The quality of parenting is critical for the development of adaptive functioning in youth (e.g., behavioral functioning, academic performance, mental health, psychological adjustment, social functioning; Luk, King, McCarty, Stoep, & McCauley, 2016). In particular, parenting is regarded as a strong predictor of life-course outcomes. Providing practitioners and researchers with a valid comprehensive measure of parenting could increase precision in determining areas of parenting that could be targeted in treatment and provide a more nuanced understanding of parental risk and protective factors that may prevent the emergence of a wide range of mental health and substance use problems. Despite wide acceptance of the multifaceted makeup of parenting, few measures exist that capture multiple aspects of overall parenting with good reliability, particularly for adolescents (Lindhiem & Shaffer, 2017; Smith, 2011). This may be due in part to the inherent complexity of effectively quantifying all relevant aspects of this critical socialization context with one measure. The aim of the current study was to develop a comprehensive and psychometrically sound measure of overall parenting quality inclusive of a variety of critical parenting dimensions for adolescent outcomes.

Parenting Dimensions

There is consensus on the existence of two multifaceted dimensions of parenting: parental warmth and parental control (e.g., Baumrind, 1978; Harkness & Super, 1992; Jacob, Moser, Windle, Loeber, & Stouthamer-Loeber, 2000). Parental warmth reflects the affective nature of the parent–child relationship as indicated by involvement, praise, warmth, and emotional availability (Barnes, Hoffman, Welte, Farrell, & Dintcheff, 2006). In contrast, the parental control dimension reflects behaviors directed at the child to shape behavior viewed as acceptable by the parent as indicated by discipline, supervision, and setting rules regarding youth behavior (Barnes et al., 2006; Darling & Steinberg, 1993). Both parental warmth and control are viewed as positive parenting behaviors that increase adaptive functioning among youth. For example, high levels of parental warmth predicted high levels of children’s empathy (Zhou et al., 2002) and social competence (Raver, Gershoff, & Aber, 2007), whereas low levels of parental warmth were associated with oppositional defiant disorder/conduct disorder (Kroneman, Hipwell, Loeber, Koot, & Pardini, 2011) and depression (Hipwell, Keenan, Kasza, Loeber, Stouthamer-Loeber, & Bean, 2008). Similarly, higher levels of parental control predicted lower rates of adolescent substance use (Alati et al., 2010) and conduct problems (Racz & McMahon, 2011).

A specific aspect of parental control, parental monitoring, has gained widespread attention given its strong association with adolescent problem behavior and substance use (Barnes et al., 2006; Trucco, Villafuerte, Heitzeg, Burmeister, & Zucker, 2016). Researchers have questioned interpretations of how the construct is operationalized (Kerr, Stattin, & Burk, 2010). Specifically, early research defined parental monitoring as active efforts by the parent to monitor and track their child’s whereabouts, activities, and associations (Dishion & McMahon, 1998). Yet, items on empirical measures of parental monitoring tend to reflect actual knowledge of children’s activities rather than active tracking efforts (Kerr & Stattin, 2000). This is a notable distinction, as parents can obtain this knowledge in multiple ways, including active monitoring, parental solicitation, as well as voluntary youth disclosure (Kerr et al., 2010). To capture the various aspects of parental monitoring, Kerr and Stattin (2000) developed a measure that assesses each of these approaches (i.e., active monitoring, parental solicitation, youth disclosure) separately and advocated for the use of the term parental knowledge in cases where active parental tracking efforts are not directly assessed. Despite strong support for parental monitoring/knowledge as a robust predictor of child and adolescent problem behavior (e.g., Abar, Jackson, & Wood, 2014; Trucco et al., 2016), this construct is often omitted from general measures of parenting.

Although parental warmth, control, and monitoring are conceptually distinct, they also have substantial overlap. Namely, parents who are high in parental warmth also tend to employ clear limit setting and expectations for their children (e.g., indicative of moderate parental control; Darling & Steinberg, 1993). Previous work demonstrates a strong correlation between parental warmth and parental control domains, which supports the co-occurrence of both practices among mothers and fathers (Barber, Olsen, & Shagle, 1994; Kuppens et al., 2009). As noted previously, parental monitoring is typically considered a subcategory of parental control, as the overarching goal of monitoring is to manage and regulate child behavior (Crouter & Head, 2002). Parental knowledge, especially through voluntary child disclosure, also has a strong association with aspects of parental warmth. Parental knowledge is acquired primarily in the context of an open and trusting parent–child relationship through parents’ interest in their child’s activities as well as a child’s comfort in disclosing this information to parents (Kerr et al., 2010). Thus, it is likely that these parental warmth and control domains are interrelated.

Assessing Parenting

To date, a valid, comprehensive, questionnaire-type measure of overall parenting quality that encompasses key domains within the broad areas of control and warmth does not exist (Hurley, Huscroft-D’Angelo, Trout, Griffith, & Epstein, 2014). In their comprehensive review of parenting measures, Hurley et al. (2014) identified 164 measures published from 1985 to 2009. Among them, the authors found 25 measures that supplied some degree of psychometric information. After thorough comparison, the authors reported that only five measures (i.e., the Child Abuse Potential Inventory, Alabama Parenting Questionnaire [APQ], Parenting Alliance Measure, Parenting Scale and Parent Child Relationship Inventory) provided comprehensive psychometric data. Except for the APQ, the other four measures focused on different domains of parenting. Assessments that focus on a specific domain of parenting fall short of assessing the wide range of attitudes and behaviors that have a significant impact on the developing child.

Two measures that attempted to capture the depth and breadth of parenting domains are The Loeber Youth Questionnaire (LYQ; Jacob et al., 2000) and the APQ (Frick 1991; Shelton, Frick, & Wootton, 1996). Both the LYQ and the APQ parenting measures do not meet acceptable psychometric properties for use in research or clinical settings consistent with prior reviews (Holden & Edwards, 1989; Locke & Prinz, 2002). The APQ includes six parenting domains: involvement, positive parenting, poor monitoring/supervision, inconsistent discipline, corporal punishment, other discipline practices. In their original psychometric paper, the authors tested four assessment formats: parent and child questionnaire and interview formats (Shelton et al., 1996). There were several issues identified with the APQ. First, both child formats failed to differentiate families of children with disruptive behavior disorder diagnoses and control group families, which calls into question the validity of these assessments. Second, although the parent-report formats distinguished between families of children with disruptive behavior disorder diagnoses from control families, the poor monitoring/supervision scale demonstrated low internal consistency (Cronbach’s α ranges from .21 to .67 across formats) and low temporal stability over a 3-day period. Third, the parent-report of the corporal punishment scale had low internal consistency across formats (Cronbach’s α ranges from .09 to .49). Thus, developing a comprehensive assessment measure of parenting that is psychometrically sound is critical in advancing the field.

Current Study

One approach to address the current lack of parenting assessments is to create a comprehensive measure of parenting practices utilizing items from already established measures. By integrating both psychometrically strong measures that tap into a specific parenting domain, as well as less psychometrically strong measures that tap into broader parenting domains, we will leverage the strengths associated with both approaches towards assessing parenting. In the current study, data from 14 established parenting subscales were analyzed to identify key domains of parenting intended for a community sample of parents and their children in their early teens. These specific subscales were selected as they represent either broad dimensions of parenting that are relevant across a wide developmental period (e.g., involvement, discipline), or parenting practices that are critical during adolescence (e.g., parental knowledge, solicitation, and reactions to child substance use). First, an exploratory factor analysis (EFA) was conducted on data from adolescents within a community sample of families to identify the underlying structure of the 14 parent-report subscales. Psychometric properties (i.e., reliability, convergent validity, discriminant validity) were examined. Second, confirmatory factor analysis (CFA) using the structure identified in the EFA was conducted on data from a separate community sample. These approaches aided us in developing the Parenting Practice Measure (PPM), a comprehensive measure of parenting quality for adolescent samples.

Method

Sample

The current study was based on a community sample of 387 families (a caregiver and adolescent from each) from Erie County, New York, enrolled in the Adolescent Family and Development Project (AFDP) (Meisel, Colder, & Hawk, 2015). The AFDP is ideal for the development of a comprehensive measure of parenting given that it contains data from parents and adolescents on a large number of parenting measures. This longitudinal study investigates behavior problems and social contextual influences as risk factors for substance use initiation. The study used a random-digit-dial sample of listed and unlisted telephone numbers generated for Erie County, New York. Calls were made by trained telephone recruiters utilizing scripts that explained the nature of the study, inclusion criteria, and compensation for participation. The participation rate was 52.4% which is well within the range of that found in population-based studies requiring extensive levels of participant involvement (Galea & Tracy, 2007). Eligibility criteria for recruitment included an English-speaking child between the ages of 10 and 12 years without any physical impairments or cognitive deficits that would preclude completion of the study procedures and a caregiver willing to participate. One caregiver and one child per household were recruited. The initial wave consisted of 387 target families. The sample was roughly evenly split on sex (N = 213, 55% female). The majority were White/non-Hispanic (83%), 9% were Black, 2% were Hispanic, 1% were Asian, and 5% were categorized by another racial category. Median family income was $70,000 and 6% received public assistance income. The majority of parents had completed college or some graduate/professional school (58%). The sample compares well to the general population in Erie County, NY across a diverse set of characteristics including sex, race/ethnicity, income, and receipt of public assistance (US Census Bureau, 2012). Participating families were asked to provide the names and contact information of the youth’s three closest friends. One of the youth’s peers, as well as one of the peer’s caregivers, were also asked to participate in the study. The initial wave consisted of 362 peer families.

Procedure

Target family interviews were conducted in a research laboratory on a university campus. Peer families were given a choice to complete the interview on campus or at their house. Prior to the interview, the caregiver and adolescent were asked to give consent and assent, respectively. Participants were taken to the same room and study procedures were explained to both the caregiver and the adolescent. The consent and the assent forms were read verbatim to both the caregiver and the adolescent. Parents were asked to give consent and adolescents were asked to give assent after study procedures were discussed and questions regarding the study were answered. After completing the consent/assent procedures, the caregiver and adolescent were taken to separate rooms to enhance privacy. Caregivers completed a variety of self-report measures reflecting their own behavior (e.g., parenting practices, substance use) as well as their perception of their child’s behavior (e.g., aggression, temperament). All questionnaires were read aloud, and responses were entered directly into a computer to minimize the occurrence of random responding and missing data points. The Institutional Review Board at the university where this study took place approved this study.

Total attrition for the study was 7.5% (29/387). This study used only the Wave 3 data. The Wave 3 assessment occurred approximately 2 years after the baseline appointment. Families that did not complete the Wave 3 assessment (N = 20) did not differ on any demographic or study variables at baseline compared to those who completed the Wave 3 assessment. In total, 370 target families and 326 peer families participated in the Wave 3 data collection. The mean age for target youth at Wave 3 was 13.6 (SD = .59) years old and peer youth was 13.6 (SD = 1.09) years old. The mean age for target caregivers was 44.9 (SD = 6.18) years old and peer caregivers was 44.3 (SD = 6.54) years old at Wave 3. Additional sample demographic information from Wave 3 is presented in Table 1.

Table 1 Youth caregiver and peer caregiver wave 3 demographic comparisons
Full size table

Measures

Parent-Report Measures Used as Indicators of Parenting

A total of 14 self-report parenting subscales were included in the EFA. The sum scores of each subscale were included in the EFA, and thus, item refers to subscale scores. The same measures were used for the CFA among peer families. The first three subscales—(1) parental monitoring, (2) child disclosure, and (3) parental solicitation—were extracted from a measure developed by Kerr and Stattin (2000). Six other subscales—(4) shared activities, (5) parental involvement with the child, (6) time spent with the child, (7) positive parenting, (8) parent–child relationships, and (9) curfew—were taken from the LYQ (Jacob et al., 2000). Three other subscales—(10) parental efficacy, (11) smoking norms, and (12) alcohol norms—were derived from the work of Kodl and Mermelstein (2004). The remaining two subscales, (13) obligations to disclose (Smetana, Metzger, Gettman, and Campione-Barr, 2006) and (14) parental authority, were derived from the Adolescents and Parents Conceptions of Parental Authority scale developed by Smetana (1988). A brief description of these 14 subscales and their reliability follow.

  1. (1)

    Parental monitoring/knowledge This subscale consisted of nine items scored using a Likert scale ranging from 1 (Never) to 5 (Always) to assess parents’ knowledge of their child’s whereabouts, activities, and associations (α = .74). Participants were asked questions such as, “Do you know what your child does during his/her free time?” and “Do you know whom your child has as friends during his/her free time?”

  2. (2)

    Child disclosure This subscale consisted of five items scored using a Likert scale ranging from 1 (Never) to 5 (Always) to assess how much their children provide information about their whereabouts, activities, and associations (α = .76). Participants were asked questions such as, “Does your child talk at home about how he/she are doing in the different subjects in school?”

  3. (3)

    Parental solicitation This subscale consisted of five items scored using a Likert scale ranging from 1 (Never) to 5 (Always) to assess if parents actively solicit information about their child’s whereabouts, activities, and friendships (α = .67). Participants were asked questions such as, “In the last month, have you talked with the parents of your child’s friends?”

  4. (4)

    Shared activities This subscale consisted of four items scored using a Likert scale ranging from 1 (More than 1 month ago) to 4 (Yesterday/today) to assess topics that parents and their children may have talked about or shared activities in the past 6 months (α = .78). Participants were asked questions such as, “When was the last time that you discussed with your child his/her plans for the coming day?” and “When was the last time you talked with your child about what he/she had done during the day?”

  5. (5)

    Parental involvement with child This subscale consisted of five items scored using a Likert scale ranging from 1 (Almost never) to 3 (Often) to assess the level of parental involvement (α = .64). Participants were asked questions such as, “How often do you have a friendly chat with your child?” and “Do you talk with your child about how he/she is doing in school?”

  6. (6)

    Time spent with child This subscale consisted of four items scored using a Likert scale ranging from 1 (Almost never) to 3 (Often) to assess parental involvement (α = .80). Participants were asked questions such as, “On the average, how much time each day are you together with your child on weekdays, that is, when you and your child are both awake?” and “On weekdays, how much of that time are you doing something together, like making something, playing a game, talking, or going out together?”

  7. (7)

    Positive parenting This subscale consisted of eight items scored using a Likert scale ranging from 1 (Almost never) to 3 (Often) to assess the type and frequency of parental praise (α = .73). Participants were asked questions such as, “In the past 6 months, when your child did something that you liked or approved of, how often did you give him/her a wink or smile?”

  8. (8)

    Parent–child relationships This subscale consisted of 15 items scored using a Likert scale ranging from 1 (Almost never) to 3 (Often) to assess the nature and quality of the parent–child relationship (α = .80). Participants were asked questions such as, “In the past 6 months, how often did you think your child was a good kid?” and “Felt proud of him/her?”

  9. (9)

    Curfew This subscale consisted of three items scored using a Likert scale ranging from 1 (No set time) to 3 (Always set time) to assess curfew policies (α = .69). Participants were asked questions such as, “Does your child have a set time to be home on school nights?”

  10. (10)

    Parental efficacy This subscale consisted of 14 items scored using a Likert scale ranging from 1 (Not at all confident) to 10 (Extremely confident) to assess how confident parents feel about their influence on child behavior (α = .89). Participants were asked questions such as, “How confident are you that you can keep your child away from the wrong kinds of kids?”

  11. (11 and 12)

    Reactions to child cigarette/alcohol use These subscales consisted of 22 items each scored using a Likert scale ranging from 1 (Not at all likely) to 4 (Very likely) to assess parental beliefs, messages, and reactions to youth smoking/drinking (α = .75 and .76, respectively). Participants were asked questions such as “If you knew your child smoked/drank alcohol or tried smoking/alcohol, how likely is it that you would yell at him/her in disapproval?”

  12. (13)

    Obligations to disclose This scale consisted of 14 items on a Likert scale that ranged from 1 (Never) to 5 (Always) to assess caregiver perceptions of their child’s duty to disclose their behavior (α = .88). Participants were asked questions such as “Without you asking, how often does your child tell you or is willing to tell you about the following things? Hanging out at a friend’s house when no adult is home.”

  13. (14)

    Parental authority This scale consisted of 20 items scored using a Likert scale ranging from 1 (Strongly disagree) to 5 (Strongly agree) to assess parents’ conceptions of the legitimacy of their parental authority across multiple domains (α = .90). Participants were asked questions such as, “It is ok for me to make rules about what my child does after school.”

Measures to Assess Convergent Validity

A child-report measure and a parent-report measure were used to assess convergent validity with factors derived from the EFA. That is, it was hypothesized that the following two constructs would be correlated with the factor scores.

Parental Demandingness

This child-report scale consisted of five items scored using a Likert scale ranging from 1 (Strongly disagree) to 5 (Strongly agree) adopted from the Parenting Style Inventory (Darling & Toyokawa, 1997; α = .66). Participants were asked to rate their level of agreement on questions such as, “If I don’t behave myself, my parent will punish me.”

Parental Control

This subscale included five items scored using a Likert scale ranging from 1 (Never) to 5 (Always; Kerr & Stattin, 2000; α = .66). Participants were asked questions such as “If your child has been out late one night, do you require that he/she explains what he/she did and who he/she was with?”

Measures to Assess Discriminant Validity

Two parent-report measures, and a child-report measure were used to assess discriminant validity with factors derived from the EFA. That is, it was hypothesized that the following three constructs would not be correlated with the factor scores.

Parental Depression

This scale was comprised of 20 items scored using a Likert scale ranging from 0 (Never) to 3 (Often). It was adopted as the Center for Epidemiologic Studies Depression Scale (Radloff, 1977; [α = .91]). Participants were asked how they have been feeling in the past month such as, “Were you bothered by things?” and “Did you feel depressed?”

Caregiver Injury and Conflict

This scale was comprised of six items scored using a Likert scale ranging from 1 (Once in the past year) to 8 (This has never happened) from the Revised Conflict Tactics Scales (Straus, Hamby, Boney-McCoy, & Sugarman, 1996). Participants were asked questions about their relationship with their significant other/romantic partner/spouse who lives in the home and who is involved with caring for their child (α = .75). For example, participants were asked how many times they “accused their partner of being a lousy lover” and “threatened to hit or throw something at their partner.”

Current Nicotine Dependence

This scale was comprised of six items adopted from the Fagerstrom Test for Nicotine Dependence (Heatherton, Kozlowski, Frecker, & Fagerstrom,1991; (α = .70) to measure child’s nicotine dependence. Participants were asked questions such as “at present, do you find it difficult to refrain from smoking in places where it is forbidden?”

Data Analyses

First, an EFA was conducted using IBM Statistical Package for Social Science (SPSS) version 22.0 (IBM SPSS Corp, 2013) to identify the underlying structure of the 14 subscales using the target family data. Prior to the EFA, the data were tested for assumptions of normality and multivariate outliers. One variable, shared activities, showed a departure from normality with skewness of − 2.69 (SE = .13) and kurtosis of 8.39 (SE = .25). Accordingly, a reciprocal transformation was conducted, which resulted in reduced skewness of − 1.67 (SE = .13) and kurtosis of 1.55 (SE = .25). Seven multivariate outliers were identified, and therefore, eliminated. No missing data existed across the variables.

After assumption testing, variables were standardized so that they were on the same metric prior to performing analyses and data were further screened to determine adequacy for factor analysis. A viable factor analysis requires each item to be correlated with at least one other item at the level of .30 or greater (Tabachnick & Fidell, 2007). A review of the factor correlation matrix showed that 12 of the 14 items had correlations greater than .30. Two items (i.e., positive parenting and curfew) with correlation values below .30 were removed from further analyses.

Following a review of the data, EFA models using principal axis factoring extraction were conducted. This approach was preferred over principal components analysis, because it does not inflate variance estimates since it only analyzes shared variance, which, in turn, has the advantage of producing more generalizable and reproducible results (Costello & Osborne, 2005). To help improve interpretability and scientific utility of the solution, oblique rotation with direct oblimin was used to maximize high correlations between the factors. Decisions on how many factors to retain were evaluated using Kaiser’s eigenvalue greater than 1 criterion (Kaiser, 1960), Cattell’s Scree Test (Cattell & Vogelmann, 1977), and cumulative variance tests. Bartlett factor scores were extracted for further analyses. Bartlett factor scores maximize validity (DiStefano, Zhu, & Mindrila, 2009) provide unbiased estimates of true factor scores (Hershberger, 2005).

Next, psychometric properties of the retained factors were examined. This included examining internal consistency as well as convergent and discriminant validity. To assess convergent validity, bivariate correlation tests were conducted between factor scores and child-report of parental demandingness and parent-report of parental control. We hypothesized that higher scores on parenting factors would be positively associated with child-report of parental demandingness and parent-report of parental control. To assess discriminant validity, bivariate correlation tests were conducted between factor scores and child-report of nicotine dependence and parent-report of parental depression and caregiver injury. We hypothesized that parenting factors would not be associated with these variables.

A CFA was used to test whether the factor structure identified in the EFA fits a separate sample. Since the current study used two different groups of caregivers in the EFA and CFA respectively, it was important to measure demographic differences between target and peer caregivers. As presented in Table 1, results show that caregivers from target families did not significantly differ from caregivers from peer families. Before running the analyses, tests were conducted for possible violations of normality, outliers, and multicollinearity. The shared activities subscale had a kurtosis value above 7 (kurtosis = 17.35) and a skewness with the absolute value above 2 (skewness = − 3.80), indicating non-normality (Fabrigar et al., 1999). Consistent with the target family data, a reciprocal transformation was conducted on shared activities. Nevertheless, multivariate non-normality was still not within acceptable normality limits based on the skewness of − 2.04 (SE = .135) and kurtosis of 3.36 (SE = .269). Thus, the CFA was conducted using Mplus version 7 (Muthén &Muthén, 1998–2015) as it offers a choice of Robust Maximum Likelihood (MLR) estimation, which allows for parameter estimates with standard errors that are robust to multivariate non-normality and multivariate outliers (Byrne, 2016).

Results

The results of the EFA demonstrate that three factors had eigenvalues over Kaiser’s (1960) criterion of 1 (Table 2). Analysis of the inflection of the scree plot also confirmed retaining three factors. The pattern matrix derived using oblique rotation (with oblimin) indicated that all variables had standardized factor scores above .32, which is the minimum utilized in most social science research (Tabachnick & Fidell, 2007). There was no evidence of double-loading. In addition, the overall measure (α = .78) and all individual factors demonstrated good internal consistency (α = .75 for Factor one, α = .72 for Factor two, and α = .71 for Factor three). The factor loadings and the Cronbach’s alphas for each factor are shown in Table 3.

Table 2 Eigenvalues and total variance explained across possible extracted factors
Full size table
Table 3 Factor loadings for the three-factor solution and Cronbach’s alpha
Full size table

The three-factor solution accounted for 59.6% of the total variance explained (Table 2). Factor one, parental knowledge and affective relationships, accounted for 31.5% of the total variance explained based on the following four items: parental monitoring, child disclosure, obligations to disclose, and parent–child relationships. Factor two, Parental Control, accounted for 19.1% of the total variance explained based on the following four items: legitimacy of parental authority, parental efficacy, and reactions to child cigarette and alcohol use. Factor three, parental communication and involvement, accounted for 9% of the total variance explained based on the following four items: shared activities, parental solicitation, time spent with the child, and parental involvement with child.

Results of correlations between factors, convergent and discriminant validity tests, demonstrated satisfactory construct validity. Factor one and factor three were highly correlated (r = .63). As expected, parental control and parental demandingness were correlated with each of the three factors. The highest correlations were found between relevant measures of parental control and the parental communication and involvement factor (r = .32). One exception to these findings involved the correlation between parental control and the child-report of parental demandingness. Although significant (p < .001), this correlation (r = .16) was lower than anticipated. Also, as expected, child nicotine dependence, parental depression, and caregiver injury and conflict were not significantly correlated with any of the three factors with one exception. Parental depression was negatively correlated with parental communication and involvement (see Table 4).

Table 4 Correlations between factors and measures selected for convergent and discriminant validity
Full size table

Next, a CFA was used to test whether the factor structure identified for the 12 retained parenting scales in the EFA fit a similar sample. That is, parental monitoring, child disclosure, obligations to disclose, and parent–child relationships were added as indicators to the first latent variable (parental knowledge and affective relationships); legitimacy of parental authority, parental efficacy, and reactions to child cigarette and alcohol use were added as indicators to the second latent variable (parental control); and shared activities, parental solicitation, time spent with child, and parental involvement with child were added as indicators to the third latent variable (parental communication and involvement). The loading for the first item of each factor was fixed to one. Given the correlation between parental knowledge and affective relationships and parental communication and involvement from the EFA (r = .63), these factors were set to covary.

Overall, the CFA model using a similar sample demonstrated acceptable fit (Fig. 1). Although the Chi square was significant (χ2 (53) = 170, p < .01) as expected given the large sample size, the relative normed Chi square (170/53 = 3.2) indicated good model fit. The CFI (.89) and TLI (.87) both indicated acceptable fit. Lastly, the RMSEA (.08) also indicated acceptable model fit. All parameter estimates loaded in the expected direction with adequate size (p < .05). The only factor loading value below the recommended cutoff of .30 was parental authority (.26) within the parental control factor. However, this is consistent with the estimates derived from the EFA results. The loadings for all other variables were between .33 and .91. Consistent with the EFA results, parental knowledge and affective relationships and parental communication and involvement were strongly correlated (r = .83).

Fig. 1
figure 1

CFA model path diagram. PKAR parental knowledge and affective relationships (Factor one), PC parental control (Factor two), and PCI parental communication and involvement (Factor three). Goodness of Fit Indices: Chi Square = χ2 (53) = 170, p < .01; Relative/normed Chi square (χ2/DF) = 3.2; Standardized root mean square residual (SRMR) = .11; Comparative fit index (CFI) = .89; Tucker–Lewis Index (TLI) = .87; Root mean square error of approximation (RMSEA) = .08. *p < . 05

Full size image

Discussion

The primary objective of this study was to develop a comprehensive and psychometrically sound assessment measure of parenting quality for adolescents. The results from the EFA and an independent sample CFA were consistent and provided evidence for a three-dimensional structure for the PPM. The resulting three factors identified were: (1) parental control, (2) parental knowledge and affective relationships, and (3) parental communication and involvement. Parental knowledge and affective relationships reflected parents’ attempts to gain knowledge of their youth’s activities through monitoring and youth disclosure, as well as the quality of the parent–youth relationship. Parental control reflected parental reactions to their youth’s substance use as well as parents’ beliefs about their abilities and authority to set rules. Parental communication and involvement reflected parents’ attempts to solicit information from their youth, as well as the frequency and type of their involvement with their youth. By and large, the literature has utilized a binary (i.e., warmth/responsiveness, control/demandingness) and non-overlapping framework for conceptualizing parenting practices based on the seminal work of Baumrind (1978). The current study indicates that the quality of parenting is best conceptualized as an integration of these domains. This is consistent with more recent work indicating that the effectiveness of control-oriented parenting practices, such as monitoring and setting limits for youth, are largely contingent on relationship quality between the parent and the adolescent. For example, in their reinterpretation of parental monitoring, Kerr et al. (2010) concluded that parental knowledge of adolescent behavior is acquired primarily in the context of an open and trusting relationship with their child, the parent’s ability to actively monitor, as well as the child’s willingness to disclose information to the parent. The PPM is a comprehensive measure of parenting that accounts for the overlapping nature of these domains.

Psychometrics

Results offer support for strong psychometric properties of the PPM. In terms of reliability, the overall measure (α = .78) and all individual factors demonstrated acceptable internal consistency (αs = .71–.76). Assessment of convergent validity through correlations with theoretically sound measures were in the expected direction. The highest level of correlation between relevant measures of parental control and the parental communication and involvement factor, which prior work has framed as a dimension of warmth (Cablova, Csemy, Belacek, & Miovsky, 2016), might indicate that involved parents who communicate effectively may also provide more discipline and rules in the home. The low level of correlation between parental control and the child-report of parental demandingness may have been due to cross-reporter discrepancies (Abar et al., 2014). Some studies suggest that parents’ reports may be biased as they are likely to overestimate levels of certain parenting practices due to social desirability (Smetana et al., 2006).

Findings related to discriminant validity tests were generally in the expected direction. However, one exception was a significant negative association (p < .05) between parental communication and involvement and parental depression. Of the different factors, parental communication and involvement is most likely to be impacted by parents’ depression. For example, prior research demonstrates that mothers with depression tend to be less involved, exhibit higher levels of negative and critical communication, and have difficulty setting limits with their children (Middleton, Scott, & Renk, 2009). In fact, much of the prior work on parental depression focuses on how depressive symptomatology negatively impacts parental involvement, also a warmth dimension, compared to parental solicitation, a control dimension (Elgar, Mills, McGrath, Waschbusch, & Brownridge, 2007). Perhaps anhedonia experienced by depressed parents is expressed as limited time spent with their child doing shared activities, more so than reduced limit setting or monitoring.

Assessing Parenting

Overall, the parenting literature has utilized a binary framework for assessment, suggesting that there are two main components of parenting: parental warmth/responsiveness and parental control/demandingness. Moreover, prior work has tended to view these parenting domains as separate and non-overlapping. For example, some researchers only assess the parental warmth domain, such as adolescent perceptions of interactions that are nurturing and supportive (Russell, & Gordon, 2017); whereas others focus solely on the parental control domain, such as how parents use disciplinary practices to gain knowledge of their child’s activities (Wang, Stanton, Li, Cottrell, Deveaux, & Kaljee, 2013) and shape acceptable behavior (Barnes et al., 2006). Although some researchers have offered a typology to delineate how different patterns of parenting practices reflecting warmth and control can be combined to reflect overall parenting styles (e.g., Baumrind, 1991), few examine the potential overlap across these two domains outside of a binary framework. Results from this study indicate that both warmth and control items load on two of the three parenting quality factors (parental knowledge and affective relationships and parental communication and involvement), providing support for a more integrative framework for assessing parenting.

Parental Control

Items that loaded onto this factor were limited to aspects of control such as parents’ reactions to youth’s use of substances and parents’ beliefs about their efficacy and authority to discipline youth. Moreover, items that loaded on this factor were categorized as both general control items and substance-specific control parenting practices. Yet, the overall parenting literature has tended to focus only on general control practices, such as the use of disciplinary practices to gain knowledge of youth activities (Wang et al., 2013) and shape acceptable youth behavior (Barnes et al., 2006). There are several reasons why these general and substance-use specific control items loaded onto the same parental control factor. First, both types of parenting practices shared a similar focus on controlling and shaping youth behavior. Second, parents with higher levels of efficacy and belief in their legitimate authority may be more likely to set and enforce rules. Prior studies have shown that parental efficacy influences parental competence and can play an essential role in enhancing parenting disciplinary practices (Dumka, Gonzales, Wheeler, & Millsap, 2010). Additionally, youths’ belief in their parents’ authority over behaviors such as substance use may also be indicative of parents’ confidence in their efficacy, making them more likely to make rules and follow through on them. In a study by Jackson (2002), adolescents were more likely to legitimize parental authority regarding substance use issues than contemporary and conventional issues. Our findings support the importance of including both components of parental control (i.e., general and substance use-specific) in a comprehensive measure of parental quality.

Parental Knowledge and Affective Relationships

Unlike the parental control factor loadings, which reflected only aspects of control, parental knowledge and affective relationships’ factor loadings represented aspects of control and warmth. parental knowledge and affective relationships included items about how parents gain knowledge about youths’ activities and whereabouts, as well as items regarding the nature of the parent–youth relationship. Results indicate the quality of the parent–child relationship may have a stronger link with parental knowledge than other traditional control dimensions. Parental knowledge likely results from creating a warm and supportive environment where youth are more willing to disclose information to their parents voluntarily. This association between active parental tracking efforts and the quality of parent–youth relationships may explain why the parental monitoring/knowledge items loaded on parental knowledge and affective relationships instead of other disciplinary constructs under parental control. These findings are consistent with prior literature indicating that voluntary youth disclosure and parental knowledge may be facilitated in part by an open and trusting parent–youth relationship and strong emotional bonds (Fletcher, Steinberg, & Williams-Wheeler, 2004; Kerr et al., 2010).

Parental Communication and Involvement

Similar to parental knowledge and affective relationships, parental communication and involvement reflected control and warmth aspects of parenting. The loadings related to the control domain included items about how often parents solicit information about youths’ activities, their friends, and their friend’s parents. Loadings related to the warmth domain included items about the frequency of communication and type of activities parents are involved in with youth. Results of this study indicate that active parental solicitation may be more related to parental warmth than control, which could explain why parental solicitation did not load on the parental control factor. Research indicates that parents who effectively communicate and are involved in shared activities improve the quality of parent–youth relationships (Ackard, Neumark-Sztainer, Story, & Perry, 2006), making it more comfortable for parents to solicit information about their youth (Lippold, Greenberg, Graham, & Feinberg, 2014). Additionally, parental solicitation may reflect an interest in bonding with youth that is characterized by mutual communication, rather than a one-sided interrogation to gain information about youth behaviors and whereabouts.

Factor Correlations

Findings demonstrated a high correlation between parental knowledge and affective relationships and parental communication and involvement (r = .63, p < .001). The high correlation between these parenting factors is consistent with prior work indicating that an increase in shared activities provides opportunities for bonding and fostering mutual communication (Crosnoe & Trinitapoli, 2008). In fact, individual items on parental knowledge and affective relationships, focused on the results of parent–youth interactions, while items on parental communication and involvement, focused on the process of parent–youth interactions.

Several limitations and future directions should be drawn. First, our findings cannot be generalized to samples with different demographic characteristics, as the caregivers in this sample were primarily White (87%), with high levels of education and income. Prior work indicates that parenting practices may operate differently across racial and ethnic groups (Smith & Krohn,1995). In addition, previous work suggests that populations characterized by a high prevalence of single-parent households, high concentrations of economic disadvantage, and low educational attainment, may be characterized by lower positive parenting practices (e.g., Mrug & Windle, 2009). Thus, the reliability and validity of the PPM may differ when used with samples of different racial and ethnic backgrounds and lower socioeconomic status. Future studies including more diverse families are necessary. Second, only parent-report measures were examined, most of which (87%) were given by mothers. Research indicates that adolescents’ reports tend to be more strongly predictive of youth behaviors than parents’ reports and perhaps less biased (Kerr & Stattin, 2000). Future studies should utilize measures from multiple reporters. Third, this study did not examine the predictive validity of the measure. Future studies should examine the correlation between parenting factors and youth’s behavioral outcomes. Lastly, the proposed measure of parenting quality consists of 12 subscales, comprising 139 individual items, which may be difficult to administer. Thus, future work that includes an individual item analysis to determine if items could be dropped, thus shortening the measure, could ease administration.

The primary aim of the current study was to assess the psychometric properties of the PPM, a comprehensive measure of parenting practices. Findings support the utility of having a broad measure of parenting that integrates multiple aspects of parenting that are critical for understanding this complex socialization context, especially during adolescence. Broad measures of parenting practices that are currently available tend to not assess parental knowledge, child disclosure, parental solicitation, and substance specific parenting practices. Prior work indicates that these parenting factors have a notable impact on adolescent problem behavior and substance use (e.g., Kerr et al., 2010; Kodl & Mermelstein, 2004; Trucco et al., 2016). Thus, a contribution of this study is the incorporation of parenting practices that are critical for adaptive functioning among adolescents with broader dimensions of parenting that are relevant across a wider age span. Findings highlight the inherent overlap between parental control and demandingness constructs and underscore the importance for practitioners to consider this when working with families. The PPM can be used as an effective tool to operationalize these integrative domains (parental knowledge and affective relationships, parental control, and parental communication and involvement) that may be most relevant to parenting.

In addition, the PPM has significant utility in providing clinicians a psychometrically valid measure of parenting quality that can aid in identifying specific problem areas as well as parenting strengths. Clinicians can use results from the PPM to develop targeted treatment plans to improve problem areas and strengthen positive areas. In the treatment process, clinicians can use the PPM to track their clients’ change on each area, which serves as a tool for clinicians to evaluate the effectiveness of their treatment and make adjustments as needed. Important future directions are translation and validation across cultures. Translating the PPM to different languages is needed in order to reach families with different ethnic and cultural backgrounds, but validating whether these constructs and this measure are reliable across more diverse cultural groups is equally important. For example, prior research indicates that dimensions of parental control are culturally variable in terms of norms and feelings towards these parenting practices (Deater-Deckard et al., 2011). Researchers and practitioners should not assume equivalence in the meaning of these parenting practices across cultures.