Applied Economics Publication details, including instructions for authors and subscription information:

This article was downloaded by: [Institutional Subscription Access] On: 21 September 2011, At: 12:02 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Applied Economics Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/raec20 Demand for carbonated soft drinks: implications for obesity policy Rigoberto A. Lopez a & Kristen L. Fantuzzi b a Department of Agricultural and Resource Economics, University of Connecticut, Storrs, CT 06269-4021, USA b Keystone Community Living, South Plainfield, NJ 07080, USA Available online: 14 Jun 2011 To cite this article: Rigoberto A. Lopez & Kristen L. Fantuzzi (2012): Demand for carbonated soft drinks: implications for obesity policy, Applied Economics, 44:22, 2859-2865 To link to this article: http://dx.doi.org/10.1080/00036846.2011.568397 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan, sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Applied Economics, 2012, 44, 2859 2865 Demand for carbonated soft drinks: implications for obesity policy Rigoberto A. Lopez a, * and Kristen L. Fantuzzi b a Department of Agricultural and Resource Economics, University of Connecticut, Storrs, CT 06269-4021, USA b Keystone Community Living, South Plainfield, NJ 07080, USA This article examines consumer choices of Carbonated Soft Drinks (CSDs) and their implications for obesity policy. Demand in relation to product and consumer heterogeneity is estimated via a random coefficients logit model (Berry et al., 1995) applied to quarterly scanner data for 26 brands in 20 US cities, involving 40 000 consumers. Counterfactual experiments show that caloric taxes could be effective in decreasing caloric CSD consumption though having little impact on obesity incidence. Keywords: demand; obesity; taxes; carbonated soft drinks; sodas JEL Classification: D12; L66; Q18; I18 I. Introduction Obesity, the leading public health crisis in the US, continues to increase at an alarming rate with an estimated social cost of over $140 billion a year and growing. The main culprits are the increase in the consumption of high-calorie foods and beverages and decrease in exercise (Kuhn, 2002). Paralleling the increase in obesity is the increase in consumption of Carbonated Soft Drinks (CSDs; DiMeglio and Mattes, 2000; Center for Science in the Public Interest, 2005). 1 Concerns about the adverse effects of excessive consumption of CSDs has generated a healthy volume of studies on the link between soda consumption, obesity and taxes, particularly from the medical and health economics literature. With 33 states having sales taxes on soft drinks (with a mean tax of about 5%), there are proposals in a few states (Maine, New York, Washington State) to significantly increase such taxes to curb consumption, and hence obesity (Brownell and Frieden, 2009). Although the empirical evidence on the effect of increased taxes on soda consumption is mixed, depending on the magnitude of the price elasticities of demand (Chouinard et al., 2007; Sturm et al., 2010), the preponderant empirical evidence point to sales taxes, having an insignificant effect on the obesity incidence (Ebbeling et al., 2006; Fletcher et al., 2010; Marlow and Shiers, 2010). This article examines US consumer CSD choices and their implications for obesity policy. Unlike previous studies, it incorporates product and consumer heterogeneity in consumer CSD choices using a random coefficients logit model (Berry et al., 1995; Berry Levinsohn Pakes, hereafter BLP) to estimate *Corresponding author. E-mail: Rigoberto.Lopez@uconn.edu 1 Carbonated soft drinks (CSDs) are a very large part of the American diet. These well-liked drinks account for approximately 30% of all beverages (alcoholic and nonalcoholic) consumed in the US (US Department of Agriculture, Economic Research Service, 2008). In 2000, Americans spent $61 billion on CSDs (National Soft Drink Association, 2003) with more than 15 billion gallons sold in the US, which equals every American consuming at least one 12-ounce can per day, or an average of 53 gallons per year (Squires, 2001). Applied Economics ISSN 0003 6846 print/issn 1466 4283 online ß 2012 Taylor & Francis 2859 http://www.informaworld.com DOI: 10.1080/00036846.2011.568397

2860 R. A. Lopez and K. L. Fantuzzi demand at the brand and consumer levels with particular reference to price and calorie content responses. The estimated parameters are then used to assess the effectiveness of a tax levied on caloric CSDs through counterfactual experiments. Empirical results point out that higher income consumers have significantly lower price elasticity of demand, therefore their CSD choices are less sensitive to price (and hence, taxes). Furthermore, lower income consumers have a preference for caloric CSDs. Combined, a tax on caloric CSDs will have a more significant impact on lowincome consumers. Empirical results also indicate that younger and male consumers have a stronger preference for caloric CSDs, further indicating that these groups among the most affected. Although CSD choices exhibit a strong degree of brand loyalty (relatively low cross-price elasticities), a 10% ad valorem tax on caloric drinks would have a significant effect on consumption (between 2 and 10% drop for caloric drinks and 3 5% increase for diet drinks, depending on the brand). However, the direct effect on the Body Mass Index (BMI) 2 is negligible. II. The Model Following the BLP model, the consumer, in choosing one unit of a CSD brand among competing products, maximizes utility, driven by the brand characteristics as well as his/her own characteristics. The indirect utility of consumer i from buying a unit of brand j is U ij ¼ x j þ p j fflfflfflfflffl{zfflfflfflfflffl} j þ D i x j þ D i p j þ i x j þ v i p j þ " ij fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} ij ð1þ which can be expressed in two components: (1) the mean utility term j, determined solely by brand characteristics, and (2) the deviation from the mean, ij, capturing the interactions between consumer and brand characteristics. In Equation 1, x j is a vector of observed product characteristics of brand j; p j the price of brand j; D i represents observed consumer characteristics (such as demographics) with a probability density function h(d); v i the unobserved consumer characteristics with a probability density function g(v), which is assumed to be normally distributed N(0, 1);,,, and are fixed parameters; and " ij an error term, which has a probability density function f("). Note that in this framework individual taste parameters with respect to price and brand characteristics are given by i ¼ þ D i þ v i and i ¼ þ D i þ v i, respectively. In order to define the market and market shares, an outside good is introduced. Let k ¼ 0 denote the outside good, which gives the consumer the option not to buy any of the J brands included in the choice set as well as excluded CSD brands and substitute beverages. The utility of the outside good is normalized to be constant over time and to equal zero. In this model, consumers choose one unit of a brand in the choice set that is assumed to yield the highest utility or the outside good. Aggregating over consumers, the market share for the jth brand equals the probability that the jth brand is chosen, given by Z s j ð p, x, Þ ¼ I ðd i, v i, " ij Þ : U ij U ik 8k ¼ 0,..., N dhðdþdgðvþdfð"þ ð2þ where H(D), G(v) and F(") are cumulative density functions for the indicated variables, assumed to be independent of each other, and ¼ð,,,,, Þ a vector of parameters. The price elasticities of the market shares for individual brands are ijk ¼ð@s j =@p k Þð p k =s j Þ 8 Z >< p j =s j i s ij ð1 s ij Þ dhðdþ dgðvþ, ¼ Z >: p k =s j i s ij s ik dhðdþ dgðvþ, for j ¼ k 9 >= otherwise >; ð3þ where each consumer has a different price elasticity for each individual brand. Given the interest in policy, the taste parameters for calories and price from the BLP model are used in a counterfactual experiment of the impact of ad valorem taxes levied on caloric CSDs on consumption and BMI. III. Data and Estimation To estimate the BLP model, we used sales data on 26 brands of CSDs in 20 cities across the US, over 20 quarters (1988 to 1992), resulting in 10 400 total 2 The National Institute of Health adopted BMI as the common public health measure (US Department of Health and Human Services, 2001). The BMI is the most convenient measure available in assessing overweightness and obesity, taking into account a person s weight and height to gauge total body fat. A BMI under 18.5 is considered underweight, a BMI of 18.5 to 24.9 is considered a healthy weight, a BMI of 25 to 29.9 is considered overweight, and a BMI of 30 and higher is considered obese.

Carbonated soft drinks and obesity 2861 product brand observations. 3 The sales data, from the Information Resources, Inc. (IRI) Infoscan database provided by the Food Marketing Policy Center at the University of Connecticut, consist of dollar sales, volume sold (in 288 ounces/case units), and the percent volume sold with any display promotion. The retail prices (p j ) were computed by dividing the dollar sales of each brand by the number of 12-ounce servings sold. The market size was assumed to be the per capita per day consumption of all CSDs, water, and fruit juices. Other definitions of market size (e.g. including fluid milk) neither significantly altered nor improved the results. Market shares were computed by dividing the number of servings sold by the market size. The outside good market share was defined as the residual between one and the sum of the observed market shares for the J brands in the choice set. The nutritional brand characteristics (x j ) were collected by examining the labels on each CSD brand for caffeine, calorie and sodium content per 12-ounce serving. The observable consumer characteristics (D) consisted of age, income and a male dummy. They were obtained by 100 random draws per market from the Behavioral Risk Surveillance System (BRFSS). 4 For estimation purposes, a market was defined as each city and quarter combination, resulting in 400 markets (20 cities 20 quarters). Another 100 random draws per market were obtained from a normal distribution with zero mean and unitary variance. Thus, this sample includes 40 000 consumer observations. All interactions of price and nutritional characteristics with consumer characteristics were considered in the model. Instrumental variables were used to control for potential endogeneity of prices due to their correlation with brand characteristics. Different sets of instrumental variables (176 in total) were interacted with error terms in the last part of the BLP estimation procedure. The first set of instruments involved 130 interactions between brand dummy variables and input prices, as in Villas-Boas (2007). Input prices included electricity prices (US Department of Energy, Energy Information Administration, 2007), wages for the different cities (US Department of Labor, 2007), the cost of sugar (US Department of Agriculture; USDA, Economic Research Service, 2007), the cost of materials (National Bureau of Economic Research, 2007) and the federal funds effective rate (Federal Reserve Board, 2007). The next set of instrumental variables used the housing price index (National City Corporation, 2007) for each city interacted with brand dummy variables. The last set of instruments consisted of 20 city dummy variables (one for each city in the sample) to capture the differences with regard to pricing of the CSD brands among the 20 cities. Finally, this article follows Berry (1994), relying on formation of a generalized method of moments estimator for estimating a proxy of the integral in Equation 2. The BLP model was estimated using Matlab and the tax simulations were done using Excel. The econometric and policy simulation results are presented in the following section. IV. Empirical Results Table 1 presents the estimated taste parameters of the demand for CSDs. The parameter estimate of the mean utility for price is statistically significant at the 5% level, indicating that for the average consumer, the CSD price creates disutility, as one would expect. The other mean utility estimates for promotion, calories, caffeine and sodium are not statistically discernable from zero. However, the distribution of the parameters for these product characteristics and price are significantly associated with consumer heterogeneity, resulting in a distribution of parameters rather than just a single point estimate. As shown in the results of Table 1, the parameters for the taste of price indicate that although the mean utility is negative, as expected, older and wealthier male consumers are less sensitive to price. The implication for tax policies is that these groups would be affected less by a general price increase in CSDs. On the other hand, calories are particularly negative valued by older wealthier and older female consumers. Thus, a caloric tax on CSDs would affect these groups less than other groups based on their calorie preference alone. Perhaps not surprisingly, the results for caffeine and sodium preferences follow a similar pattern to those for calories, which may reflect some health concerns. Since a large number of cross-price elasticities were computed, Table 2 presents only a sample of cross-price elasticities, averaged over the 20 cities in 3 Atlanta, Baltimore, Chicago, Cincinnati, Hartford, Houston, Indianapolis, Kansas City, Los Angeles, Louisville, Miami, Milwaukee, Minneapolis, Nashville, New York, Omaha, Phoenix, Raleigh, Salt Lake City, Seattle. 4 The BRFSS is a yearly telephone health survey consisting of more than 350 000 observations per year. The survey is conducted by each of the 50 state health departments with support from the Centers for Disease Control and Prevention. The BRFSS survey has been successfully used in economic analysis of obesity (Chou et al., 2004; Burke and Heiland, 2007).

2862 R. A. Lopez and K. L. Fantuzzi Table 1. Estimates of the BLP discrete choice model Deviations Variable Mean utility Age Income Male Unobservables Constant 2.078 1.807 1.483 1.472 2.203** (5.361) (16.473) (5.089) (8.388) (0.265) Price 8.086** 2.718 3.787** 1.326 1.975** (1.080) (4.090) (1.394) (2.333) (0.106) Promotion 0.345 (0.411) Calories 0.003 1.494** 1.445** 0.885** 0.751** (0.588) (0.180) (0.122) (0.162) (0.044) Caffeine 0.159 1.956** 1.192** 1.679** 1.373** (0.708) (0.884) (0.283) (0.967) (0.135) Sodium 0.886 1.740 2.206** 1.584 0.907** (3.102) (1.361) (0.337) (0.987) (0.072) Notes: SEs are given in parentheses. ** Denotes test statistic significance at the 5% level. the sample. The cross-price elasticities are all positive, as expected, implying that the brands are substitutes. Note that the cross-price elasticities are quite low when compared to the own-price elasticities. This confirms that although consumers are sensitive to CSD prices with respect to their chosen brands, they have brand loyalty and will substitute to the outside good rather than choosing another brand of CSD. Given the estimated price elasticities and the assumption of a full tax transmission rate, the effect of a 10% ad valorem calorie tax on CSDs was estimated at the product brand level. 5 These results are presented in Table 3. The magnitude of the own-price elasticities in Table 2 is in the range of previously estimated elasticities using brand-level demand with scanner data on CSDs. For example, Chan (2006) reports larger own-price elasticities that range between 5 and 11 for CSDS using household-level data on CSD purchases. On the other hand, Dhar et al. (2005) report lower own-price elasticities for broader CSD product definitions, ranging between 2.7 and 4.4. The own-price elasticities reported in Table 2 and brand-level elasticities from previous studies are much larger than the elasticities reported for aggregate CSD or soft drink categories in previous studies. For example, Andreyeba et al. (2010) report the price elasticity from 14 previous studies for the aggregate soft drink category to have a mean of 0.79 and to range between 0.8 and 1.0 (95% confidence interval). Two factors might explain much of the differences in the magnitude: (1) the much lower of aggregation used in this study, namely at the brand (and city) level of CSDs rather than an aggregate soft drink category which includes CSDs and other products (e.g. fruit juices); (2) Table 2 reports partial own-price elasticities, i.e. elasticity of only changing the price of a brand leaving the other prices constant. In terms of the scope of this study, we are interested in changing the price of all caloric CSDs not only changing the price of one brand assuming all other CSD prices constant. Since there is no analytical expression to aggregate our elasticity results to a national soft drink category, we compute the net elasticity assuming all CSD prices in the sample increase by 1% and also the ones driven by a tax policy only targeting caloric CSDs. Given the foregoing, it would be appropriate to compare instead the net elasticity of response; that is, the elasticity if all brands experience the same price increase. To partially address this, the results in Table 3 show the quantity response for a 10% price increase only for all caloric CSDs due to a 10% ad valorem tax (resulting in a 10% increase in price under full transmission, the price of diet CSDs remaining constant). These results suggest an average of 0.58 net price elasticity of demand for caloric CSDs (17 brands) and 0.42 for diet CSDs (which, of course, will show an increase due to substitution as they would not be subject to the tax). These net elasticities with respect to an increase in the price of caloric CSDs are lower than the aggregate elasticities reported by Andreyeba et al. (2010) for all soft drinks indicating a lower consumption response to taxes than their price elasticities would suggest. 5 Simulated calorie tax reductions are based on per capita annual CSD consumption in ounces and calories (US Department of Agriculture, Economic Research Service, 2007) using 3500 calories per pound of body weight (American Diabetic Association, 2003).

Carbonated soft drinks and obesity 2863 Table 2. Sample of own- and cross-price elasticities Brands 7 UP A&W Canada Dry Cherry 7UP Cherry Coke Coke Coke Classic Diet 7UP Diet Cherry 7UP Diet Coke Diet Dr Pepper Diet Pepsi 7UP 10.119 1.124 0.127 0.108 0.082 0.204 1.586 0.057 0.019 0.061 0.005 0.061 A&W 0.392 6.377 0.129 0.117 0.081 0.202 1.487 0.017 0.013 0.005 0.000 0.003 Canada Dry 0.183 0.056 3.747 0.574 0.066 0.181 0.156 0.101 0.030 0.476 0.014 0.409 Cherry 7 UP 0.408 1.157 0.129 7.405 0.081 0.221 1.670 0.063 0.021 0.052 0.004 0.060 Cherry Coke 0.466 1.198 0.191 0.827 9.101 0.911 0.665 0.008 0.002 0.137 0.006 0.044 Coke 0.349 0.715 1.870 0.706 0.276 10.119 0.902 0.005 0.001 0.289 0.054 0.047 Coke Classic 0.347 0.773 1.977 0.708 0.276 1.309 10.058 0.007 0.003 0.368 0.057 0.058 Diet 7 UP 0.106 0.011 0.047 0.021 0.003 0.004 0.067 3.886 0.403 0.448 0.082 0.526 Diet Cherry 7 UP 0.119 0.013 0.056 0.023 0.003 0.005 0.085 1.751 3.182 0.396 0.083 0.463 Diet Coke 0.020 0.000 0.030 0.003 0.007 0.054 0.051 0.014 0.022 4.850 0.259 1.339 Diet DR Pepper 0.043 0.001 0.022 0.006 0.007 0.022 0.016 0.500 0.097 0.569 4.098 1.335 Diet Pepsi 0.050 0.001 0.057 0.007 0.005 0.025 0.023 0.033 0.059 3.053 0.151 3.109

2864 R. A. Lopez and K. L. Fantuzzi Table 3. CSD consumption and BMI changes from a 10% ad valorem calorie tax Brand Percentage change in quantity Change in ounces consumed (per capita yearly) Such a tax would have the effect of reducing the quantity of caloric CSDs consumed and increase the consumption of noncaloric (diet) CSDs. However, the substitution is imperfect and consumers would also switch to the outside good (water, fruit juices and residual CSDs). Assuming brand loyalty and that consumers reduce their consumption of caloric CSDs by switching to noncaloric soft drinks, the reductions in caloric CSDs would translate to less than a pound of weight loss for the average consumer, or approximately one tenth of their BMIs. This is an upper bound as consumers could also switch to caloric alternatives that are not taxed (e.g. fruit juices), especially among those who are price sensitive such as low-income consumers. The results for the impact of a 10% ad valorem tax on changes in BMIs or obesity rates indicate that such a tax would have a very weak effect on reversing the obesity epidemic, as Kuchler et al. (2005) also found. In addition, such a tax, like other nutrition regulation policies, is likely to face stiff political opposition from a well-organized industry lobby (Nestle, 2002). Finally, a 10% tax is greater than most state sales tax rates, which may be interpreted as excessive, given the ongoing food price inflation. Change in calories consumed (per capita yearly) In addition, given the choices of low-income consumers, this tax would be regressive with respect to income, since low-income consumers spend a larger portion of their income on food. Therefore, any policy that increases the price of food will have the greatest impact on lower income consumers. In sum, this article adds to the mounting empirical evidence that point that taxes on caloric CSDs are not an effective policy tool to curb obesity, although they are effective in generating revenue. If the policy objective is the former, then alternative programmes (e.g. educational campaigns advertising policy), perhaps funded by a soda tax, are probably more effective than using taxes as a stand-alone policy instrument. As demonstrated by their preferences, these policies can be more effective when targeting younger, low-income consumers. V. Concluding Remarks Change in body weight (pounds) (per capita yearly) Change in average consumer BMI 7UP 9.936 568.51 7106.32 2.030 0.322 A&W 5.611 321.03 4761.88 1.361 0.216 Canada Dry 2.683 153.49 1790.70 0.512 0.081 Cherry 7 UP 1.450 82.96 1036.95 0.296 0.047 Cherry Coke 9.405 538.12 6995.53 1.999 0.317 Coke 9.775 559.27 6804.42 1.944 0.308 Coke Classic 8.063 461.36 5613.22 1.604 0.254 DR Pepper 4.517 258.44 3230.55 0.923 0.146 Minute Maid 7.663 438.44 6028.49 1.722 0.273 Mountain Dew 1.740 99.56 1410.39 0.403 0.064 Pepsi 4.819 275.73 3492.58 0.998 0.158 Pepsi Free 2.942 168.32 2103.97 0.601 0.095 RC 4.567 261.33 3484.44 0.996 0.158 Schweppes 7.782 445.26 4452.55 1.272 0.202 Slice 5.418 309.99 5114.77 1.461 0.232 Sprite 6.668 381.51 4514.49 1.290 0.204 Sunkist 5.677 324.81 5142.84 1.469 0.233 Diet 7 UP 4.176 238.91 0 0 0 Diet Cherry 7 UP 5.084 290.87 0 0 0 Diet Coke 4.694 268.57 0 0 0 Diet DR Pepper 4.811 275.27 0 0 0 Diet Pepsi 3.260 186.53 0 0 0 Diet Pepsi Free 3.699 211.65 0 0 0 Diet Rite 3.727 213.26 0 0 0 Diet Slice 4.185 239.44 0 0 0 Diet Sprite 4.277 244.73 0 0 0 This article examined consumer choices of CSDs using a random coefficients logit demand model (BLP, 1995) and scanner data, and then examined the

Carbonated soft drinks and obesity 2865 impact of a tax policy to address the obesity epidemic. Empirical demand results indicate that consumer choices of CSDs are driven by both product and consumer characteristics. More specifically, lower income and younger consumers, as well as male consumers, tend to have a more positive valuation of calories, suggesting that they are less concerned about obesity consequences. Furthermore, higher income and older consumers as well as male consumers are less sensitive to price changes, suggesting they care less about higher prices. By the same token, younger, low-income consumers who are male have relative preference for caloric CSDs. A counterfactual experiment shows that an ad valorem tax on caloric CSDs would be effective in decreasing consumption of CSDs but would have a hardly discernable effect on the incidence of obesity. Thus, a more comprehensive programme is needed rather than a stand-alone tax policy on caloric drinks. References American Diebetic Association (2003) A simple 100 calories a day can be the difference in weight maintenance versus gain or loss. Available at http://www.eatright. org/public/nutritioninformation/index_17622.cfm (accessed 23 August 2006). Andreyeba, T., Long, M. and Brownell, K. (2010) The impact of food prices on consumption: a systematic review of research on the price elasticity of demand for food, American Journal of Public Health, 100, 216 22. Berry, S. (1994) Estimating discrete choice models of product differentiation, RAND Journal of Economics, 25, 242 62. Berry, S., Levinsohn, J. and Pakes, A. (BLP) (1995) Automobile prices in market equilibrium, Econometrica, 63, 841 89. Brownell, K. and Frieden, T. (2009) Ounces of prevention: the public policy case for taxes on sugared beverages, New England Journal of Medicine, Perspective, 30, 1 4. Burke, M. and Heiland, F. (2007) Social dynamics of obesity, Economic Inquiry, 45, 571 91. Center for Science in the Public Interest (2005) Soft drinks by the numbers. Available at http://www.saveharry. com/bythe numbers (accessed 2 June 2007). Chan, T. (2006) Estimating a continuous hedonic-choice model with an application to demand for soft drinks, RAND Journal of Economics, 27, 1 17. Chou, S., Grossman, M. and Saffer, H. (2004) An economic analysis of adult obesity: results from the behavioral risk factor surveillance system, Journal of Health Economics, 23, 565 87. Chouinard, H., Davis, D., LaFrance, J. and Perloff, J. (2007) Fat taxes: big money for small change, Forum for Health Economics and Policy, 10, 1 28. Dhar, T., Chavas, J. P., Cotterill, R. and Gould, B. (2005) An econometric analysis of brand-level strategic pricing between Coca Cola Company and Pepsico, Journal of Economics and Management Strategy, 14, 905 31. DiMeglio, D. P. and Mattes, R. (2000) Liquid versus solid carbohydrates: effects on food intake and body weight, International Journal of Obesity, 24, 794 800. Ebbeling, C., Feldman, H., Osganian, S., Chomitz, V., Ellenbogen, S. and Ludwig, D. (2006) Effects of decreasing sugar-sweetened beverage consumption on body weight in adolescents: a randomized, controlled pilot study, Pediatrics, 117, 673 80. Federal Reserve Board (2007) Selected interest rates. Available at http://www.federalreserve.gov/releases/ h15/data.htm (accessed 15 August 2007). Fletcher, J., Frisvold, D. and Tefft, N. (2010) Can soft drink taxes reduce population weight?, Contemporary Economic Policy, 28, 23 35. Kuchler, F., Tegene, A. and Harris, J. M. (2005) Taxing snack foods: manipulating diet quality or financing information programs?, Review of Agricultural Economics, 27, 4 20. Kuhn, B. (2002) Weighing in on obesity, Food Review, 25, 38 49. Marlow, M. and Shiers, A. (2010) Would soda taxes really yield health benefits?, Regulation, 33, 34 8. National Bureau of Economic Research (2007) Manufacturing industry database. Available at http:// www.nber.org/nberces/nbprod96.htm (accessed 5 September 2007). National City Corporation (2007) Housing valuation analysis. Available at http://nationalcity.com (accessed 15 October 2007). National Soft Drink Association (2003) About soft drinks. Available at http://www.nsda.org (accessed 16 October 2007). Nestle, M. (2002) Food Politics, University of California Press, Berkeley, CA. Squires, S. (2001) Soft drinks, hard facts, Washington Post, 27 February 2001, Health Section, p. 15. Sturm, R., Powell, L., Chriqui, J. and Chaloupka, F. (2010) Soda taxes, soft drink consumption, and children s body mass index, Health Affairs, 29, 1 7. US Department of Agriculture, Economic Research Service (2007) Sugars and sweeteners: data tables. Available at http://www.ers/usda.gov/sugar/data/htm (accessed 15 August 2007). US Department of Agriculture, Economic Research Service (2008) Beverages: per capita availability. Available at http://www.ers.usda.gov/data/foodconsumption/ spreadsheets/beverage.xls (accessed 13 August 2008). US Department of Energy, Energy Information Administration (2007) Energy prices and trends. Available at http:// www.eia.doe.gov/emeu/aer/contents.html (accessed 15 August 2008). US Department of Health and Human Services (2001) The Surgeon General s Call to Action to Prevent and Decrease Overweight and Obesity, US Government Printing Office, Washington, DC. Available at http:// www.surgeongeneral.gov/topics/obesity (accessed 12 August 2008). US Department of Labor (2007) Bureau of labor statistics: wages, earnings, and benefits. Available at http:// www.bls.gov (accessed 5 September 2007). Villas-Boas, B. (2007) Vertical contracts between manufacturers and retailers: inference with limited data, Review of Economic Studies, 74, 625 52.