soc paper summary

Journal of Public Economics 94 (2010) 967–974

Contents lists available at ScienceDirect

Journal of Public Economics

journal homepage: www.elsevier.com/locate/jpube

The effects of soft drink taxes on child and adolescent consumption and weight outcomes☆

Jason M. Fletcher a, David E. Frisvold b,⁎, Nathan Tefft c

a School of Public Health, Division of Health Policy and Administration, Yale University, 60 College Street, New Haven, CT 06510, USA b Department of Economics, Emory University, 1602 Fishburne Drive, Atlanta, GA 30322-2240, USA c Department of Economics, Bates College, 4 Andrews Road, Lewiston, ME 04240, USA

☆ This study was supported, in part, by the Robert Wo Emory Global Health Institute. We thank Alexandra Ehrl support with the restricted NHANES data and Jeremy Gre John Zimmerman for excellent research assistance. We Lori Kowaleski-Jones, seminar participants at Bowdoin Rudd Center for Food Policy and Obesity at Yale Univer University, and participants at the 2009 International World Congress, 2009 Robert Wood Johnson Foundatio 2009 Association for Policy Analysis and Management fa Association of America Annual Meeting, and 2010 Economists conference for helpful comments. The fin paper are those of the authors and do not necessaril Research Data Center, the National Center for Health Disease Control and Prevention. ⁎ Corresponding author.

E-mail addresses: [email protected] (J.M. Fletch (D.E. Frisvold), [email protected] (N. Tefft).

0047-2727/$ – see front matter © 2010 Elsevier B.V. Al doi:10.1016/j.jpubeco.2010.09.005

a b s t r a c t

a r t i c l e i n f o

Article history: Received 7 September 2009 Received in revised form 8 September 2010 Accepted 8 September 2010 Available online 17 September 2010

JEL classifications: I18 H75

Keywords: Obesity Soft drink taxation

Childhood and adolescent obesity is associated with serious adverse lifetime health consequences and its prevalence has increased rapidly. Soft drink consumption has also expanded rapidly, so much so that soft drinks are currently the largest single contributors to energy intake. In this paper, we investigate the potential for soft drink taxes to combat rising levels of child and adolescent obesity through a reduction in consumption. Our results, based on state soft drink sales and excise tax information between 1989 and 2006 and the National Health Examination and Nutrition Survey, suggest that soft drink taxation, as currently practiced in the United States, leads to a moderate reduction in soft drink consumption by children and adolescents. However, we show that this reduction in soda consumption is completely offset by increases in consumption of other high-calorie drinks.

od Johnson Foundation and the ich and Stephanie Robinson for en, Alejandra Carrillo Roa, and also thank Meena Fernandes, College, Mt. Holyoke College, sity, and Washington and Lee Health Economics Association n Healthy Eating Conference, ll conference, 2010 Population American Society of Health dings and conclusions in this y represent the views of the Statistics, or the Centers for

er), [email protected]

l rights reserved.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

While soft drink taxes have been used for decades as a way to raise revenues, lately there have been an increasing number of proposals to considerably increase these taxes in order to combat the rapidly growing obesity epidemic in the United States. These proposals have often framed soda taxation as a “sin tax” and made comparisons between soft drink taxation and cigarette taxation, which has both lowered tobacco consumption and raised considerable revenues. The price elasticity for soda is estimated to range between −0.8 and −1.0 (Andreyeva et al., 2010), which indicates that further taxation could

lead to substantial consumption reduction. Additionally, soft drink revenues in the United States are approximately $70.1 billion per year, suggesting a relatively large tax revenue potential, even accounting for the drop in consumption from a soda tax (Sicher, 2007); in comparison tobacco revenues are approximately $93.1 billion per year (Tobacco-NAFTA Industry Guide, 2009). Thus soft drink taxation could improve health by lowering consumption, as well as generate substantial revenues to relax government budget constraints and could potentially be used for further obesity prevention or reduction. However, the potential behavioral responses to increasing soda taxes have not been fully examined. There is relatively weak information on the tax elasticity on consumption of soft drink taxes, and, importantly, the substitution patterns between soft drinks and other (potentially high calorie) beverages has not been adequately explored in this context. Thus soda taxes could have unintended consequences and may be an ineffective “obesity tax”.

There are multiple examples documenting how behavioral responses to public policies can counteract the intent of the policies or lead to unintended consequences. One of the best known examples in the economics literature is an evaluation of a set of automobile safety regulations in the late 1960s, where Peltzman (1975) shows that drivers responded to increased safety regulations by driving faster, which completely offset any reductions in highway fatalities. More recently, Adams and Cotti (2008) show that smoking bans in bars have lead to increases in fatal accidents due to an increase in the distance driven to bars that allow smoking. Adda and Cornaglia (2010) find that smoking bans increase the exposure of non-smokers to tobacco smoke

968 J.M. Fletcher et al. / Journal of Public Economics 94 (2010) 967–974

as a result of the relocation of smokers away from bars and restaurants. There is also some evidence documenting an unintended rise in obesity and weight due to smoking bans (Fletcher, 2009) and higher cigarettes taxes, although there is mixed evidence on taxes (Chou et al., 2004, 2006; Gruber and Frakes, 2006; Baum, 2009). Courtemanche (2009) seeks to harmonize these disparate findings and has shown a counterintuitive reduction in obesity following cigarette taxation, where the proposed behavioral mechanisms are through changes in exercise and food consumption. Additionally, Evans and Farrelly (1998) find that smokers respond to higher cigarette taxes by smoking cigarettes with higher tar and nicotine content. Similarly, Adda and Cornaglia (2006) find that smokers adjust their behavior to smoke more intensely in response to higher taxes. Finally, recent theoretical and simulation research by Schroeter et al. (2008) shows there are plausible scenarios where taxes on certain types of food (e.g. food away from home) could lead to increases in body weight simply due to substitution among types of food. Likewise, soda taxes could cause substitution to other, high-calorie beverages and increase or have no effect on net caloric intake. In the extreme, these potential substitution patterns could result in a case where individuals consume no soda but offset soda consumption with other high-calorie beverages, such as fruit juice, juice drinks, or whole milk. The impact of these behavioral responses would be no tax revenues as well as no weight reduction and the policy would accomplish neither of its proposed goals.

In this paper we combine newly collected soft drink tax data between 1989 and 2006 with the restricted-access version of the nationally representative National Health Examination and Nutrition Survey (NHANES) in order to examine the effects of soft drink taxes on child and adolescent soft drink consumption, substitution patterns, and weight outcomes. We use standard empirical specifications from the cigarette taxation literature (Chaloupka and Warner, 2000), including year and state fixed effects, and conduct a series of robustness checks to increase confidence in the results. Overall, we find evidence of moderate reductions in soft drink consumption from current soda tax rates. However, we also show that reductions in calories from soda are completely offset by increases in calories from other beverages. Thus, we find that, as currently practiced, soda taxes do not reduce weight in children and adolescents and is, therefore, likely an ineffective “obesity tax”. These results suggest that public health policymakers should consider behavioral responses when crafting policies to reduce obesity.

1 See Chriqui et al. (2008) for an overview of U.S. soft drink taxes in 2007. 2 http://online.wsj.com/article/SB124208505896608647.html#mod=rss_Health

(last accessed May 12, 2009). 3 http://www.cnn.com/2008/HEALTH/12/18/paterson.obesity/ (last accessed May

11, 2009). 4 In addition to attempting to reduce obesity, the goals of these proposals were also

to raise revenue. The tax in New York was proposed to raise revenues to help balance the state budget and the Center for Science in the Public Interest proposed to raise revenues to pay for national health care reform.

5 Powell et al. (2009) also take a reduced form approach and focus on adolescent weight. Like Fletcher et al. (2010a), they find no evidence of reduced form effects on adolescent weight. However, the interpretation of their results is unclear because they are unable to control for unobserved state-level characteristics that could be correlated with soft drink taxes. They also are only able to examine a small set of ages (8th, 10th, and 12th graders), and can provide no evidence of whether adolescents are price sensitive in their soda consumption decisions.

2. Background literature

The rise in childhood and adolescent obesity in the US and other developed countries has been the source of considerable debate and public policy effort. The effects of obesity on chronic health conditions have been compared to the effects of aging twenty years in adults, and the health costs associated with obesity are even greater than two other behaviors widely recognized to cause significant harm: cigarette smoking and alcohol consumption (Sturm, 2002). Since childhood and adolescent obesity may increase the risk of adult obesity (Nader et al., 2006), prevention, rather than treatment, of adult obesity may be the more effective tool in limiting lifetime health care costs. Finkelstein et al. (2009) estimate that the medical costs attributable to obesity were as high as $147 billion in 2008, which is nearly 10% of all medical spending nationally. The additional medical costs of obesity represent an externality because the obese do not fully pay for their higher costs within pooled health insurance and public health insurance programs and because being insured increases obesity (Bhattacharya et al., 2009). There have been many efforts at reducing the burden of obesity, some focus on food prices and advertising (Chou, Rashad, and Grossman 2008) and others highlight the need to increase exercise opportunities (Cawley et al., 2007). Recently, there have been a large number of proposals that focus on reducing

consumption of products with little nutritional value, such as soft drinks.

Soft drinks were the single largest contributor to energy intake during the last decade (7%) (Block, 2004) and soft drink consumption increased by almost 500% during the past 50 years (Putnam and Allshouse, 1999). While soft drinks are a large and growing category of caloric intake, it may be unclear whether focusing on a single class of consumption can lead to weight change. In fact, there is emerging evidence that small net changes in caloric consumption can lead to substantial changes in the prevalence of obesity over time (Cutler, Glaeser, and Shapiro, 2003). For example, Hill et al. (2003) suggest that reducing energy intake by only 100 cal per day, which is less than one fewer can of soda per day, could prevent weight gain in over 90% of the population. Ludwig et al. (2001) demonstrate that consuming one additional sugary drink per day over a period of eighteen months increased the odds of being obese in children by 60%. Harnack et al. (1999) find that total energy intake is positively associated with the consumption of nondiet soft drinks. Thus, it may be possible to effectively reduce weight by targeting a single food item.

A policy that has garnered recent attention as a possible method for reducing weight and thus improving health is a tax on soft drinks.1

The taxation of soft drinks by states, which dates back to at least 1920 (New York Times, 1920), has historically been used to raise revenue (Caraher and Cowburn, 2005). Recently, however, taxes on soft drinks have been proposed at federal, state, and local levels as an indirect tax to reduce obesity, while continuing to raise revenue (Chouinard et al., 2007; Brownell and Frieden, 2009). For example, in May 2009 the Center for Science in the Public Interest planned to propose a federal excise tax on sweetened soft drinks and other beverages.2 Brownell and Frieden (2009) suggest a tax equal to a penny per ounce, which is larger than all prior taxes on soft drinks, to our knowledge. Governor David Paterson of New York proposed an “obesity tax” in December 2008, which consisted of an 18% tax on sugared beverages.3,4

These tax proposals have been offered with little evidence to assess their potential effectiveness. In fact, most of the rigorous evidence that exists appears to suggest limited effectiveness. For example, in the context of adult outcomes, Fletcher et al. (2010a) use repeated cross-sections of the Behavioral Risk Factor Surveillance System (BRFSS) data combined with state and year fixed effects to show that the reduced from effect of soft drink taxes on adult weight is negligible. The reduced form approach was unable, though, to shed light on whether there were consumption responses to taxation.5

While the current evidence suggests no effects from soft drink taxation on adult weight, it is possible that there may be differential effects of soda taxation on children. For example, Chaloupka and Warner (2000) review multiple studies of cigarette price elasticity, noting that youth have often been found to be more responsive to price than adults (although these conclusions are complicated by the fact that youth are also subject to access restrictions). One explanation for this possibility is that children and adolescents may spend a larger share of their budget on soft drinks than adults or households and are

969J.M. Fletcher et al. / Journal of Public Economics 94 (2010) 967–974

therefore more responsive to changes in price. If youth exhibit a relatively strong price response as in the case of cigarette consump- tion, policies that moderately change soft drink prices could lead to substantial declines in obesity over time among children and adolescents.

3. Conceptual and empirical framework

We conceptualize the demand for soft drinks by child i in state s as a function of individual and family socio-demographic characteristics (X), household income (I), and the prices of soda (Ps) and other beverages (Po):

sodais = f Ps τsð Þ; Po; Xi; Iið Þ; ð1Þ

where soft drink prices are a function of soft drink taxes imposed by the state (τs). Our focus in this paper is estimating the effects on soda consumption and weight from increasing soda tax rates.6 But in order to gain a complete understanding of the effects, it is useful to recall the decomposition of the effects:

∂sodais ∂τs

= ∂sodais ∂Ps

∂Ps ∂τs

; ð2Þ

where ∂sodais ∂Ps

is the price elasticity of demand of soda and ∂Ps ∂τs

is the

proportion of tax pass-through or shifting from suppliers to consumers. Because we do not have adequate price data for the individuals in this sample and because price is endogenous in demand analysis (Gruber and Frakes, 2006), we focus on the reduced form effects of (exogenous) taxes on consumption. Before presenting our empirical model, we outline what is known about the two components of the reduced form estimates. This discussion also highlights the uncertainty of calculating consumption responses when not using evidence from direct tax effects.

Estimates of the price elasticity of demand for soft drinks are available, but there does not seem to be a consensus. While a recent survey of the literature suggest typical price elasticities of −0.8 to −1.0 (Andreyeva et al., 2010), more recent examples of compensated price elasticity estimates range from −0.15 (Zheng and Kaiser, 2008) to −1.90 (Dharmasena and Capps, 2009). The differences in estimates are, in part, due to the specificity of food and drink categories that might be considered candidates for complements or substitutes. In any case, there is a relatively large range of uncertainty for policy makers to predict demand responses to price changes given the evidence accumulated thus far.

In addition to the uncertainty surrounding price elasticities, there is scant evidence on how soft drink taxes might affect behavior through prices. Since policy makers often do not set prices directly, they must rely on indirect methods such as taxation. Kotlikoff and Summers (1987) describe the theory of tax incidence and show that tax shifting, or the proportion of a tax that is reflected by a change in price, varies by market. In fact, Besley and Rosen (1999) show that the change in soft drink price exceeds a tax change by 29% and suggest that this overshifting of the tax burden is the result of imperfect competition in the soft drink industry.7 Overall, however, as a result of the recent changes in states' soft drink taxes, additional work is needed in this area.

Even after establishing the effect of taxes on consumption behavior through prices, the further question remains regarding how effectively reducing soft drink consumption will improve child weight and obesity prevalence. In fact, if youths reduce their consumption of caloric and non-caloric soft drinks in response to a

6 We assume the income effect is unimportant in these decisions. 7 Kenkel (2005) also finds evidence of overshifting between alcohol taxes and

prices.

tax but increase their consumption of non-taxed high-calorie beverages such as juice or whole milk, then a tax on soft drinks could actually increase weight (Schroeter et al., 2008). For example, while a 12 ounce can of soda contains 140 cal, 12 oz of whole milk contains 219 cal. The empirical uncertainty surrounding soft drink price elasticity and tax shifting combined with the theoretical ambiguity of the effect of soft drink taxes on obesity warrants an analysis of the direct tax effects on both consumption and weight outcomes when considering the dual policy goals of revenue generation and child health.

3.1. Empirical framework

Our empirical strategy regarding soft drink taxes follows the literature on cigarette taxation (Chaloupka and Warner, 2000) and work examining adult BMI (Fletcher et al., 2010a) by estimating state and year fixed effects models using the repeated cross-sections of the NHANES datasets. Specifically, we estimate the empirical models,

Yistq = β′1Xistq + β2Tstq + μs + δt + γq + εistq; ð1Þ

where Yistq is the consumption or weight outcome of individual i in state s in year t in quarter q, which is determined by individual and environmental characteristics (Xistq), the soft drink tax in state s in year t in quarter q (Tstq), and state, year, and quarter of year fixed effects.8 Using this strategy, the impacts of soft drink tax rates are identified from changes in the tax rate within states over time. To estimate the direct impact of soft drink taxes on soda consumption, we examine whether any soft drinks are consumed, the amount of calories from soft drinks, and the total grams of soft drinks that are consumed. To estimate whether children respond to soft drink taxes by changing their consumption of other beverages, we examine the amount of calories consumed from juice, juice-related drinks, and whole milk. To estimate the impact of soft drink taxes on weight outcomes, we examine an age and sex-specific normalized measure of body mass index and the weight classifications of obese, overweight, and underweight.

A limitation of this strategy is the potential that unobserved characteristics that vary within states and over time are related to soft drink taxes and BMI. To address this limitation, we estimate additional specifications that also include variables that measure state economic and health conditions.

4. Data

4.1. Soft drink taxes

States currently tax soft drinks through excise taxes, sales taxes, and special exceptions to food exemptions from sales taxes. For this paper, we define the soft drink tax as the tax on soft drinks net of taxes on other food items.

To determine the soft drink tax rate, we combine sales tax information with details on other soft drink taxes. The “Book of the States” (The Council of State Governments, 1990–2007) is used to identify state sales tax information. Published annually, it lists tax information and whether each state has a sales tax food exemption. This resource, LexisNexis Academic, and state departments of revenue web sites allow us to compile state sales tax rates and effective food tax rates by quarter. To understand how states tax soft drinks, we conduct further research with the above resources. Using LexisNexis Academic and departments of revenue web sites, we are able to

8 With our relatively short panels of states (1989–1994 and 1999–2006) and because not every state is included in the survey in every year, we do not include state-specific time trends in our preferred results, but we do perform several robustness checks outlined below.

Table 2 Descriptive statistics. Sources: NHANES 1989–1994 and 1999–2006.

Variable Mean Standard error

Sample size

BMI Z-score 0.418 0.011 22,132 Obese 0.148 0.004 22,132 Overweight 0.297 0.005 22,132 Underweight 0.033 0.002 22,132 Total calories 2063.071 11.293 21,040 Total calories from soft drinks 115.247 2.531 21,040 Consumed any soft drink 0.564 0.006 21,040 Total grams of soft drink consumption 314.876 6.645 21,040 Total calories from juice 51.626 1.270 21,040 Consumed any juice 0.338 0.006 21,040 Total grams of juice consumption 110.739 2.621 21,040 Total calories from juice drinks 66.414 1.613 21,040 Consumed any juice drinks 0.369 0.006 21,040 Total grams of juice drinks consumption 170.059 4.713 21,040 Total calories from whole milk 58.617 1.525 21,040 Consumed any whole milk 0.254 0.005 21,040 Total grams of whole milk consumption 94.329 2.489 21,040 Dietary recall is based on a weekday 0.626 0.005 21,040 Vitamin C intake (dietary recall measure) (mg) 88.438 1.045 21,040 Caffeine intake (dietary recall measure) (mg) 34.499 1.042 12,707 Serum intake of vitamin C (mg/dL) 1.110 0.008 9265 Serum intake of vitamin D (ng/mL) 25.923 0.116 10,449 Female 0.491 0.005 22,438 Age 10.513 0.043 22,438 Black 0.149 0.002 22,438 Other race/ethnicity (including Hispanic)

0.241 0.004 22,438

White 0.610 0.005 22,438 Soft drink tax rate 2.707 0.029 22,438

Notes: Descriptive statistics are weighted using the NHANES survey weights.

Table 3 The impact of soft drink taxes on soft drink consumption and calories consumed from soft drinks.

Consumed a soft drink

Total grams of soft drink consumption

Calories from soft drinks

Soft drink tax rate

−0.005 −18.052** −5.920** (0.005) (7.333) (2.834)

Observations 21,040 21,040 21,040 R-squared 0.089 0.169 0.161

Notes: Heteroskedasticity-robust standard errors in parentheses that allow for clustering within states. Additional variables include female, age, age squared, black, other race, weekday, state, year, and quarter. All regressions utilize NHANES survey weights. *Significant at 10%; **significant at 5%; ***significant at 1%.

Table 1 Summary statistics: soft drink tax rates, 1989–2006.

Year All states States with a positive tax rate

Mean tax rate

Standard deviation

Count Mean tax rate

Standard deviation

1989 1.623 2.335 20 4.139 1.841 1990 1.839 2.526 21 4.465 1.906 1991 1.971 2.591 22 4.569 1.882 1992 2.067 2.587 23 4.583 1.776 1993 2.334 2.919 24 4.960 2.220 1994 2.334 2.919 24 4.960 2.220 1995 2.084 2.618 23 4.621 1.822 1996 2.076 2.608 23 4.604 1.815 1997 2.076 2.608 23 4.604 1.815 1998 1.954 2.603 21 4.745 1.742 1999 1.934 2.584 21 4.698 1.748 2000 1.875 2.544 20 4.783 1.549 2001 1.758 2.488 19 4.718 1.564 2002 1.728 2.550 18 4.897 1.642 2003 1.755 2.589 18 4.974 1.663 2004 1.895 2.676 19 5.087 1.661 2005 1.888 2.667 19 5.067 1.658 2006 1.890 2.674 19 5.074 1.677

Note: Column variables represent means or percents across all states for the given year.

970 J.M. Fletcher et al. / Journal of Public Economics 94 (2010) 967–974

determine if soft drinks fall under food exemptions, or if they are taxed wholly or additionally by sales or excise taxes.

We convert all tax descriptions such that they may be incorporated into tax rates.9 For taxes that are quoted in dollars per quantity, we

9 “Soft drinks” are commonly defined similarly across states, although this is not always true. The broadest definitions include nonalcoholic beverages with added natural or artificial sweeteners. In most cases the definitions include both carbonated and noncarbonated beverages as well as soda water. In some cases there is a maximum percent content of fruit or vegetable juice, ranging anywhere from 10% to 50%, above which beverages may be excluded. Milk, milk products, and milk substitutes such as soy or rice are explicitly excluded in most cases.

convert any level tax into a percent of expenditure using the Bureau of Labor Statistics annual nationwide price index for a quantity of soda. We do not include taxes on soft drink syrups because we do not have information on the expected amount of syrup per quantity of soft drink.

Using the effective dates of each tax, we are able to combine information on excise taxes and sales taxes to calculate quarterly soft drink tax rates for each state. We confirm this data collection effort by comparing with soft drink tax descriptions in Jacobson and Brownell (2000) and Chriqui et al. (2008).

Table 1 shows the average annual tax rates across all states for 1989 through 2006. The average soft drink tax rate varies between 1.5 and 2.3% during this period. The number of states with any tax on soft drinks in each year varies between 19 and 24 and, among states with a tax, the average rate varies between 4.1 and 5.1%. Also, there were 53 tax rate changes within states over time.

4.2. NHANES

The NHANES surveys are administered by the National Center for Health Statistics (NCHS) of the CDC to assess the health and nutritional status of the civilian, non-institutionalized population of the United States using a complex, multistage probability sample design. NHANES III includes nearly 34,000 respondents and was conducted between 1988 and 1994. In 1999, the NHANES program changed to consist of a nationally representative sample of about 5000 persons each year; however, the sampling design remained similar to NHANES III. Of these recent surveys, we will utilize information from 1999 to 2006.

NHANES data is collected through survey questionnaires and physical examinations that occur primarily in a mobile examination center. Importantly for this analysis, the NHANES data contain information on body mass index (BMI), youth's soft drink and other beverage consumption, and demographic characteristics. Additional- ly, state of residence information is available through the NCHS Restricted Data Center, which allows us to merge our state-level tax information with the individual level data. As a result of a possible disclosure risk as deemed by the NCHS staff with the restricted-access data in our analysis, we exclude 1988 from our sample.

Height and weight were measured by trained health technicians during the physical examinations and BMI was calculated as weight in kilograms divided by height in meters squared. We construct dichotomous measures of obese (BMI≥95th percentile of the historical age of month- and sex-specific distribution), overweight and obese (which we call overweight throughout the text) (BMI≥85th percentile), and underweight (BMIb5th percentile) and the continuous measure of BMI Z-score10 from BMI for all individuals

10 The BMI Z-score for an individual is calculated as the BMI minus the mean BMI of the reference population, which is then divided by the standard deviation of the reference population, where the reference population is all individuals of the same sex and month of age. Thus, the units of the BMI Z-score are standard deviations from the mean.

13 Serum levels of vitamin C are only available for ages 6 and older and are not available between 1999 and 2002. Missing values between 1988 and 1994 occurred because of “inadvertent misdilution of the ascorbic acid–serum ratio” (CDC, 2006, p. 123). Serum levels of vitamin D are available for ages 12 and older in 1988 through 1994, ages 6 and older in 2001 through 2002, ages 3 and older in 2003 through 2006, and are not available in 1999 through 2000. The specific measure in NHANES is 25 hydroxyvitamin D (25(OH)D), which is the most reliable index of vitamin D nutritional status and has a half-life of 2 to 3 weeks (CDC, 2009, 2010). There are at least two sources of measurement error in serum levels of vitamin D in NHANES (CDC, 2010). The first source is a reformulation of the DiaSorin radioimmunoassay (RIA) kit, which lead to differences in the measurement of 25(OH)D between 1988 through

Table 4 The impact of soft drink taxes on non-soft drink beverage consumption and calories consumed from non-soft drink beverages.

Juice consumption Juice drink Whole milk

Panel A: Caloric intake Soft drink tax rate −0.058 1.857 7.670***

(1.521) (2.332) (2.156) Observations 21,040 21,040 21,040 R-squared 0.033 0.032 0.066

Panel B: Grams of consumption Soft drink tax rate 1.312 5.499 11.153***

(3.099) (6.830) (3.532) Observations 21,040 21,040 21,040 R-squared 0.031 0.034 0.064

Notes: Heteroskedasticity-robust standard errors in parentheses that allow for clustering within states. Additional variables include female, age, age squared, black, other race, weekday, state, year, and quarter. All regressions utilize NHANES survey weights. *Significant at 10%; **significant at 5%; ***significant at 1%.

Table 6 The impact of soft drink taxes on nutrient intake.

Serum intake of vitamin C

Serum intake of vitamin D

Vitamin C intake (dietary recall measure)

Caffeine intake (dietary recall measure)

Soft drink tax rate

0.008 0.072 1.634 −3.164*** (0.011) (0.372) (2.019) (1.141)

Observations 9260 10,445 21,040 12,707 R-squared 0.172 0.371 0.043 0.160

Notes: Heteroskedasticity-robust standard errors in parentheses that allow for clustering within states. Additional variables for the serum intake outcomes include female, age, age squared, black, other race, the length of the fasting period in minutes, the time of the day that blood was drawn (morning, afternoon, and evening), state, year, and quarter. For the dietary recall outcomes, additional variables include female, age, age squared, black, other race, weekday, state, year, and quarter. All regressions utilize NHANES survey weights. *Significant at 10%; **significant at 5%; ***significant at 1%.

Table 5 The impact of soft drink taxes on total caloric intake.

Total caloric intake

Soft drink tax rate −7.840 (12.353) [−32.944, 17.264]

Observations 21,040 R-squared 0.145

Notes: Heteroskedasticity-robust standard errors in parentheses and confidence interval in brackets that allow for clustering within states. Additional variables include female, age, age squared, black, other race, weekday, state, year, and quarter. All regressions utilize NHANES survey weights. *Significant at 10%; **significant at 5%; ***significant at 1%.

Table 7 The impact of soft drink taxes on BMI, obese, and overweight.

BMI Z-score Obese Overweight Underweight

Soft drink tax rate 0.015 0.009 0.002 −0.002 (0.016) (0.006) (0.011) (0.003)

Observations 22,132 22,132 22,132 22,132 R-squared 0.028 0.022 0.027 0.008

Notes: Heteroskedasticity-robust standard errors in parentheses that allow for clustering within states. Additional variables include female, age, age squared, black, other race, state, year, and quarter. All regressions utilize NHANES survey weights. *Significant at 10%; **significant at 5%; ***significant at 1%.

971J.M. Fletcher et al. / Journal of Public Economics 94 (2010) 967–974

between the ages of 3 and 18 using the sex-specific BMI for age thresholds from the 2000 CDC Growth Charts.

During the physical examinations, survey respondents completed a 24-hour dietary recall with a trained dietary interviewer that detailed all foods and beverages, except water, that were consumed in the previous 24 h.11 Children aged five years and younger, and many children aged six to 11 years, completed the dietary recall through a proxy respondent. The dietary interviewers contacted schools and other care providers to obtain complete dietary intake information. Individual foods were coded and classified using the U.S. Department of Agriculture's Survey Nutrient Database System. Using the dietary intake information, we construct measures of soft drink consumption, including whether the youth consumed a soft drink during the recall period, the total grams consumed, and the total calories consumed from soft drinks. To explore the possibility of substitution effects, we also construct similar measures for juice, juice-like drinks, and whole milk.12

To further examine the nutritional impacts of soft drink taxation, we examine nutrient intake derived from the dietary recall information from the laboratory component of NHANES through blood collection of a subsample of respondents. As two composite measures of nutrition related to soft drink and juice consumption, we examine total caffeine intake and vitamin C intake from the dietary recall module (informa- tion for many other vitamins and nutrients, such as vitamin D, is not

11 To be consistent with NHANES III and NHANES 1999–2004, we use recall data from only the first day of the dietary interview for the 2005–2006 survey wave. 12 These categories are defined in the Appendix. Although we are unable to distinguish between diet and regular soft drinks between 1999 and 2006, we have information on grams of consumption as well as calories, which allows us some ability to distinguish diet and regular soda consumption since diet soda does not have calories. Under most state definitions of soft drinks with respect to taxation (discussed in an earlier footnote), soft drinks defined using the appendix categories would be subject to taxation while juice and whole milk would not. Juice drinks may be subject to taxation depending on percent juice content, but this is not measured in NHANES.

available in this module). However, we also are able to supplement the dietary recall information with data from the laboratory module, allowing blood serum measures of vitamin D and C. Although these single-nutrient analyses have several limitations,13 we focus on elements that reflect our focus on soft drink, juice, milk, and other beverage consumption patterns.

We merge NHANES III data with the NHANES 1999–2006 data. Most relevant survey questions are asked similarly across the survey years, with the exception of race and ethnicity. We measure race and ethnicity as black non-Hispanic, white non-Hispanic, and other race or ethnicity to construct categories which are consistent throughout the survey. We restrict the sample to children and adolescents between the ages of 3 and 18 with non-missing height and weight or soft drink consumption information.14

Table 2 displays the summary statistics. Fifteen percent of children are obese and 30% are overweight or obese. Fifty nine percent of children consume any soft drink during the day with an average of

1994 and 2001 through 2006 (CDC, 2010). Following the analytical recommendations in CDC (2010), we adjusted the values from 1988 through 1994 to be comparable to the measurements in 2001 through 2006. The second source of measurement error is drifts in the assay performance that occurred between 2001 and 2006 (CDC, 2010). The year dummy variables in our empirical specifications are likely to control for some, but not all, of the assay drifts during these years. 14 We focus on children in this paper in part because of the availability of previous studies showing little to no effects of soft drink taxation on adult weight. We note, though, that we find similar results to Fletcher et al. (2010a), using the NHANES data and find no statistically significant effects of soft drink on adult weight outcomes. Additionally, our findings suggest very small effects on soft drink consumption, which we plan to investigate further in future research.

Table 8 Alternative specifications that control for additional state characteristics.

BMI Z-score

Obese Overweight Soft drink calories

Juice calories

Juice drink calories

Whole milk calories

Total Calories

Controlling for lagged state average BMI 0.021 0.009 0.004 −5.000* 0.626 2.644 7.815*** −6.006 (0.016) (0.007) (0.011) (2.726) (1.370) (2.034) (2.411) (12.031)

N 21,872 21,872 21,872 20,804 20,804 20,804 20,804 20,804 Controlling for lagged state unemployment rate 0.015 0.009 0.002 −6.107** −0.069 1.708 7.643*** −8.175

(0.017) (0.006) (0.010) (2.644) (1.523) (2.119) (2.201) (12.328) N 22,132 22,132 22,132 21,040 21,040 21,040 21,040 21,040 Controlling for cigarette taxes 0.017 0.010 0.004 −4.193 0.113 1.413 7.280*** −9.166

(0.017) (0.007) (0.011) (2.898) (1.484) (2.238) (2.251) (12.465) N 22,132 22,132 22,132 21,040 21,040 21,040 21,040 21,040 Controlling for all of the above 0.024 0.010 0.006 −4.178 0.755 1.569 7.422*** −8.728

(0.018) (0.007) (0.011) (2.778) (1.345) (1.915) (2.677) (12.523) N 21,872 21,872 21,872 20,804 20,804 20,804 20,804 20,804

Notes: Heteroskedasticity-robust standard errors in parentheses that allow for clustering within states. Each cell represents a separate regression. Additional variables include female, age, age squared, black, other race, state, year, and quarter. All regressions utilize NHANES survey weights. *Significant at 10%; **significant at 5%; ***significant at 1%.

972 J.M. Fletcher et al. / Journal of Public Economics 94 (2010) 967–974

332 g or 12 oz and 122 cal per day. Although the average caloric intake from soda represents only 6% of the average total caloric intake, soft drinks represent a significant component of children's diets compared to other beverages. Children are approximately twice as likely to drink any soft drink during the day as juice, juice-related drinks, or whole milk and the calories consumed from soft drinks are also approximately twice the amount of calories consumed from these other drinks.

5. Results

Table 3 presents the baseline associations between soft drink tax rates and soft drink consumption. The soft drink tax coefficients represent the effects of a one percentage point increase in the tax rate on the probability of consuming a soft drink, the total grams of soft drinks consumed, and the total calories from soft drinks consumed. All regression models throughout the paper include year, quarter, and state fixed effects (unreported), and standard errors are clustered at the state level.15 Regressions are estimated using ordinary least squares and NHANES survey weights are used throughout the paper.16,17

As shown in Table 3, we find little influence of a tax on soft drinks on the probability that a youth consumes soda. However, a tax on soft drinks does influence the amount of soda that youths consume. A one percentage point increase in the soft drink tax rate reduces the amount of calories consumed by soda by nearly 6 cal, which is about 5% of the average calories from soda. This reduction in calories is likely not caused by a switch to diet soft drinks as there is an 18 gram decrease in soft drink consumption from a one percentage point increase in the soft drink tax rate, which is also about 5% of the average grams from soda.

Thus, our initial findings suggest that increasing the taxes on soft drinks will lead to reductions in soft drink consumption by children

15 We follow results from Cameron et al. (2008) that suggest the use of “Moulton- type” standard error estimators in the case of data with 30 or more clusters. An alternative is the “pairs-cluster bootstrap” estimator, though Cameron et al. show that it performs no better with more than 30 clusters. In addition, Rao and Wu (1988) show that bootstrap techniques produce inconsistent estimates when used with sampling weights. 16 The outcome variables that measure the grams and calories consumed of different beverages are naturally censored at 0. An alternative to linear regression in the presence of censored dependent variables is the Tobit model. Our results are qualitatively similar to estimates based on Tobit models. However, estimates from Tobit models are not consistent in the presence of heteroskedastic errors or fixed effects. Following Angrist (2001), we report the linear regression estimates as our preferred estimates. 17 For regressions with one of the consumption outcomes as the dependent variable, we use the survey weights for the dietary recall data. For regressions with one of the weight outcomes as the dependent variable, we use the survey weights for the mobile examination center data. To construct survey weights for our combined NHANES dataset, we follow the NCHS recommended analytic guidelines and divide the NCHS- supplied weights for each cycle by the number of cycles in our dataset. These weights incorporate the probability of being surveyed and adjustments for non-response.

and adolescents. The magnitude of the reduction is somewhat modest. As discussed before, there is evidence that reducing consumption by 100 cal per day could prevent weight gain in 90% of the population (Ludwig et al., 2001), but typical increases in the soda tax of 1 to 2 percentage points would not dramatically affect caloric intake. The point estimates suggest, though, that an increase in the tax rate of over 16 percentage points would be required to affect a reduction in soda of approximately 100 cal per day; however, an increase of this magnitude is outside the support of our data. Indeed, several recent proposals have called for this magnitude of tax rate increase—for example Governor Patterson of New York recently suggested an 18% “obesity tax” on soft drinks. However, while soda consumption may be reduced with a large tax increase, it is important to understand whether this represents a reduction in total calories or whether individuals may respond to the tax by increasing consumption of non- taxed items of similar calories.

We next examine potential substitution patterns between soft drinks and other high-calorie consumption items that are not typically included in soft drink tax definitions, such as juice, juice- related drinks, and whole milk. As shown in Table 4, there is some suggestive evidence that soft drink taxes affect the consumption of juice or juice-related drink. The results do show that whole milk is a substitute for soft drinks; a one percentage point increase in the soft drink tax rate increases caloric intake from whole milk by nearly 8 cal per day, which is 13% of the average calories from whole milk. Similarly, a one percentage point increase in the soft drink tax rate increases whole milk consumption by 11 g or 12% of the average grams of whole milk. The decrease in calories from soft drinks in response to an increase in the soft drink tax rate is completely offset by the increase in calories from whole milk. Thus, in Table 5, the results show that there is no statistically significant impact of the soft drink tax rate on total calories.

To provide additional evidence of the substitution patterns we find in Table 4, in Table 6 we examine the impacts of the soft drink tax rate on serum levels of vitamins C and D from the laboratory module as well as composite intakes of caffeine and vitamin D from the dietary recall module. Both our main results in Table 4 and those in columns 1 and 3 of Table 6 show effects consistent with positive but insignificant increases in juice and other items containing vitamin C. We also find in column 2 a positive but insignificant increase in vitamin D blood serum levels. The unit increase (measured in nanograms/milliliter) is consistent with a small increase in whole milk consumption.18 Finally,

18 Vitamin D supplements with a dose of 400 IU (international units) per day have been shown to increase blood levels of vitamin D by approximately 10 ng/dl (Vieth, 1999). One serving (244 g, or 146 cal) of whole milk includes 97.6 IU of vitamin D, so that an 8 calorie increase in whole milk consumption per day may be expected to increase vitamin D by approximately 0.13 ng/ml.

21 A few states have introduced restrictions on the sale of soft drinks in vending machines in schools in recent years. Consistent with the results of Fletcher et al. (2010b), which show that states' restrictions on vending machines in schools do not reduce soft drink consumption, we find that there is no interactive influence of soft drink taxes and vending machine restrictions, either full or partial bans, on soft drink consumption. However, an important caveat to this conclusion is that state-wide restrictions on vending machines began in 2003, which is near the end of our sample

Table 9 The impact of soft drink taxes on consumption by day of the week/school year.

Calories from soft drinks

Total grams of soft drink consumption

Juice calories

Juice drink calories

Whole milk calories

Panel A: Weekday Soft drink tax rate

−6.889** −21.568** −1.607 2.463 7.794*** (3.179) (8.678) (1.361) (3.097) (2.567)

Observations 14,308 14,308 14,308 14,308 14,308 R-squared 0.154 0.163 0.050 0.035 0.071

Panel B: Weekend Soft drink tax rate

−5.439 −17.019 3.798 0.690 4.389 (5.679) (13.808) (3.539) (2.232) (2.665)

Observations 6732 6732 6732 6732 6732 R-squared 0.186 0.191 0.032 0.054 0.079

Panel C: School year Soft drink tax rate

−6.514** −20.787*** 0.739 4.161 6.847*** (2.408) (5.924) (2.695) (4.030) (2.084)

Observations 12,164 12,164 12,164 12,164 12,164 R-squared 0.143 0.152 0.024 0.049 0.060

Panel D: On vacation from school Soft drink tax rate

−5.103 −9.634 −2.747 −0.725 11.046* (3.931) (11.852) (2.338) (5.666) (5.530)

Observations 2981 2981 2981 2981 2981 R-squared 0.159 0.163 0.040 0.070 0.070

Notes: Heteroskedasticity-robust standard errors in parentheses that allow for clustering within states. Additional variables include female, age, age squared, black, other race, state, year, and quarter. Regressions in panels C and D also include weekday as an additional covariate. All regressions utilize NHANES survey weights. *Significant at 10%; **significant at 5%; ***significant at 1%.

973J.M. Fletcher et al. / Journal of Public Economics 94 (2010) 967–974

consistent with our soft drink consumption findings from Table 4, our results show a 3 milligram decrease in caffeine consumption as a result of a 1% soft drink tax increase.19

Given the results that a change in the soft drink tax rate induces youths to substitute whole milk for soft drinks and that the magnitude of these effects is similar, it is not likely that an increase in soft drink taxes would decrease obesity. Indeed as shown in Table 7, the results confirm that soft drink taxes have little influence on BMI, overweight, or obesity among children and adolescents. Similar to Fletcher et al.'s (2010a) estimates for adults, the estimates are small in magnitude and not statistically significant.20

Overall, our estimates, which are identified using variation within state over time, demonstrate that soft drink taxes reduce soft drink consumption. However children and adolescents are found to respond to the taxes by shifting consumption to other high-calorie beverages, such as whole milk. Therefore, the net effect of soda taxes on caloric intake is minimal, and we find no effect on weight outcomes in children and adolescents. Soda taxes seem to be an ineffective “obesity tax” due to a standard behavioral response to the policy, where children and adolescents consume more calories of relatively cheaper beverages, which is milk in this case.

5.1. Robustness checks

In addition to our baseline results, we also perform several robustness checks in order to increase confidence that we are estimating the causal effects of soft drink taxes on outcomes. One concern about the validity of these estimates is whether changes in state soft drink taxes are endogenous. To address this possibility, we include time-varying state characteristics as additional covariates. As

19 A 12 ounce Coca-Cola contains 140 cal and 35 mg of caffeine. 20 As shown in Appendix Table 1, restricting the sample to states that have ever had a soft drink tax has little influence on the results shown in Tables 3, 4, and 7.

shown in Table 8, the main results are robust to the inclusion of the one-year lagged state mean of adult BMI, one-year lagged unemploy- ment rates, and cigarette tax rates. These additional results demon- strate that the main findings in this paper are unaffected by the possibility that states raise soft drink taxes in response to changes in population weight or changes in the macro economy.

Given that children spend a large portion of their time in school, an additional concern is that the changes in soft drink tax rates coincide with changes in school food policies.21 For example, Clark and Gleason (2006) find that participation in the National School Lunch Program is associated with a decrease in soft drink consumption and an increase in whole milk consumption. To address this possibility, we examine the impact of soft drink taxes on beverage consumption on weekdays and on weekends and we compare consumption during the school year with consumption during school vacations.22 As shown in Table 9, we find evidence of a decrease in calories consumed from soft drinks and an increase in calories consumed from whole milk both during school and away from school, although the estimates on weekdays and during school vacations are generally measured with less precision.

6. Conclusion

In this paper, we present the first evidence of whether soft drink taxes are linked with consumption decisions and weight outcomes of children and adolescents. We use a national sample that contains weight outcomes and consumption patterns of children and adoles- cents between 1989 and 2006. We then merge newly collected state- level soft drink tax data for this time period with the survey data in order to use a quasi-natural experimental design to estimate the short term effects of soft drink taxation. Our results suggest that soft drink taxation, as currently practiced in the United States, leads to a moderate decrease in the quantity of soft drinks consumed by children and adolescents. As a result, soft drink taxation may yield lower revenues for states than expected if behavioral responses to the tax are not accounted for. Additionally, soft drink taxes do not appear to have countered the rise in obesity prevalence because any reduction in soft drink consumption has been offset by the consumption of other calories. Cast in this light, the revenue generation and health benefits of soft drink taxes appear to be weaker than expected.

Despite this evidence against the effectiveness of soft drink taxes to reduce obesity, we believe that there are at least two directions for further inquiry in this area. First, although there is no evidence that soft drink taxes improve weight outcomes in children and adoles- cents, the fact that children and adolescents substitute more nutritious whole milk for soft drinks when taxed suggests that there may be broader health benefits that are not yet understood. Second, most historical tax rates are considerably lower than those that have been recently proposed, so that extrapolating our results to much larger increases in tax rates may not be appropriate. It is possible that there is a tax rate threshold at which consumers' reactions are greatly magnified, so it is unclear whether consumer substitution patterns would be sufficiently different with large tax changes to reduce total

period. 22 In each NHANES wave, individuals between the ages of 6 and 19 are asked whether they are in school or on vacation between grades. Ninety two percent of individuals report being on vacation between grades during May through August; defining school vacations as vacation from school during these months yields similar results.

974 J.M. Fletcher et al. / Journal of Public Economics 94 (2010) 967–974

caloric intake. Findings from these areas of inquiry could suggest that there are pathways by which it is possible that soft drink taxes could indirectly improve child and adolescent health. However, the evidence to date is that soft drink taxes are ineffective as an “obesity tax”.

Appendix A. Supplementary data

Supplementary data to this article can be found online at doi:10.1016/ j.jpubeco.2010.09.005.

References

Adams, Scott, Cotti, Chad, 2008. Drunk driving after the passage of smoking bans in bars. Journal of Public Economics 92 (5–6), 1288–1305.

Adda, Jerome, Cornaglia, Francesca, 2006. Taxes, cigarette consumption, and smoking intensity. American Economic Review 96 (4), 1013–1028.

Adda, Jerome, Cornaglia, Francesca, 2010. The effect of bans and taxes on passive smoking. American Economic Journal: Applied Economics 2 (1), 1–32.

Andreyeva, Tatiana, Long, Michael, Brownell, Kelly, 2010. The impact of food prices on consumption: a systematic review of research on price elasticity of demand for food. American Journal of Public Health 100 (2), 216–222.

Angrist, Joshua D., 2001. Estimation of limited dependent variable models with dummy endogenous regressors: simple strategies for empirical practice. Journal of Business & Economic Statistics 19 (1), 2–16.

Baum, C., 2009. The effects of cigarette costs on BMI and obesity. Health Economics 18 (1), 3–19.

Besley, T.J., Rosen, H.S., 1999. Sales Taxes and Prices: An Empirical Analysis. National Tax Journal 52 (2), 157–178.

Bhattacharya, Jay, Bundorf, Kate, Pace, Noemi, Sood, Neeraj, 2009. Does health insurance make you fat? NBER Working Paper 15163.

Block, G., 2004. Foods contributing to energy intake in the US: data from NHANES III and NHANES 1999–2000. Journal of Food Composition and Analysis 17, 439–447.

Brownell, K.D., Frieden, T.R., 2009. Ounces of prevention — the public policy case for taxes on sugared beverages. New England Journal of Medicine 360, 18.

Cameron, A.C., Gelbach, J.B., Miller, D.L., 2008. Bootstrap-based improvements for inference with clustered errors. Review of Economics and Statistics 90 (3), 414–427.

Caraher, M., Cowburn, G., 2005. Taxing Food: Implications for Public Health Nutrition. Public Health Nutrition 8 (8), 1242–1249.

Cawley, John, Meyerhoefer, Chad, Newhouse, David, 2007. The impact of state physical education requirements on youth physical activity and overweight. Health Economics 16 (12), 1287–1301.

Chaloupka, F., Warner, K., 2000. The economics of smoking. In: Cuyler, A., Newhouse, J. (Eds.), Handbook of Health Economics. Elsevier Science, North-Holland, Netherlands.

Centers for Disease Control and Prevention, 2006. NHANES III Laboratory Data File Documentation. U.S. Department of Health and Human Services, Washington, DC.

Centers for Disease Control and Prevention, 2009. Laboratory Procedure Manual, Analyte: 25-Hydroxyvitamin D, Matrix: Serum, Method: Diasorin (Formerly Incstar) 25-OH-D assay for NHANES 2001–2002. U.S. Department of Health and Human Services, Washington, DC.

Centers for Disease Control and Prevention, 2010. Analytical Note for NHANES 2000– 2006 and NHANES III (1988–1994): 25-Hydroxyvitamin D Analysis. U.S. Depart- ment of Health and Human Services, Washington, DC.

Chou, S.-Y., Grossman, M., Saffer, H., 2004. An economic analysis of adult obesity: results from the behavioral risk factor surveillance system. Journal of Health Economics 23, 565–587.

Chou, S., Grossman, M., Saffer, H., 2006. Reply to Jonathan Gruber and Michael Frakes. Journal of Health Economics 25, 389–393.

Chou, S., Rashad, I., Grossman, M., 2008. Fast-food restaurant advertising on television and its influence on childhood obesity. Journal of Law and Economics 51 (4), 599–618.

Chouinard, H., Davis, D., LaFrance, J., Perloff, J., 2007. Fat Taxes: Big Money for Small Change. Forum for Health Economics and Policy 10 (2) Article 2.

Chriqui, J.F., Eidson, S.S., Bates, H., Kowalczyk, S., Chaloupka, F.J., 2008. State sales tax rates for soft drinks and snacks sold through grocery stores and vending machines, 2007. Journal of Public Health Policy 29 (2), 226–249.

Clark, Melissa, Gleason, Phil, 2006. School meal participation, beverage consumption, and child obesity: evidence from the CSFII. Working Paper.

The Council of State Governments, 1990–2007. The Book of the States. Council of State Governments, Lexington, KY.

Courtemanche, Charles, 2009. Rising cigarette prices and rising obesity: coincidence or unintended consequence? Journal of Health Economics 28 (4), 781–798.

Cutler, D.M., Glaeser, E.L., Shapiro, J.M., 2003. Why Have Americans Become More Obese? Journal of Economic Perspectives 17 (3), 93–118.

Dharmasena, S., Capps, O., 2009. Demand interrelationships of at-home nonalcoholic beverage consumption in the United States. Presented at the Agricultural & Applied Economics Association's 2009 AAEA & ACCI Joint Annual Meeting, Milwaukee, Wisconsin, July 26–29, 2009.

Evans, William N., Farrelly, Matthew C., 1998. The compensating behavior of smokers: taxes, tar, and nicotine. RAND Journal of Economics 29 (3), 578–595.

Finkelstein, E.A., Trogdon, J.G., Cohen, J.W., Dietz, W., 2009. Annual medical spending attributable to obesity: payer- and service-specific estimates. Health Affairs 28 (5), w822–w831.

Fletcher, J.M., 2009. The effect of smoking cessation on weight gain: evidence using workplace smoking restrictions. Working Paper.

Fletcher, J.M., Frisvold, D., Tefft, N., 2010a. Can soft drink taxes reduce population weight? Contemporary Economic Policy 28 (1), 23–35.

Fletcher, J.M., Frisvold, D., Tefft, N., 2010b. Taxing soft drinks and restricting access to vending machines to curb child obesity. Health Affairs 29 (5), 1059–1066.

Gruber, J., Frakes, M., 2006. Does falling smoking lead to rising obesity? Journal of Health Economics 25 (2), 183–197.

Harnack, L., Stang, J., Story, M., 1999. Soft drink consumption among US children and adolescents: nutritional consequences. Journal of the American Dietetic Association 99, 436–441.

Hill, J.O., Wyatt, H.R., Reed, G.W., Peters, J.C., 2003. Obesity and the environment: where do we go from here? Science 299, 853–857.

Jacobson, M., Brownell, K., 2000. Small taxes on soft drinks and snack foods to promote health. American Journal of Public Health 90 (6), 854–857.

Kenkel, D.S., 2005. Are alcohol tax hikes fully passed through to prices: evidence from Alaska. American Economic Review 95 (2), 273–277.

Kotlikoff, L.J., Summers, L.H., 1987. Tax Incidence. In: Auerbach, A.J., Feldstein, M. (Eds.), Handbook of Public Economics. Elsevier Science, North-Holland, Netherlands, pp. 1043–1092.

Ludwig, D.S., Peterson, K.E., Gortmaker, S.L., 2001. Relation between consumption of sugar-sweetened drinks and childhood obesity: a prospective, observational analysis. The Lancet 357, 505–508.

Nader, P.R., O'Brien, M., Houts, R., Bradley, R.H., Belksy, J., Crosnoe, R., Friedman, S., Mei, Z., Susman, E.J., Development National Institute of Child Health and Human Development Study of Early Child Care and Youth, 2006. Identifying risk for obesity in early childhood. Pediatrics 118, e594–e601.

New York Times, 1920. City's Soda Cost Millions Monthly: Tax on Soft Drinks in Manhattan and Bronx for February and March is $465,445. New York Times, May 23, 1920.

Peltzman, Sam, 1975. The effects of automobile safety regulation. The Journal of Political Economy 83 (4), 677–726.

Powell, L.M., Chriqui, J., Chaloupka, F.J., 2009. Associations between state-level soda taxes and adolescent body mass index. Journal of Adolescent Health 45 (3), S57–S63.

Putnam, J.J., Allshouse, J.E., 1999. Food Consumption, Prices, and Expenditures, 1970– 97. Food and Consumers Economics Division, Economic Research Service, US Department of Agriculture, Washington, DC.

Rao, J.N.K., Wu, C.F.J., 1988. Resampling inference with complex survey data. Journal of the American Statistical Association 83 (401), 231–241.

Schroeter, C., Lusk, J., Tyner, W., 2008. Determining the impact of food price and income changes on body weight. Journal of Health Economics 27 (1), 45–68.

Sicher, John, 2007. Special issue: top-10 CSD results for 2006. Beverage Digest 50 (5). Sturm, R., 2002. The effects of obesity, smoking, and drinking on medical problems and

costs. Health Affairs 21, 245. Tobacco-NAFTA Industry Guide, 2009. Datamonitor. February 2009, available at: http://

www.researchandmarkets.com/reports/706797/tobacco_north_america_nafta_ industry_guide.

Vieth, Reinhold, 1999. Vitamin D supplementation, 25-hydroxyvitamin D concentra- tions, and safety. American Journal of Clinical Nutrition 69 (5), 842–856.

Zheng, Y., Kaiser, H.M., 2008. Advertising and U.S. nonalcoholic beverage demand. Agricultural and Resource Economics Review 37 (2), 147–159.