Family Background, Family Income, Maternal Work and Child Development - October 1998

2.1 Research Questions

The parental circumstances in which children are raised have changed. Two major observable phenomena illustrate this point. The proportion of currently "married" (registered or common-law) mothers of preschool age children who are working has increased dramatically since the 1980's. By 1994, over half (56 percent) of women with children less than age 3 were employed, up from 39 percent in 1981 (see Statistics Canada 1996 for these statistics and the following). The employment rate of women whose youngest child was aged 3 to 5 also increased during this period, rising from 47 percent in 1981 to 59 percent in 1994. Second, the proportion of female headed lone-parent families has risen in the last two decades. In 1991, they represented 16 percent of all families with children, up from 10 percent in 1971.¹ This implies that there is a rising number of children who spend a significant proportion of their childhood in single-parent families. With the absence of a spouse, the family environment is altered and thus begins a long period of adjustment into poorer economic conditions with spells of poverty and welfare dependence. Female lone parents are considerably less likely than women in two-parent families to be employed. In fact, the proportion of female lone parents with jobs in 1994 was lower than in 1981: 50 percent versus 55 percent. This decline can be traced to a drop in employment among young lone mothers (Dooley 1996). In 1994, only 27 percent of lone mothers with children under age 3 and 47 percent of those whose youngest child is aged 3 to 5 were employed, compared with 60 percent of those whose youngest child is aged 6 to 15.

In many respects children are better off than they have ever been. The typical child is more likely to have been the result of a planned decision by his or her parents, has a better chance of survival through infancy and childhood, completes more years of schooling, has more material goods, and has fewer siblings with whom to compete for the time and attention of parents. These parents in turn are better educated, dedicate more hours to market work, and are having their children later when earning power is greater.²

If family income is used as a summary measure of children's material well-being there are large differences according to the type of family children live in and whether the mother is strongly committed or not to the labour market. For example, in 1994-1995, for children aged 4-5 years in Canada (see Table A1 and following sections for this particular sample used in the empirical analysis), mean family income in a two-parent family was $64,000 with a working mother and $46,000 with a "non-working" mother; in a single mother family, mean income was $26,000 with a working mother and $15,000 with a "non-working" mother.

At the same time, for the same children and for a larger window of children's ages, developmental outcomes can be assessed. In the areas of emotional and behavioural problems, social functioning and cognitive skills, it is observed that positive (scores) outcomes for children are strongly correlated with their parents' income. Moreover, a number of children are experiencing difficulties. They are likely to be assessed as developmentally "delayed". The children at risk are mostly those in single-parent families, those at the lower end of the income scale, those whose parents' educational levels are low, and those whose parents fare lowest in terms of parenting skills.³ Thus, with respect to child well-being, the move into the work force by the mother has potential influences depending on the outcomes of interest.

The preceding observations (and the existing evidence in the research literature which is reviewed in the next section) concerning the effects parental employment across all population groups suggest that three main research questions can be raised.

First, does maternal employment have positive or negative implications for children's well-being independently from both child and family background characteristics that make some women more likely to work?

Parental employment will affect family life in two ways. First, decisions involving the employment of the mother, if we suppose that there is a spouse/partner who is already a labour force participant, invariably implicate increased (or decreased if she decides not to work) financial resources and their use to purchase market goods and services enhancing the well-being of the family. Second, given the absolute constraint on the amount of time available, although it is "renewable" each day, there is the problem of the allocation of each family member's time. Many sources of family well-being cannot be directly purchased in the market. They must be produced by combining family members' time with market goods purchased with financial resources obtained with each family member's labour market income. Although maternal employment increases the amount of market goods the family can accumulate, it may decrease the opportunity for mother's to provide nonmarket commodities, especially in the form of attention, monitoring, learning activities, and cognitive stimulation for her children. On the other hand, the additional market goods and services made available by the mother's earnings may fill the gap created by her decreased time at home when employed. Purchased substitute child care is an example of services increased financial resources permit.

Moreover, family work decisions not only determine the allocation of the family members' time between market and nonmarket activities, but they also cause changes in the intra-allocation of time spent on nonmarket activities (leisure, education, organization, home production, personal time). It is reasonable to expect that maternal employment will change household production and the allocation of available time. If the mother participates in the labour market, family members will generally shy away from labour-intensive home production techniques and move towards goods-intensive techniques. Time given up for paid work may have low returns and, in its place, employed mothers substitute other kinds of time that have higher returns. Alternately, to give up time for a low wage job instead of dedicating it to child care may also have low returns. Although we can measure parents' time spent with children in non-direct child-care-related activities (e.g. going to cinema, family eating together at restaurant)⁴ we have little or no knowledge of the process by which direct or indirect time spent with children relates to child development. When both parents are employed, men are likely to contribute somewhat more to home production.⁵

Finally, it has also been found (Duncan, 1990; and Strauss, 1992; Phipps and Burton, 1993) that mothers' relative income (in two-parent families) has a definite impact on children expenditures. A one dollar increase in mothers' income translates into more expenditures on children goods than a similar increase in spouses' income.⁶ Thus, families in which mothers earn a relatively important share of total income may spend more on goods and services that are correlated with children development.

Second, do better outcomes for children actually reflect differences between families?

Observed differences in child outcomes - there are gaps not only by parental income but also according to parental education - may show the effects of preexisting differences in families' characteristics, especially the mother's. Mothers who gain employment differ in many important ways from mothers who do not. For example, on average, the former have more education directly influencing both child development and mothers' employment. These factors, that predispose some mothers towards employment, may be the real reason for the better outcomes.

A major general finding from social science research is the strong association between children's well-being and families socioeconomic background consisting evidently of financial resources but more importantly of own human capital (such as innate or learned skills, educational attainment, psychological and health status), personal or psychological resources (such as resiliency, positive outlook, motivation), and "social capital" (Coleman, 1988) (such as community ties, relations with neighbours and friends). These parental characteristics are resources used by parents for the socialization process of children and the provision of cognitive stimulation. These resources include cultural values inherited from the parents' own family of origin and life experiences. It is plausible that some of these resources might be positively affected by work experiences. On that account, Parcel and Menaghan (1990, 1994) argue that occupational conditions matter.⁷ On the one hand, parents' experience in paid work activities can bring to children forms of social control. Because market work imposes social discipline, having a job influences parental socialization efforts with children and the kinds of behaviour encouraged in children. On the other hand, long and low paid working hours may be accompanied by feelings of parental stress that may interfere with effective and positive parenting.

Evidently, families will differ in terms of their basic parental abilities and skills, in their endowments of financial and human capital, and in their values. Parents of different socioeconomic conditions will raise their children in different ways, largely due to differences in parental characteristics and life experiences. Parents with more meagre economic resources may still efficiently use them in the child-rearing process. For instance, less educated and/or less financially well-off parents may not possess a great deal observable endowments. However, they can spend much quality time interacting with their children, so the children may have access to whatever human capital the parents do possess, including personal resources.

Although researchers differ in their estimates of the contribution of parental characteristics to differences in children's outcomes (see, for example, Heckman, 1995), there is general agreement that controlling for these differences is essential. The fundamental problem for empirical analysis is that some of these important parental characteristics are unmeasured or unobservable (see below).

Third, do better outcomes for children appear to be more strongly related to income from earnings than to income from government transfers?

This third question is incidental to this research and reaches far beyond it (not the least by the data that would be required to answer it). However, a "naive" and simple income model, in which children's well-being was seen as depending only on the level of their families' economic resources as well as the amount of time parents "invest" in them, would imply that one might expect to observe better outcomes with a transfer system in place than without it. Since income transfers, such as welfare assistance and child tax benefits, increase the income of poor families and make it possible for mothers to spend time with their children rather than in labour market, they should have a positive contribution. Allusion has already been made to the reverse causation that may exist between the observation that some families have low income that places them below the poverty level and so may have to depend on welfare. On this issue the following questions have been raised: do "culture of poverty" or underclass affiliation have effects independent of income? Will raising a child in a single-parent family and his or her exposure to a spell of poverty and welfare dependence cause delayed development? Alternate questions are: does low-wage maternal employment have detrimental outcomes effects for children living in poverty? Does combining welfare and work generate better children's outcomes? All these questions are difficult to answer because they are linked to the second research question.

2.2 Conceptual and Methodological Issues

In economics, the conceptual framework of the analysis to address these questions and to sort out these effects is dominated by the "investment theory" developed by Becker (1981; Becker and Tomes, 1976, 1979). In this theory, children's outcomes as adult are strongly related to the inter-generational transmission of endowments (genetic, human, cultural and financial) that parents transmit to their children, combined with what parents invest in their children. In particular, the decision to participate in the labour market and the choice of hours of work for each member in a family are the result of the maximization of a family utility (or well-being) function, given the budget and time constraints of the family. Time plays an unique role in this process, both as an input in the production of family services, but also as an argument in the family utility function. One of the family services produced are "child services" from which parents derive utility (such as the joy of having them, of raising them and of seeing them becoming "successful" adults). These "services" depend on both the quantity and the "quality" of children in the family. The quality of children is produced in part by the time parents are willing to allocate for the production of family services. Within this framework, the quantity and quality of children, labour participation and hours of work are jointly chosen variables. The data requirements to implement empirically such a "structural" model in its full generality are formidable. Rather than identifying all the mechanisms and measuring all variables, the studies employ some simplifying assumptions and most estimate a "reduced-form" model. In this latter approach, a "counterfactual" is used to describe the alternative (such as alternative children's outcomes) to actual events used as a basis of comparison.

2.2.1 Problems with the identification of the effects of work on child outcomes

The main goal of this paper is to identify whether children with mothers who work, all other thing equal, do worse than children with a mother that stays home. Of course, if working was the result of a random draw, we could simply compare the mean scores of children with working mothers with the mean scores of children with non-working mothers and measure whether the differences between the means are statistically different from each other. Clearly, this is not the case. Mothers who work generally hold more human capital than mothers without work. For example, they possess more years of education. Therefore, simply comparing means could confound the effects of education with the effects of work. A positive difference between means of working mothers with means of non-working mothers would be attributed to working, when it would be due to differences in education. But this is not a major problem, because regression analysis can control for years of education which is available in the data set. However, there remain other human capital variables that are not observable to the statistician. For example, mothers can have the same years of education but one can be an A student and the other a C student. We do not have any information on the quality of the education received and the work put into schooling by the mother. If we suppose that these non-observed human capital variables are positively correlated with the probability of working and with positive outcomes for children again we could be attributing positive effects to work while they are the positive effects of unobservable human capital variables. Finally, women who do not work may precisely not work because they love child-rearing while mothers who do work have no particular preferences for child-rearing. In this case, negative effects of work could be due to differences in preferences towards child-rearing. It is very difficult to find a proxy for child-rearing preferences in the data.

Ideally to identify the effects of work, one would need to find a variable (an instrument) that affects the probability of working but does not affect the children's outcomes. We initially thought that local area unemployment rates could be that variable. Unfortunately, this variable is very correlated with another variable used in the regression analysis which is the percentage of low-income families in the child's neighbourhood. This makes unfeasible the use of instrumental variable methods or selection bias correction methods for the identification of the effects of work on outcomes. However, the analysis has clearly identified what could be important sources of bias in the work effects on outcome and make important statements about the effects of work on outcomes. Section 4 will present the optimal estimation methodology while reiterating the caveats found in this section. In future work with panel data, more of these unobservable human capital effects will be taken into account since it will be possible to control for fixed family effects.

The same type of identification problems are also true of income effects, however regressions will be presented for two measures of income, one will be total family income, while the other will be sources of income that are not from the mother, providing the analysis with an income variable that can vary independently from the mother's labour supply variable.

2.2.2 Timing effects

One important issue is the timing of the effects of maternal employment on child outcomes. Some psychologists emphasize the first years of life as a determinant for subsequent child outcomes (Belsky and Eggebeen 1991). But current circumstances in children's lives are also likely to influence measurable child outcomes such as cognitive ability and social adjustment. It is not obvious which set of influences is stronger and how to control for the persistence of earlier circumstances. It is not clear also that the first year of a child's life is more important than the second or third year.⁸ On this issue, the data set released from Cycle 1 of the NLSCY has one limitation, as it does not provide a complete history of work patterns (labour participation, vacation, sick leave, unpaid and paid maternity leaves) of women since the birth of the children (for the first and subsequent years of life for those children older than one year). So, the timing of the mother's return to work relative to the birth of her child cannot be taken into account, for those working women. Also, there is no information on the child's exposure to income insecurity (such as family welfare dependence and poverty spells) since all labour force and income information relate to the year preceding the interview.

• ¹Data from the NLSCY indicates that, in 1994-1995, 84.2% of children in Canada aged 0 to 11 years lived in a two-parent family, 15.7% lived with a single parent (among them most - 92.8% - lived with a single mother).
• ²Garnett and Myles (1996) present the following evidence for children aged 0-6 between 1973 and 1991: "The share of children in families with two or more adult earners rose from 38 to 62 percent; the proportion of children in families where the highest earner has at least some postsecondary education rose from 25 to over 40 percent; the percentage of children with more than two children declined from 40 to 29 percent; and, the age at which families are having children rose (the proportion of children with the highest earner under age 26 fell from 18 to 11 percent)", 253.
• ³In this area there is the important problem of sorting out cause and effect. For example, children living in persistent poverty or in low income families are more likely to have lower developmental outcomes. However, family income is a "marker" for other observables and non- observables variables. Parents in low income families are more likely to have less education which is associated with lack of success in the job market, to show symptoms of depression, to have less self-esteem, to adopt less competent parenting practices that indirectly benefits children. Causation could runs in both directions.
• ⁴From Statistics Canada survey's on the use of time.
• ⁵See Lefebvre and Merrigan (1998) who also show using two surveys on the use of time in Canada that the proportion of total time spent with children in families where the youngest child is four or younger also rose in the 80's, whether mothers were employed or not. The increase is certainly related to the growing knowledge of the high payoff of investing time in children, particularly when they are very young.
• ⁶Qualitatively, the same effect has been observed for a child benefit (allowance) that does not depend on household income, marital status or labour market status (see Kooreman, 1998).
• ⁷The relationship between working parents' jobs characteristics and working conditions and a child's outcomes has never been addressed in the Canadian literature. The rich information of the NLSCY on different dimensions of jobs and parent and children's outcomes represents a unique opportunity to disentangle the many channels by which parents' work affects children. This particular aspect is the subject of a companion forthcoming research paper.
• ⁸In Canada, most employed mothers giving birth to a child either give up temporarily paid work or go on maternity leave. In 1991, there were 164,000 maternity leaves from work (in the same year 403,000 Canadians were born), almost double the number in 1980.