(1) = 126.99, p <

.001 for responsiveness and χ2(1) = 52.14,p < .001 for demandingness. As such, alllatent constructs were modeled separately inour analyses.

Structural models: APIMs

To test our research questions and hypothe-ses, we conducted several APIM structuralequation models involving (a) the relationsbetween each of the five latent predictorsand parents’ demandingness and (b) the rela-tions between each of the five latent predic-tors and parents’ responsiveness. We choseto use single latent predictors of both moth-ers and fathers in our models (e.g., mothers’and fathers’ childrearing stress) rather thanmultiple latent predictors (e.g., mothers’ andfathers’ childrearing stress and role restric-tion), as the latter would have generated mod-els with too many manifest variables in rela-tion to the sample size of this study.

For each relation, analyses were con-ducted in three steps. First, we investigatedthe possibility of significant actor and part-ner effects of the independent variables onparenting style. The models included twolatent predictors and two latent outcome vari-ables (e.g., the relation between mothers’childrearing stress and mothers’ responsive-ness and the relation between fathers’ chil-drearing stress and fathers’ responsiveness),thus allowing us to test for the actor and

partner effects of both mothers and fatherssimultaneously.

Second, we tested whether the actor orpartner effects differed significantly betweenfathers and mothers by specifying equalityconstraints (i.e., nested models). Because con-straining one path to be equal to anotherpath yields a gain of one degree of freedom,a statistically significant change in the chi-square value as compared with the model withno equality constraints indicates that actor orpartner effects are statistically different fromeach other and stronger for one parent. A non-significant change in the chi-square value ascompared with the model with no equalityconstraints indicates no differences betweenthe two parents.

Finally, we examined whether the modelsare more actor, partner, couple, or contrast ori-ented. To do so, we followed the proceduredescribed by Kenny and Ledermann (2010):We calculated the ratio of the partner effectto the actor effect, also called the k param-eter. If k is near zero, the actor-only patternis indicated; if k is near 1, we have a cou-ple pattern; and the contrast pattern occursif k is −1. However, if the partner effectswere much stronger than the actor effects,we defined k as the ratio of actor to part-ner effects, whereby a k near zero indicates apartner-only effect (see also Kenny and Led-ermann, 2010, p. 364). We then computed thebootstrap confidence interval (CI) for k. TheseCIs provide direct information on whether aspecific pattern takes place. Defining k as thepartner–actor ratio, the actor-only pattern isverified when 0 but not 1 and −1 is in theCI, the couple pattern is supported when 1but not 0 is in the interval, and the contrastpattern is verified when −1 but not 0 is inthe interval. Defining k as the actor–partnerratio, the partner-only pattern is verified when0 but not 1 and −1 is in the CI. Next, allks that support a specific pattern are fixedto 0 (actor-only or partner-only pattern), −1(contrast pattern), or 1 (couple pattern). Then,we reestimate this simpler model and com-pare it with the more general model implyingno specific pattern. A nonsignificant changein the chi-square value indicates that themore parsimonious model is consistent with


the data. We refer to Kenny and Ledermann(2010) and Ledermann, Macho, and Kenny(2011) for a more detailed description of theprocedure.

Demandingness: Testing the APIM

We evaluated five SEM APIMs in whichstress or support on the part of mothersand fathers predicted demandingness on thepart of mothers and fathers. As shown inTable 1, all models had an adequate to goodfit. The models revealed a significant actoreffect between fathers’ childrearing stress andfathers’ demandingness (β = .22, p < .05),indicating that higher levels of childrear-ing stress on the part of fathers results inmore demandingness. Constraining the actoreffects to be equal revealed no gender dif-ferences, χ2(1) = .37, ns, indicating that theactor effects of childrearing stress on demand-ingness are similar for mothers and fathers(b = .05, p < .05). The k, defined as the part-ner–actor ratio, for the effects of childrear-ing stress was .06. The 95% CI ranged from−.001 to .131, which supports the actor-onlypattern as zero is included in the CI. Afterplacing constraints on k, the model compar-ison test supported the more parsimoniousmodel, which indicates that the model can bedescribed as an actor-only model.

Furthermore, the APIM tests revealed asignificant actor effect of mothers’ qualityof marriage (QMI) on mothers’ demanding-ness (β = .23, p < .05). To test for genderdifferences, we constrained the actor pathsto be equal. The χ2 difference test revealedno significant differences, χ2(1) = .93, ns,with b = .04, ns. In other words, the actoreffects of QMI on demandingness that werefound are similar for mothers and fathers,although the constrained model revealed thatthese spillover effects are not significant.

Responsiveness: Testing the APIM

Table 2 shows five SEM APIM in which thestress or support levels of mothers and fatherspredict the responsiveness of mothers andfathers. The fits of all models were adequateto good. The models showed a significant

actor effect between fathers’ role restrictionand responsiveness (β = −.21, p < .05). Thepartner effect of mothers’ role restriction onfathers’ responsiveness was significant at thetrend level (β = .16, p = .07). Constrainingthe actor effects to be equal revealed no gen-der differences, χ2(1) = 1.05, ns, indicatingthat the actor effects of role restriction onresponsiveness are similar for mothers andfathers (b = −.05, p < .01). Constraining thepartner effects to be equal revealed no genderdifferences, χ2(1) = 1.99, ns, and it resultedin the absence of the above-mentioned trendeffect (b = .05, ns). The k, defined as the part-ner–actor ratio, for the effects of role restric-tion was −.059. The 95% CI ranged from−.130 to −.007. Albeit the near-zero value ofk indicated an actor-oriented model, the actor-only pattern was not supported (i.e., zero isnot included in the CI).

With regard to marital relationship, signifi-cant partner effects were found between theQMI of mothers (β = .34, p < .01), moth-ers’ perception of partner support (β = .22,p < .05), ineffective arguing on the part ofmothers (β = −.20, p < .05), and responsive-ness on the part of fathers. Further tests, how-ever, revealed no gender differences in thesepartner paths, with χ2(1) = 2.60, ns, b = .08,p < .01 for the QMI; χ2(1) = 1.13, ns, b =.11, p < .05 for partner support; and χ2(1) =2.35, ns; b = −.07, ns for ineffective argu-ing. The k, defined as the actor–partner ratio,was .086 for QMI and .116 for partner sup-port. The 95% CIs ranged from −.007 to.221 for QMI and from −.041 and .349 forpartner support, which support partner-onlypatterns as zero is included in the CIs. Afterplacing constraints on the ks, model compar-ison tests supported the more parsimoniousmodels, which indicates that both models canbe described as partner-only models.

Discussion

Over the past decades, the parenting literaturehas contained frequent calls for the systematicinclusion of both mothers and fathers. Inthis study, we investigated how two sourcesof stress and support experienced by fathersand mothers influence their own parenting


styles, as well as the parenting styles of theirpartners. Focusing on the APIM, the aim wasto assess actor and partner effects of parentingstress, positive and negative aspects of maritalrelationship on parenting styles, to examinepossible gender differences in the pathwaysand to assess whether the models are moreactor, partner, contrast, or couple oriented.Although the results provide evidence for bothactor- and partner-oriented patterns, no genderdifferences were found. The interpretationof the data is complicated, however, as theeffects seem to depend on both type ofdeterminants and parenting style.

More specifically, we found significantactor effects of parenting stress on parentingstyles for both mothers and fathers, whereasno actor effects were found between posi-tive and negative aspects of marital relation-ship and parenting styles. The latter find-ing is inconsistent with other studies, whichhave reported modest but consistent inter-relatedness between global marital qualityand parenting (Erel & Burman, 1995), aswell as between marital conflict and parent-ing (Krishnakumar & Buehler, 2000). Con-versely, we found partner effects betweenpositive aspects of marital relationship andresponsive parenting style for both fathersand mothers, whereas no partner effects werefound between parenting stress and parentingstyles. The results thus suggest that the pat-terns for the effects of parenting stress ondemanding and responsive parenting stylesare actor oriented, whereas the patterns for theeffects of positive aspects of marital relation-ship on responsive parenting style are partneroriented, in that an individual’s parenting isinfluenced by the partner’s feelings of maritalquality and support, instead of the individ-ual’s own feelings of marital quality andsupport. One possible explanation is that par-enting stress is more covert for the partneras it may be something individually deter-mined, and therefore leading to actor effectsand not to partner effects. The partner rela-tionship, on the other hand, is shared betweenpartners and is more overt. An alternativeexplanation for the absence of actor effects ofmarital relationship on parenting is that moststudies on the influence of marital relationship

on parenting use a single score or constructto represent the marital relationship, therebyneglecting differences in the feelings of thetwo spouses with regard to the marital rela-tionship. Indeed, the few parenting studiesthat use reports from both fathers and mothersregarding marital relationship and/or maritaldiscord either total or average the scores ofboth spouses (Sturge-Apple, Davies, Boker, &Cummings, 2004) or use the summed scoresas manifest indicators to create a single latentconstruct (Kaczynski et al., 2006). In addi-tion, many parenting studies assess maritalrelationship according to one reporter (Davies,Sturge-Apple, Woitach, & Cummings, 2009)or according to one spouse and an indepen-dent observer (Buehler et al., 2006). As such,these studies are not able to assess truly rela-tional phenomena (Kenny et al., 2006). More-over, the actor effects found in prior studiesthat used summed or average scores mightactually reflect both actor and partner effects.In this study, we view marital relationships aswell as parenting styles as climates in whichbehaviors are expressed. Although parentingstyle is a constellation of attitudes towardthe child (Darling & Steinberg, 1993), mar-ital relationship can be seen as a compilationof feelings and attitudes between two spouses,with each parent experiencing the quality ofhis or her marriage somewhat differently. Itcan be further assumed that men and womenwant to maintain the same climate in theirroles as parents and in their roles as spouses.Although parents might acquiesce in theirown feelings of marital quality and support,this does not implicate that they acquiescein their partner’s feelings of marital qualityand support. It is therefore plausible that anindividual’s parenting style might be moreinfluenced by the partner’s perceptions of themarital quality and support than it is by theindividual’s own perception of marital qual-ity and support. Another possible explanationfor the lack of actor effects between maritalrelationship and parenting styles is that ourmeasurement of marital relationship and par-enting styles differs from previous studies. Inthis study, we were interested in marital qual-ity and partner support as positive measures


of marital relationship and ineffective argu-ing as a negative measure of marital rela-tionship. Although the meta-analytic study byErel and Burman (1995) provided evidencefor linkages between positive aspects of mar-ital relationship and parenting, the greaterpart of the studies on the association betweenmarital relationship and parenting, however,focused on negative aspects of marital rela-tionship, whereby Krishnakumar and Buehler(2000) found that the strongest associationsbetween interparental conflict and parentingoccur when interparental conflict is measuredas a combination of disagreement and overtconflict style or purely as overt conflict style.Furthermore, we focused on demandingnessand responsiveness, two parenting styles thatare not necessarily ineffective, whereas someother scholars focused on harsh punishmentor lax control. In future studies, it might beinteresting to further investigate the relativesize of actor and partner effects, taking intoaccount different dimensions of marital rela-tionship and parenting behaviors.

Interestingly, actor effects were found forthe demandingness and responsiveness of par-enting styles, whereas partner effects werefound only for responsive parenting. Theseresults are somewhat similar to those reportedin a study by Nelson and colleagues (2009),albeit in a different context. Nelson and col-leagues investigated spillover and crossovereffects between family stress and the sup-portive and nonsupportive techniques thatparents use to teach children about emo-tions. The authors found spillover effectsfor both supportive and nonsupportive par-ent responses to children’s negative emo-tions, whereas crossover effects were foundonly with supportive responses. Althoughdemandingness might be somewhat similarto nonsupportive parenting discussed in thestudy by Nelson and colleagues, responsive-ness is more similar to their notion of support-ive parenting. As suggested by Nelson andcolleagues, one possible explanation for thefindings is that “spillover, which results fromstress in other contexts of one’s own life, ismore personal and therefore, has the possibil-ity to be more negative than stress stemming

from one’s partner” (p. 677). We concur withthis possibility.

We also investigated whether the strengthof the pathways differs between mothers andfathers. Previous studies have suggested that,for men, the role of father and husband maybe less distinct than are the roles of motherand wife for women. It is therefore possi-ble that fathering may be more likely deter-mined by the state of the marital relation-ship (Coiro & Emery, 1998; Parke, 2002).This hypothesis has been endorsed mostly bystudies that do not directly test the effectsof parent gender (Cummings et al., 2010). Ina review on mothering and fathering, how-ever, Coiro and Emery (1998) suggested thatthe pattern of supportive evidence for thefathering-vulnerability hypothesis may not beas pervasive as it may seem at first glance, asfew analyses of mother–father differences arebased on within-couple comparisons and, assuch, are not able to grasp the interrelationaluniqueness of the dyads. Indeed, our resultsclearly demonstrate that, at first glance, whenwe did not constraint the pathways of thedyads to be equal (i.e., in the unrestricted anal-yses), fathering appears to be more affectedthan mothering. When testing for gender dif-ferences, however, the strength of the path-ways between stress and parenting appearedequally strong for mothers and fathers. Assuch, our findings do not lend support to thefathering-vulnerability hypothesis.

This study contributes to the current bodyof research using rigorous methods toexamine the associations of parenting stress,positive and negative aspects of marital rela-tionship with the parenting styles of mothersand fathers. Among others features, our studyuses latent constructs for parenting stress,marital relationship, and parenting styles,adjusted for measurement error. Another ad-vantage of this research involves the use ofparent reports of parenting stress and maritalrelationship, combined with child reports ofparenting styles, thus avoiding the problem ofcommon-method variance. Furthermore, theframework of the APIM enabled us to exploregender differences and to test whether themodels are more actor, partner, contrast, orcouple oriented. Nonetheless, it is important


to note a number of limitations that atten-uate the clarity of the current results. First,given the small sample size and the use oflatent constructs, it was impossible to inves-tigate different measures of marital relation-ship and parenting stress simultaneously asdeterminants of parenting styles. Given theactor effects of parenting stress on parent-ing styles and the partner effects of maritalrelationship on parenting styles, future studies(with larger sample sizes) should consider themutual influence of various sources of stresson mothering and fathering. Second, althoughit was not the focus of this study, the influ-ence of several background variables shouldbe examined in a more sophisticated way. Forexample, in this study, the gender of the childwas not significantly related to parenting ofeither the mother or the father, despite thesuggestions of some authors that child gendermoderates the relation between interparentaldiscord and changes in parenting practices(Cummings et al., 2010; Sturge-Apple et al.,2004). It might be interesting for future stud-ies to conduct more thorough investigationof the influence of the gender of the child,using multiple-group SEM with larger sam-ple sizes. Finally, the cross-sectional nature ofthe data makes causality difficult to establish.Although associations between stress and par-enting styles can be examined, the time order-ing among the variables is not clear. Despitethe fact that results from the few availablelongitudinal studies indicate that marital dis-cord and parenting stress seem to proceedparenting (Abidin, 1990; Floyd, Gilliom, &Costigan, 1998), corroboration of our findingsproduced by longitudinal data would lendcredibility to the findings.

Despite its limitations, this study con-tributes to the literature by demonstratingactor effects of parenting stress on demand-ing and responsive parenting styles, as wellas partner effects of positive aspects of mari-tal relationship on responsive parenting style.This study further shows that the strength ofthese pathways is similar for both mothers andfathers, and thus do not support the hypothesisthat the parenting of fathers is more vulnera-ble than the parenting of mothers is to stressderived from the marital relationship.

References

Abidin, R. (1990). Introduction to the special issue:The stresses of parenting. Journal of Clinical ChildPsychology, 19, 298–301.

Amato, P. R., Meyers, C. E., & Emery, R. E. (2009).Changes in nonresident father–child contact from1976 to 2002. Family Relations, 58, 41–53.

Aunola, K., Stattin, H., & Nurmi, J. E. (2000). Parentingstyles and adolescents’ achievement strategies. Jour-nal of Adolescence, 23, 205–222.

Barber, B. K. (1996). Parental psychological control:Revisiting a neglected construct. Child Development,67, 3296–3319.

Barnett, M., Deng, M., Mills-Koonce, W., Willoughby,M., & Cox, M. (2008). Interdependence of parentingin mothers and fathers of infants. Journal of FamilyPsychology, 22, 561–573.

Baumrind, D. (1991). The influence of parenting style onadolescent competence and substance use. Journal ofEarly Adolescence, 11, 56–95.

Belsky, J. (1984). The determinants of parenting. Aprocess model. Child Development, 55, 83–96.

Belsky, J., Gilstrap, B., & Rovine, M. (1984). The Penn-sylvania Infant and Family Development Project: Sta-bility and change in mother–infant and fahter–infantinteraction in a a family setting at one, 3 and 9months. Child Development, 55, 692–705.

Belsky, J., Woodworth, S., & Crnic, K. (1996). Troublein the second year: Three questions about familyinteraction. Child Development, 67, 556–578.

Belsky, J., Youngblade, L., Rovine, M., & Volling, B.(1991). Patterns of marital change and parent–childinteraction. Journal of Marriage and the Family, 53,487–498.

Benson, M. J., Buehler, C., & Gerard, J. M. (2008). Inter-parental hostility and early adolescent problem behav-ior—Spillover via maternal acceptance, harshness,inconsistency, and intrusiveness. Journal of EarlyAdolescence, 28, 428–454.

Bentler, P. M. (1990). Comparative fit indexes in struc-tural models. Psychological Bulletin, 107, 238–246.

Bronfenbrenner, U. (1989). Ecological systems theory.Annual Child Development, 6, 187–249.

Bronte-Tinkew, J., Horowitz, A., & Carrano, J. (2010).Aggravation and stress in parenting: Association withcoparenting and fahter engagement among residentfathers. Journal of Family Issues, 31, 525–555.

Brown, T. (2006). Confirmatory factor analysis forapplied research. New York, NY: Guilford.

Buehler, C., Benson, M. J., & Gerard, J. M. (2006).Interparental hostility and early adolescent problembehavior: The mediating role of specific aspects ofparenting. Journal of Research on Adolescence, 16,265–291.

Buehler, C., & Gerard, J. (2002). Marital conflict, ineffec-tive parenting, and children’s and adolescent’s mal-adjustment. Journal of Marriage and the Family, 64,78–92.

Byrne, B. (2001). Structural equation modeling withAmos: Basic concepts, applications and prgramming.Mahwah, NJ: Erlbaum.


Coiro, M. J., & Emery, R. E. (1998). Do marriageproblems affect fathering more than mothering? Aquantitative and qualitative review. Clinical Child andFamily Psychology Review, 1, 23–40.

Cooper, C., McLanahan, S., Meadows, S., & Brooks-Gunn, J. (2009). Family structure transitions andmaternal parenting stress. Journal of Marriage andthe Family, 71, 558–574.

Cox, M. J., & Paley, B. (1997). Families as systems.Annual Review of Psychology, 48, 243–267.

Crnic, K., Arbona, A. P. Y., Baker, B., & Blacher,J. (2009). Mothers and fathers together: Contrastsin parenting across preschool to early school agein children with developmental delays. Interna-tional Review of Research in Mental Retardation, 37,3–30.

Crnic, K., Gaze, C., & Hoffman, C. (2005). Cumulativeparenting stress across the preschool period: Relationsto maternal parenting and child behaviour at age 5.Infant and Child Development, 14, 117–132.

Crnic, K., & Greenberg, M. (1990). Minor parentingstresses with young children. Child Development, 61,1628–1637.

Crnic, K., Greenberg, M., Ragozin, A., Robinson, N., &Basham, R. (1983). The effects of stress and socialsupport on mothers of premature and full-term infants.Child Development, 45, 209–217.

Crnic, K., & Low, C. (2002). Everyday stresses andparenting. In M. Bornstein (Ed.), Handbook of par-enting: Vol. 5. Practical issues in parenting (2nd ed.,pp. 243–267). Mahwah, NJ: Erlbaum.

Cummings, E. M., Goeke-Morey, M. C., & Raymond, J.(2004). Fathers in family context: Effects of maritalquality and marital conflict. In M. Lamb (Ed.), Therole of the father in child development (4th ed.,pp. 196–221). New York, NY: Wiley.

Cummings, E. M., Merrilees, C. E., & George, M. W.(2010). Fathers, marriages and families: Revisitingand updating the framework for fathering in familycontext. In M. Lamb (Ed.), The role of the father inchild development (5th ed., pp. 154–176). Hoboken,NJ: Wiley.

Cummings, E. M., & O’Reilly, A. W. (1997). Fathersin family context: Effects of marital quality on childadjustment. In M. Lamb (Ed.), The role of the fahterin child development (3rd ed., pp. 49–65). New York,NY: Wiley.

Darling, N., Cumsille, P., & Pena-Alampay, L. (2005).Rules, legitimacy of parental authority, and adolescentautonomy in Childe, the Phillipines and the UnitedStates. New Directions in Child and Adolescent Devel-opment, 108, 47–60.

Darling, N., & Steinberg, L. (1993). Parenting style ascontext. An integrative model. Psychological Bulletin,113, 487–496.

Darling, N., & Toyokawa, T. (1997). Construction andvalidation of the Parenting Style Inventory II. Unpub-lished manuscript, The Pensylvania State University.

Davies, P. T., Sturge-Apple, M., Woitach, M., &Cummings, E. M. (2009). A process analysis of thetransmission of distress from interparental conflict to

parenting: Adult relationship security as an explana-tory mechanism. Development Psychology, 45,1761–1773.

Deater-Deckard, K. (1998). Parenting stress and childadjustment: Some old hypotheses and new questions.Clinical Psychologist, 5, 314–332.

Deater-Deckard, K., Dodge, K. A., Bates, J. E., & Pettit,G. S. (1998). Multiple risk factors in the developmentof externalizing behavior problems: Group and indi-vidual differences. Development and Psychopathol-ogy, 10, 469–493.

Deater-Deckard, K., & Scarr, S. (1996). Parenting stressamong dual-earner mothers and fathers: Are theregender differences? Journal of Family Psychology, 10,45–59.

Deater-Deckard, K., Smith, J., Ivy, L., & Petrill, S. A.(2005). Differential perceptions of and feelings about-sibling children: Implications for research on par-enting stress. Infant and Child Development, 14,211–225.

Dykstra, P. A., van Tilburg, T. G., & Gierveld, J. D.(2005). Changes in older adult loneliness—Resultsfrom a seven-year longitudinal study. Research onAging, 27, 725–747.

Erel, O., & Burman, B. (1995). Interrelatedness of maritalrelations and parent–child relations. A meta-analyticreview. Psychological Bulletin, 118, 108–132.

Fagan, J., Bernd, E., & Whiteman, V. (2007). Adolescentfathers’ parenting stress, social support, and involve-ment with infants. Journal of Research on Adoles-cents, 17, 1–22.

Fincham, F. D., & Beach, S. R. (2010). Marriage inthe new millenium: A decade in review. Journal ofMarriage and the Family, 72, 630–649.

Floyd, F., Gilliom, L., & Costigan, C. (1998). Marriageand the parenting alliance: Longitudinal prediction ofchange in parenting perceptions and behaviors. ChildDevelopment, 69, 1461–1479.

Frias-Armenta, M., & McCloskey, L. A. (1998). Deter-minants of harsh parenting in Mexico. Journal ofAbnormal Child Psychology, 26, 129–139.

Grych, J. H. (2002). Marital relationships and parenting.In M. H. Bornstein (Ed.), Handbook of parenting(Vol. 4, pp. 231–252). Mahwah, NJ: Erlbaum.

Hakvoort, E., Bos, H., Van Balen, F., & Hermanns, J.(2011). Spillover between mothers’ postdivorce rela-tionships: The mediating role of parenting stress. Per-sonal Relationships. doi: 10.1111/j.1475-6811.2011.01351.x

Hu, L., & Bentler, P. (1999). Cut off criteria for fitindexes in covariance structure analysis: Conventionalcriteria versus new alternatives. Structural EquationModeling, 6, 1–55.

Hutch-Bocks, A., & Hughes, H. (2008). Parenting stress,parenting behavior, and children’s adjustment in fam-ilies experiencing intimate partner violence. Journalof Family Violence, 23, 243–251.

Isacco, A., Garfield, C., & Rogers, T. (2010). Correlatesof coparental support among married and nonmarriedfathers. Psychology of Men and Masculinity, 11,262–278.


Kaczynski, K. J., Lindahl, K. M., Mailik, N. M., &Laurenceau, J. P. (2006). Marital conflict, maternaland paternal parenting, and child adjustment: A testof mediation and moderation. Journal of FamilyPsychology, 20, 199–208.

Kashy, D. A., & Kenny, D. A. (2000). The analysisof data from dyads and groups. In H. T. Reis &C. M. Judds (Eds.), Handbook of research methodsin social psychology. (pp. 451–477). New York, NY:Cambridge University Press.

Kenny, D., Kashy, D., & Cook, W. (2006). Dyadic dataanalysis. New York, NY: Guilford.

Kenny, D. A., & Ledermann, T. (2010). Detecting, mea-suring, and testing dyadic patterns in the actor–partnerinterdependence model. Journal of Family Psychol-ogy, 24, 359–366.

Kjobli, J., & Hagen, K. A. (2009). A mediation model ofInterparental collaboration, parenting practices, andchild externalizing behavior in a clinical sample.Family Relations, 58, 275–288.

Kline, R. (2005). Principles and practices of structuralequation modeling (2nd ed.). New York, NY: Guil-ford.

Krishnakumar, A., & Buehler, C. (2000). Interparentalconflict and parenting behaviors: A meta-analyticreview. Family Relations, 49, 25–44.

Kurdek, L. A. (1994). Conflict-resolution styles in gay,lesbian, heterosexual nonparent and heterosexual par-ent couples. Journal of Marriage and the Family, 56,705–722.

Lamb, M. (2010). How do father influence children’sdevelopment? Let me count the ways. In M. Lamb(Ed.), The role of the father in child development (5thed., pp. 1–26). Hoboken, NJ: Wiley.

Ledermann, T., & Macho, S. (2009). Mediation in dyadicdata at the level of the dyads: A structural equationmodeling approach. Journal of Family Psychology,23, 661–670.

Ledermann, T., Macho, S., & Kenny, D. A. (2011).Assessing mediation in dyadic data using the actor–partner interdependence model. Structural EquationModeling, 18, 595–612.

Litovsky, V., & Dusek, J. (1985). Perceptions of childrearing and self-concept development during the earlyadolescent years. Journal of Youth and Adolescence,14, 373–387.

Minuchin, S. (1974). Families and family therapy. Cam-bridge, MA: Harvard University Press.

Muthen, L., & Muthen, B. (2008). Mplus user’s guide.Version 5.2. Los Angeles, CA: Author.

Nelson, J. A., O’Brien, M., Blankson, A. N., Calkins,S. D., & Keane, S. P. (2009). Family stress andparental responses to children’s negative emotions:Tests of the spillover, crossover, and compensatoryhypotheses. Journal of Family Psychology, 23,671–679.

Norton, R. (1983). Measuring marital quality: A criticallook at the dependent variable. Journal of Marriageand the Family, 45, 141–151.

Parke, R. (2002). Fathers and families. In M. H. Bornstein(Ed.), Handbook of parenting: Vol. 3. Being andbecoming a parent (2nd ed., pp. 27–74). Mahwah,NJ: Erlbaum.

Pleck, J. (2010). Paternal involvement: Revised conceptu-alization and theoretical linkages with child outcomes.In M. Lamb (Ed.), The role of the father in child devel-opment (5th ed., pp. 58–93). Hoboken, NJ: Wiley.

Purdie, N., Carroll, A., & Roche, L. (2004). Parenting andadolescent self-regulation. Journal of Adolescence,27, 663–676.

Raykov, T. (2009). Evaluation of scale reliability forunidimensional measures using latent variable model-ing. Measurement and Evaluation in Counseling andDevelopment, 42, 223–232.

Raykov, T., & Marcoulides, G. A. (2006). On mul-tilevel model reliability estimation from the per-spective of structural equation modeling. StructuralEquation Modeling—A Multidisciplinary Journal, 13,130–141.

Repetti, R. L. (1987). Individual and common compo-nents if the social-environment at work and psycho-logical well-being. Journal of Personality and SocialPsychology, 52, 710–720.

Rodgers, A. (1998). Multiple sources of stress and par-enting behavior. Children and Youth Services Review,20, 525–546.

Simons, R. L., Whitbeck, L. B., Conger, R. D., & Melby,J. N. (1990). Husband and wife differences in deter-minants of parenting: A social-learning and exchangemodel of parental behavior. Journal of Marriage andthe Family, 52, 375–392.

Steiger, J. H. (1990). Structural model evaluation andmodification: An interval estimation approach. Multi-variate Behavioral Research, 25, 173–180.

Sturge-Apple, M., Davies, P. T., Boker, S., & Cum-mings, E. M. (2004). Interparental discord and par-enting: Testing the moderating roles of child andparent gender. Parenting—Science and Practice, 4,361–380

Tucker, L. R., & Lewis, C. (1973). Reliability coefficientsfor maximum likelihood factor-analysis. Psychome-trika, 38(1), 1–10.

Van den Troost, A. (2005). Marriage in motion. Leuven:University Press.

Wentzel, K. R. (1994). Relations of social goals pursuit tosocial acceptance, classroom behavior and perceivedsocial support. Journal of Educational Psychology,86, 173–182.

Woodworth, S., Belsky, J., & Crnic, K. (1996). Thedeterminants of fathering during the child’s secondand third years of life: A developmental analysis.Journal of Marriage and the Family, 58, 679–692.

Yucel, D., & Gassanov, M. A. (2010). Exploring actorand partner correlates of sexual satisfaction amongmarried couples. Social Science Research, 39,725–738.

Documents

Parenting stress and marital relationship as determinants of mothers' and fathers' parenting