Mothers' Employment, Parental Involvement, and the Implications for Intermediate Child Outcomes

Mothers' Employment, Parental Involvement, and the Implications for Intermediate Child Outcomes

Social Science Research 30, 25– 49 (2001) doi:10.1006/ssre.2000.0685, available online at http://www.idealibrary.com on Mothers’ Employment, Parental...

83KB Sizes 0 Downloads 16 Views

Social Science Research 30, 25– 49 (2001) doi:10.1006/ssre.2000.0685, available online at http://www.idealibrary.com on

Mothers’ Employment, Parental Involvement, and the Implications for Intermediate Child Outcomes Cathleen D. Zick University of Utah

W. Keith Bryant Cornell University

and ¨ sterbacka Eva O Åbo Akademi University, Turku/Åbo, Finland Data from the National Survey of Families and Households are used to investigate how married mothers’ work patterns affect the frequency of potentially human capital enriching parent– child activities and, in turn, if these parent– child activities and work patterns are related to children’s subsequent behavior and academic achievements. The analyses suggest that both parents in employed-mother households engage in reading/homework activities with their children more frequently than do parents in households where the mother is not employed. Increases in the frequency of reading/homework activities and playing/project activities are found to be related to fewer behavioral problems and higher grades. At the same time, the direct effect of a mother’s employment during the preschool years generally has no effect on intermediate child outcomes. © 2001 Academic Press Key Words: mothers’ employment; parent– child time; parental involvement; child outcomes.

As married mothers’ labor force participation rates have steadily climbed over the past 30 years, questions have arisen regarding what this trend has meant for This research was supported in part by a University of Utah Faculty Research Grant. David Huth and Nancy McLeod provided valuable programming assistance in the early stages of this project. Address correspondence and reprint requests to Cathleen D. Zick, 225 S. 1400 E., Rm 228, Department of Family and Consumer Studies, University of Utah, Salt Lake City, UT 84112. E-mail: [email protected]. 25 0049-089X/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved.

26

¨ STERBACKA ZICK, BRYANT, AND O

the welfare of children. One frequently expressed concern in the popular literature is that children with employed mothers may spend less time with both parents compared to children who are reared in a family where the father works outside of the home and the mother is a full-time homemaker (Hewlett, 1991; Hochschild, 1989). This supposition has been countered by the argument that increases in mothers’ labor force participation rates may lead fathers to substitute for mothers in some child care activities (see Pleck, 1996, for a review of this literature). Some also claim that while employed mothers and fathers may both spend less time with their children, they need not be sacrificing “quality” time (Albert and Popkin, 1987; Bryant and Zick, 1996b). Moreover, while mothers’ employment may affect parent– child time, there has been little research on the question of how such a shift affects children’s well-being, if at all. In this article, we use nationally representative data to examine how married mothers’ work patterns affect the frequency of shared parent– child activities and, in turn, if these parent– child activities and work patterns are related to children’s behavior and academic achievements. THE LITERATURE Time diary studies are typically used to investigate how mothers’ employment influences the amount of time both mothers and fathers spend in the physical care (e.g., bathing a young child) and nonphysical care (e.g., helping a child work through a problem) of children. Within this literature, family sociologists and home economists argue that the added time commitments created by the mothers’ employment are generally met by reducing the time she spends in traditional activities, such as child care. Families may choose to deal with this reduction by shifting some child care to the father, purchasing market substitutes, or simply leaving some tasks undone (Gershuny and Robinson, 1988; Nock and Kingston, 1988; Sanik, 1981). In contrast, economists posit that the amount of time mothers spend caring for children is partially a function of the opportunity costs of her time. If employed mothers have higher opportunity costs than otherwise comparable nonemployed mothers, than they will spend less time doing all types of household activities, including child care (Kooreman and Kapteyn, 1987). In such economic models, the father is predicted to increase his involvement with children when the mother’s opportunity costs rise only if this time is seen as a substitute for her time spent with children. The empirical literature on mothers’ employment and parental time spent in physical and nonphysical care of children paints a fairly consistent picture in the case of mothers. The majority of the studies find a negative relationship between the mother’s employment and/or her opportunity cost of time and the time she spends caring for children (Baydar, Greek, and Gritz, 1999; Bryant and Zick, 1996a; Coverman and Sheley, 1986; Gershuny and Robinson, 1988; Hill and Stafford, 1985; Kooreman and Kapteyn, 1987; Maret and Finlay, 1984; Medrich et al., 1982; Robinson, 1977; Sanik, 1981). In contrast, the findings with respect

INTERMEDIATE CHILD OUTCOMES

27

to the impact of maternal employment on fathers’ child care time are mixed. Two empirical studies find a positive relationship (Ishii-Kuntz and Coltrane, 1992; O’Connell, 1993), one finds a negative relationship (Barnett, 1987), and the majority find no statistically significant relationship (Bryant and Zick, 1996a; Cooney et al., 1993; Coverman and Sheley, 1986; Marsiglio, 1991). Taken together, the evidence suggests that as mothers increase their labor supply, they reduce the time they spend in physical and nonphysical care of children, and for the most part, fathers do not increase their time in a compensatory fashion. What are the implications of these shifts in parental time allocation for the welfare of children? On this point, the literature is much less forthcoming for two reasons. First, the measure of parental child care that is typically used— diary estimates of a parent’s time spent in physical and nonphysical child care—may be too gross to provide meaningful predictions about what such time means for children’s well-being. Lamb, Pleck, Charnov, and Levine (1987) argue that there are three aspects of parental involvement that may be important for child development: interaction, accessibility, and responsibility. In this context, diary reports of time spent in physical and nonphysical care may collapse time that is high on one of these dimensions with time that is low on the same dimension. Second, most studies stop short of directly testing the relationship between maternal employment, parental time spent with children, and child outcomes. In part, this is due to the fact that most time diary data sets contain little if any information on child outcomes. The few studies that have examined the relationship between mothers’ employment, specific types of parent– child time, and child outcomes focus exclusively on the role of mother– child time and, when taken together, provide an inconclusive picture. Stafford (1987) presents evidence that the more hours a mother works, the worse a child does in school and that, holding employment hours constant, increases in the mothers’ time spent in educationally related child care activities (e.g., helping, teaching, talking, and reading to a child) do little to offset the negative impact of employment. In contrast, while Moorehouse (1991) finds a negative effect of maternal employment on a child’s schooling outcomes, this effect is somewhat mitigated if the mother frequently engages in various child-oriented activities. Also holding employment status constant, Datcher-Loury (1988) finds that the more hours a highly educated mother spends with her child, the higher the child’s educational attainment. A number of studies have used mother’s employment as a proxy measure of time spent with children (i.e., they have capitalized on the finding that employed mothers typically spend less time in child care than nonemployed mothers) and examined its relationship to children’s academic outcomes. The findings of these studies are generally mixed with some concluding that children of employed mothers have lower levels of educational attainment compared to children of nonemployed mothers (Fleisher, 1977; Hill and Duncan, 1987; Milne et al., 1986), while others find no significant relationship between mother’s employment and a child’s scholastic achievement (Alwin and Thornton, 1984; D’Amico,

28

¨ STERBACKA ZICK, BRYANT, AND O

Haurin, and Mott, 1983; Datcher-Loury, 1988; Haveman and Wolfe, 1994; Leibowitz, 1977; Murnane et al., 1981; Sewell, Hauser, and Wolf, 1980). Parcel and Menaghan (1994) and Desai, Chase-Lansdale, and Michael (1989) examine the impact of parental employment on more short-term behavioral and cognitive child outcomes and conclude that in some instances employment appears to matter and at other times it does not. Research within the developmental literature on the impact of father’s involvement on child outcomes (reviewed in Pleck, 1996) provides convincing evidence that higher levels of father involvement (measured in a variety of ways) have numerous positive consequences for children. As an example, Amato and Rivera (1999) find that higher paternal involvement—measured by father– child closeness, father’s support, and father’s frequency of engaging in child-related activities—is associated with fewer behavioral problems in children. Unfortunately, none of these studies of father involvement explore the link between mothers’ employment, paternal and maternal involvement, and child outcomes as we do here. MATERIALS AND METHODS The Framework Our empirical work is grounded in a household production framework (Becker, 1965, 1991). In the context of this framework, parents gain satisfaction from raising happy, healthy, well-behaved children. With this goal in mind, they allocate their time and money resources among various household activities subject to a series of income, time, and technology constraints. For example, a father may spend time helping a child with a computer game. The amount of time he spends helping the child depends on the opportunity cost of his time, the opportunity cost of his spouse’s time (an alternative source of help), the price of the computer software used by the child, the type of computer the family owns, and the relative value the family places on the child acquiring computer skills versus other skills. More generally, demand functions can be written for the time each household member spends in each activity including the times fathers (F c) and mothers (M c) spend in child-related activities. Such time is hypothesized to be a function of the price of the father’s time, the prices of goods and services used in the activity, the prices of time for a family member who might substitute for the individual in question (e.g., the mother), the family’s nonwage income (V), the determinants of the underlying technology pertinent to each activity (Z g), and the family’s preferences (Z p). In the context of the current formulation, time spent with children and time spent in market work are simultaneously determined at any point, implying that if parent– child time is estimated as a function of the mother’s (and father’s) actual employment, all of the coefficients will contain simultaneous equations bias. To avoid such bias, two modifications are made to the model. In the case of the fathers, we restrict our sample to those families where the father is

29

INTERMEDIATE CHILD OUTCOMES

employed full-time and we then assume that for any given household the father’s hours of market work are fixed. That is, the family treats the father’s work hours as a given (although the father’s work hours may still vary across families). This assumption is supported by the findings in a number of studies that show that males’ hours of market work are much less sensitive to changes in exogenous factors than are their labor force participation decisions (Heckman, 1993). Thus, a father’s work influences both parents’ involvement in child-related activities only through household income (V). In the case of the mothers, we use an instrumental variables technique (Kmenta, 1986) where the mother’s employment status (M 1 ) is estimated as a function of factors both inside (Z f) (e.g., the mother’s education and prior work experience) and outside (Z o) of the immediate family environment (e.g., rural/ urban location). Then the estimated parameters from this equation are used to generate predicted employment status for the mother (i.e., the instrument), which is entered as an independent variable in the parent– child involvement equations in lieu of the mother’s value of time. In the cross section, it is also assumed that the price of goods is a constant. The family’s production of children’s human capital at a point in time, t, (C t ) (typically measured in the literature by one or more indicators including developmental benchmarks, aptitude, behavioral problems, and/or scholastic achievement) is posited to be a function of the cumulative time the mother and the father have spent with the child to date, the child-related goods and services (G) that have been purchased, nonwage income, and a set of predetermined variables that influence the parents’ preferences and technical ability to produce child-related human capital (e.g., their education levels). Mathematically, the model we estimate has the following general form: F c ⫽ f共M 1 , V, Z g, Z p兲,

(1)

M c ⫽ m共M 1 , V, Z g, Z p兲,

(2)

M 1 ⫽ m 1 共Z f, Z o兲,

(3)

and

冘 冘 t

C t ⫽ c共

t

F ci ,

i⫽0

i⫽0

冘 G , V , Z , Z 兲. t

M ci ,

i

t

gt

pt

(4)

i⫽0

Data and Measures Data to test our model come from waves 1 and 2 of the National Survey of Families and Households (NSFH) (National Survey of Families and Households Documentation, 1994). A sample of 13,008 households were interviewed in wave 1 of the NSFH during 1987 and 1988. In 1992–1994, 10,008 of the original 13,008 households were reinterviewed in wave 2. The wave 1 survey asked extensive questions regarding various aspects of individual and family well-

30

¨ STERBACKA ZICK, BRYANT, AND O

being. For the 5,677 households where one or more minor children were present, one child was randomly selected and the main respondent was asked a battery of questions about this “focal child.” Questions about parental involvement were asked for the 3,771 households where the focal child was age 11 or younger. Our sample is further reduced because we restrict our analysis to households where (1) there was a husband and wife in 1987–1988 who both participated in the NSFH (N ⫽ 2155), (2) the father reported that he worked 35 or more h per week (N ⫽ 1597), (3) a reinterview took place in 1992–1994 and both parents were still married to each other at that time (N ⫽ 1088), (4) the selected focal child is the biological child of both parents (N ⫽ 1002), and (5) both parents are non-Hispanic Caucasians (N ⫽ 811). These first three restrictions are mandated in order to test the model we have described. The latter two restrictions are made so that we do not have to estimate different parameters by race/ethnicity and stepparenting status. Our final sample size is in keeping with other analyses that have used the NSFH parental involvement measures (see for example Amato and Rivera, 1999, and Marsiglio, 1991). In wave 1 of the NSFH, an individual age 18 or older in the household was selected to be the primary respondent. An extended in-person interview was done with this primary respondent and in addition, s/he filled out a self-enumerated questionnaire. If the primary respondent was married, then his/her spouse completed a “secondary” respondent self-enumerated questionnaire in wave 1. In some instances, the questions asked of the primary respondent and the secondary respondent are identical; in others, they are not. The lack of comparability across the primary and secondary respondents interviews in wave 1 coupled with the fact that a primary respondent can be either the mother or the father create some variable construction challenges for the current investigation that are discussed below. In wave 2, the primary and secondary respondents are asked identical sets of questions in the survey, making incomparability across fathers and mothers a nonissue. Measuring parental involvement. Measurement of parental involvement is both conceptually and operationally a difficult task. Ideally, a researcher should be measuring those parental activities that have a strong potential of enhancing a child’s human capital (i.e., what the popular press refers to as “quality time”). Findings from the child development literature suggest that activities that involve intensive parent– child interaction (reading a book to a young child), activities that signal a parent’s accessibility (e.g., parental supervision of siblings’ play), and activities that reflect a heightened sense of responsibility for the child (e.g., arranging for work-related child care) should all be included as potentially human capital enhancing for the child (Blair and Hardesty, 1994; Lamb et al., 1987; Pleck, 1996). While the NSFH does not contain detailed time diary information, questions are asked that allow for an approximation of parental involvement within the domains of interaction and accessibility. Parent– child interaction and parent accessability are measured by three variables in the NSFH during the wave 1 interviews. The first two variables measure

INTERMEDIATE CHILD OUTCOMES

31

the frequency with which parents spend time with any of their children (a) playing and/or working on projects together (FREQ OF MOTHER PLAYING, FREQ OF FATHER PLAYING) and (b) reading to children and/or helping with homework (FREQ OF MOTHER READING &/OR HELPING, FREQ OF FATHER READING &/OR HELPING). These measures have been used by Marsiglio (1991) and Amato and Rivera (1999) in the measurement of paternal engagement and they probably come the closest to examples of what Albert and Popkin (1987) refer to as “quality time.” The wording of the questions for the parental involvement variables differs slightly depending on whether the parent is the primary or secondary respondent and the age composition of the children in the home. Primary respondents who had children only under age 5 answered questions about the frequency of playing together with the child(-ren) at home and reading to the child(-ren). Primary respondents who had one or more children age 5 or older answered questions about the frequency of working with the child(-ren) on a project or playing together at home and helping the child(-ren) with reading or homework. Secondary respondents who had one or more children age 3 or older also answered questions about the frequency of working on a project or playing together at home and helping with reading or homework. Secondary respondents who had children only under age 3 were not asked any questions about parental involvement. This creates a missing data problem for the mothers and fathers who are secondary respondents and whose only children are between the ages of 0 and 2 in wave 1. (Note that responses are not missing, however, for their spouses who are primary respondents.) We estimate the regressions for these parental involvement measures using only the mothers and fathers who were asked about parental involvement. That is, our regressions exclude all secondary respondents (some of whom are mothers and some of whom are fathers) who were in households where all children were between the ages of 0 and 2 and those with older children who chose not to answer these questions. Then we use the estimated parameters to generate predicted values for all 812 fathers and mothers in our sample. Having accounted for the frequency of potentially enriching parent– child activities, incremental information regarding parental accessibility is obtained by including the mother’s labor force participation status. That is, if the mother is employed she is likely to be less accessible to a child than a mother who is not employed. We use the instrumental variable approach described above to construct the predicted probability of labor force participation in 1987–1988 for each mother (MOTHER’S LFP). This measure is then entered as an independent variable in the parental involvement regressions. As noted earlier, fathers’ employment effects are captured exclusively by our measure of household income (INCOME–WAVE 1), which is the sum of the husband’s annual earnings plus all nonwage income reported by the family in the wave 1 interview. Preference and productivity shifters (i.e., elements of Z 1 and Z 2 ) used as additional regressors in the parental involvement equations include the mother’s

32

¨ STERBACKA ZICK, BRYANT, AND O

age (MOTHER’S AGE), father’s completed years of schooling (FATHER’S EDUC), a dummy variable that measures whether the mother has more than a high school education (MOTHER’S EDUC), and dummy variables that assess each parent’s health in wave 1 (FATHER’S HEALTH and MOTHER’S HEALTH). Initially, we had planned to include the mother’s age and years of schooling and the father’s age and years of schooling among the regressors but standard collinearity diagnostic tests (i.e., variance inflation factors and condition indices) using thresholds identified by Belsey, Kuh, and Welsch (1980) revealed that this would create muticollinearity problems. As a consequence, we enter only the mother’s age, the father’s years of schooling, and the dummy variable that measures whether the mother has more than a high school education in the current formulation. Information about the focal child that may also affect productivity (Z 1 ) and/or preferences (Z 2 ) includes the primary respondent’s assessment of the focal child’s temperament in wave 1 (FOCAL CHILD EASY and FOCAL CHILD DIFFICULT) and a series of dummy variables that measure the focal child’s age at wave 1 (FOCAL CHILD AGE 1, FOCAL CHILD AGE 2, FOCAL CHILD AGE 3, FOCAL CHILD AGE 4, and FOCAL CHILD AGE 5–11). We use a series of dummy variables to measure the focal child’s age rather than a continuous variable to allow for the possibility that age may have nonlinear effects. In addition, because questions about shared activities are asked with respect to all children in the household, we include among the regressors the number of children under age 18 in the home in wave 1 (#KIDS) and the proportion who are male (FRACTION BOYS). Thus, the specification allows for the possibility that parental preferences for engaging in child-focused activities may vary by the number and sex composition of children in the home. Measuring intermediate child outcomes. Both mothers and fathers are asked about behavioral problems and grades for the focal child in the wave 2 interview and we use these as the dependent variables in the second portion of the analysis. The simple correlation between mothers’ and fathers’ reports of grades is very high (␳ ⫽ 0.88), but it is somewhat lower for their reports of behavioral problems (␳ ⫽ 0.60). This may reflect the fact that the reporting of behavioral problems is inherently more subjective than is the reporting of grades. We chose to use both parents’ reports as outcome measures, given that we do not have strong reasons to believe that one parent’s report is more valid than the other’s. The behavioral problems measures are derived from the mothers’ and fathers’ responses to a series of questions regarding whether the focal child exhibits one or more of 15 different behavioral problems. The answers to these questions are summed to create a behavioral problems index (fathers’ reports denoted by BPI-F and mothers’ reports denoted by BPI-M) that has frequently been used in the literature (Yeung, Duncan, and Hill, 1999; Parcell and Menaghan, 1994; Peters, Averett, and Gennetian, 1996; Teachman, Day, Call, and Carver, 1998). High scores on the scale reflect greater numbers of behavioral problems and/or greater severity of any particular problem. Lower numbers reflect fewer behavioral

INTERMEDIATE CHILD OUTCOMES

33

problems and/or less severity with regard to any particular problem. The reliability of the BPI-F and BPI-M scales as measured by the standardized Cronbach’s ␣ are 0.865 and 0.860, respectively. The academic outcome measures are taken from the fathers’ and mothers’ reports of the focal child’s typical academic grades at the time of the wave 2 interview (GRADES-F and GRADES-M). These reports were gathered only for those focal children who are age 11 or older at the time of the second interview (N ⫽ 230) and thus the analysis that focuses on grades is restricted to this smaller sample. The numerical range of their reports is 1 to 5, with higher values being associated with better grades. In our model, behavioral problems and typical grades as reported in the wave 2 interview are each posited to be a function of parental involvement as measured by the forecasted frequency with which the parents engaged in playing and/or working on projects with children and reading to and/or helping children with homework. While these measures of parental involvement may matter throughout childhood, the developmental literature suggests that their impact may be particularly strong in the preschool years (Mantzicopoulos, 1997; Miedel and Reynolds, 1999; Pettit, Brown, Mize, and Lindsey, 1998; Reynolds, Mavrogenes, Bezruczko, and Hagemann, 1996). Yet, our measures are at one point in calender time: the wave 1 interview. To deal with this problem, we use the estimates from our parental time-use equations along with information from the mother’s work history and fertility modules to generate predictions of the frequency of shared parent– child activities during the 5-year period during which the focal child was age 0 – 4. (Recall that all focal children in our sample will be age 5 or older at the time of the wave 2 interview.) These forecasts are made using time-varying information on the focal child’s age, the mother’s employment, the father’s age, and number of children in the home in each of the focal child’s preschool years, but they assume that other independent variables in the equations do not change (i.e., a father who identifies his health as excellent in the wave 1 interview is a father whose health is assumed to have been excellent throughout the focal child’s preschool years). While this assumption is somewhat restrictive, it is necessary to make the calculations tractable, since neither the wave 1 nor wave 2 interviews contain retrospective, time-varying information on the other independent variables. In the case of the mother’s employment status, we use information from the mother’s work history module in conjunction with information on the focal child’s birth date to construct a measure of the number of years the mother worked outside of the home during the focal child’s first 5 years of life (#YRS MOTHER WORKED 0 – 4). Thus, the mother’s labor force participation status may have two effects on the focal child’s behavioral and scholastic outcomes in our formulation. First, it may directly influence the child because employed mothers are presumably less accessible than nonemployed mothers. Second, it may indirectly influence the focal child through the role it plays in shifting the

34

¨ STERBACKA ZICK, BRYANT, AND O

frequency with which both mothers and fathers engage in potentially enriching parent– child activities (i.e., via the parental involvement regressions). Originally, variables measuring the focal child’s time spent in paid child care were also to be included among the regressors in the BPIW2 equation (as the only measure of purchased child-related goods available in the NSFH). But, information on the use of paid child care is fairly limited in the wave 1 interview and the data that are available are highly collinear with the mother’s employment status in wave 1. Family income at the time of the second interview is again measured by the sum of the father’s labor earnings plus all nonwage income (INCOME—WAVE 2). Other variables measuring the effects of preference and technology shifters in the intermediate child outcome equations include a measure of whether the focal child is firstborn (FOCAL CHILD’S BIRTH ORDER), the focal child’s health at the time of wave 2 (FOCAL CHILD’S HEALTH), the focal child’s sex (FOCAL CHILD’S SEX), the focal child’s age in wave 2 (FOCAL CHILD’S AGE— WAVE 2), and the number of siblings less than 18 years of age in the home in wave 2 (#SIBLINGS—WAVE 2). In addition, we include two 4-item scales that assess the parenting practices of the mother (MOTHER’S PARENTING) and the father (FATHER’S PARENTING) and two 12-item scales that assess the father’s goals (FATHER’S GOALS) and the mother’s goals (MOTHER’S GOALS) for their children. These scales are all measured during the wave 1 interview. Similar to the parent– child interaction questions, the parenting and goals questions are not asked of the secondary respondent if all of their children are under age 3 at the time of this first interview. We replace the missing data with the sex-specific means plus a random number that is generated from a normal distribution that has a mean of zero and a standard deviation equal to the sex-specific standard deviation for the nonmissing values. This strategy of mean imputation minimizes any potential bias in our coefficient estimates and leaves the standard errors relatively unaffected. Perhaps surprisingly, the correlations between MOTHER’S GOALS and FATHER’S GOALS and MOTHER’S PARENTING and FATHER’S PARENTING are not excessively high. The simple correlation between mothers’ and fathers’ goals for their children in the current sample is 0.21. The simple correlation between mothers’ and fathers’ parenting styles is 0.36. The reliability of FATHER’S GOALS and MOTHER’S GOALS as measured by the standardized Cronbach’s ␣ are 0.83 and 0.81, respectively, while the Chronbach’s ␣s for MOTHER’S PARENTING and FATHER’S PARENTING are 0.47 and 0.42, respectively. Some formulations in the child outcomes literature estimate separate equations for boys and girls (e.g., Desai, Chase-Lansdale, and Michael, 1989; Yeung, Duncan, and Hill, 1999), while others group boys and girls together and include only a dummy variable for the child’s sex among the regressors (e.g., Peters, Averett, and Gennetian, 1996; Parcel and Menaghan, 1994). With no strong theoretical guidance to draw on with respect to this issue, we estimate the equations both ways and conduct F tests to see which formulation should be

INTERMEDIATE CHILD OUTCOMES

35

preferred on statistical grounds. The F tests reveal that the more parsimonious formulation is to be preferred and so we present the estimates that include only the main effect of the focal child’s gender (FOCAL CHILD’S SEX). RESULTS Descriptive Findings Definitions and descriptive statistics for all of the variables used in the analyses appear in Table 1. The NSFH oversampled several population subgroups. Thus, the descriptive statistics and all of the multivariate analyses make use of the wave 2 weights so that the results may be generalized to the large population of families in 1992–1994 who meet our sampling criteria. Table 1 reveals that the families in our study reflect characteristics typical of young, White, intact families where the father was employed in the late 1980s and early 1990s. The parents are well educated and in good health. They average slightly less than two children in the home and more often than not the mother is employed. Mothers typically reported being more involved in both playing and reading/helping with homework than did fathers. These relative differences in fathers’ and mothers’ child-related activities obtained in the NSFH are consistent with the findings of previous studies that are based on time diary data (Bryant and Zick, 1996a, 1996b; Kooreman and Kapteyn, 1987; Nock and Kingston, 1988; Robinson and Godbey, 1997; Zick and Bryant, 1996). The mean age of the focal child in wave 2 was between 9 and 10 years. And, most often the focal child is reported to be in good or excellent health and s/he is reported to be neither easy nor difficult to raise. Both mothers and fathers also generally report that the focal child typically has few behavioral problems and does well in school. Parents’ Involvement in Child-Related Activities The ordinary least-squares regression results for the frequency of parental involvement with playing and/or working on projects with children and reading to children and/or helping with homework appear in Table 2. We limit our narrative of the findings in this table to those variables that influence multiple types of parental involvement in child-related activities. Turning first to the mother’s probability of employment (MOTHER’S LFP), the estimates reveal that as her probability of employment rises, the frequency with which both parents engage in reading and homework activities rises. This supports the contention that while employed mothers may decrease the time they spend in physical and nonphysical care, they attempt to compensate by increasing other types of potentially enriching parent– child interactions. In addition, it also supports the argument that fathers may increase certain types of parent– child activities in response to the mother’s employment. Next we turn to the effects of father’s education (FATHER’S EDUC). In interpreting these estimates, it is important to remember that father’s and moth-

¨ STERBACKA ZICK, BRYANT, AND O

36

TABLE 1 Variable Definitions and Descriptive Statistics (N ⫽ 811)

Variable FATHER’S EDUC MOTHER’S EDUC

MOTHER’S AGE FATHER’S HEALTH

MOTHER’S HEALTH

# KIDS

INCOME—WAVE 1

FRACTION BOYS

FOCAL CHILD AGE 1 a

FOCAL CHILD AGE 2 a

FOCAL CHILD AGE 3 a

Standard deviation

Definition

Mean

Father’s education (years) Dummy variable: 1 ⫽ mother has more than 12 years of education; 0 ⫽ otherwise Mother’s age at the wave 1 interview (years) Dummy variable: 1 ⫽ father rates his health as good, fair or poor at the wave 1 interview; 0 ⫽ very good or excellent Dummy variable: 1 ⫽ mother rates her health as good, fair or poor at the wave 1 interview; 0 ⫽ very good or excellent Number of children in the home at the wave 1 interview Total annual household income minus the wife’s annual earnings reported in wave 1 (dollars/1000) Proportion of children in the home at the wave 1 interview who are boys Dummy variable: 1 ⫽ focal child is age 1 at the time of the wave 1 interview; 0 ⫽ otherwise Dummy variable: 1 ⫽ focal child is age 2 at the time of the wave 1 interview; 0 ⫽ otherwise Dummy variable: 1 ⫽ focal child is age 3 at the time of the wave 1 interview; 0 ⫽ otherwise

14.18 0.06

2.57 0.24

31.92

5.74

0.10

0.31

0.11

0.32

1.99

1.01

33.55

25.58

0.51

0.43

0.11

0.33

0.12

0.34

0.10

0.31

37

INTERMEDIATE CHILD OUTCOMES TABLE 1—Continued

Variable FOCAL CHILD AGE 4 a

FOCAL CHILD AGE 5–11 a

FOCAL CHILD EASY b

FOCAL CHILD DIFFICULT b

FREQ OF MOTHER PLAYING c

FREQ OF FATHER PLAYING c

FREQ OF MOTHER READING &/OR HELPING c

FREQ OF FATHER READING &/OR HELPING c

MOTHER’S LFP d

Definition

Mean

Standard deviation

Dummy variable: 1 ⫽ focal child is age 4 at the time of the wave 1 interview; 0 ⫽ otherwise Dummy variable: 1 ⫽ focal child is age 5 to 11 at the time of the wave 1 interview; 0 ⫽ otherwise Dummy variable: 1 ⫽ primary respondent rated the focal child as an easy child to raise in the wave 1 interview; 0 ⫽ otherwise Dummy variable: 1 ⫽ primary respondent rated the focal child as a difficult child to raise in the wave 1 interview; 0 ⫽ otherwise Number of times in the past year the mother has played with her children Number of times in the past year the father has played with his children Number of times in the past year the mother has read to her children or helped them with homework Number of times in the past year the father has read to his children or helped them with homework Mother’s estimated probability of being in the labor force

0.11

0.32

0.44

0.52

0.34

0.49

0.05

0.23

242.62 (N ⫽ 700)

129.01

198.14 (N ⫽ 604)

134.33

232.83 (N ⫽ 693)

130.35

135.79 (N ⫽ 600)

133.14

0.59

0.16

¨ STERBACKA ZICK, BRYANT, AND O

38

TABLE 1—Continued Standard deviation

Variable

Definition

Mean

# YRS MOTHER WORKED 0–4

Number of years the mother worked when the focal child was between the ages of 0 and 4 Dummy variable: 1 ⫽ focal child is a girl; 0 ⫽ focal child is a boy Dummy variable: 1 ⫽ focal child is the firstborn (or the only) child; 0 ⫽ otherwise Age of focal child at the time of the wave 2 interview Number of siblings the focal child has in the household at the time of the wave 2 interview Dummy variable: 1 ⫽ focal child’s physical health is rated as good, fair or poor by the primary respondent in the wave 2 interview; 0 ⫽ otherwise Total annual household income minus the wife’s and husband’s annual labor earnings reported in wave 2 (dollars/1000) Four-item scale measuring the mother’s parenting style in the wave 1 interview; ranges from 4 to 12, with higher scores reflecting a more authoritative parenting style

2.93

2.21

0.50

0.52

0.53

0.52

10.16

3.41

1.30

0.98

0.14

0.36

2.36

7.11

7.28

1.40

FOCAL CHILD’S SEX

FOCAL CHILD’S BIRTH ORDER

FOCAL CHILD’S AGE— WAVE 2 # SIBLINGS—WAVE 2

FOCAL CHILD’S HEALTH

INCOME—WAVE 2

MOTHER’S PARENTING

er’s years of education are highly correlated and this is the reason why we substituted a dummy variable (1 ⫽ greater than high school education) for the continuous measure of the mother’s years of schooling. This means that the

39

INTERMEDIATE CHILD OUTCOMES TABLE 1—Continued

Variable FATHER’S PARENTING

MOTHER’S GOALS

FATHER’S GOALS

BPI-F e

BPI-M e

Standard deviation

Definition

Mean

Four-item scale measuring the father’s parenting style in the wave 1 interview; ranges from 4 to 12, with higher scores reflecting a more authoritative parenting style Twelve-item scale measuring the mother’s goals for her children in the wave 1 interview; responses range from 7 to 84, with higher scores reflecting higher goals for the children Twelve-item scale measuring the father’s goals for his children in the wave 1 interview; responses range from 7 to 84, with higher scores reflecting higher goals for the children Fifteen-item scale measuring the behavioral problems of the focal child as assessed by the father in wave 2; scale ranges from 15 to 45 Fifteen-item scale measuring the behavioral problems of the focal child as assessed by the mother in wave 2; scale ranges from 15 to 45

7.19

1.45

66.80

6.94

66.49

7.22

21.44

5.11

21.23

5.02

estimated coefficients associated with father’s education may be reflecting the effect of both parents’ education levels to some extent. In this context, the estimates reveal that as the father’s education rises, both mothers and fathers

¨ STERBACKA ZICK, BRYANT, AND O

40

TABLE 1—Continued

Variable GRADES-F

GRADES-M

Standard deviation

Definition

Mean

Father’s assessment of the school performance of the focal child (age ⬎12 in 1992) in wave 2; ranges from 1 to 5: 5 ⫽ mostly A’s and 1 ⫽ F’s Mother’s assessment of the school performance of the focal child (age ⬎12 in 1992) in wave 2; ranges from 1 to 5: 5 ⫽ mostly A’s and 1 ⫽ F’s

4.01 (N ⫽ 230)

1.08

3.96 (N ⫽ 230)

1.07

a

The omitted group in this series of dummy variables are focal children under the age of 1 at the time of the wave 1 interview. b The omitted group in this series of dummy variables are those focal children who were identified as neither difficult nor easy to raise at the time of the wave 1 interview. c Responses are recorded to arrive at annual frequencies as follows: 2 ⫽ ever or rarely, 12 ⫽ once a month or less, 48 ⫽ several times a month, 52 ⫽ about once a week, 182 ⫽ several times a week, and 350 ⫽ almost every day. d This is the instrumental variable used in place of the mother’s actual wave 1 labor force participation status. Among the regressors in the estimating logit equation were the mother’s education, the mother’s age, household income minus her labor earnings (if any), her health status, residential location (i.e., how urban the area is), and the father’s and mother’s views about parents’ roles in the family. The logit parameter estimates used to generate this instrument are available from the authors upon request. e These behavioral problems include “has sudden changes in mood or feelings,” “feels or complains that no one loves him/her,” “is rather high strung, tense and nervous,” “cheats or tells lies,” “argues too much,” “has difficulty concentrating, cannot pay attention for long,” “is easily confused, seems to be in a fog,” “bullies or is cruel or mean to others,” “is disobedient at home,” “does not seem to feel sorry after he/she misbehaves,” “has trouble getting along with other children,” “is impulsive, or acts without thinking,” “feels worthless or inferior,” and “is not liked by other children.” Parents are asked in each case to identify if the statement was “not true,” “sometimes true,” or “often true” for the focal child. The responses are then coded “1,” “2,” or “3,” respectively, and summed to create the focal child’s behavioral problems index.

increase their frequency of playing with children and/or working on projects together. In addition, fathers also increase their reading and/or helping with homework activities. This suggests that, in general, more highly educated parents may place a premium on devoting time to such parent– child activities. Not surprisingly, the coefficients associated with age of the focal child indicate that the frequency of playing with children and/or working on projects is inversely related to age of the focal child. In contrast, the older the focal child,

41

INTERMEDIATE CHILD OUTCOMES TABLE 2 OLS Parameter Estimates of the Parental Involvement Equations (t Ratios in Parentheses)

Independent variables

FREQ OF MOTHER PLAYING

FREQ OF FATHER PLAYING

FREQ OF MOTHER READING &/OR HELPING

INTERCEPT MOTHER’S LFP MOTHER’S EDUC FATHER’S EDUC # KIDS FATHER’S HEALTH MOTHER’S HEALTH INCOME—WAVE 1 FOCAL CHILD AGE 1 FOCAL CHILD AGE 2 FOCAL CHILD AGE 3 FOCAL CHILD AGE 4 FOCAL CHILD AGE 5–11 MOTHER’S AGE FOCAL CHILD EASY FOCAL CHILD DIFFICULT FRACTION BOYS Adjusted R 2 N

228.61 (5.76)*** 3.94 (0.12) 34.84 (1.79)* 7.09 (3.46)*** ⫺6.02 (⫺1.20) 22.45 (1.51) ⫺28.99 (⫺2.07)** ⫺0.11 (⫺0.57) ⫺0.65 (⫺0.03) ⫺22.00 (⫺1.11) ⫺8.19 (⫺0.41) ⫺54.75 (⫺2.78)*** ⫺83.50 (⫺4.07)*** ⫺1.92 (⫺1.85)* 12.61 (0.92) 4.57 (0.22) 7.07 (0.67) 0.16 699

238.86 (5.40)*** ⫺11.14 (⫺0.30) 10.25 (0.49) 8.68 (4.11)*** ⫺10.78 (⫺2.01)** ⫺22.35 (⫺1.45) 5.79 (0.38) ⫺0.20 (⫺0.89) ⫺1.08 (⫺0.05) ⫺36.67 (⫺1.69)* ⫺42.58 (⫺2.03)** ⫺50.07 (⫺2.34)** ⫺88.25 (3.99)*** ⫺2.85 (⫺2.53)** 9.44 (0.61) ⫺79.44 (⫺3.24)*** 31.59 (2.76)*** 0.19 603

56.55 (1.31) 88.42 (2.47)** ⫺12.71 (⫺0.61) 3.42 (1.54) 9.52 (1.77)* ⫺37.48 (⫺2.34)** ⫺5.41 (⫺0.36) 0.50 (2.38)** 88.86 (4.07)*** 76.49 (3.58)*** 96.39 (4.50)*** 81.79 (3.83)*** 48.76 (2.20)** ⫺0.06 (⫺0.06) ⫺5.85 (⫺0.39) ⫺23.71 (⫺1.03) 0.74 (0.06) 0.05 692

FREQ OF FATHER READING &/OR HELPING ⫺5.81 (⫺0.12) 70.21 (1.78)* ⫺27.80 (⫺1.23) 9.62 (4.25)*** ⫺1.26 (⫺0.22) ⫺33.02 (⫺2.01)** ⫺6.77 (⫺0.42) ⫺0.12 (⫺0.48) 20.58 (0.84) 59.13 (2.51)** 14.99 (0.66) 22.26 (0.96) 19.46 (0.81) ⫺0.91 (⫺0.76) 15.05 (0.90) ⫺27.53 (⫺1.04) 9.83 (0.80) 0.05 599

* p ⬍ 0.10. ** p ⬍ 0.05. *** p ⬍ 0.01.

the greater the frequency with which the mother reads to her children and/or helps with homework. In part, this result likely reflects the fact that interest in reading to/with children is greater after infancy (the omitted category in this series of dummy variables). As the number of children in the household increases, fathers decrease their frequency of playing with children and mothers increase their frequency of reading and/or helping with homework. Presumably the decline in the father– child play activities occurs as the number of children rises in part because children are more likely to substitute sibling play for play with the father. Reading and/or helping with homework activities may increase because reading in particular is an activity that can be done simultaneously with more than one child. The Focal Child’s Intermediate Outcomes After estimating the equations that focus on parents’ frequencies of childrelated activities, the next step is to use these equations to generate predicted values for frequencies in each category during the period when the focal child was ages 0 – 4 years. We sum these forecasted values across mothers and fathers. The totals (PREDICTED FREQ OF PARENTS’ READING/HELP WITH HOMEWORK and PREDICTED FREQ OF PARENTS’ PLAYING/WORKING ON PROJECTS), scaled by 1000, are then entered as independent variables

42

¨ STERBACKA ZICK, BRYANT, AND O

along with the proportion of the total in each category that is attributable to the mother (FRACTION OF READING/HELPING W/HOMEWORK DONE BY MOTHER and FRACTION OF PLAYING/WORKING ON PROJECTS DONE BY MOTHER) in the child outcome equations. In our conceptual framework, intermediate child outcomes are posited to be a function of the total time that parents have historically (rather than concurrently) spent with their children in human capital enriching activities. We operationalize this model by first estimating the parental involvement equations and then using these estimated parameters to forecast parental involvement during the 5-year period when the focal child was ages 0 to 4 years. 1 The predicted values are then used as independent variables in the child outcome equations where the outcome in question is measured after the child is age 5 or older. Thus, the time ordering of the data that we use help to insure that while parental involvement may affect child outcomes, the reverse is not likely to be true. The modest adjusted R 2 ’s associated with the four equations (ranging from 0.05 to 0.19) reduce the reliability but not the validity of using them to forecast parental involvement over the preschool years. That is, while the associated coefficient estimates are likely to be unbiased, they will have generally large standard errors associated with them—suggesting that they will provide a con1

This procedure is perhaps best illustrated with an example. Suppose we have a female focal child whose mother was not employed outside of the home while the child was ages 0 –3 years, but was employed outside of the home when the child was age 4 years. In addition, suppose the focal child had no siblings from ages 0 –2 years, but a female sibling was born when the child was age 3 years. The mother and father are both college graduates and both parents are in excellent health. The primary respondent rates the focal child as neither easy nor difficult to raise at the time of the wave 1 interview. Finally, the mother is 33 years old and the earnings of the husband plus the household’s nonwage income total $40,000 per year at the time of the wave 1 interview. In this example, the forecast of the mother’s playing activities over the focal child’s preschool years would be as follows: Year 0: 228.62 ⫹ 共34.84 ⫻ 1兲 ⫹ 共7.09 ⫻ 16兲 ⫺ 共6.02 ⫻ 1兲 ⫺ 共0.11 ⫻ 40兲 ⫺ 共1.92 ⫻ 33兲 ⫽ 303.12 Year 1: 228.62 ⫹ 共34.84 ⫻ 1兲 ⫺ 共0.65 ⫻ 1兲 ⫹ 共7.09 ⫻ 16兲 ⫺ 共6.02 ⫻ 1兲 ⫺ 共0.11 ⫻ 40兲 ⫺ 共1.92 ⫻ 34兲 ⫽ 300.55 Year 2: 228.62 ⫹ 共34.84 ⫻ 1兲 ⫺ 共22 ⫻ 1兲 ⫹ 共7.09 ⫻ 16兲 ⫺ 共6.02 ⫻ 1兲 ⫺ 共0.11 ⫻ 40兲 ⫺ 共1.92 ⫻ 35兲 ⫽ 277.28 Year 3: 228.62 ⫹ 共34.84 ⫻ 1兲 ⫺ 共8.19 ⫻ 1兲 ⫹ 共7.09 ⫻ 16兲 ⫺ 共6.02 ⫻ 2兲 ⫺ 共0.11 ⫻ 40兲 ⫺ 共1.92 ⫻ 36兲 ⫽ 283.15 Year 4: 228.62 ⫹ 共3.94 ⫻ 1兲 ⫹ 共34.84 ⫻ 1兲 ⫺ 共54.75 ⫻ 1兲 ⫹ 共7.09 ⫻ 16兲 ⫺ 共6.02 ⫻ 2兲 ⫺ 共0.11 ⫻ 40兲 ⫺ 共1.92 ⫻ 37兲 ⫽ 238.61. Thus, in this example, the forecast of the mother’s total number of playing activities over the focal child’s preschool years would be 303.12 ⫹ 300.55 ⫹ 277.28 ⫹ 283.15 ⫹ 238.61 ⫽ 1,402.71.

INTERMEDIATE CHILD OUTCOMES

43

servative estimate of the effects of parental involvement on intermediate child outcomes. The behavioral problems scales are censored on both tails of the distribution, with the lower observable limit being 15 and the upper observable limit being 45. Similarly, both parents’ reports of the focal child’s grades are censored, with the lower observable limit being 1 and the upper observable limit being 5. Technically, this means that the estimation should be done using a two-limit tobit estimation routine. This estimation technique produces coefficients that have a complicated interpretation (McDonald and Moffitt, 1980). As a consequence, we estimate the equations using both two-limit tobit and ordinary least-squares (OLS) regression. Comparisons across the two estimating techniques reveal little differences in the parameter estimates, suggesting that censoring is not a serious problem in these equations. We elect to present the OLS results because of their straightforward interpretation. The tobit results are available from the authors on request. Variance inflation factor and condition index calculations both reveal that there are no multicollinearity problems in our equation specifications using the thresholds suggested by Belsey, Kuh, and Welsch (1980). The OLS parameter estimates for both the mothers’ and the fathers’ reports of the focal child’s behavioral problems and his/her typical grades appear in Table 3. Comparisons between the parameter estimates generated by mothers’ and fathers’ reports of behavioral problems reveal that the signs for estimated coefficients across the equations are the same and the magnitude of the estimated coefficients are generally similar. Tests of statistical significance differ in some instances (e.g., FRACTION OF READING/HELPING W/HOMEWORK DONE BY MOTHER not reaching the 0.05 threshold in the case of the BPI-M equation, but exceeding the 0.05 threshold in the case of the BPI-F equation) but are generally parallel. The same observations can also be made when comparing the parameter estimates of the mothers’ and fathers’ reports of typical grades. These similarities suggest that the choice of parent reporter does not have a significant impact on the conclusions we draw from our multivariate analyses. Table 3 indicates that as our forecast of parental involvement in reading/ homework activities during the preschool years rises, behavioral problems decline and grades improve (with all but the estimated coefficient in the fathers’ reports of grades equation reaching conventional levels of statistical significance). The estimated coefficients associated with the mother’s share of reading/ helping with homework in each equation suggest that mothers’ involvement is moderately more effective than fathers—although both matter. The frequency of playing with children and/or working on projects also appears to be positively related to school performance as measured by typical grades, but in this case there is no gender difference. Finally, we find that the number of years the mother is in the labor force has little impact on intermediate child outcomes as measured by behavioral problems and reported grades. The one exception is in the equation where the fathers’ reports of behavioral problems in wave 2 is the dependent

⫺10.5 (⫺1.47) 5.8 (0.89) 0.25 (1.31) ⫺1.03 (⫺2.95)** 0.90 (2.28)** ⫺0.06 (⫺1.09) 1.41 (2.99)** ⫺0.02 (⫺0.13) 0.48 (3.61)** 0.34 (2.70)** ⫺0.04 (⫺1.63) ⫺0.07 (⫺2.91)** ⫺0.002 (⫺0.07) .10

⫺15.09 (⫺2.12)** 8.86 (1.35) 0.47 (2.50)** ⫺0.50 (⫺1.43) 1.55 (3.93)** 0.04 (0.74) 1.88 (3.96)** 0.31 (1.61) 0.29 (2.19)** 0.52 (4.11)** ⫺0.05 (1.91)* ⫺0.07 (⫺2.97)** 0.02 (0.87) .13

3.2 (0.87) ⫺0.06 (⫺0.90) 0.48 (3.21)** 0.23 (1.66) ⫺0.10 (⫺2.42)** ⫺0.31 (⫺1.75)* 0.10 (1.70)* ⫺0.009 (⫺0.17) 0.02 (0.40) 0.013 (1.38) 9.91 ⫻ 10 ⫺ 4 (1.17) ⫺1.95 ⫻ 10 ⫺ 4 (⫺0.28) .20

5.26 (1.94)*

0.16 (3.74)**

0.055 (1.45)

⫺6.53 (⫺2.22)**

GRADES-F (N ⫽ 230)

2.40 (0.68) ⫺0.86 (⫺1.25) 0.45 (3.08)** 0.28 (2.11)** ⫺0.09 (⫺2.21)** ⫺0.56 (⫺3.54)** 0.10 (1.65) ⫺0.07 (⫺1.55) 0.04 (0.92) 0.02 (2.04)** 1.90 ⫻ 10 ⫺ 4 (0.23) ⫺6.2 ⫻ 10 ⫺ 4 (⫺0.92) .21

6.17 (2.29)**

0.10 (2.31)**

0.088 (2.31)**

⫺5.54 (⫺1.90)*

GRADES-M (N ⫽ 230)

PREDICTED FREQ OF PARENTS’ READING/HELP WITH HOMEWORK (IN 1000’S) and PREDICTED FREQ OF PARENTS’ PLAYING/WORKING ON PROJECTS (IN 1000’S) are the sum of the forecasts of mothers’ and fathers’ frequencies of reading and/or helping with homework (playing with children and/or helping with projects) over the 5-year period when the focal child was a preschooler. FRACTION OF READING/HELPING W/HOMEWORK DONE BY MOTHER and FRACTION OF PLAYING/WORKING ON PROJECTS DONE BY MOTHER are the proportion of this total attributable to the mother. * p ⬍ 0.10. ** p ⬍ 0.05. *** p ⬍ 0.01.

a

⫺0.26 (⫺2.26)**

⫺1.0 (⫺1.14)

37.5 (4.43)** ⫺0.26 (⫺2.34)**

36.43 (4.30)**

INTERCEPT PREDICTED FREQ OF PARENTS’ READING/HELP WITH HOMEWORK (IN 1000’S) a PREDICTED FREQ OF PARENTS’ PLAYING/ WORKING ON PROJECTS (IN 1000’S) a FRACTION OF READING/HELPING W/HOMEWORK DONE BY MOTHER a FRACTION OF PLAYING/WORKING ON PROJECTS DONE BY MOTHER a # YRS MOTHER WORKED 0–4 FOCAL CHILD’S SEX FOCAL CHILD’S BIRTH ORDER FOCAL CHILD’S AGE—WAVE 2 FOCAL CHILD’S HEALTH # SIBLINGS—WAVE 2 MOTHER’S PARENTING FATHER’S PARENTING MOTHER’S GOALS FATHER’S GOALS INCOME—WAVE 2 Adjusted R 2

BPI-M (N ⫽ 812)

⫺4.0 (⫺3.77)**

BPI-F (N ⫽ 812)

Independent variables

TABLE 3 OLS Parameter Estimates of Child Outcomes (t Ratios in Parentheses)

44 ¨ STERBACKA ZICK, BRYANT, AND O

INTERMEDIATE CHILD OUTCOMES

45

variable. In this equation, increases in the number of years the mother works while the focal child is a preschooler is associated with a significant increase in the father’s reporting of behavioral problems. In the case of fathers’ reports of behavioral problems, the role that mother’s employment plays appears to be somewhat complex. The direct effect of employment during the preschool years is associated with an increase in fathers’ reports of behavioral problems. At the same time, these negative employment effects are offset by employment-induced increases in both the mother’s and father’s parent– child activities—which in turn are associated with fewer behavioral problems. To assess the net impact, we calculated the average predicted BPI scores by the mother’s employment patterns during the preschool years. Children whose mothers did not work outside of the home at all during their first 5 years of life had a mean predicted BPI score of 21.19 (N ⫽ 206). The corresponding mean for children whose mothers worked 1– 4 years was 21.41 (N ⫽ 258), while the mean predicted score for children whose mothers worked all 5 years was 21.22 (N ⫽ 347). We also looked at the preschool work patterns of the mothers for those focal children whose predicted BPI scores were in the lowest and the highest quartiles and again we found no statistically significant differences— suggesting that on balance, mothers and fathers may be compensating for the loss in accessibility when mothers enter the labor force by increasing the frequency with which they engage in potentially enriching parent– child activities such as reading and/or helping with homework. A number of consistently significant findings with regard to preference and productivity shifters across the equations are also worth noting. Specifically, we find consistent evidence that firstborn children have higher BPI scores and higher grades compared to children who are not firstborn. Girls generally appear to have significantly fewer behavioral problems and significantly higher grades than do boys. If the focal child has serious health problems, then the parents report significantly more behavioral problems and significantly lower grades. This finding is consistent with previous research that has found an association between children’s health and behavioral problems (Goldberg et al., 1997; Parcel and Menaghan, 1994). Finally, parenting practices and parental goals generally appear to influence reported behavioral problems but not grades. The higher each parent’s aspirations for the focal child (as measured in 1987–1988), the lower the reported BPI score 5 years later. While the more authoritative the parenting style (as measured in 1987–1988), the higher the reported BPI score 5 years later. DISCUSSION There has been a long, rich history of household time-use research within the social sciences. Most of this work focuses on identifying the correlates of time allocation and only implicitly assumes that the time spent in various activities (e.g., playing with children) is related to family outcomes (e.g., a child’s later behavioral problems). Rarely have the connections between time use and family

46

¨ STERBACKA ZICK, BRYANT, AND O

outcomes been empirically tested. In this article we explore what possible links there may be between mothers’ employment time, the frequency with which parents engage in playing and reading and/or homework activities with their children, and children’s subsequent behavior and grades. As more mothers have entered the labor force, the debate regarding what this means for young children has grown. Our analyses suggest that employed mothers engage in reading/homework activities with their children more frequently than do nonemployed mothers. In addition, when the mother is employed, the father also engages in reading and/or helping with homework more often than when the mother is not employed. In part, this may occur because in those households where both parents are employed, less time is spent in direct physical and/or nonphysical care (Bryant and Zick, 1996a, 1996b) and as a consequence, mothers and fathers attempt to compensate by doing more of those activities they perceive to be enriching. An equally important question asked in this study is “Does the frequency of parent– child interactions affect intermediate child outcomes?” And, the answer to that question appears to be “Yes.” Shared activities between parents and children, such as reading to a child, working on a project together, helping with homework, and actively playing together, generally require high levels of parental involvement. And, as a consequence, when these activities are shared more frequently with a father or mother they appear to result in significant human capital enrichment for the child. Our analysis is not forthcoming about the reasons why the mother’s reading and/or help with homework decreases behavioral problems and increases academic performance more so than the father’s reading and/or help with homework. It may be that mothers’ interaction styles in these activities are more effective than are fathers’ interaction styles. Or, perhaps the duration of the mothers’ time in these activities is typically longer than that of the fathers. (See Bryant and Zick, 1996a, for some confirmation of this.) Recall that the NSFH contains information on frequency but not duration of interaction. Interestingly, most of the parent– child time-use literature places considerable emphasis on physical and nonphysical care time (Bryant and Zick, 1996b; Coverman and Sheley, 1986; Gershuny and Robinson, 1988; Hill and Stafford, 1985; Zick and Bryant, 1996) or the effect of the mother’s employment status on particular child outcomes (Alwin and Thornton, 1984; Datcher-Loury, 1988; Desai, Chase-Lansdale, and Michael, 1989; Fleisher, 1977; Haurin and Mott, 1983; Haveman and Wolfe, 1994; Hill and Duncan, 1987; Milne et al., 1986). If the goal is to make assessments of the correlates with and trends in developmentally and socially meaningful parent– child time, our findings suggest that analyses of parent– child time should probably desegregate these relatively large categories of time use into more precise categories that capture the types of focused, potentially human capital enriching activities of the type examined here. Our analyses also show that parental goals and parenting practices play important roles in intermediate behavioral outcomes of children. Separate from

INTERMEDIATE CHILD OUTCOMES

47

the frequency of focused parent– child interactions, our findings suggest that fathers and mothers affect child behavior through the goals they set for their children and through the parenting practices that they use. What’s more, the effects of fathers’ and mothers’ parenting styles and goals, while both in the same direction, have somewhat different magnitudes. In particular, fathers’ goals appear to have larger effects on the children’s BPI scores than do mothers’ goals. As an initial investigation, our work has several limitations that should be noted in order to place the study’s findings in context. The NSFH subsample used in this study is a somewhat homogeneous group of families in that they are all Caucasian with employed fathers who are married and who did not experience a divorce or parental death during the 5 years that elapsed between the wave 1 and wave 2 interviews. The time-use measures we have for these parents come from recall questions regarding typical frequency of activities asked at one point in time. As such, we had to make assumptions about the relative validity and reliability of these measures and about household characteristics that did not change over time (e.g., a parent’s health) in order to generate estimates of the frequency of potentially enriching parent– child activities during the preschool years. Finally, we use only two intermediate child outcomes in our analyses–a behavioral problems index reported when the child is between the ages of 5 and 17 and a report of typical grades received in school. It remains to be seen if other intermediate child outcomes (e.g., a child’s reading/math abilities and vocabulary) and long-run child outcomes (e.g., a child’s educational attainment and earnings as an adult) are similarly affected by parental involvement. REFERENCES Albert, L., and Popkin, M. (1987). Quality Parenting, Random House, New York. Alwin, D. F., and Thornton, A. (1984). “Family origins and schooling process,” American Sociological Review 49, 784 – 802. Amato, P. R., and Rivera, F. (1999). “Paternal involvement and children’s behavior problems,” Journal of Marriage and the Family 61, 375–384. Baydar, N., Greek, A., and Gritz, R. M. (1999). “Young mothers’ time spent at work and time spent caring for children,” Journal of Family and Economic Issues 20, 61– 84. Becker, G. S. (1991). A Treatise on the Family (rev. ed), Harvard Univ. Press, Cambridge, MA. Becker, G. S. (1965). “A theory of the allocation of time,” Economic Journal 75, 493–517. Belsey, D. A., Kuh, E., and Welsch, R. E. (1980). Regression Diagnostics, Wiley, New York. Blair, S. L., and Hardesty, C. (1994). “Parental involvement and the well-being of fathers and mothers of young children,” Journal of Men’s Studies 3, 49 – 68. Bryant, W. K., and Zick, C. D. (1996a). “An examination of parent– child shared time,” Journal of Marriage and the Family 58, 227–237. Bryant, W. K., and Zick, C. D. (1996b). “Are we investing less in the next generation? Historical trends in time spent caring for children,” Journal of Family and Economic Issues 17, 365–392. Cooney, T. M., Pedersen, F. A., Indelicato, S., and Palkovitz, R. (1993). “Timing of fatherhood: Is ‘on time’ optimal?,” Journal of Marriage and the Family 55, 205–215. Coverman, S., and Sheley, J. F. (1986). “Change in men’s housework and child-care time, 1966 – 1975,” Journal of Marriage and the Family 48, 413– 422. D’Amico, R. K., Haurin, J., and Mott, F. L. (1983). “The effects of mothers’ employment on adolescent and early adult outcomes of young men and women,” in Children of Working

48

¨ STERBACKA ZICK, BRYANT, AND O

Parents: Experiences and Outcomes (C. D. Hayes and S. B. Kamerman, Eds.), National Academy Press, Washington, DC. Datcher-Loury, L. (1988). “Effects of mother’s home time on children’s schooling,” Review of Economics and Statistics 70, 367–373. Desai, S., Chase-Lansdale, P. L., and Michael, R. T. (1989). “Mother or market? Effects of maternal employment on the intellectual ability of 4-year-old children,” Demography 26, 545–561. Fleisher, B. M. (1977). “Mother’s home time and the production of child quality,” Demography 14, 197–212. Gershuny, J., and Robinson, J. P. (1988). “Historical changes in the household division of labor,” Demography 25, 537–552. Goldberg, S., Janus, M., Washington, J., Simmons, R. J., MacLusky, I., and Fower, R. S. (1997). “Prediction of preschool behavioral problems in healthy pediatric samples,” Journal of Developmental and Behavioral Pediatrics 18, 304 –313. Haveman, R., and Wolfe, B. (1994). “Succeeding Generations—On the Effects of Investments in Children,” Russell Sage Foundation, New York. Heckman, J. J. (1993). “What has been learned about labor supply in the last twenty years?,” American Economic Review 83, 116 –121. Hewlett, S. A. (1991). When the Bough Breaks—The Cost of Neglecting our Children, Basic Books, New York. Hill, C. R., and Stafford, F. P. (1985). “Parental care of children: Time diary estimates of quantity, predictability, and variety,” in Time, Goods and Well-Being (F. T. Juster and F. P. Stafford, Eds.), Institute for Social Research, Ann Arbor, MI. Hill, M. S., and Duncan, G. J. (1987). “Parental family income and the socioeconomic attainment of children,” Social Science Research 16, 39 –73. Hochschild, A. (1989). The Second Shift—Working Parents and the Revolution at Home, Viking, New York. Ishii-Kuntz, M., and Coltrane, S. (1992). “Predicting the sharing of household labor: Are parenting and housework distinct?,” Sociological Perspectives 35, 629 – 649. Kmenta, J. (1986). Elements of Econometrics, 2nd ed., Macmillan, New York. Kooreman, P., and Kepteyn, A. (1987). “A disaggregated analysis of the allocation of time within the household,” Journal of Political Economy 95, 223–249. Lamb, M. E., Pleck, J. H., Charnov, E. L., and Levine, J. A. (1987). “A biosocial perspective on paternal behavior and involvement,” in Parenting Across the Lifespan: Biosocial Perspectives, (J. B. Lancaster, J. Altman, and A. Rossi, Eds.), Academic Press, New York. Leibowitz, A. (1977). “Parental inputs and children’s achievement,” The Journal of Human Resources 12, 242–251. Mantzicopoulos, P. Y. (1997). “The relationship of family variables to Head Start children’s preacademic competence,” Early Education and Development 8, 357–375. Maret, E., and Finlay, B. (1984). “The distribution of household labor among women in dual-earner families,” Journal of Marriage and the Family 46, 357–364. Marsiglio, W. (1991). “Paternal engagement activities with minor children,” Journal of Marriage and the Family 53, 973–986. McDonald, J. F., and Moffitt, R. A. (1980), “The use of tobit analysis,” Review of Economics and Statistics 57, 318 –321. Medrich, E. A., Roizen, J., Rubin, V., and Buckley, S. (1982). The Serious Business of Growing Up—A Study of Children’s Lives Outside of School, Univ. of California Press, Berkeley, CA. Miedel, W. T., and Reynolds, A. J. (1999). “Parental involvement in early intervention for disadvantaged children: Does it matter?,” Journal of School Psychology 37, 379 – 402. Milne, A. M., Myers, D. E., Rosenthas, A. S., and Ginsburg, A. (1986). “Single parents, working mothers, and the educational achievement of school children,” Sociology of Education 59, 125–139.

INTERMEDIATE CHILD OUTCOMES

49

Moorehouse, M. J. (1992). “Linking maternal employment patterns to mother– child activities and children’s school competence,” Developmental Psychology 27, 295–303. Murnane, R. J., Maynard, R. A., and Ohls, J. C. (1981). “Home resources and children’s achievement,” Review of Economics and Statistics 63, 369 –377. National Survey of Families and Households Documentation (1994). Center for Demography and Ecology, Madison, WI. Nock, S. L., and Kingston, P. W. (1988). “Time with children: The impact of couples’ work-time commitments,” Social Forces 67, 59 – 85. O’Connell, M. (1993). Where’s Papa? Fathers’ Role in Child Care, Population Trends and Public Policy, Bulletin No. 20, Population Reference Bureau, Washington, DC. Parcel, T. L., and Menaghan, E. G. (1994). Parents’ Jobs and Children’s Lives, Aldine de Gruyter, New York. Peters, H. E., Averett, S., and Gennetian, L. (1996). Fathers as Providers of Child Care, presented at the NICHD-sponsored Conference on Father Involvement, October 10 –11, Bethesda, MD. Pettit, G. S., Brown, E. G., Mize, J., and Lindsey, E. (1998). “Mothers’ and fathers’ socializing behaviors in three contexts: Links with children’s peer competence,” Merril Palmer Quarterly 44, 173–193. Pleck, J. H. (1996). “Paternal involvement: Levels, sources, and consequences,” in The Role of the Father in Child Development (M. E. Lamb, Ed.), 3rd ed., Wiley, New York. Reynolds, A. J., Mavrogenes, N. A., Bezruczko, N., and Hagemann, M. (1996). “Cognitive and family-support mediators of preschool effectiveness: A confirmatory analysis,” Child Development 67, 1119 –1140. Robinson, J. P. (1977). How Americans Use Time: A Social-Psychological Analysis of Everyday Behavior, Praeger, New York. Sanik, M. M. (1981). “Division of household work: A decade comparison—1967–1977,” Home Economics Research Journal 10, 175–180. Sewell, W. H., Hauser, R. M., and Wolf, W. C. (1980). “Sex, schooling and occupational status,” American Journal of Sociology 86, 551–583. Stafford, F. P. (1987). “Women’s work, sibling competition, and children’s school performance,” American Economic Review 77, 972–980. Teachman, J., Day, R., Call, V., and Carver, K. (1998). “Sibling resemblance in behavioral and cognitive outcomes: The role of father presence,” Journal of Marriage and the Family 60, 835– 848. Yeung, W. J., Duncan, G. J., and Hill, M. S. (2000). “Putting fathers back in the picture: Parental activities and children’s attainment,” Marriage and Family Review 29, 97–113. Zick, C. D., and Bryant, W. K. (1996). “A new look at parents’ time spent in child care: Primary and secondary time use,” Social Science Research 25, 260 –280.