Evaluation of the sugar-sweetened beverage tax in Oakland, United States, 2015–2019: A quasi-experimental and cost-effectiveness study

JournalFeeds

Authors:
Justin S. White , et al.

Abstract

Methods and findings

An SSB tax ($0.01/oz) was implemented on July 1, 2017, in Oakland. The main sample of sales data included 11,627 beverage products, 316 stores, and 172,985,767 product-store-month observations. The main analysis, a longitudinal quasi-experimental difference-in-differences approach, compared changes in beverage purchases at stores in Oakland versus Richmond, California (a nontaxed comparator in the same market area) before and 30 months after tax implementation (through December 31, 2019). Additional estimates used synthetic control methods with comparator stores in Los Angeles, California. Estimates were inputted into a closed-cohort microsimulation model to estimate quality-adjusted life years (QALYs) and societal costs (in Oakland) from 6 SSB-associated disease outcomes. In the main analysis, SSB purchases declined by 26.8% (95% CI −39.0 to −14.7, p < 0.001) in Oakland after tax implementation, compared with Richmond. There were no detectable changes in purchases of untaxed beverages or sweet snacks or purchases in border areas surrounding cities. In the synthetic control analysis, declines in SSB purchases were similar to the main analysis (−22.4%, 95% CI −41.7% to −3.0%, p = 0.04). The estimated changes in SSB purchases, when translated into declines in consumption, would be expected to accrue QALYs (94 per 10,000 residents) and significant societal cost savings (>$100,000 per 10,000 residents) over 10 years, with greater gains over a lifetime horizon. Study limitations include a lack of SSB consumption data and use of sales data primarily from chain stores.

Author summary

1. Introduction

Sugar-sweetened beverages (SSBs) constitute a major, nonnutritional source of calories and added sugars in the US diet and are associated with obesity, type 2 diabetes, and overall mortality [13]. As of 2021, excise taxes had been levied on beverage distributors in 7 US localities and more than 35 countries to reduce the risk of diet-sensitive chronic disease and increase government revenue for health promotion [4,5].

SSB taxation has been considered by many to be among the most promising population-level interventions to prevent and control chronic disease [6]. Following implementation of Berkeley’s first-in-the-nation SSB tax in 2015, a growing number of studies have documented the benefits of US municipal-level taxes on consumers [7]. This body of literature has shown that SSB taxes increase retail SSB prices for consumers, leading to short-term reductions in purchase and intake of SSBs [811]. What remains unclear is whether SSB taxation produces sustained declines in SSB intake, and the extent to which consumers substitute SSBs for other foods or beverages or cross borders to shop in untaxed locations. In addition, the net effect of SSB taxes on health and costs when considering the health profiles of exposed populations has not been estimated [1214]. A 2021 federal commission on diabetes recommended that the US government levy a federal SSB tax to improve diabetes prevention and control efforts [15]. This study aimed to provide critical and timely evidence on beverage purchasing patterns and health and cost impacts to inform such a policy.

On July 1, 2017, the city of Oakland, California implemented a $0.01/oz tax on the distribution of all beverages ≥12 ounces containing added sugar. We evaluated whether the SSB tax in Oakland, California was associated with reduced SSB purchases up to 30 months after tax implementation. Prior research has found that the Oakland tax increased retail prices of SSBs by approximately 14%, although evidence of changes in SSB purchasing and intake following the tax has been mixed [1620]. We also assessed changes in purchases of untaxed beverages, cross-border purchases, compensatory food purchases, and subgroup differences by beverage category, beverage size, store type, and area-level household income. Finally, we performed a microsimulation to estimate the health and cost impact across major disease outcomes associated with SSB intake, ensuring that our results are informative to ongoing policymaking.

2. Methods

This study used retail sales data and difference-in-differences and synthetic control approaches to estimate the impact of the SSB tax on beverage purchases. A microsimulation model used the quasi-experimental estimates and a variety of survey data (described below) to implement a microsimulation model of the cost-effectiveness of the SSB tax. All analyses were prespecified (S1 Prospective Plan). The study follows Consolidated Health Economic Evaluation Reporting Standards (S1 Checklist).

2.1. Data

Retail sales data (described elsewhere [21]) were licensed from Information Resources Inc. (IRI), which obtains data from major US retailers. IRI establishes relationships with store chains and independent stores to share electronic data regarding all sales transactions at participating stores. Studies have found that sales in IRI retail scanner data align well with other sources of retail sales data [22]. Longitudinal sales data were provided in 4-week periods, hereafter monthly, for beverages purchased from January 1, 2015 (30 months before-tax in Oakland) through December 31, 2019 (30 months after-tax). Data were provided at the individual product level based on a unique universal product code (UPC) and contained monthly store-level volume and dollar sales of each beverage purchased (11,627 UPCs). Our data files contain the total sales value and volume by store for each beverage product and selected food products during a 4-week period.

2.2. Beverage classification

Beverages were classified as taxed or untaxed based on Oakland tax regulations and classified into 8 taxed and 15 untaxed beverage categories (see S1 Text). Nutrient and ingredient data were drawn from Label Insight, the then-largest publicly available branded food composition database in the US [23,24], and a hand-coding procedure from online searches (S1 Text). Beverages were also classified as individual- or family-sized (defined as ≥36 oz) [25].

2.4. Statistical analyses

This study used a difference-in-differences approach to compare beverage purchases before versus after an SSB tax in Oakland (the intervention city) compared with those in Richmond (a noncontiguous comparator city 12 miles from Oakland with a similar sociodemographic and health profile [S1 Table]). Difference-in-differences is a common quasi-experimental technique used to evaluate health policies, including SSB taxes, when the exposure is not randomly assigned, because of its ability to restrict the scope of potential confounding from secular trends [11,26,27]. Secondarily, we conducted a synthetic control approach that constructed a “synthetic” Oakland by assigning weights to stores from our comparator location, Los Angeles (LA), using pretax SSB purchases and a set of location-specific prognostic factors [28]. Synthetic control methods take a data-driven approach to constructing a comparison group and may reduce susceptibility of estimates to unmeasured time-varying confounders [29].

2.4.1. Main analysis of volume purchases.

The main analysis, conducted at the UPC level, uses a difference-in-differences approach to assess changes in volume of SSBs purchased before versus after tax for Oakland compared with Richmond. By adjusting for the counterfactual change in outcomes as measured in the comparator city, differences-in-differences allows one to attribute any differential changes in Oakland to the tax. The primary independent variable in the difference-in-differences analysis is an indicator variable for whether the observation was recorded in Oakland after implementation of the tax. The difference-in-differences analysis also included fixed effects (i.e., indicators) for store, year-quarter, and UPC. The regressions therefore adjusted for measured or unmeasured confounders that were unchanged during the study period within store (e.g., store type), within the store’s neighborhood (e.g., racial/ethnic composition), within year-quarter (e.g., seasonality), or within UPC (e.g., package size, calories) [27,30]. The fixed effects account for any changes over time in sample composition of stores or UPCs. In particular, UPCs discontinued prior to the tax or introduced after the tax are not a source of variation used to identify (or bias) the difference-in-differences estimate. Standard errors were clustered by zip code to account for correlated outcomes over time. P values were corrected for multiple hypothesis testing using Anderson’s sharpened False Discovery Rate procedure [31]. See S1 Text for additional details.

Beverage substitution was also assessed with a difference-in-differences regression of purchases of untaxed beverages. It compares pre-post changes in purchases of untaxed beverages in Oakland while differencing out secular trends in purchases of untaxed beverages in comparator city Richmond. Cross-border shopping was assessed with a difference-in-differences regression of SSB purchases in the border areas within approximately 3 miles [32] of Oakland versus in equidistant border areas of Richmond. Compensatory purchases of sweet foods have been hypothesized to increase following an SSB tax as consumers search for cheaper sources of sugar [3335]. Compensatory food purchases were assessed with a difference-in-differences regression of selected sweet snacks (packaged baked goods, 5,130 UPCs) in Oakland compared with Richmond. Subgroup analyses were performed by beverage category (listed in S1 Text), store type (convenience, pharmacy, supermarket/grocery store), beverage size (individual- versus family-sized), and income around the store’s location (store zip code with above- versus below-median proportion of households with <$35,000 income). All difference-in-differences analyses included fixed effects for store, year-quarter, and UPC.

2.4.2. Secondary and sensitivity analyses of volume purchases.

To examine dynamic treatment effects, we estimated an event study difference-in-differences regression of the time-varying association between the tax and SSB purchases [27,36]. The event study provided quarter-by-quarter estimated effects [37]. The significance of the before-tax coefficients provided a diagnostic of parallel pre-event outcome trends, an indicator of whether the identifying assumption of difference-in-differences estimation may be satisfied [37,38].

Finally, while selecting nearby Richmond as a comparison city accounted for any potential effects of SSB tax-related media on purchases, its proximity meant that SSB purchases in Richmond could have increased after the tax due to cross-border purchasing from Oakland into Richmond city proper or border areas. Therefore, we also conduct a secondary analysis that estimated a synthetic control model with a matched comparison group constructed from stores in LA, California. The synthetic control method algorithmically found the weighted average of LA stores that best matched the trend of before-tax values of the outcome and prognostic covariates in Oakland [28,29]. In this way, the method may be more robust to unmeasured time-varying confounders [28], although it relies on aggregated data by store. We considered as prognostic covariates several sociodemographic characteristics of the store zip code from the 2016 American Community Survey and the 2010 US census, including: population size (2010), median household income (2016), racial/ethnic composition (proportion non-Hispanic White, non-Hispanic Black, Hispanic, and Asian, 2010), proportion in poverty (household income <$10,000, 2016), proportion aged 18 to 24 (2010), and number of housing units (2010). Prognostic variables were selected due to their hypothesized relationship with SSB purchases, informed by prior literature [3941]. Uncertainty estimates and statistical significance of the model were based on placebo treatment effects for all stores in the donor pool, as if each control store had been treated. A randomization inference approach was used to calculate whether or not the estimated effect of the actual treated unit is large relative to the distribution of placebo effects (i.e., the proportion of control units that have an estimated placebo effect as large as that of the treated unit) [28].

We conducted sensitivity analyses for the synthetic control estimates for specifications that included Richmond (as well as LA) stores in the donor pool of comparator stores, adjustment for average retail price of taxed products per ounce as an additional covariate (omitted from the main model because it is endogenous to price), a cross-validation procedure of the four quarters before the tax with the early pretax period acting as a training period [29] and restricting the sample to a set of products that were available for all time periods.

2.4.3. Microsimulation model.

We simulated the population of Oakland to determine the anticipated effect of the tax-induced decline in SSB intake on quality-adjusted life years (QALYs) and societal costs from SSB-associated disease outcomes. We used demographic data from the American Community Survey (2019) from Oakland’s Alameda County and assigned simulated persons before-tax SSB intake levels and disease risk factors/history (S6 Table) by repeatedly sampling from age-, sex-, race-, and ethnicity-matched participants in the National Health and Nutrition Examination Survey (NHANES, 2015 to 2018). Following our previously published simulation for modeling the impact of SSB intake changes on health and cost outcomes [42], we estimated 6 disease outcomes: obesity (body mass index [BMI] of ≥30 kg/m2) [43], coronary heart disease (angina, myocardial infarction, cardiac arrest, ischemic heart disease, or heart failure) [44], cerebrovascular accident (ischemic or hemorrhagic stroke) [45], type 2 diabetes mellitus (hemoglobin A1c of ≥6.5%, fasting plasma glucose of ≥126 mg/dL, or 2-hour oral glucose tolerance test result of ≥200 mg/d) [46], chronic kidney disease (estimated glomerular filtration rate of <90 mL/min [47], using the Chronic Kidney Disease-Epidemiology [CKD-EPI] equation) [48], and both dental caries and periodontal disease (tooth decay of the permanent teeth, periodontitis, or more advanced disease, based on clinical attachment loss and periodontal probing depth) [49]. We used prior estimates of the demographic-specific incidence rate and mortality associated with these outcomes (S6 Table) [50] and then reduced the incidence rates to reflect the relative risk reductions associated in each incidence rate and mortality from reduced SSB intake. Binomial random variables were drawn with probability equal to the incidence rate of each condition by demographics (S9 Table), and fatality rates calibrated to match the demographic mortality rate estimates (S10 Table). Without the SSB tax, the baseline levels of SSB consumption and associated outcome incidence were assumed. With the SSB tax, the levels of tax reduction were first assumed to equal the level of SSB consumption reduction, and the relative risk shown in the main text from meta-analytic estimates were used to reduce the incidence rate for each outcome (see S1 Text).

We first simulated reductions in SSB intake in Oakland based on the point estimates and 95% confidence intervals from our difference-in-differences estimates, assuming the percent reduction in purchasing equaled the percent reduction in consumption [51,52]. We repeatedly sampled from normal distributions constructed around the difference-in-differences point estimates and their 95% confidence intervals to construct uncertainty estimates around our simulated outcomes. We also simulated how much cross-border purchasing and substitution to other foods and untaxed beverages would have to occur to negate the tax effects (the “point of cost neutrality” in a threshold analysis [53]). The level of SSB consumption reduction was varied until we found the point which the healthcare cost savings from the tax was equal to the cost of the tax itself, estimating the tax cost as the increased price per ounce observed in the scanner data, multiplied by the level of consumption reflected in the posttax environment for each county. This inherently assumes that the tax itself produces a deadweight loss of market inefficiency that amounts to a societal cost [54,55]. We note the work of Alcott and colleagues [56] that indicates the deadweight loss itself can be challenging to identify for an SSB tax due to difficulty calculating how much the tax corrects for internalities (i.e., the costs that consumers impose on themselves).

For all health outcomes except obesity, we estimated the disutility of each outcome to calculate QALYs based on a survey-based assessment [57]. For obesity, we adopted a prior review of BMI-specific disutility [58], weighted by the BMI distribution in NHANES to assess quality-of-life losses beyond those related to the aforementioned outcomes (e.g., for obesity-related liver disease).

We obtained healthcare costs (inpatient, outpatient, and prescription drug spending) per patient from the Medical Expenditure Panel Survey (MEPS, 2018) and attendant surveys for other obesity complications (S6 Table). We included healthcare costs specific to the type of primary insurance coverage (commercial, Medicare, Medicaid/other public, or uninsured; S6 Table). We adjusted costs for inflation to January 2021 US dollars using the Consumer Price Index [59]. We then computed cost and disutility from both a healthcare perspective (including healthcare costs) and societal perspective (including both healthcare costs and the costs of the tax) on both 10-year and lifetime horizons and discounted at a standard 3% annual rate (S1 Checklist) [60,61]. We performed uncertainty analyses by repeated Monte Carlo sampling from the distributions of each input parameter, by specifically constructing a log-normal distribution based on the mean input parameter estimate and the standard deviations around those means for each input parameter (see S6 Table). After rerunning the model 10,000 times, the number of iterations that resulted in stability of the model results to rounding error, we tabulated the 95% credible intervals around the distribution of each outcome for each city’s population to provide a sense of the uncertainty bounds around the modeling results.

The research was determined not to meet the criteria for human participant research of the institutional review board at the University of California San Francisco.

3. Results

3.2. Main analysis of volume purchases

SSB purchases declined by 26.8% (95% CI −39.0 to −14.7, p < 0.001) in the 30 months after tax implementation in Oakland compared with Richmond, an average monthly change of −45,220 oz per store (95% CI −65,667 to −24,773, p < 0.001) according to the difference-in-differences estimate (Table 2). There was no significant change in untaxed beverage purchases, cross-border SSB purchases compared with Richmond border areas, or compensatory purchases of sweet snacks (Table 2).

Following the tax, SSB purchases in Oakland declined for several beverage categories, including sweetened soda (−23.1%; −8,664 oz per store, 95% CI −12,244 to −5,081, p < 0.001), fruit drinks (−30.4%; −8,383, 95% CI −14,114 to −2,654, p = 0.006), sports drinks (−42.4%; −21,415, 95% CI −31,883 to −10,949, p < 0.001), and sweetened teas (−24.4%; −7,484, 95% CI −11,713 to −3,256, p < 0.003). SSB purchases declined by 38.4% in supermarkets (−183,767 oz per store, 95% CI −229,894 to −137,640, p < 0.001) but did not significantly change in convenience stores or pharmacies. We found similar declines in SSB purchases for individual-sized (−28.3%; −26,296 oz per store, 95% CI −38,221 to −14,372, p < 0.001) and family-sized (−25.3%; −19,052, 95% CI −28,936 to −9,168, p < 0.001) products, and for stores in lower-income (−25.0%; −30,902, 95% CI −39,533 to −22,272, p = 0.003) and higher-income (−26.5%; −57,012, 95% CI −92,624 to −21,400, p = 0.003) areas.

3.3. Secondary analyses of volume purchases

3.3.1. Change in volume purchases over time (event study).

The event study estimates in the pretax period are not statistically different from zero, suggesting that volume sales in the intervention and comparator cities did not change differentially prior to tax implementation (Fig 1). This finding supports the parallel trends condition for identification of the difference-in-differences model. The magnitude of the association between the tax and purchases grew in Oakland during the first 2 year-quarters after-tax, remaining stable thereafter.

thumbnail

Fig 1. Time-varying association between the Oakland tax and volume sales of SSBs (plots from event study difference-in-differences models).

Note: This plot shows the change in average monthly ounces sold per product in a store, based on an event study difference-in-differences regression. Quarterly estimates are relative to the quarter before the tax (quarter −1, red dotted line). Gray dotted lines are 95% confidence intervals from robust standard errors clustered by zip code.


https://doi.org/10.1371/journal.pmed.1004212.g001

3.3.2. Synthetic control estimates of volume purchases.

SSB purchases in Oakland declined by 22.4% (95% CI −41.7% to −3.0%, p = 0.04) following tax implementation compared with a synthetic control group of LA stores (Fig 2, Table 3). In sensitivity analyses, estimates were similar in all specifications except for a muted response when using a set of products available in all time periods (−15.1%) (Table 3). Restricting the data to a consistent set of products provides a stable comparison group, but it could deliver biased estimates if the rate of products entering the market changes over time.

In synthetic control analyses, cross-border SSB purchases did not change in Oakland border areas relative to LA border areas (Table 3).

thumbnail

Fig 2. Synthetic control estimates of the association between the Oakland tax and the volume of SSB purchases.

Note: This figure displays the proportionate change in the volume of SSB purchases in Oakland compared with a synthetic control group composed of a data-driven weighted average of stores in LA. Gray dotted lines are 95% confidence intervals.


https://doi.org/10.1371/journal.pmed.1004212.g002

3.4. Microsimulation estimates

Summary statistics of model input parameters are provided in S6 Table. The simulated Oakland population, based on the ACS (2019), had a median age of 36.9 years old, with 66% of the population aged 18 to 64; 52% male; and 29% White, 24% Black, 14% Asian, and 27% Hispanic (including all races under Hispanic ethnicity; S6 Table).

The simulated population, based on NHANES populations sampled and weighted to reflect the American Community Survey demographics of Alameda County, had a 17.4% prevalence of obesity, 5.2% prevalence of coronary heart disease, 3.0% prevalence of cerebrovascular accident, 7.9% prevalence of diabetes, 11.3% prevalence of chronic kidney disease, and 82.1% prevalence of dental disease (S6 Table); the first 4 statistics were possible to validate against the California Health Interview Survey (CHIS) and were found to be within 0.1% absolute percentage points of estimates from CHIS for Alameda County (2019) [62].

The simulated Oakland population, in the context of the SSB tax, was estimated to reduce net SSB intake by 26.8%, or 1.33 oz per person per day. The reduced intake was estimated to accrue 94.0 QALYs per 10,000 population over 10 years, and 967.6 QALYs over a lifetime (Table 4). From a societal perspective, the net tax impact per 10,000 people would save an estimated $102,154 over 10 years ($1.02 per person per year) and $1,251,817 over a lifetime ($2.81 per person per year). The long-term cost savings would be fully negated if the 26.8% reduction in SSB purchases resulted in a reduction in actual intake of only 0.3% (e.g., due to cross-border purchases or dietary substitution).

4. Discussion

This study used rigorous quasi-experimental methods to examine the association between the Oakland SSB tax and sustained changes in volume purchases up to 30 months after-tax. Following the tax, SSB purchases declined by 26.8% in Oakland relative to a comparator city in the same market. This effect size, when combined with a retail price increase of 14% [18], corresponds to a price elasticity estimate of −1.9, in line with meta-analytic evidence from other studies (mean −1.6, 95% CI −2.1 to −1.1) [11]. The estimated decline in SSBs from the difference-in-differences analysis was similar when using a matched sample of LA stores in the synthetic control analysis (22.4%). These SSB reductions would be expected to deliver significant health improvements over 10-year and lifetime horizons, respectively. The QALYs gained would represent an increase of approximately 0.15% over 10 years, growing to 0.35% over a lifetime. The SSB reductions were also projected to save $1,088 per QALY saved (>$100,000 per 10,000 people) over a 10-year horizon and $1,294 per QALY saved (>$1M per 10,000 people) over a lifetime horizon. This adds to the evidence base that SSB taxes are projected to be highly cost-effective or even cost-saving [1214,63].

To give a sense of how our cost-effectiveness results compare to other public health interventions, prior studies estimate that a clinical child obesity intervention would have a cost of $774 per QALY saved; eliminating the tax subsidy for television advertising would generate cost savings of $0.10 per QALY saved; a $1 cigarette tax (20% to 40% increase in price across states compared with approximately 15% associated with Oakland’s SSB tax [18]) plus anti-smoking media campaign would generate cost savings of $573,346 to $797,701 per QALY saved; smoke-free workplace policies have a cost of $726 per QALY saved; and traffic-related air pollution measures such as low-emission zones have a cost of $3,023 (£2,465) per QALY saved [6467]. Thus, the SSB tax is estimated to be less cost-effective than the cigarette tax plus media campaign but more cost-effective than the smoke-free workplace policy and air pollution measures.

We investigated several potential ways that consumers may have compensated for the tax—substituting to untaxed beverages or high-sugar foods or shopping in untaxed neighboring jurisdictions. We did not find significant increases in purchases of untaxed beverages, neither in aggregate nor for different categories of untaxed beverages. Furthermore, we did not observe offsetting caloric compensation in Oakland, adding to prior evidence finding little or no caloric compensation following SSB taxes [3335]. In addition, we did not find significant cross-border shopping, a point we return to in the next paragraph.

Prior studies have found that Oakland’s tax increased SSB prices by roughly 14% to 15% [1620] and reported a more muted after-tax decline in SSB sales. One study found no decline in SSB purchases 1 year after-tax, although it included only 34 households in Oakland [20]. A pair of studies found declines in retail sales in Oakland of 14% to 18% 1 and 2 years posttax, compared with Sacramento, which was offset by roughly half due to cross-border shopping [19,68]. In contrast, we found greater reductions in SSB purchasing and no significant cross-border shopping across several model specifications. We speculate that estimates may, in part, be sensitive to estimation approach, data set, and the choice of comparator. We assessed Richmond’s validity by conducting diagnostic tests of parallel before-tax trends and found Richmond’s city limits and border area to be a well-matched comparator to Oakland’s city limits and border areas. We were able to adjust for UPC and store characteristics, removing many potential confounders. Our findings were further supported in our complementary analysis that used a synthetic control; the use of the alternative, data-driven control group yielded a similar estimate of the tax’s effect and no significant changes in cross-border shopping. That said, even if our estimated tax effects were offset by half due to cross-border shopping, the effect size would remain substantial.

We observed heterogeneity in the response to the SSB tax, with SSB purchases declining more for several beverage categories (e.g., soda and sports drinks) than in others (sweetened coffee and flavored water). SSB purchases also declined most in supermarkets, where SSB prices in Oakland increased somewhat more after-tax than in other stores [16,18]. This finding is particularly important given that two-thirds of calories in the US diet come from grocery stores [69].

Insofar as recent policy recommendations have included scaling SSB taxation to the state or federal levels, the projected improvements in quality of life and cost savings we find are particularly relevant. While our findings are not necessarily generalizable to all US regions, studies of other US cities have found similar reductions in SSB purchases [25,33,70]. This suggests that our health and cost predictions may approximate the benefits that could accrue through a nationwide SSB tax. An SSB tax at the state or federal levels would be expected to provide fewer opportunities for cross-border shopping than a municipal tax, as most consumers would face high time and travel costs by trying to avoid the tax. Thus, our estimates contribute to understanding the benefits of state or federal taxation, regardless of whether the extent of cross-border shopping approximates our estimate or higher estimates from the literature.

4.1. Limitations

This study has several limitations. First, the tax status and beverage category could not be determined for a small proportion of products (constituting 3% of sales volume). However, any resultant bias was likely small, as non-ascertainment did not differ across cities. Second, our study relied on beverage purchases rather than consumption. Retail purchases closely correlate with consumption, although this correlation has been studied at the household, not individual level [51,52]. Survey-based measures of SSB intake may be subject to reporting bias [71] and are costly and challenging to collect across an entire population. We also found that the estimated decline in SSB purchases was 100-fold greater than the consumption decline needed for the tax to be cost-neutral, according to our sensitivity analysis. Third, because of the distance between Richmond to Oakland, cross-border shopping from Oakland to Richmond is possible and could lead to overestimation of tax effects in the difference-in-differences regressions. However, we failed to observe a crude increase in SSB purchasing in Richmond (S2 Table) or significant increases in the more proximate border areas of intervention cities, making this unlikely. This concern, in part, motivated the synthetic control analyses with an LA-only synthetic control group to avoid this potential bias; it too revealed no significant changes in cross-border shopping compared with LA border areas (Table 3). Fourth, the IRI data primarily reflect purchases in chain stores; our results may not apply to smaller and independent stores, such as corner liquor or convenience stores. In Philadelphia, independent stores experienced a 38% decline in SSB purchases following its SSB tax, which fell at the high end of estimates for SSB declines in large chains [33]. Insofar as independent stores are more prevalent in low-income neighborhoods, our estimates may understate differences in tax effects by area-level income. The fact that we observed similar reductions in higher- versus lower-income neighborhoods suggests that the SSB tax had substantial impact across the neighborhood income spectrum. Fifth, the availability of 2 cities only for the difference-in-differences analysis did not allow for clustering standard errors by city; rather, we cluster standard errors by zip code that may understate the degree of uncertainty in our estimates. Sixth, the microsimulation omitted any health or economic benefits of public investments using SSB tax revenues and did not estimate costs due to lost productivity and work absenteeism because we lacked data on which individuals purchased these products and their associated work histories. As such, the net benefits of SSB taxation we derived likely represent an underestimate [4,42,72]. Further research might also consider the cost-effectiveness of SSB taxation for key population subgroups, including differences by race/ethnicity, income, and chronic disease risk. A further limitation of our modeling analysis is that we cannot incorporate into the societal cost estimates the potential costs of taxation in terms of declining revenue to SSB companies (and associated surplus accumulation, or employment or economic contractions). However, such revenue effects remain debated as SSB company literature suggests that these companies rapidly pivot to new product entrances and alternatives such as zero-calorie beverages when SSB consumption declines [73].

References

  1. 1.
    Dietary Guidelines Advisory Committee. Scientific report of the 2015 Dietary Guidelines Advisory Committee. Washington, DC: U.S. Department of Agriculture, Agricultural Research Service; 2015.
  2. 2.
    Malik VS, Li Y, Pan A, Koning LD, Schernhammer E, Willett WC, et al. Long-Term Consumption of Sugar-Sweetened and Artificially Sweetened Beverages and Risk of Mortality in US Adults. Circulation. 2019;139(18):2113–2125. pmid:30882235
  3. 3.
    Imamura F , O’Connor L, Ye Z, Mursu J, Hayashino Y, Bhupathiraju SN, et al. Consumption of sugar sweetened beverages, artificially sweetened beverages, and fruit juice and incidence of type 2 diabetes: systematic review, meta-analysis, and estimation of population attributable fraction. BMJ. 2015;351:h3576. Epub 2015/07/23. pmid:26199070; PubMed Central PMCID: PMC4510779.
  4. 4.
    Krieger J, Magee K, Hennings T, Schoof J, Madsen KA. How sugar-sweetened beverage tax revenues are being used in the United States. Prev Med Rep. 2021;23:101388. pmid:34040929
  5. 5.
    Hattersley L, Thiebaud A, Fuchs A, Gonima A, Silver L, Mandeville K. Taxes on Sugar-Sweetened Beverages: International Evidence Experiences. The World Bank; 2020.
  6. 6.
    Andreyeva T, Chaloupka FrJ, Brownell KD. Estimating the potential of taxes on sugar-sweetened beverages to reduce consumption and generate revenue. Prev Med. 2011;52(6):413–416. Epub 2011/03/30. pmid:21443899.
  7. 7.
    Falbe J, Thompson HR, Becker CM, Rojas N, McCulloch CE, Madsen KA. Impact of the Berkeley excise tax on sugar-sweetened beverage consumption. Am J Public Health. 2016;106(10):1865–1871. PubMed Central PMCID: PMC5024386. pmid:27552267
  8. 8.
    Cawley J, Thow AM, Wen K, Frisvold D. The Economics of Taxes on Sugar-Sweetened Beverages: A Review of the Effects on Prices, Sales, Cross-Border Shopping, and Consumption. Annu Rev Nutr. 2019;39:317–338. pmid:31116649.
  9. 9.
    Teng AM, Jones AC, Mizdrak A, Signal L, Genç M, Wilson N. Impact of sugar-sweetened beverage taxes on purchases and dietary intake: Systematic review and meta-analysis. Obes Rev. 2019;20(9):1187–1204. Epub 2019/06/21. pmid:31218808.
  10. 10.
    Redondo M, Hernández-Aguado I, Lumbreras B. The impact of the tax on sweetened beverages: a systematic review. Am J Clin Nutr. 2018;108(3):548–563. pmid:30535085
  11. 11.
    Andreyeva T, Marple K, Marinello S, Moore TE, Powell LM. Outcomes Following Taxation of Sugar-Sweetened Beverages: A Systematic Review and Meta-analysis. JAMA Netw Open. 2022;5(6):e2215276–e. pmid:35648398
  12. 12.
    Wang YC, Coxson P, Shen Y-M, Goldman L, Bibbins-Domingo K. A penny-per-ounce tax on sugar-sweetened beverages would cut health and cost burdens of diabetes. Health Aff (Millwood). 2012;31(1):199–207. pmid:22232111.
  13. 13.
    Lee Y, Mozaffarian D, Sy S, Liu J, Wilde PE, Marklund M, et al. Health Impact and Cost-Effectiveness of Volume, Tiered, and Absolute Sugar Content Sugar-Sweetened Beverage Tax Policies in the United States. Circulation. 2020;142(6):523–534. pmid:32564614
  14. 14.
    Long MW, Gortmaker SL, Ward ZJ, Resch SC, Moodie ML, Sacks G, et al. Cost Effectiveness of a Sugar-Sweetened Beverage Excise Tax in the U.S. Am J Prev Med. 2015;49(1):112–23. Epub 2015/06/22. pmid:26094232.
  15. 15.
    National Clinical Care Commission. Report to Congress on Leveraging Federal Programs to Prevent and Control Diabetes and its Complications. US Department of Health and Human Services; 2021.
  16. 16.
    Leider J, Li Y, Powell LM. Pass-through of the Oakland, California, sugar-sweetened beverage tax in food stores two years post-implementation: A difference-in-differences study. PLoS ONE. 2021;16(1):e0244884. pmid:33395444
  17. 17.
    Cawley J, Frisvold D, Hill A, Jones D. Oakland’s sugar-sweetened beverage tax: Impacts on prices, purchases and consumption by adults and children. Econ Hum Biol. 2020;37:100865. pmid:32126505
  18. 18.
    Falbe J, Lee MM, Kaplan S, Rojas NA, Ortega Hinojosa AM, Madsen KA. Higher Sugar-Sweetened Beverage Retail Prices After Excise Taxes in Oakland and San Francisco. Am J Public Health. 2020:e1–e7. Epub 2020/05/22. pmid:32437271.
  19. 19.
    Léger PT, Powell LM. The impact of the Oakland SSB tax on prices and volume sold: A study of intended and unintended consequences. Health Econ. 2021;30(8):1745–1771. pmid:33931915
  20. 20.
    Cawley J, Frisvold D, Jones D. The impact of sugar-sweetened beverage taxes on purchases: Evidence from four city-level taxes in the United States. Health Econ. 2020;29(10):1289–1306. pmid:33463850
  21. 21.
    Muth MK, Sweitzer M, Brown D, Capogrossi K, Karns S, Levin D, et al. Understanding IRI household-based and store-based scanner data. United States Department of Agriculture Economic Research Service, 2016 April 2016. Report No.: Technical Bulletin 1942.
  22. 22.
    Levin D, Noriega D, Dicken C, Okrent AM, Harding M, Lovenheim M. Examining food store scanner data: A comparison of the IRI InfoScan data with other data sets, 2008–2012. United States Department of Agriculture Economic Research Service. 2018;TB-1949.
  23. 23.
    The Open Data Initiative [Internet]. 2020 [cited 2020 Dec 9]. Available from: https://www.labelinsight.com/open-data.
  24. 24.
    Coyne KJ, Baldridge AS, Huffman MD, Jenner K, Xavier D, Dunford EK. Differences in the sodium content of bread products in the USA and UK: implications for policy. Public Health Nutr. 2018;21(3):632–636. Epub 2017/11/21. pmid:29157327
  25. 25.
    Roberto CA, Lawman HG, LeVasseur MT, Mitra N, Peterhans A, Herring BJ, et al. 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. JAMA. 2019;321(18):1799–1810. Epub 2019/05/16. pmid:31087022; PubMed Central PMCID: PMC6518342.
  26. 26.
    Dimick JB, Ryan AM. Methods for evaluating changes in health care policy: the difference-in-differences approach. JAMA. 2014;312(22):2401–2402. pmid:25490331.
  27. 27.
    Wing C, Simon K, Bello-Gomez RA. Designing difference in difference studies: best practices for public health policy research. Annu Rev Public Health. 2018;39:453–469. pmid:29328877
  28. 28.
    Abadie A, Diamond A, Hainmueller J. Synthetic control methods for comparative case studies: Estimating the effect of California’s tobacco control program. J Am Stat Assoc. 2010;105(490):493–505.
  29. 29.
    Abadie A. Using Synthetic Controls: Feasibility, Data Requirements, and Methodological Aspects. J Econ Lit. 2021;59(2):391–425.
  30. 30.
    Wooldridge JM. Econometric analysis of cross section and panel data. MIT Press; 2010.
  31. 31.
    Anderson ML. Multiple Inference and Gender Differences in the Effects of Early Intervention: A Reevaluation of the Abecedarian, Perry Preschool, and Early Training Projects. J Am Stat Assoc. 2008;103(484):1481–1495.
  32. 32.
    Seiler S, Tuchman A, Yao S. The Impact of Soda Taxes: Pass-Through, Tax Avoidance, and Nutritional Effects. J Market Res. 2021;58(1):22–49.
  33. 33.
    Bleich SN, Dunn CG, Soto MJ, Yan J, Gibson LA, Lawman HG, et al. Association of a Sweetened Beverage Tax With Purchases of Beverages and High-Sugar Foods at Independent Stores in Philadelphia. JAMA Netw Open. 2021;4(6):e2113527–e. pmid:34129022
  34. 34.
    Gibson LA, Lawman HG, Bleich SN, Yan J, Mitra N, LeVasseur MT, et al. No Evidence of Food or Alcohol Substitution in Response to a Sweetened Beverage Tax. Am J Prev Med. 2021;60(2):e49–e57. pmid:33349471
  35. 35.
    Oddo VM, Leider J, Powell LM. The Impact of Seattle’s Sugar-Sweetened Beverage Tax on Substitution to Sweets and Salty Snacks. J Nutr. 2021;151(10):3232–3239. pmid:34159364
  36. 36.
    Freyaldenhoven S, Hansen C, Pérez JP, Shapiro JM. Visualization, Identification, and Estimation in the Linear Panel Event-Study Design. National Bureau of Economic Research Working Paper Series. 2021;No. 29170.
  37. 37.
    Taylor RLC, Kaplan S, Villas-Boas SB, Jung K. Soda Wars: The Effect of a Soda Tax Election on University Beverage Sales. Econ Inq. 2019;57(3):1480–1496.
  38. 38.
    Freyaldenhoven S, Hansen C, Shapiro JM. Pre-event trends in the panel event-study design. Am Econ Rev. 2019;109(9):3307–3338.
  39. 39.
    Han E, Powell LM. Consumption Patterns of Sugar-Sweetened Beverages in the United States. J Acad Nutr Diet. 2013;113(1):43–53. pmid:23260723
  40. 40.
    Powell ES, Smith-Taillie LP, Popkin BM. Added Sugars Intake Across the Distribution of US Children and Adult Consumers: 1977–2012. J Acad Nutr Diet. 2016;116(10):1543–1550.e1. Epub 2016/08/06. pmid:27492320; PubMed Central PMCID: PMC5039079.
  41. 41.
    Rosinger A, Herrick KA, Gahche JJ, Park S. Sugar-sweetened beverage consumption among U.S. youth, 2011–2014. NCHS Data Brief. 2017;(271). cdc:44039. pmid:28135184
  42. 42.
    Basu S, Jacobs LM, Epel E, Schillinger D, Schmidt L. Cost-effectiveness of a workplace ban on sugar-sweetened beverage sales: A microsimulation model. Health Aff (Millwood). 2020;39(7):1140–1148. pmid:32634357.
  43. 43.
    Visscher TL, Snijder MB, Seidell JC. Epidemiology: Definition and classification of obesity. In: Kopelman PG, Caterson ID, Dietz WH, editors. Clinical obesity in adults and children. Hoboken (NJ): Wiley; 2009. p. 3–14.
  44. 44.
    Sánchez-Romero LM, Penko J, Coxson PG, Fernández A, Mason A, Moran AE, et al. Projected Impact of Mexico’s Sugar-Sweetened Beverage Tax Policy on Diabetes and Cardiovascular Disease: A Modeling Study. PLoS Med. 2016;13(11):e1002158. PubMed Central PMCID: PMC5089730 pmid:27802278
  45. 45.
    Sacco RL, Kasner SE, Broderick JP, Caplan LR, Connors J, Culebras A, et al. An updated definition of stroke for the 21st century: A statement for healthcare professionals from the American Heart Association/American Stroke Association. Stroke. 2013;44(7):2064–2089. pmid:23652265
  46. 46.
    American diabetes Association. 2. Classification and Diagnosis of Diabetes: Standards of Medical Care in Diabetes—2021. Diabetes Care. 2021;44(Supplement 1):S15–S33. pmid:33298413
  47. 47.
    Levey AS, Coresh J, Bolton K, Culleton B, Harvey KS, Ikizler TA, et al. K/DOQI clinical practice guidelines for chronic kidney disease: Evaluation, classification, and stratification. Am J Kidney Dis. 2002;39(2, Suppl 1):S1–S266. pmid:11904577
  48. 48.
    Matsushita K, Mahmoodi BK, Woodward M, Emberson JR, Jafar TH, Jee SH, et al. Comparison of Risk Prediction Using the CKD-EPI Equation and the MDRD Study Equation for Estimated Glomerular Filtration Rate. JAMA. 2012;307(18):1941–1951. pmid:22570462
  49. 49.
    Eke PI, Dye BA, Wei L, Slade GD, Thornton-Evans GO, Borgnakke WS, et al. Update on Prevalence of Periodontitis in Adults in the United States: NHANES 2009 to 2012. J Periodontol. 2015;86(5):611–622. pmid:25688694
  50. 50.
    Institute for Health Metrics and Evaluation. Global Burden of Disease Results Tool. Seattle, Washington: University of Washington; 2021 [cited 2021 Jan 21]. Available from: http://ghdx.healthdata.org/gbd-results-tool.
  51. 51.
    Appelhans BM, French SA, Tangney CC, Powell LM, Wang Y. To what extent do food purchases reflect shoppers’ diet quality and nutrient intake? Int J Behav Nutr Phys Act. 2017;14(1):46. pmid:28399887
  52. 52.
    Einav L, Leibtag ES, Nevo A. On the accuracy of Nielsen Homescan data. U.S. Department of Agriculture, Economic Research Service. Economic Research Report. 2008;No. 69.
  53. 53.
    Neumann PJ, Sanders GD, Russell LB, Siegel JE, Ganiats TG. Cost-effectiveness in health and medicine: Oxford University Press; 2017.
  54. 54.
    Hausman JA. Exact Consumer’s Surplus and Deadweight Loss. Am Econ Rev. 1981;71(4):662–676.
  55. 55.
    Diewert WE. The Measurement of Deadweight Loss Revisited. Econometrica. 1981;49(5):1225–1244.
  56. 56.
    Allcott H, Lockwood BB, Taubinsky D. Regressive sin taxes, with an application to the optimal soda tax. Q J Econ. 2019;134(3):1557–1626.
  57. 57.
    Salomon JA, Haagsma JA, Davis A, de Noordhout CM, Polinder S, Havelaar AH, et al. Disability weights for the Global Burden of Disease 2013 study. Lancet Glob Health. 2015;3(11):e712–e723. pmid:26475018
  58. 58.
    Craig BM, Tseng DS. Cost-effectiveness of gastric bypass for severe obesity. Am J Med. 2002;113(6):491–498. pmid:12427499
  59. 59.
    CPI Inflation Calculator. CPI Inflation Calculator 2021 [cited 2021 Jan 21]. Available from: https://cpiinflationcalculator.com/
  60. 60.
    Husereau D, Drummond M, Petrou S, Carswell C, Moher D, Greenberg D, et al. Consolidated Health Economic Evaluation Reporting Standards (CHEERS)—Explanation and Elaboration: A Report of the ISPOR Health Economic Evaluation Publication Guidelines Good Reporting Practices Task Force. Value Health. 2013;16(2):231–50. pmid:23538175
  61. 61.
    Sanders GD, Neumann PJ, Basu A, Brock DW, Feeny D, Krahn M, et al. Recommendations for Conduct, Methodological Practices, and Reporting of Cost-effectiveness Analyses: Second Panel on Cost-Effectiveness in Health and Medicine. JAMA. 2016;316(10):1093–1103. Epub 2016/09/14. pmid:27623463.
  62. 62.
    California Health Interview Survey. AskCHIS Los Angeles, CA: University of California Los Angeles Center for Helath Policy Research; 2021 [cited 2021 Jan 21]. Available from: http://ask.chis.ucla.edu/.
  63. 63.
    Wilde P, Huang Y, Sy S, Abrahams-Gessel S, Jardim TV, Paarlberg R, et al. Cost-Effectiveness of a US National Sugar-Sweetened Beverage Tax With a Multistakeholder Approach: Who Pays and Who Benefits. Am J Public Health. 2019;109(2):276–284. pmid:30571305.
  64. 64.
    Kahende JW, Loomis BR, Adhikari B, Marshall L. A review of economic evaluations of tobacco control programs. Int J Environ Res Public Health. 2009;6(1):51–68. Epub 2008/12/28. pmid:19440269.
  65. 65.
    Ballinger A, Chowdhury T, Crombie H, Owen L. Modelling the cost-effectiveness of interventions to reduce traffic-related air-pollution. Lancet. 2017;390:S7.
  66. 66.
    Sharifi M, Franz C, Horan CM, Giles CM, Long MW, Ward ZJ, et al. Cost-Effectiveness of a Clinical Childhood Obesity Intervention. Pediatrics. 2017;140(5). pmid:29089403
  67. 67.
    Gortmaker SL, Long MW, Resch SC, Ward ZJ, Cradock AL, Barrett JL, et al. Cost Effectiveness of Childhood Obesity Interventions: Evidence and Methods for CHOICES. Am J Prev Med. 2015;49(1):102–111. pmid:26094231
  68. 68.
    Leider J, Powell LM. Longer-term impacts of the Oakland, California, sugar-sweetened beverage tax on prices and volume sold at two-years post-tax. Soc Sci Med. 2021:114537. pmid:34838326
  69. 69.
    Liu J, Micha R, Li Y, Mozaffarian D. Trends in Food Sources and Diet Quality Among US Children and Adults, 2003–2018. JAMA Netw Open. 2021;4(4):e215262–e. pmid:33844000
  70. 70.
    Powell LM, Leider J. Impact of a sugar-sweetened beverage tax two-year post-tax implementation in Seattle, Washington, United States. J Public Health Policy. 2021;42(4):574–588. pmid:34732842
  71. 71.
    Knox MA, Oddo VM, Walkinshaw LP, Jones-Smith J. Is the public sweet on sugary beverages? Social desirability bias and sweetened beverage taxes. Econ Hum Biol. 2020;38:100886. pmid:32505894
  72. 72.
    Jones-Smith JC, Knox MA, Coe NB, Walkinshaw LP, Schoof J, Hamilton D, et al. Sweetened beverage taxes: Economic benefits and costs according to household income. Food Policy. 2022;110:102277.
  73. 73.
    Williams JD, Goldsworthy P. Breaking Down the Chain: A Guide to the Soft Drink Industry. Public Health Law & Policy; 2011.
  74. 74.
    Crosbie E, Pomeranz JL, Wright KE, Hoeper S, Schmidt L. State Preemption: An Emerging Threat to Local Sugar-Sweetened Beverage Taxation. Am J Public Health. 2021;111(4):677–686. pmid:33600243.
  75. 75.
    Crosbie E, Schillinger D, Schmidt LA. State Preemption to Prevent Local Taxation of Sugar-Sweetened Beverages. JAMA Intern Med. 2019;179(3):291–292. Epub 2019/01/23. pmid:30667481; PubMed Central PMCID: PMC7083242.
  76. 76.
    Hattersley L, Thiebaud A, Silver L, Mandeville K. Countering Common Arguments Against Taxes on Sugary Drinks. The World Bank; 2020.

Source link