Abstract

This study examines the effects of college on weight over much of the life cycle. I compare weights for college students with their weights before and after college and with the weights of noncollege peers using data from the National Longitudinal Survey of Youth (NLSY). I also examine the longer-term effects of college measured almost three decades later. I find that college freshmen gain substantially less than the 15 pounds rumored to be typical for freshmen. Using difference models, individual-specific fixed-effects models, and instrumental variables models to control for various sources of potential bias, I find that freshman year college attendance is estimated to cause only about a one-pound increase. Supplemental results show that those from lower socioeconomic backgrounds gain more weight during the freshman college year. Longer term, having a college education consistently decreases weight. These negative effects have faded over the last 20 years, and they diminish as respondents approach middle age. These trends are more prevalent for whites and Hispanics than for blacks.

Introduction

Recent estimates suggest that 35 % of adult Americans are obese, which is roughly a 100 % increase from 25 years prior (Flegal et al. 2012). Those aged 18–29 (who include college students) form the age cohort experiencing the largest increase in obesity rates (Mokdad et al. 1999; Yakusheva et al. 2011): as many as 35 % of college students are either overweight or obese (Douglas et al. 1997; Lowry et al. 2000). Identifying the causes of excessive weight gain in adolescents is important because obese adolescents are more likely to be obese as adults (Nader et al. 2006; Ogden et al. 2010). In addition, the best approach to weight management may be to avoid gaining too much weight in the first place because losing unwanted weight permanently is difficult, with about 75 % of the lost weight typically being regained within five years (Anderson et al. 2001; Franz et al. 2007). Determining the point of the life cycle when weight changes are most likely to occur may aid in developing the best intervention and prevention techniques.

College entry from high school may be one of these critical periods associated with weight gain because it is likely a time when youths experience significant lifestyle changes. First, college students may experience a significant decrease in physical activity and exercise, which could be due to an increase in the largely sedentary activity of studying. Second, many college students have poor eating habits (American College Health Association 2005)—including skipping breakfast (Debate et al. 2001; Huang et al. 1994) and consuming limited variety of foods (Haberman and Luffey 1998)—perhaps because of a lack of cooking skills and inadequate cooking facilities. Third, college students may consume more alcohol. Drinking alcohol potentially increases weight because it is an energy-dense macronutrient, promotes fat storage, and enhances appetite (Colditz et al. 1991; Prentice 1995; Westerterp-Plantenga and Verwegen 1999). Drinking patterns could be affected by college attendance because of either newfound autonomy from parents or peer pressure. The effects of college attendance on weight may differ according to living arrangements. Students living in university-provided housing on campus may be more likely to eat in university dining halls with a food court atmosphere and limited food variety. On-campus food offerings may be more likely to include high-fat or fried foods.

Previous studies have found that education is associated with better health, which presumably would include medically recommended weights (Cutler and Lleras-Muney 2008, 2010; Grossman 2006). Specifically, Grossman argued that education improves health by enhancing the ability to use and respond to knowledge, including medical information, to produce good health (Grossman 1972a, b, 1975; Grossman and Kaestner 1997). Education may indirectly improve health by increasing household income, which facilitates access to health care, better housing, and a healthier environment (Kemptner et al. 2011). Indeed, education has been found to decrease mortality (Lleras-Muney 2005), reduce the probability of smoking (Kenkel et al. 2006; Sander 1995a, b), and enhance infant health (Currie and Moretti 2003). Given these findings, attending college might actually enable students to avoid gaining unhealthy amounts of weight.

Negative (and potentially deleterious) influences of college attendance on weight and weight changes would not necessarily contradict the literature’s hypothesized protective health effects of education. The beneficial health effects from education found empirically occur after college graduation. Perhaps a college education largely enhances the ability to produce health after the college education is complete. Thus, college attendance could have a positive influence on weight and weight change while in college (say, during the freshman year) but confer beneficial health effects after college completion.

Studies examining socioeconomic disparities have found positive relationships between socioeconomic status (SES) and various health outcomes (Marmot et al. 1991; Smith 2004), with advantaged individuals being healthier than their disadvantaged counterparts. In turn, several studies have found that in developed countries, advantaged individuals are less likely to be obese than their disadvantaged peers (for reviews, see McLaren 2007; Sobal and Stunkard 1989). This finding has been attributed to the factors such as the greater likelihood of poor individuals to cover caloric requirements with high-fat, high-calorie foods (Drewnowski and Specter 2004), to disadvantaged neighborhoods, where walking and biking are less feasible because of fewer sidewalks and safety concerns (Chang 2006; Chang et al. 2009; Chung and Myers 1999; Poortinga 2006), and to the cumulative effects of social and economic adversity—known as the weathering hypothesis (Geronimus et al. 2006). The negative association between SES and obesity has been observed for both genders—and is particularly strong for females (Scharoun-Lee et al. 2011)—and for multiple ethnic groups, although not always for nonwhites (Baum and Ruhm 2009; Chang and Christakis 2005; Chang and Lauderdale 2005; Mokdad et al. 2001; Ogden et al. 2010). This relationship tends to strengthen with age, at least though midlife, but it may have weakened over the last several decades (Chang and Lauderdale 2005).

This is the first study to examine rigorously the effects of college on weight over much of the life cycle. First, I examine weight and annual weight changes during college using data from the National Longitudinal Survey of Youth 1997 (NLSY97), which is a large, nationally representative data set that annually surveyed youths over a 14-year period (1997–2010). I document the weight of college students and the amount of weight they gain during their freshman, sophomore, junior, and senior years. I compare these data with weight and weight changes in the year prior to college attendance and in the year following college graduation, as well as with weight and weight changes for similarly aged noncollege NLSY97 youths. I attempt to estimate the effects of college attendance with multivariate regression analysis, using weight-change models, individual-specific fixed-effects models, and instrumental variables models. I also conduct several falsification tests, and I estimate several sets of supplemental models to explore the robustness of the results.

Next, I examine the longer-term effects of college on weight measured several years to almost 30 years after graduation, and I compare these longer-term effects of college on weight today with these effects from approximately 20 years earlier for those at similar ages. I use the 1997 NLSY cohort to examine how attending college around the turn of the twenty-first century is related to the weights of young adults aged 25–31 in 2010. I use the 1979 NLSY cohort (NLSY79) to examine how completing college in the late 1970s and 1980s is related to the weights of adults aged 45–52 in 2010. I compare the effects of college on the weight changes of 25- to 31-year-old NLSY97 respondents with corresponding effects for similarly aged NLSY79 respondents approximately 20 years earlier, in 1992, to explore whether the effects of college on weight have changed over time.

Literature Review

Several studies have examined weight for college freshmen (Crombie et al. 2009); however, all have used small “convenience” samples of students attending a particular university, typically where the study’s researchers are employed. The samples are further self-selected in that students participating in the survey have volunteered to do so and ultimately have volunteered to be later reweighed, if multiple weight observations are collected. Most of these studies have examined sample averages, and none have attempted to identify the causal effect of college on weight by controlling for potential sources of bias.

In the economics literature, researchers have examined the longer-term effects of education on weight—that is, after college graduation. Most have used instruments to identify causal effects, but their results differ markedly. Kenkel et al. (2006) used NLSY data from the 1979 cohort to examine the effects of high school completion on being overweight and obese. Their second-stage instrumental variables estimates for education on weight are statistically insignificant. In a multicountry framework using European data, Brunello et al. (2013) found that education has beneficial effects on body mass index (BMI) for females but not males. Just the opposite, using Australian twin data, Webbink et al. (2010) found no effects of schooling on being overweight for women but did find effects for men. Lundborg (2013), also using data on twins, found no effects of schooling on being overweight. Using German data, Jurges et al. (2011) found positive effects of schooling on obesity for men in some models but statistically insignificant effects for men in others; they found that schooling does not significantly affect overweight and obesity for women. In another study using German data, Kemptner et al. (2011) identified negative effects of education on male body weight. Arendt (2005), examining Danish data, found inconclusive evidence for education’s effects on BMI.

In this study, I attempt to build on the literature in seven ways. First, I examine the amount of weight gained during college using a large, nationally representative panel data set (NLSY97) that annually collected weight information on each respondent over a 14-year period. Second, I separately examine the amount of weight gained by college students during their freshman, sophomore, junior, and senior years. Third, I compare weight and annual weight changes for college students with weight measures for similarly aged noncollege NLSY97 youths. Fourth, this study is the first to compare annual weight changes in college with weight changes in the year prior to college attendance and in the year following college graduation. Fifth, I attempt to estimate the causal effects of college attendance on weight using first-differenced models and individual-specific fixed-effects models, in addition to instrumental variables models to control for various potential sources of bias. Sixth, I compare effects of college attendance on weight while in college with longer-term effects of college on weight measured almost 30 years later. Seventh, I examine how the effects of having a college degree on weight have changed over time for similarly aged individuals and for the same cohort of individuals over time as they approach midlife. In sum, this study provides the most complete picture of the effects of a college education on weight over most of the life cycle.

Data

I use NLSY97 data to examine college attendance and weight. The NLSY97 is a large, nationally representative panel data set that annually collects information about each respondent’s weight and educational attainment.1,2 The NLSY97 began annually interviewing 8,984 youths in 1997, when they were between the ages of 12 and 16, and the survey is ongoing. The NLSY97 sample contains 6,748 nationally representative respondents and an oversample of 2,236 additional African American and Hispanic respondents.

First, I identify respondent weight (measured in pounds) at the survey date. I exclude pregnant females and new mothers (women who gave birth within the past year) because their reported weight may not be representative of their nonpregnancy weight. I also exclude respondents with weights below 80 pounds or above 400 pounds (and those with heights below 48 inches or above 84 inches); this restriction eliminates very few observations and does not appreciably affect the results. I include observations only for those whose weight is reported during the months of March, April, or May—at approximately the end of an academic year and, fortunately, when many of the NLSY97 interviews were conducted.3 I examine two measures of annual weight change: (1) average annual weight change since age 16, and (2) weight change since the prior survey. When consecutive interviews do not occur exactly 52 weeks apart, I divide weight change between surveys by weeks between surveys and multiply by 52 to convert the data to year equivalents.

Because, all else equal, taller people tend to weigh more, I include height as a covariate in my weight regressions. To adjust further for height, I also examine BMI and BMI changes, where BMI is defined as weight divided by squared height (CDC 2006a, b). In supplemental models, I also examined obesity, defined as a BMI greater than or equal to 30 (CDC 2006b).

I also control for standard demographic characteristics. This includes dummy variables for gender and race. I control for age using a dummy variable for each age represented in each model. I also adjust for marital status, the number of children in the household, family size, work experience measured in years, whether the respondent lives in an urban area, and the educational attainment of the respondent’s mother. I also include state dummy variables and year dummy variables (one for each year represented in each model) as covariates throughout the analysis.

For the first part of my analysis—examining weight while in college—I select NLSY97 high school graduates between the ages of 17 and 23, identifying the years, if any, in which each attended college as a freshman, sophomore, junior, and senior. NLSY97 respondents who graduated from high school but were not enrolled in college serve as a comparison group.4 I also identify the pre-freshman year (the year before enrolling in college as a freshman) and the post-senior year (the year after completing the college senior year) as additional points of comparison.

I also examine longer-term effects of college on weight after graduation. To do this, I examine the weights of NLSY97 respondents from the 2010 interview wave, when respondents were aged 25–31. I compare these longer-term effects of college, which are relatively current, with the longer-term effects of college on weight from approximately 20 years earlier using data on similarly aged respondents (e.g., respondents aged 25–31) from the 1992 interview wave of the NLSY79. I also examine even longer-term effects of college on weight using data on NLSY79 respondents collected in 2010, when these respondents were aged 45–52.5 I use these data to explore the possibility that observable weight differences are more apparent when respondents are older because gaining weight occurs gradually.

Empirical Methodology

I use multivariate regression analysis to estimate the relationship between college attendance and weight. The key variables are a measure of weight (W) and college attendance (C). Formally, consider an empirical model in which weight is regressed on college attendance:
Wit=β0+β1Xit+β2Cit+εWit,
1
for observation i in year t, where X is a vector of standard demographic covariates and εW is the error term in the weight equation. Ordinary least squares (OLS) estimation of college attendance on weight will produce biased results if variables unobserved to the researcher are correlated with both weight and college attendance.
To control for time-invariant unobserved factors, I examine weight changes (Wit − Wit − 1):
WitWit1=β0+β1Xit+β2Cit+εWit.
2

In a first specification of weight change, I examine average annual weight change since age 16; in a second specification, I examine annual weight change since the prior survey. I normalize both measures by the time interval (e.g., the number of weeks) between them and then annualize so that, for example, a weight change of 2.0 means 2 pounds of weight gain over a year. Models explaining the change in weight between periods will, by definition, control for time-invariant individual-specific factors.

In addition to the “weight differencing” approach, I estimate individual-specific fixed-effects models that compare multiple observations from the same respondent. If the individual-specific unobserved component is the same across observations from the same respondent (time-invariant for each respondent), it can be identified and controlled for with respondent-specific dummy variables. In practice, the individual-specific fixed-effects model essentially compares observations over time from the same respondent.

Unfortunately, the first-differenced and fixed-effects models have limitations. If the respondent’s unobserved component is not constant over time, the estimates may still be biased. To explore the potential for remaining bias due to time-varying unobserved factors, I estimate instrumental variables (IV) models in which I predict college attendance and then estimate the effect of predicted college attendance (Ĉ) on weight:
Cit=α0+α1Xit+α2Zit+εCit
3
WitWit1=β0+β1Xit+β2C^it+εWit.
4

I identify IV models by including instruments (Z) in the college attendance model that are not included in the weight models. Rather than attempt to discover a new instrument with which to identify college attendance, I use an instrument that has been widely and successfully used in the literature: proximity of residence to a land-grant institution (see, e.g., Moretti 2004; Shapiro 2006).6 Under the 1862 and 1890 Morrill Acts, land-grant institutions were established and supported financially with endowments from the sale of federal land in an effort to teach agriculture (and science and engineering, as opposed to the classic liberal arts curriculum), at least partially in response to changes initiated by the industrial revolution. Specifically, I use as instruments dummy variables measuring (1) whether the respondent resides in a county with a land-grant institution, (2) whether the respondent resides in a county that borders a county in the same state with a land-grant institution, and (3) whether the respondent resides in a county that borders a county in another state with a land-grant institution. These instruments will be valid if they significantly explain college attendance and they do not affect weight gain independently from college attendance. Although the federal government placed land-grant institutions in apparent random fashion long before the NLSY97 cohort entered college (more than 100 years ago), the areas in which they were placed may have since developed differently than areas without these institutions in ways that could affect health and investments in public health and, consequently, weight. Furthermore, areas with land-grant institutions have a larger proportion of college graduates and a smaller proportion of high school graduates with some college, reputedly because the presence of such a university lowers the cost of attending college (Moretti 2004; Shapiro 2006).

To control for county-specific characteristics that may be correlated with whether a county has a land-grant institution and weight gain, I include in both the college attendance and the weight gain models (as part of Xit, not as instruments) additional covariates measuring county population density, the percentage of the population aged 18–20, the percentage of births to mothers younger than 20, the percentage of the population aged 25 or older with a bachelor’s degree, the unemployment rate, per capita income, and the percentage of families in poverty. Descriptive statistics suggest these county-level characteristics are similar for NLSY respondents who enroll and those who do not enroll in college.

Results

Descriptive Statistics

First, I present weighted descriptive statistics for 17- to 23-year-old NLSY97 respondents in Table 1 for the full sample and separately for those who are enrolled and are not enrolled in college. In my full sample, weight averages 163.5 pounds, and these college-aged respondents gain between 3 and 4 pounds annually (between surveys, and yearly since age 16). About 47 % of the observations are from respondents attending college (or are from the year before the respondent’s freshman year or from the year after the respondent’s senior year). More respondents are enrolled as college freshmen than other college years, perhaps reflecting college attrition (dropouts). Those who are not enrolled in college weigh more (168.3 vs. 157.9 pounds), but they do not necessarily gain more weight per year. Those not enrolled have gained less weight annually since age 16 than their peers in college (3.6 pounds vs. 3.7 pounds) but more weight since the last survey (3.2 pounds vs. 3.0 pounds).

For the next set of descriptive statistics, I select a subsample of NLSY97 respondents who directly entered college upon high school graduation, advanced one grade per year until college graduation four years later, and provided valid weight and height and education information during the freshman, sophomore, junior, and senior college years as well as for the pre-freshman and post-senior years; I also include NLSY97 respondents who never entered college during the four years following high school graduation and who provided valid weight and height and education information during these four post–high school years (that is, the freshman, sophomore, junior, and senior year equivalents, and the pre-freshman and post-senior year equivalents). Figure 1 shows the weights of college students during their freshman, sophomore, junior, and senior years of college, as well as their weights during the year prior to college enrollment and the year after college graduation, by gender, using the sampling weights. Figure 1 also depicts the weights of comparably aged noncollege youths by gender. Both college and noncollege students tend to gain weight over time. Males weigh more than females, as would be expected given that they tend to be taller. Noncollege students weigh more than college students. Weight seems to increase more over this six-year period for noncollege respondents than for their enrolled peers, particularly for females. Noncollege males gain 14.8 pounds, and college males gain 14.1 pounds; corresponding figures for females are 14.6 and 8.3 pounds, respectively.

Figure 2 presents annual weight changes for college students and noncollege youths by gender. Male college students gain less weight than noncollege males each year except the freshman year. In the freshman year, college males gain 5.1 pounds compared with 3.4 pounds in the freshman year equivalent for noncollege males. Female college students gain more than their noncollege female peers in both the pre-freshman year and freshman year, but the difference in these years is less than 1 pound. Weight increases the most during the freshman year for male and female college students. Some evidence suggests that noncollege peers gain more weight after the freshman year equivalent for females. In sum, the freshman year seems to be unique. First, college students gain more during this year than in any other, but similarly aged noncollege students do not. Second, the freshman year is the only year (with the exception of the pre-freshman year for females) in which college students gain more weight than similarly aged noncollege students.

Table 2 presents descriptive statistics for NLSY97 respondents in 2010, NLSY79 respondents in 1992, and NLSY79 respondents in 2010 separately for those with and those without a college degree. Every measure of weight and weight change is larger for those without a college degree than for college graduates, except annual weight changes in 2010 for NLSY79 respondents. The likelihood of having a college degree is also increasing over time (this probability is higher for NLSY97 respondents in 2010 than for NLSY79 respondents).

Regression Results

Next, I estimate models for weight change, presenting coefficient results for the college-related covariates in Table 3. As discussed earlier, examining average annual weight change since age 16 represents an attempt to control for time-invariant sources of unobserved characteristics. Results for Model 1, which examines the effect of freshman year college attendance for 18- and 19-year-olds, indicate that freshman year attendance increases weight by 1.187 pounds relative to nonattendance. This effect is statistically significant at the 1 % level. Examining the effect of college attendance collectively, across all college grades, on weight change using a sample of 18- to 22-year-olds in Model 2, I find that college attendance significantly increases weight by about half a pound. The effects of college attendance on weight may be different depending on the year of college attendance. For example, college attendance may have its largest effects on weight during the freshman year because this is when college students first experience increased autonomy from their parents. Therefore, I next reestimate the models allowing freshman, sophomore, junior, and senior college attendance to have different effects, using the same sample of 18- to 22-year-olds. For Model 3, I find that the freshman year significantly increases weight by about 1 pound and the sophomore year has a marginally significant effect, increasing weight by about two-fifths of a pound. Junior and senior year attendance continue to have statistically insignificant effects.

I also examine annual weight change since the prior survey, which is another attempt to control for unobserved individual characteristics. Presented as Models 4, 5, and 6 using the same format as before, some results are similar to those regarding weight change since age 16. Freshman year attendance increases weight by about 1.3 pounds for 18- and 19-year-olds, and this effect remains significant at the 1 % level. Other results are different. College attendance no longer has a statistically significant effect on weight changes when attendance is examined across all college grades in Model 5. The results from Model 6 are somewhat similar to those from Models 4 and 5: when the effects of each year of college attendance are identified separately, freshman year attendance significantly increases weight changes by about 1 pound, but college enrollment in other years does not have effects that are significantly different than zero.

Another way to control for unobserved characteristics is to examine weight in levels including individual-specific dummy variables. I identify individual-specific fixed-effects models because NLSY respondents potentially provide multiple weight observations. Fixed-effects results are presented as Models 7, 8, and 9 in Table 3, using the same format as for the other weight models. For identification, I examine a sample of 18- to 22-year-olds in each fixed-effects model. Freshman year attendance again consistently increases weight by almost 1 pound when just the freshman college year is identified in Model 7 and by about 1.5 pounds when other college years are identified separately in Model 9. Attending college, when it is considered jointly across all college years, has a significant, positive effect on weight in Model 8, but this may be largely due to the freshman college year.

Instrumental Variables

I next explore the potential for unobserved heterogeneity bias due to time-varying factors, by predicting college attendance using instruments measuring whether the respondent’s county of residence, an in-state bordering county, or an out-of-state bordering county contain a land-grant university. I reestimate the weight change and individual fixed-effects models, which already purportedly control for time-invariant factors, to explore whether also controlling for time-varying factors changes the results. The first-stage instruments are sufficiently powerful, with an F statistic for joint significance of 29.56, which is well above the often-cited threshold of 10 (Staiger and Stock 1997). However, the second-stage results (not shown) are mixed, and none of the point estimates on the covariate for predicted college attendance are statistically significant. Hausman tests are unable to reject the null hypothesis that the IV estimates are statistically different than the OLS estimates in any of the models, so I conclude that examining weight changes and including individual-specific fixed effects sufficiently control for unobserved factors. However, differences in OLS and IV estimates are more likely to appear to be statistically insignificant when the IV estimates are imprecisely measured.

Falsification Tests

If unobserved factors have been adequately controlled, the weights of college students should not differ from the weights of their noncollege peers in the year before college. I test this in a type of falsification test by estimating the effect of the pre-freshman year, where the pre-freshman year dummy variable equals 1 if the respondent was first a freshman in college in the subsequent year. Falsification results are presented in Table 4 for the weight-change models and for the fixed-effects weight-level models. As shown in Model 1, which examines weight changes since age 16 for 17- and 18-year-olds, the pre-freshman year does not have a statistically significant effect on weight. This suggests the weight changes of college freshmen are not significantly different than those of their noncollege peers prior to freshman year. Next, I expand the sample to those aged 17–23 in Model 2 and add dummy variables for the post-senior year and for the other years of college attendance. In this model, being a freshman significantly increases weight changes by three-fourths of a pound, and this effect is statistically significant at the 1 % level. Model 2 also provides some evidence that sophomore year college attendance significantly increases weight, but this effect is smaller (about three-eighths of a pound). Somewhat surprisingly, the pre-freshman year also significantly increases weight changes since age 16, by about half a pound. The post-senior year is not significantly associated with weight changes, nor is junior year or senior year college attendance. Results for models examining weight changes since the prior survey suggest the freshman year increases weight by almost 1 pound, and this effect remains statistically significant; however, weight does not change significantly in any of the other years, including the pre-freshman year (in either Model 3 or Model 4) and the post-senior year. The pre-freshman year and the post-senior year also do not affect weight when measured in levels with individual fixed effects included in Models 5 and 6, although evidence remains that freshman year and sophomore year college attendance significantly increase weight. In sum, the results generally suggest that college attendance significantly affects weight in years in which it plausibly could (e.g., in the freshman and sophomore years) and that the weights (and changes in weight) of college students and their noncollege peers do not significantly differ prior to potential college attendance.

Robustness Tests

Statistically significant weight effects have consistently been found for the college freshman year but not for other years, so I next estimate several alternative model specifications, focusing exclusively on the freshman year. Because weight is not reported on exactly the day the academic year ends, freshman year effects could change with the portion of the academic year completed (as of the survey date). Thus, I next include dummy variables for the survey month. As shown in Model 1 of Table 5, the effect of the freshman year covariate is essentially unchanged. In Model 2, I interact the freshman year covariate with each survey month included in the analysis. Results show that the college freshman year significantly increases weight by more than 1 pound for each survey month. This effect is slightly larger for April and May than for March, suggesting that weight continues to increase through the end of the freshman academic year.

The effects of freshman year attendance could be different for older freshman. To explore this, I first reestimate the model on 17- and 18-year-olds in Model 3, for 20- and 21-year-olds in Model 4, and for 22- and 23-year-olds in Model 5. Indeed, I find significant, positive effects of freshman year college attendance only for younger freshman. Results for freshmen aged 22 and older are not statistically significant. Correspondingly, a freshman attendance × age interaction term in Model 6 (with 17- through 23-year-olds) has statistically significant negative effects, again indicating that the effects of freshman year attendance pertain to younger college freshmen.

Long-Term Effects

Next, I examine longer-term implications of education on weight. First, I estimate the effects of having a college degree on weight change for NLSY97 respondents in 2010. I compare these results with similarly aged NLSY79 respondents in 1992 to explore whether the longer-term effects of education have changed over time. Last, I examine the effects of college on weight change for NLSY79 respondents in 2010, when they were aged 45–52, for an estimation of even longer-term effects of education.

Table 6 presents these estimates of the effects of having a college degree on weight changes. As before, I examine weight changes in an attempt to control for time-invariant individual-specific factors that may be unmeasurable. My first measure of weight change covers the interval since the respondent was 16 years of age. The second measure is weight changes since the prior survey. In both cases, I normalize to annual weight changes because the period between weight observations will vary for respondents.

As shown in column 1 of Table 6, a college degree decreases average annual weight changes by about one-third pound (over the period since age 16) for NLSY97 respondents in 2010. Contrary to earlier results showing that attending college may increase weight, this finding suggests that any negative weight effects of college occur after college completion, not while enrolled. This effect is larger in absolute value for similarly aged NLSY79 respondents in 1992 but smaller for 45- to 52-year-old NLSY79 respondents in 2010. Negative longer-term effects of college on weight appear to fade (e.g., decrease in absolute value) over time for specific ages and to fade over time as individuals age. Although weight changes gradually, the effects of college on weight do not appear to grow over time.

Results for weight changes since the preceding survey, presented in column 2, are only marginally significant for NLSY97 respondents in 2010. College does not significantly affect weight changes between surveys for the NLSY79 respondents in either 1992 or 2010. Unlike for weight-change models in column 1, those in column 2 indicate that, if anything, the negative weight effects of college are growing over time in absolute value for similarly aged individuals.

Longer-term effects of having a college education on weight are not identified in individual-specific fixed-effects models because, as specified, each respondent provides only one weight observation (e.g., a weight observation for 2010). Including weight observations from additional survey years (e.g., from 2008) would not solve this problem because education does not vary across surveys for respondents whose education is complete.

To explore the potential for unobserved heterogeneity bias, I estimate IV models, using the same instruments described earlier. In Table 7, for each of the samples (for NLSY97 respondents in 2010 and NLSY79 respondents in 1992 and 2010), I present second-stage results for weight changes since age 16 and since the prior survey. This table also shows the effects of the instruments in first-stage models for each sample.

Examining first-stage results, the instruments have substantially more power in explaining college education for NLSY97 respondents than for NLSY79 respondents. For this sample, a joint test for the statistical significance of the instruments produces a test statistic of 37.47, which is well above the 10 threshold recommended by Staiger and Stock (1997). The test statistic for NLSY79 respondents is between 5.45 and 7.00. In each first-stage model, residing in a county with a land-grant institution significantly increases the probability of earning a college degree. Some evidence also suggests that living in a bordering county in a neighboring state significantly increases the odds of having a college degree.

In every case but one, second-stage weight-change results indicate that having a college degree decreases the weight measure. Many of these effects are statistically significant and larger in absolute value than corresponding OLS weight-change results. In half of the cases, Hausman tests indicate that negative IV estimates are significantly larger (in absolute value) than negative OLS estimates. In most of the remaining half, Hausman tests cannot reject the null hypothesis that the IV and OLS estimates are not statistically different from one another. Taken together, this suggests, if anything, that negative long-term effects are underestimated in OLS weight-change models and represent lower bounds. Specifically, results in the second-stage IV weight-change models indicate that the negative effects of a college education are larger in absolute value for 25- to 31-year-olds, both in 2010 and in 1992, than for 45- to 52-year-olds in 2010.

Effects by Gender

Results could differ by gender because (1) SES is negatively related to weight for females but not necessarily for males, (2) employment rates may differ by gender, (3) physical activity on the job may be different for women, (4) women may face greater penalties in the labor market and marriage market for being overweight, (5) women may be more concerned about their appearance, (6) weight gain associated with pregnancy affects only women, or (7) peer influences may affect women differently (Argys and Rees 2008; Brunello et al. 2013; Cawley 2004; Chang and Lauderdale 2005; Conley and Glauber 2007; Webbink et al. 2010). Because education could influence each of these factors, I next estimate the effects of college attendance on weight by gender. For brevity, I present only the effects of key covariates (freshman year college attendance and having a college degree) for one of the model specifications (average annual weight changes since age 16).

Sample results are presented in Table 8. The results are mixed. Freshman year college attendance increases weight by about a half a pound more for males than females (1.4 pounds for males vs. 0.9 pounds for females) in the table’s top panel. In the long-term analysis, having a college degree decreases weight changes more (in absolute value) for 25- to 31-year-old females in 2010 and 1992 than for their male counterparts, but more for 45- to 52-year-old males in 2010 than for corresponding females.

Effects Along Other Demographic Dimensions

Education may be an important factor influencing disparities in weight and in the prevalence of obesity found in the literature between whites, blacks, and Hispanics (Flegal et al. 2012; Johnston and Lee 2011) and by SES (Chang and Christakis 2005; Chang and Lauderdale 2005; Scharoun-Lee et al. 2011). To explore these relationships, I repeat the gender-specific analysis, presented in Table 8, by race, marital status, and socioeconomic background (proxied by maternal education). Following the same format as for gender, results show that whites gain more weight in their freshman year of college than blacks and Hispanics (almost 1.4 pounds vs. 1.0 and 0.4 pounds, respectively). However, in the longer term, a college education significantly decreases weight for whites but not for blacks. Some long-term evidence also suggests that having a college degree decreases weight most for Hispanics.

Although too few college students are married to conduct a meaningful analysis of freshman year weight gain by marital status, longer-term effects in Table 8 show that, if anything, a college education decreases weight more for those who are married than for their nonmarried counterparts. As for the gender- and race-specific results, the effects of a college education have faded over time for particular ages and shrink with age for those who are and who are not married.

Also shown in Table 8 are results by SES. Low SES is defined as having a mother with a high school education or less; those whose mother earned a college degree or more are classified as from a high socioeconomic background. A small group of those whose mother has some college are in neither category. Those from low socioeconomic backgrounds gain significantly more weight in their freshman year of college than those from high socioeconomic backgrounds. The difference is almost 1.5 pounds versus an amount of freshman year weight gain that is not statistically different from zero. Thereafter, the socioeconomic background results are mixed. The negative weight effects from having a college education are stronger for those whose mother has a high school diploma or less in 1992 but are somewhat weaker in 2010.

BMI and Obesity

Last, I examine BMI to explore the robustness of the results to the choice of weight measure, and obesity to explore the effects of college on other parts of the weight distribution. A representative set of results from preferred models (not shown) is available upon request. Effects on BMI changes since age 16 are similar to those for weight, with freshman year attendance significantly increasing BMI and having a college degree significantly decreasing BMI. For obesity, I divide the samples into those not obese at age 16 and those obese at age 16. Then, I estimate (using linear probability models) the effects of freshman year attendance and having a college degree on obesity transitions: becoming obese and no longer being obese. Freshman year college attendance has statistically insignificant effects, potentially because relatively few 18- to 19-year-olds are obese at those ages. In the longer term, having a college degree decreases the probability of becoming obese, by about 7.5 percentage points for 25- to 31-year-olds in 1992 and 2010 and by about two-thirds that amount for 45- to 52-year-olds in 2010. The sample size for those already obese at age 16 with which to estimate the probability of no longer being obese is relatively small, which may explain why results from this model are often statistically insignificant.

Conclusions

This analysis provides new estimates of the effects of college on weight. College students from the NLSY97 gain about 4 pounds during their freshman year and between 2 and 3 pounds during subsequent college years. However, because results in this analysis show that these students gain a couple of pounds in the year before and in the year after college, this weight gain is not necessarily caused by attending college. Furthermore, similarly aged youth who are not in college gain several pounds per year. College students likely would have gained at least some weight, on average, even if they had not enrolled in college.

Using nationally representative NLSY data, weight-change models and individual-specific fixed-effects models suggest college enrollment increases weight by about a pound during the freshman year but not in the later college years. IV results are not significantly different. Although each estimation method has its own weaknesses, together they provide compelling evidence for contemporaneous effects of college attendance on weight. These results are also largely robust to examining different outcomes (i.e., BMI), different samples of respondents (i.e., balanced and unbalanced panels), analyses by gender, and falsification tests (e.g., examining the pre-freshman year).

In the longer term, after graduation, a college education seems to provide some of the protective benefits for weight typically found for other health measures. In particular, those with a college degree weigh significantly less—as many as 12 pounds less—than their peers without a college degree, and college graduates gain as much as half a pound less per year than those without a degree. In turn, college graduates are less likely to be obese. This set of findings differs from a literature that has not found beneficial effects of education on weight (e.g., BMI, obesity) for either males or females in preferred model specifications (Arendt 2005; Brunello et al. 2013; Jurges et al. 2011; Kemptner et al. 2011; Kenkel et al. 2006; Lundborg 2013). Some of the evidence in this study suggests that the effects of college have faded in magnitude over the last 20 years for people at similar ages, and these effects diminish with age within cohorts. By contrast, Webbink et al. (2010) found that the effects of college for males increase with age. Perhaps other factors affecting weight that develop in adulthood gradually dilute the impact of college.

This analysis contributes to the literatures on population health, cohort change, and health disparities by examining the relationship between college—which influences subsequent SES—and obesity at several key ages in the life cycle and over time for particular ages. In sum, the results show that over the life cycle, college may initially increase weight during the first year of enrollment, with beneficial effects on weight after graduation that diminish somewhat with age. These findings have several important implications. First, the long-run results support the belief that education offers protective weight benefits. In addition, college years represent a period when many youths become responsible for the first time for self-regulating diet, physical activity, and other behaviors. Thus, college entrance may be a crucial period determining weight trajectories. Even if attending college alters students’ lifestyles in only small ways, minor and seemingly harmless changes in behavior potentially have significant effects on weight over time. Perhaps this explains why adolescent weight is one of the most significant predictors of adult obesity (Nader et al. 2006; Ogden et al. 2010).

Examining the determinants of weight remains important because obesity is a primary risk factor for both premature death (Allison et al. 1999; Flegal et al. 2005; Fontaine et al. 2003)7 and a host of health problems, including diabetes, heart disease, high cholesterol, and hypertension (McTigue et al. 2006; Mokdad et al. 2001; Must et al. 1999). In addition, obesity reduces quality of life (Haomiao and Lubertkin 2005; Stewart et al. 2013), exacerbates late-life disabilities and physical limitations (Alley and Chang 2007; Freedman et al. 2007; Freedman and Martin 2000; Martin et al. 2010), and results in financial costs from increased medical expenditures and lost worker productivity (Andreyeva et al. 2004; Finkelstein et al. 2003; Quesenberry et al. 1998).

Acknowledgments

I would like to thank the Faculty Research and Creative Activity Committee (FRCAC) and Michael D. Allen at Middle Tennessee State University for financial support.

Notes

1

I do not use National Health and Nutrition Examination Survey data (NHANES) or Behavioral Risk Factor Surveillance System (BRFSS) data because these are not panel data sets and are therefore unable to identify year-to-year weight changes. Further, BRFSS data do not include youths, instead examining adults at least 20 years of age.

2

The NLSY97 measures of weight (and height) are self-reported and are potentially measured with error. Following Cawley’s (2000) procedure, I use self-reported and measured weights and heights in NHANES data to adjust my measures of weight and height.

3

I do not include respondents who reported their survey weight during summer months because I do not want to confound the effects of college with the effects of summer and summer’s various activities on weight. Sensitivity analyses show that including respondents who reported their weight in June does not appreciably change the reported results and conclusions.

4

Because the NLSY97 began surveying 12- to 17-year-olds in 1997, some of the respondents were in high school before the survey began and cannot provide body weight information for those years. I use an unbalanced panel in my analysis, where a “balanced” panel is defined as a consistent sample of respondents who provide a weight observation for every year (or age) included in the analysis. Otherwise, respondents who do not provide valid weight (and height) information in the six requisite post–high school graduation years (e.g., the pre-freshman, freshman, sophomore, junior, senior, and post-senior years) would be eliminated from the analysis. With an unbalanced panel, changes in sample average weight over time might reflect attrition such that, for example, low-income individuals who weigh more might be more likely to drop out, downwardly biasing the sample average weight in later survey years. On the other hand, a balanced panel discards potentially useful information. Regardless, I conduct sensitivity analyses to explore whether instead using a balanced panel would substantively change the results. Results from key models are largely unchanged, so I do not present results using the balanced panel in the tables.

5

The NLSY79 collected information about weight in 1981, 1982, 1985, 1986, 1988, 1989, 1990, 1992, 1993, 1994, and every two years thereafter. I do not use NLSY79 data to examine weight in college because weight was reported sporadically during the early 1980s, when most NLSY79 respondents were in college. Furthermore, not all NLSY79 respondents provided the information needed to identify weight during college because some began college before the NLSY79 began surveying; this would be true, for example, for a NLSY79 respondent aged 21 in 1979.

6

Similarly, Card (1995), Kling (2001), and Currie and Moretti (2003) used the presence of all two- and/or four-year colleges and universities.

7

For somewhat different results, see Mehta and Chang (2009).

References

Alley, D. E., & Chang, V. W. (
2007
).
The changing relationship of obesity and disability, 1988–2004
.
Journal of the American Medical Association
,
298
,
2020
2027
. 10.1001/jama.298.17.2020.
Allison, D. B., Fontaine, K. R., Manson, J. E., Stevens, J., & VanItallie, T. B. (
1999
).
Annual deaths attributable to obesity in the United States
.
Journal of the American Medical Association
,
282
,
1530
1538
. 10.1001/jama.282.16.1530.
American College Health Association
. (
2005
).
The American College Health Association National College Health Assessment (ACHA-NCHA): Spring 2003 reference group report
.
Journal of American College Health
,
53
,
199
210
.
Anderson, J. W., Konz, E. C., Frederick, R. C., & Wood, C. L. (
2001
).
Long-term weight-loss maintenance: A meta-analysis of U.S. studies
.
American Journal of Clinical Nutrition
,
74
,
579
584
.
Andreyeva, T., Sturm, R., & Ringel, J. S. (
2004
).
Moderate and severe obesity have large differences in health care costs
.
Obesity Research
,
12
,
1936
1943
. 10.1038/oby.2004.243.
Arendt, J. N. (
2005
).
Does education cause better health? A panel data analysis using school reforms for identification
.
Economics of Education Review
,
24
,
149
160
. 10.1016/j.econedurev.2004.04.008.
Argys, L. M., & Rees, D. I. (
2008
).
Searching for peer effects: A test of the contagion hypothesis
.
Review of Economics and Statistics
,
90
,
442
458
.
Baum, C. L., & Ruhm, C. J. (
2009
).
Age, socioeconomic status, and obesity growth
.
Journal of Health Economics
,
28
,
635
648
. 10.1016/j.jhealeco.2009.01.004.
Brunello, G., Fabbri, D., & Fort, M. (
2013
).
The causal effect of education on body mass: Evidence from Europe
.
Journal of Labor Economics
,
31
,
195
223
. 10.1086/667236.
Card, D. E. (
1995
).
Using geographic variation in college proximity to estimate the returns to schooling
. In L. N. Christofides, & R. Swidinsky (Eds.),
Aspects of labour market behavior: Essays in honor of John Vanderkamp
(pp.
201
221
).
Toronto, Canada
:
University of Toronto Press
.
Cawley, J. (
2000
).
Body weight and women’s labor market outcomes
(NBER Working Paper No. 7841).
Cambridge, MA
:
National Bureau of Economic Research
.
Cawley, J. (
2004
).
The impact of obesity on wages
.
Journal of Human Resources
,
39
,
451
474
. 10.2307/3559022.
Centers for Disease Control and Prevention (CDC)
. (
2006a
).
Body mass index (BMI)
. Retrieved from http://www.cdc.gov/nccdphp/dnpa/bmi/index.htm
Centers for Disease Control and Prevention (CDC)
. (
2006b
).
Overweight and obesity: Defining overweight and obesity
. Retrieved from https://www.cdc.gov/obesity/adult/defining.html
Chang, V. W. (
2006
).
Racial residential segregation and weight status among US adults
.
Social Science & Medicine
,
63
,
1289
1303
. 10.1016/j.socscimed.2006.03.049.
Chang, V. W., & Christakis, N. A. (
2005
).
Income inequality and weight status in US metropolitan areas
.
Social Science & Medicine
,
61
,
83
96
. 10.1016/j.socscimed.2004.11.036.
Chang, V. W., Hillier, A. E., & Mehta, N. K. (
2009
).
Neighborhood racial isolation, disorder, and obesity
.
Social Forces
,
87
,
2063
2092
.
Chang, V. W., & Lauderdale, D. S. (
2005
).
Income disparities in body mass index and obesity in the United States, 1971–2002
.
Archives of Internal Medicine
,
165
,
2122
2128
.
Chung, C., & Myers, S. L. (
1999
).
Do the poor pay more for food? An analysis of grocery store availability and food price disparities
.
Journal of Consumer Affairs
,
33
,
276
296
. 10.1111/j.1745-6606.1999.tb00071.x.
Colditz, G. A., Giovannucci, E., Rimm, E. B., Stampfer, M. J., Rosner, B., Speizer, F. E., . . . Willett, W. C. (
1991
).
Alcohol intake in relation to diet and obesity in women and men
.
American Journal of Clinical Nutrition
,
54
,
49
55
.
Conley, D., & Glauber, R. (
2007
).
Gender, body mass, and socioeconomic status: New evidence from the PSID
.
Advances in Health Economics and Health Services Research
,
17
,
253
275
. 10.1016/S0731-2199(06)17010-7.
Crombie, A. P., Ilich, J. Z., Dutton, G. R., Panton, L. B., & Abood, D. A. (
2009
).
The freshman weight gain phenomenon revisited
.
Nutrition Reviews
,
67
,
83
94
. 10.1111/j.1753-4887.2008.00143.x.
Currie, J., & Moretti, E. (
2003
).
Mother’s education and the intergenerational transmission of human capital: Evidence from college openings
.
Quarterly Journal of Economics
,
118
,
1495
1532
. 10.1162/003355303322552856.
Cutler, D., & Lleras-Muney, A. (
2008
).
Education and health: Evaluating theories and evidence
. In R. F. Schoeni, J. S. House, J. A. Kaplan, & H. Pollack (Eds.),
Making Americans healthier: Social and economic policy as health policy
(pp.
29
60
).
New York, NY
:
Russell Sage Foundation
.
Cutler, D., & Lleras-Muney, A. (
2010
).
Understanding differences in health behaviors by education
.
Journal of Health Economics
,
29
,
1
28
. 10.1016/j.jhealeco.2009.10.003.
Debate, R. D., Topping, M., & Sargent, R. G. (
2001
).
Racial and gender differences in weight status and dietary practices among college students
.
Adolescence
,
36
,
819
833
.
Douglas, K. A., Collins, J. L., Warren, C., Kann, L., Gold, R., Clayton, S., . . . Kolbe, L. J. (
1997
).
Results from the 1995 National College Health Risk Behavior Survey
.
Journal of American College Health
,
46
,
55
66
.
Drewnowski, A., & Specter, S. E. (
2004
).
Poverty and obesity: The role of energy density and costs
.
American Journal of Clinical Nutrition
,
79
,
6
16
.
Finkelstein, E. A., Fiebelkorn, I. C., & Wang, G. (
2003
).
National medical spending attributable to overweight and obesity: How much, and who’s paying
.
Health Affairs
, Advance online publication. doi:10.1377/hlthaff.w3.219.
Flegal, K. M., Carroll, M. D., Kit, B. K., & Ogden, C. L. (
2012
).
Prevalence of obesity and trends in the distribution of body mass index among US adults, 1999–2010
.
Journal of the American Medical Association
,
307
,
491
497
.
Flegal, K. M., Graubard, B. I., Williamson, D. F., & Gail, M. H. (
2005
).
Excess deaths associated with underweight, overweight, and obesity
.
Journal of the American Medical Association
,
293
,
1861
1867
.
Fontaine, K. R., Redden, D. T., Wang, C., Westfall, A. O., & Allison, D. B. (
2003
).
Years of life lost due to obesity
.
Journal of the American Medical Association
,
289
,
187
193
.
Franz, M. J., VanWormer, J. J., Crain, A. L., Boucher, J. L., Histon, T., Caplan, W., . . . Pronk, N. P. (
2007
).
Weight-loss outcomes: A systematic review and meta-analysis of weight loss clinical trials with a minimum 1-year follow-up
.
Journal of the American Dietetic Association
,
107
,
1755
1767
.
Freedman, V. A., & Martin, L. G. (
2000
).
Contribution of chronic conditions to aggregate changes in old-age functioning
.
American Journal of Public Health
,
90
,
1755
1760
. 10.2105/AJPH.90.11.1755.
Freedman, V. A., Schoeni, R. F., Martin, L. G., & Cornman, J. (
2007
).
Chronic conditions and the decline in late-life disability
.
Demography
,
44
,
459
477
. 10.1353/dem.2007.0026.
Geronimus, A. T., Hicken, M., Keene, D., & Bound, J. (
2006
).
‘Weathering’ and age patterns of allostatic load scores among blacks and whites in the United States
.
American Journal of Public Health
,
96
,
826
833
. 10.2105/AJPH.2004.060749.
Grossman, M. (
1972
).
On the concept of health capital and the demand for health
.
Journal of Political Economy
,
80
,
223
255
. 10.1086/259880.
Grossman, M. (
1972
).
The demand for health: A theoretical and empirical investigation
.
New York, NY
:
Columbia University Press
.
Grossman, M. (
1975
).
The correlation between health and schooling
. In N. E. Terleckyj (Ed.),
Household production and consumption
(pp.
147
211
).
New York, NY
:
Columbia University Press
.
Grossman, M. (
2006
).
Education and nonmarket outcomes
.
Handbook of the Economics of Education
,
1
,
577
633
. 10.1016/S1574-0692(06)01010-5.
Grossman, M., & Kaestner, R. (
1997
).
Effects of education on health
. In J. R. Behrman, & N. Stacey (Eds.),
The social benefits of education
(pp.
69
123
).
Ann Arbor
:
University of Michigan Press
.
Haberman, S., & Luffey, D. (
1998
).
Weighing in college students’ diet and exercise behaviors
.
Journal of American College Health
,
46
,
189
191
. 10.1080/07448489809595610.
Haomiao, J., & Lubertkin, E. I. (
2005
).
The impact of obesity on health-related quality-of-life in the general adult US population
.
Journal of Public Health
,
27
,
156
164
. 10.1093/pubmed/fdi025.
Huang, Y., Song, W. O., Schemmel, R. A., & Hoerr, S. M. (
1994
).
What do college students eat? Food selection and meal patterns
.
Nutrition Research
,
14
,
1143
1153
. 10.1016/S0271-5317(05)80242-8.
Johnston, D. W., & Lee, W.-S. (
2011
).
Explaining the female black-white obesity gap: A decomposition analysis of proximal causes
.
Demography
,
48
,
1429
1450
. 10.1007/s13524-011-0064-x.
Jurges, H., Reinhold, S., & Salm, M. (
2011
).
Does schooling affect health behavior? Evidence from the educational expansion in Western Germany
.
Economics of Education Review
,
30
,
862
872
. 10.1016/j.econedurev.2011.04.002.
Kemptner, D., Jurges, H., & Reinhold, S. (
2011
).
Changes in compulsory schooling and the causal effect of education on health: Evidence from Germany
.
Journal of Health Economics
,
30
,
340
354
. 10.1016/j.jhealeco.2011.01.004.
Kenkel, D., Lillard, D., & Mathios, A. (
2006
).
The roles of high school completion and GED receipt in smoking and obesity
.
Journal of Labor Economics
,
24
,
635
660
. 10.1086/504277.
Kling, J. R. (
2001
).
Interpreting instrumental variables estimates of the returns to schooling
.
Journal of Business and Economic Statistics
,
19
,
358
364
. 10.1198/073500101681020006.
Lleras-Muney, A. (
2005
).
The relationship between education and adult mortality in the United States
.
Review of Economic Studies
,
72
,
189
221
. 10.1111/0034-6527.00329.
Lowry, R., Galuska, D. A., Fulton, J. E., Wechsler, H., Kann, L., & Collins, J. L. (
2000
).
Physical activity, food choice, and weight management goals and practices among U.S. college students
.
American Journal of Preventive Medicine
,
18
(
1
),
18
27
. 10.1016/S0749-3797(99)00107-5.
Lundborg, P. (
2013
).
The health returns to schooling—What can we learn from twins?
Journal of Population Economics
,
26
,
673
701
. 10.1007/s00148-012-0429-5.
Marmot, M., Smith, G. D., Stansfeld, S., Patel, C., North, F., Head, J., . . . Feeny, A. (
1991
).
Health inequalities among British civil servants: The Whitehall II Study
.
Lancet
,
337
,
1387
1393
.
Martin, L. G., Schoeni, R. F., & Andreski, P. M. (
2010
).
Trends in health of older adults in the United States: Past, present, future
.
Demography
,
47
,
S17
S40
. 10.1353/dem.2010.0003.
McLaren, L. (
2007
).
Socioeconomic status and obesity
.
Epidemiologic Reviews
,
29
,
29
48
. 10.1093/epirev/mxm001.
McTigue, K., Larson, J. C., Valoski, A., Burke, G., Kotchen, J., Lewis, C. E., . . . Kuller, L. (
2006
).
Mortality outcomes and cardiac and vascular outcomes in extremely obese women
.
Journal of the American Medical Association
,
296
,
79
86
.
Mehta, N. K., & Chang, V. W. (
2009
).
Mortality attributable to obesity among middle-aged adults in the United States
.
Demography
,
46
,
851
872
. 10.1353/dem.0.0077.
Mokdad, A. H., Bowman, B. A., Ford, E. S., Vinicor, F., Marks, J. S., & Koplan, J. P. (
2001
).
The continuing epidemic of obesity and diabetes in the United States
.
Journal of the American Medical Association
,
286
,
1195
1200
.
Mokdad, A. H., Serdula, M. K., Dietz, W. H., Bowman, B. A., Marks, J. S., & Koplan, J. P. (
1999
).
The spread of the obesity epidemic in the United States, 1991–1999
.
Journal of the American Medical Association
,
282
,
1519
1522
.
Moretti, E. (
2004
).
Estimating the social return to higher education: Evidence from longitudinal and repeated cross-sectional data
.
Journal of Econometrics
,
121
,
175
212
. 10.1016/j.jeconom.2003.10.015.
Must, A., Spadano, J., Coakley, E. H., Field, A. E., Colditz, G., & Dietz, W. D. (
1999
).
The disease burden associated with overweight and obesity
.
Journal of the American Medical Association
,
282
,
1523
1529
.
Nader, P. R., O’Brien, M., Houts, R., Bradley, R., Belsky, J., Crosnoe, R., . . .
National Institute of Child Health and Human Development Early Child Care Research Network
. (
2006
).
Identifying risk for obesity in early childhood
.
Pediatrics
,
118
,
e594
e601
.
Ogden, C. L., Carroll, M. D., Curtin, L. R., Lamb, M. M., & Flegal, K. M. (
2010
).
Prevalence of high body mass index in US children and adolescents, 2007–2008
.
Journal of the American Medical Association
,
303
,
242
249
.
Poortinga, W. (
2006
).
Perceptions of the environment, physical activity, and obesity
.
Social Science & Medicine
,
63
,
2835
2846
. 10.1016/j.socscimed.2006.07.018.
Prentice, A. M. (
1995
).
Alcohol and obesity
.
International Journal of Obesity and Related Metabolic Disorders
,
19
,
S544
S550
.
Quesenberry, C. P., Caan, B., & Jacobson, A. (
1998
).
Obesity, health services use, and health care costs among members of a health maintenance organization
.
Archives of Internal Medicine
,
158
,
466
472
. 10.1001/archinte.158.5.466.
Sander, W. (
1995a
).
Schooling and quitting smoking
.
Review of Economics and Statistics
,
77
,
191
199
.
Sander, W. (
1995b
).
Schooling and smoking
.
Economics of Education Review
,
14
,
23
33
. 10.1016/0272-7757(94)00033-3.
Scharoun-Lee, M., Gordon-Larsen, P., Adair, L. S., Popkin, B. M., Kaufman, J. S., & Suchindran, C. M. (
2011
).
Intergenerational profiles of socioeconomic (dis)advantage and obesity during transition to adulthood
.
Demography
,
48
,
625
651
. 10.1007/s13524-011-0024-5.
Shapiro, J. M. (
2006
).
Smart cities: Qualify of life, productivity, and the growth effects of human capital
.
Review of Economics and Statistics
,
88
,
324
335
.
Smith, J. P. (
2004
).
Unraveling the SES-health connection
. In L. J. Waite (Ed.),
Aging, health, and the public policy: Demographic and economic perspectives
(Vol.
30
, pp.
108
132
).
New York, NY
:
Population Council
.
Sobal, J., & Stunkard, A. J. (
1989
).
Socioeconomic status and obesity: A review of the literature
.
Psychological Bulletin
,
105
,
260
275
. 10.1037/0033-2909.105.2.260.
Staiger, D., & Stock, J. (
1997
).
Instrumental variables regression with weak instruments
.
Econometrica
,
65
,
557
586
. 10.2307/2171753.
Stewart, S. T., Cutler, D. M., & Rosen, A. B. (
2013
).
US trends in quality-adjusted life expectancy from 1987 to 2008: Combining national surveys to more broadly track the health of the nation
.
Journal of Public Health
,
103
(
11
),
e78
e87
. doi:10.2105/AJPH.2013.301250
Webbink, D., Martin, N. G., & Visscher, P. M. (
2010
).
Does education reduce the probability of being overweight?
Journal of Health Economics
,
29
,
29
38
. 10.1016/j.jhealeco.2009.11.013.
Westerterp-Plantenga, M. S., & Verwegen, C. R. T. (
1999
).
The appetizing effect of an aperitif in overweight and normal-weight humans
.
American Journal of Clinical Nutrition
,
69
,
205
212
.
Yakusheva, O., Kapinos, K., & Weiss, M. (
2011
).
Peer effects and the freshman 15: Evidence from a natural experiment
.
Economics and Human Biology
,
9
,
119
132
. 10.1016/j.ehb.2010.12.002.