Cawley et al (2021) PDF: The Pass-Through of a Tax on Sugar-Sweetened Beverages in Boulder, Colorado

T HE P ASS -T HROUGH OF A T AX ON S UGAR S WEETENED B EVERAGES IN B OULDER , C OLORADO J OHN C AWLEY , D AVID F RISVOLD , D AVID J ONES , AND C HELSEA L ENSING This paper estimates the pass-through of the largest tax on sugar-sweetened beverages enacted in the U.S., which is two cents per ounce in Boulder, Colorado. A strength of the paper is that, to achieve as complete a perspective as possible, we estimate the pass-through of the tax not only to beverage prices in retail stores but also to those in restaurants, and we examine data for the treatment community of Boulder and a comparison community of Fort Collins, Colorado, from four sources: (a) handcollected data on prices from stores; (b) Nielsen Retail Scanner Data of store prices; (c) hand-collected data on prices in restaurants; and (d) web-scraped data from online restaurant menus. Across the multiple datasets, we ﬁnd consistent evidence that the tax was largely, but not completely, passed through to consumers. In both the hand-collected store data and restaurant data, pass-through is slightly less than 75%, whereas pass-through is just over 50% using the scanner data; consumers bear most, but not all, of the largest tax on sugar-sweetened beverages in the United States. Key words: pass-through, soda, sugar-sweetened beverages, tax. JEL codes: H22, H75, I18. John Cawley is a professor at the Cornell University and NBER, Department of Policy Analysis and Management, and Department of Economics, 2312 Martha Van Rensselaer Hall, Ithaca, NY 14853. David Frisvold is an associate professor at the University of Iowa and NBER, Department of Economics, 21 E. Market St., Iowa City, IA 52240. David Jones is an associate director at the Mathematica Policy Research, P.O. Box 2393, Princeton, NJ 08543. Chelsea Lensing is an assistant professor at the Coe College, Department of Economics, 1220 1st Ave NE, Cedar Rapids, IA, 52402. We thank Amelia Hulbert, Kathryn Panega, Colter Dewitt, Nikki Dee, Meagan Stephenson, Carolina Ramirez, Jessica Hardison, Kelsey White, Marley Vasquez, Kristin Milardo, Raina Benford, Taylor Daniel, and Rachel Arndt for their assistance with the data collection. We would like to acknowledge Rachel Arndt from the Boulder County Public Health and Sara Cooper and Emily Burns from the Colorado School of Public Health for collaboration on data collection of receipts and for helpful comments and discussions. We acknowledge funding from the Colorado Health Foundation. We thank Elizabeth Botkins, Rachel Grifﬁth, Benjamin Lockwood, Mark Stehr, and participants at the American Society of Health Economists conference and Society for Beneﬁt–Cost Analysis conference for helpful comments. Researchers own analyses calculated (or derived) based in part on data from The Nielsen Company (US), LLC and marketing databased provided through the Nielsen Datasets at the Kilts Center for Marketing Data Center at The University of Chicago Booth School of Business. The conclusions drawn from the Nielsen data are those of the researchers and do not reﬂect the views of Nielsen. Nielsen is not responsible for, had no role in, and was not involved in analyzing and preparing the results reported herein. Correspondence to be sent to: [email protected] The incidence of taxes is a classic topic in public ﬁnance. Economic theory indicates that the relative burdens of a tax are determined by the market power of ﬁrms and the elasticities of supply and demand (Kotlikoff and Summers 1987; Fullerton and Metcalf 2002; Weyl and Fabinger 2013). For example, in a perfectly competitive market, if demand is completely inelastic or if ﬁrms face constant marginal costs, pass-through would be 100% and consumers would bear the entire burden of the tax. In contrast, when demand is elastic, consumers share the tax burden with ﬁrms, with the shares depending on the relative elasticity of supply and demand (which is inﬂuenced by factors such as the socioeconomic status of consumers). If the market is imperfectly competitive, taxes can be over shifted (price may rise by more than the tax) if oligopolists ﬁnd it optimal to reduce output and charge higher prices in response (Anderson, de Palma, and Kreider 2001; Bonnet and Réquillart 2013). Numerous studies have estimated the pass-through of taxes on Amer. J. Agr. Econ. 103(3): 987–1005; doi:10.1111/ajae.12191 Published online January 22, 2021 © 2021 Agricultural and Applied Economics Association 988 May 2021 products such as cigarettes and gasoline, typically ﬁnding that 100% or more of the taxes are passed through to consumers (e.g., Poterba 1996; Besley and Rosen 1999), but a smaller body of literature ﬁnds partial pass-through, in the range of 45 to 85% (e.g., Doyle and Samphantharak 2008; Harding, Leibtag, and Lovenheim 2012). We estimate the pass-through of the largest tax on sugar-sweetened beverages (SSBs) in the US. Numerous organizations, such as the World Health Organization, Institute of Medicine, American Academy of Pediatrics, and the American Public Health Association, have called for taxes on SSBs because SSBs contribute to obesity and poor health (Rudd Center for Food Policy and Obesity 2014). In addition to being high calorie and zero nutrient, SSBs have a high glycemic load (i.e., they raise blood sugar), which, independently of obesity, contributes to insulin resistance and diabetes (Malik and Hu 2011). SSB taxes are hypothesized to reduce consumption by increasing prices of the taxed beverages; thus, the degree of passthrough and subsequent price increases are critical. Many countries recently implemented taxes on SSBs, including Chile, Finland, France, Ireland, Mexico, and the United Kingdom (Thow et al. 2018). Within the U.S., several cities have adopted taxes on SSBs: ﬁrst Berkeley, CA, in 2015; followed by Philadelphia, Boulder, and Oakland in 2017; and San Francisco and Seattle in 2018.1 All of these city-level taxes are imposed on beverage distributors who sell to retailers. Given the relative newness of the taxes, their effects are not well understood.2 Studies of these excise taxes in the U.S. estimate the impact on prices by examining hand-collected price data observed in retail stores or scanner data of sales in retail stores.3 As shown in ﬁgure 1, three cities have estimates of pass1 Many states also impose sales taxes on soft drinks, although they are very small, are primarily a tool to increase revenue, and apply to diet as well as caloric soft drinks (Fletcher, Frisvold, and Tefft 2010, 2015) 2 There is also a literature examining the impact of SSB taxes outside of the U.S. Several studies ﬁnd that more than 100% of the SSB tax in Mexico was passed through to consumers, although the studies lack geographic control groups and rely on pre-post comparisons and comparisons to untaxed non-substitute products (Colchero et al. 2015; Grogger 2017). 3 Although this paper focuses on the pass-through of an SSB tax, other research has examined the impact on purchases and consumption. For example, Cawley et al. (2019) estimate that Philadelphia’s tax reduced purchases of taxed beverages by 8.5 ounces per shopping trip and reduced adults’ soda consumption by 31% but had no impact on the consumption of children. Cawley et al. (2020b) ﬁnd no change in SSB purchases or SSB consumption in Oakland following that city’s SSB tax. There may also be heterogeneity in responses to the tax; using scanner data from Berkeley, Debnam (2017) concludes that high-SSB-consuming households are less price elastic than low-SSB-consuming households. Amer. J. Agr. Econ. through from SSB taxes using both types of data: Berkeley, Philadelphia, and Seattle.4 In Berkeley, Falbe et al. (2015) and Cawley and Frisvold (2017) estimate that 43–47% of the SSB tax was passed on to consumers, based on hand-collected data from retail prices in Berkeley and San Francisco. Using Nielsen scanner data to examine the Berkeley tax, Rojas and Wang (2021) estimate pass-through at 16% and not statistically signiﬁcant for all SSBs, whereas Bollinger and Sexton (2018) estimate pass-through at 23% for one grocery store chain and no change in prices for one pharmacy chain. In Philadelphia, Cawley et al. (2020a) estimate 103% pass-through using hand-collected data from a wide range of stores throughout the city. Cawley, Willage, and Frisvold (2018) estimate 55% pass-through using handcollected data from stores in the airport terminals in Philadelphia. Bleich et al. (2020) estimate pass-through at 121% based on handcollected data from small, independent stores. Using scanner data, Roberto et al. (2019) estimate pass-through at 90%, whereas Seiler, Tuchman, and Yao (2020) estimate passthrough at 97%. In Seattle, Jones-Smith et al. (2020) estimate pass-through at 90% based on hand-collected data, whereas Powell and Leider (2020) ﬁnd pass-through of 59% using scanner data. Thus, based on comparing estimates of the impact of the same tax across types of data, studies using scanner data tend to estimate lower pass-through rates than studies using hand-collected data. This difference could potentially be due to scanner data excluding prices from smaller store types. For example, Cawley et al. (2020a), based on their analysis of hand-collected data in Philadelphia, ﬁnd that pass-through is highest for convenience stores with gas stations and for independent stores (i.e., not part of a chain). We contribute to the literature on the passthrough of taxes on SSBs by providing the most comprehensive assessment of SSB tax pass-through to date. We use multiple data sources, including hand-collected data and scanner data, enabling us to provide a more complete understanding of how the tax affects pass-through than has been possible in previous studies that used just a single type of data. 4 In Oakland, using hand-collected data, Cawley et al. (2020b) ﬁnd pass-through of 61% and Falbe et al. (2020) estimate passthrough at 92%. In San Francisco, Falbe et al. (2020) estimates pass-through at 100% using hand-collected data. In Cook County, IL, Powell, Leider, and Léger (2020) ﬁnd pass-through of 119% using scanner data. Cawley, Frisvold, Jones and Lensing Sugar-Sweetened Beverage Tax 989 Figure 1. Estimated pass-through of SSB taxes, by city, study, and data type Note: This ﬁgure shows the pass-through estimates with the 95% conﬁdence interval for papers in the literature on SSB taxes, including the estimates from this paper. The pass-through estimates are calculated as the point estimate of the change in price, in cents per ounce, divided by the amount of the local tax. The ﬁgure shows the primary estimate for all stores or restaurants in the sample for all SSBs, if available; otherwise, the row headings describe the store types or beverage types corresponding to the estimate. Having results from these different data sources is of practical value to policymakers. The hand-collected data cover all retail stores, including smaller convenience stores, whereas the scanner data consist of larger chain stores. Given the different customer base at different store types, policymakers may prefer one estimate more than another depending on the policy objectives. By providing a comprehensive assessment of SSB tax pass-through with both types of data, we are able to inform policymakers about the variation in the impact of the tax on prices by data source and store type. We estimate pass-through in retail stores and restaurants, the latter a novel addition to the literature. The study uses data at multiple points in time before and after the tax, which permits us to estimate how the estimated impact of the tax varies over time. In addition, this paper is the ﬁrst to estimate the passthrough of the largest city-level tax on SSBs in the US, which is the tax of 2 cents per ounce in Boulder, CO, that was implemented on July 1, 2017.5 Boulder’s tax on SSBs is substantial; it represents 22% of the pretax price of a 20-ounce bottle, 68% of the pretax price of a 2-liter bottle, and 53% of the pretax price of a 12-pack of 12-ounce cans.6 An important contribution of this paper is that it is the ﬁrst to estimate the pass-through of SSB taxes to prices in restaurants.7 5 The tax in Boulder passed by ballot initiative in November 2016, with 54% of voters in favor of the tax. It is an excise tax on distributors and took effect on July 1, 2017. The tax applies to SSBs with at least 5 grams of caloric sweetener per 12 ﬂuid ounces. It does not apply to diet soda, products in which milk is the primary ingredient, alcoholic mixers, or coffee drinks. The tax is applied to the size of the prepared product; for example, the tax on the syrup used to prepare a 32-ounce fountain drink is 64 cents. 6 These percentages were calculated using the mean price of SSBs in Boulder in April 2017, according to our hand-collected store data. 7 A recent study of the pass-through of SSB taxes on restaurant prices was published that cites the working paper version of the present study as the ﬁrst paper to estimate the impact on restaurant prices (Marinello et al. 2020). Marinello et al. (2020) estimate pass-through of 82% for bottled soda and 29%, but not statistically signiﬁcant, for fountain drinks. 990 May 2021 Restaurants are an important location for beverage purchase and consumption in the U.S. During 2011–12, U.S. adults on average consumed 50.1 calories of SSBs per day, or 23.5% of their total SSB consumption, at restaurants (An and Maurer 2016). The passthrough of a SSB tax may be different in restaurants than stores for a variety of reasons. The elasticity of demand of restaurant patrons may differ from that for store shoppers; Okrent and Alston (2012) found that the demand for food away from home, particularly at fullservice restaurants, was much more sensitive to price increases than the demand for food at home. In addition, differences in elasticity of supply between restaurants and retailers could affect the extent of pass-through. Bottled drinks are identical at all locations (restaurants and retailers), but restaurant meals are highly differentiated and have a lot more heterogeneity, so consumers may see them as imperfect substitutes for each other. In addition, there may be differences in price salience; people may not scrutinize the price of an SSB as much in a restaurant—a drink might be part of their standard order. Bundling may also be a factor; some restaurants (especially fast food) may offer meal deals or children’s meals that include a drink with other items for a single price. In addition, some restaurants provide free reﬁlls of SSBs. The net effect of all of these factors on passthrough is ambiguous, and determining whether the pass-through of SSB taxes to prices in restaurants differs from retailers is ultimately an empirical question. Another important strength of the paper is its rich and varied data, which provide a more comprehensive view of the impact of SSB taxes than ever before. The literature has shown that ﬁndings from hand-collected and scanner data can vary at times, likely due to the store and product types included in the data. Estimating pass-through in Boulder using multiple types of data provides a more complete picture of price responses across the range of stores and products. We analyze four types of data, three of which were original data collection. First, we collected price data in person from hundreds of stores of all types, sizes, and and both chain-afﬁliated and independent stores in Boulder and a control community in multiple periods before and after the tax. After the tax, we not only recorded posted (shelf) prices, but we also purchased taxed and untaxed beverages in order to determine whether the tax was added at the register Amer. J. Agr. Econ. as opposed to included in the posted (shelf) price. The tax was levied on beverage distributors, in part because excise taxes are more salient and, thus, more likely to reduce consumption (Chetty, Looney, and Kroft 2009). However, we ﬁnd that not all retailers included the tax in the posted, or shelf, prices; some instead added it at the register, where it is less salient, which also means that previous studies that only examine posted prices could underestimate pass-through. Although the advantages of these data are the coverage for all retail stores and the data on posted and register prices, limitations are that we collected prices on the most commonly purchased items instead of every product, and have data from every two months instead of more frequently. The second data source is the Nielsen Retail Scanner Data. These data have the strength of covering the full set of products and continuously measuring transacted prices, allowing us to both test for parallel trends in prices before the tax in Boulder and a control community that includes Fort Collins, CO, and to test for variation in pass-through over time after the tax was implemented. The data are limited in that they contain information for mostly chain stores. Both the hand-collected store data and the Nielsen data have the advantages of containing information about the prices of a wide range of beverages: various sizes (e.g., 20 ounce and 2-liter bottles), various containers (bottles, cans, and fountain drinks), and a wide range of brands and products, both taxed and untaxed. The third dataset consists of hand-collected prices from hundreds of restaurants in the same communities before and after the tax. Similar to the hand-collected retail store data, advantages are that we collect data from all limited-service restaurants and coffee shops, but the limitations are that we collected prices on the most commonly purchased items instead of every product, and have data from every two months instead of more frequently. The fourth dataset are web-scraped, weekly data from online restaurant menus in these communities. These data contribute frequent measures of posted beverage prices before and after the tax. Thus, by using multiple data sources for retail stores and restaurants with different strengths and weaknesses, we are able to provide a more comprehensive analysis of the impact of SSBs taxes than the prior literature. For a wide variety of reasons, the passthrough of the 2.0 cent per ounce tax in Cawley, Frisvold, Jones and Lensing Boulder may be different from that of the smaller taxes of 1 cent per ounce in Berkeley and Oakland and 1.5 cents per ounce in Philadelphia. In perfectly competitive markets, the pass-through of a tax is determined by the relative elasticities of supply and demand. The elasticity of demand tends to vary along the demand curve, and thus, there may be a different elasticity of demand (and thus tax passthrough) for a larger tax. A larger tax may be more salient or noticeable to consumers (Taubinsky and Rees-Jones 2018) and may result in more dramatic responses by consumers (who may either reduce their total intake or decide to evade the tax through cross-border shopping), which would lead to lower pass-through in a perfectly competitive market. On the other hand, Kroft et al. (2020) ﬁnd that under imperfect competition, greater salience increases the incidence of taxes on consumers for products in grocery stores. Thus, a larger, more salient, tax could have higher pass-through. In either case, because salience has shown to increase with the magnitude of a tax, and incidence varies by salience, the empirical ﬁndings reported in this paper of the largest SSB tax to date are an important contribution to the study of tax policy. Moreover, Allcott, Lockwood, and Taubinsky (2019) estimate that an optimal nationwide tax (considering both the aim to change consumers’ behavior and potential regressivity) is between 1 and 2.1 cents per ounce, with the upper bound falling just above the Boulder tax, so the pass-through observed for the largest existing tax may be the best evidence regarding the pass-through of a socially optimal tax.8 Finally, a large tax could also lead to more dramatic responses by suppliers, who may change their advertising or stocking strategies. We estimate the pass-through of the SSB tax to consumers using a difference-in-differences design, comparing the change in prices per ounce in Boulder to that in a comparison city of Fort Collins, CO. The results indicate that the Boulder tax increased retail prices immediately after its implementation on July 1, 2017, and that this increase remained relatively constant for at least the next four months. Based on the hand-collected retail data, posted prices of taxed beverages increased by 0.84 cents per ounce on 8 However, Allcott, Lockwood, and Taubinsky (2019) also note that optimal city-level taxes could be substantially lower if there is cross-border shopping. Sugar-Sweetened Beverage Tax 991 average, a 42% pass-through rate. However, 20% (primarily, convenience stores) of the stores in Boulder do not include the tax in their posted prices but instead add it at the register. As a result, pass-through is larger when measured by the register prices: 1.4 cents per ounce, or 71% of the tax. We also estimate lower pass-through from the Nielsen scanner data at 1.064 cents per ounce, or 53% of the tax. Based on the handcollected restaurant data, prices of fountain drinks rose by 0.87 cents per ounce, or 43.5% of the tax, shortly after the implementation of the tax and continued to increase for the next two months to a pass-through rate of 74.2%. The estimates from the webscraped data on restaurant prices are consistent with these estimates. Methods and Data Overview To estimate the pass-through of the SSB tax to retail prices, we use a difference-in-differences design and compare the change in prices (in cents per ounce) over time in Boulder to the change in prices in the control community of Fort Collins, CO. An important assumption underlying this design is that, in the absence of the tax, the trends in prices in Boulder would be the same as the trends in Fort Collins. The geographic proximity of these areas, similarities in demographic characteristics and locations of large, public universities in Boulder and Fort Collins are consistent with this assumption.9 In addition, Fort Collins is forty-ﬁve miles to the north, which makes cross-border shopping from Boulder unlikely.10 To investigate the plausibility of our identifying assumption of parallel trends in prices in the treatment and control areas, we assess the trends in prices in these areas over time prior 9 As shown in the online supplementary Appendix in Table A1, the demographic characteristics of these two cities are similar. The city of Boulder is home to the University of Colorado at Boulder and is fully enclosed within Boulder County. Fort Collins, in Larimer County, Colorado, is home to Colorado State University. 10 We also collected data from a second control community, Boulder County. Boulder County is an appealing control group because it has the advantage of proximity; any unobserved shocks to demand in Boulder around the time of the tax are likely experienced by the rest of the county. However, there may be spillover effects of the tax due to cross-border shopping by Boulder residents seeking to avoid the tax. Therefore, we use Fort Collins as the primary control group throughout this paper. 992 May 2021 to the tax. Evidence from four data sources, which we describe later in the Results section, is consistent with the parallel trends assumption. We estimate the impact of the tax for taxed and untaxed products separately. We estimate the impact of the SSB tax on untaxed products because the tax could cause substitution from taxed to untaxed products (e.g., from Coke to Diet Coke), which would alter the price of the untaxed products. For our primary estimates, we pool all products and sizes. However, because the price elasticity, and thus the pass-through, may vary by product size, we also estimate pass-through separately for the most common product sizes in retail stores. Pass-through could vary by size if demand is more inelastic for individual servings (e.g., 20-ounce bottles) than for larger volumes (e.g., 2-liter bottles) that may be part of planned, larger shopping trips. We also examine heterogeneity by retail store type. Pass-through could vary by store type if the elasticities of demand and supply differ across store type because of differences in the stores’ marginal costs or because of differences in their clientele. To get as complete a perspective as possible across products, store types, and data collected methods, we assembled four datasets on beverage prices: (a) hand-collected data of posted prices and purchase prices of beverages from all retail stores, (b) scanner data of prices from retail stores in the Nielsen Retail Scanner Data, (c) hand-collected data of listed prices of fountain drinks and coffee drinks from all limited-service restaurants, and (d) webscraped data of prices from a selected sample of restaurant menus. Appendix Figures A1 and A2, in the online Supplementary Appendix, show the location of each retailer store and restaurant where we gathered prices in Boulder and Fort Collins, respectively.11 In the following subsections, we provide additional information on each of these datasets and explain the exact empirical speciﬁcation for each of them. Hand-Collected Data of Beverage Prices from Stores We collected beverage prices from stores at four points in time, twice before the tax (April and June 2017) and twice after the tax 11 Appendix Figure A3, in the online supplementary Appendix, shows the location of each retail store and restaurant where we gathered prices in Boulder County. Amer. J. Agr. Econ. (August and October 2017). The four time points enable us to examine the change in prices over time before the tax and to compare the pass-through of the tax at two points in time after implementation. We collected data from all grocery stores, pharmacies, and convenience stores in Boulder and Fort Collins. We identiﬁed these stores and their addresses using the ReferenceUSA database, which includes approximately twenty four million U.S. businesses and is updated monthly.12 Data collectors visited and recorded prices from 98 retailers (38 in Boulder) in April, 176 retailers (73 in Boulder) in June, 173 retailers (71 in Boulder) in August, and 174 retailers (71 in Boulder) in October.13 After the data collection in April, we expanded the set of retailers to include liquor stores. We collected the prices of soft drinks, energy drinks, sports drinks, iced tea, juice, water, mixers for alcoholic drinks, and fountain drinks. We chose the most common sizes and brands to maintain consistency among the products and reduce the burden on data collectors in the ﬁeld. We selected a mix of products that are taxed and untaxed. For example, we selected 20 oz. bottles, 2 liter bottles, and 12 packs of 12 oz. cans of Pepsi (taxed), Diet Pepsi (untaxed), Coke (taxed), and Diet Coke (untaxed). We also selected products that are consumed more commonly in Boulder, such as Hansen’s soda (taxed), San Pellegrino (untaxed), and GT’s Organic Raw Kombucha (untaxed).14 For all products, we collected the posted price and whether the product was on sale. If a store did not post prices, data collectors asked an employee for the price of the products. We collected this information for all products in each of the four periods, except that we began collecting the prices of Hansen’s, San Pellegrino, and alcohol mixers in June (the second of the two pre-tax periods). The full list of products is shown in the online supplementary Appendix, Table A2. We collected a total of 7219 prices 12 Speciﬁcally, we included all retailers with veriﬁed listings in Boulder and Fort Collins, CO that are classiﬁed as supermarkets or other grocery stores (NAICS code 445110); convenience stores (NAICS code 445120); pharmacies and drug stores (NAICS code 446110); gasoline stations with convenience stores (NAICS code 447110); warehouse clubs and supercenters (NAICS code 452311); and beer, wine, and liquor stores (NAICS code 445310). 13 More details on data collection are presented in the online supplementary Appendix, Table A2. 14 Fermented beverages with less than 11 grams of caloric sweetener per 12 ﬂuid ounces were exempt from the tax. The GT’s kombucha products that were collected meet this criterion. Cawley, Frisvold, Jones and Lensing for 2505 products in Boulder and Fort Collins. The balanced sample of products with prices for all four months consists of 3636 productspeciﬁc, store-speciﬁc prices for 909 products. Failing to consider the register price could lead to an underestimate of the overall passthrough of the tax to consumers if retailers add the tax at the register instead of including it in the posted price.15 To test this possibility, we construct the register price, which is equal to the posted price plus the amount of the tax that is itemized on the receipt, before sales tax is included. Speciﬁcally, in October (after the tax), in addition to collecting posted prices, data collectors purchased 20 oz. bottles of Pepsi and Diet Pepsi from each retailer and kept the receipt. If the store did not sell these products, the data collectors purchased another taxed SSB and a comparable untaxed product. Based on the receipts, we determine whether the posted price matches the price that retailers charge consumers (excluding sales tax).16 For most retailers, the posted price is equal to the register price. However, sixteen out of seventy nine Boulder retailers (20.2%) did not include the tax in the posted price, and instead, added the tax at the register and itemized the amount of the tax on the receipt. If a retailer adds the tax at the register for the SSB we purchased, we assume that the retailer does the same for all SSBs in both periods after the tax was implemented.17 In the online supplementary Appendix, Table A4 displays the characteristics of retail stores that itemize the tax on the receipt at the register and stores that incorporate the tax into the posted price on the shelf. Most (81%) of the retail stores that itemize the tax on the receipt at the register are convenience 15 The SSB tax is levied on distributors, who are the wholesalers who sell the SSBs to retailers. Any and all distributors who sell SSBs within city limits are required by law to remit the tax to the city government. Large retailers like supermarkets buy from distributors and thus the SSB tax is typically incorporated into their shelf prices. However, some small, independent stores may not buy from a distributor but may instead drive to a superstore and buy cases of drinks that they bring back and sell individually. If that superstore is outside city limits then the tax is not imposed on that sale. In that case, the small store owner is a “self-distributor” and is obligated to pay the beverage tax to the city themselves. This is one potential reason that, combined with the desire to reduce the salience of the tax, could lead a store owner to add the tax at the register. 16 One retailer includes sales tax in the posted price. As a result, the receipt price, before the sales tax is included, is less than the posted price in all periods for this retailer. 17 Consistent with this assumption, results based on the set of products that were purchased are similar to the results from all products, as shown in the online supplementary Appendix, Table A3. Sugar-Sweetened Beverage Tax 993 stores; in contrast, only 13% of the retailers that incorporate the tax into the posted price in Boulder are convenience stores. As a result, the annual sales and number of employees in retailers that pass on the tax at the register are lower compared to other retailers in Boulder. Using the hand-collected store data, we estimate the following regression: Yisct = β0 + β1 ðBoulderc × Aprilt Þ + β2 ðBoulderc × August t Þ + β3 ðBoulderc × Octobert Þ + δt + λm + μw + θs + φi + εisct ð1Þ where Yisct denotes the price per ounce of product i in store s in community c in month t; Boulder is a binary variable equal to one if store s is located in the City of Boulder (and 0 if the store is located in Fort Collins); and April, August, and October are binary variables equal to one if the price is recorded in that month; June is the omitted reference month. To reiterate, the data consist of two pre-tax periods (April and June) and two periods after the tax was introduced (August and October). δt represents month ﬁxed effects. λm represents day-of-the-month ﬁxed effects and μw represents day-of-the-week ﬁxed effects. θs represents store ﬁxed effects. φi represents product ﬁxed effects.18 ε is a stochastic error term. In the equation listed above, β2 and β3 are the coefﬁcients of interest; they represent the difference-in-differences estimates of the impact of the Boulder tax on prices in the two post-tax periods of August and October respectively, relative to the pre-tax period of June. Comparing β3 to β2 indicates whether the estimate of pass-through changed over time after the tax. The data include two geographic clusters: Boulder and Fort Collins. As a result, it is not possible to cluster standard errors by geography; with only two clusters, the standard errors would be degenerate (Donald and Lang 2007). As a result, we share with the most previous studies of city-level beverage taxes the inability to cluster standard errors by city. Instead, 18 We deﬁne a product based on the size and the name. Examples of products are a 20 oz. bottle of Pepsi, a 2 liter bottle of 7Up, a 12 pack of 12 oz. cans of Diet Coke, a 8.4 oz. can of Red Bull, and a small fountain drink. 994 May 2021 we follow the previous literature (e.g., Cawley and Frisvold 2017) and cluster standard errors at the store level, with the acknowledgement that the resulting standard errors may be underestimated (Cameron and Miller 2015).19 Nielsen Retail Scanner Data The second data source we use is the Nielsen Retail Scanner Data. The sample includes weekly prices of all beverages sold in stores in the City of Boulder and Larimer County, which contains Fort Collins, for all weeks in 2017.20 Prices are measured as the volume-weighted average prices of the product for that week and are reported each Saturday. We construct the price per ounce for each product for each week based on the weekly price and the volume of each product. There are 912 taxed soft drinks and energy drinks sold in thirty six stores in the City of Boulder and Larimer County in the sample. Because each product is not sold in every store, there are 292,330 weekly prices on SSBs in these stores. We also separately examine the impact on prices of bottled water (562 products in 36 stores) and diet soft drinks (666 products in 36 stores), which are untaxed. The types of stores in the sample in the geographic areas include grocery stores (11), pharmacies (15), convenience stores with gas stations (3), and mass merchandise stores (7).21 Similar to equation (1), using the Nielsen Retail Scanner Data, we estimate the following speciﬁcation: Yiscw = α0 + α1 ðBoulderc × Postw Þ + ϕw + θs + φi + ηisct ð2Þ where Yiscw denotes the average price per ounce of product i in store s in community 19 To put our limited number of clusters into context, several previous studies of the pass-through of taxes on SSBs (e.g., Grogger 2017) had data only for the treated country or state with no geographic control. When we include the area of Boulder County outside of the city of Boulder as an additional control group, clustering standard errors at the community level, using the wild cluster bootstrap method as recommended by Cameron, Gelbach, and Miller (2008), yields similar, but slightly smaller standard errors on the coefﬁcients of interest. As a result, we report the more conservative standard errors, clustered at the store level. 20 Store locations are identiﬁed based on the county and 3-digit zip code. Given this information, we can identify which stores are located in the city of Boulder. However, we cannot identify which stores are located in Fort Collins because all areas in Larimer County share the same 3-digit zip code. 21 Because convenience stores are not located in both the treatment and control areas, we are not able to estimate the impact on prices within convenience stores. We are able to examine heterogeneity in other store types. Amer. J. Agr. Econ. c in week w. Boulder is a binary variable equal to one if store s is located in the City of Boulder (and 0 if the store is located in Fort Collins). Post is a binary variable equal to one for all weeks ending after July 1.22 ϕw represents week ﬁxed effects. θs represents store ﬁxed effects. φi represents product (deﬁned as a unique UPC code) ﬁxed effects. η is a stochastic error term. Similar to equation (1), we cluster standard errors at the store level. Compared to equation (1), equation (2) includes weekly instead of monthly ﬁxed effects. Because prices in the Nielsen data are recorded on Saturday for each week, we exclude day-of-the-month and day-of-theweek ﬁxed effects. The coefﬁcient of interest is α1, which represents the change in prices after the implementation of the tax on July 1 in Boulder relative to the change in prices after July 1 in Fort Collins. In addition to estimating the average impact across all weeks after the tax was implemented, we modify equation (2) to include interaction terms of Boulder and each week. Hand-Collected Data from Restaurants and Coffee Shops For the third source of data, we collected the price and number of ounces of all sizes of fountain drinks from restaurants, which are taxed if the drink is caloric (not diet). We also collected the prices of a 12 oz. drip coffee, a 12 oz. latte, a 12 oz. mocha latte, and a 12 oz. hot chocolate from coffee shops, which are all untaxed. Although a mocha latte and a hot chocolate are sweetened beverages, the city council exempted milk-based products from the tax. We collected data from all limited-service restaurants and coffee shop locations in city of Boulder and Fort Collins.23 Data collectors visited each of these restaurants to determine whether the restaurant sold fountain drinks or coffee drinks and to record the prices and sizes. We collected this information from restaurants in April, June, August, and October 22 The prices are recorded on Saturday for each week. July 1, 2017 was a Saturday. 23 Speciﬁcally, using the ReferenceUSA database, we included all restaurants with veriﬁed listings in Boulder and Fort Collins, CO that are classiﬁed as limited-service restaurants (NAICS code 722513) and snack and non-alcoholic beverage bars (NAICS code 722515), which includes all coffee shops listed under SIC code 581228. Limited-service restaurants are restaurants in which customers order at the counter. Cawley, Frisvold, Jones and Lensing 2017, and from coffee shops in June, August, and October 2017. Data collectors visited 152 restaurants in April, 211 restaurants and coffee shops in June, 208 restaurants and coffee shops in August, and 205 restaurants and coffee shops in October.24 We collected 1400 prices for 423 fountain drinks and 1055 prices for 432 coffee drinks in Boulder and Fort Collins. The balanced sample includes 295 fountain drinks with prices for four periods for a total of 1180 prices of fountain drinks and 288 coffee drinks with prices for three periods for a total of 864 prices of coffee drinks. With these data, we estimate the impact of the SSB tax on restaurant prices using equation (1). We estimate equation (1) separately for fountain drinks (taxed) and coffee drinks (untaxed). OrderUp Data of Restaurant Beverages For the fourth source of data, we collected beverage prices from the online menus of restaurants that participate in the OrderUp.com delivery platform in the city of Boulder and the Fort Collins area. OrderUp is an online restaurant food ordering and delivery company that was founded in 2009 and serves customers in over sixty locations (including Boulder and Fort Collins) across twenty two states. We were able to collect these data more frequently because we collected these data by web scraping as opposed to in-person recording. We scraped the OrderUp data weekly, beginning every Wednesday, from March 22, 2017, through October 25, 2017. The frequency of the data provides us with greater detail on the timing and consistency of price changes after the introduction of the tax and of the trends in prices prior to the tax. The data collection began with 219 restaurants, of which 158 appeared in all waves of data collection. Reasons for a restaurant not remaining in the sample include termination of use of the OrderUp system, closures, name or address changes (these are the two identifying variables for a restaurant), and technical errors occurring when the website is updated and the scrape incorrectly reads or saves a menu. Of the 158 restaurants consistently in the sample every week, 114 consistently 24 The number of restaurants selling each product in each time period are shown in the online supplementary Appendix, Table A5. Sugar-Sweetened Beverage Tax 995 have beverage items throughout the entire period.25 Of the 114 restaurants, forty two are located within the city of Boulder and seventy two are located in the Fort Collins area.26 The types of beverages on the OrderUp menus are more varied than the handcollected restaurant data. The OrderUp beverage items in the ﬁnal sample range from speciﬁc branded items (e.g., Coke, Oogave Rootbeer) to general types of drinks (e.g., apple juice, tea). The full list of items is shown in the online supplementary Appendix, Table A6. We categorize each beverage item into one of three categories based on the Boulder SSB tax law: taxed, untaxed, or unknown. Most OrderUp beverage items have names that we can categorize as taxed or not under the Boulder SSB law, but some items have generic names that we cannot deﬁnitively categorize (e.g., “Coke products” may include both taxed regular Coke and untaxed Diet Coke). Of the 877 beverage items in the balanced sample, we can identify 688 products as either taxed or untaxed. Some beverage items contain information on ﬂuid ounces, but the majority only contain the name of the item. The number of ounces of the product is known for only 67 of the 877 items. As such, for the OrderUp items, we report price per drink instead of price per ounce. We assume that the number of ounces did not change over time for the drinks for which size is not listed. Although this is untestable for all items, there was no change in size after the tax for any of the sixty seven drink items of known size, which supports the plausibility of this assumption. When we estimate the difference-in-differences equation using the weekly online menu data from OrderUp, we replace the month ﬁxed effects with weekly ﬁxed effects. To be speciﬁc, we estimate: Yircw = χ0 + πðBoulderc × Montht Þ + ϕw + θr + φi + ξircw ð3Þ where Yircw denotes the price per drink of product i in restaurant r in community c in 25 We identify products by item name, and size when applicable, thus menu updates that change either variable exclude the item from the balanced sample. 26 For this sample, the Fort Collins area includes Fort Collins, Evans, Garden City, Greeley, Loveland, and Windsor. 996 May 2021 week w. Month refers to a set of binary variables, spanning from March through October, equal to one if the price is recorded in that month; June is the omitted reference month. ξ is a stochastic error term. All other variables are deﬁned as above. The coefﬁcients of interest are the vector π, which represents the difference in the price per drink of beverages in restaurants in each month, relative to the difference in the month before the tax was implemented (June). Results Amer. J. Agr. Econ. bottle, 2-liter bottle, 12 pack of 12 ounce cans, and fountain drinks) and speciﬁc brands (Pepsi products, Coke products, and other brands) sold in stores show similar patterns. The trends in the price per ounce of fountain drinks in restaurants and the price per drink from OrderUp are also stable in Boulder and parallel to the trends for taxed products outside of Boulder prior to the introduction of the tax (online supplementary Appendix, Figures A4 and A5). Figure 3 displays the results from an eventstudy design using the Nielsen Retail Scanner Data, which conﬁrms that the trends in prices in Boulder and Larimer County leading up to the implementation of the tax are similar. Evidence Regarding Parallel Trends To investigate the plausibility of the assumption that Fort Collins is a valid counterfactual for the outcomes experienced in Boulder, we examine whether the trends in the outcome (prices per ounce) between Boulder (treatment) and Fort Collins (comparison) were parallel prior to the implementation of the tax on July 1. We present the trends for taxed and untaxed drinks, for the hand-collected store data (ﬁgure 2), Nielsen scanner data (ﬁgure 3), hand-collected restaurant data (online supplementary Appendix, Figure A4), and web-scraped restaurant data (online supplementary Appendix, Figure A5). Previous research on SSB taxes using handcollected data typically only collect one period of prices prior to the tax (e.g., Cawley et al. 2020a, 2020b). Our hand-collected store and restaurant data include two periods, spanning three months, of pre-tax prices. Although a beneﬁt of this data collection is that the two periods allow us to examine the change in prices over time prior to the introduction of the tax, we acknowledge that there could be changes in the trends that occur outside of these two periods. A strength of our other data sources are their multiple periods of data collection prior to the introduction of the tax. The web-scraped restaurant data include ﬁfteen weeks of pre-tax prices. The Nielsen scanner data include six months of weekly pre-tax prices. Based on our hand-collected data, the trends in prices of all taxed products in Boulder are stable prior to the introduction of the tax in July and are comparable to the trends in prices of taxed products outside of Boulder over this same period (ﬁgure 2). Graphs of the trends in prices for speciﬁc sizes (20-ounce Difference-in-Differences Estimates for Retail Stores Table 1 presents the difference-in-differences estimates for taxed and untaxed items, based on the hand-collected data, separately for the entire sample (i.e., unbalanced panel) and the balanced panel of products and stores. Results for taxed items are shown for both posted prices and register prices. Column 1 presents results based on posted prices for the entire sample. The posted prices of SSBs increased from June (one month prior to the tax) to August (one month after the tax) by 0.839 cents per ounce in Boulder, relative to Fort Collins. The tax is 2 cents per ounce, so the price increase represents a pass-through of 41.95%. In October (six months after the tax), prices were 0.795 cents per ounce higher than in June. Thus, prices rose from one month before to one month after the tax and then remained ﬂat for the next two months. Importantly, the coefﬁcient on the interaction term for Boulder * April suggests that there was not a differential trend in prices between Boulder and Fort Collins prior to the tax.27 Column 2 of table 1 shows the difference-indifferences estimates using the register prices for all stores. Prices in Boulder increased by 1.422 cents per ounce from June to August, for an estimated pass-through rate of 71.1%. Again, the estimate for October is very similar to that for August, implying that pass-through 27 The difference-in-differences estimates including the second control community, Boulder County, are presented in online supplementary Appendix in Table A7. Estimates are similar, at 46.9% pass-through of the tax by August. Sugar-Sweetened Beverage Tax 997 8 Price Per Ounce in Cents 9 10 11 12 Cawley, Frisvold, Jones and Lensing Apr May Jun Boulder: Taxable (Posted) Boulder: Non-Taxable Fort Collins: Non-Taxable Jul Aug Sep Oct Boulder: Taxable (Register) Fort Collins: Taxable Figure 2. Trends in the price per ounce of SSBs and other beverages at retailers Note: Price per ounce is reported in cents. Taxed and not taxed items are deﬁned according to whether the item is taxed under the law in Boulder. Posted prices are the prices shown on the shelf for each item. Register prices are constructed to account for stores that do not include the SSB tax in the posted price, and are equal to the posted price plus the amount of the tax that is itemized on the receipt. The data are balanced at the store-item level across all four waves of the data collection. remained roughly constant in the months after the tax.28 The third column of table 1 reports results for untaxed beverages. The effect of the Boulder tax on the price of untaxed items is small in magnitude and not statistically signiﬁcant. There is some evidence of a differential trend in the prices of untaxed products from April to June, but this suggests that prices for untaxed items were decreasing in Boulder relative to Fort Collins, which is in the opposite direction of the positive pass-through observed for taxed beverages and would 28 As shown in the online supplementary Appendix in Table A4, approximately 20% of stores itemize the tax at the register; more than half of these (13 out of 16) are convenience stores. In contrast, only 8 out of 63 stores that only incorporate the tax into the shelf price are convenience stores. Stores that itemize the tax at the register also increased their prices on the shelf. The mean shelf price of taxed beverages in these stores increased by 0.438 cents per ounce (with a standard error of 0.101) from June to August, whereas the mean price for untaxed items increased by only 0.147 cents per ounce (with a standard error of 0.078). Because these stores also itemized the tax at the register, the mean price paid at the register of taxed beverages increased by 2.439 cents per ounce. In contrast, in stores that only incorporated the tax into the shelf price (and did not itemize the tax), mean prices increased by 0.973 cents per ounce (with a standard error of 0.148) for taxed beverages and 0.397 cents per ounce (with a standard error of 0.144) for untaxed beverages. indicate that the small and not statistically signiﬁcant result is an overestimate. In the last three columns of table 1, we restrict the sample to the balanced panel of products that are consistently in the sample during all four periods. The sample is balanced over products and stores. This does not meaningfully affect the results; thus, changes in products or stores do not drive the estimates for the entire sample. We next examine whether the extent of pass-through varies by the size of the beverage, whether it is a fountain drink, and by store type. Table 2 displays difference-in-differences estimates using the entire sample of handcollected retail data and register prices for beverages by size (20 ounce bottles, 2 liter bottles, and 12 packs of 12 ounce cans), for fountain drinks, and for store types (convenience, grocery, pharmacies, and liquor). One month after the tax was implemented, passthrough was greatest for 12 packs of 12-ounce cans at 1.729 cents per ounce or 86.5% of the tax. Seventy-three percent of the tax was passed through for 20-ounce bottles, and 67% of the tax was passed through for 2-liter bottles. Three months after the tax, in October, there were not major May 2021 Amer. J. Agr. Econ. 12 /0 2 4 7 /0 02 05 01 /0 11 10 9/ 8/ 03 7/ 06 6/ 5/ 01 4/ 04 04 2/ 3/ 1/ 07 −.5 0.5 1 1.5 2 998 Figure 3. Estimates of the impact on retail prices using the Nielsen Retail Scanner Data Note: This ﬁgure shows the impact of the SSB tax in Boulder on weekly prices of SSBs throughout 2017. The x-axis shows the month and date of the Saturday denoting the end of the week. The solid line displays the coefﬁcient estimates of interaction terms between a binary variable denoting that the store is in Boulder and a binary variable denoting the week. Additional variables include week ﬁxed effects, store ﬁxed effects, and product ﬁxed effects. The dashed lines represent the 95% conﬁdence interval. The tax was implemented on July 1. The ﬁgure shows that prior to July 1, the trends in prices in Boulder were similar to those in Larimer County. By the second week after July 1, the price per ounce of SSBs in Boulder is 1.08 cents per ounce higher than in Larimer County. differences in pass-through by the size of the beverage; it is roughly 70% for each beverage size. Prices on fountain drinks rose by roughly 2.6 cents per ounce or 130% of the tax. However, for fountain drinks there is evidence of a pre-existing trend; the estimate prior to the tax in April is 1.725 cents per ounce and is statistically signiﬁcant.29 The second half of table 2 reports results separately by store type. The SSB tax is approximately fully passed through at convenience stores: prices increase by 1.978 cents per ounce or 99% of the tax. In contrast, the tax is passed through at 75% for grocery stores and 68% for liquor stores. The estimate for pharmacies is 2.939 cents per ounce in August; however, the estimate in April is 2.703 cents per ounce and is statistically signiﬁcant, suggesting that for pharmacies there may have been a pre-existing trend in prices. These estimates are based on separate regressions for each store type. We also include products from all store types and interaction terms with each store type to test whether the differences by store type are statistically signiﬁcant, as shown 29 Pass-through rates do not vary based on the distance of the retailer within Boulder to the nearest competitor in an untaxed area. Pass-through rates are similar for soda, energy drinks, and sweetened teas, but lower for sports drinks at 53% in August. Appendix Table A8, in the online supplementary Appendix, displays estimates for speciﬁc products. in the online supplementary Appendix, Table A9. The only statistically signiﬁcant difference across store types is that prices increased by 0.767 cents per ounce more in convenience stores than grocery stores. Table 3 displays the difference-in-differences estimates from the Nielsen Retail Scanner Data. As shown in the ﬁrst column, the price of taxable beverages increased by 1.064 cents per ounce following the tax in Boulder. Thus, 53.2% of the tax was passed through to consumers. As shown in ﬁgure 3, the price increase occurs by the second week after the implementation of the tax, and the pass-through of the tax remains constant for the next six months. The estimate of the price increase from the Nielsen data is slightly larger than the estimate from the posted prices of the hand-collected data (1.064 vs. 0.839 in August) but smaller than the estimate from the register prices (1.064 vs. 1.422 in August). When we restrict the Nielsen sample to products that are also included in the hand-collected data, the pass-through estimate rises to 1.260 cents per ounce or 63%. There is no change in the prices of (untaxed) diet soft drinks, and there is a small statistically signiﬁcant increase in the price of (untaxed) bottled water of 0.178 cents per ounce. In the Nielsen data, prices increased by 0.781 cents per ounce for 20-ounce bottles (39.1% of the tax), 1.150 cents per ounce for 2-liter bottles (57.5% of the tax), and 0.985 cents per ounce for 12 packs of 12-ounce ounces (49.3% of the tax). Prices increased by 1.533 cents per ounce in grocery stores (76.7% of the tax), which is similar to the estimate in the hand-collected data. In contrast to the hand-collected data, in the scanner data, the estimate for pharmacies is 0.420 cents per ounce (21% of the tax) and is not statistically signiﬁcant. Nielsen includes a category of retailers called mass merchandise stores, which are large retail stores that include a grocery within the store, but grocery is not the focus of the retailer. Prices did not change following the tax in these stores.30 Difference-in-Differences Estimates for Restaurants Table 4 reports results using the handcollected data on fountain drinks and coffee drinks from restaurants. The price of fountain drinks increased by 0.870 cents per ounce in 30 When we separate these stores from grocery stores in the hand-collected data, we ﬁnd a similar result. Cawley, Frisvold, Jones and Lensing Sugar-Sweetened Beverage Tax 999 Table 1. Estimates of the Change in Retail Prices in Boulder after the SSB Tax Boulder x Apr Boulder x Aug Boulder x Oct N NxT Mean R2 Unbalanced taxable Posted prices Unbalanced taxable Register prices −0.391 (0.205) 0.839 (0.146) 0.795 (0.146) 2505 7218 7.890 0.969 −0.080 (0.239) 1.422 (0.165) 1.450 (0.178) 2505 7218 7.890 0.969 Balanced taxable Register prices Balanced untaxed Unbalanced untaxed Balanced taxable Posted prices −0.451 (0.262) 0.234 (0.176) 0.216 (0.186) 1605 4512 11.416 0.940 −0.476 (0.199) 0.811 (0.192) 0.703 (0.226) 909 3636 8.08 0.978 −0.098 (0.390) 1.550 (0.174) 1.437 (0.223) 909 3636 8.08 0.978 −0.367 (0.215) 0.239 (0.273) −0.024 (0.200) 544 2176 10.898 0.953 Results in this table are estimated using the hand-collected retail data. The dependent variable is the price in cents per ounce. The estimates show the change in the number of cents per ounce of the retail price relative to the prices in June in Fort Collins. Posted prices are the prices shown on the shelf for each item. Register prices are equal to the posted price plus the amount of the tax that is itemized on the receipt. Standard errors, in parentheses, are clustered at the store level. Additional variables that are included, but not shown, are month ﬁxed effects, day of the month ﬁxed effects, day of the week ﬁxed effects, store ﬁxed effects, and product ﬁxed effects. N represents the number of unique store speciﬁc items, N x T represents the number of unique store speciﬁc item observations across all waves. Mean is the pre-tax average price per ounce in cents. Estimates in the last three columns are balanced over products and stores. Boulder from June to August, relative to the price in Fort Collins, implying a pass-through of 43.5%. In contrast to retail prices, the prices of fountain drinks in restaurants continued to rise after August. In October, the relative price per ounce in Boulder was 1.483 cents higher than in June, for a pass-through of 74.2%. As also shown in the table, there is no statistically signiﬁcant change in the prices of untaxed products (coffee) as a result of the tax on SSBs, and the coefﬁcient in August is small in magnitude (−0.145). Estimates for the balanced sample of stores are similar to, but slightly smaller than, those for the entire (unbalanced) sample. Table 5 displays results using the price data scraped from restaurant menus on OrderUp. An advantage of these data, relative to the hand-collected restaurant data, is that they could be collected more frequently, so we have greater ability to test for differences in pretrends, as well as changes in pass-through over time after the tax. A limitation of the OrderUp data is that we generally do not observe the size of the drink in ounces, so we observe price per drink rather than price per ounce, and while we can estimate the change in overall price, we cannot estimate percent passthrough without assuming the amount of ounces for each beverage. The interaction of the indicator variable for Boulder with months prior to the tax (March, April, and May) yields no evidence of a differential trend between the treatment and control communities, which is consistent with the identifying assumption of the difference-indifferences design. For taxed beverages, the tax increased prices by 17.3 cents in August, 21.1 cents in September, and 20.2 cents in October. Prices also rose for untaxed beverages following the SSB tax: by 6.5 cents in August, 8.4 cents in September, and 7.8 cents in October. Beverages of unknown tax status (listed in column 3) experienced changes in price similar to those of untaxed items. If we assume that all beverages sold by restaurants through the OrderUp delivery service are 12-ounce servings, then prices of taxed beverages increased by 1.68 cents per ounce or 84.2% of the tax. If all beverages sold are 20-ounce servings, then prices increased by 1.01 cents per ounce or 50.5% of the tax. This range of 51 to 84% includes the estimate of 74% from the hand-collected data. Although we cannot precisely estimate the percentage pass-through of the tax, these data serve the important purposes of conﬁrming parallel trends for Boulder and the control communities prior to the tax, and for conﬁrming that the restaurant prices of taxed drinks rose more in Boulder than in the control communities after the tax. Discussion and Conclusion This paper makes several important contributions. We use four different datasets to N NxT Mean R2 Boulder x Oct Boulder x Aug Results in this table are estimated using the full sample of taxed products from the hand-collected retail data and the prices charged at the register. The dependent variable is the price in cents per ounce. The estimates show the change in the number of cents per ounce of the retail price relative to the prices in June in Fort Collins. Standard errors, in parentheses, are clustered at the store level. Additional variables that are included, but not shown, are month ﬁxed effects, day of the month ﬁxed effects, day of the week ﬁxed effects, store ﬁxed effects, and product ﬁxed effects. N represents the number of unique store speciﬁc items, N x T represents the number of unique store speciﬁc item observations across all waves. Mean is the pre-tax average price per ounce in cents. 2.703 (0.976) 2.939 (0.948) 3.096 (0.953) 357 1083 9.733 0.967 0.075 (0.558) 1.498 (0.222) 1.471 (0.178) 628 1877 9.733 0.972 −0.214 (0.280) 1.978 (0.219) 1.929 (0.279) 1076 3108 8.904 0.982 0.252 (0.318) 1.729 (0.238) 1.541 (0.235) 222 694 3.816 0.868 Boulder x Apr 0.287 (0.242) 1.461 (0.188) 1.422 (0.188) 854 2440 8.861 0.719 −0.119 (0.417) 1.342 (0.204) 1.343 (0.208) 416 1188 3.200 0.808 1.725 (0.867) 2.644 (0.470) 2.528 (0.493) 215 640 4.143 0.908 Pharmacy Grocery 12pk 20 oz 2L Fountain Convenience Store type Package size Table 2. Heterogeneity in Estimates of the Change in Retail Prices in Boulder after the SSB Tax Amer. J. Agr. Econ. 1.366 (0.350) 1.677 (0.321) 515 1150 9.057 0.965 May 2021 Liquor 1000 provide the most comprehensive assessment of how a city SSB tax inﬂuences prices of beverages in stores and restaurants. In doing so, we provide the ﬁrst evidence regarding the pass-through of the largest tax on SSBs in the United States, the 2 cent per ounce tax in Boulder, CO. Further, we provide the ﬁrst estimates of the pass-through of SSB taxes to prices of beverages in restaurants. Using two different datasets, the ﬁrst of which is hand-collected data from hundreds of retailers and the second of which is the Nielsen Retail Scanner Data, we estimate that the tax was substantially, but not fully, passed through to consumers in the form of higher prices at retailers. Data from transactions at store registers indicate that 71.1% of the tax was passed through one month after the tax was instituted and that the pass-through remained roughly constant for the next several months. The pass-through was similar across sizes of SSBs within three months of the implementation of the tax, including single-use 20-ounce bottles and larger 2-liter bottles and 12 packs of 12-ounce cans. Additionally, pass-through was larger for convenience stores than grocery stores. Importantly, we ﬁnd that estimated pass-through is lower when using scanner data than the hand-collected data (53.2% vs. 71.1%), which is largely consistent with comparisons across studies focusing on the same cities. This paper provides the ﬁrst direct comparison of the results across data sources and provides evidence that differences in the store types included in the data sources contribute to the differences in estimated pass-through; that is, pass-through at convenience stores in Boulder was higher than larger stores, and the scanner data has a much lower percentage of convenience stores and small store types. Finally, there is little evidence of any impact of the tax on the store prices of untaxed beverages. Overall, our estimates suggest that the tax on SSBs in Boulder was substantially, but not fully, passed through to consumers. With the exception of convenience stores, the 95% conﬁdence intervals rule out 100% pass-through. The estimates of the pass-through of the tax in Boulder are larger than estimates of the pass-through of the SSB tax in Berkeley (Falbe et al. 2015; Cawley and Frisvold 2017), Oakland (Cawley et al. 2020b), and Seattle (Powell and Leider 2020), but less than the pass-through of the tax in Philadelphia (Roberto et al. 2019; Seiler, Tuchman, and Yao 2020; Cawley et al. 2020a). Our estimates Observations R2 Number of products Number of stores Mean Results in this table are estimated using the Nielsen Retail Scanner Data. The dependent variable is the price in cents per ounce. The estimates show the change in the number of cents per ounce of the retail price relative to the prices before July in Larimer County, Colorado. Standard errors, in parentheses, are clustered at the store level. Additional variables that are included, but not shown, are week ﬁxed effects, store ﬁxed effects, and product ﬁxed effects. Mean is the pre-tax average price per ounce in cents. −0.014 (0.0293) 43,447 0.974 476 7 7.409 0.420 (0.304) 57,999 0.967 273 15 8.869 1.533 (0.025) 169,929 0.961 737 11 7.792 0.985 (0.202) 48,471 0.570 104 36 2.726 1.150 (0.202) 35,546 0.671 70 36 2.105 0.178 (0.0372) 134,780 0.696 562 36 7.245 Boulder x Post 1.064 (0.191) 292,330 0.957 912 36 8.169 −0.033 (0.0857) 213,665 0.957 666 36 6.634 0.781 (0.264) 30,666 0.493 65 36 8.675 Grocery stores 12 packs 2L bottles 20 oz. bottles Bottled water Diet soda Taxable products Table 3. Estimates of the Change in Retail Prices in Boulder after the SSB Tax Using Nielsen Retail Scanner Data Pharmacy Mass stores Cawley, Frisvold, Jones and Lensing Sugar-Sweetened Beverage Tax 1001 are lower than the estimates of the passthrough of taxes on SSBs in other countries (e.g., Colchero et al. 2015; Berardi et al. 2016; Grogger 2017; Bergman and Hansen 2019); however, this may be partly due to those studies lacking geographic control groups. Another major contribution is that we provide the ﬁrst evidence for any city on the pass-through of SSB taxes to the prices of beverages in restaurants. Data hand-collected from hundreds of restaurants indicate that the pass-through of the tax was 74.2% on fountain drinks (similar to the 71.1% found for beverage purchases at retailers), and the tax had no detectable impact on the prices of untaxed coffee drinks. For restaurants, the increase in prices is slightly more gradual than retailers; this could be due to restaurants in general changing their prices less frequently than retailers. These results have implications beyond Boulder. Many cities have recently enacted taxes on SSBs, and their effects are not well understood. This paper contributes to the growing literature on the impacts of these taxes. These results also have implications for simulations of the effect of SSB taxes on consumption, which have often assumed that taxes are fully passed through to consumers (e.g., Wang et al. 2012; Dharmasena, Davis, and Capps Jr 2014; Long et al. 2015). The results of this paper imply that consumers do not always bear the full burden of SSB tax (e.g., pass-through is not necessarily full) and that pass-through rates can vary across different localities and store types. It is commonly assumed that an excise tax will be incorporated into the shelf price (e.g., Chetty, Looney, and Kroft 2009). However, we ﬁnd that not all retailers increase the posted price of SSBs in response to the tax. Among retailers in Boulder selling SSBs, 21%, which primarily were convenience stores, chose to add the tax at the register and itemize it on the receipt. Ignoring these decisions of retailers would lead to a substantial underestimate of the pass-through rate. The estimated pass-through based on posted prices is 41.95%; whereas, pass-through based on register prices is 71.1%. Increasing the price at the register compared to the shelf could have important implications for the impact of the tax on passthrough, purchases, and the regressivity of the tax. The tax is more salient when it is included in the shelf price because it is observed at the point of decision making; 1002 May 2021 Amer. J. Agr. Econ. Table 4. Estimates of the Change in Hand Collected Restaurant Prices in Boulder after the SSB Tax Fountain full sample Boulder x Apr Boulder x Aug Boulder x Oct N NxT Mean R2 Coffee full sample Fountain balanced sample Coffee balanced sample −0.145 (0.412) 0.714 (0.545) 432 1055 23.315 0.905 −1.067 (0.598) 0.728 (0.398) 1.234 (0.390) 292 1168 7.853 0.700 −0.205 (0.415) 0.370 (0.540) 288 864 24.048 0.909 −0.853 (0.576) 0.870 (0.360) 1.483 (0.396) 423 1399 7.963 0.735 Results in this table are estimated using the hand-collected restaurant data. The dependent variable is the price in cents per ounce. The estimates for Boulder x August and Boulder x October show the change in the number of cents per ounce of the restaurant price relative to the prices in June in Fort Collins. Standard errors, in parentheses, are clustered at the store level. Additional variables that are included, but not shown, are month ﬁxed effects, day of the month ﬁxed effects, day of the week ﬁxed effects, restaurant ﬁxed effects, and product ﬁxed effects. N represents the number of unique restaurant speciﬁc items, N x T represents the number of unique restaurant speciﬁc item observations across all waves. Mean is the pre-tax average price per ounce in cents. consumers may not notice it being added at the register. Consistent with this, Chetty, Looney, and Kroft (2009) ﬁnd that alcohol purchases decrease more in response to the tax when the tax is incorporated into the posted price instead of added at the register. Goldin and Homonoff (2013) suggest that cigarette taxes imposed at the register could be less regressive than similar taxes incorporated into the posted prices if low-income consumers are more attentive to prices at the register than high-income consumers. Thus, the impact of the tax on SSB purchases could differ for the stores that imposed the tax at the register compared to the majority of stores in Boulder where the tax was incorporated into the posted price. The size of the tax can also inﬂuence the salience of the tax. For example, Taubinsky and Rees-Jones (2018) ﬁnd that consumers are less responsive to taxes that are not as salient on low-priced items. The tax in Boulder Table 5. Estimates of the Change in OrderUp Restaurant Prices in Boulder after the SSB Tax Boulder x Mar Boulder x Apr Boulder x May Boulder x Jul Boulder x Aug Boulder x Sept Boulder x Oct N NxT Mean R2 Taxed Untaxed Unknown 0.013 (0.021) 0.010 (0.009) 0.000 (0.003) 0.082 (0.041) 0.173 (0.067) 0.211 (0.087) 0.202 (0.089) 343 10976 2.448 0.921 0.011 (0.006) 0.011 (0.006) 0.007 (0.005) 0.027 (0.032) 0.065 (0.038) 0.084 (0.040) 0.078 (0.040) 345 11040 2.840 0.753 0.015 (0.018) −0.007 (0.012) 0.008 (0.014) 0.003 (0.020) 0.066 (0.032) 0.090 (0.039) 0.087 (0.039) 189 6048 3.447 0.745 Results in this table are estimated using the balanced sample of the OrderUp restaurant data. The dependent variable is the price in dollars per drink. The estimates show the change in the dollars per drink of the restaurant price relative to the prices in June in Fort Collins. Standard errors, in parentheses, are clustered at the store level. Additional variables that are included, but not shown, are month ﬁxed effects, restaurant ﬁxed effects, and product ﬁxed effects. N represents the number of unique restaurant speciﬁc items, N x T represents the number of unique restaurant speciﬁc item observations across all waves. Mean is the pre-tax average price per drink in dollar. Cawley, Frisvold, Jones and Lensing is the largest in the U.S., and Kroft et al. (2020) show that greater salience increases the incidence of taxes on consumers under imperfect competition. Thus, the size of the tax, with a greater salience, could potentially explain why the estimates of pass-through for retail stores are larger in Boulder than Berkeley, Oakland, or Seattle.31 Strengths of this analysis include: (a) a wide variety of data from stores and restaurants that were collected through various modes; (b) multiple periods of prices prior to the implementation of the tax, which allow us to assess whether the trends in prices are similar in the treated and the multiple comparison communities; (c) multiple periods of prices after the implementation of the tax, which allow us to determine how quickly restaurants and retailers respond to the tax; (d) prices from a wide range of products; (e) prices from all retailers and limited-service restaurants in the two communities, which minimizes sampling error; (f ) large sample sizes of hundreds of stores and hundreds of restaurants; (g) weekly prices from online restaurant menus; and (h) both posted and receipt prices from retailers. The study has limitations as well. Although we ﬁnd evidence consistent with the parallel trends assumption, we acknowledge that Fort Collins may be an imperfect control for Boulder. We have rich data on retail prices, but ideally, we would like to observe wholesale prices as well, to see the pass-through of the tax from the distributor to the retailer. We also lack data on sales, SSB consumption, or consumer weight. Another limitation of this study is that we have a small number of clusters; in particular, we have two geographic clusters (Boulder and Fort Collins). For the hand-collected data, we also have limited clusters in time, with four time periods; however, we have many more time points in the web-scraped data and continuous observations over time in the Nielsen scanner data. Despite these limitations, this paper presents the most comprehensive evidence regarding the incidence of a tax on SSBs, and it provides important contributions by examining the largest such tax in the United States, implications for tax salience, comparisons of 31 Although the tax in Philadelphia at 1.5 cents per ounce is smaller than the tax in Boulder, the tax in Philadelphia also includes diet soft drinks and received signiﬁcant media attention after implementation, in part due to legal challenges that did not occur in Boulder, Oakland, or Seattle. The tax in Cook County, IL also included diet soft drinks; it was particularly contentious and was repealed shortly after it was implemented. Sugar-Sweetened Beverage Tax 1003 results using different modes of data collection, and effects in restaurants and retailers. Supplementary Material Supplementary material are available at American Journal of Agricultural Economics online. References Allcott, Hunt, Benjamin B Lockwood, and Dmitry Taubinsky. 2019. Regressive Sin Taxes, with an Application to the Optimal Soda Tax. Quarterly Journal of Economics 134(3): 1557–626. An, Ruopeng, and Grace Maurer. 2016. Consumption of Sugar-Sweetened Beverages and Discretionary Foods among US Adults by Purchase Location. European Journal of Clinical Nutrition 70(12): 1396–400. Anderson, Simon P, Andre de Palma, and Brent Kreider. 2001. Tax Incidence in Differentiated Product Oligopoly. Journal of Public Economics 81(2): 173–92. Berardi, Nicoletta, Patrick Sevestre, Marine Tepaut, and Alexandre Vigneron. 2016. The Impact of a ‘Soda Tax’on Prices: Evidence from French Micro Data. Applied Economics 48(41): 3976–94. Bergman, U Michael, and Niels L Hansen. 2019. Are Excise Taxes on Beverages Fully Passed Through to Prices? The Danish Evidence. Public Finance Analysis 75 (4): 323–56. Besley, Timothy J, and Harvey S Rosen. 1999. Sales Taxes and Prices: An Empirical Analysis. National Tax Journal 52(2): 157–78. Bleich, Sara N, Hannah G Lawman, Michael T LeVasseur, Jiali Yan, Nandita Mitra, Caitlin M Lowery, Ana Peterhans, Sophia Hua, Laura A Gibson, and Christina A Roberto. 2020. The Association of a Sweetened Beverage Tax with Changes in Beverage Prices and Purchases at Independent Stores. Health Affairs 39(7): 1130–9. Bollinger, Bryan, and Steven E. Sexton. 2018. “Local Excise Taxes, Sticky Prices, and Spillovers: Evidence from Berkeley’s Soda Tax.” SSRN Working Paper. Bonnet, Celine, and Vincent Réquillart. 2013. Tax Incidence with Strategic Firms in the Soft Drink Market. Journal of Public Economics 106: 77–88. 1004 May 2021 Cameron, A Colin, Jonah B Gelbach, and Douglas L Miller. 2008. Bootstrap-Based Improvements for Inference with Clustered Errors. Review of Economics and Statistics 90(3): 414–27. Cameron, A Colin, and Douglas L Miller. 2015. A Practitioner’s Guide to ClusterRobust Inference. Journal of Human Resources 50(2): 317–72. Cawley, John, and David E Frisvold. 2017. The Pass-Through of Taxes on Sugar-Sweetened Beverages to Retail Prices: The Case of Berkeley, California. Journal of Policy Analysis and Management 36(2): 303–26. Cawley, John, David Frisvold, Anna Hill, and David Jones. 2019. The Impact of the Philadelphia Beverage Tax on Purchases and Consumption by Adults and Children. Journal of Health Economics 67: 102225.. 2020a. The Impact of the Philadelphia Beverage Tax on Prices and Product Availability. Journal of Policy Analysis and Management 39(3): 605–28.. 2020b. Oakland’s Sugar-sweetened Beverage Tax: Impacts on Prices, Purchases and Consumption by Adults and Children. Economics & Human Biology 37: 100865.. 2020c. The Pass-Through of a Tax on Sugar-Sweetened Beverages in Boulder, Colorado. American Journal of Agricultural Economics forthcoming, https://doi. org.10.1111/ajae.12191. Cawley, John, Barton Willage, and David Frisvold. 2018. Pass-Through of a Tax on Sugar-Sweetened Beverages at the Philadelphia International Airport. Journal of the American Medical Association 319(3): 305–6. Chetty, Raj, Adam Looney, and Kory Kroft. 2009. Salience and Taxation: Theory and Evidence. American Economic Review 99(4): 1145–77. Colchero, M Arantxa, Juan Carlos Salgado, Mishel Unar-Munguía, Mariana Molina, Shuwen Ng, and Juan Angel RiveraDommarco. 2015. Changes in Prices after an Excise Tax to Sweetened Sugar Beverages Was Implemented in Mexico: Evidence from Urban Areas. PLOS ONE 10(12): e0144408. Debnam, Jakina. 2017. Selection Effects and Heterogeneous Demand Responses to the Berkeley Soda Tax Vote. American Journal of Agricultural Economics 99(5): 1172–87. Amer. J. Agr. Econ. Dharmasena, Senarath, George Davis, and Oral Capps, Jr. 2014. Partial Versus General Equilibrium Calorie and Revenue Effects Associated with a SugarSweetened Beverage Tax. Journal of Agricultural and Resource Economics 39 (2): 157–73. Donald, Stephen G, and Kevin Lang. 2007. Inference With Difference-in-Differences and Other Panel Data. Review of Economics and Statistics 89(2): 221–33. Doyle, Joseph, Jr, and Krislert Samphantharak. 2008. $2.00 Gas! Studying the Effects of a Gas Tax Moratorium. Journal of Public Economics 92(3–4): 869–84. Falbe, Jennifer, Nadia Rojas, Anna H Grummon, and Kristine A Madsen. 2015. Higher Retail Prices of Sugar-Sweetened Beverages 3 Months after Implementation of an Excise Tax in Berkeley, California. American Journal of Public Health 105(11): 2194–201. Falbe, Jennifer, Matthew M Lee, Scott Kaplan, Nadia A Rojas, Alberto M Ortega Hinojosa, and Kristine A Madsen. 2020. Higher Sugar-Sweetened Beverage Retail Prices after Excise Taxes in Oakland and San Francisco. American Journal of Public Health 110: 1017–23. Fletcher, Jason M, David E Frisvold, and Nathan Tefft. 2010. The Effects of Soft Drink Taxes on Child and Adolescent Consumption and Weight Outcomes. Journal of Public Economics 94: 967–74.. 2015. Non-Linear Effects of Soda Taxes on Consumption and Weight Outcomes. Health Economics 24(5): 566–82. Fullerton, Don, and Gilbert E Metcalf. 2002. Tax Incidence. In Handbook of Public Economics, ed. Alan J Auerbach and Martin Feldstein, 1787–872.Amsterdam: Elsevier B.V. Goldin, Jacob, and Tatiana Homonoff. 2013. Smoke Gets in Your Eyes: Cigarette Tax Salience and Regressivity. American Economic Journal: Economic Policy 5(1): 302–36. Grogger, Jeffrey. 2017. Soda Taxes and the Prices of Sodas and Other Drinks: Evidence from Mexico. American Journal of Agricultural Economics 99(2): 481–98. Harding, Matthew, Ephraim Leibtag, and Michael F Lovenheim. 2012. The Heterogeneous Geographic and Socioeconomic Incidence of Cigarette Taxes: Evidence Cawley, Frisvold, Jones and Lensing from Nielsen Homescan Data. American Economic Journal: Economic Policy 4 (4): 169–98. Jones-Smith, Jessica C, Lina Pinero Walkinshaw, Vanessa M Oddo, Melissa Knox, Marian L Neuhouser, Philip M Hurvitz, Brian E Saelens, and Nadine Chan. 2020. Impact of a Sweetened Beverage Tax on Beverage Prices in Seattle, WA. Economics and Human Biology 39: 100917. Kotlikoff, Laurence J, and Lawrence H Summers. 1987. Tax Incidence. In Handbook of Public Economics, ed. Alan J Auerbach and Martin Feldstein, 1043–92. Amsterdam: Elsevier B.V. Kroft, Kory, Jean-William P. Laliberte, Rene Leal-Vizcaino, and Matthew J. Notowidigdo. 2020. “Salience and Taxation with Imperfect Competition.” National Bureau of Economic Research Working paper w27409. Long, Michael W, Steven L Gortmaker, Zachary J Ward, Stephen C Resch, Marj L Moodie, Gary Sacks, Boyd A Swinburn, Rob C Carter, and Y Claire Wang. 2015. Cost Effectiveness of a Sugar-Sweetened Beverage Excise Tax in the US. American Journal of Preventive Medicine 49(1): 112–23. Malik, Vasanti S, and Frank B Hu. 2011. Sugar-Sweetened Beverages and Health: Where Does the Evidence Stand? American Journal of Clinical Nutrition 94(5): 1161–2. Marinello, Samantha, Andrea A Pipito, Julien Leider, Oksana Pugach, and Lisa M Powell. 2020. The Impact of the Oakland Sugar-Sweetened Beverage Tax on Bottled Soda and Fountain Drink Prices in Fast-Food Restaurants. Preventive Medicine Reports 17: 101034. Okrent, Abigail, and Julian M. Alston. 2012. “The Demand for Disaggregated FoodAway-from-Home and Food-at-Home Products in the United States.” Economic Research Report (ERR) 139. Washington, DC: U.S. Department of Agriculture, Economic Research Service. Poterba, James M. 1996. Retail Price Reactions to Changes in State and Local Sales Taxes. National Tax Journal 49(2): 165–76. Powell, Lisa M, and Julien Leider. 2020. The Impact of Seattle’s Sweetened Beverage Tax on Beverage Prices and Volume Sold. Economics & Human Biology 37: 100856. Sugar-Sweetened Beverage Tax 1005 Powell, Lisa M, Julien Leider, and Pierre Thomas Léger. 2020. The Impact of the Cook County, IL, Sweetened Beverage Tax on Beverage Prices. Economics & Human Biology 37: 100855. Roberto, Christina A, Hannah G Lawman, Michael T LeVasseur, Nandita Mitra, Ana Peterhans, Bradley Herring, and Sara N Bleich. 2019. Association of a Beverage Tax on Sugar-Sweetened and Artiﬁcially Sweetened Beverages with Changes in Beverage Prices and Sales at Chain Retailers in a Large Urban Setting. Journal of the American Medical Association 321(18): 1799–810. Rojas, Christian, and Emily Wang. 2021. Do Taxes for Soda and Sugary Drinks Work? Scanner Data Evidence from Berkeley and Washington State. Economic Inquiry 59(1): 95–118. Rudd Center for Food Policy and Obesity. 2014. Sugar-Sweetened Beverage Taxes and Sugar Intake: Policy Statements, Endorsements and Recommendations, Hartford, CT: University of Connecticut.. http://www.uconnruddcenter.org/ﬁles/ Pdfs/SugaryDrinkTaxStatements.pdf Seiler, Stephan, Anna Tuchman, and Song Yao. 2020. The Impact of Soda Taxes: PassThrough, Tax Avoidance, and Nutritional Effects. Journal of Marketing Research. https://doi.org/10.1177/0022243720969401. Taubinsky, Dmitry, and Alex Rees-Jones. 2018. Attention Variation and Welfare: Theory and Evidence from a Tax Salience Experiment. Review of Economic Studies 85(4): 2462–96. Thow, Anne Marie, Shauna M Downs, Christopher Mayes, Helen Trevena, Temo Waqanivalu, and John Cawley. 2018. Fiscal Policy to Improve Diets and Prevent Noncommunicable Diseases: From Recommendations to Action. Bulletin of the World Health Organization 96(3): 201. Wang, Y Claire, Pamela Coxson, Yu-Ming Shen, Lee Goldman, and Kirsten BibbinsDomingo. 2012. A Penny-per-Ounce Tax on Sugar-Sweetened Beverages Would Cut Health and Cost Burdens of Diabetes. Health Affairs 31(1): 199–207. Weyl, E Glen, and Michal Fabinger. 2013. Pass-tThrough as an Economic Tool: Principles of Incidence under Imperfect Competition. Journal of Political Economy 121 (3): 528–83.

Cawley et al (2021) PDF: The Pass-Through of a Tax on Sugar-Sweetened Beverages in Boulder, Colorado

Document Details

Tags

Related

Summary

Full Transcript

Upgrade to continue