Foster Parents’ Parenting and the Social-Emotional Development and Adaptive Functioning of Children in Foster Care: A PRISMA-Guided Literature Review and Meta-Analysis

Considering factors such as maltreatment experiences, former placement changes, and age at placement, the influence of foster parents’ parenting is fundamental for the development of children in foster care (CFC; Orme and Buehler 2001). Identifying important variables for effective foster parenting (i.e., parenting that facilitates adaptive child development, regulation, and adjustment, including cognitive, social-emotional functioning, and attachment security, and reduces maladjustment symptoms and developmental pathways) may help with the selection of appropriate foster parents for CFC (Washington et al. 2018). This can have positive effects on the development of CFC in particular and on the costs of youth welfare institutions in general.

Regarding the development of CFC, Leve et al. (2012) found that CFC with maltreatment experiences have a higher risk of developing mental disorders. They also often show developmental delays (Oswald et al. 2010). In addition, alterations in neuroendocrine stress-response functioning (Dozier et al. 2006; Fisher and Stoolmiller 2008) increase the risk of executive functioning deficits (Pears et al. 2008; Bruce et al. 2009), and alterations in social information processing and emotional regulation abilities (Price and Landsverk 1998; Kay et al. 2016) have been described in CFC across multiple studies. Maltreatment and neglect experiences of children may lead to adaptive behavior in an abusive family but may lead to problems in a foster family. For example, multiple studies have shown that children who have experienced maltreatment or abuse in the past are overly wary of angry faces (da Silva Ferreira et al. 2014). This might be adaptive when living with parents whose anger may be an important threat cue (Belsky et al. 2012); however, it comes at the cost of assuming hostile intent too readily under benign conditions and might thus lead to aggressive responses that would not have been evoked if the child’s attributions had been different (Dodge et al. 1995). Therefore, such behaviors may require tailored parenting responses among foster parents (Solomon et al. 2017).

This paper addresses the question of which kind of parenting dimensions are relevant to CFC’s developmental outcomes. It is based on a literature review and meta-analysis as important bases for systematically exploring the associations of parenting factors with the developmental outcomes in CFC.

Darling and Steinberg (1993) noted that it is essential to give a clear definition of the term “parenting” because of the use of different indicators for parenting in previous research. The authors introduced an integrative parenting model, which includes three different categories of parenting: parenting goals, parenting style, and parenting practices (or behavior). The present review follows these terms and definitions, as further explained below. Furthermore, we discriminate between functional and dysfunctional parenting. As a result, functional parenting can be seen within the dimensions of responsiveness and demandingness/control (e.g., Wolfe and McIsaac 2011), which includes provisions of warmth and limit setting matched to the children’s needs. This also includes the appropriate setting of boundaries on the one hand and encouraging children’s competencies on the other hand. By contrast, dysfunctional parenting is defined as either overly strict and rigid or unclear and inconsistent parental behavior, which therefore misses the appropriate orientation towards children’s needs. Furthermore, dysfunctional parenting includes the component of psychological control of children (i.e., parental attempts “to control their child’s behavior using psychological tactics aimed at undermining their emotional security or sense of self”; ibid., p. 804). Parenting goals include socialization goals for the child, such as the “acquisition of specific skills and behaviors” (e.g., appropriate manners or academic abilities) and of more global qualities (e.g., curiosity or critical thinking; Darling and Steinberg 1993, pp. 492–493). Therefore, parenting goals do not necessarily imply direct links to behavior, but instead refer to their fundamental attitudes towards parenting. Parenting style is defined as a “constellation of attitudes toward the child that are communicated to the child and create a specific emotional climate” (e.g., a combination of tone of voice and body language; ibid.). Baumrind (1991) identified four different kinds of parenting styles, as a result of combinations of parenting behavioral expressions. Authoritative parents show high responsiveness towards their child, and they practice appropriate control to encourage their child’s adaptive development. Parents with an authoritarian style show high expressions of appropriate control, as well as non-appropriate control (e.g., psychological control), to influence the child. They show low levels of responsiveness. High expressions of responsiveness without the use of any control characterize a permissive parenting style. Finally, neglecting parents show low levels of responsiveness and control. Parenting behaviors are “behaviors defined by specific content and socialization goals” (e.g., helping with homework and asking about hobbies; Darling and Steinberg 1993, pp. 492–493). Because of the immediate relation to specific situations, parenting behaviors are the easiest to observe.

Foster families are considered caring environments that enable the CFC to thrive despite past adverse experiences (Comas-Diaz et al. 2012). Thus, parenting in foster families might support resilience through role modeling processes and the experience of positive, trustworthy, and stable relationships. Schofield and Beek (2005) found qualitative data to support a successful foster family model in which foster parents’ parenting behavior promotes the trust of CFC in availability, reflective function, self-esteem, autonomy, and family membership. Moreover, Oswald et al. (2010) argued that even though there is much research on children in foster care regarding maltreatment, abuse, neglect, and mental health, very few studies have focused on the children’s long-term development after they transitioned into foster care.

Regarding parenting and children’s development in the general population, meta-analyses showed small to moderate positive associations between functional parenting and adaptive child developmental variables and attachment security, respectively (Karreman et al. 2006; Pinquart 2015; De Wolff and van Ijzendoorn 1997). In contrast, dysfunctional parenting showed small to moderate negative associations with adaptive functioning (Karreman et al. 2006; Pinquart 2015). Additionally, externalizing and internalizing problem behaviors in children showed small to moderate negative associations with functional parenting (Rothbaum and Weisz 1994; Pinquart 2017) and small to moderate positive associations with dysfunctional parenting (Pinquart 2017; McLeod et al. 2007a, b).

Even though these results may give the first hint of how parenting and developmental outcomes relate to each other in foster families, it is challenging to answer this question based on these research results as they mainly come from studies that include biological families. On the one hand, foster parents’ parenting behavior might have a similarly strong, or even stronger association (compared to biological families) to the development of the CFC, given their adverse early life events (Heller et al. 1999). In addition, foster care placement is also often associated with a change in the context of living (e.g., by moving to an area with lower occurrences of community-level violence), which also influences children’s problem behaviors, and may thus increase the association between parenting and developmental outcomes in CFC (Lynch and Cicchetti 1998). Furthermore, it is considered particularly difficult to change experiences and learning processes during early childhood when they are adverse experiences. Additionally, children’s genetic predispositions and their parents’ childrearing regimes are known to be closely interwoven (Maccoby 2000). Thus, dissimilarities between the genetic make-up of foster parents and their nonbiological children might also contribute to smaller associations between parenting behavior and outcomes in these populations. McCrory and Viding (2015) proposed a latent vulnerability model that explains why children with maltreatment experiences more often show problem behavior and why interventions (such as living in a foster family) can be less effective for some children with maltreatment experiences than for others. The authors propose that persons with childhood maltreatment experiences adapt their neurocognitive threat processing, which may be vital in threatening environments. However, those persons may “overattribute threat in ways that increase the frequency of reactive aggression” (ibid., p. 500).

Furthermore, negative expectations of CFC regarding their foster parents’ behavior (because of adverse experiences in the biological family) may change the association between foster parents’ parenting and children’s development (Milan and Pinderhughes 2000), compared to the association between biological parents’ parenting and children’s behavior. This is in line with the results from Kemmis-Riggs et al. (2018). They found that more traditional parenting training for parents of biological children had little effect on foster families. Additionally, Gardenhire et al. (2019) noted that foster parents often miss information regarding CFC’s former experiences. This may also have a reducing effect on the association of foster parents’ parenting and the development of the CFC, because it is more difficult for foster parents to tailor their parenting in response to the needs of the CFC.

As indicated above, associations between developmental outcomes for children and parenting may be moderated by several proximal and distal factors. Such effects may have to be considered, especially in foster families. Several variables were identified regarding child characteristics that could influence foster care disruptions, which are supposed to be the result of maladaptive developmental pathways in CFC (Oosterman et al. 2007). As a result of this, CFC more often experienced disruptions when placed at an older age. Regarding the gender of the CFC, Leathers (2002) found a different prevalence of conduct problems for boys and girls in foster care, with boys showing fewer symptoms of conduct disorder. Additionally, placement instability is associated with more behavioral problems (Leathers 2002), developmental delays, and mental disorders in CFC (Oswald et al. 2010). However, because of the above noted assumed reciprocal longitudinal effects of parents’ and children’s behavior, the length of residence with the actual foster family may influence this association (ibid.).

Regarding foster family characteristics, the age of parents is one well-documented factor in the research of parenting (e.g., Van Holland De Graaf et al. 2018). Considering that there are often differences between the mean ages of foster and biological parents (Chodura et al. 2019), the age of foster parents is a potentially relevant variable for the association of parenting and development of CFC. Foster parents further can be differentiated into professional and nonprofessional foster caregivers. Professional foster caregivers often have special educational training (also called therapeutic foster care, e.g., Murray et al. 2010). In contrast, nonprofessional foster caregivers are not explicitly trained in these areas. However, they nevertheless often receive support from caseworkers. Oosterman et al. (2007) found that foster parents’ professionalism was associated with more adaptive developmental outcomes for CFC. Furthermore, Leslie et al. (2000) showed that the mental and physical needs of foster children are associated with the socioeconomic status (income, education, and employment status; APA, Task Force on Socioeconomic Status 2007) of the foster family and may moderate the associations between parenting and child development for CFC.

Given the potentially challenging behaviors and special needs of CFC, it is crucial to investigate further the association between foster parents’ parenting and developmental outcomes of CFC. Although other reviews in the past have focused either on the efficacy of foster parents’ interventions as a whole (e.g., Goldman Fraser et al. 2013; Kerr and Cossar 2014; Kinsey and Schlösser 2013), different intervention components, or certain delivery formats (Kemmis-Riggs et al. 2018; Gubbels et al. 2019), or have descriptively summarized factors associated with outcomes for CFC (Jones et al. 2011; Goemans et al. 2015; Washington et al. 2018), none of them have attempted to provide a quantitative analysis of the associations between parenting behaviors, styles, and goals of foster parents and various outcome measures for CFC. Given the substantial individual and societal implications of foster care, it is of the highest importance to establish sound empirical evidence that improves the long-term developmental outcomes of this burdened population.

Therefore, in the present literature review and meta-analysis, we aim to answer the research questions of whether the parenting of foster parents is associated with the development of CFC, and more specifically, whether various components of foster parents’ parenting are differently related to specific developmental outcomes of CFC, including cognitive as well as social-emotional functioning. Furthermore, we aim to identify the characteristics of foster parents, the foster family, or the CFC that moderate these associations.

Figure 1 displays the research process. The initial literature search was done in August 2017 in the ERIC, PsycINFO, Psyndex, and PsycARTICLES databases. The following search terms were used in all variations over several search attempts: (foster children OR foster care) AND ([longitudinal OR repeated measures OR pretest posttest] OR [parent* OR rear* or care*]) AND relation*. No date limits were specified for the inclusion. The search was updated in May 2020. Overall, the search attempts led to 3761 potential publications. Then, the titles and abstracts of the publications were scanned. Full texts of conceivable studies were checked further for criteria regarding inclusion or exclusion. The lists of references of included articles were checked carefully for more potential primary studies, and 113 additional potential publications were found. Full texts were requested directly from the authors or purchased if not otherwise available on the various platforms.

For the search in 2017, authors of studies with missing data for the meta-analysis were contacted by email to obtain the missing data and ensure the inclusion of a maximum number of primary studies. Four additional studies were added by doing this. Twenty-one full texts could not be obtained online or by contacting the authors directly by mail; twenty of these were doctoral dissertations. In the 2020 update, all potentially interesting primary studies could be retrieved from the search platforms.

Publications had to meet the following inclusion criteria: (a) primary study about children in foster care, (b) outcome reflecting the children’s development or behavior, (c) at least one variable related to foster parents’ parenting, (d) study written in German or English, (e) sample group with at least ten participants, and (f) quantitative measures reported to compute an effect size.

To increase the number of potential studies, and therefore, to increase the review’s representativeness, cross-sectional and longitudinal studies without treatment were both identified as appropriate. Waitlist or care-as-usual control groups in intervention studies were also identified as adequate study designs. It was assumed that the results of control groups (without the results of the intervention groups) would show longitudinal pathways, as seen in other studies without interventions. Studies investigating CFC samples with medical conditions (e.g., substantiated prenatal drug exposure) were excluded due to potential biases.

The reasons for excluding a primary study were mostly related to “parenting not examined” (N = 104), “a qualitative design of the primary study” (N = 38), or “not enough quantitative information” (N = 37).

The coding manual for the meta-analytical computation consisted of two variable areas: one in which the moderators that had been shown in earlier research to potentially influence the parenting-development association were included. Coded variables for this area were Children in foster care: the examined area of development of the CFC, mean age of the CFC, the gender distribution of the CFC, length of residence in the current foster family, and the number of placements of the CFC; Foster parents: area of parenting (parenting goals, parenting style, parenting behavior; functional vs. dysfunctional), and the mean age of the foster parents; Foster Families: professionalism, kinship relationships between the foster parents and the CFC, the number of children living in the family, the family income, the highest education level of the foster parents, and their employment status.

The second area contained general study descriptions and potential study artifacts that may also affect the reported associations. These consisted of General characteristics: the publication year, authors, scientific area, country of study, and publication type; Children in foster care: the information source; Foster parents: the information source; Measurements: the study design, the statistical measurement of the association, the sample size, the time between measurements (for longitudinal studies), and an examination of additional variables; Study quality. The study quality estimation followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement (von Elm et al. 2014). For every item on the primary study checklist, it was rated (yes = 1, no = 0) in terms of whether the primary authors followed the recommendations. Studies could achieve a maximum of 22 points according to the STROBE statement’s list of recommendations (ibid.). The complete coding manual can be found in the supplemental materials related to this article.

Two independent raters did the coding following a coding manual generated by the first author. One study (Ackerman and Dozier 2005) was coded by both raters and then discussed to ensure the coding sheet’s unambiguity. Differences and misunderstandings were deliberated before coding the rest of the studies. This study, therefore, was not included in the computation of interrater agreement. Interrater agreement across all variables in the computation was 92.06% and therefore excellent. It was lower but still satisfactory for coding of additional variables (72.29%), area of parenting (75.80%), and area of child development (86.94%). The agreement was highest for type of measurement and professionalism (99.36% each), as well as for publication year and study design (98.41% each).

The effect sizes were computed with the statistics program R (R Core Team 2017) using the packages metafor (Viechtbauer 2010) and robumeta (Fisher et al. 2017). The moderator analyses were done as meta-regressions with robumeta. A random-effects model was conducted (Hedges 1981). Although there was no restriction for publication type, it was checked for publication bias by the trim-and-fill method (Duval and Tweedie 2000). This method assumes that funnel plots will show left and right symmetrical variance from the overall effect size of a meta-analysis if there is no publication bias. It, therefore, shows how many studies need to be reported to obtain the assumed symmetry. In some studies, sample sizes of CFC and foster parents were different, so the sample size of CFC was included in the analysis.

In some studies, the same construct was presented with different outcomes or independent variables. Therefore, effect sizes were integrated through robumeta to clusters according to the recommendation of Hedges et al. (2010). Following this step, an intercorrelation ρ had to be estimated to define clusters. A conservative correlation of ρ  = 0.99 between studies of the same cluster was assumed in the study at hand. Clusters had been defined when effect sizes were provided in the same study or when characteristics (e.g., authors, sample sizes, and descriptions) in studies showed that the same sample was used for different papers. Therefore, the results can be reported on the variable level or study level. The variable level results were all integrated according to effect sizes, which could be more than one per study, corrected by clustering. The results on the study level (e.g., descriptions of the studies) were based on the number of studies. Effect sizes were interpreted according to Cohen (1988), with r = 0.1 as small, r = 0.3 as moderate, and r = 0.5 as high effect.

For dysfunctional and functional parenting behaviors, separate meta-analyses were computed for variables of child developmental outcomes (Tables 2 and 3).

The studies were published between 1994 and 2019. Fifteen studies came from the field of psychology, and twenty-five studies were cross-sectional. The study quality for all 31 studies ranged from 15 to 21.5, and the sample sizes ranged from n = 20 to n = 5516. As expected, more functional parenting was significantly associated with more adaptive developmental outcomes for CFC (t = 2.99, p = 0.031), along with less externalizing (t = 2.51, p = 0.026) and internalizing problems (t = 3.42, p = 0.004), respectively. Additionally, more dysfunctional parenting behavior was significantly associated with more externalizing (t = 2.54, p = 0.039) and internalizing problems (t = 2.54, p = 0.044) for CFC. Note that some random-effects models have less than four degrees of freedom, which indicates few available effect sizes. Therefore, the analysis has low statistical power, and those results must be considered with caution.

The forest plots were inspected and showed similar and expected distributions for all combinations of variables. Forest plots of the associations of parenting behavior and child development variables can be retrieved as supplemental electronic material (Figures E1 to E10).

For reliable moderator analyses, child developmental outcome variables were combined to increase statistical power. Therefore, adaptive functioning and attachment security were combined to indicate adaptive child development, and internalizing problems, externalizing problems, and total problem behavior were combined to indicate maladaptive child development (Table 4).

Funnel plot and trim-and-fill methods were conducted with the package metafor (Figures E11 to E20, electronic supplementals). This was not possible with the clustered meta-analytical model from robumeta, but effect sizes for both methods overall showed the same directions. Therefore, this might be the best estimation. No indices for publication bias were found except for one effect size. For the association between functional parenting behavior and internalizing problem behavior, ten studies were computed as missing on the left side (Z = − 6.89, p < 0.001). However, for this effect size, missing studies on the right side would have been expected for publication bias to indicate a trend to zero. Adding the hypothetical effects to the overall effect size would lead to higher associations and would support the reported results. Overall, no evidence was found to support the assumption of unpublished effects with zero or small effect sizes.

The authoritative parenting style could be identified when reviewing the literature for functional parenting styles. Three studies with 13 effect sizes could be interpreted from the literature review. Two studies were from the field of psychology. All studies were cross-sectional and were published between 1999 and 2014. The study quality ranged from 16 to 18, and the sample sizes ranged from 38 to 104. The age range of foster parents was low (45.14–47.18 years) compared to a high age range for the CFC (4.44–11 years¹). Additionally, the range for the CFC’s length of residence was rather high (22.3–44.76 months). Eight associations between the authoritative parenting style and internalizing, externalizing, and total problem behavior were negative but small (r = − 0.13 to r = − 0.21), according to Cohen (1988). Four associations were smaller than 0.10 (r = − 0.03 to r = − 0.07). Associations between the authoritative parenting style and the adaptive functioning of CFC were positive; two were small (r = 0.12 and r = 0.25), and one only r = 0.01. This may indicate that the effect sizes were found to aggregate into a small overall effect (at most).

Authoritarian and permissive parenting styles were identified as dysfunctional parenting styles. Three studies and 21 effect sizes could be used in the literature summary. All studies had a cross-sectional design and were from the field of psychology. They were published between 2008 and 2015, with sample sizes ranging from 33 to 104, and the study quality ranging from 16 to 18. A low range in the ages of CFC (11–13.21 years) could be identified, in comparison to a high range in the length of residence (25.97–83.21 months) and the age of foster parents (45.14–61.03). The authoritarian parenting style was associated slightly positively with internalizing, externalizing, and total problem behavior in CFC in one case (r = 0.17) and moderately associated in seven cases (r = 0.25 to r = 0.42), indicating an overall small to a moderate positive association between the authoritarian parenting style and maladaptive child development outcomes. However, a similar relation was also shown for the authoritarian parenting style and placement stability (r = 0.16 and r = 0.20 in one study). A permissive parenting style was associated slightly negatively with placement stability in two studies (r = − 0.12 and r = − 0.18) and, in one case, near zero with r = − 0.06. The association with total problem behavior was also very low in one calculation (r = − 0.04) and small in three cases (r = − 0.15 to r = − 0.13). Additionally, the permissive parenting style showed slight positive associations with internalizing problems in three calculations with effect sizes between r = 0.13 and r = 0.16, and in one effect size close to zero with r = − 0.01.

It was possible to compute effect sizes for functional parenting goals and CFC externalizing and internalizing problem behaviors, respectively, across four studies (Table E5 in the supplemental material). As a result of this, foster parent attachment goals were the most investigated parenting goal variables. However, all associations were not significant.

Three studies examined dysfunctional parenting goals. Specifically, these goals were criticism, rejection, overprotection, and the challenging aspects of being a foster parent (such as looking after a child with special needs). One study showed a medium association with total problem behavior in CFC (r = 0.35 and r = 0.46) and one study showed a high association between overprotection and placement stability (r = 0.99). The challenging aspects of fostering were not associated with externalizing problems for CFC (r = − 0.01).

Regarding foster parenting behavior, more functional parenting was associated with more adaptive developmental outcomes for CFC and less maladaptive developmental outcomes, respectively. Additionally, more dysfunctional parenting behavior was associated with more maladaptive developmental outcomes for CFC. This shows the potential association of foster parenting behaviors and CFC’s development, especially for functional parenting behavior. It may indicate a buffering or correcting effect of functional parenting behavior for CFC despite former maltreatment or experiences of neglect before foster placement. However, according to the categorization provided by Cohen (1988), all associations were small to moderate, suggesting that many additional variables must be considered when exploring the development of CFC.

When computing possible moderators, it was evident that many variables could not be computed due to a lack of variance in these measures. However, the meta-regression indicated a stronger positive association of functional parenting behavior and adaptive child development in longitudinal studies. This may indicate a stronger relationship between foster parents’ and children’s outcomes over time, resulting in a greater influence on parental behavior. Additionally, CFC may first need to learn adequate reactions and cognitions and adapt to their foster parents’ different parenting behaviors. Foster parents might also need to learn to better adapt their parenting according to the CFC’s special needs.

Looking at the association between dysfunctional parenting behavior and adaptive child development, we found that the minor negative association was weaker for studies with nonprofessional foster parents, or when professionalism was not reported. One reason for this finding might be that parental and familial characteristics (e.g., the age of foster parents, the number of siblings in the family) may have been higher in nonprofessional foster families, or that children with more extreme needs had been selected for placement in foster families with professionals. Therefore, the association between parenting and child development may be moderated more differentially and therefore ended up smaller. The same assumption may apply to kinship states in foster families, where the association was strongest for nonkinship foster families. Furthermore, the negative association between dysfunctional parenting behavior and adaptive child development became stronger when more children lived in the family. This may be because other buffering elements in the relationship between foster parents and children might be reduced, and the dysfunctional characteristics of parenting behavior may more easily interfuse within the parent–child relationship.

In summary, parenting behavior was the best-examined parenting variable in foster family studies, and our meta-analysis supports the association and complex interactions with children’s developmental outcomes. The results could indicate a model of one dimension describing functional and dysfunctional parenting behavior, as Wolfe and McIsaac (2011) described for community samples, because the two linear models could also show one linear effect. However, different potential moderators for the associations were found for functional and dysfunctional parenting behavior. Dysfunctional parenting behavior was also shown to be influenced by more moderator variables. In summary, the results show the high importance of well-elaborated definitions and concrete assessments of parenting behaviors in primary studies. An investigation of the emergence of functional and dysfunctional parenting behaviors here seems to be imperative in further research to address this issue. Additionally, further investigations of parenting behavior dimensions (e.g., responsiveness, demandingness/control, and psychological control; Wolfe and McIsaac 2011) are needed to better distinguish their impact on children’s development. This may also be important for foster parent interventions. Kemmis-Riggs et al. (2018) provided support for foster parent training that promotes functional parenting behaviors.

Many moderators could not be analyzed, even though there is some evidence in former research that other variables influence the association between foster parenting behaviors and the development of CFC. Considering the results of the meta-analysis at hand, the following moderators should be considered in future studies: the professionalism of foster care providers, the kinship status between foster parents and the CFC, the number of children living in the foster family, and the number of placement disruptions of the CFC. Furthermore, more longitudinal studies are necessary to investigate CFC’s development and infer the causes and effects of identified associations.

With respect to parenting styles, the results of the current review were mixed. Authoritative and authoritarian parenting styles were mostly related to child development outcomes, the same as one would expect for functional and dysfunctional parenting styles in biological families. These associations were small to moderate, just as they are for parenting behavior. This also applied to permissive parenting style and placement stability, as well as to internalizing problems. Surprisingly, the authoritarian parenting style was also associated with higher placement stability. One could argue that the longer that the children lived in their foster family, this parenting style in foster parents might have been expressed more due to their behavior. Therefore, foster parents may think a more authoritarian style is necessary. On the other hand, authoritarian foster parents may question the continuation of fostering less than parents who exhibit other parenting styles. Furthermore, permissive parenting was—unexpectedly—associated with less externalizing and total problem behavior. An explanation may be that permissive parents do not limit externalizing and total problem behaviors as much as parents with different parenting styles. Therefore, they may report those child behaviors less. By contrast, it also may be possible that the CFC truly do show less problematic behavior when experiencing those parenting styles. This is in line with previous research regarding the impact of former maltreatment and experiences of neglect (e.g., Bruce et al. 2009; Price and Landsverk 1998). By this, a permissive or authoritarian parenting style would be better tailored to the expectations, and therefore reactions, of the CFC (e.g., da Silva Ferreira et al. 2014). However, the limited evidence of associations between parenting styles and the developmental paths of CFC makes it difficult to explain those rather counter-intuitive findings. Goemans et al. (2015) found little impact on the course of behavioral or emotional problems in CFC, which may be explained by the long-term altering effects of maltreatment experiences on functional and structural brain development (e.g., Teicher et al. 2002). This is still a mostly uninvestigated area in previous research, and further investigation is highly needed.

Regarding the associations between foster parenting behavior and CFC’s developmental outcomes, our findings were generally in line with those reported previously for biological families (Table E6 in the supplemental material). Therefore, one could assume that the associations of parenting and children’s developmental outcomes are similar in foster and biological families. This may be a promising result because similar parenting interventions might also help foster parents improve the developmental outcomes of the children in their care. However, a descriptive comparison of the effect sizes may lead to the false assumption that the underlying processes are comparable in both types of families. For example, each family type’s different composition may lead to different moderating effects, as noted above. As a result, the effect sizes may be similar but also moderated by unrelated variables. Additionally, there is a high variance in effect sizes for these results. This may support the assumption that the relationship between foster parenting behavior and the development of CFC runs in the same direction as it does in biological families but is also influenced by additional or different moderators (for example, the number of placements of the CFC beforehand). Therefore, further research is needed to better distinguish between parenting and child development associations in biological and foster families. This goes along with Kemmis-Riggs et al. (2018), who identified that specific intervention components focusing on the special needs of CFC were effective. However, the differential susceptibility hypothesis (e.g., Pluess and Belsky 2010) may argue that CFC, who are most susceptible to environmental adversity, may also benefit most from developmentally supportive rearing conditions. Unfortunately, for the meta-analysis at hand, it was not possible to distinguish between the groups of potentially more or less susceptible children. Hence, this is another area that should be investigated in future research.

Additionally, the parenting style results, as combinations of parenting behavioral dimensions, show similar effects only for authoritative parenting styles. This may indicate that this parenting style is favored in the training of foster parents. The current evidence is rather limited concerning drawing firm conclusions, however. For instance, the influence of the foster parenting style and goals, especially, could not be meta-analytically computed entirely due to the lack of data. Future research should better differentiate those concepts to understand possible associations. In addition, the bidirectional character of the association, as stated for biological families (Brenner and Fox 1999), must be examined in more detail.

Furthermore, we chose to compute the interrater agreement instead of more conservative measurements. The coding sheet contained numeric, ordinal, nominal, and open-answer variables. It was, therefore, not possible to select a single statistical measurement for interrater reliability. Additionally, it is unlikely that coding decisions were made by chance because of the high number of open-answer variables. Therefore, a conservative interrater reliability measurement may not be necessary. The two variables with the lowest agreement (86.94% and 75.80%, respectively) were areas of child development and parenting, although the agreement still can be interpreted as satisfactory. In future research, a clearer definition of these central variables would be helpful to increase the comparability of different studies and research approaches.

Moreover, 21 studies could not be retrieved, and the full text of these could not be screened for suitability with the meta-analysis and review. This should be noted in terms of publication bias, even though publication bias testing did not flag this as an issue. Finally, it should be noted that we decided to include the attachment security of CFC in the analysis while searching for eligible studies because of the large appearance of those variables. However, we did not specify these terms in the database searches. Therefore, we cannot guarantee that all studies with eligible research questions for CFC attachment security were found.