|Home | About | Journals | Submit | Contact Us | Français|
Based on 13,694 mother-child dyads from the Early Childhood Longitudinal Kindergarten Cohort (ECLS-K), this study examined the bidirectional relations between parental and child functioning from kindergarten through third grade. Results from the cross-lagged models demonstrated that child-driven effects co-occurred with parental effects and these effects were comparable in size. At the same time, however, results from the latent profile analysis revealed idiosyncratic patterns of parent and child functioning. Compared with children in the least optimal functioning profiles, those in the average and above average profiles elicited greater improvements in parents’ functioning over time. Although children characterized by poor academic performance at kindergarten appeared to precede parents characterized by harsh parenting at third grade, there was a threshold in the evolving strength of the overall child-driven effects. Taken together, the results from this study underscore the importance of considering reciprocal processes in the parent-child dynamic while also underscoring individual differences in these processes across the early to middle childhood years.
Bidirectionality is a fundamental property of dyadic relationships and this holds true for the relationships between parents and their young children (Bell, 1968; Lerner, 2006). Thanks to the development of advanced longitudinal modeling techniques, there has been an accumulation of evidence on the existence of bidirectional effects between parents and children; that parenting is not just directed at children, it is also elicited from them (Ansari & Crosnoe, 2015b; Crosnoe, Augustine, Huston, 2012; Lugo-Gil, & Tamis-LeMonda, 2008; Yan & Dix, 2014). Despite the theoretical and empirical evidence, there remains doubt regarding the role of child effects in the fields of family and developmental psychology and, if they exist, whether these effects are comparable in size when compared with the effects of parents on their children.
Just as parent-child relationships are dynamic, so too is child development (Lerner, 2006); yet few studies have explored the link between various typologies of parenting behaviors in conjunction with various types of child behaviors. Theoretically, however, both homogeneity and heterogeneity would be observed in these relationships, especially in the study of child-driven effects where individual differences in the strength and types of parenting behaviors are observed. In other words, parents may react to different types of child adjustment behaviors with a different set of parenting practices. To unravel this question on the specificity of child effects, sub-patterns of child behaviors across domains need to be examined.
The current study aims to address these gaps in knowledge by: (a) examining the bidirectional effects between child adjustment and various parenting practices across the early- to middle-childhood years; (b) examining the strength of child-driven effects in the presence of parental effects; and (c) presenting a more nuanced analysis of the strength of these child-driven effects among children and parents with different typologies of adjustment and functioning.
Motivated by Bell’s (1968) reinterpretation of the directionality in parent-child relationships, many theorists in the decades since have endeavored to elucidate the phenomena of the reciprocal relations between children and their parents (Belsky, 1984; Patternson, 1982; Sameroff, 1975). As predicted by the seminal analyses of these pioneers, child development has been increasingly viewed as part of a dynamic and transactional system that involves complex interplay between children and their parents. Child-driven effects, which represent one component in the transactional system, may occur in moment-to-moment parent-child interactions (Del Vecchio, & Rhoades, 2010) and can potentially explain a wide variety of individual differences among children and parents (Elgar, McGrath, Waschbusch, Stewart, & Curtis, 2004; Van Bakel, & Riksen-Walraven, 2002); thus, child effects should be considered just as important as the frequently assumed impacts of parents on their children (Pardini, 2008).
Despite the growing body of theoretical and empirical evidence on child-driven effects, these effects have remained an area of much needed research because of the implicit assumption that the impacts of parents on children are by no means rivaled in importance by the effects of children on parents (Lytton 1990; Pardini, 2008). These biases have been reflected in the extant literature. For example, much of the empirical inquiry into child effects has been conducted using samples of children with particular characteristics that may render these effects more salient: children with special needs or children with severe behavioral problems (Burke, Pardini, & Loeber, 2008; Harrison, & Sofronoff, 2002; Pettit & Arsiwalla, 2008), children from high risk families (Yates, Obradović, & Egeland, 2010), boys (Fite, Colder, Lochman, & Wells, 2006; Keijsers, Loeber, Branje, & Meeus, 2011; Pardini, Fite, & Burke, 2008), or adolescents (Coley, Votruba-Drzal, & Schindler, 2008; Gault-Sherman, 2012; Willoughby, & Hamza, 2011). These findings are undoubtedly important in expanding our understanding of child effects; however, these studies are not able to mitigate the existing doubt regarding whether: (a) normally developing children would elicit differential parenting responses as, for example, children with special needs do; and (b) whether these child-driven effects, if they exist, would be strong enough not to be ignored, similar to the effects of parents on their children. In this study, we investigate these child-driven effects between kindergarten and third grade, which is a critical period for shaping children’s short- and long-term school success and well-being (Duncan et al., 2007; Jones, Greenberg, Crowley, 2015), among a nationally representative sample to determine whether the effects of children on their parents are similar to the largely assumed effects of parents on their children.
Child adjustment is multi-faceted (Lerner, 2006), yet most theorists have conceptualized the reciprocal influences between children and parents through the lens of children’s aversive behaviors and temperamental-based traits (Kiff, Lengua, & Zalewski, 2011; Patterson, 1982). For example, in Bell and Harper’s (1977) seminal analysis of bidirectional effects, it was suggested that when children’s actions exceeded parental standards in terms of frequency or intensity (e.g., aggression, hyperactivity), then they elicited new parental responses (e.g., physical punishment). Similarly, in Patterson’s coercive model of parenting (1982), children’s aversive and coercive noncompliance was conceptualized as the trigger that was reinforced by a loss of control by parents, resulting in a cycle of negative parent-child interactions. Thus, this integrative model of parenting suggests that children’s externalizing behavior problems render them more or less difficult to be cared for, and these individual differences shape the quantity and quality of parenting that they receive (Belsky, 1984).
Under these theoretical configurations, most scientists have studied children’s disruptive behaviors and how they, in turn, affect different facets of parent functioning. For example, children’s temperament-based attention span, attentional control, and irritability have been found to predict less sensitive and consistent parenting (Belsky, Pasco Fearon, Bell, 2007; Lengua, & Kovacs, 2005; Gross, Shaw, Moilanen, Dishion, & Wilson, 2008) and postnatal depression (Sugawara, Kitamura, Toda, & Shima, 1999); children’s oppositional behaviors and early externalizing problems have been associated with an increase in mothers’ depressive symptoms (Choe, Shaw, Brennan, Dishion, & Wilson, 2014) and parents’ psychological control during adolescents (Pettit, Laird, Dodge, Bates, & Criss, 2001); and the DRD2 Aþ1 polymorphism, associated with low vagal withdrawal and high aggressive behaviors, has been found to affect child behavior by exerting an evocative influence on parenting behavior (Mills-Koonce, Propper, Gariepy, Blair, Garrett-Peters, & Cox, 2007).
Child-driven effects, however, are unlikely to be restricted to the domain of children’s disruptive and aversive behaviors. Child inhibition, for instance, has been suggested to influence mothers’ perception of their children as vulnerable (Shamir-Essakow, Ungerer, Rapee, & Safier, 2004). These perceptions may, in turn, facilitate differential parental reactivity, such as intrusiveness and over-protection (Rubin, Coplan, & Bowker, 2009). As another example, attentive children who respond positively to their mother’s bids for interaction tend to elicit more sensitive care giving, warmth, and attention when compared with children who are highly reactive (Harach, & Kuczynski, 2005; Mills-Koonce, et al., 2007; Shamir-Essakow, et al., 2004). Moreover, children’s cognitive functioning has also been found to predict higher quality and sensitive parenting (Ansari & Crosnoe, 2015a; Lugo-Gil, & Tamis-LeMonda, 2008). The current study not only examines children’s functioning across multiple domains to represent parents’ overall perception of their children abilities, but also on unique patterns of child adjustment.
To date, much of the empirical inquiry into child effects has been conducted in the variable-oriented approach: linking specific child behaviors to parenting practices at the mean level (e.g., Ansari & Crosnoe, 2015a; Gershoff, Aber, Clements, 2009; Lugo-Gil & Tamis-LeMonda, 2008). This approach is informative and builds on developmental systems theory (Lerner, 2006)—pointing to the direction and strength of certain associations between children and their parents. Child development and parenting, however, are dynamic; children’s behaviors do not occur in isolation (Nagin & Tremblay, 1999). What is missing from the variable-centered approach, therefore, is how the configural patterns of children’s adjustment that naturally occur may correspond to the configural patterns of parenting practices that also naturally occur.
This issue can be addressed by using person-centered methodologies (Hart, Atkins, Fegley, Robins, & Tracy, 2003). On one hand, child driven effects may vary based on different typologies of children's skills. Typologies of child behavior problems have been investigated before (Edelbrock, & Achenbach, 1980) and offer a fine-grained differentiation among children in addition to the global internalizing versus externalizing dichotomy. These typologies, however, fail to consider the role of children’s academic abilities, and how these typologies condition the child-driven effects are also unclear. For example, children with positive behavior coupled with high academic performance may elicit more benign parenting practices compared with children both high in externalizing behavior and high in academic achievement (Ansari & Crosnoe, 2015b). On the other hand, child behaviors may not only affect specific parenting functioning, but, rather, they may elicit particular patterns of parenting practices that may co-occur in natural contexts. Indeed, it has been shown that certain parenting practices tend to co-occur (Ansari & Crosnoe, 2015b); however, questions such as whether certain child behaviors may render parents to be both psychologically stressed and reliant on harsh parenting or psychologically stressed but reliant on positive parenting are less clear. For these reasons, variable-oriented approaches need to be complemented with person-oriented techniques; they are complementary, but not competing. Thus, this study employed two approaches–cross-lagged models and profile analysis–to provide a comprehensive understanding of child driven effects.
To address the limitations in the existing literature, the current study used a large and nationally representative sample of children followed from kindergarten through third grade to test four hypotheses. First, bidirectional effects between child adjustment and various parenting practices exist across the early-to-middle-childhood years. Second, the strength of child-driven effects is comparable to that of parental effects. Third, different profiles of child functioning differentially elicit improvements in parenting. Fourth, child functioning at kindergarten differentially predicts different configurations of parent functioning in third grade.
Data for the current investigation were drawn from the Early Childhood Longitudinal Kindergarten Cohort (ECLS-K; for sampling information see, Tourangeau et al., 2006), a nationally representative sample of 21,409 kindergarten children (cohort of 1998) who were followed from kindergarten entry through the end of the eight-grade year. Data were collected from parents, teachers, and school administrators as well as direct assessments of children. For the purposes of this study, we restricted our sample to all children and families who participated in data collection during the kindergarten (1998; T1) and third grade (2002; T2) waves, resulting in a final sample of 13,694 children (49% female) and families (57% White, 18% Latino, 13% Black, 6% other). For sample demographics, see Table 1.
Descriptives of the focal parent and child functioning variables are provided in Table 2.
Five aspects of parent functioning were considered as part of a multi-dimensional construct that tapped into parents’ well-being, harshness and discipline, and warmth across the kindergarten and third grade years. First, parents reported on their depressive symptoms using 12 items drawn from the Center for Epidemiological Studies-Depression Scale (CES-D; Radloff, 1977). Items were scored on a 4-point scale (1= never, 4= most of the time), with higher scores reflective of less optimal mental health (average α = .88). Six items were also drawn from the Parenting Stress Index (Abidin, 1983; e.g., “being a parent is harder than I thought it would be”) that were also based on a 4-point scale (1= never, 4= most of the time) and used to create a composite of parenting stress (average α = .62).
At the same time points, parents responded to 11 vignettes about their harshness in dealing with child misbehavior, which we used to create a continuous scale ranging from 0 (e.g., give warning) to 4 (e.g., spank or hit back; see Crosnoe & Cooper, 2010; Gershoff, Aber, Raver, Lennon, 2007). Parents also reported on the frequency with which they spanked their child. Similar to past conventions (Gershoff et al., 2012), we created a scale so that parents who reported that they never spanked their child received a score of 0. Parents who reported that they had spanked their child, but not in the last week, were coded as 1, and those who reported to having spanked their child in the last week received a score of their stated frequency plus 1.
Finally, a composite was created for parental warmth by averaging four items from the Home Observation for Measurement of the Environment scale (HOME; Caldwell & Bradley, 1984), involving parents’ warmth toward their children (e.g., “I express affection by hugging, kissing, and holding child”, “Child and I often have warm, close times together”). The measure demonstrated adequate reliability (average α = .63).
Across both the kindergarten and third grade years, teachers were asked to report on children’s social skills and behavior using the Social Rating Scale (SRS: Gresham & Elliot, 1990), which was based on a 4-point scale (1= never to 4= very often). The SRS tapped into five different domains of child behavior, namely: externalizing behavior (e.g., gets angry, fights; α’s = .89–.90), internalizing behavior (e.g., anxiety, loneliness; α’s = .76–78), interpersonal skills (e.g., ability to make friends, tendency to help others; T1 α’s = .89), self-control (e.g., attentiveness, ability to control temper; α’s = .79–.80), and approaches toward learning (e.g., eagerness to learn; T1 α’s = .89–.91).
Children’s math and reading skills were directly assessed using standardized assessments developed by NCES (α’s = .92–.94; Rock & Pollack, 2002). Content from the reading assessment covered letter recognition, early reading, and phonological awareness, whereas the math assessment covered number sense, properties, and measurement. Due to high correlation between the subscales (r = .67 for T1; r = .73 for T2), and similar to past conventions (Crosnoe, Bonazzo, & Wu, 2015), we created a composite for academic skills.
To reduce the possibility of spurious associations, all models controlled for a theoretically relevant set of child and household variables, namely: child race/ethnicity, child gender, child age, parents’ educational history, parents’ marital status, mothers’ age, mothers’ employment status, household language, household income, region, and urbanicity. All covariates were drawn from the kindergarten year.
The first aim of this study was to determine whether children elicited changes in their parents’ functioning over time. To address this question, we employed cross-lagged structural equation models (SEM) with latent variables using the Mplus program (Muthén & Muthén, 1998–2013). A benefit of cross-lag models is that they allow for the estimation of change in our focal dependent variables; therefore, we examined whether children elicited change in parent functioning and whether parent functioning predicted change in child functioning over time. Model fit was evaluated using the chi-square statistic, the comparative fit index (CFI; recommendation, >.90), and the root-mean square error of approximation (RMSEA; recommendation, < .05; Hu & Bentler, 1999).
The second aim of this study was to identify different combinations of parent and child functioning using latent profile analyses (LPA). To determine the optimal number of profiles, we conducted a series of models that fit one to seven profiles to the data. Model fit was evaluated using: (a) The Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC), and the Adjusted Bayesian Information Criterion (ABIC; lower values are better; Nylund, Asparouhov, & Muthén, 2007); (b) the entropy statistic (ranging from 0–1) with values greater than .70 indicating appropriate classification (Jung &Wickrama, 2008); and (c) a log-likelihood test (Vuong-Lo-Mendell-Rubin; LMRT) that compares k classes to a model with k-1 classes (Nylund et al., 2007). After identifying the appropriate number of classes, we then conducted cross-tabulation analyses to examine the preliminary relations between profiles of child functioning at kindergarten and parental functioning at third grade. Then, we used the profiles to determine whether the “child effects” varied across different typologies of child functioning. Finally, multinomial logistic regressions were used to determine whether children’s functioning at kindergarten was associated with different typologies of parent functioning over time.
All models were conducted using: (a) longitudinal child level weights to ensure that our sample was representative of the nation’s children while also accounting for cross-wave attrition; (b) clustering variables to account for non-independence and the nested nature of the data; and (c) full information maximum likelihood estimation to address missing data.
We first tested a single measurement model for children’s and parents’ functioning across the kindergarten and third grade waves. As can be seen in Table 1, all factor loadings were significant at a minimum probability of .001 and all loadings were comparable across time. The measurement model fit the data well: CFI = .95, RMSEA = .03, and x2 (df = 180) = 2329.47, p < .001. We then proceeded to testing the full structural model which also demonstrated good fit: CFI = .90, RMSEA = .03, and x2 (df = 486) = 5090.35, p < .001.
As can be seen in Figure 1, parents’ functioning at kindergarten was associated with change in children’s functioning over the course of the next three years (β = −.11, p < .001). That is, parents with less optimal functioning at kindergarten had children who demonstrated less optimal functioning in third grade. We also found, however, that children’s functioning was predictive of change in parents’ functioning (β = −.10, p < .001), suggesting that children’s difficulties translated into less optimal parenting over time. The effect size for the association between parents’ impact on children and children’s impact on their parents were not statistically different (confirmed with Wald’s Test). Even though these effect sizes were modest, they reveal that the family system is dynamic and that the effect of children on their parents is equivalent to the effects of parents on their children.
The second aim of this study was to determine whether the observed “child effects” varied according to the different typologies of parent and child functioning. We used LPA to identify the most common configurations of functioning at the national level during the early elementary school years.
According to the statistics in Table 3, the five-class solution for child functioning at kindergarten best fit the data and, in all cases, the class probabilities were above 80%, indicating that children were likely assigned to the correct profile. Thus, we settled on the five-class model, and each of the five profiles was labeled according to the relative scores on the child functioning measures. The top panel of Figure 2 displays the mean z-scores for each of the child functioning profiles at the kindergarten wave. Overall, 20% of children demonstrated below average academic and behavioral skills, which we labeled as poor behavior and academic. Seven percent of children exhibited high levels of behavior problems and below average academic skills; thus, they were labeled very poor behavior and academic. Over a third (37%) of children displayed average levels of child functioning, and were labeled accordingly, whereas another third (35%) of children exhibited above average levels of behavior and academic skills, which we labeled as strong behavior and above average academic. Finally, 2% of children exhibited high levels of functioning across all domains, which we labeled as strong academic but above average behavior.
We conducted the same LPAs for the measures of parent functioning during third grade. As can be seen in Table 3, all fit indices suggested that the six-class solution best fit the data (AIC, BIC, ABIC, entropy, LRT), which we labeled accordingly (see bottom panel of Figure 2). Overall, 2% of parents exhibited above average depressive symptoms but were not harsh, which we labeled as moderate psychological risks. Nine percent of parents demonstrated average levels of well-being, but were not harsh toward their children, which we labeled as low psychological risk and competent parenting. Approximately 3% of parents reported high levels of depression and demonstrated negative behaviors toward their children, and thus, this group was labeled as high psychological risk and problematic parenting. Most parents (49%), however, reported average levels on all the outcomes, which we labeled as average psychological risk and parenting and a third of parents (32%) reported average levels of stress and depression, but exhibited above average harshness, which we labeled average psychological risk and harsh parenting. Finally, 2% of parents reported extremely high levels of depression, but did not report any negative behaviors toward their children, which we labeled as high psychological risks and competent parenting.
After identifying profiles of parents’ and children’s functioning, we conducted cross-tabulation analyses to examine the preliminary relations between profiles of child functioning at kindergarten and parent functioning at third grade. As demonstrated in Table 4, in general, as children’s profiles shift from poor to competent functioning at kindergarten, parents were less likely to be in profiles characterized as harsh, negative, or depressed during the third grade year.
Having established the different profiles of children’s and parents’ functioning, we proceeded to the third and fourth aims of the study: (a) to determine whether different profiles of child functioning differentially elicited improvements in parenting, and (b) to determine whether child functioning at age 5 differentially predicted different configurations of parent functioning in third grade.
As can be seen in Table 5, there were no differences between the two least optimal profiles (i.e., very poor behavior and academic versus poor behavior and academic; see Model 1); however, children in the average (β = −.23, p < .05), strong behavior but above average academic (β = −.36, p < .001), and strong academic but above average behavior (β = −.44, p < .01) profiles elicited improvements in parents’ functioning over time. Importantly, we also identified a significant gradient—the children with more optimal functioning elicited greater improvements in parents’ functioning, but there was a threshold, as no differences emerged between the two most optimal profiles. In sum, despite coming to the same conclusion—children elicit improvements in their parents’ functioning—we find that there is considerable variation in the size of this effect and there appears to be a threshold where these elicitation processes level off, both at the bottom and top end of child functioning distribution.
Next, we examined whether children’s skills predicted the different configurations of parents’ functioning in third grade, net of parents’ functioning in kindergarten (see Table 6). Results from these analyses revealed that children’s functioning was, for the most part, only associated with parents’ likelihood of entering the most optimal profile in third grade (average but not harsh), but did not consistently differentiate between the remaining parenting profiles.
Guided by multiple transactional theories from the fields of developmental and family science (Belsky, 1984; Patternson, 1982; Sameroff & Chandler, 1975), this study tested the bidirectional relations between parental and child functioning from kindergarten through third grade. By conducting cross-lagged models, we found the co-occurrence of bidirectional effects between parents’ and children’s overall functioning across domains. Importantly, the child-driven effects were as large as the parental effects. At the same time, however, results from the profile analyses revealed idiosyncratic patterns of parent and child functioning. Although children characterized by poor academic performance at kindergarten appeared to precede parents characterized by harsh parenting at third grade, there was a threshold in the evolving strength of the overall child-driven effects.
Results from the cross-lagged model supported our hypothesis of the bidirectional effects between parents and their children. Similar to the existing literature on the effects of parenting (e.g., Davis-Kean, 2005; Gershoff et al., 2007), we found that parents who exhibited less optimal functioning at kindergarten had children who demonstrated less optimal functioning at third grade. Just as importantly, however, children’s functioning at kindergarten was also predictive of changes in parents’ functioning over time. In other words, children’s difficulties during early childhood translated into less optimal parenting through the early elementary school years. Unlike prior studies that have examined isolated domains of parent or child functioning (e.g., Ansari & Crosnoe, 2015a; Crosnoe et al., 2012; Gershoff et al., 2009; Yan & Dix, 2014), the current study used indicators of the overall functioning of parents and their children. This methodology is in line with recent studies in the fields of developmental and family psychology that have advocated for a more holistic analysis of the family system (Bornstein, 2005; Lerner, 2006) in part because parenting behaviors in reaction to children’s behaviors are dependent on parents’ judgment of children’s overall functioning, rather than discrete behaviors.
Moreover, the documented effect sizes for the parent- and child-driven effects were not statistically different. Although somewhat surprising, this is consistent with prior research on the family system that have shown that child-driven effects are not any weaker than parent-driven effects (Ansari & Crosnoe, 2015a; Lugo-Gil & Tamis-LeMonda, 2008). These results, therefore, emphasize the importance of considering the often-overlooked influence that children have on their parents in family and developmental studies (Pardini, 2008). To be noted, these effects persist from early to middle childhood–a period of three to four years. This is of note because this developmental period is rarely studied compared with adolescence during which children are believed to have gained greater independence to exert observable influence on their environment (Pardini et al., 2008; Scarr & McCartney, 1983). Taken together, the results from this study underscore the potential children have to influence their parents and the fact that these effects are comparable in magnitude to the parent-driven effects. These findings are particular important for the field of family science as they imply that children’s well-being and functioning are potential policy levers for altering intra-familial dynamics. This possibility is critical when considering that these processes are difficult to manipulate through large-scale policy initiatives. Thus, intervention programs can improve the parenting that children receive by targeting children’s own behavior (see also, Ansari & Crosnoe, 2015a).
Distinctive typologies of child functioning at kindergarten were identified across children’s externalizing behavior, internalizing behavior, interpersonal skills, self-control, approaches toward learning, and academic performance. First, some homogeneity across these developmental domains was observed: children who were competent in one domain were more likely to exhibit similar performance in other areas. Children’s eagerness to learn co-occurs with their academic performance (Hair, Halle, Terry-Humen, Lavelle, & Calkins, 2006); self-control accompanies children’s interpersonal skills (Quirk, Nylung-Gibson, & Furlong, 2013); and their internalizing and externalizing behavior problems tend to occur simultaneously (Gilliom & Shaw, 2004). Second, the two groups identified as strong performance across domains displayed distinctive patterns of functioning, with one group excelling in behavioral competence and the other group excelling in academic performance. These nationally representative typologies extend our knowledge of the dynamic nature of child development (Edelbrock, & Achenbach, 1980; Nagin & Tremblay, 1999) and offer a fine-grained differentiation among children in addition to the global internalizing versus externalizing dichotomy.
More interestingly, however, cross-tabulation analyses demonstrated that different child functioning typologies at the kindergartens wave were associated with different parenting profiles at third grade. In general, parents were less likely to experience psychological risks, exhibit harshness and problematic parenting functioning, and more likely to exhibit warmth at third grade if their children were in the better functioning group across developmental domains at kindergarten. However, children’s academic performance appears as a potential antecedent of parents’ harsh parenting practices. Specifically, compared with the average-functioning group, children with strong behavior but poor academic skills were more likely to experience harsh parenting three years later. Academic performance might still drive parents’ harshness as harshness is more frequent among poorly performing children. Why might this be? During the transition to elementary school parents’ place a stronger emphasis on children’s academic functioning and treat it more as an overall indicator of children’s competence; thus, parents are more emotionally affected by children’s academic performance. Harsh parenting, characterized by immediate reactivity and intensified negative affect, is more likely to be influenced by these strongly academic motivated processes.
Results from the profile analyses indicated that the sizes of the child-driven effects varied substantially across groups. In general, children with more optimal functioning elicited greater improvements in parents’ functioning. Compared with children in the two least optimal functioning profiles, those in the average, strong behavior and above average academic, and strong academic and above average behavior profiles elicited greater improvements in parents’ functioning over time. For example, children with strong academic skills and above average behavior had parents who demonstrated 44% of a standard deviation improvement in parental functioning over time as compared with children with very poor behavior and academic skills. However, there appears to be a threshold where these elicitation processes level off, both at the bottom and top end of child functioning distribution. Thus, even though academic performance might elicit harsher parenting, as long as the child is in the above-average functioning profiles, the strength of the child-driven effects on parental functioning would not differ significantly.
In sum, the present study incorporates the advantages of two different approaches of exploring child driven effects. On the one hand, the traditional cross-lagged model approach informs us of the strength of child-driven effects; on the other hand, the profile analysis informs us how the strength and direction of these effects would change as the function of the children’s functioning profiles at the earlier time point. This potentially explains the variability in the child-driven effects that have been demonstrated in prior developmental and family studies. In other words, the magnitude of the child-driven effects might depend on the typologies of children’s developmental functioning and, therefore, require continued attention in family studies. Thus, similar to the recent call by Duncan (2015), our results have implications for the field of family and developmental psychology more broadly by underscoring the importance of applying multiple estimation methods to the same data. In doing so, not only did we replicate our findings across methodologies—lending support to our general conclusions—but, just as importantly, the profile analyses provided nuances that were not possible to capture with the cross-lag models.
Despite these important contributions, the results of the study should be interpreted it light of its limitations. First, even though the study spans approximately three to four years— from kindergarten to third grade—small time intervals would be ideal to test the exact timing of when these child-driven effects are observed. Along these same lines, continued research is necessary to test the strength of these child driven effects throughout different developmental stages. Second, the measures available in the ECLS-K were somewhat limited when compared with other community samples and, thus, these data do not capture the complexity of family functioning. Third, although we tapped into the heterogeneity of the child driven effects through the lens of parent and child functioning typologies, there are other factors and stratification systems that require continued empirical inquiry. To name just a few, continued research is necessary to determine how child effects might vary as a function of families’ socio-economic and cultural backgrounds that might amplify (or weaken) the child elicitation process (Ansari & Crosnoe, 2015a, 2015b). Fourth, our sample consisted of almost entirely mothers and we did not look at the role of fathers and siblings, and how they might affect these transactional processes; thus, continued research is necessary to evaluate these dyadic processes with fathers and determine how child-driven affects influence parenting styles in families with multiple children. Fifth, the effects sizes of the cross-lagged model were modest and, thus, generalizations of these results should be made with caution. We do note, however, that the heterogeneity of these effects may have resulted in an underestimation of effect sizes as we did observe larger effects when examining individual differences in these processes. Finally, although person-centered modeling provides the flexibility of identifying distinct subgroups within the population, these methods remain fairly subjective and, thus, should be interpreted with caution (Bauer & Curren, 2003).
With these associated limitations in mind, the current study leveraged developmental theory (Lerner, 2006; Sameroff & Chandler, 1975) to examine a relatively understudied phenomenon—the effects of children on their parents. Taken as a whole, this study demonstrated the importance of considering “child effects” in the fields of developmental and family psychology while also underscoring the potential for individual differences in these processes across the early to middle childhood years. Just as importantly, the findings reported herein have potential implications for intervention and prevention programs aimed at improving parents’ functioning and other family dynamics, suggesting that children’s own skills and behaviors might be a potential point of intervention.
The authors acknowledge the support of grants from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R24 HD42849, PI: Mark Hayward; T32 HD007081-35, PI: Kelly Raley) and the Research Fund for the Doctoral Program of Higher Education of China (SWU2120121851, PI: Ni Yan). Opinions reflect those of the authors and not necessarily the opinions of the granting agencies.
Ni Yan, Southwest University, P.R.C.
Arya Ansari, University of Texas at Austin, U.S.A.