PDF Calorie Labeling in Chain Restaurants and Body Weight ...

Calorie Labeling in Chain Restaurants and Body Weight: Evidence from New York

Brandon J. Restrepo*

U.S. Food and Drug Administration

May 21, 2016

Abstract.--This study analyzes the impact of local mandatory calorie labeling laws implemented by New York jurisdictions on body weight. The analysis indicates that on average the point-of-purchase provision of calorie information on chain restaurant menus reduced body mass index (BMI) by 1.5% and lowered the risk of obesity by 12%. Quantile regression results indicate that calorie labeling has similar impacts across the BMI distribution. An analysis of heterogeneity suggests that calorie labeling has a larger impact on the body weight of lower income individuals, especially lower income minorities. The estimated impacts of calorie labeling on physical activity, smoking, and the consumption of alcoholic beverages, fruits, and vegetables are small in magnitude, which suggests that other margins of adjustment drive the body-weight impacts estimated here.

Keywords: Calorie labeling, chain restaurants, body mass index, obesity

JEL Classification Codes: H75, I12, I14, I18

* I would like to thank J?r?me Adda, David Blau, Anoshua Chaudhuri, Christian Dustmann, Javier Espinosa, Matthew Jones, Daeho Kim, Audrey Light, Trevon Logan, Corbin Miller, Matthew Neidell, Matthias Rieger, Soumyajit Sukul, Bruce Weinberg, seminar participants at Ko? University, European University Institute, U.S. Food and Drug Administration, USDA Economic Research Service, and San Francisco State University, as well as conference participants at the 2015 Southern Economic Association and Agricultural & Applied Economics Association Annual Meetings for helpful comments and suggestions. And, finally, I would like to give special thanks to the EUI's Max Weber Program for providing excellent research facilities at Villa La Fonte and the Badia Fiesolana over the 2012-2014 period, where much of the preliminary research for this project was carried out. Any opinions, findings, conclusions, or recommendations expressed are those of the author and do not necessarily reflect the views of the U.S. Food and Drug Administration.

1. Introduction Obesity remains a major public health problem in the U.S. In 2009-2010, one in three

adults was classified as obese, and no state in the nation had met the Healthy People 2010 objective of reducing the adult obesity rate to 15% (Ogden et al. 2012). Obesity increases the risk of morbidity and treating obesity-related illness imposes substantial healthcare costs on society. A recent study estimated that, in 2006, among Medicare and Medicaid beneficiaries, per capita medical spending was 36-47% higher for obese individuals than for non-obese individuals (Finkelstein et al. 2009). Cawley and Meyerhoefer (2012) estimated that obesity causes annual medical costs to rise by $3,022 (in 2008 dollars), which amounts to about 6% of median household income in 2008.

Changes in the food environment and unhealthy eating habits are important to understanding the recent rise in obesity.1,2 For example, there has been a dramatic increase in the consumption of food from restaurants, which tend to offer energy-dense and nutrient-poor food (Currie et al. 2010; Anderson and Matsa 2011).3 The estimated share of daily calories consumed coming from restaurants and fast-food establishments more than tripled between 1977 and 2008 (Lin and Guthrie 2012).

While the provision of nutrition information on packaged foods has been mandatory in the U.S. since the Nutrition Labeling and Education Act of 1990 (NLEA) took effect, foods sold or served in restaurants were exempted from this requirement. Recently, several

1 It has been shown that increased caloric intake accounted for about 75% of the rise in adult obesity in the U.S. between 1990 and 2001 (Bleich et al. 2008). 2 In an analysis of the impact of economic factors on obesity, Courtemanche et al. (2016a) found that changes in a host of economic factors explain 43% of the rise in obesity over the 1990-2010 period in the U.S., which, in large part, they found to be driven by factors related to the time costs of caloric intake (e.g. restaurant density). 3 A recent review of the literature concluded that while causality is difficult to establish, there is a wealth of evidence indicating that the consumption of restaurant food is strongly associated with increased caloric intake and a higher risk of weight gain and obesity (Rosenheck 2008).

2

U.S. jurisdictions have mandated that chain restaurants post calorie counts on menus in order to improve access to nutrition information at the point of purchase and to facilitate better informed and healthier choices. The New York City (NYC) health department was the first to implement a local calorie labeling law in July 2008 and six New York (NY) county health departments quickly followed suit by implementing similar laws in 2009 and 2010.4,5

The first contribution of this study is that it provides the first estimates of the impact of calorie labeling laws on body mass index (BMI) and the probability of obesity. Previous work has focused on estimating the response of purchase behavior to calorie information posted on menus in restaurant settings, e.g. by studying whether consumers choose lower calorie meals or buy fewer items.6 However, behavioral changes may occur outside the restaurant setting as well. For example, individuals may use the calorie information they observe on menus to decide how much to eat later in the day, they may substitute consumption towards non-chain restaurant meals, and there are many other potentially important margins of adjustment. And in addition to demand-side changes, supply-side responses to calorie labeling laws (e.g. the introduction of low-calorie menu items or reformulation to reduce the caloric content of existing products) could also have an impact on body weight.

4 NYC is composed of 5 counties: the Bronx, Kings, New York, Queens, and Richmond. Thus, a total of 11 of 62 counties in NY implemented a calorie labeling law between 2008 and 2012. The local laws in NY apply to all chain restaurants with 15 or more locations nationwide, including fast-food and full-service restaurants. Not all restaurants are chains, but chains are responsible for a disproportionate fraction of restaurant traffic. For example, in 2007, only 10% of NYC's 23,000 restaurants were chains, but they accounted for 33% of all restaurant traffic (Farley et al. 2009). 5 This study focuses its analysis on the impact of mandatory calorie labeling laws in NY because a subset of its counties has had regulations in place for longer than any other U.S. jurisdiction and there is substantial variation in the mandate across and within NY counties over the study period to exploit in estimation. Also, the legal requirements of calorie labeling laws in areas outside of NY differed from the law implemented in NY counties. For example, unlike NY jurisdictions, some other U.S. jurisdictions required a dietary statement be posted on menus along with calorie counts. 6 See Littlewood et al. (2015) and Long et al. (2015) for recent reviews of this literature.

3

This study exploits within-county variation in the availability of calorie information posted on chain restaurant menus over time brought on by implementation of mandatory calorie labeling laws and the differential timing of implementation across NY counties to identify the effect of calorie labeling on BMI. This empirical approach allows estimation of the overall impact of calorie labeling on body weight, which may operate through a wide variety of behavioral responses to calorie information posted on menus, both inside and outside of chains, as well as supply-side responses. The analysis indicates that on average implementation of calorie labeling laws in NY led to economically important and statistically significant reductions in BMI and the risk of obesity.

The second contribution that this study makes to the literature is that it adds to the understanding of the channels through which calorie labeling affects consumer behavior, by analyzing whether calorie labeling induces changes in exercise, smoking, or dietary behavior as measured by a limited set of food and beverage items captured in the BRFSS. The estimated effects of calorie labeling on physical activity, smoking participation, and alcohol, fruit, and vegetable consumption are statistically insignificant and too small to explain the body-weight impacts of calorie labeling estimated here.

The third contribution of this study is that it sheds additional light on whether estimation of the average effect of calorie labeling masks heterogeneity in the responsiveness to calorie information posted on chain restaurant menus (Robert Wood Johnson Foundation 2009; 2013). Quantile regression point estimates are similar in size across the BMI distribution and are not significantly different across quantiles. While I find that the estimated effects of calorie labeling on body weight are larger for some groups relative to others (e.g. women versus men), the estimates from different pairs of subsamples are generally not

4

significantly different from each other.7 An important exception is suggestive evidence that calorie labeling has a larger impact on the body weight of lower income individuals, especially lower income minorities.

The rest of this paper is organized as follows. First, I review the literature on the effectiveness of calorie labeling on menus in chain restaurants. Second, I summarize the data sets used in the analysis. Third, I describe the empirical approach employed in the study, explain the results of the analysis, and explore several mechanisms that may drive the results. Lastly, I provide a discussion of the results and conclude.

2. Previous Literature Many studies have examined whether calorie labeling induces individuals to make

healthier choices in restaurant settings.8,9 Elbel et al. (2009) found that calorie labeling had no impact on the calories purchased in several fast-food chain restaurants, despite the fact that 27% of those seeing calorie counts reported using them. Similarly, while Tandon et al. (2011) found that calorie labeling caused a significant increase in parents seeing nutrition information, they found no evidence that calorie labeling decreased calories purchased for either children or parents. Finkelstein et al. (2011) used transaction data from a Mexican fastfood chain and found that calorie labeling had no impacts on in-store or drive-through purchase behavior.

7 It is important to note, however, that I may be lacking power in the subsample analyses to establish that the body-weight impacts of calorie labeling are larger for some groups than others. 8 There are also many studies analyzing hypothetical menu item choices and purchase intentions. These studies used survey or laboratory experiment data and generally found evidence suggesting that calorie labeling decreases the calories of hypothetical purchases, decreases purchase intentions, and increases intentions to purchase lower calorie meals (Robert Wood Johnson Foundation 2013). 9 A recent meta-analysis by Long et al. (2015) found that calorie labeling is associated with a statistically significant reduction of 18 calories ordered per meal; among controlled studies, however, calorie labeling is found to be associated with a statistically insignificant reduction of 8 calories per meal. Another recent metaanalysis (Littlewood et al. 2015) found that calorie labeling is associated with a statistically significant reduction of 78 calories ordered per meal.

5

Bollinger et al. (2011) found that calorie labeling in Starbucks resulted in a modest average reduction of 14 calories purchased per transaction, which was driven by changes in consumers' food choices and not beverage choices.10 Wisdom et al. (2010) found that assigning calorie-labeled menus to diners at a fast-food sandwich chain caused them to order about 61 fewer calories--a reduction that was due to side-dish and drink choices and not sandwich choices. Ellison et al. (2013) found that, while assignment of menus with calorie counts alone in a full-service restaurant reduced entr?e calories, it did not significantly reduce calories from other sources such as drinks and desserts.11

The studies discussed above suggest that the impact of calorie labeling on calories ordered may depend on the menu items or type of establishments under consideration, which creates some ambiguity regarding the overall impact of calorie labeling laws. Compensatory behavior may also have important implications for the overall impact of providing calorie information. For example, Roberto et al. (2010) found that, in an experiment that took place in a university classroom, diners assigned a calorie-labeled menu ordered fewer calories during a study meal but offset this calorie reduction by consuming more calories later in the day.12

There is also evidence that supply-side responses to calorie labeling laws may have a beneficial impact on the nutrient content of restaurant foods. Namba et al. (2013) found that

10 They also found that calorie labeling had larger impacts on the purchase behavior of women and individuals who were high-calorie purchasers before calories were posted on menus. 11 In addition, they found that, among diners assigned a calorie-labeled menu, the reduction in calories ordered was larger for those who were less "health conscious" compared with those who were more "health conscious". In similar studies, Ellison et al. (2014a; 2014b) found that random assignment of calorie-labeled menus did not significantly reduce total calories ordered but the addition of a symbolic traffic light did significantly reduce total calories ordered. 12 A third group of diners was assigned a menu with calorie information and a statement about the recommended daily caloric intake for an average adult. These diners also ordered fewer calories than those who were assigned a menu with no calorie information, but this reduction was not offset by increases in calorie consumption later in the day.

6

implementation of local calorie labeling laws caused a 5% increase in what they refer to as "healthier adult entr?es" on fast-food chain restaurant menus.13 And in a survey of NYC chain restaurant managers, Bollinger et al. (2010) found that, among managers that reported changing their menus at least once a year, the probability of managers indicating that a lowcalorie option was added to their menu in the past 6 months was higher for NYC chains that were required to comply with calorie labeling requirements (chains with 15-20 locations nationwide versus those with 10 to 14).

In sum, evidence that calorie labeling reduces the amount of calories purchased in chain restaurants is mixed.14 Unlike previous studies that focus on the first-stage impact of calorie labeling, this study evaluates whether calorie labeling laws lead to a reduction in body weight. The strength of the empirical approach used here is that it allows measurement of the overall impact of calorie labeling on body weight, which may operate through a variety of demand-side and supply-side responses. To complement the body-weight analysis, I also investigate the importance of mechanisms related to dietary behavior, smoking, and physical activity. The literature suggests that the impact of calorie labeling on consumers may not be

13 Bleich et al. (2015a) and Bleich et al. (2016) found that in recent years large chain restaurants have significantly reduced the number of calories in newly introduced menu items, which they argue may be in anticipation of the federal menu labeling regulations. Bleich et al. (2015b) found that restaurants that voluntarily posted calorie information had lower average per-item calorie content than those that did not. Bruemmer et al. (2012) found that, among menu items that were on menus 6 and 18 months after calorie labeling requirements were implemented in King County WA, there were improvements in the nutrient content of chain restaurant entr?es. 14 A closely related literature examines whether the provision of nutrition information on packaged foods has beneficial impacts on health as measured by body weight. The findings in this literature are also mixed. Using a differences-in-differences estimation approach that compares nutrition label users to non-users, Variyam and Cawley (2006) found that that implementation of NLEA was associated with a decrease in BMI among only one group--non-Hispanic white females. Drichoutis et al. (2009) employ a propensity score matching approach and found no evidence that nutrition labeling affects body weight. And Loureiro et al. (2012) estimate switching regression models and found that nutrition labeling reduces the body weight of both men and women, but has a larger impact on the body weight of women.

7

uniform across individuals, which motivates an analysis of heterogeneity in the impact of calorie labeling on body weight.

3. Data The main analysis draws on data from selected state files of the 2004-2012 Behavioral

Risk Factor Surveillance System (BRFSS). The analysis sample is composed of individuals who reside in NY counties and counties in the NY-NJ-PA Metropolitan Statistical Area (MSA) that did not implement a calorie labeling law over the study period. The total number of observations in the 2004-2012 BRFSS for these counties is 136,471. I drop 6,109 observations because county information could not be identified.15 Self-reported height and weight are used to calculate an individual's BMI.16 I drop 7,080 observations due to missing information on BMI. To address the concern that outliers are driving the results, I drop 127 observations for which BMI is below 10 or above 60.17 The main regression analysis controls for the following individual-level information: age, gender, race/ethnicity, educational attainment, family income, the number of children, and marital status. The main estimation sample consists of 103,220 individuals, for whom information on county of residence, BMI, and all the above-mentioned demographics is available.

I obtained county-level information on the timing of calorie labeling laws from the Center for Science in the Public Interest. The adoption and effective dates of these laws were

15 The county identifier is suppressed for BRFSS respondents who reside in a county with fewer than 50 respondents or adult populations less than or equal to 10,000 residents. 16 Cawley (1999) developed a procedure to address empirical problems associated with self-reported height and weight data. Studies that have employed this correction have found that coefficient estimates in regressions involving measures of body weight as a dependent variable are not sensitive to using the correction (Gruber and Frakes 2006; Lakdawalla and Philipson 2002). Below, I also examine the sensitivity of my results to correcting for reporting error in height and weight, using the NHANES. I choose not to employ this correction in the main analysis because the NHANES is representative of the U.S. non-institutionalized civilian population and not representative of NY state. 17 Dropping these individuals does not affect the results of the analysis.

8

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download