Childhood Obesity and Education Levels

Childhood Obesity and Education Levels:
An Econometric Exploration




























By Kanishq Baweja
BBA- B
130010201095
Executive Summary

America’s health outlook is exacerbated by the childhood obesity epidemic. This paper
examines the effect of education level (a high school degree or above) on childhood obesity
using econometric methods. Ten models of bivariate and multivariate regressions were used,
including state and time fixed effects. The results demonstrated that education level is not
statistically significantly associated with childhood obesity.
The policy implications of these findings along with some suggestions of potential
prospective policies are explained.

Childhood Obesity and Educational Attainment in the United States
According to the Center for Disease Control (2013), the prevalence of childhood obesity
(children ages 2-19) has tripled since 1980 from 6% to 17%, exacerbating the already
considerable consequences of obesity on America’s health. More than ever, childhood
obesity demands innovative solutions to address multifaceted causes involving the individual,
family, community, and environment. Obesity is measured as a ratio of one’s weight and
height to calculate an age and gender-specific number that determines body mass index
(BMI). The commonly accepted notion is that eating too many calories and not getting
enough physical activity cause childhood obesity. The underlying factors,costs, and solutions,
however, are much more nuanced.

Childhood obesity warrants the priority of national policymakers, as well as individuals and
communities alike. Its effects on the cost of health care and quality of life for affected
individuals are profound. Annual direct health care costs (including annual prescription drug,
emergency room, and outpatient costs) related to childhood obesity totaled $14.1 billion, plus
inpatient costs of $237.6 million, according to a 2009 study (Cawley, 2010). These costs will
continue to climb as obesity during childhood persists into adulthood. Total costs of obesity
measured $147 billion in 2008

Additionally, research suggests that childhood obesity is also a socioeconomic issue
correlated with income, food access, and ethnicity. The prevalence of obesity in 2004 was
20% of African American children, 19% of Mexican American children, and 39% of Native
American boys, compared with 16% of non-Hispanic white children. There is also a racial
and socioeconomic gap in physical activity. In California, only 11% of male Latino teens and
18% of male African American teens engage in moderate physical activity at least five times
a week, compared to 20% of male white teens (Grissom, 2005). In terms of food access, a
multi-state study cites findings that people with access to fresh food markets have lower rates
of obesity. In California, obesity and diabetes rates are 20% higher for those living in the
least healthy food environments, controlling for household income, race/ethnicity, age,
gender, and physical activity levels.

Figure 1: Childhood Obesity Rates in the U.S

The educational attainment of the U.S. population is similar to that of many other
industrialized countries. The vast majority of the population has completed secondary
education. Today, a rising number of college graduates outnumber high school dropouts.
On average, residents of the United States are spending more time working toward their
education. (Reed & Cochrane, 2012)
Figure 2: Percentage of the Population 25 and Over Who Have Completed High
School or College


However, education is also one of the main determinants of class and status in American
society. Much like income, education level also differs by race, age, size of household and
geography. According to 2010 Census Data, demographics with the highest educational
attainment in the United States are also among those with the highest per capita and
household income (U.S. Census, 2010). Thus, it is important to note that while the population
as a whole is proceeding further in educational programs, income and educational attainment
remains highly correlated.

Literature Review


Academic literature has explored the relationship between socioeconomic factors, such as,
paternalism, and knowledge of obesity on obesity. The studies span the globe and a variety of
topics. However, none of them directly explore the relationship between education levels and
childhood obesity. “Paternalism, Obesity, and Tolerable Levels of Risk” by Michael Merry,
(2012) explores the relationship between paternalism and childhood obesity. He does this by
looking at the risks of paternalistic intervention in order to prevent childhood obesity. His
main argument is that there is a correlation between poverty and obesity, thus a correlation
between education levels and obesity. He makes this connection by arguing that poorer
people aren’t as well educated as wealthier people because they cant afford private school
tuition/ live in good school districts, an have less access to reliable health information and
preventative health care. This is especially important when considering the interactions
between culture, socioeconomic factors, and education and their effect on childhood obesity.
“Parental perception of their children’s weight status and its association with their nutrition
and obesity knowledge”, examines paternal perceptions of children’s weight status and its
association with their knowledge on nutrition and obesity. (Muhammad, et al., 2008)
Administered in Kuala Lumpur, parents responded to a self-administered questionnaire that
contains parental perception of their child’s weight status as well as knowledge of nutrition
and obesity. The parents’ perception of the children’s weight status was then compared with
the actual measured weight status. The study found that parents have a good knowledge of
weight status, however they need to be educated about how to cook in more healthy, low fat
ways. These findings indicate a potentially correlative link between obesity, culture, and
information, but not education levels.“Parents are the exclusive agents of change in the
treatment of childhood obesity” compares the efficacy of a family-based approach for the
treatment of childhood obesity. (Golan, et al., 1998) Typically, in the traditional approach,
the children are the agents of change. In addition to using physiological statistics, the study
included a questionnaire that incorporates socioeconomic and demographic questions such as
family eating and activity habits. All participants in the treatment and control groups received
an hour-long support and educational counseling session. In the treatment group, parents
received 14 sessions, where as in the control group, children received 30 sessions. This study
concluded that the dropout rate amongst parents was nine times higher than those in the
control group. Additionally, the percentile of weight reduction was significantly higher in the
treatment group than in the control group. Thus, treatment of childhood obesity using parents
as the exclusive agents of change was more effective than the control group.
Another interesting study is “Social class, parental education, and obesity prevalence in a
study of six-year-old children in Germany”. (Lamerz, et al., 2005) This representative cross-
sectional survey examined the effects of socioeconomic status, and parental education on
obesity among six year olds in Aachen, Germany. The study showed that there was a
significant relationship between parents' years of education and childhood obesity. When
looking at other socioeconomic status variables, parental education was the most important
socioeconomic status variable that accounts for this relationship. The exploration also argued
that future studies should focus on early childhood obesity and prevention programs. The
authors expressed that prevention programs should target undereducated parents and their
young children who are at high risk. Finally, the study concluded social status is inversely
associated with childhood obesity as early as age six.

Hypothesis
The hypothesis being tested here states that levels of education in a state effect the
levels of childhood obesity in the state

Econometric Model

Y= β1 + B2(X) + µ

Where Y is the obesity in the state, and X is the Average literacy in the state
µ here is the standard error term that is used to standardise the results from the above
economic model have standardised results.

Data and Methodology
The objective of this study is to determine if there is a causal link between education levels
and obesity rates.
The study uses cross-sectional county level data extrapolated from the USDA “Food
Environment Atlas” from 2012. There were a total of 3,282 counties observed in the data set.
County level data is used for two main reasons. First, it provides a breadth of data that
increases the sample size such that bolsters the internal validity of the tests run in the study.
Secondly, many socioeconomic, demographic, and educational issues are local issues.
Demographics vary within states and from region to region, thus county level data allows my
tests to capture this through its wide array of data points. There are eleven variables explored
in this study. Variables observed include: the dependent variables childhood obesity rate
(2008, 2011, and the percent change between the two years) and the independent variables
state, education level in 2009 (a high school education or above), the number of fast food
restaurants per 1,000 people in 2009, the percent of the population that was white, the percent
of the population that participates in the Supplemental Nutrition Assistance Program (SNAP),
the percent of obese adults in 2009, the number of work-out/recreational facilities per 1,000
people in 2009, and finally the interaction between education level and participation in the
SNAP program.
Additionally, culture plays and important role in the prevalence of obesity, especially among
children. However, there is no prescribed way to control forcultural factors. Race was used in
the study to try to control for some of the omitted cultural aspects, but this method is not
foolproof. Race can also be associated with several of the other socioeconomic factors like
income, education, and food security, making it an increasingly complicated and involved
situation.

Despite attempting to be as comprehensive as possible, the models are not faultless. Problems
related to interaction effects may arise as a result of the variables used. Since education,
participation in the SNAP program, and the amount of fast food restaurants are all associated
with socioeconomic status, each of these variables could potentially have an effect on one
another. For example, those who participate in the SNAP program may live in impoverished
areas that have a high number of fast food restaurants and a small amount of people who have
high school degrees or above.
There may also be some problems related to omitted variable bias. Though genetics
contribute for some amount of childhood obesity, there is not a good way to test for genetics
through the data I was able to access. Doing so would require panel data that includes the
percentage of obese parents and children, over a series of years. Also, cross sectional data
that includes the percentage of total population or of children with diabetes would have been
a useful proxy for genetics or an obesity risk factor. Culture and cultural factors, explained
above, may contribute to some amount of omitted variable bias, as well.
Consideration also needs to be given to political and economic factors. Some states
implemented comprehensive childhood obesity policies between 2008 and 2011.
Implementation of these policies could negatively bias the regression results for some states.
States that did not implement policies would not be experiencing this bias.
Additionally, it is probable that states have made changes to their education systems within
the three-year window that was examined. These changes would not have an instant impact
but have the potential to skew the results. Lastly, the years used in the study coincided with
the Great Recession in the United States. Since education is closely related to socioeconomic
status, education levels may be swayed as a result.

Findings



State and Time Fixed Effects – Effect on Childhood Obesity Rates
Year 2008 2009 2008-2011
β
2
0.002 0.002 -0.001
β
1
-0.068 -0.064 -0.059
Constant 14.18** 13.98** -0.087*
R-squared 0.153 0.147 0.055
** p < 0.01 * p < 0.05
Durbin Watson 1.453


The above model time fixed effects to examine this relationship, as well. The constant in this
model is 14.18%, which is about the same as the regression without fixed effects (an increase
of 0.01 percentage points). Education level is associated with a 0.002 percentage point
increase in childhood obesity, only a 0.001 percentage point decrease from the other bivariate
model for this year. This figure is not statistically significant. The p-value is 0.519. However,
it is important to note that although they still are not strong, the rsquared values are much
higher when running fixed effects, than they were before. The rsquared value is 0.15,
meaning that 15 percent of the relationship between education and childhood obesity is
explained when using state fixed effects. This improvement in the rsquared value could be
attributed to the addition of some unobserved (and potentially unobtainable) confounding
variables by using state fixed effects.

In fixed effects for 2011, the constant is 13.98%. This is a 0.01 percentage point decrease
from the same regression without fixed effects. In this model, education is also associated
with a 0.004 percentage point increase in childhood obesity.
This number is also not statistically significant. The p-value is 0.656, meaning that we can
only be 34.4% confident of this relationship. The r-squared is also .147 and has the same
meaning as the previous fixed effects regression.

The last case, of the percent change in obesity rates between 2008 and 2011, uses both state
and time fixed effects. The constant in this model is -0.087%. This is about 1.8 percentage
points more than the same regression without state fixed effects. This could be attributed to
the fact that only a fraction of states introduced comprehensive childhood obesity policies in
those three years. We can now assume that in model 7, states had a higher percentage
reduction in childhood obesity between 2008 and 2011 were artificially inflating the results.
However, these findings are not statistically significant. The p-value is 0.648 meaning that we
can only be 64.8% confident of this value. The r-squared value in this model is 0.055,
meaning that only 5.5% of the effect of education on childhood obesity levels is accounted
for by this regression using state and time fixed effects.


Inference

These findings seem to prove that there is no evidence that demonstrates the relationship
between education level and childhood obesity rate is causal. But rather, the relationship
between the two can be correlated when considering the other important factors like race and
socioeconomics. Thus, we can accept the null hypothesis. The study’s results are especially
interesting because they disagree with some popular notions that those with high education
levels, a high school degree or above, are systematically less obese than those who are less
educated, especially in poorer areas. For example, Merry’s argument in “Paternalism,
Obesity, and Tolerable Levels of Risk”, is in direct contrast with my findings. Although his
argument may be salient in the Netherlands, it is disproved in my study of the United States. I
agree that there may be a correlative link between education levels and childhood obesity
levels. Childhood obesity levels are not statistically significantly related to education levels,
even in poor areas.
There is not a causal link.

Policy Context

After looking at the models, it can be concluded that education level does not have a
statistically significant effect on childhood obesity rates, across the country. Also, generally
speaking, comprehensive childhood obesity bills implemented between 2008 and 2011 have
effectively reduced the rate of childhood obesity rates in their respective states. However,
determining the statistical significance of this would be the result of an entirely different
study. The results necessitate the exploration of several policy implications.
As we have seen, comprehensive policies for solving childhood obesity seem to be effective
when implemented on a state level. Currently there are about 20 states that have some type of
legislation that aims to effectively reduce the prevalence of childhood obesity. These policies
include information campaigns, some sort of information tracking, a physical education
mandate for schools, Body Mass Index screening in schools, parent education programs and a
task force created to help solve the childhood obesity problem. If more states were to pass
comprehensive childhood obesity legislation, then it could be assumed that the rate of
childhood obesity would continue to decline and may even pick up the pace.
Education campaigns that seek to inform students about healthy eating, exercise, and the risks
associated with obesity will not work by themselves. Although, when implemented in
conjunction with other policy elements that seek to provide support to the impoverished or
provide adequate time/ spaces for exercise and play, the results could prove to be more
fruitful. Examples of these types of comprehensive policy solutions could include
information mandates to educate the public along with reforming the SNAP program, or
setting physical activity standards in schools.


Conclusion

The problem of obesity in America today has profound consequences for the health
of our citizens and the national economy. The growing number of overweight
children will only magnify those impacts over time, increasing the health care costs
and reducing the overall efficiency of the workforce. This study concluded using
bivariate and multivariate regressions and state and time fixed effects that there is
not a causal link between education levels and childhood obesity rates. Obesity
arises from a number of complex factors, including those that weren’t controlled for
in this study, so any effective solution to reduce childhood obesity rates must be
multifaceted and comprehensive.

References

 Cawley, J. (2010). The economics of childhood obesity. Health Affairs, 29(3), 364-
371.
 Cawley J, Meyerhoefer C. (2012). The medical care costs of obesity: an instrumental
 variables approach.
 Journal of Health Economics, 31, 219-30.
 Centers for Disease Control and Prevention. (2013) Overweight and obesity: Data and
statistics.
 Lamerz et al. (2005). “Social class, parental education, and obesity prevalence in a
study of six-year-old children in Germany”. The International Journal of Obesity.
Volume 29, pgs. 373-380.
 Merry, M. (2012). “Paternalism, Obesity, and Tolerable Levels of Risk”. The
Democracy
 Education Journal. Volume 20, number 1, pgs. 1-6. Muhammad et al. (2008).
“Parental perception of their children’s weight status and its association with their
nutrition and obesity knowledge”. Asia Pacific Journal of Clinical Nutrition. Volume
17, number 4, pgs. 597-602.
 Reed & Cochrane. (2012). Student Loan Debt and the Class of 2011. The
Institute for College Access and Success.
 Treuhaft, S. & Karpyn, A. (2010). The grocery gap: who has access to healthy food
and why it matters. Policy Link.
 U.S. Census Bureau. Current Population Reports - Table POP1. (Series P-25, No.
917).
 U.S. Department of Agriculture. (2013). Supplemental nutrition assistance program:
 Healthy incentives pilot. Retrieved from http://www.fns.usda.gov/snap/hip/
 “USDA Food Environment Atlas”. (2012). United States Department of Agriculture.