Unidimensional versus multidimensional perspectives on father involvement.

The purpose of this study was to evaluate and compare the utility of unidimensional and multidimensional models of the construct of father involvement and to provide measurement advice for researchers. Participants were 1,139 two-parent families from the 1997 Child Development Supplement of the Panel Study of Income Dynamics (PSID-CDS). Results from Structural Equation Modeling procedures supported the conceptualization of paternal involvement as a multidimensional construct, consisting of distinct domains of involvement. In particular, paternal cognitive monitoring was not related to the other aspects of involvement included in the model. Findings are discussed in terms of implications for further development and use of measures of father involvement and for further study of paternal involvement using the PSID-CDS data set.

Keywords: father involvement, fathers, multidimensional construct

**********

Fathers, as active, involved parents, began to attract the attention of social scientists in the '70s (e.g., Lamb, 1975). This interest was in contrast to research on fathers conducted in the prior few decades that, for the most part, focused on the consequences for children of father absence (Biller, 1971). The refocused interest on fathers as active parents increased dramatically in the ensuing years with considerable attention given to father involvement in direct child-rearing activities (e.g., Hofferth, Pleck, Stueve, Bianchi, & Sayer, 2002; Larson, 1993; Marsiglio, 1991; McBride & Rane, 1997). In the past decade researchers were joined by policymakers and the popular media in a continued and explicit focus on fathers and fathering (Coltrane & Allan, 1994; Lamb, 1997; LaRossa, 1997; Tamis-LeMonda & Cabrera, 1999; Thomas, 1998).

Several factors have contributed to this increased interest in fathers and fatherhood, including changing societal conceptions of parental roles, increases in maternal employment, shifts in the demographic profile of modern families, and policy debates over the well-being of children as well as a growing body of literature outlining the impact of father involvement on child development (Lamb, 2004; Marsiglio, 1995). Accompanying this increased interest has been a rapidly expanding research literature on fathers and fatherhood, especially during the past decade (Cabrera, Tamis-LeMonda, Bradley, Hofferth, & Lamb, 2000; Day & Lamb, 2004; Lamb, 2004; Tamis-LeMonda & Cabrera, 2002).

As fathering research has flourished, one important issue has come to light: how should father involvement be conceptualized and measured? As discussed below, many researchers have made fundamental contributions to the study of father involvement by trying to clarify and, more recently, broaden conceptualizations of fathering (e.g., Day & Lamb, 2004; Doherty, Kouneski, & Erickson, 1998; Hawkins & Dollahite, 1997; Lamb, 2004; Lamb, Pleck, Charnov, & Levine, 1987; Palkovitz, 1997; Pleck & Masciadrelli, 2004). However, as conceptualizations of fathering have expanded to more fully include the many ways men can parent, ways of thinking about and measuring father involvement have been challenged. For example, do multiple aspects of fathering behavior yet reflect a unitary construct of "father involvement," or do these broader conceptualizations necessitate viewing father involvement as multidimensional, composed of separate facets of men's parenting? In parallel, can brief, comprehensive measures of the general construct of "father involvement" be created and used effectively, or is the best approach to measure different aspects of involvement separately, depending on the goals of a particular study? The aim of the current investigation was twofold: (a) to explore the construct of father involvement, with a focus on identifying the strengths and limitations of unidimensional versus multidimensional approaches to thinking about involvement and (b) to make measurement recommendations for fathering researchers based on the results of the current study.

CONCEPTUALIZATIONS OF FATHER INVOLVEMENT

One of the earliest, clear conceptualizations of father involvement to emerge was the Lamb et al. (1987) model. This model conceptualizes parental involvement as consisting of three distinct categories: (a) engagement (parent interacting with child in one-on-one activity such as playing, feeding, etc.), (b) accessibility (parent is physically and psychologically available to child), and (c) responsibility (parent assumes responsibility for welfare and care of child--e.g., making childcare arrangements, knowing when the child needs to go to the pediatrician, etc.).

Although the Lamb et al. (1987) conceptualization is quite useful and has served as the basis for much of the research on father involvement, it has not escaped criticism. Hawkins (Hawkins & Palkovitz, 1999; Hawkins et al., 2002), Palkovitz (1997), Daly (2001), and others have argued that research on fatherhood conducted during the past 30 years has conceptualized and operationalized paternal involvement as consisting of temporal and readily observable events. These scholars argue that an overdependence on the Lamb et al. (1987) conceptualization of father involvement has led to the limited, more unidimensional view of this construct that is prevalent in the research literature. In their critique of this "ticks and clicks" approach to conceptualizing and measuring father involvement, Hawkins and Palkovitz (1999) argue that paternal involvement is a multidimensional construct that includes affective, cognitive, and ethical components as well as the observable behavioral components more often studied in research on fatherhood.

In particular, Palkovitz (1997) called for an expansion and reconstruction of the concept of father involvement. He identified 15 major categories of parental involvement: communication (talking, listening, expressing love), teaching (role modeling, disciplining, encouraging interests and hobbies), monitoring (looking after schoolwork, knowing who child's friends are), thought processes (worrying, planning, praying), errands (driving child someplace, picking up needed items), caregiving (feeding, bathing, caring for a sick child), child-related maintenance (cooking, doing laundry, making repairs), shared interests (reading together, developing expertise), availability (attending/leading activities, spending time together), planning (birthdays, vacations, saving for the future), shared activities (shopping, playing together, working together), providing (housing, clothing, food, healthcare), affection (hugging, cuddling, tickling), protection (monitoring child's safety, providing safe home environment and activities), and supporting emotionality (encouraging the child, developing interests). Palkovitz's re-conceptualization of involvement emphasizes its multidimensionality and also highlights cognitive and affective aspects of parenting, which had not been central to previous conceptualizations of involvement such as the Lamb et al. (1987) model.

Others have also worked to expand our conceptualization of father involvement by examining fathering behavior from generative-fathering (Dollahite, Hawkins, & Brotherson, 1997), social-constructionist (Marsiglio, Amato, Day, & Lamb, 2000) and social-capital perspectives (Amato, 1998). In addition, there has been a recent call for increased attention to the quality of fathers' involvement with their children as well as the quantity of involvement within the father involvement literature (Cabrera et al., 2000; Hawkins & Dollahite, 1997). Clearly, there is no longer a single way of approaching and thinking about the study of father involvement.

Although the broadening of the involvement construct is timely, the expansion of this construct brings several challenges for researchers. Most notable among these is the challenge researchers face in developing reliable and valid measures of father involvement that tap into the multiple dimensions that constitute these more recent conceptualizations (Day & Lamb, 2004; Hawkins et al., 2002; Marsiglio, Day, & Lamb, 2000; Palkovitz, 2002; Parke, 2002; Roggman, Fitzgerald, Bradley, & Raikes, 2002). This challenge has been difficult to balance with the need for researchers to develop measures of father involvement of a manageable length and/or format so as not to be cumbersome and prohibitive for participants. Many researchers have addressed these challenges by attempting to develop relatively brief, global measures of father involvement that adequately reflect the complexity of this construct (e.g., Hawkins et al., 2002). As noted by Palkovitz (2002), though, these attempts have fallen short of capturing the full range of behaviors that reflect the diversity we know exists in how men approach their parenting roles.

A related and perhaps more central question involves whether "father involvement" can encompass all of the important domains recently identified by researchers and still be conceived of as a unitary construct. This question essentially concerns whether father involvement is a unidimensional of multidimensional construct. If father involvement is unidimensional, this means that involvement, as a general construct, is something that should be measured using a single score. (For example, one could measure a father's involvement and conclude overall that he is a more of less involved father.) Statistically speaking, as a unidimensional construct, involvement's different domains (e.g., interaction, monitoring, affection) can be summarized by a single unitary factor. This means that a one-factor analytic model should fit the data. If, on the other hand, the different domains of involvement cannot be represented by a single factor (i.e., a one-factor model fails to fit the data), it may suggest that father involvement is multidimensional, meaning that it consists of separate facets reflecting different aspects of men's parenting. In this case, a multiple (first-order) factor analytic model should fit the data.

If the factors in such a multiple-factor model are correlated, one could further explore the conceptual structure of involvement by examining the fit of a multidimensional model with multiple first-order factors representing different aspects of involvement and one second-order factor representing what the different dimensions of involvement have in common. If the inclusion of a second-order factor improves (of does not compromise) the fit of the model, separate aspects of father involvement could be said to possess both specific qualities and general commonalities. Practically, this outcome would imply that researchers could compute and interpret an overall involvement score as well as separate scores for different aspects of involvement. In contrast, if the inclusion of a second-order factor worsens the fit of the model, the separate aspects of involvement may be viewed as distinct. In this case, it would be inadvisable to measure aspects of involvement and conclude that a father is more or less involved--one could only conclude that a father is less involved in interaction and affection but more involved in monitoring, for example. Thus, the issue of measurement of father involvement necessitates a reexamination of conceptualizations of involvement, which in turn can provide recommendations for measurement. Interested readers can refer to McDonald (1999) for further discussion of unidimensional and multidimensional factor analytic models.

THE PRESENT STUDY

The purpose of the current study was to evaluate the utility of unidimensional versus multidimensional conceptualizations of father involvement in child-rearing and to provide advice to fathering researchers about the proper use of measures with differing dimensional structures. Consistent with the idea that father involvement is multidimensional, we predicted that different domains of involvement would represent multiple first-order factors. This prediction was based on recent conceptualizations of involvement as multidimensional (e.g., Palkovitz, 1997) as well as the results of Hawkins et al. (2002), who sought to develop a relatively brief inventory of father involvement that could capture its multiple aspects (e.g., behavioral, cognitive, affective, and moral/ethical dimensions of involvement as well as indirect and direct involvement). Using exploratory principal components analysis coupled with confirmatory factor analysis, Hawkins et al. identified nine factors representing different conceptually consistent dimensions of father involvement, thereby substantiating that involvement is multidimensional.

However, we also predicted that the best model to fit our data would not include a second-order "father involvement" factor, implying not only that involvement should be viewed as multidimensional, but also that the different aspects of involvement should be viewed as distinct and measured as such. This prediction follows from the recent broadening of conceptualizations of fathering (Hawkins & Dollahite, 1997; Palkovitz, 1997); by expanding our view to recognize the various ways in which fathers approach the parenting role, it becomes less likely that different dimensions of fathering will neatly cohere. Here our predictions diverge somewhat from their basis on Hawkins et al. (2002), who performed further analyses to test a higher-order model including a global father involvement factor. Hawkins et al. Concluded that this was the best fitting and most parsimonious model for their data, a conclusion that implies the interpretability of an overall involvement score as well as scores within particular domains of involvement.

Although we recognize the need for a simple paper-and-pencil assessment tool that provides a global measure of father involvement, we believe that such measures fall short of adequately capturing the complexity of paternal involvement and can obscure important distinctions between particular domains of involvement. Thus, the use of these measures as indicators of global involvement may generate false or misleading conclusions about the relationship of this construct with other phenomena that are of interest to researchers and practitioners. Drawing upon data from the Child Development Supplement of the Panel Study of Income Dynamics (Hofferth, 1998), an important study which has already been used to investigate paternal involvement (e.g., Hofferth, 2003; Hofferth & Anderson, 2003; McBride, Schoppe, Ho, & Rane, 2004) and will continue to be an important source of data for those who study fathering behavior, the analyses described below evaluate and compare the utility of unidimensional and multidimensional models of the father involvement construct. Based on our results, we are able to provide not only general measurement recommendations, but also specific recommendations for researchers who plan to use the PSID-CDS data set to investigate father involvement in the future.

METHOD

BACKGROUND AND PROCEDURE

Data for this investigation were obtained from the 1997 Child Development Supplement (CDS) of the Panel Study of Income Dynamics (PSID), conducted by the Survey Research Center of the Institute for Social Research at the University of Michigan (Hofferth, 1998). The Child Development Supplement (CDS) is a supplement to the core PSID data begun in 1997 with funding from the National Institute of Child Health and Human Development with the goal of providing researchers with a nationally representative database of children and their families. For the CDS, data from approximately 2,500 families (a subset of the larger PSID sample) were collected regarding 3,563 children aged 0-12 years (including up to two randomly selected children per family). The data provided by the CDS include a large body of rich information on children's family environments and well-being provided by the child's primary caregiver, other (secondary) caregiver and/or father outside of the home, teacher or childcare provider, school or childcare administrator, and the child him/herself (for those aged three years and older).

The CDS survey was conducted over the period from March 1997 to December 1997 (Hofferth & Anderson, 2003). PSID families with children in the designated age range were contacted regarding their interest in participating in the CDS. Ninety percent of the families in the core PSID sample contacted about the CDS agreed to participate. Eighty-four percent of families who were contacted for the first time in 1997 as part of an immigrant "refresher" sample of families who emigrated to the U.S. post-1968 agreed to participate (Hofferth & Anderson, 2003). The combined response rate for the CDS sample was 88%, and the combined sample was made nationally representative by using weights based on the 1997 Current Population Survey (Hofferth & Anderson, 2003).

PARTICIPANTS IN THE PRESENT STUDY

For the purposes of this study, we are focusing on the fathers or father-figures living in the homes of children in the CDS sample and the questionnaire measures they completed. This focus resulted in the use of a subsample of 1,139 children (569 boys, 570 girls; average age = 6.1 years) who were identified as living with a secondary caregiver who is the child's biological or adoptive father, stepfather, or father-figure. The families of children included in this sample had an average annual income of $56,173 (SD = $5,893). In addition, the current sample contains only children from the two largest ethnic groups represented in the CDS data--Caucasian (69%) and African-American (31%). These children's fathers completed two questionnaire instruments: the other caregiver-child questionnaire and the other caregiver-household questionnaire. The other caregiver-child questionnaire included items about the target child's well-being and behavior, in addition to items about parental affection and interaction with the target child. The other caregiver-household questionnaire included items about the division of household labor, parenting beliefs, gender-role beliefs, and the parent's mental health. For these instruments, the response rate was approximately 50% (Hofferth, Davis-Kean, Davis, & Finkelstein, 1998). Weights calculated by the CDS were used to adjust for the lower response rate by other caregivers to the survey instruments (Hofferth & Anderson, 2003; for more information, see the Description of the 1997 Child Supplement Weights (1999) at: http://psidonline.isr.umich.edu/CDS/weightsdoc.html).

VARIABLES REPRESENTING ASPECTS OF FATHER INVOLVEMENT

As described in Hofferth et al. (1998), items included in the CDS were designed, in part, to assess the social and psychological resources of families--including aspects of parenting such as warmth, disciplinary practices, monitoring/ supervision, and engagement. The items reflecting parenting were originally chosen for the CDS from a number of different large-scale studies (e.g., the National Longitudinal Survey of Youth). We examined the CDS database and selected items theorized to represent different important aspects of father involvement highlighted by previous investigators. In particular, we focused on items that reflected engagement, accessibility, and responsibility forms of involvement (Lamb et al., 1987) as well as other involvement domains such as communication, shared activities, monitoring, and affection (Palkovitz, 1997). For the purposes of this study, we focused on single-informant data (the questionnaires the fathers completed themselves) because these questionnaires contained the information most relevant to assessing involvement and because similar questions about fathers living in the home were not asked of mothers and children. We found sets of items tapping four important facets of paternal involvement: responsibility, affection/communication, participation in activities, and cognitive monitoring. The sets of items included under each conceptual aspect are detailed below, including the number of items constituting each set, item examples, and the sources of the original items.

Responsibility (Rel-Re3). The first set of items reflects responsibility forms of involvement and consists of three items assessing fathers' reports of which partner had responsibility (self vs. other or shared) for child-related household tasks (e.g., selecting the child's pediatrician and making appointments). These items were drawn from those used in the National Survey of Families and Households (Hofferth et al., 1998).

Affection/communication (Wal-Wa4). The second set of items reflects the amount of affection fathers expressed to their children and the amount of communication they had with them. Fathers reported on their own affection/communication using four items (e.g., "How often in the past month have you told the child that you love him/her?"). These items were rated on a five-point scale (1 = not in the last month; 5 = every day). This set of items was originally developed by Child Trends, Inc., for use in the JOBS study (Hofferth et al., 1998).

Participation in activities (Pal-Pal3). The next two sets of items reflect paternal participation in two categories of specific activities. Fathers reported on how often they had engaged in household-centered and child-centered activities with their children within the last month using 13 items. Six of these items reflected household-centered activities (e.g., "How often have you done dishes together?"), and seven of these items reflected child-centered activities (e.g., "How often have you worked on homework with him/her?"). All of these items were rated using a five-point scale (1 = not in the last month; 5 = every day). These items were used previously in the National Survey of Families and Households and the National Longitudinal Survey of Youth and were designed to tap parental interaction with and cognitive stimulation of their children (Hofferth et al., 1998).

Cognitive monitoring (Col-Co4). The last set of items represents paternal monitoring or the extent to which fathers are involved in setting limits and guidelines for their children's activities. Fathers reported on the extent to which they monitored their children using four items (e.g., "How often do you try to control which children your child(ren) spend(s) time with?"). These items were also rated using a five-point scale (1 = never; 5 = very often). The Detroit Area Study was the source of these items (Hofferth et al., 1998).

ANALYSIS PLAN

Our main goal for the analyses was to compare the fit of different models of father involvement. In order to analyze these data, we employed Confirmatory Factor Analysis (CFA) techniques. Maximum likelihood estimation was used to obtain the parameter estimates. The fit of each of the models was evaluated using the chi-square test statistic, root mean square of approximation (RMSEA; Steiger & Lind, 1980), and standardized root-mean-square residual (SRMR; Joreskog & Sorbom, 1993). Evaluating the fit of the models based on the chi-square measure is problematic for several reasons (i.e., chi-square is a test for exact fit, and it is sensitive to sample size and violations of normality; Loehlin, 1998). Thus, for these analyses, we relied more on a test of close fit based on the RMSEA, which measures discrepancy per degree of freedom and imposes a penalty for adding complexity to a model without substantially improving its fit. We have followed Browne & Cudeck's (1993) suggestion that RMSEA [less than or equal to] .05 indicates a close fit, RMSEA between .05 and .08 indicates a fair fit, and values in the range of .08 to. 10 indicate mediocre fit. RMSEAs larger than .10 indicate poor fit. We also used SRMR as an additional index of model fit, and a cutoff value of .08 for SRMR (SRMR < .08) is recommended for acceptable models. Moreover, we reported two additional goodness-of-fit indices to ease the interpretation of our model fitting results: the Tucker-Lewis index (TLI) and Comparative Fit index (CFI) (Loehlin, 1998; McDonald & Marsh, 1990). Larger values of these indices denote better fit.

RESULTS

PRELIMINARY ANALYSES

We first used CFA to check the dimensionality of each set of items chosen to represent a particular aspect of father involvement. Because the responsibility set only contained three items, tests of dimensionality using [chi square] or other statistics described above were not possible. However, Cronbach's alpha for this set of items was computed ([aplha] = .62). For the affection/ communication item set, CFA revealed that a 2-factor solution fit this set of items best: [chi square](1) = 10.44, p < .01; RMSEA = .092; SRMR = .008, with one factor containing two items focused on love and physical affection and the second factor containing two items focused on time spent talking with the child. Cronbach's alphas for these two factors were .81 and .74, respectively. CFA confirmed the division of items tapping paternal participation in activities into household-centered and child-centered sets: [chi square](64) = 435.64, p < .05; RMSEA = .083; SRMR = .044. Cronbach's alphas were .77 for the household-centered factor and .76 for the child-centered factor. For the set of items representing cognitive monitoring, CFA confirmed a one-factor structure: [chi square](2) = .61, p = .74; RMSEA = .00; SRMR = .004. The cognitive monitoring factor had an alpha of .68. Thus, checking each of the sets of items for dimensionality using CFA resulted in six (somewhat more differentiated) sets of items tapping various aspects of father involvement: responsibility, expressed love and physical affection, talking with the child, participation in household-centered activities, participation in child-centered activities, and cognitive monitoring.

MODEL FITTING

We first tested if father involvement is a unidimensional construct by fitting a single factor model with all the items loading on one factor simultaneously. This hypothesis was strongly rejected: [chi square] (252) = 3779.88, p <.01; RMSEA = .11; SRMR =. 11 (TLI = .549; CFI = .589). We then examined the multidimensional hypothesis for father involvement. A six first-order factor model was fitted. This analysis essentially tested how well the factors hung together. CFA was again employed: [chi square] (237) = 919.81, p < .05; RMSEA = .050; SRMR = .046 (TLI = .907; CFI = .920). The intercorrelations among the factors for this model are shown in Figure 1. The results from simultaneously fitting all of the sets of items indicate that the fit of this model can be considered close to fair (using RMSEA; the model also falls into the acceptable range for SRMR). Some lack of fit was primarily due to the cognitive monitoring factor, which was not significantly associated with the other factors (see Figure 1).

[FIGURE 1 OMITTED]

In determining whether father involvement can be considered an overarching construct for these sets of items, a second-order factor analysis model was further tested, which was conceptually equivalent to using CFA to conduct an additional factor analysis on the factor correlation matrix from the first-order model. By conducting this further analysis, a single second-order factor (which we might call something like father involvement) was created, and then the factor loadings of the factors created in the first-order model on the second-order factor were examined. The fit of the second-order factor model was actually slightly worse than the fit of the first-order model: [chi square](246) = 1108.80, p < .05; RMSEA = .055; SRMR = .056 (TLI = .887; CFI = .899). The RMSEA for this model indicates a fair fit, and although the SRMR is still acceptable, it is higher for this model than for the first-order model, indicating a poorer fit. Furthermore, the difference in fit between the first- and second-order models was statistically significant: [[chi square].sub.diff](9) = 188.99, p < .01. The standardized factor loadings for the individual items constituting the factors and the second-order factor loadings of the factors on the second-order father involvement factor are shown in Figure 2.

[FIGURE 2 OMITTED]

Generally, the results favor the first-order model, suggesting that father involvement, at least as represented in this data set, is multidimensional. Furthermore, the poorer fit of the second-order model suggests that a global father involvement construct representing commonalities among the distinct domains of involvement is not supported. It is important to note that, similar to the results of the fitting of the first-order model, the cognitive monitoring factor was the only factor not to have a significant factor loading on the second-order father involvement factor. This point and additional implications of the results are addressed below.

DISCUSSION

Overall, the results of this investigation support the conceptualization and measurement of paternal involvement as a multidimensional construct characterized by distinct facets of involvement. Although the sets of items we chose to represent different important aspects of father involvement were intercorrelated and the first-order model proved to be a reasonable fit for the data, the second model (which included a higher-order factor to represent father involvement) exhibited a poorer fit. Our results are both consistent with and divergent from those of Hawkins et al. (2002), who concluded that father involvement is a multidimensional construct, but in contrast to our results found evidence for commonalities among the different domains of involvement they measured as represented by a global involvement factor.

It appears that the factor representing cognitive monitoring, which was not associated with the other factors in the first-order model and did not load significantly on the second-order father involvement factor, did not fit as well with the other aspects of involvement in both models. This monitoring factor may have been most responsible for the poorer fit of the higher-order model. Cognitive aspects of involvement have only been recently addressed in the father involvement literature, in an effort to expand the conceptualization of involvement beyond the Lamb et al. (1987) more time-based model (McBride et al., 2004; Palkovitz, 1997). Cognitive involvement reflects the presence of children in their fathers' thoughts and includes such processes as reasoning, planning, evaluating, and monitoring (Palkovitz, 1997). Interestingly, Hawkins et al. (2002) acknowledge that cognitive aspects of involvement are not prominent in their Inventory of Father Involvement. Perhaps if more items assessing paternal cognitions had been included, the fit of their higher-order model would have been compromised, leading Hawkins et al. to conclude that the first-order model was a better fit for their data. However, it is also important to acknowledge that although the set of items used to represent cognitive monitoring in this study achieved adequate reliability, the operationalization of monitoring in this study may have been weak. Paternal monitoring has become a focus of fathering researchers more recently, and thus the CDS items (which were developed earlier) may not have been ideally suited to measuring monitoring. Despite this possibility, we suggest that the best course of action for future researchers may be to measure cognitive aspects of involvement using instruments specifically designed to tap paternal cognitions.

Our findings have important implications for the measurement of father involvement. Because our results indicate that involvement consists of distinct domains, we suggest that the best way to measure different aspects of father involvement may be with different instruments. Although a brief questionnaire such as the Inventory of Father Involvement (IFI; Hawkins et al., 2002) has its advantages, given our findings we are not convinced that a global father involvement construct or a unitary way of measuring such a construct actually exists. We do concur with Hawkins et al. that it is helpful to examine the strengths and weaknesses of individual fathers. Thus, there are certain conditions under which such an approach and a brief, yet comprehensive questionnaire such as the IFI is highly useful (e.g., for practitioners who work with individual fathers and families, are interested in father involvement as one important aspect of the family system, and likely don't have the time to administer separate involvement instruments). Also, for researchers interested in how fathers allocate their resources to different aspects of involvement or how fathers structure their identities as fathers (Ihinger-Tallman, Pasley, & Buehler, 1995; Minton & Pasley, 1996), the examination of different aspects of involvement within the same instrument can provide an efficient and clear comparison of different types of fathering behaviors.

However, for researchers more interested in antecedents of father involvement or consequences of involvement for children and families (e.g., Amato, 1998; McBride, Schoppe, & Rane, 2002), all facets of father involvement may not be of equal interest. For such researchers, we fear that the use of a brief, comprehensive questionnaire (in an attempt to capture too many aspects of involvement under the umbrella of "father involvement") may be alluring, but could, in fact, lead to results that are difficult to interpret and may actually further obscure our understanding of fathering behavior. Moreover, for some practitioners whose work centers on fathers, a focus on specific aspects of involvement may be more fruitful. Thus, we suggest an approach where researchers and practitioners first decide which aspects of involvement are important for their purposes and then take action to collect information about those particular aspects of involvement from multiple sources.

The current results also hold more specific implications for the further study of father involvement using the PSID-CDS data set, an already rich source of information on children and their families that continues to grow in breadth and depth with the recent collection of follow-up longitudinal data on participating children and families (for more information, see What's New with CDS? [2003] at: http://psidonline.isr.umich.edu/CDS/whatsnew.html). A number of investigations have already used the PSID-CDS data set to address questions related to father involvement (e.g., Hofferth, 2003; Hofferth & Anderson, 2003; McBride et al., 2004; Yeung, Sandberg, Davis-Kean, & Hofferth, 2001), and surely more will follow. The CDS contains relatively rich data with respect to different aspects of involvement, including the information used in this investigation on paternal responsibility, participation in activities, affection, communication, and cognitive monitoring as well as information about actual time spent with children, which has been included in other studies (Hofferth, 2003; Hofferth & Anderson, 2003; Yeung et al., 2001). Thus, the CDS allows researchers the unique opportunity to investigate multiple facets of fathering behavior using a large, nationally representative data set. We hope that in the current study we have highlighted some of the ways to measure fathering behaviors within the CDS and that future researchers can use our investigation as a useful, albeit incomplete, guide to studying paternal involvement using the CDS. With consistent use of the CDS data to investigate questions relevant to paternal involvement, greater progress can be made to understand involvement and its antecedents and consequences. Above all, based on the findings of the current study, we implore investigators to exercise caution in studying multiple aspects of involvement using the CDS data and not "lump together" diverse involvement behaviors under the heading "father involvement."

Although we have made recommendations based on the current results, it is also important to acknowledge some of the limitations of our sample and approach. Because of the unavoidable limits of a large study like the CDS, we were notable to include all of the ways fathers can be involved in their children's lives in our conceptualization of paternal involvement. For example, we weren't able to include information about fathers' self-identification as financial providers for their children and families. Items tapping the "provider" role are included in Hawkins et al.'s (2002) Inventory of Father Involvement, but it is interesting to note that these items were not as strongly associated with their global father involvement factor as were items tapping other aspects of father involvement. We suspect that including providing in our analysis would have only worsened the fit of the higher-order model. Also, we were notable to use the CDS time-diary information in the current study because of large amounts of missing data. Thus, we were unable to include detailed, time-based measures of engagement and accessibility, important aspects of the influential Lamb et al. (1987) conceptualization, in our model of father involvement. However, we were able to include fathers' reports of time spent in child- and household-centered activities with their children as one way of tapping fathers' more direct interactions with their children. Our understanding of the structure of the father-involvement construct would benefit from the inclusion of provider information, specific time-based interaction information, and other relevant aspects of father involvement (e.g., supporting the child's mother) in future studies.

In addition, we must also acknowledge that we didn't examine the structure of involvement for nonresidential fathers in this study. The essential aspects of involvement and the construct's structure are likely different for this population (Hawkins et al., 2002). Also, to assess father involvement in this study, we relied on fathers' reports of their own cognitions and behaviors. A more accurate picture of father involvement would be achieved by using multiple reports from a variety of sources (e.g., the child's mother and the child him/herself), which was not consistently possible in the CDS data set. Furthermore, some of the factors in our model were indicated using fewer items than would have been ideal (e.g., responsibility). It is our hope that future research will be able to address these concerns.

Taken together, our results and those of others (e.g., Hawkins et al., 2002) highlight the multidimensional nature of father involvement and emphasize the need to balance the attraction of brief, yet comprehensive measures of father involvement with the need for more directed, intensive inquiries into particular aspects of fathering behavior using more specific measures. We hope that our work continues to encourage researchers to move beyond singular conceptions of involvement while at the same time prioritizing focus for researchers and practitioners. Both approaches are necessary and useful, and together they will lead us to a better understanding of the work that fathers do.

Subscription rates (current volume):

                         Individual

Print                     $50.00          (USA)
                          $65.00       (International)

Electronic                $30.00       (All sites)

Print & electronic        $70.00          (USA)
                          $85.00       (International)

                       Institution

Print                    $150.00
                         $165.00

Electronic                $90.00

Print & electronic       $180.00
                         $195.00

Send name, address, and payment to:

Fathering, P.O. Box 32, Harriman, IN 37748-0032 USA
www.mensstudies.com / [phone], Fax [phone]

REFERENCES

Amato, P. (1998). More than money? Men's contributions to their children's lives. In A. Booth & A.C. Crouter (Eds.), Men in families: When do they get involved? What difference does it make? (pp. 241-278). Mahwah, NJ: Erlbaum.

Biller, H.B. (1971). Father, child and sex role. Lexington, MA: Heath.

Browne, M.W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K.A. Kollen & J.S. Long (Eds.), Testing Structural Equation Models (pp. 136- 162). Thousand Oaks, CA: Sage.

Cabrera, N.J., Tamis-LeMonda, C.S., Bradley, R.H., Hofferth, S., & Lamb, M.E. (2000). Fatherhood in the twenty-first century. Child Development, 71, 127-136.

Coltrane. S., & Allan, K. (1994). "New" fathers and old stereotypes: Representations of masculinity in 1980's television advertising. Masculinities, 2, 1-25.

Daly, K. (2001). Minding the time in family experience: Emerging perspectives and issues. Oxford, UK: Elsevier Science.

Day, R., & Lamb, M.E. (Eds.) (2004). Conceptualizing and measuring father involvement. Mahwah, NJ: Erlbaum.

Description of the 1997 PSID child supplement weights. (1999). Retrieved September 17, 2003 from PSID-CDS Web site: http://psidonline.isr.umich.edu/CDS/ weightsdoc.html

Doherty. W.J., Kouneski, E.F., & Erickson, M.F. (1998). Responsible fathering: An overview and conceptual framework. Journal of Marriage and the Family, 60, 277-292.

Dollahite. D.C., Hawkins, A.J., & Brotherson, S.E. (1997). Fatherwork: A conceptual ethic of fathering as generative work. In A.J. Hawkins & D.C. Dollahite (Eds.), Generative fathering : Beyond deficit perspectives (pp. 200-216). Thousand Oaks, CA: Sage.

Hawkins, A.J., Bradford, K.P., Palkovitz, R., Christiansen, S.L., Day, R.D., & Call, V.R.A. (2002). The inventory of father involvement: A pilot study of a new measure of father involvement. The Journal of Men's Studies, 10, 183-196.

Hawkins, A.J., & Dollahite, D.C. (1997). Generative fathering: Beyond deficit perspectives. Thousand Oaks, CA: Sage.

Hawkins, A.J., & Palkovitz, R. (1999). Beyond ticks and clicks: The need for more diverse and broader conceptualizations and measures of father involvement. The Journal of Men's Studies, 8, 11-32.

Hofferth, S.L. (1998). Report on 1997 data collection for the PSID Child Development Supplement. Retrieved April 25, 1999 from PSID-CDS Web site: http://www.isr.umich.edu.src/child-development/home.html

Hofferth, S.L. (2003). Race/ethnic differences in father involvement in two-parent families: Culture, context, or economy? Journal of Family Issues, 24, 185-216.

Hofferth, S.L., & Anderson, K.G. (2003). Are all dads equal? Biology versus marriage as a basis for paternal investment. Journal of Marriage and Family, 65, 213-232.

Hofferth, S.L., Davis-Kean, P.E., Davis, J., & Finkelstein, J. (1998). User guide for the Child Development Supplement of the Panel Study of Income Dynamics. Retrieved April 25, 1999 from PSID-CDS Web site: http://www.isr.umich.edu.src/child-development/home.html.

Hofferth, S.L., Pleck, J.H., Stueve, J.L., Bianchi, S., & Sayer, L. (2002). The demography of fathers: What fathers do. In C. Tamis-LeMonda & N. Cabrera (Eds.), The handbook of father involvement: Multidisciplinary perspectives (pp. 63-90). Mahwah, NJ: Erlbaum.

Ihinger-Tallman, M., Pasley, K., & Buehler, C.L. (1995). Developing a middle-range theory of father involvement postdivorce. In W. Marsiglio (Ed.), Fatherhood: Contemporary theory, research and social policy (pp. 57-77). Thousand Oaks, CA: Sage.

Joreskog, K.G., & Sorbom, D. (1993). LISREL 8: Structural equation modeling with the SIMPLIS command language. Hillsdale, NJ: Erlbaum.

Lamb, M.E. (1975). Fathers: Forgotten contributors to child development. Human Development, 18, 254-266.

Lamb, M.E. (Ed.) (1997). The role of the father in child development (3rd ed.). New York: Wiley.

Lamb, M.E. (Ed.) (2004). The role of the father in child development (4th ed.). New York: Wiley.

Lamb, M.E., Pleck, J.H., Charnov, E.L., & Levine. J.A. (1987). A biosocial perspective on paternal behavior and involvement. In J. Lancaster, J. Altmann, A. Rossi, & L. Sherrod (Eds.), Parenting across the lifespan: Biosocial dimensions (pp. 111-142). New York: Aldine de Gruyter.

LaRossa, R. (1997). The modernization of fatherhood: A social and political history. Chicago: University of Chicago Press.

Larson, R.W. (1993). Finding time for fatherhood: The emotional ecology of adolescent-father interactions. New Directions for Child Development, 62, 7-25.

Loehlin, J.C. (1998). Latent variable models: An introduction to factor, path and structural analysis (3rd ed.). Hillsdale, NJ: Erlbaum.

Marsiglio, W. (1991). Paternal engagement activities with minor children. Journal of Marriage and the Family, 53, 973-986.

Marsiglio, W. (Ed.). (1995). Fatherhood: Contemporary theory, research, and social policy. Thousand Oaks, CA: Sage.

Marsiglio, W., Amato, P., Day, R., & Lamb, M.E. (2000). Scholarship on fatherhood in the 1990s and beyond: Past impressions, future prospects. Journal of Marriage and the Family, 62, 1173-1191.

Marsiglio, W., Day, R.D., & Lamb, M.E. (2000). Exploring fatherhood diversity: Implications for conceptualizing father involvement. Marriage and Family Review, 29, 269-293.

McBride, B.A., & Rane, T.R. (1997). Role identity, role investments, and paternal involvement: Implications for parenting programs for men. Early Childhood Research Quarterly, 12, 173-197.

McBride, B.A., Schoppe, S.J., Ho, M., & Rane, T.R. (2004). Multiple determinants of father involvement: An exploratory analysis using the PSID-CDS data set. In R.D. Day & M.E. Lamb (Eds.), Conceptualizing and measuring father involvement (pp. 321-340). Mahwah, NJ: Erlbaum.

McBride, B.A., Schoppe, S.J., & Rane, T.R. (2002). Child characteristics, parenting stress, and parental involvement: Fathers versus mothers. Journal of Marriage and Family, 64, 998-1011.

McDonald, R.P. (1999). Test theory: A unified treatment. Hillsdale, NJ: Erlbaum.

McDonald, R.P., & Marsh, H.W. (1990). Choosing a multivariate model: Noncentrality and goodness-of-fit. Psychological Bulletin, 107, 247-255.

Minton, C., & Pasley, K. (1996). Fathers' parenting role identity and father involvement: A comparison of nondivorced and divorced, nonresident fathers. Journal of Family Issues, 17, 26-45.

Palkovitz, R. (1997). Reconstructing "involvement": Expanding conceptualizations of men's caring in contemporary families. In A.J. Hawkins & D.C. Dollahite (Eds.), Generative fathering: Beyond deficit perspectives (pp. 200-216). Thousand Oaks. CA: Sage.

Palkovitz, R. (2002). Involved fathering and child development: Advancing our understanding of good fathering. In C.S. Tamis-LeMonda & N. Cabrera (Eds.), Handbook of father involvement: Multidisciplinary perspectives (pp. 119-140). Mahwah, NJ: Erlbaum.

Parke, R.D. (2002). Fathers and families. In M. Bornstein (Ed.), Handbook of parenting: Vol. 3. Being and becoming a parent (2nd ed., pp. 27-73). Mahwah, NJ: Erlbaum.

Pleck, J.H., & Masciadrelli, B. (2004). Paternal involvement by U.S. residential fathers: Levels, sources, and consequences. In M.E. Lamb (Ed.), The role of the father in child development (4th ed., pp. 222-271). New York: Wiley.

Roggman, L.A., Fitzgerald, H.E., Bradley, R.H., & Raikes, H. (2002). Methodological, measurement, and design issues in studying fathers: An interdisciplinary perspective. In C. Tamis-LeMonda & N. Cabrera (Eds.), Handbook of father involvement: Multidisciplinary Perspectives (pp. 1-30). Mahwah, NJ: Lawrence Erlbaum.

Steiger, J.H., & Lind, J. (1980). Statistically based tests for the number of common factors. Paper presented at the annual meeting of the Psychometric Society, Iowa City, IA.

Tamis-LeMonda, C., & Cabrera, N. (1999). Perspectives on father involvement: Research and policy. Social Policy Report: Society For Research in Child Development, 13, 1-26.

Tamis-LeMonda, C., & Cabrera, N. (Eds.)(2002). Handbook of father involvement: Multidisciplinary perspectives. Mahwah, NJ: Lawrence Erlbaum.

Thomas, N.G. (Ed.). (1998). U.S. policy initiative on fathering. Social Policy Report: Society For Research in Child Development, 12, 22-23.

What's new with CDS? (2003). Retrieved September 17, 2003 from PSID-CDS Web site: http://psidonline.isr.umich.edu/CDS/whatsnew.html

Yeung, W.J., Sandberg, J.F., Davis-Kean, P.E., & Hofferth, S.L. (2001). Children's time with fathers in intact families. Journal of Marriage and the Family, 63, 136-154.

This research was supported in part by a grant to the second author from the American Educational Research Association, which receives funds for its "AERA Grants Program" from the National Science Foundation and the National Center for Education Statistics and Office of Educational Research and Improvement (U.S. Department of Education) under NSF Grant #RED-9980573. Opinions reflect those of the authors and do not necessarily reflect those of the granting agencies.

Correspondence concerning this article should be addressed to Brent A. McBride, Department of Human and Community Development, University of Illinois at Urbana-Campaign, 1105 W. Nevada, Urbana, IL 61801. Electronic mail: [e-mail].