Economic Stressors and Childhood Obesity: Differences by Child Age and Gender

Childhood obesity is a public health challenge in the United States (U.S.) and elsewhere in the world. Additionally, those who are obese are heavier than in the past (Anderson & Butcher, 2006). In the U.S., one in three children is overweight or obese (Ogden et al., 2010), a prevalence that has tripled since 1970 (Anderson & Butcher, 2006; Kumanayika & Grier, 2006; Wang & Zhang, 2006). In response to this public health issue, Healthy People 2010 (US/DHHS, 2000) and President Obama (US/Office of the President, 2010) have identified childhood obesity as a national health priority as it has immediate consequences for a child’s physical and psychological health (Puhl & Latner, 2007; Raman, 2002; Strauss, 1999; US/DHHS, 2000), as well as implications for future health (Freedman et al., 2007; Raman, 2002; Strauss, 1999; US/DHHS, 2000). Beyond negative health outcomes, there are also economic costs (e.g., greater need for health care) associated with childhood obesity (Marder & Chang 2006; Skinner et al., 2008). Thus, identifying factors related to childhood obesity not only has implications for the health and quality of life of children, but it also has important implications for family expenditures and health care costs.


Introduction
Childhood obesity is a public health challenge in the United States (U.S.) and elsewhere in the world. Additionally, those who are obese are heavier than in the past (Anderson & Butcher, 2006). In the U.S., one in three children is overweight or obese (Ogden et al., 2010), a prevalence that has tripled since 1970 (Anderson & Butcher, 2006;Kumanayika & Grier, 2006;Wang & Zhang, 2006). In response to this public health issue, Healthy People 2010 (US/DHHS, 2000) and President Obama (US/Office of the President, 2010) have identified childhood obesity as a national health priority as it has immediate consequences for a child's physical and psychological health (Puhl & Latner, 2007;Raman, 2002;Strauss, 1999;US/DHHS, 2000), as well as implications for future health (Freedman et al., 2007;Raman, 2002;Strauss, 1999;US/DHHS, 2000). Beyond negative health outcomes, there are also economic costs (e.g., greater need for health care) associated with childhood obesity (Marder & Chang 2006;Skinner et al., 2008). Thus, identifying factors related to childhood obesity not only has implications for the health and quality of life of children, but it also has important implications for family expenditures and health care costs.
It is commonplace to focus on physical inactivity and dietary factors as the cornerstones of the childhood obesity epidemic, but stress is another common feature of the landscape facing American families today. Stress manifests itself across numerous dimensions at both the individual and family level. While stress can be managed successfully by many individuals and families, in some cases stress can become severe enough to lead to serious health consequences. A vast literature has demonstrated the effects of stress on numerous health outcomes for children and adults (e.g., Dearing et al., 2006;Evans & English, 2002;Gee & Walsemann, 2009;Kort-Butler, 2009;Schilling et al., 2008), including childhood obesity (Eisenmann, 2006;Garasky et al., 2009;Gundersen et al., 2008;Lohman et al., 2009).
While this work has provided policymakers with important insights, a central issue related to this research must be addressed. Through regression analyses this work has found positive associations between stress and childhood obesity after controlling for a host of potential confounding factors, but it has not ascertained whether unobserved factors correlated with stress (e.g., a child's ability to cope with adverse conditions) may be the cause of the association. That is, these studies have implicitly assumed that time invariant unobserved factors were similar between different children and have ignored the possibility that children in households experiencing stress are different in unobserved ways from children in households not experiencing stress. To address this issue, the current study examined panel data using fixed effects models that controlled for time-invariant differences between children by using only within-individual variation to estimate the regression coefficients (Allison, 2005).
We examined data from the first two waves of the Child Development Supplement (CDS) of the Panel Study of Income Dynamics (PSID). Our results indicated that exposure to housingrelated economic stressors leads to a higher probability of a child being obese and to higher levels of obesity, especially for younger females. Other forms of stress examined here (financial and neighborhood) were not related to child weight status. These findings were robust across a range of model specifications and suggest that efforts to reduce housing stress may also lead to reductions in childhood obesity.

Theoretical framework
Child health is an important aspect of family well-being. The ecological theory of human development identifies four levels of influence faced by families attempting to maximize their well-being (Bronfenbrenner & Morris, 1998). These levels are the microsystem (e.g., individuals and families), the mesosystem (e.g., neighborhoods and social networks), the exosystem (e.g., community), and the macrosystem (i.e., larger cultural context). Theoretical work on the "stress process" defines stress as a negative physiological response and stressors as the external factors that cause this negative response (e.g., Aneshensel, 1992;Boss, 1988;Chrousos & Gold, 1992;Pearlin, Menaghan, Lieberman & Mullan, 1981;Pearlin, Schieman, Fazio & Meersman, 2005).
While genetic factors have consistently been shown to be central to whether a child is obese (e.g., Crossman et al., 2006;Gibson et al., 2007), environmental factors at the microsystem and mesosystem levels such as family, parental, and economic influences matter as well (e.g., Anderson et al., 2003;Cutler et al., 2003;Dietz & Robinson, 2008;Loureiro & Nayga, 2005). Environmental factors combined with a biological predisposition toward obesity provide the conditions for one's propensity for obesity to come to fruition (Anderson et al., 2003).

Stress and health
Stress is an environmental factor that often leads to reduced psychological and physiological health. One of the most common consequences of stress exposure is psychological distress, especially depression (e.g., Brooks et al., 2002;Kort-Butler, 2009;Schilling et al., 2008). Daily hassles and chronics stressors are the typical antecedents to depressive symptoms with stress leading to anxiety and aggression (Evans & English, 2002;Krause et al., 2003;Kort-Butler, 2009).
Physiologically, higher levels of stress have been associated with higher levels of selfreported illness (Gee & Walsemann, 2009;Goodman, 1999;Larson et al., 2008). Early and chronic exposure to stressors has been linked with cardiovascular disease and obesity.
The effects of economic stressors on health and well-being have been examined under a range of headings including economic stress, financial stress, economic hardship, economic strain, economic pressure, and material hardship (e.g., Fletcher et al., 2005;Kim & Garman, 2003;Kim et al., 2006). Lower income children are at an elevated risk of exposure to varied forms of these stressors (Gershoff et al., 2007;Pearlin et al., 2005;Wickrama et al., 2007) and may have more severe health consequences stemming from stress due to their greater exposure and vulnerability (Sampson et al., 1997;Spencer, 2001). Recently, scholars have shown an association between household-and individual-level indicators of stress and childhood obesity Gundersen et al., 2008;Lohman et al., 2009).

Stress and childhood obesity
Recent evidence suggests that stress faced by family members may lead to childhood obesity. Gundersen et al. (2008) found using data from the National Health and Nutrition Examination Survey (NHANES) that higher values of an index measuring cumulative stress exposure led to higher probabilities of obesity for food secure children in comparison to food insecure children. This result held for younger children, but the effect was statistically insignificant for older children. Lohman et al. (2009) found using data from the Three-City Study that higher levels of individual stress experienced by a child between the ages of 10 and 15 were statistically significantly associated with higher probabilities of childhood obesity. In contrast to Gundersen et al. (2008), they found that food insecure children with higher levels of maternal stress had higher probabilities of childhood obesity in comparison to food secure children. In should be noted that Gundersen et al. (2008) and Lohman et al. (2009) examined data from different surveys and employed different protocols for measuring food insecurity and stress which may have led to their finding somewhat conflicting results. Garasky et al. (2009) found using data from the PSID that stress was associated with childhood obesity, but the type of stress had differential effects for younger and older children. For younger children, lack of cognitive stimulation and emotional support in the household were associated with higher probabilities of obesity. However, for older children, mental and physical health problems and financial strain in the household were associated with higher probabilities of obesity. Finally, van Jaarsveld et al. (2009) conjectured that perceived stress in pre-adolescence may set adiposity trajectories, with no accentuation of differences due to perceived stress in adolescence, when they did not find an association between perceived stress and weight gain among British adolescents. Important to this research, few studies have investigated whether associations between stressor exposure and weight status differ between females and males, nor have they examined whether the associations differ by gender within age groups (an exception being van Jaarsveld et al. (2009) who examined potential differences among adolescents by gender). Among adolescents, boys and girls have the same likelihood of being overweight (Anderson et al., 2003) with no differences in the effect of perceived stress on weight gain (van Jaarsveld et al., 2009). However, some research indicates that girls and boys cope with stress differently (e.g., Frydenberg & Lewis, 2004;Rudolph, 2002), although others have found no gender differences in adolescent coping (Wadsworth & Compas, 2002). Different coping styles for boys and girls may affect their risks of obesity. For example, eating-related responses to stress differ for boys and girls (Mikolajczyk et al., 2009) with adolescent girls being more likely than boys to suffer from disordered eating (Hepworth, 2004). Additionally, greater consumption of sweets and fast food consistent with higher rates of "emotional eating" has been found among women relative to men (Larsen et al., 2006).
The current study expanded on previous work by considering three types of economic stressors most commonly examined in the literature and most relevant to the home environments of children -housing, financial and neighborhood stressors -within a fixed effects framework. We examined separate groups of younger and older children and, given the mixed results by age and gender discussed above, compared the relationship between stress and obesity for girls versus boys within each age group. Previous studies did not systematically examine the potential effects of these economic factors on a child's propensity to be obese in this way. Additionally, previous work concentrated on binary measures of obesity (an exception being van Jaarsveld et al. (2009)). We utilized the obesity gap Jolliffe, 2004) to depict the extent of a child's obesity. A central advantage to using the gap measure is that it addresses a key disadvantage associated with binary measures of child weight status. A binary measure of obesity treats all children with a body mass index (BMI) ≥ 95 th percentile for age and sex the same. With the obesity gap, these children are treated differently within the context of the models. From a policy perspective, these additional analyses are important as one may be especially interested in children with relatively high levels of obesity (US/Office of the President, 2010).

Methods
We controlled for time invariant unobserved factors through the following fixed effects model: where i denotes a child, t denotes the interview wave (t=1 or 2), OB it α denotes the measures of the weight status of the child (in manners described below), I is a vector of the three economic stressor indices described below, Y is a vector of time varying covariates, μ is a child-specific fixed effect, and ε is an error term. We estimated logit fixed effects models for the binary specification of weight status and linear regression fixed effects models for our continuous (i.e., gap) measure. With respect to the direction of the influence of the economic stressors on childhood obesity, these models implicitly assumed that the stressors affected weight status. While in theory the relationship between the economic stressors and weight status could be bidirectional, our model was consistent with the vast majority of research in this area and every study discussed above.

Data
Our analyses were conducted with data from the first two waves of the Child Development Supplement (CDS-I and CDS-II) of the PSID. The PSID, begun in 1968, is a longitudinal study of a nationally representative sample of U.S. individuals and the families in which they reside. In 1997, a refresher sample of post-1968 immigrant families and their adult children was introduced to keep the study representative of the U.S. population (PSID, 2005). Currently, PSID interview waves are conducted biannually.
The CDS, a research component of the PSID focusing on children age 0-12 years in PSID families was introduced in 1997. The CDS examines a range of developmental outcomes within the context of family, neighborhood, and school environments. Assessments of cognition, behavior, and health status are obtained from a variety of sources including the child and the child's primary caregiver. The multi-method CDS survey design includes computer-assisted personal interviews (CAPI) and audio computer-assisted self-interviews (ACASI) (PSID, 2008b). Trained personnel measure the child's height without shoes using a rafter's square and tape measure. The child's weight is measured using a digital scale (PSID, 2008a). In 2002-2003, interviewers recontacted families that participated in CDS-I and remained active in the PSID as of 2001. Of those families, 91% were successfully reinterviewed (PSID, 2008b).
Information from CDS-I and CDS-II constituted the basis of our study. We supplemented these data with income and household composition data from contemporaneous PSID interview waves. Together, these data were well-suited for this analysis as they provided a large sample of households and detailed longitudinal information on child characteristics, family stressors, and relevant covariates. This data set had other strengths as well. In particular, the use of directly measured child height and weight strengthened this study as other large, national studies (e.g., Youth Risk Behavior Survey; National Longitudinal Survey of Youth) tend to rely on less reliable self-report or parental-report methods.
Our analytic sample consisted of 1,263 youths who at the time of the CDS-I interview were between 2 and 14 years of age. This research examined the full analytic sample, as well as two age-based subsamples. Consistent with other research, the full sample was split at age 8 years at the time of the CDS-I interview. Children less than 8 years of age (i.e., 24-95 months of age) at the time of the CDS-I interview were the younger sample (n = 677 children). Children at least 8 years of age (i.e., 90-167 months of age) at the time of the CDS-I interview comprised the older sample (n = 586 children). Youths under age 2 years (less than 24 months) at the time of the CDS-I interview and over age 18 years (over 228 months) at the time of the CDS-II interview were removed. Also, youths classified as underweight (body mass index < 5 th percentile for age and gender as defined below) at either their CDS-I or CDS-II interview were removed from the sample since underweight children were not the focus of the referent group (i.e., normal weight youth).

Dependent variables
The measures used to delineate child weight status began with the calculation of a child's body mass index (BMI, kg/m 2 ). The BMI for each child was then mapped into a percentile based on age (in months) and gender using the Centers for Disease Control and Prevention (CDC) growth charts for the United States (e.g., Kuczmarski et al., 2002;Ogden et al., 2002). Our measures of obesity (OB BIN and OB GAP ) were derived from these BMI percentiles and employed the definitions of the American Academy of Pediatrics (Barlow, 2007). That is, we set the obesity cutoff at the 95 th percentile for age and gender.
Our binary measure of child weight status (OB BIN ) was defined as follows: where BMIPER was the child's BMI percentile for age and gender.
Our measure of obesity severity (OB GAP , the obesity gap measure) was defined as follows: (3) OB GAP = 0 otherwise To better understand how OB BIN and OB GAP were defined, consider three children who have BMI percentiles for age and gender of 50, 96 and 99, respectively. Regarding OB BIN , the first child, and any child with a BMIPER below the 95 th percentile, will have OB BIN set equal to 0. For the second and third child, and any child with BMIPER ≥ 95 th percentile, OB BIN will equal 1. As for OB GAP , the first child, and any child with a BMIPER below the 95 th percentile, OB GAP will be set to 0 as well. Key to these analyses is that the second and third children, both of whom are classified as obese, will have different values for OB GAP . OB GAP will equal 0.2 for the second child, while OB GAP will equal 0.8 for the third child. That is, children with more severe obesity (higher BMI percentiles for age and gender) will have higher obesity gap scores. Summary statistics for the variables in the multiple regressions for each wave are reported in Table 1 with changes in the variables across interviews reported in Table 2. About onefifth of the analytic sample was categorized as obese in each wave. The group of younger children had a higher proportion that was obese compared to the older children. About onefourth (22.7%) of all of the children experienced a change in weight status between interviews with a higher percentage of younger children (28.5%) experiencing change compared to the older children (15.9%). Older children, however, were more likely to become obese (9.6%) between interviews than to no longer be obese (6.3%) at the second interview. Younger children slightly more frequently became non-obese between waves (13.7% became obese, 14.8% were no longer obese at CDS-II). By definition, there was more change in weight status when status was assessed via the obesity gap (31.8% for all children, 37.8% for younger children, and 24.9% for older children) than when obesity was assessed through the standard binary measure.

Economic stressors
Stress is a difficult concept to measure. Both subjective (i.e., directly asking individuals how much stress they feel from a given stressor) and objective (i.e., observing stressors within an individual's environment or assessing physiologic markers such as cortisol) assessments of stress have been used in research. For example, Gottholmseder et al. (2009)  This study employed objective measures of stress for three categories of economic stressors (housing, financial and neighborhood). Consistent with individuals experiencing "stress pile-up" as a result of dealing with multiple stressors at once (McGuigan, 1999;White & Klein, 2002) and aggregate economic risk being a more important correlate than any single economic risk (MacFadyen et al., 1996), a stressor index was calculated for each category by summing the dichotomous response values for the variables in the category.
Housing stressors. Economic stressors related to housing were measured via four variables. CDS respondents when asked about economic problems in the last 12 months were asked three questions about housing experiences. "Did they move to cheaper living quarters?" "Did they move in with other people?" "Did they send one or more of the children to live with someone else?" A 'yes' response to any of these items was considered an indicator of housing-related economic stress. The last housing indicator regarded mortgage and rent expenditures. Respondents were queried about monthly first mortgage, second mortgage and rent payments. An indicator of economic stress was calculated by summing these payments and multiplying by 12 to arrive at an annual housing expenditure. The annual housing expenditure was divided by annual family income to determine the share of annual income spent on housing. If the share of income spent on housing was greater than 30 percent, an amount commonly considered to represent a household that is housing cost burdened (US/HUD, 2007), we considered this to be an indicator of economic stress. The housing stressors index ranged in value from 0 to 4.

Financial stressors.
Economic stressors related to finances were measured via ten variables.
Respondents were asked a series of questions regarding possible financial problems they may have experienced in the past 12 months. Specifically, respondents were asked: Have you done any of the following or have any of the following happened as a result of economic problems in the last 12 months?: (1) Sold possessions or cashed in life insurance; (2) Postponed major purchases; (3) Postponed medical care; (4) Borrowed money from friends or relatives; (5) Filed for or taken bankruptcy; (6) Fallen behind in paying bills; (7) Had a creditor call or come to see you to demand payment; (8) Had your wages attached or garnished by a creditor; (9) Had a lien filed against your property because you could not pay a bill; and (10) Had your home, car or other property repossessed. For each question, a 'yes' response was considered an indicator of economic stress. The financial stressors index ranged in value from 0 to 10.
Neighborhood stressors. Economic stressors related to the household's neighborhood were measured via two variables. Respondents were asked to rate their neighborhood as a place to raise children on a five-point scale ranging from 'excellent' to 'poor.' Families residing in neighborhoods self-assessed as being a 'fair' or 'poor' place to raise children were considered to be experiencing neighborhood-related economic stress. Respondents also were asked how safe it was to walk around alone in their neighborhood after dark. Responses ranged from 'completely safe' to 'extremely dangerous.' Families residing in neighborhoods self-assessed as being 'somewhat dangerous' or 'extremely dangerous' also were considered to be experiencing neighborhood-related economic stress. The neighborhood stressors index ranged in value from 0 to 2.
One-third (31.20%) of the respondents reported they had experienced some aspect of housing stress (had an index value > 0) at one or both CDS interviews. Similarly, about onethird (33.57%) reported experiencing some facet of neighborhood stress during the study. It was more common to experience an aspect of financial stress. Three-fourths (74.43%) of the respondents indicated they had experienced an element of financial stress during one or both of the CDS interviews. As seen in Table 1, the mean values for each of the three stressor indices decreased from CDS-I to CDS-II for all children. Housing fell from 0.245 (out of 4) to 0.196; financial decreased from 1.451 (out of 10) to 1.182; neighborhood declined from 0.299 (out of 2) to 0.295. Index values declined across age groups as well, except for neighborhood stressors increasing between interviews from 0.298 to 0.301 for younger children. As seen in Table 2, some children experienced change in these measures. The most change occurred with financial stressors with over 60 percent of the sample experiencing a change. In contrast, about one-fourth of all children in the analytic sample experienced a change in their level of housing or neighborhood stressors.

Time varying covariates
Two time varying covariates were included in these analyses. These were the number of persons in the household and annual total family income. As seen in Table 1, there were about four members on average in each household. The slight mean decrease in household size between waves stemmed from three times as many of the households of older children experiencing a decrease in size versus an increase in size (38.1% and 13.7%, respectively). Mean annual income increased from $50,565 at CDS-I to $70,725 at CDS-II with approximately three-fourths of the sample gaining income between waves.

Results
Results from our estimation of equation (1) on the samples of all children, younger children and older children are displayed in Table 3. The columns in each table reflect results for the full analytic sample (columns 1-2), the younger children (columns 3-4) and the older children (columns 5-6) for the various specifications of weight status (OB). In columns (1), (3) and (5), we set α=0 in equation (1) (i.e., the binary measure of obesity) and estimated logit fixed effects models. In columns (2), (4) and (6), we set α=1 (i.e., the obesity gap). We estimated linear regression fixed effects models for the obesity gap analyses.
www.intechopen.com Estimated coefficients with standard errors in parentheses. Obese (BMI ≥ 95th percentile). Superscripts of *, **, and *** indicate that the p-value of the coefficient is less than 0.10, 0.05, or 0.01, respectively. Table 3. Effects of stressor indices and time varying covariates on weight status and gap measures.
The housing stressors index was significantly, positively related to the likelihood of being obese across both specifications of weight status for all children. With respect to the magnitude of the relationship, a one unit increase from the average value of the housing stressors index across both interview waves (from 0.221 to 1.221) led to a 10.7 percentage point increase in the probability of being obese from a mean probability of 20% at CDS-I, and 32.6% increase in the depth (obesity gap) of obesity for those who were obese.
(Derived from Tables 1 and 3.) We found no evidence of a relationship between the other forms of economic stressors -financial and neighborhood stressors -and child weight status. Estimated coefficients with standard errors in parentheses. Obese (BMI ≥ 95th percentile). Superscripts of *, **, and *** indicate that the p-value of the coefficient is less than 0.10, 0.05, or 0.01, respectively. The coefficients on the time varying covariates are suppressed for brevity. The relationships between economic stressors and child weight status were considered for younger (age 2 to 8 years at the CDS-I interview) and older (age 8 to 14 years) children separately and are reported in Table 3, as well. For younger children, exposure to housing stressors was positively and significantly related to obesity. Examining the magnitude of the relationships as above, a one unit increase in the housing stressors index led to a 10.7 percentage point increase in the probability of being obese. The percent increase for the obesity gap measure was 36.9%. While the percentage point increases in the probability of being obese were the same for all children and only the younger children, the magnitude of the association for the depth of obesity for those who were obese was larger for the sample of younger children than for all children. Unlike younger children, for older children the housing stressors index was significantly positively associated with weight status for only www.intechopen.com the binary obesity measure. This result indicated that a one unit increase in the housing stressor index was associated with a 9.5 percentage point increase in the likelihood of being obese for older children. The other two measures of economic stress were statistically insignificant for both age groups. Table 4 provides additional depth to the age-based analyses by examining the relationships between the three economic stressors and weight status by the gender of the child. These analyses revealed that the significant relationship between housing stressors and weight status for younger children was confined to females. For younger females, a one unit increase in the housing stressors index increased the likelihood that a younger female was obese by 19.3 percentage points. The percentage increases in the depth (obesity gap) of obesity for those young females who were obese stemming from a unit increase in the housing stressor index was 57.9. The relationship between housing stressors and weight status for younger males was statistically insignificant. The remaining results were generally consistent with those reported in Table 3. Financial and neighborhood stressors were not related to weight status for younger females or males. For older females and males, none of the stressor indices were related to weight status except for exposure to neighborhood stressors unexpectedly being negatively associated with weight status for older females in the obesity gap examinations.

Conclusions
Although efforts to prevent and treat obesity have traditionally emphasized physical activity and diet, it is now becoming more widely recognized that the causes of childhood obesity are complex and multifactorial (Dietz & Robinson, 2008;Eisenmann, 2006). In this paper, we considered the impacts of economic stressors on childhood obesity. Employing an ecological theoretical framework (Bronfenbrenner & Morris, 1998), we considered factors within the microsystem and mesosystem associated with childhood obesity. Specifically, we examined the relationships between economic stress and obesity during childhood and adolescence. In contrast to previous work on this topic, we explicitly controlled for the effects of time invariant unobserved factors (e.g., a child's ability to cope with adverse conditions) which may be correlated with stress through the estimation of fixed effects models. Additionally, we included measures that portray the incidence of obesity. After controlling for these factors, we found statistically significant effects of economic stress on childhood obesity. Consistent with others (e.g., Garasky et al. (2009)), we found that the relationships between economic stressors and child weight status differed across age groups. More specifically, experiencing housing stress was significantly positively associated with a younger girl's probability of being obese, and the depth of her obesity.
The differing associations between stressors and child weight by age groups can be considered within either a developmental family process framework or a developmental neuropsychoendocrine perspective (i.e., maturation of the stress systems). Younger children are more dependent on their caregivers. Middle childhood youth and early adolescents (ages 8 to 14 years in this study) may have additional/alternative sources of support such as peers, coaches, teachers and romantic partners to whom they may turn in times of stress. Additionally, we conjecture that housing stressors may be more apparent to younger children than other stressors. Compared to the other sources of economic stress considered here (financial and neighborhood stressors), housing-related stress may have the most serious ramifications in terms of altering a child's day-to-day routine. There is evidence, for example, showing that frequent moves are negatively related to the well-being of children (Astone & McLanahan, 1994;Tucker et al., 1998). Children and families experiencing the types of housing stress examined here may have less control over their food choices and physical activities. Girls in particular are less physically active (Sallis et al., 2000), are more susceptible to stress (Rudolph, 2002), and are more likely to use food as a coping mechanism (Larsen et al., 2006.) We offer five recommendations for future research based on the results of this study. First, other indicators of economic stress should be considered. For example, families often contend with finding adequate and affordable child care which may impact the health and well being of their children, especially their younger children. Second, although examining the mechanisms underlying the relationship between economic stressors and obesity was outside the scope of this study, several potential pathways could be examined. One possibility is parenting. Several studies suggest that economic pressure is associated with lower marital quality, lower parenting quality, and higher levels of depression in children's caretakers, each of which is associated with worse child outcomes (e.g., Kalil & Dunifon, 2007;Robila & Krishnakumar, 2006). Third, our models implicitly assumed that exposure to economic stressors affected child weight status, but that child weight status did not affect the economic stress experienced in the child's household. While the consensus in the literature is that our models reflect the appropriate direction of the relationship when considering economic stressors, future research may want to consider whether the direction of the relationship is reversed when examining non-economic stressors. For example, parents with overweight children may become stressed due to the stigmatization and discrimination associated with weight in our society (e,g., Puhl & Brownell, 2006) or may experience stress as a result of negative perceptions of their appearance (Kraig & Keel, 2001). Fourth, researchers may wish to investigate the effects of longer-term economic shifts within families on child weight status using the longitudinal study design employed here. For example, the effects of income volatility, changes in family structure, and job loss could be considered. Similarly, longer term exposure to stressors could be examined. Lastly, while this article has not examined the effect of the social safety net on childhood obesity, future research may wish to consider how the numerous assistance programs in the U.S. interact to help families mitigate economic challenges and stress. Dietz and Robinson (2008) contend that it is unlikely that the problem of obesity will respond to a single intervention. A clear policy implication from this research is the identification of a new avenue for reducing the likelihood that a child will be obese. Based on our findings, reductions in housing stress will lead to reductions in the extent and depth of childhood obesity. From a policy perspective, efforts to address housing stress such as the provision of rent vouchers and current programs targeting families facing foreclosure may have a potential added benefit of reducing childhood obesity and associated health care costs to U.S. government programs such as Medicaid and State Child Health Insurance Programs.

Acknowledgment
We wish to thank Yemisi Kuku for tremendous research assistance. www.intechopen.com