The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Federal Reserve Bank of Chicago Maternal Employment and Overweight Children Patricia M Anderson, Kristin F. Butcher and Phillip B. Levine WP 2002-10 Maternal Employment and Overweight Children August 2002 Patricia M. Anderson Kristin F. Butcher Phillip B. Levine Department of Economics Dartmouth College Hanover, NH 03755-3514 and National Bureau of Economic Research Federal Reserve Bank of Chicago Research Department 230 S. LaSalle Street Chicago, IL 60604 Department of Economics 106 Central Street Wellesley College Wellesley, MA 02481 and National Bureau of Economic Research patricia.m.anderson@dartmouth.edu kbutcher@frbchi.org plevine@wellesley.edu We would like to thank Kate Baicker, David Card, Anne Case, Jon Gruber, Luojia Hu and Doug Staiger, for helpful discussions, two anonymous referees for their comments and seminar participants at Princeton University, Harvard University, Dartmouth College, George Washington University, the University of Chicago, the University of Illinois-Chicago, the University of Illinois at UrbanaChampaign, The Federal Reserve Bank of Chicago, and the National Bureau of Economic Research Health Economics summer meeting. We are also grateful to Joe Hotz and Rebecca Kilburn for sharing their data on state childcare regulations. Rachel Roth provided research assistance through the Presidential Scholars program at Dartmouth College. ABSTRACT This paper seeks to determine whether a causal relationship exists between maternal employment and childhood overweight. We use matched mother/child data from the National Longitudinal Survey of Youth and employ econometric techniques to control for observable and unobservable differences across individuals and families that may influence both children’s weight and their mothers’ work patterns. Our results indicate that a child is more likely to be overweight if his/her mother worked more hours per week over the child’s life. Analyses by subgroups show that it is higher socioeconomic status mothers whose work intensity is particularly deleterious for their children’s overweight status. I. INTRODUCTION Childhood overweight may be one of the most significant health issues facing American children today.1 Over the past three decades, it has grown so dramatically that observers routinely describe the trend as an epidemic. In the 1963 to1970 period, 4 percent of children between the ages of 6 and 11 were defined to be overweight; that level had more than tripled by 1999, reaching 13 percent (Centers for Disease Control, 2001). This general trend masks important differences in the incidence of childhood weight problems by socioeconomic status. For example, only 1.7 percent of black boys were recorded as being overweight in the 1963 to 1965 period, but that rate has grown almost 10 times to reach 15.1 percent in 1988-1994 (National Center for Health Statistics, 1998). By 1988-1994, 18.8 percent of Mexican-American boys and 17.4 percent of black girls were overweight compared to 11.7 percent of white girls. As we show subsequently, rates of childhood overweight are much higher for the poor and for those with less educated parents as well. Researchers and public health officials are currently at a loss to explain the rapid rise in childhood obesity. At some basic physiological level, the cause of this increase in overweight among children is clear: weight gain is attributable to taking in more energy than one expends. What is unclear is what has upset this balance between energy intake and expenditure over the last three decades. For example, while genetics clearly play a role in determining overweight, it is hard to imagine that such dramatic and rapid changes have taken place in our genetic makeup that this factor alone could be the culprit. 1 The Surgeon General David Satcher said, “This crisis is stealing youth, innocence, and health from our children, and yet as a nation we have been badly remiss in addressing it.” (Mashberg, 1999). 1 Thus, it is important to consider other causes of overweight, including the environmental factors that may affect either the intake or expenditure of energy. In this regard, analysts have tended to point to factors like the availability and consumption of calorie-rich fast foods along with increased television viewing and decreased exercise. These explanations beg the question of why these behaviors have changed, however, since fast food and television have both been available for decades. Popular opinion routinely draws a direct link between mothers working and poor outcomes for children. Typical comments express concern about the effects of child care, for instance warning that “parents who casually warehouse their kids could use a healthy does of anxiety (Feder, 1999).” According to the Washington Post, “two-thirds of the people surveyed said that although it may be necessary for a mother to work, it would be better for her family if she could stay home and care for the house and children (Grimsley and Melton, 1998).” Popular news reports on the topic of childhood obesity are similarly peppered with comments from health practitioners who either implicitly or explicitly attribute changes in children’s diet and exercise to the increased likelihood that both parents work outside the home. For example, a 1999 Boston Herald article cited a pediatric nutrition specialist who “noted in particular that dual-career couples are spending less time monitoring their latchkey children, who consequently snack after school, using their often liberal allowances on candy, ice cream, or soda pop (Mashberg, 1999).” Popular nutrition author Dr. Andrew Weil in an interview on CNN attributed the increased reliance on prepared and processed foods to the fact that “typically, people say they don’t have time to cook.” The interviewer attributed this time constraint to the prevalence of dual-career families (Weil, 2002). Those who believe that dual-career families may be contributing to changes in children’s diet and exercise habits have a compelling prima facie case. The rise in women working outside the 2 home coincides with the rise in childhood weight problems. From 1970 to 1999, the fraction of married women with children under six who participate in the labor force doubled, rising from 30 percent to 62 percent. Married women with children ages 6 to 17 dramatically increased their labor force participation as well, rising from 49 percent to 77 percent over this period (U.S. Bureau of the Census, 2000). There are several potential mechanisms through which children’s eating patterns and level of physical activity may be affected by having parents who work outside the home. Children may eat differently if child care providers are more likely to give them food that is highly caloric and of poor nutritional value. Further, parents who work outside the home may serve more high-calorie prepared or fast foods, and unsupervised children may make poor nutritional choices when preparing their own after-school snacks. Similarly, unsupervised children may spend a great deal of time indoors, perhaps due to their parents’ safety concerns, watching television or playing video games rather than engaging in more active outdoor pursuits. Alternatively, the increase in working mothers may have no adverse effect on childhood weight problems. First, any correlation between working mothers and overweight children may be spurious, if, for example, mothers who work are those who would be less attentive to their children’s nutrition and exercise in any case. On the other hand, there may be a negative impact of maternal work on childhood overweight if households where the mother works have more money with which to purchase more healthful meals, and children of these households participate in after-school sports, thereby increasing their activity levels. Finally, increases in maternal work may be a small component of the myriad environmental changes affecting children’s health. The United States might have faced the current epidemic in childhood overweight, even if women’s labor force activity had not dramatically increased. 3 The purpose of this paper is to explore whether the observed coincident rise in maternal employment and childhood overweight represents a causal relationship between these two phenomena. We focus on the role of maternal employment rather than parental employment more generally for three reasons. First, it is mothers’ labor supply that has changed dramatically over recent decades. Second, despite the dramatic increase in women’s paid market employment, they still bear the bulk of responsibility for child rearing. Third, data limitations in the analysis reported below only enable us to link the employment histories of mothers and children. Using the National Longitudinal Survey of Youth (NLSY), we first document a simple correlation between maternal employment and overweight. The remainder of the paper attempts to identify whether this simple relationship reflects more than a spurious correlation in which children whose mothers work full-time would still be overweight even if their mothers did not work. To this end, five techniques are employed. First, we estimate standard probit models including a full range of observable characteristics of the mother and child. Second, we estimate “long-difference” models that compare changes in child overweight status at the beginning and end of the panel to changes in maternal work history, thus differencing out any unobserved child-specific fixed effect. We also estimate sibling difference models, both comparing weight outcomes for siblings at the same time and at the same age, thus differencing out any unobserved family-specific fixed effects. Finally, we estimate instrumental variables models. We also estimate the models by income, maternal education, and race/ethnicity. We analyze subgroups separately since public policy may interact with childhood overweight in different ways for different groups. For example, if maternal work is particularly deleterious for overweight among poor children, one might worry that welfare reform and its attendant work requirements will have unintended adverse consequences for childhood obesity. 4 Our results lead us to several conclusions. First, mothers who work more intensively, in the form of more hours per week over the child’s life, are significantly more likely to have an overweight child. There is no evidence that mothers who work are simply those who are inherently less attentive to their children’s health outcomes. In other words, we do not find any support for the notion that these differences are driven by unobservable heterogeneity. Interestingly, the aggregate relationship is entirely driven by the relationship between maternal work and children’s weight outcomes among higher socioeconomic status families. These children have the lowest incidence of obesity. For example, if the mother in a top income quartile family works an extra 10 hours per week (on average while working since the child’s birth), the child is between 1.3 and 3.8 percentage points more likely to be overweight. Thus for high socioeconomic status families, increases in mothers’ average weekly hours of work over the last three decades can explain between 12 and 35 percent (depending on the specification and under certain assumptions) of the increase in the incidence of overweight among children in these families. Finally, while our results indicate that maternal employment has a significant impact on children’s overweight for some groups, those who would blame maternal employment for the deterioration in children’s health overall need to look elsewhere for the whole story. Particularly for the subpopulations with the highest incidence of childhood obesity, mothers’ employment does not appear to be a factor. II. PREVIOUS RESEARCH Being overweight as a child has both immediate consequences and long-term implications for individuals, as well as for society as a whole. For example, the increase in childhood overweight has been accompanied by a marked increase in the number of children developing type II diabetes, which has serious health risks (Thompson, 1998). In addition, studies have shown that overweight children are much more likely to become overweight adults than normal weight children (Bouchard, 5 1997; and Dietz, 1997). Being overweight may have serious health consequences for adults including diabetes, coronary heart disease, atherosclerosis, and colorectal cancer (Power, et al., 1997). Furthermore, being overweight may have social and economic consequences. For example, studies have shown that obesity is negatively related to education and earnings (Averett and Korenman, 1996; Gortmaker, et al., 1993; and Cawley, 2000). Moreover, the health consequences for individuals place additional pressure on the scarce resources of the nation’s health care system. Thus, the importance of a better understanding of the determinants of childhood obesity, including the potential role of maternal employment, is clear. Although little research has directly examined the impact of maternal employment on childhood overweight, past work on other determinants may help inform this issue. Many studies have found a strong correlation between parent and child weight problems, (c.f. Vuille and Mellbin, 1979; Dietz, 1991), although such a correlation could be due to either genetic or behavioral factors. As indicated above, though, while a genetic explanation for overweight is compelling, other factors must play a role as well given the dramatic trends in overweight in the United States over the last few decades. Thus, researchers have turned to environmental factors ( c.f. Locard, et al., 1992; Woolston, 1987; Bar-Or, et al., 1998). The evidence shows a positive correlation between television viewing and overweight among children (c.f. Gortmaker, et al., 1996; and Dietz and Gortmaker, 1985). Findings regarding the relationship between family structure, socioeconomic status, and childhood overweight are more mixed. Sobal and Stunkard (1989) find a weak correlation between low socioeconomic status and obesity for children, but Dietz (1991), Gerald, et al. (1994) and Wolfe, et al. (1994) find a stronger one. Similarly, studies have tended to find a significant relationship between family structure and obesity, although results across studies are not always consistent about 6 the sign of the effect (c.f. Dietz, 1991; Wolfe, et al., 1994; and Gerald, et al., 1994). Researchers have also examined the influence of the types of foods children eat, but the role of parental involvement in this regard is also mixed (c.f. Klesges, et al., 1991; and Birch and Fisher, 1998). Finally, recent work on breastfeeding suggests that infants who are breastfed may be less likely to be overweight later in life than those who are not (von Kries, et al. 1999, Gilman, et al. 2001). Research specifically examining the link between maternal employment and childhood overweight is very limited. Takahashi, et al. (1999) finds a positive relationship between mothers’ employment and children’s probability of being overweight, but the data are only for 3-year-old Japanese children. Additionally, Johnson, et al. (1992) study US children age 2-5 in 1987-88 and find no significant effect of maternal employment on nutrient intake. III. DATA AND DESCRIPTIVE STATISTICS To conduct our analysis, we mainly use the matched mother-child data from the National Longitudinal Survey of Youth (NLSY), which are described briefly here and in more detail in the Data Appendix. The NLSY first surveyed 12,686 individuals, of whom 6,283 were women, between the ages of 14 and 22 in 1979 and has continued to survey them annually through 1994 and biennually since then. Beginning in 1986, the children of those women have been surveyed biennually as well. At the time we began this project matched data through 1996 had been released, giving us six survey years of data.2 The key outcome variable, an indicator for whether the child is overweight, is based on body mass index (BMI). BMI is defined as weight in kilograms divided by height in meters squared (kg/m2) and is a commonly used measure to define obesity and overweight in adults. The Centers for Disease Control (CDC) has recently endorsed the use of BMI to assess overweight in children, and 7 has produced sex-specific BMI percentile charts for children aged 2 to 20 for this purpose. We follow CDC nomenclature and classify children with a BMI above the 95th percentile of the BMI distribution for their sex-age group as “overweight.”3 We use the height and weight measures for children between the ages of 3 and 11 in the NLSY to calculate BMI. We truncate the age distribution prior to adolescence, since the younger children are likely to have less choice about the composition of their diet than do adolescents. In addition, this truncation means that children in our sample have not gone through puberty, with its attendant body changes, which may make BMI a less accurate approximation to measures of adiposity. Since the National Health and Nutrition Examination Survey (NHANES) is the source of the CDC’s official measures of overweight, we calculate overweight for the same age children in the NHANES and find that our data are comparable.4 The second key component to our analysis is the mother’s employment history. While current employment status is available, it is less appropriate than a measure of long-term exposure. Current employment may fluctuate, and with it the flows of calories consumed and expended, but it is really the stock of net calories that will determine overweight status. Using the child’s lifetime exposure to maternal employment will let us more closely approximate this stock concept. Fortunately, the NLSY provides a virtually complete work history for each mother, allowing us to calculate total weeks worked and total hours worked starting from the date of the child’s birth until each survey date. We use these data to construct a measure of average hours worked per week by mothers during the weeks in which they worked at all. We then use this hours per week measure 2 Appendix Table 1 presents sample means for all of the variables used in the analysis. Children above the 85th percentile are referred to as “at-risk of overweight.” 4 In the NHANES III, 10.3 percent are overweight, while in our analysis sample 10.6 are so classified. 3 8 along with total weeks worked per year to capture lifetime exposure.5 Including both of these measures rather than simply a total hours measure helps us to distinguish between those mothers who work at a high intensity, but intermittently, from those who consistently work, but at a lower intensity. Mothers who work at a high intensity may face time constraints during the period in which they are working and may be less able to provide a daily routine that includes nutritious foods and regular exercise. Table 1 presents simple descriptive statistics for the fraction of children who are overweight by three measures of socioeconomic status and by intensity of mothers’ work per week. We have three broad measures of socioeconomic status: quartile of the family income distribution, mother’s education, and race/ethnicity. In the first column, we can see that children who are poorer, whose mother’s are less educated, and who are members of racial/ethnic minority groups are more likely to be overweight. We have divided mothers’ work into three categories capturing intensity of work per week: those who never worked, those who worked fewer than 35 hours per week, and those who worked at least 35 hours per week, on average, since the child’s birth. In the first row, we see that the more hours a mother works per week, the more likely her child is to be overweight. The analysis by subgroups, however, shows that this pattern is not universal. The adverse relationship between mother’s working and childhood overweight seems to hold only for wealthier families, with better educated, or non-Hispanic mothers. 5 One potential difficulty in using these two specific maternal employment measures is that it may lead to identification of our models from individuals exhibiting unusual behavior. For instance, holding constant weeks worked and examining changes in hours worked per week, it is possible that women who work for just a few weeks per year but very intensively in those weeks may be identifying these models. Alternatively, variation in weeks worked holding constant hours worked per week could come from individuals who work 40 hours per week, but just for one week per year. To examine this issue, we created ranked categories of weeks worked and hours worked per week, cross-tabulated them, and found very few observations far off the diagonals. In other words, virtually no one has an extreme value of one measure, but not the other. 9 IV. ECONOMETRIC APPROACHES Although the descriptive statistics in Table 1 make a compelling prima facie case that there is a relationship between maternal work and childhood overweight, particularly for higher socioeconomic status groups, mothers who work are likely to differ from mothers who do not in both observable and unobservable ways. These omitted variables may bias (either up or down) the relationship between maternal work and childhood overweight across all subgroups. In our analysis, we use five techniques to address these concerns. First, we estimate standard probit models for whether the child is overweight. The effect of mothers’ work is identified by variation across children and over time.6 While these models can account for observable differences across individuals, there still may be unobservable differences that bias the relationship between a mother’s work intensity and her child’s weight. For example, if mothers who work more hours are those who are less attentive to their children’s health regardless of their work effort, that will induce a spurious positive correlation between hours of work and overweight. Alternatively, suppose permanent family income is related to both child health and maternal work. With only imperfect proxies for permanent income the unobservable components of family income may load onto the maternal work coefficient. We have four techniques for addressing unobservable heterogeneity. The first three exploit the longitudinal and family-based nature of the survey. This allows us to “difference out” any permanent unobservable characteristics of individual children over time or within families that might influence both a mother’s work intensity and her children’s weight. For example, if a mother suffers from chronic depression that affects her ability to monitor her children’s health and her ability to 6 Note that in these models, each child may have multiple observations over time, and may have siblings in the sample, who also have multiple observations overtime. Our reported standard errors are corrected for heteroskedasticity and an arbitrary covariance structure over time and within families. 10 work, these techniques will account for it. Child fixed effects can be eliminated using “long differences,” where we take the difference between the first and the last observation for each individual in the sample. Here the effect on the child’s probability of being overweight is identified by variation over time in the mother’s work behavior within each child’s lifetime. Family fixed effects can be eliminated in two alternative ways. We use data on sibling pairs in the sample to create both “point in time” sibling differences and “at the same age” sibling differences. For pairs observed at the same time, identification is provided if the mother changed her work behavior between the births of the two siblings. For pairs at the same age, work intensity varies if the mother changed her work behavior in the years it took the younger sibling to reach the same age as the older sibling at a particular point in time. Each of these “difference” methods has strengths and weaknesses. A well-known drawback with any difference method is that it may exacerbate attenuation bias due to measurement error (Greene 1993). It is for this reason that we have chosen to estimate child fixed effect models in longdifferences rather than first differences.7 If a mother’s work behavior is highly serially correlated, then much of the observed variation in work intensity over short periods of time may be due to measurement error. Long-differences reduce this problem (Griliches and Hausman 1986). Sibling difference methods at the same time also have an advantage in this regard relative to sibling differences at the same age. The former utilizes the mother’s work patterns averaged over both siblings’ entire lives. The latter throws out some of this information since some of the mother’s work behavior during the older sibling’s life is discarded while “waiting” for the younger sibling to “catch up.”8 Since another method of reducing problems associated with measurement error is to 7 8 The median difference is 6 years (3 surveys) between the first and last observation. Because we have restricted the NLSY child sample to include those between the ages of 3 and 11 and because the 11 average data over longer periods (Zimmerman 1992), sibling differences at the same time and long differences offer this advantage.9 Beyond that, although all the difference methods control for fixed unobservable characteristics (either over time or within the same family), each has different vulnerabilities to factors that change over time or within the family. If, for instance, a mother becomes depressed during our sample period and this affects both her work and her children’s health, long differences cannot control for this. Sibling differences at the same age may or may not difference out a shock like this, depending upon when it occurs, but for pairs at the same time the shock to both siblings would be the same (assuming that the shock has the same effect for children of different ages). Alternatively, if one child in the family suffers from chronic health problems, long-differences would control for this, but both forms of sibling difference methods could be biased. Because it is impossible to specify the form of unobservable changes over time or within family, it is impossible to determine which method is preferable, a priori. Instrumental variables estimation is the fourth method we use to control for unobservable heterogeneity. In theory, this method can account for unobservable heterogeneity, whether it is fixed or variable, and for measurement error bias. Instruments must be related to mothers’ work behavior, but have no effect on children’s probability of being over weight. Here, we use the variation between states and over time in the unemployment rate, child care regulations, wages of child care workers, welfare benefit levels, and the status of welfare reform in the state.10 Higher unemployment NLSY child assessments are only conducted every other year, only about half as many sibling pairs are available when we compare siblings at the same age. 9 Children’s ages (and thus the number of years of data available to calculate the lifetime exposure to maternal work) vary systematically across these methods. For the “at the same age” estimates, the siblings’ average age is 6.6 years. For the “at the same time” estimates, on average the younger sibling is 5.9 years old and the older sibling is 9.2 years old. For the “long difference” estimates, in the second observation, the child is on average 9.2 years old. 10 First stage regression results are presented in Appendix Table 2. Data on child care regulations were graciously 12 makes it more difficult for mothers to find work. Child care regulations and higher child care wages may reflect higher costs and less utilization. Women in states with more restrictive welfare rules may be more likely to work. Because the residuals in a model of children’s overweight status are unlikely to be related to these geographic variables, our model should be appropriately identified. In practice, instrumental variables estimation has two drawbacks. First, it is often difficult to come up with variables that satisfy the exogeneity requirement. Second, even if one can come up with such variables, they are often only weakly related to the variable of interest, leading to weak second stage results. Keeping the strengths and weaknesses of each of these approaches in mind, we turn to the results. V. ECONOMETRIC ANALYSES A. Probit Analysis The first column of Table 2 presents the result of estimating a simple probit on the probability of child overweight, based on average hours worked per week (if working) and average weeks worked per year, over the child’s lifetime.11 Here we find a positive and significant relationship between the average hours a mother works per week (if working) and childhood overweight. The point estimate indicates that mothers who work 10 hours more per week increase the likelihood that their children will be overweight by 1.2 percentage points.12 The number of weeks worked are negatively and significantly related to childhood overweight in this specification. provided by Joe Hotz and Rebecca Kilburn and were used in Hotz and Kilburn (1996). Because each of our instruments only differs across states and time, the fact that we use so many of them makes it difficult to interpret any one particular coefficient. Multicollinearity will lead to imprecise parameter estimates because so little variation in the data is available to identify any specific coefficient. 11 Throughout the paper, all models include variables indicating whether the child’s height and weight were physically measured or reported by the child’s mother. Mothers appear to underestimate the child’s height, resulting in those children being more likely to be classified as overweight. 12 We have also estimated comparable models using an indicator for at-risk of overweight and BMI as a continuous measure as dependent variables and obtained qualitatively similar results. 13 Given that there are no controls for socioeconomic status, it is likely that weeks worked is picking up some of the positive effect of income on health status. Column 2 adds controls for a number of demographic variables. Regarding race/ethnicity, if black and Hispanic mothers have fewer employment opportunities than white mothers, and black and Hispanic children are more likely to be overweight for a variety of reasons, excluding race and ethnicity controls will understate the effect of working on children’s weight. For similar reasons we control for mother’s education (coefficient shown) and AFQT score (coefficient not shown), since a mother’s education and ability may affect both her employment patterns and her children’s health. A mother may continue to work outside the home with her first child, and choose to reduce her outside work effort with the birth of subsequent children, implying that birth order and number of children in a family may have an effect on children’s weight. Thus, we control for whether a child is firstborn and the number of children in the family. Although we do not focus on their coefficients, we control for a number of other variables either because of their likely link to children’s weight or to maternal employment patterns.13 As with the work variables, there is a question of how to interpret these additional coefficients. For instance, maternal education may have a causal impact on childhood overweight, or mothers with more education may have other attributes that are different and reduce the likelihood that their children will be overweight. The results in Column 2 show that including these control variables reduces the coefficient on average hours worked per week from 1.2 percentage points to 0.8, but it is still statistically significant. However, the number of weeks worked becomes insignificant. Several of the other coefficients are worth noting. Black children are significantly more likely to be overweight than other 13 These include the child’s birth weight, whether the child is female, both the child’s and the mother’s age in years, controls for education levels of the mother’s parents, and controls for whether they were present when the mother was 14. We also include dummy variables for each year of the NLSY. 14 groups. Mother’s education is negatively and significantly related to the probability that her child is overweight – an extra year of education reduces the probability that a child will be overweight by 0.6 percentage points.14 Children from larger families are also less likely to be overweight, although being the first-born child is not significantly related to the probability of being overweight. Column 3 includes all the controls in Column 2 as well as additional controls for whether the child was ever breastfed, and the mother’s weight status. As discussed earlier, recent evidence (von Kries, et al. 1999, Gilman, et al. 2001) suggests that children who were breastfed are less likely to develop weight problems by the time they are school aged. Mothers with demanding work schedules may be less likely to find the time to breastfeed, and this is a possible pathway through which maternal work may affect childhood overweight. Additionally, mother’s weight status may reflect either the impact of genetics on the child’s likelihood of being overweight or the effects of the common home environment on the family’s weight status.15 Although the additional variables included in Column 3 do not change the estimated impact of hours per week or weeks worked much, they are significantly related to childhood overweight. We estimate that children who are breastfed are about 2.3 percentage points less likely to be overweight. Again, the interpretation of this finding is unclear – there may be nutritional value in breast milk that affects children’s health later in life, or it may simply be that mothers who breastfeed are more attentive to their children’s nutrition throughout the children’s lives. Mother’s weight is also found to have a large impact on children’s weight status. Note that an obese mother (BMI of at 14 AFQT score (not shown) also has a negative and significant coefficient. 15 In a sense, including mother’s weight may be “overcontrolling” for the home environment. If working mothers are time constrained and are more likely to rely on calorie-rich prepared and fast foods, then we would expect everyone in the family to be more likely to be overweight when the mother works. 15 least 30) is also overweight (BMI of at least 25). Therefore, a child whose mother is obese is a full 8.1 percentage points (the sum of the two coefficients) more likely to be overweight. Column 4 adds controls for average family income since the child’s birth and the percent of the child’s life that the mother was married. Average family income since the child’s birth is a proxy for permanent income, which measures parents’ long-term ability to meet their children’s needs. Similarly, the percent of the child’s life the mother was married is a measure of the long-term resources available to the family. Again, the inclusion of the additional variables in Column 4 does not affect the estimated impact of the maternal employment variables. Moreover, we find no effect of family income or mother’s average marital status on childhood overweight. The insignificance of family income appears to be due to the fact that socioeconomic status is well controlled for by race, education, AFQT score and grandparents’ education. If we include income without these other variables, as expected, we find that overweight is negatively correlated with income. So far we have considered the impact of exposure to maternal employment over the child’s entire life for children in the 3-11 age range, yet recent research emphasizes the importance of the early childhood environment on subsequent outcomes (see c.f. Nelson, 1999; and Shonkoff and Phillips, 2000). In Table 3, we investigate whether the timing of maternal employment makes a difference for children’s overweight status. Table 3 presents two sets of analyses. First, it separates the sample into children age 3-5 and those age 6-11. This allows us to examine whether the effect of mother’s work differs between pre-school and elementary school-aged children. If the earlier result was driven by day care quality, one might expect there to be a bigger effect of mothers working more hours per week for pre-school children. Second, Table 3 allows mother’s work in the child’s first three years of life to have a separate impact from overall work. 16 The first column simply reprints the results from Table 2, Column 4 for comparison. Columns 2-5 include the same control variables used generating this result. In the second and third columns, we can see that an increase in mothers’ hours worked per week has a larger effect on the likelihood of childhood overweight for school-age children, although the difference between the two groups is not statistically significant (p-value=0.42). The fourth column shows that neither hours per week nor weeks worked in the child’s first three years of life has a significant impact on the likelihood that a child is overweight when included in the model with no other maternal work measures. In Column 5 we include our overall lifetime exposure measures along with exposure during the first three years of life, and we see that it is only hours per week since the child’s birth that matters. Thus, the impact seems cumulative, rather than concentrated in a particular period. B. Correcting for Unobservable Heterogeneity The preceding section shows that mothers who work more intensively, in the form of more hours per week, on average, over their child’s life, are more likely to have an overweight child. This result holds when we control for a wide range of observable characteristics. In Table 4, we present results for models that control for unobservable differences across individuals and families that may affect both mother’s work intensity and children’s weight. The results from the long-difference model are shown in the first column of Table 4. In this specification, we find that children of mothers who work an additional 10 hours per week while working face a 1.5 percentage point increase in the likelihood of overweight. This estimate is actually larger than that obtained from the probit model and is statistically significantly different from zero at conventional levels. We continue to find weak results regarding the relationship between the number of weeks worked since birth and childhood overweight. It is the intensity of the 17 work effort that seems to be most important. There is no evidence from these estimates, then, that our earlier result was driven primarily by unobserved heterogeneity in fixed characteristics. The results of sibling differences are reported in Columns 2 (“at the same time”) and 3 (“at the same age”) of Table 4. Here, point estimates are slightly larger than those obtained from the probit model, but they are somewhat more imprecise. Standard errors are roughly twice the size, and this greater imprecision renders the point estimates insignificant in these specifications. Nevertheless, these estimates again provide no real indication of serious bias from unobserved heterogeneity in the probit specifications. One of the advantages of sibling differences over individual differences is that they allow us to examine the impact of attributes that do not vary for an individual. Here, for example, we can examine the impact of breastfeeding on overweight, while holding constant any fixed maternal characteristics. The coefficient on breastfeeding is unstable across the specifications in Columns 2 and 3. In neither case is it significantly different from zero, although in Column 3 the point estimate is similar to that in Table 2. As these results suggest that mothers who breastfeed their children may simply be different in many ways from mothers who do not, more research is needed to determine if there is, in fact, a causal impact of breastfeeding on overweight. Column 4 presents the results of the instrumental variables model. For computational simplicity, we have applied linear probability models to estimate the factors affecting childhood overweight despite the discrete nature of this outcome.16 The IV method results in point estimates on our maternal employment measures that are similar to those reported in other models, but the standard errors are several times larger. Thus, while the employment effects are not significant, the 16 We have also estimated a linear probability model in which we do not instrument for maternal employment and found that the parameter estimates in this model are virtually identical to the derivatives from the analogous probit model reported in Column 4 of Table 2. 18 fact that the pattern of the point estimates is similar to those reported earlier provides another indication that the earlier results were not driven primarily by omitted variable bias. However, our instruments do not allow enough power to reject that the true coefficient is zero. Overall, then, these results suggest that mothers who are more time constrained due to working more hours per week may have a more difficult time ensuring their children get nutritious meals and regular exercise. If anything, the point estimates from the difference and IV models suggest the probit models may provide an underestimate of the true effect. C. Estimates for Subgroups In this section, we present estimates of each of the earlier models by income quartile, mother’s education and race/ethnicity, respectively.17 In Table 5, we divide the sample by income (measured over the child’s lifetime) quartiles. Interestingly, the positive effect of more intensive employment is driven almost exclusively by those in the top quartile. In the probit models, the point estimate on hours per week is always positive, but increases with income quartile. Only the estimate for the highest quartile is statistically significant. For this group, the results indicate that a 10-hour increase in average hours worked per week (if working) since the child’s birth increases the likelihood that the child will be overweight by 1.3 percentage points. Turning to the estimates that control for unobserved fixed individual or family effects, we see that the impact of having a mother who works more intensively is always greatest for children in the highest income quartile. Consistent with our measurement error discussion above, the estimates for this group are largest for the long difference and “at the same time” estimates. Across all methods, however, the results for the highest income quartile are always positive, at least as big as the probit estimates, and statistically 17 We exclude the instrumental variables estimates because the estimates become even more imprecise with smaller subgroups. 19 significant in two out of three cases. For this group, there is no evidence that unobservable heterogeneity drives the relationship between hours worked per week and childhood overweight. Table 6 reports the results separating the sample by mother’s level of education. Here the impact of the intensity of maternal employment is consistently positive and significant for children of more educated mothers. Among these children, probit estimates indicate that if their mothers worked an additional 10 hours per week while working since they were born, they were 1.1 percentage points more likely to be overweight. Again, estimates obtained from long differences and sibling differences support this finding. Children of more highly educated mothers are significantly more likely to be overweight if their mothers work more hours per week in two of the three additional specifications; point estimates are larger for this group in all specifications that control for unobserved heterogeneity. The effect of hours per week for children of mothers who are high school dropouts and graduates are erratic and generally statistically insignificant. The effect of weeks worked is insignificant virtually throughout. We have also separated the sample by racial/ethnic group and report the results of these estimates in Table 7. In probit models we find that the overall effect of more intensive working appears to be driven by the experience of whites. Point estimates on hours worked per week in models estimated exclusively for whites using long differences and sibling differences are at least as big as the probit estimate, although only the long difference estimates provide sufficient precision to reject the null hypothesis of no effect. Estimates for other groups show no consistent pattern and coefficients on weeks worked are uniformly statistically insignificant. Overall, then the subgroup analysis shows that a measure of the intensity of mother’s work over the child’s lifetime has a positive effect on a child’s likelihood of being overweight if the child is in a high income family, with a well-educated, or white mother. This is consistent with time 20 constraints affecting these mother’s ability to supervise their child’s eating and exercise patterns. For these subgroups, a 10-hour increase in average hours worked per week over a child’s life is estimated to increase the likelihood that the child is overweight by between one and four percentage points, depending on the specification. Thus, a mother of this type moving from part-time (20 hours per week) work to full-time work (40 hours) is expected to increase the probability that her child is overweight by between 2 and 8 percentage points. VI. DISCUSSION Intuitively, one might have thought that higher socioeconomic status mothers would be those for whom working would matter least, because they are the mothers in the best position to purchase high quality child care in their absence. Instead, we find that it is only for this group that mother’s work matters, implying that when these mothers spend more time per week with their children, they are doing something that promotes a nutritious diet and exercise for their children. There are alternative interpretations of this finding. First, it may be difficult to find caregivers who have skills equal to those of these mothers. Absent direct information on childcare provider skills, however, it is impossible to determine if this is the case. Alternatively, it may be the case that lower socioeconomic status women are more time constrained whether or not they work. For these women, shadow prices of nutritious meals and exercise may be high. Changes in work patterns may not sufficiently shift the time constraint for us to observe changes on these margins. For example, if there are few grocery stores and safe places to play near one’s house and one faces transportation difficulties, then one may not have time to provide nutritious meals and active play time regardless of whether one works. To put the magnitude of our findings in context, we consider the extent to which the effect of mothers’ work can explain the increased prevalence of overweight among children over the past few 21 decades. First, recall that childhood overweight has increased across all income, race, and education groups. Since maternal work is only related to childhood overweight among relatively advantaged families, and because even when we control for a large number of variables we can only explain around 6 percent of the variation in childhood overweight, there are clearly other factors besides working mothers contributing to this epidemic. Here we examine how much of the increase in childhood overweight can be explained by increases in mothers’ average hours per week, for a subgroup where maternal work has an impact. This analysis is necessarily inexact because we must use several different data sets that cover slightly different time periods and use somewhat different data definitions. Since the exercise is simply for illustrative purposes, we conduct the analysis for just one subgroup: the top quartile of the income distribution. For this exercise, we used data from the March 1976 and March 1995 CPS to estimate the increase in hours worked per week over the past calendar year for women, 16 years or older, who had children under 18 living at home, in family’s with incomes in the top quartile of the income distribution. We also use data from the 1971-1975 and the 1988-1994 NHANES to estimate the change in the percentage of these children who are overweight.18 Average hours worked per week increased from 20.1 in 1975 to 27.2 in 1994 for this group. The results from our analyses above predict that this change (7.1) in average hours per week will lead to an increase in the probability of a child being overweight of between 0.9 to 2.7 percentage points.19 In 1976, the probability that a child in a top-income-quartile household was overweight was 2.1 percent. By 1994, this had risen to 9.9 percent. Thus, the probability that a child from one of these families was overweight increased by 7.8 percentage points. Based on these calculations, the increase in average hours worked by mothers 18 We describe the CPS, NHANES I and NHANES III data sources in the Data Appendix. See Appendix Table 3 for sample means of these data. 22 in high-income families can account for between 11.8 to 34.6 percent of the trend in the prevalence of childhood overweight for children in these families. VII. SUMMARY AND CONCLUSIONS The contribution of this work is several fold. First, much of the research on childhood overweight reports simple correlations between overweight and various characteristics of the child or the family. This research is among the first to grapple with issues of causality. It presents robust evidence of a positive and significant impact of maternal work on the probability that a child is overweight. It is not simply that mothers who work are those who would have overweight children regardless of their employment behavior. In addition, contrary to some other childhood outcomes, the effect of maternal work on childhood overweight is not sensitive to whether the mother works during a child’s early years of development. Further, this work presents prima facie evidence that the mechanism through which maternal employment affects childhood overweight is constraints on mother’s time; it is hours per week, not the number of weeks worked, that affects children’s probability of overweight. This result makes sense if it is the day-to-day routines that matter for a mother’s ability to supervise her child’s nutritional intake and energy expenditure. Working fewer hours per week allows more time for shopping, cooking, and energy expending play dates or organized sports. Finally, we show that it is important to examine explanations for childhood overweight separately for subgroups. Working more hours per week only has a deleterious effect on children in higher socioeconomic status households. While clearly there is much more to learn about causal factors related to the epidemic of overweight among children in the United States, this project lays the groundwork for future research 19 The estimated coefficient ranges between 0.013 and 0.038. Since average hours per week are in units of 10, we first multiply the coefficient by 10. Then we multiply this by the change in average hours per week. 23 into the causes of childhood overweight. Further work is needed in understanding the mechanisms through which mothers’ working translates into overweight children. For example, how does child care quality affect children’s nutrition and energy expenditure? A deeper understanding of other direct contributors to childhood overweight is also imperative. For example, we need to know more about children’s opportunities for vigorous exercise, including physical education in school, afterschool programs, and access to parks or other recreational facilities. This work demonstrates that it is critical to examine these contributors separately for different population subgroups. A deeper understanding is important if society is going to develop appropriate policy responses to this important public health issue. 24 References Averett, Susan and Sanders Korenman, “The Economic Reality of ‘The Beauty Myth’: Economic Differentials by Body Mass, “ Journal of Human Resources, vol. 31, no. 2, 1996, page 304. Bar-Or, Oded, John Foreyt, Claude Bouchard, Kelly D. Brownell, William H. Dietz, Eric Ravussin, Arline D. Salbe, Sandy Schwenger, Sachico St. Jeor, and Benjamin Torun, “Physical Activity, Genetic, and Nutritional Considerations in Childhood Weight Management,” Roundtable Discussion, The Official Journal of the American College of Sports Medicine, vol 30, no. 1 1998, pp. 2-10. Birch, Leann L., and Jennifer O. Fisher. "Development of Eating Behaviors Among Children and Adolescents." Pediatrics, vol. 101, no. 3, March 1998, pp. 539-549. Bouchard, Claude, “Obesity in Adulthood – the Importance of Childhood and Parental Obesity,” New England Journal of Medicine, vol. 337, no. 13, September 25, 1997, p. 926-927. Cawley, John. “Body Weight and Women’s Labor Market Outcomes.” NBER working paper, no. 7841. August 2000. Centers for Disease Control and Prevention, National Center for Health Statistics. Division of Health Examination Statistics. data available at: http://www.cdc.gov/nchs/products/pubs/pubd/hestats/overwght99.htm. Council of Economic Advisers. “Technical Report: Explaining the Decline in Welfare Receipt, 1993-1996. Washington, DC: White House Council of Economic Advisers. 1997. Dietz, William. “Factors Associated with Childhood Obesity.” Nutrition. Vol. 7, No. 4 (July-August 1991) pp. 290-1. Dietz, William, “Periods of Risk in Childhood for the Development of Adult Obesity – What Do We Need to Learn?,” Journal of Nutrition, vol. 127, Supplement, 1997, p. 1884-1886. Dietz, William H. and Mary C. Bellizzi. “Introduction: The Use of Body Mass Index to Assess Obesity in Children. American Journal of Clinical Nutrition. Vol. 70 (suppl). pp. 123S125S. Dietz, William H. and Steven L. Gortmaker. “Do We Fatten Our Children at the Television Set? Obesity and Television Viewing in Children and Adolescents.” Pediatrics Vol. 75 No. 5 (May 1985) pp. 807-12. Feder, Don. “Day-care Study Defies Common Sense.” Boston Herald (March 8, 1999), p. O25. 25 Freeman, J. V, and C. Power, and B. Rodgers. “Weight-for-Height Indices of Adiposity: Relationships with Height in Childhood and Early Adult Life.” International Journal of Epidemiology. Vol. 24, No. 5 (October 1995), pp. 970-976. Gerald LB, A. Anderson, GD Johnson GD, C Hoff and RF Trimm. “Social Class, Social Support and Obesity Risk in Children.” Child Care, Health & Development. Vol. 20 No. 3 (May-June 1994) pp. 145-63. Gilman, Matthew, Sheryl L. Rifas-Shiman, Carlos A. Camargo Jr., Catherine S. Berkey, A. Lindsay Frazier, Helaine R. H. Rockett, Alison E. Field, Graham A. Colditz, “Risk of Overweight Among Adolescents Who Were Breastfed as Infants,” Journal of the American Medical Association, May 16, 2001, Vol. 285 No. 19, pp.2461-2467. Gortmaker, Steven L., Aviva Must, James M. Perrin, Arthur M. Sobol, and William H. Dietz, “Social and Economic Consequences of Overweight in Adolescence and Young Adulthood, “ New England Journal of Medicine, vol. 329, September 1993, p.1008-12. Gortmaker Steven L., Aviva Must, Arthur M. Sobol, Karen Peterson, Graham A. Colditz, William H. Dietz. “Television Viewing as a Cause of Increasing Obesity among Children in the United States, 1986-1990.” Archives of Pediatrics & Adolescent Medicine Vol. 150 No. 4(April 1996) pp. 356-62. Greene, William H. Econometric Analysis. Englewood Cliffs, NJ: Prentice Hall. 1993. Griliches, Zvi and Jerry A. Hausman. “Errors in Variables in Panel Data.” Econometrics. Vol. 31, No. 1 (February 1986), pp. 93-118. Journal of Grimsley, Kristen Downey and R. H. Melton. “Homemakers Get Plenty of Support.” The Washington Post (March 22, 1998). p. A16. Hotz, V. Joseph and Rebecca Kilburn. “Regulating Child Care: The Effects of State Regulations on Child Care Demand and Its Cost.” Unpublished manuscript. 1996. Johnson, Rachel K., Helen Smiciklas-Wright, Ann C. Crouter and Fern K. Willits, "Maternal Employment and the Quality of Young Children's Diets - Empirical Evidence Based on the 1987-1988 Nationwide Food-consumption Survey" Pediatrics. Vol 90 No. 2 Part 1 (August 1992), pp. 245-249. Klesges Robert C., Risa J. Stein, Linda H. Eck, Terry R. Isbell and Lisa M. Klesges. “Parental Influence on Food Selection in Young Children and its Relationships to Childhood Obesity” American Journal of Clinical Nutrition. Vol. 53 No. 4 (April 1991) pp. 859-64. 26 Locard, Elisabeth, Nicole Mamelle, Agathe Billette, Michell Migniac, Francoise Munoz, and Sylvie Rey, “Risk Factors of Obesity in a Five Yar Old Population: Parental versus Environmental Factors,” International Journal of Obesity, vol 16, p. 721-729, 1992. Mashberg, Tom. "Childhood Obesity Reaches Epidemic Proportions." Boston Herald. May 23, 1999. p. 1. National Center for Health Statistics. Health, United States, 1998 with Socioeconomic Status and Health Chartbook. Hyattsville, Maryland: 1998. National Institutes of Health. Clinical Guidelines on the Identification, Evaluation and Treatment of Overweight in Adults: The Evidence Report. June 1998. Available from the National Heart, Lung and Blood Institute web site at http://www.nhlbi.nih.gov/guidelines/obesity/ob_gdlns.pdf Nelson, C.A. “How important are the first 3 years of life?” Applied Developmental Science, 3, 235238, 1999. Power, C., J.K. Lake, and T.J. Cole, “Measurement and Long-term Health Risks of Child and Adolescent Fatness,” International Journal of Obesity, vol 21, 1997, p.507-526. Shonkoff, Jack P. and Deborah A. Phillips, Editors, From Neurons to Neighborhoods: The Science of Early Childhood Development, National Academy Press, Washington DC, 2000. Sobal, Jeffery and Albert J. Stunkard, “Socioeconomic Status and Obesity: A Review of the Literature,” Psychological Bulletin, vol. 105, no. 2, 1989, p.260-275. Takahashi, E., K. Yoshida, H. Sugimori, M. Miyakawa, T. Izuno, T. Yamagami, S. Kagamimori. “Influence Factors on the Development of Obesity in 3-year-old Children Based on the Toyama Study.” Preventive Medicine, vol. 28, no. 3 (March 1999) pp. 293-6. Thompson, “Diabetes: Childhood Obesity Hastens Type II Onset,” Health Line (The National Journal Group, Inc.), December 4, 1998, section: Trends and Timelines. United States Bureau of the Census. Statistical Abstract of the United States 2000. Washington, DC: Government Printing Office. 2000. von Kries Rudiger, Berthold Koletzko, Thorsten Sauerwald , Erika von Mutius, Dietmar Barnert, Veit Grunert, Hubertus von Voss. “Breast Feeding and Obesity: Cross Sectional Study” British Medical Journal. Vol. 319 (7203) (July 17, 1999) pp. 147-150. Vuille, Jean-Claude and Tore Mellbin. “Obesity in 10-year-olds: an Epidemiologic Study.” Pediatrics. Vol. 64, No. 5 (November 1979), pp. 564-72. 27 Weil, Andrew. Interview with Paula Zahn, CNN.com/Health, May 28, 2002. Whitaker, Robert C., Jeffrey A. Wright, Margaret S. Pepe, Kristy D. Seidel, and William H. Dietz. “Predicting Obesity in Young Adulthood from Childhood and Parental Obesity.” The New England Journal of Medicine. Vol. 337, No. 13 (September 25, 1997), pages 869-873. Wolfe Wendy S., Cathy C. Campbell, Edward A. Frongillo Jr, Jere D. Haas and Thomas A. Melnik. “Overweight Schoolchildren in New York State: Prevalence and Characteristics.” American Journal of Public Health. Vol. 84 No. 5 (May 1994) pp. 807-13. Woolston, Joseph L., “Obesity in Infancy and Early Childhood,” Journal of the American Academy of Child Adolescent Psychiatry,” Special Article, vol. 26, 1987, p.123-126. Zimmerman, David J., “Regression Toward Mediocrity in Economic Stature.” The American Economic Review, Vol. 82, No. 3 (June 1992), pp. 409-429. 28 Table 1: Rates of Overweight for Children Age 3 to 11 in the NLSY, by Maternal Employment and Socioeconomic Status Mother Worked 35 Hours/Week Since Birth All Mother Never Worked Mother Worked < 35 Hours/Week Since Birth 10.6 9.4 10.1 12.9 1st Quartile 12.4 13.3 11.4 13.0 2nd Quartile 11.1 8.5 11.0 11.5 3rd Quartile 11.7 12.1 11.0 12.2 4th Quartile 8.5 3.2 7.3 10.6 Less than High School Degree 12.8 13.0 14.3 11.3 High School Degree 10.7 8.1 10.2 11.7 Some College or More 9.5 7.9 7.6 11.6 Hispanic 13.3 17.0 13.2 12.5 Black, non-Hispanic 15.0 11.7 14.2 16.0 White, non-Hispanic 9.5 7.6 9.0 10.3 All By Income Quartile Since Birth By Maternal Education By Race/Ethnicity Notes: Hours per week relate to weeks in which some work occurred. Sampling weights are used to provide nationally representative estimates. Table 2: Probit Estimates of the Impact of Maternal Employment Since Child was Born on Childhood Overweight (Estimates Represent Derivatives, Robust Standard Errors in Parentheses) Average Hours per Week if Working Since Child’s Birth (in units of 10) (1) 0.012 (0.002) (2) 0.008 (0.003) (3) 0.007 (0.002) (4) 0.007 (0.003) Number of Weeks Worked Since Child’s Birth (in units of 52) -0.003 (0.001) -0.001 (0.002) -0.001 (0.002) -0.001 (0.002) Black, non-Hispanic 0.051 (0.012) 0.035 (0.005) 0.024 (0.011) Hispanic 0.016 (0.013) 0.009 (0.007) 0.011 (0.013) Mother’s Highest Grade Completed -0.006 (0.002) -0.005 (0.002) -0.004 (0.002) Child Was First Born -0.010 (0.009) -0.007 (0.004) -0.008 (0.008) Number of Children -0.010 (0.004) -0.009 (0.004) -0.010 (0.004) Child Was Breast Fed -0.020 (0.008) -0.018 (0.008) Mother’s BMI 25 (Overweight or Obese) 0.034 (0.008) 0.031 (0.009) Mother’s BMI 30 (Obese) 0.047 (0.011) 0.039 (0.012) Average Family Income Since Birth (in units of $10,000) -0.002 (0.002) Percent of Child’s Life Mother was Married -0.013 (0.011) Psuedo R-Squared 0.020 0.046 0.059 0.060 Number of Observations 16,650 16,650 16,650 16,650 10.6 10.6 10.6 10.6 3 to 11 3 to 11 3 to 11 3 to 11 % of Children Overweight in Sample Age of Children in Sample Notes: The dependent variable is a binary variable equal to 1 if child’s BMI is above the 95th percentile for his/her age and sex. The standard errors are robust, clustered on the mother’s identification code, as there are multiple children per mother over time. All columns include dummies for mother reported height and weight. Columns 2-4 also include child’s birth weight, mother’s afqt score, both the child’s and mother’s age in years, dummy variables for the year of the survey, controls for education levels of the mother’s parents, dummy variables indicating whether mother’s parents were present when she was 14, and a dummy variable indicating whether the child is female. All estimates are weighted using the child’s sampling weight. Table 3: Probit Estimates of the Impact of Maternal Employment at Different Points in Child’s Life on Childhood Overweight (Estimates Represent Derivatives, Robust Standard Errors in Parentheses) Average Hours per Week if Working Since Child’s Birth (in units of 10) (1) 0.007 (0.003) (2) 0.005 (0.003) (3) 0.008 (0.004) Number of Weeks Worked Since Child’s Birth (in units of 52) -0.001 (0.002) 0.003 (0.003) -0.001 (0.002) (4) (5) 0.007 (0.003) -0.001 (0.003) Average Hours per Week if Working in Child’s First 3 Years of Life (in units of 10) 0.003 (0.003) 0.0003 (0.003) Number of Weeks Worked in Child’s First 3 Years of Life (in units of 52) 0.001 (0.004) 0.002 (0.005) Psuedo R-Squared 0.059 0.069 0.073 0.059 0.060 Number of Observations 16,650 6,565 10,085 16,650 16,650 % Overweight in Sample 10.6 10.9 10.4 10.6 10.6 3 to 11 3 to 5 6 to 11 3 to 11 3 to 11 Age of Children in Sample Notes: All specifications include the same covariates as Column 4 of Table 2. See notes to that table. Table 4: Alternative Methods to Control for Unobservable Heterogeneity in the Impact of Maternal Employment on Childhood Overweight (Robust Standard Errors in Parentheses) Average Hours per Week Since Child’s Birth (in units of 10) Individual Long Difference (1) 0.015 (0.007) Sibling Differences at Same Time (2) 0.009 (0.008) Sibling Differences at Same Age (3) 0.008 (0.006) Instrumental Variables (4) 0.009 (0.024) Number of Weeks Worked Since Child’s Birth (in units of 52) -0.005 (0.004) -0.006 (0.005) 0.003 (0.006) 0.004 (0.016) Mother’s Highest Grade Completed -0.028 (0.011) — 0.026 (0.013) -0.005 (0.003) Child Was First Born — 0.002 (0.013) 0.014 (0.018) -0.007 (0.009) Number of Children -0.015 (0.012) — 0.019 (0.013) -0.007 (0.008) Child Was Breast Fed — 0.012 (0.017) -0.029 (0.022) -0.016 (0.010) Mother’s BMI 25 (Overweight or Obese) 0.006 (0.020) — -0.016 (0.020) 0.031 (0.010) Mother’s BMI 30 (Obese) 0.046 (0.023) — 0.023 (0.024) 0.049 (0.014) Average Family Income Since Birth (in units of $10,000) -0.001 (0.001) 0.007 (0.008) -0.018 (0.005) -0.003 (0.003) Percent of Child’s Life Mother was Married 0.018 (0.048) -0.025 (0.056) 0.016 (0.038) -0.014 (0.015) — -0.0004 (0.004) 0.002 (0.006) 0.009 (0.003) R-Squared 0.016 0.011 0.019 0.041 Number of Observations 4,159 7,919 4,775 15,050 Birth Weight (in pounds) Age of Children in Sample 3 to 11 3 to 11 3 to 11 3 to 11 Notes: Column 1: The dependent variable represents the difference between the last and first observation for each child in a binary variable equal to 1 if child’s BMI is above the 95th percentile for his/her age and sex. Other control variables include differences in: whether the mother reported the child’s height and weight, mother’s level of education, age, whether the mother was overweight or obese, the number of children in the family, income since birth, and mother’s marital status since birth. Estimates are computed using OLS and are weighted using the child’s sampling weight. The standard errors are robust, clustered on mother’s identification code as there are multiple observations in each household over time. Column 2: The dependent variable represents the difference between siblings at the same point in time in a binary variable equal to 1 if child’s BMI is above the 95th percentile for his/her age and sex. Other control variables include differences in: the child’s sex and age, whether the mother reported the child’s height and weight, whether the child was breastfed, whether the child was firstborn, birth weight, income since birth, mother’s marital status since birth, and year fixed effects. Estimates are computed using OLS and are weighted using the child’s sampling weight. The standard errors are robust, clustered on a sibling pair identification code, as there are multiple observations for each sibling pair. Column 3: The dependent variable represents the difference between siblings at the same age in a binary variable equal to 1 if child’s BMI is above the 95th percentile for his/her age and sex. Other control variables include differences in: the child’s sex, the years between the two surveys, the number of children in the family, mother’s level of education, whether or not the mother was overweight or obese, whether the child was breastfed, whether the child was first born, income since birth, mother’s marital status since birth, and whether the mother reported the child’s height and weight. Estimates are computed using OLS and are weighted using the child’s sampling weight. The standard errors are robust, clustered on a sibling pair identification code, as there are multiple observations for each sibling pair. Column 4: The dependent variable is a binary variable equal to 1 if child’s BMI is above the 95th percentile for his/her age and sex. The same additional explanatory variables are included as in Column 4 of Table 1. The standard errors are robust, clustered on mother’s identification code, as there are multiple observations in each household over time. Estimates are weighted using the child’s sampling weight. Table 5: Probit Estimates of the Impact of Maternal Employment Since Child was Born on Childhood Overweight by Average Family Income Since Birth Quartiles (Estimates Represent Derivatives, Robust Standard Errors in Parentheses) Percent Overweight in Sample 1st Quartile (1) 2nd Quartile (2) 3rd Quartile (3) 4th Quartile (4) 12.4 11.1 11.7 8.5 PROBIT Average Hours per Week if Working Since Child’s Birth (in units of 10) 0.001 (0.004) 0.003 (0.005) 0.004 (0.006) 0.013 (0.005) Number of Weeks Worked Since Child’s Birth (in units of 52) -0.001 (0.005) 0.001 (0.004) -0.005 (0.004) 0.001 (0.003) 4,161 4,165 4,161 4,163 Number of Observations INDIVIDUAL LONG DIFFERENCES Average Hours per Week if Working Since Child’s Birth (in units of 10) 0.001 (0.010) -0.001 (0.011) 0.025 (0.014) 0.035 (0.017) Number of Weeks Worked Since Child’s Birth (in units of 52) 0.001 (0.011) -0.017 (0.009) -0.008 (0.009) -0.0004 (0.007) 1,040 1,040 1,040 1,039 Number of Observations SIBLING DIFFERENCES -SAME TIME Average Hours per Week if Working Since Child’s Birth (in units of 10) -0.011 (0.011) -0.013 (0.013) -0.004 (0.011) 0.038 (0.013) Number of Weeks Worked Since Child’s Birth (in units of 52) 0.021 (0.020) 0.005 (0.013) -0.020 (0.011) 0.003 (0.008) 1,980 1,980 1,980 1,979 Number of Observations SIBLING DIFFERENCES - SAME AGE Average Hours per Week if Working Since Child’s Birth (in units of 10) 0.012 (0.010) 0.001 (0.013) 0.002 (0.013) 0.014 (0.011) Number of Weeks Worked Since Child’s Birth (in units of 52) 0.012 (0.018) 0.009 (0.010) 0.010 (0.013) -0.003 (0.010) 1,118 1,118 1,118 1,117 Number of Observations Notes: Probit estimates are obtained from models comparable to Table 2, Column 4. Long difference and sibling difference estimates are obtained from models comparable to Table 4, Columns 1 through 3. See the notes to those tables. In all specifications, children are between the ages of 3 and 11. Table 6: Probit Estimates of the Impact of Maternal Employment Since Child was Born on Childhood Overweight by Mother’s Education (Estimates Represent Derivatives, Robust Standard Errors in Parentheses) Some College/ High School Dropout High School Graduate College Grad (1) (2) (3) 12.8 Percent Overweight in Sample 10.7 9.5 PROBIT Average Hours per Week if Working Since Child’s Birth (in units of 10) -0.007 (0.005) 0.009 (0.004) 0.011 (0.005) Number of Weeks Worked Since Child’s Birth (in units of 52) -0.004 (0.004) 0.001 (0.003) -0.003 (0.003) 3,106 8,169 5,375 Number of Observations INDIVIDUAL LONG DIFFERENCES Average Hours per Week if Working Since Child’s Birth (in units of 10) -0.010 (0.013) 0.016 (0.010) 0.029 (0.011) Number of Weeks Worked Since Child’s Birth (in units of 52) 0.005 (0.012) -0.003 (0.006) -0.009 (0.006) 731 1,996 1,432 Number of Observations SIBLING DIFFERENCES - SAME TIME Average Hours per Week if Working Since Child’s Birth (in units of 10) -0.016 (0.010) -0.011 (0.009) 0.044 (0.012) Number of Weeks Worked Since Child’s Birth (in units of 52) -0.001 (0.024) -0.011 (0.007) 0.007 (0.009) 1,888 3,675 2,356 Number of Observations SIBLING DIFFERENCES - SAME AGE Average Hours per Week if Working Since Child’s Birth (in units of 10) 0.017 (0.011) 0.006 (0.008) 0.012 (0.015) Number of Weeks Worked Since Child’s Birth (in units of 52) -0.037 (0.016) 0.010 (0.009) 0.005 (0.008) 1,130 2,108 1,233 Number of Observations Notes: Probit estimates are obtained from models comparable to Table 2, Column 4. Long difference and sibling difference estimates are obtained from models comparable to Table 4, Columns 1 through 3. See the notes to those tables. In all specifications, children are between the ages of 3 and 11. Table 7: Probit Estimates of the Impact of Maternal Employment Since Child was Born on Childhood Overweight by Racial/Ethnic Group (Estimates Represent Derivatives, Robust Standard Errors in Parentheses) Hispanic (1) Black (non-Hispanic) (2) White (non-Hispanic) (3) 13.3 15.0 9.5 Average Hours per Week if Working Since Child’s Birth (in units of 10) -0.011 (0.006) 0.005 (0.005) 0.008 (0.003) Number of Weeks Worked Since Child’s Birth (in units of 52) 0.001 (0.004) -0.005 (0.003) 0.0002 (0.002) 2,946 4,959 8,745 Average Hours per Week if Working Since Child’s Birth (in units of 10) 0.023 (0.017) 0.001 (0.010) 0.019 (0.008) Number of Weeks Worked Since Child’s Birth (in units of 52) -0.0004 (0.010) 0.002 (0.007) -0.007 (0.005) 741 1,265 2,153 Average Hours per Week if Working Since Child’s Birth (in units of 10) -0.019 (0.013) 0.001 (0.013) 0.013 (0.010) Number of Weeks Worked Since Child’s Birth (in units of 52) 0.005 (0.011) 0.007 (0.011) -0.009 (0.007) 1,696 2,417 3,806 Average Hours per Week if Working Since Child’s Birth (in units of 10) -0.022 (0.016) 0.014 (0.011) 0.010 (0.008) Number of Weeks Worked Since Child’s Birth (in units of 52) 0.020 (0.016) -0.007 (0.012) 0.004 (0.008) 991 1,361 2,119 Percent Overweight in Sample Number of Observations Number of Observations Number of Observations Number of Observations Notes: Probit estimates are obtained from models comparable to Table 2, Column 4. Long difference and sibling difference estimates are obtained from models comparable to Table 4, Columns 1 through 3. See the notes to those tables. In all specifications, children are between the ages of 3 and 11. DATA APPENDIX Our primary data source is the National Longitudinal Survey of Youth 1979 (NLSY); its main characteristics are described in the text. This Appendix provides additional detail on these data as well as some supplemental data used in our analysis. We briefly describe the data used from the NHANES I and III data along with the CPS. Our key measure of whether a child is overweight is based on the child’s body mass index (BMI). BMI is defined as weight in kilograms divided by height in meters squared (kg/m2) and is a commonly used measure to define obesity and overweight in adults. According to guidelines in National Institutes of Health (1998), adults are considered underweight if their BMI is less than 18.5, overweight if their BMI is 25 or more, and obese if their BMI is 30 or more. Use of the BMI to assess children has been more controversial, although its use is fairly widespread.20 The Centers for Disease Control (CDC) has recently endorsed the use of BMI to assess overweight in children, and has produced sex-specific BMI charts for children aged 2 to 20 for just this purpose.21 We use these charts to define overweight cutoffs for children in our samples. For children, however, the nomenclature is somewhat different than for adults. Children with a BMI above the 85th percentile of the BMI distribution for their sex-age group are defined as “at-risk of overweight;” those with a BMI above the 95th percentile for this distribution are termed “overweight.” It is important to note that these percentile cutoffs are based mainly on data from 20 Ideally, one would prefer to measure overweight using a measure that reflects adiposity. Since it is impractical to do so in large scale surveys, researchers have employed the BMI, which only requires the measurement of height and weight. It is somewhat controversial when used to assess overweight among children because children experience growth spurts at individual-dependent ages and this can weaken the relationship between height and weight-based measures to adiposity. See Freeman, et al. (1995) and Whitaker, et al. (1997) for a discussion of the use of BMI in children. Recently, Dietz and Bellizi (1999) reporting on a conference convened by the International Obesity Task Force, noted that the BMI “offered a reasonable measure with which to assess fatness in children and adolescents.” Additionally, they conclude that a BMI above the 85th percentile for a child’s age and sex group is likely to accord with the adult definition of overweight, and above the 95th percentile with the adult definition of obese. 21 See http://www.cdc.gov/growthcharts/ for general information, and see http://www.cdc.gov/nchs/about/major/nhanes/growthcharts/bmiage.txt for specific BMI percentiles. years before our survey began, so that trends in overweight can be detected.22 Not every child in our sample is actually weighed and measured. Rather, about 15 percent have mother reported weight and 22 percent have mother reported height. All of our models included indicators for these mother reports, since they are more likely to result in errors in BMI, and hence in the classification of overweight status. Our key explanatory variables are measures of the child’s exposure to maternal employment at different periods of life. To create these variables, we use employment history data available in each year of the survey.23 The starting and ending dates of up to 5 jobs that the mother has held since the last interview are recorded in each survey. For each of these jobs, the usual hours worked per week or per day are recorded. For each job worked, the starting and ending dates of up to 4 periods of unpaid leave are also recorded. Additionally, each survey year contains created variables equal to the total number of hours and weeks worked since the last interview.24 We begin with the child’s birth date, and then look week-by-week at whether that date is between the start and end dates of reported job. If so, then the week is coded as being at work, and the usual hours per week are added to the total number of weeks worked.25 If however, this week is between the start and ending date of a leave period, the week and hours are subtracted back off. This accumulation of weeks and hours is continued until the first interview date after the child’s birth. For each successive survey, the NLSY-created number of weeks and hours since the last interview are added on to our cumulative measure to create weeks and hours worked since birth as of that survey. The variable labeled “Average Hours per Week if Working Since Child’s Birth” is then created by dividing hours worked by 22 These percentile cutoffs refer to the BMI distribution from an earlier period in order to provide a fixed standard for assessing overweight. Thus while the new CDC growth charts covering a large number of developmental markers are based on data from 1963-1994, the 1988-1994 data from NHANES III is generally excluded from the BMI chart. Prior to the release of these charts, percentiles based entirely on NHANES I from 1971-1974 had been available for older children. The newly released cutoffs are similar. 23 Only children born after the start of the survey are included in the sample. 24 These created variables occasionally cover slightly less than 100% of the elapsed time period. However, since our key variable is hours divided by weeks, even for these individuals our measure should be highly correlated with the true lifetime average. weeks worked, trimmed at 80 to avoid outliers.26 If weeks worked is equal to zero, the hours per week variable is set to zero. We also continue with the week-by-week formation of hours and weeks worked for each year up to age 6. This allows us to create similar average hours per week measures for weeks worked in the first year of life, second year, first three years, first six years, etc. in order to investigate timing issues. The NLSY data must be modified to create the long differences and sibling differences used in estimation and described in the text. To create long differences, we start with our analysis sample and subtract the first observation for a child from the last observation, to obtain one difference per child. The median (and modal) observation looks back 3 surveys (and thus 6 years) to create the long difference. Sibling differences are similar except rather than creating differences over time, the differences are within family. For these models, each child-year observation is joined pair wise with child-year observations for all available siblings in either the same year, or at the same age. Differences are then created across these pairs. Note that both siblings must be within our age range. To estimate instrumental variable models, we need to supplement the NLSY data with outside data that we use as instruments. The instruments we use to predict maternal employment are state child care regulations, wages of child care workers, welfare benefit levels, the status of welfare reform in the state, and the annual unemployment rate in the state. The state child care regulations were generously provided by Joe Hotz and Rebecca Kilburn, who used these data in Hotz and Kilburn (1996). These regulation data include indicators for requirements for liability insurance, for training beyond high school for directors, and for more than one inspection per year. We also average the maximum caregiver/child ratio at each of 6 ages (0-11 months, 12-23, 24-35, 36-47, 48-59 and 60+) to create one variable. Each type of regulation is reported separately for center-based care and for family-based care. 25 26 If instead usual hours per day are reported, that number is multiplied by 5 to obtain weekly hours worked. Just 28 observations (less than 0.2 percent of the analysis sample) are recoded in this step. In addition to these regulation variables, the hourly wage of workers in the child care sector is calculated for each state from the CPS’s monthly outgoing rotation group datafile available from the National Bureau of Economic Research. Two variables relating to the state welfare system are also employed as instruments. These variables include the maximum AFDC benefit level in the state for a family of three in a given year and an indicator variable for whether a state in a given year had a waiver to implement pre-TANF reforms. They are obtained from the data file used in Council of Economic Advisers (1997) and are described there. A final instrument is the state unemployment rate for each year, obtained from the Bureau of Labor Statistics. The source data for the instruments is available annually starting in 1983 for the regulation variables and in 1979 for all other variables. Our instruments are created as a weighted average value over the child's lifetime (or since 1983 for the regulations). The results of the first stage regressions using these instruments are shown in Appendix Table 2. Information from the NHANES I and NHANES III along with data from the CPS are used at the end of the paper to simulate the extent to which increased maternal employment can explain the increase in the rate of childhood overweight. Estimates from these sources used as inputs to that simulation are reported in Appendix Table 3. The NHANES I and NHANES III were conducted in 1971-1975 and 1988-1994, respectively. Each is a national survey that over-represents certain population subgroups, but provides sampling weights to generate national representative estimates. Weight and height are physically measured for each survey respondent. In these data, we calculate income quartiles using the categorical measures of income in the past calendar year. A comparison of rates of overweight by race over time is complicated by changes in racial/ethnic categories over time. In particular, the earlier survey did not separate Hispanics from whites and blacks and the later survey separates out “Mexican-Americans” rather than all Hispanics. An examination of maternal education is hindered by difficulties in linking children with mothers in these data. Instead, we use the educational attainment of the head of the household. We also used data from the March 1976 and March 1995 CPS. These data provide information on the work behavior of respondents in the previous calendar years. We chose the 1976 and 1995 surveys because they provide employment data for the 1975 and 1994 calendar years, which correspond to the final years of the two NHANES surveys. To measure maternal employment we used the employment patterns of all women age 16 and over who have a child under the age of 18 living in their home. Income quartiles are defined based on a continuous income measure of the past calendar year. For educational attainment in 1976, we defined high school dropouts to be those with less than 12 years of schooling, high school graduates to be those with 12 years of schooling, and some college as those with more than 12 years of schooling. In 1995 these categories are obtained directly. Race and ethnicity is defined consistently across the two surveys. Appendix Table 1: Means of Variables Included in Regression Analyses (Standard Deviations in Parentheses) Ages 3-11 Ages 3-5 Overweight 0.106 0.109 (0.308) (0.311) Average Hours per Week if Working 2.968 2.799 Since Child’s Birth (in units of 10) (1.491) (1.640) Number of Weeks Worked 3.774 2.373 Since Child’s Birth (in units of 52) (2.894) (1.792) African American 0.154 0.141 (0.361) (0.348) Hispanic 0.066 0.062 (0.249) (0.242) Mother’s Highest Grade Completed 12.666 12.878 (2.058) (2.116) Mother’s AFQT Score 43.602 45.776 (27.198) (27.477) Child Was First Born 0.480 0.444 (0.500) (0.497) Number of Children 2.493 2.336 (1.066) (1.023) Child Was Breast Fed 0.521 0.547 (0.500) (0.498) Mother’s BMI 25 0.411 0.386 (Overweight or Obese) (0.492) (0.487) Mother’s BMI 30 0.166 0.147 (Obese) (0.372) (0.354) Average Family Income 40.224 42.827 Since Birth (in units of $10,000) (26.550) (29.080) Percent of Child’s Life 0.721 0.754 Mother was Married (0.367) (0.367) Mother Reported Child’s Weight 0.219 0.228 (0.413) (0.420) Mother Reported Child’s Height 0.149 0.148 (0.356) (0.356) Child’s Birth Weight in Pounds 7.410 7.443 (1.307) (1.330) Hours per Week Trimmed at 80 0.001 0.002 (0.039) (0.049) Child’s Age in Years 6.676 4.031 (2.478) (0.812) Child is Female 0.483 0.483 (0.500) (0.500) Mother’s Age in Years 31.537 30.416 (3.612) (3.700) Mother’s Mother’s Highest Grade 10.363 10.616 (3.851) (3.792) Mother’s Father’s Highest Grade 8.298 8.676 (5.833) (5.869) Mother’s Mother Present 0.932 0.938 When She was 14 (0.253) (0.241) Mother’s Father Present when 0.735 0.750 She was 14 (0.441) (0.433) Year 1 0.083 0.161 (0.276) (0.367) Year 2 0.139 0.180 (0.346) (0.383) Year 3 0.183 0.175 (0.387) (0.380) Year 4 0.200 0.176 (0.400) (0.381) Year 5 0.206 0.182 (0.405) (0.385) Number of Observations 16650 6565 Notes: All estimates are weighted using the child’s sampling weight. Ages 6-11 0.104 (0.305) 3.075 (1.377) 4.671 (3.102) 0.162 (0.368) 0.069 (0.254) 12.531 (2.008) 42.210 (26.928) 0.503 (0.500) 2.593 (1.082) 0.504 (0.500) 0.427 (0.495) 0.178 (0.382) 38.557 (24.652) 0.699 (0.365) 0.213 (0.409) 0.150 (0.357) 7.388 (1.292) 0.001 (0.030) 8.205 (1.686) 0.483 (0.500) 32.254 (3.364) 10.202 (3.879) 8.055 (5.191) 0.927 (0.260) 0.725 (0.446) 0.033 (0.179) 0.113 (0.316) 0.188 (0.391) 0.215 (0.411) 0.222 (0.416) 10085 Appendix Table 2: First Stage Regressions for the IV Estimates Reported in Table 4 (Robust Standard Errors in Parentheses) Mother’s Work: Average Hours per Week Since Child’s Birth Maximum AFDC/TANF Benefit for a Family of 3 Pre-TANF Welfare Reform Implemented (Waiver State) Years of Education Required for Director of Day Care Center Years of Education Required for Family Day Care Provider Average caregiver/child Ratio, Center based Average caregiver/child Ratio, Family based Is Center Required to Carry Liability Insurance Is Family Care Required to Carry Liability Insurance Number of Annual Inspections Conducted by Licensing Agency, Center Based Number of Annual Inspections Conducted by Licensing Agency, Family based More than One Inspection per Year, Center based More than One Inspection per year, Family based Training Beyond H.S. Required for Director, Center based Training Beyond H.S. Required for Director, Family based Hourly Wage of Workers in Child Care Sector Unemployment Rate Mother Reported Child’s Weight Mother Reported Child’s Height Child’s Birth Weight in Pounds Black Hispanic Child’s Age in Years Child is Female Mother’s Highest Grade Completed Mother’s AFQT Score Mother’s Age in Years Mother’s BMI 25 (Overweight or Obese) Mother’s BMI 30 (Obese) Mother’s Mother’s Highest Grade Mother’s Father’s Highest Grade Mother’s Mother Present When She was 14 Mother’s Father Present when She was 14 Child Was Breast Fed Child Was First Born -0.00006 (0.0003) -0.336 (0.275) -0.003 (0.010) 0.033 (0.020) 0.050 (0.016) 0.022 (0.026) -0.105 (0.099) 0.180 (0.201) -0.040 (0.115) -0.024 (0.034) 0.332 (0.269) 0.284 (0.176) -0.058 (0.108) 8.702 (2.207) -0.005 (0.063) -0.077 (0.023) -0.085 (0.054) -0.013 (0.065) 0.013 (0.022) 0.176 (0.088) 0.172 (0.115) 0.084 (0.012) 0.014 (0.046) 0.014 (0.023) 0.002 (0.002) -0.046 (0.016) 0.116 (0.059) -0.073 (0.081) -0.003 (0.016) -0.033 (0.014) 0.040 (0.203) 0.327 (0.158) -0.130 (0.062) -0.094 (0.047) Mother’s Work: Average Weeks per Year Since Child’s Birth 0.0004 (0.0005) -1.392 (0.407) 0.021 (0.017) -0.032 (0.034) 0.120 (0.029) 0.083 (0.049) -0.247 (0.174) 0.196 (0.325) -0.632 (0.184) 0.283 (0.149) 1.328 (0.433) -0.353 (0.291) 0.028 (0.181) -6.109 (6.168) -0.162 (0.106) -0.105 (0.039) -0.075 (0.094) -0.067 (0.112) 0.070 (0.035) 0.501 (0.141) 0.371 (0.174) 0.571 (0.022) 0.210 (0.078) 0.078 (0.035) 0.014 (0.003) 0.030 (0.026) 0.056 (0.101) -0.267 (0.134) 0.037 (0.025) -0.039 (0.022) -0.438 (0.324) 0.692 (0.251) -0.283 (0.108) -0.251 (0.076) Number of Children Average Family Income Since Birth (in units of $10,000) Percent of Child’s Life Mother was Married Year 1 Year 2 Year 3 Year 4 Year 5 R-squared Number of Observations -0.297 (0.038) 0.003 (0.002) -0.200 (0.096) -0.424 (0.225) -0.343 (0.177) -0.310 (0.135) -0.244 (0.096) -0.110 (0.058) 0.104 15050 -0.623 (0.056) 0.015 (0.003) 0.090 (0.160) -0.884 (0.355) -0.837 (0.288) -0.778 (0.224) -0.501 (0.161) -0.301 (0.102) 0.378 15050 Appendix Table 3: Rates of Overweight in NHANES I and NHANES III and Hours Worked per Week by Mothers in the March 1976 and March 1995 CPS by Socioeconomic Status Rates of Overweight Average Hours Worked Per Week NHANES I (1971-1975) NHANES III (1988-1994) March 1976 CPS March 1995 CPS 4.5 10.3 17.9 23.9 5.7 14.9 15.3 17.2 2 Quartile 4.2 9.6 17.4 24.6 3rd Quartile 5.6 8.8 18.6 26.5 4th Quartile 2.1 9.9 20.1 27.2 Less than High School Degree 4.9 12.0 13.8 13.4 High School Degree 5.2 12.0 19.7 26.9 Some College or More 3.0 8.1 20.6 27.8 Hispanic --- --- 16.2 19.8 Black 4.4 12.9 19.6 24.0 White 4.5 10.1 17.7 24.7 All By Income Quartile 1st Quartile nd By Maternal Education By Race/Ethnicity Notes: Income quartiles are created based on categorical measures of family income in the preceding calendar year. In NHANES I, blacks and whites include Hispanics, who could be of either race, but in NHANES III these categories represent non-Hispanics. Working Paper Series A series of research studies on regional economic issues relating to the Seventh Federal Reserve District, and on financial and economic topics. Extracting Market Expectations from Option Prices: Case Studies in Japanese Option Markets Hisashi Nakamura and Shigenori Shiratsuka WP-99-1 Measurement Errors in Japanese Consumer Price Index Shigenori Shiratsuka WP-99-2 Taylor Rules in a Limited Participation Model Lawrence J. Christiano and Christopher J. Gust WP-99-3 Maximum Likelihood in the Frequency Domain: A Time to Build Example Lawrence J.Christiano and Robert J. Vigfusson WP-99-4 Unskilled Workers in an Economy with Skill-Biased Technology Shouyong Shi WP-99-5 Product Mix and Earnings Volatility at Commercial Banks: Evidence from a Degree of Leverage Model Robert DeYoung and Karin P. Roland WP-99-6 School Choice Through Relocation: Evidence from the Washington D.C. Area Lisa Barrow WP-99-7 Banking Market Structure, Financial Dependence and Growth: International Evidence from Industry Data Nicola Cetorelli and Michele Gambera WP-99-8 Asset Price Fluctuation and Price Indices Shigenori Shiratsuka WP-99-9 Labor Market Policies in an Equilibrium Search Model Fernando Alvarez and Marcelo Veracierto WP-99-10 Hedging and Financial Fragility in Fixed Exchange Rate Regimes Craig Burnside, Martin Eichenbaum and Sergio Rebelo WP-99-11 Banking and Currency Crises and Systemic Risk: A Taxonomy and Review George G. Kaufman WP-99-12 Wealth Inequality, Intergenerational Links and Estate Taxation Mariacristina De Nardi WP-99-13 Habit Persistence, Asset Returns and the Business Cycle Michele Boldrin, Lawrence J. Christiano, and Jonas D.M Fisher WP-99-14 Does Commodity Money Eliminate the Indeterminacy of Equilibria? Ruilin Zhou WP-99-15 A Theory of Merchant Credit Card Acceptance Sujit Chakravorti and Ted To WP-99-16 1 Working Paper Series (continued) Who’s Minding the Store? Motivating and Monitoring Hired Managers at Small, Closely Held Firms: The Case of Commercial Banks Robert DeYoung, Kenneth Spong and Richard J. Sullivan WP-99-17 Assessing the Effects of Fiscal Shocks Craig Burnside, Martin Eichenbaum and Jonas D.M. Fisher WP-99-18 Fiscal Shocks in an Efficiency Wage Model Craig Burnside, Martin Eichenbaum and Jonas D.M. Fisher WP-99-19 Thoughts on Financial Derivatives, Systematic Risk, and Central Banking: A Review of Some Recent Developments William C. Hunter and David Marshall WP-99-20 Testing the Stability of Implied Probability Density Functions Robert R. Bliss and Nikolaos Panigirtzoglou WP-99-21 Is There Evidence of the New Economy in the Data? Michael A. Kouparitsas WP-99-22 A Note on the Benefits of Homeownership Daniel Aaronson WP-99-23 The Earned Income Credit and Durable Goods Purchases Lisa Barrow and Leslie McGranahan WP-99-24 Globalization of Financial Institutions: Evidence from Cross-Border Banking Performance Allen N. Berger, Robert DeYoung, Hesna Genay and Gregory F. Udell WP-99-25 Intrinsic Bubbles: The Case of Stock Prices A Comment Lucy F. Ackert and William C. Hunter WP-99-26 Deregulation and Efficiency: The Case of Private Korean Banks Jonathan Hao, William C. Hunter and Won Keun Yang WP-99-27 Measures of Program Performance and the Training Choices of Displaced Workers Louis Jacobson, Robert LaLonde and Daniel Sullivan WP-99-28 The Value of Relationships Between Small Firms and Their Lenders Paula R. Worthington WP-99-29 Worker Insecurity and Aggregate Wage Growth Daniel Aaronson and Daniel G. Sullivan WP-99-30 Does The Japanese Stock Market Price Bank Risk? Evidence from Financial Firm Failures Elijah Brewer III, Hesna Genay, William Curt Hunter and George G. Kaufman WP-99-31 Bank Competition and Regulatory Reform: The Case of the Italian Banking Industry Paolo Angelini and Nicola Cetorelli WP-99-32 2 Working Paper Series (continued) Dynamic Monetary Equilibrium in a Random-Matching Economy Edward J. Green and Ruilin Zhou WP-00-1 The Effects of Health, Wealth, and Wages on Labor Supply and Retirement Behavior Eric French WP-00-2 Market Discipline in the Governance of U.S. Bank Holding Companies: Monitoring vs. Influencing Robert R. Bliss and Mark J. Flannery WP-00-3 Using Market Valuation to Assess the Importance and Efficiency of Public School Spending Lisa Barrow and Cecilia Elena Rouse Employment Flows, Capital Mobility, and Policy Analysis Marcelo Veracierto Does the Community Reinvestment Act Influence Lending? An Analysis of Changes in Bank Low-Income Mortgage Activity Drew Dahl, Douglas D. Evanoff and Michael F. Spivey WP-00-4 WP-00-5 WP-00-6 Subordinated Debt and Bank Capital Reform Douglas D. Evanoff and Larry D. Wall WP-00-7 The Labor Supply Response To (Mismeasured But) Predictable Wage Changes Eric French WP-00-8 For How Long Are Newly Chartered Banks Financially Fragile? Robert DeYoung WP-00-9 Bank Capital Regulation With and Without State-Contingent Penalties David A. Marshall and Edward S. Prescott WP-00-10 Why Is Productivity Procyclical? Why Do We Care? Susanto Basu and John Fernald WP-00-11 Oligopoly Banking and Capital Accumulation Nicola Cetorelli and Pietro F. Peretto WP-00-12 Puzzles in the Chinese Stock Market John Fernald and John H. Rogers WP-00-13 The Effects of Geographic Expansion on Bank Efficiency Allen N. Berger and Robert DeYoung WP-00-14 Idiosyncratic Risk and Aggregate Employment Dynamics Jeffrey R. Campbell and Jonas D.M. Fisher WP-00-15 Post-Resolution Treatment of Depositors at Failed Banks: Implications for the Severity of Banking Crises, Systemic Risk, and Too-Big-To-Fail George G. Kaufman and Steven A. Seelig WP-00-16 3 Working Paper Series (continued) The Double Play: Simultaneous Speculative Attacks on Currency and Equity Markets Sujit Chakravorti and Subir Lall WP-00-17 Capital Requirements and Competition in the Banking Industry Peter J.G. Vlaar WP-00-18 Financial-Intermediation Regime and Efficiency in a Boyd-Prescott Economy Yeong-Yuh Chiang and Edward J. Green WP-00-19 How Do Retail Prices React to Minimum Wage Increases? James M. MacDonald and Daniel Aaronson WP-00-20 Financial Signal Processing: A Self Calibrating Model Robert J. Elliott, William C. Hunter and Barbara M. Jamieson WP-00-21 An Empirical Examination of the Price-Dividend Relation with Dividend Management Lucy F. Ackert and William C. Hunter WP-00-22 Savings of Young Parents Annamaria Lusardi, Ricardo Cossa, and Erin L. Krupka WP-00-23 The Pitfalls in Inferring Risk from Financial Market Data Robert R. Bliss WP-00-24 What Can Account for Fluctuations in the Terms of Trade? Marianne Baxter and Michael A. Kouparitsas WP-00-25 Data Revisions and the Identification of Monetary Policy Shocks Dean Croushore and Charles L. Evans WP-00-26 Recent Evidence on the Relationship Between Unemployment and Wage Growth Daniel Aaronson and Daniel Sullivan WP-00-27 Supplier Relationships and Small Business Use of Trade Credit Daniel Aaronson, Raphael Bostic, Paul Huck and Robert Townsend WP-00-28 What are the Short-Run Effects of Increasing Labor Market Flexibility? Marcelo Veracierto WP-00-29 Equilibrium Lending Mechanism and Aggregate Activity Cheng Wang and Ruilin Zhou WP-00-30 Impact of Independent Directors and the Regulatory Environment on Bank Merger Prices: Evidence from Takeover Activity in the 1990s Elijah Brewer III, William E. Jackson III, and Julapa A. Jagtiani WP-00-31 Does Bank Concentration Lead to Concentration in Industrial Sectors? Nicola Cetorelli WP-01-01 On the Fiscal Implications of Twin Crises Craig Burnside, Martin Eichenbaum and Sergio Rebelo WP-01-02 4 Working Paper Series (continued) Sub-Debt Yield Spreads as Bank Risk Measures Douglas D. Evanoff and Larry D. Wall WP-01-03 Productivity Growth in the 1990s: Technology, Utilization, or Adjustment? Susanto Basu, John G. Fernald and Matthew D. Shapiro WP-01-04 Do Regulators Search for the Quiet Life? The Relationship Between Regulators and The Regulated in Banking Richard J. Rosen Learning-by-Doing, Scale Efficiencies, and Financial Performance at Internet-Only Banks Robert DeYoung The Role of Real Wages, Productivity, and Fiscal Policy in Germany’s Great Depression 1928-37 Jonas D. M. Fisher and Andreas Hornstein WP-01-05 WP-01-06 WP-01-07 Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy Lawrence J. Christiano, Martin Eichenbaum and Charles L. Evans WP-01-08 Outsourcing Business Service and the Scope of Local Markets Yukako Ono WP-01-09 The Effect of Market Size Structure on Competition: The Case of Small Business Lending Allen N. Berger, Richard J. Rosen and Gregory F. Udell WP-01-10 Deregulation, the Internet, and the Competitive Viability of Large Banks and Community Banks WP-01-11 Robert DeYoung and William C. Hunter Price Ceilings as Focal Points for Tacit Collusion: Evidence from Credit Cards Christopher R. Knittel and Victor Stango WP-01-12 Gaps and Triangles Bernardino Adão, Isabel Correia and Pedro Teles WP-01-13 A Real Explanation for Heterogeneous Investment Dynamics Jonas D.M. Fisher WP-01-14 Recovering Risk Aversion from Options Robert R. Bliss and Nikolaos Panigirtzoglou WP-01-15 Economic Determinants of the Nominal Treasury Yield Curve Charles L. Evans and David Marshall WP-01-16 Price Level Uniformity in a Random Matching Model with Perfectly Patient Traders Edward J. Green and Ruilin Zhou WP-01-17 Earnings Mobility in the US: A New Look at Intergenerational Inequality Bhashkar Mazumder WP-01-18 The Effects of Health Insurance and Self-Insurance on Retirement Behavior Eric French and John Bailey Jones WP-01-19 5 Working Paper Series (continued) The Effect of Part-Time Work on Wages: Evidence from the Social Security Rules Daniel Aaronson and Eric French WP-01-20 Antidumping Policy Under Imperfect Competition Meredith A. Crowley WP-01-21 Is the United States an Optimum Currency Area? An Empirical Analysis of Regional Business Cycles Michael A. Kouparitsas WP-01-22 A Note on the Estimation of Linear Regression Models with Heteroskedastic Measurement Errors Daniel G. Sullivan WP-01-23 The Mis-Measurement of Permanent Earnings: New Evidence from Social Security Earnings Data Bhashkar Mazumder WP-01-24 Pricing IPOs of Mutual Thrift Conversions: The Joint Effect of Regulation and Market Discipline Elijah Brewer III, Douglas D. Evanoff and Jacky So WP-01-25 Opportunity Cost and Prudentiality: An Analysis of Collateral Decisions in Bilateral and Multilateral Settings Herbert L. Baer, Virginia G. France and James T. Moser WP-01-26 Outsourcing Business Services and the Role of Central Administrative Offices Yukako Ono WP-02-01 Strategic Responses to Regulatory Threat in the Credit Card Market* Victor Stango WP-02-02 The Optimal Mix of Taxes on Money, Consumption and Income Fiorella De Fiore and Pedro Teles WP-02-03 Expectation Traps and Monetary Policy Stefania Albanesi, V. V. Chari and Lawrence J. Christiano WP-02-04 Monetary Policy in a Financial Crisis Lawrence J. Christiano, Christopher Gust and Jorge Roldos WP-02-05 Regulatory Incentives and Consolidation: The Case of Commercial Bank Mergers and the Community Reinvestment Act Raphael Bostic, Hamid Mehran, Anna Paulson and Marc Saidenberg WP-02-06 Technological Progress and the Geographic Expansion of the Banking Industry Allen N. Berger and Robert DeYoung WP-02-07 Choosing the Right Parents: Changes in the Intergenerational Transmission of Inequality Between 1980 and the Early 1990s David I. Levine and Bhashkar Mazumder WP-02-08 6 Working Paper Series (continued) The Immediacy Implications of Exchange Organization James T. Moser WP-02-09 Maternal Employment and Overweight Children Patricia M. Anderson, Kristin F. Butcher and Phillip B. Levine WP-02-10 7
@FedFRASER