The effects of food stamps on obesity.

Author:Baum, Charles L.
  1. Introduction

    Increasing numbers of Americans are obese, where obese is defined as having a body mass index (BMI) of 30 or more and where BMI equals weight in kilograms divided by height in meters squared (CDC 2006a). In fact, the latest estimates indicate that about 30% of adult Americans are currently obese, which is roughly a 100% increase from 25 years ago (Flegal et al. 1998, 2002; Ogden et al. 2006). These increases are found for both males and females, as well as for various races. As a result of this dramatic trend, adult Americans are now more likely to be obese than to smoke cigarettes.

    Public health officials in the United States have become increasingly alarmed about the growing prevalence of obesity because the medical literature finds that obesity increases morbidity and mortality (Stevens et al. 1998; Calle et al. 1999) by increasing the prevalence of diabetes, cardiovascular disease, stroke, cancer, hypertension, dyslipidemia, gout, sleep apnea, and osteoarthritis (Must et al. 1999; Chow et al. 2000; Rauscher 2000; Castro-Rodriguez et al. 2001; Field 2001; Michaud et al. 2001; Kenchaiah et al. 2002). Some have asserted that obesity will soon overtake tobacco as the leading preventable cause of death (Mokdad et al. 2004). Currently, estimates suggest that obesity contributes to between 111,909 and 365,000 premature adult deaths in the United States each year, compared to 435,000 premature deaths due to tobacco (Allison et al. 1999; Mokdad et al. 2004; Flegal et al. 2005; Mokdad 2005).

    Many societal changes do not initially appear to explain why the prevalence of obesity is increasing. For example, an increasing portion of Americans are exercising and dieting, and Americans are estimated to spend over 30 billion dollars on weight loss programs annually (Philipson and Posner 1999). Furthermore, Americans currently possess more knowledge of the consequences of obesity than ever before (Philipson and Posner 1999; Philipson 2001), and yet, Americans are more likely to be obese now than ever.

    Economists have examined various causes of obesity. For example, Philipson and Posner (1999), Philipson (2001), and Lakdawalla and Philipson (2002, 2007) suggest that increased obesity is the result of jobs becoming more sedentary; Anderson, Butcher, and Levine (2003) find evidence that maternal employment increases childhood obesity because working mothers have less time to prepare healthy meals; Cutler, Glaeser, and Shapiro (2003) assert that technological advances in food preparation, making food readily available, have caused hyperbolic consumers (defined as those who lack self-control) to overeat; and Chou, Grossman, and Saffer (2004) find that BMI and obesity have significantly increased due to increases in the number of restaurants and decreases in food prices.

    Here, I examine the effect of the Food Stamp Program on obesity. Prior to the Food Stamp Act of 1964 (and other food assistance programs passed during the twentieth century), poverty was assumed to be associated with a decrease in food consumption. Various twentieth-century government programs changed this by constructing a safety net that helps prevent starvation of those in poverty. The Food Stamp Act does this by guaranteeing an allotment of food for those below the poverty level (USDA 2003). In 2005, Food Stamp Program participants averaged $92.70 in monthly benefits at a cost of $31.0 billion to the government (USDA 2006a). It is in the period since the Food Stamp Act's passage that the prevalence of obesity has increased so dramatically. Between 1971 and 1974, the Food Stamp Program served between 9.3 and 12.8 million participants annually (USDA 2006a), and the prevalence of obesity in the United States was 14.5% (Flegal et al. 2002). These statistics have doubled; in 2005, the Food Stamp Program served an estimated 25.7 million participants (USDA 2006a), and the prevalence of obesity is currently over 30%.

    The Food Stamp Program potentially increases obesity by increasing food consumption, resulting in excessive caloric intake. Food stamps potentially increase food consumption by making the monetary cost of food zero for eligible individuals up to their food stamp allotment (though since Food Stamp Program participation rates are well below 100%, nonmonetary costs such as stigma and the opportunity cost of applying and recertifying for the benefits likely remain significant). A survey of the literature suggests that a dollar of food stamps increases food consumption between $0.17 and $0.47, which is more than an equivalent amount of cash would (Fraker 1990). It is not surprising that this would be true for constrained households, but this also appears to be true for the other 85 to 95% of food stamp households that are unconstrained (Fraker 1990).

    Although recipients could potentially use food stamps to buy healthier foods, recent evidence by Wilde, McNamara, and Ranney (1999) suggests that food stamp recipients consume significantly more sugar and fat than eligible nonrecipients. Additional evidence by Whitmore (2002) indicates that food stamp recipients in San Diego and Alabama in the 1990s consumed more soft drinks than peers who instead received cash benefits. If so, then it is possible that recipients not only consume more food, but they also consume more of the foods likely to lead to weight gain.

    Food stamps might also exacerbate obesity by promoting binge eating. Townsend et al. (2001) suggest that abundant food at the beginning of each monthly food stamp cycle leads to overeating, with food becoming scarce at the end of each cycle. They argue that the net effect of this cycle is weight gain. However, Townsend and colleagues also note that food insecurity and Food Stamp Program participation are related. (1) Food stamps could affect obesity by affecting (reducing) food insecurity; alternatively, those who are food insecure could be more likely to apply for and receive food stamp benefits, in which case food stamps and obesity would at least partially be correlated with each other through a third, potentially unmeasurable, factor (food insecurity). (2)

    Only a couple of studies have examined the effects of Food Stamp Program participation on obesity. In the nutrition literature, seminal work by Gibson first examined this link. (3) Gibson (2003) finds that Food Stamp Program participation among low-income women (but not men) is significantly associated with increased obesity. In concurrence, economists Meyerhoefer and Pylypchuk (2008) find that contemporaneously measured food stamp receipt has statistically significant positive effects on low-income women but not on low-income men.

    In this project, I estimate the relationship between food stamp benefits and the probability of being obese and the obesity gap (which is the prevalence of obesity multiplied by the average amount by which BMI exceeds the obesity threshold) using National Longitudinal Survey of Youth (NLSY79) data. I focus the analysis on subsamples of income-eligible males and females and control for possible omitted variable bias using an individual-specific fixed effects estimator. I attempt to build on the work of Gibson (2003) and Meyerhoefer and Pylypchuk (2008) by exploring the dynamic relationship between food stamps and obesity, where current weight is linked to past weight and past food stamp receipt. This seems like an important area for research because contemporaneously measured food stamp receipt would not be expected to have an instantaneous and substantial effect on weight. Instead, since current weight is not independent from past weight, the food stamp-weight relationship is likely much more complex. To do this, I estimate (i) models that explain weight changes over time with controls for initial weight status, (ii) models that identify the effects of current and past program receipt to explore whether food stamps have lagged effects, (iii) models that identify the effects of patterns of food stamp receipt, including short-term, medium-term, and long-term receipt, as well as the effects of receiving benefits in multiple spells, and (iv) models that explain the hazard rates for becoming obese at particular times during a 15 year period conditional on not yet being obese.

  2. Data

    Here, I use 1979 cohort National Longitudinal Survey of Youth (NLSY79) data to estimate the effects of food stamp benefits on obesity. In 1979, the NLSY79 began annually interviewing a cohort of 12,686 respondents (6283 of whom were female) who were between the ages of 14 and 21. The cohort initially included oversamples of blacks, Hispanics, low-income whites, and military personnel. The military sample was dropped in 1984, and the low-income whites were dropped in 1990, and I do not include either in my analysis for that reason. (4) After the 1994 survey, the NLSY79 began interviewing biennially, and these respondents have since been re-interviewed on that basis.

    The NLSY79 identifies the weight of each respondent in pounds at the time of the 1981, 1982, 1985, 1986, 1987, 1989, 1990, 1992, 1993, 1994, 1996, 1998, and 2000 surveys. The NLSY79 identifies height in inches at the time of the 1981, 1982, and 1985 surveys. I assume that height does not change after the 1985 survey because NLSY79 respondents are at least 20 years of age at that time. To measure obesity, I first calculate each individual's BMI. Obesity is then defined as a BMI of 30 or more (CDC 2006b). This calculation results in a sample of 25,249 male and 24,328 female person-year observations that provide valid weight and height information and valid responses required to create the other covariates used throughout the analysis.

    To maintain a consistent sample across all survey years, I eliminate respondents who did not provide valid weight (and height) information in all surveys that collect weight information. Otherwise, changes in sample average weight over time might reflect attrition (where, for...

To continue reading