Cost-Effectiveness of a US National Sugar-Sweetened Beverage Tax With a Multistakeholder Approach: Who Pays and Who Benefits

Abstract

Objectives. To estimate the health impact and cost-effectiveness of a national penny-per-ounce sugar-sweetened beverage (SSB) tax, overall and with stratified costs and benefits for 9 distinct stakeholder groups.

Methods. We used a validated microsimulation model (CVD PREDICT) to estimate cardiovascular disease reductions, quality-adjusted life years gained, and cost-effectiveness for US adults aged 35 to 85 years, evaluating full and partial consumer price pass-through.

Results. From health care and societal perspectives, the SSB tax was highly cost-saving. When we evaluated health gains, taxes paid, and out-of-pocket health care savings for 6 distinct consumer categories, incremental cost-effectiveness ratios ranged from $20 247 to $42 662 per quality-adjusted life year for 100% price pass-through (incremental cost-effectiveness ratios similar with 50% pass-through). For the beverage industry, net costs were $0.92 billion with 100% pass-through (largely tax-implementation costs) and $49.75 billion with 50% pass-through (largely because of partial industry coverage of the tax). For government, the SSB tax positively affected both tax revenues and health care cost savings.

Conclusions. This stratified analysis improves on unitary approaches, illuminating distinct costs and benefits for stakeholders with political influence over SSB tax decisions.

Sugar-sweetened beverage (SSB) consumption increases the risk of weight gain, obesity, type 2 diabetes, heart disease, and other chronic diseases.^1–4 Despite recent decline, SSB consumption in the United States remains high, with a median intake of 1.1 serving (8 ounces) per day for adults, contributing to an estimated 52 000 cardiometabolic deaths annually.³

Taxation is considered an effective policy tool to curb SSB consumption.^5–7 In 2014, Berkeley, California, became the first US jurisdiction to pass an SSB excise tax (1 cent per ounce); 6 other localities followed suit and approved similar taxes in 2016,⁸ and Seattle, Washington, enacted a 1.75-cent-per-ounce SSB excise tax in June 2017.⁹ Preliminary evaluation from Berkeley indicated a 21% decrease in SSB consumption after tax implementation,¹⁰ with similarly promising results from Philadelphia, Pennsylvania,¹¹ and other countries that have implemented SSB taxes, such as Mexico.¹²

In previous analyses of the potential effectiveness and cost-effectiveness of SSB taxes,^13–16 tax payments in societal and health care perspectives were treated as mere transfers from one stakeholder to another, with no net cost to society. However, both consumers and industry may be financially affected by a tax while government also gains important revenues that are typically not considered. In addition, societal and health care perspectives do not identify the ultimate bearers of those costs. Stakeholder groups each will be very interested in their own costs and effects.¹⁷ Government adoption of SSB taxes depends on multiple factors and stakeholders that affect political support.⁸

Therefore, we analyzed the distinct perspectives of 9 stakeholder groups including 6 consumer categories classified by insurance status, the beverage industry, the government, and other private-sector payers. This allows an understanding of the costs and benefits for each stakeholder, including costs of tax payments for consumers and the beverage industry, tax revenues for the government, and health care costs stratified by the ultimate cost bearers. We performed this investigation as part of the Food Policy Review and Intervention Cost-Effectiveness (Food-PRICE) Study.

METHODS

We used a validated miscrosimulation model, the CVD PREDICT model,¹⁸ to simulate the cardiovascular health and cost consequences of a national SSB tax of $0.01 per ounce. We modeled both a full and partial pass-through to consumer prices. Rationale for these scenarios and the modeling logic pathway are provided in Appendix A, Figure A, and Figure B (available as supplements to the online version of this article at http://www.ajph.org). Based on an average pretax price of $0.059 per ounce,¹⁹ a 100% pass-through increases the price by $0.01 per ounce (a 16.9% increase), as assumed in 2 recent analyses,^13,14 and a 50% pass-through increases the price by $0.005 per ounce (an 8.5% increase) and reduces beverage industry net revenue by $0.005 per ounce.²⁰

Study Population

We investigated 9 distinct stakeholder groups, including 6 categories of adult consumers, classified by health insurance status, the beverage industry, the government, and other private-sector payers (including employers that offer health care benefits). The 6 insurance categories were

1. private (private single-service plan, private plus other government, other coverage),

2. Medicare (Medicare alone, Medicare plus other government, Medicare plus private),

3. Medicaid alone,

4. dual-eligible (Medicare plus Medicaid),

5. other government (other government, state sponsored), and

6. no coverage.

We obtained insurance status, baseline risk factor, and SSB intake data from the National Health and Nutrition Examination Survey (NHANES) 2005 through 2012 cycles; further details are provided in Appendix E and Table B (available as supplements to the online version of this article at http://www.ajph.org). We defined SSBs as any beverage with added caloric sweeteners (e.g., soft drinks, sports and energy drinks, fruit drinks, and sweetened tea) and excluding 100% juice, milk, and diet beverages, consistent with our previous work.³

Policy Effects

We derived the effects of price changes on SSB intakes from a nationally representative US analysis of the association between SSB prices and intake.²¹ Average own-price elasticity for SSBs was estimated at −0.66 (0.66% decline in intake in response to each 1% increase in price), in agreement with a recent meta-analysis of prospective observational and interventional studies (mean elasticity for SSBs of −0.67).²² In sensitivity analyses (Appendix B and Table C, available as supplements to the online version of this article at http://www.ajph.org), we considered differential elasticities for low-income (defined as income-to-poverty ratio < 1.85) and higher-income consumers (income-to-poverty ratio ≥ 1.85).²¹ We did not model specific replacement scenarios to water, diet soda, juice, and milk, because we expected the net substitution effects to be small, and because explicitly modeling the substitution would be double-counting in cases in which the relative risk input coefficients already accounted for substitution (Appendix B).

We derived the evidence and magnitude of SSBs’ etiologic effect on cardiometabolic outcomes from meta-analyses of prospective cohorts.^3,23 Evidence suggests that SSBs are associated with increased cardiometabolic risk through relationships with body mass index plus additional body mass index–independent relationships of SSBs with coronary heart disease, stroke, and diabetes (Table A, available as a supplement to the online version of this article at http://www.ajph.org).

Model Structure

The CVD PREDICT microsimulation model projects the lifetime health outcomes and cardiovascular disease (CVD)–related costs through updates of individual-level CVD history (Figure C, available as a supplement to the online version of this article at http://www.ajph.org). The model is populated by weighted sampling with replacement over time of US adults aged 35 to 85 years from NHANES to create a representative synthetic population of 1 000 000 adults in each consumer category. Risk factors included sex, age, systolic blood pressure, total cholesterol, high-density lipoprotein cholesterol, smoking, and diabetes status.

For each individual, established CVD risk factors predict annual risks of CVD events (coronary heart disease including cardiac arrest, myocardial infarction, angina, and stroke) and their sequelae including acute (i.e., from 30 days up to 1 year) and postacute (i.e., all subsequent years) mortality, morbidity, and health care costs. For each relative change in SSB intake induced by the tax policy, the model generated predicted changes in the probability of each health outcome at the individual level. We aggregated these changes to provide mean and total estimates for the entire population and for each consumer category. Further details are available in Appendix C (available as a supplement to the online version of this article at http://www.ajph.org), and a comprehensive description of the model and validation performance has been published elsewhere.¹⁸ All individuals were simulated in 1-year cycles until age 100 years or death. We conducted probabilistic sensitivity analysis to assess model uncertainty and the robustness of our cost-effectiveness analysis results. In the probabilistic sensitivity analysis, we drew 1000 random values for key parameters from a prespecified probability distribution (Table C, available as a supplement to the online version of this article at http://www.ajph.org).

Health Care Costs

We calculated each individual’s total health care costs with and without the tax intervention, incorporating any reductions in acute, chronic, and drug costs associated with CVD. We summed individual costs for national and stakeholder-specific costs. For the latter, we estimated different health care costs by using data from the Medical Expenditure Panel Survey and the Centers for Medicare and Medicaid Services according to their ultimate cost bearers including (1) consumers (e.g., out-of-pocket costs, some insurance premiums), (2) government (e.g., costs through public health insurance), and (3) the private sector (e.g., employers that subsidize insurance premiums, hospitals that absorb unreimbursed medical spending, charitable entities). Full methods on health care saving attribution are presented in Appendix F (available as a supplement to the online version of this article at http://www.ajph.org).

Cost-Effectiveness Analysis

We conducted cost-effectiveness analysis from multiple perspectives including (1) health care sector, (2) societal, and (3) distinct perspectives of the 9 stakeholder groups. We followed recommendations from the Second Panel on Cost-Effectiveness in Health and Medicine¹⁷ with adaptations because of distinctive features of this tax policy application. For the health care–sector perspective, we excluded tax-implementation costs as outside of the sector. For the societal perspective, we included tax-implementation costs but excluded tax payments as mere transfers (with net zero impact on society).¹⁷ We assumed the implementation costs, including government tax-collection costs and industry compliance costs, to be 2% of SSB tax payments (details in Appendix D, available as a supplement to the online version of this article at http://www.ajph.org). We based tax payments on posttax SSB intake for simulated individuals. Notably, all perspectives excluded indirect health care savings (e.g., productivity gains). This exclusion underestimates cost-effectiveness (or cost-savings) of the intervention for all perspectives, but allows parallel comparisons of the different stakeholder groups, a key objective of this study. We discounted costs and quality-adjusted life years (QALYs) at 3% annual rate.¹⁷ We reported all costs in constant 2013 US dollars.

We calculated incremental cost-effectiveness ratios (ICERs) as the net monetary cost divided by QALYs gained. When net monetary costs were positive, we compared ICERs with a willingness-to-pay threshold of $50 000 to $150 000 per QALY as recommended by the American College of Cardiology and American Heart Association.²⁴ In the cost-effectiveness analysis by distinct stakeholders, we attributed QALYs gained to the 6 consumer categories as the groups most rationally interested in these gains. Net monetary costs included SSB tax payments and the consumer portion of health care savings (for consumers), tax revenue (for the government), insurance premium costs (for employers in the private sector), and implementation costs and possible revenue losses (for the beverage industry). In an additional analysis, assuming a linear supply function, we also computed the net present value of the reduction in producer surplus attributable to the tax (see Appendix B, available as a supplement to the online version of this article at http://www.ajph.org, for details). Given the potential for offsetting sales in diet soda and bottled water,^10,25 we did not attempt to further model the beverage industry’s profit losses attributable to lower SSB sales.

RESULTS

Baseline mean pretax SSB intake was 1.1 (SD = 1.6) servings per day (Table 1). As expected, those with private insurance (57% of adults) had the highest income, with close to average SSB intakes. Adults with no insurance coverage had below-average income, youngest mean age, and highest SSB intake.

TABLE 1—

Baseline Demographics and Health Characteristics Among US Adults, Overall and by 6 Consumer Categories: National Health and Nutrition Examination Survey, United States, 2005–2012

		6 Consumer Categories by Insurance Statusa
Characteristic	Overall, % or Mean ±SD	Private, % or Mean ±SD	Medicare, % or Mean ±SD	Medicaid, % or Mean ±SD	Dual-Eligible, % or Mean ±SD	Other Government, % or Mean ±SD	No Coverage, % or Mean ±SD
Population, millionsb	166.5	95.6	32.5	4.8	2.2	7.7	23.7
Percentage of populationb	100.0	57.44	19.54	2.86	1.30	4.64	14.22
Demographics
Age, y	54.7 ±12.9	50.8 ±10.1	71.6 ±8.3	48.8 ±10.6	63.2 ±12.0	51.8 ±10.5	48.3 ±9.2
Female	52.78	52.26	54.85	65.15	63.50	52.79	48.53
White	74.36	79.4	84.43	38.06	56.17	65.83	51.94
African American	9.8	7.96	7.62	29.78	24.69	12.09	14.07
Hispanic	10.52	7.29	5.13	22.37	15.05	13.36	27.25
Other race/ethnicity	5.32	5.36	2.82	9.79	4.09	8.73	6.74
Poverty–income ratioc	3.2 ±1.6	3.9 ±1.4	2.8 ±1.5	1.0 ±0.9	1.2 ±0.8	3.0 ±1.7	1.9 ±1.4
SSB intake (8-oz serving/d)	1.1 ±1.6	1.1 ±1.6	0.7 ±1.3	1.5 ±2.1	1.4 ±2.0	1.1 ±1.7	1.6 ±1.9
Health at baseline
Current smoker	17.6	14.6	11.1	41.5	23.2	22.4	32.0
Current hypertension treatment	34.4	27.6	58.1	46.8	62.8	40.0	22.8
Prevalent diabetes	10.8	7.6	18.7	19.4	31.3	13.1	8.7
Angina	2.8	1.3	6.9	3.7	9.1	2.7	2.3
MI	4.3	2.0	11.1	6.7	12.4	4.2	3.0
Stroke	3.6	1.6	8.8	7.4	19.0	3.8	2.1
CVD-freed	91.6	95.9	79.4	85.6	69.4	91.7	93.7
SBP, mm Hg	124.4 ±17.8	121.7 ±15.5	132.6 ±20.7	122.5 ±17.2	132.0 ±22.3	123.2 ±17.5	123.9 ±18.6
DBP, mm Hg	69.9 ±13.6	70.3 ±13.3	67.2 ±14.3	70.3 ±13.2	70.2 ±13.3	71.1 ±13.0	71.5 ±13.4
BMI, kg/m2	29.1 ±6.5	29.0 ±6.4	28.7 ±6.2	31.4 ±8.1	31.9 ±7.1	30.0 ±6.8	29.2 ±6.5
Total cholesterol, mg/dL	203.0 ±41.9	205.0 ±41.0	193.9 ±41.8	200.2 ±46.2	193.8 ±43.8	201.2 ±39.2	208.9 ±43.1
HDL, mg/dL	53.9 ±16.6	54.5 ±16.9	55.1 ±16.4	52.1 ±16.7	53.9 ±17.0	52.3 ±15.9	50.7 ±15.4
LDL, mg/dL	119.0 ±36.4	120.4 ±35.1	110.6 ±36.6	118.8 ±37.3	114.3 ±41.0	115.6 ±33.9	128.0 ±38.3
Triglycerides, mg/dL	141.3 ±110.4	139.2 ±117.5	137.3 ±78.8	146.1 ±156.2	145.0 ±109.8	144.1 ±120.2	153.2 ±105.1

Note. BMI = body-mass index; CVD = cardiovascular disease; DBP = diastolic blood pressure; HDL = high-density lipoprotein; LDL = low-density lipoprotein; MI =  myocardial infarction; mm Hg = millimeters of mercury; SBP = systolic blood pressure; SSB = sugar-sweetened beverage.

^aConsumers categorized by insurance status: private = private + (single service plan) + (private + other government) + (other coverage); Medicare = Medicare + (Medicare + other government) + (private + Medicare); dual-eligible = Medigap + (Medicare + Medicaid); other government = other government + (state sponsored). See Appendix E and Table B, available as supplements to the online version of this article at http://www.ajph.org, for details.

^bPopulation percentages by consumer category from CVD PREDICT based on National Health and Nutrition Examination Survey (2005–2012). Total population from US Census Bureau for 2014.

^cPoverty status calculated based on annual Department of Health and Human Services’ poverty guidelines for a household of 4 persons (2005–2006 = $19 971; 2007–2008 = $21 203; 2009–2010 = $21 954; 2011–2012 = $23 021). No information on poverty-to-income ratio for all (n = 94 558).

^dCVD-free indicates absence of angina, MI, or stroke.

Health Outcomes

With 100% price pass-through, we estimated the SSB tax to prevent 4494 (95% uncertainty interval [UI] = 2640, 6599) lifetime myocardial infarction events (2.06% reduction from the no-tax case) and 1540 (95% UI = 995, 2118) lifetime total ischemic heart disease deaths per million adults (1.42% reduction from the no-tax case; Table 2 for reductions, and Table D, available as a supplement to the online version of this article at http://www.ajph.org, for complete counts with and without tax). With 50% price pass-through, changes in health outcomes were approximately half as large. When evaluated by consumer category, health impacts per million adults were largest for those with no coverage (8375 [95% UI = 4905, 12 083] MI events prevented) and Medicaid (7714 [95% UI = 4407, 11 105] MI events prevented) consumers, who each also had highest baseline SSB intakes. Health impacts were smallest for the Medicare group, which had lower mean SSB intake and higher starting age.

TABLE 2—

Lifetime Changes in Health Outcomes Among US Adults Attributable to a National Sugar-Sweetened Beverage Tax, Overall and for 6 Consumer Categories

		6 Consumer Categories by Insurance Statusa
	Overall	Private	Medicare	Medicaid	Dual-Eligible	Other Government	No Coverage
100% price pass-through to consumers (95% UI)
Reduction in no. MI events, per million adults	4 494 (2 640, 6 599)	4 679 (2 802, 6 560)	1 820 (1 063, 2 595)	7 714 (4 407, 11 105)	6 196 (3 722, 8 733)	4 773 (2 817, 6 821)	8 375 (4 905, 12 083)
Reduction in no. stroke events,b per million adults	60 (−81, 181)	233 (103, 348)	21 (−36, 94)	139 (−83, 315)	65 (−157, 369)	−411 (−566, −298)	−379 (−619, −203)
Reduction in total stroke deaths,b per million adults	−70 (−28, 127)	−163 (114, 222)	−145 (124, 168)	118 (14, 205)	−221 (−189, −256)	−129 (−210, −56)	−343 (−452, −227)
Reduction in total IHD deaths, per million adults	1 540 (995, 2 118)	1 232 (629, 1 866)	579 (360, 803)	1 617 (831, 2 472)	907 (553, 1 302)	1 384 (820, 1 966)	2 200 (1 237, 3 236)
Increase in life expectancy, y	0.04 (0.02, 0.05)	0.04 (0.02, 0.05)	0.01 (0.01, 0.02)	0.05 (0.03, 0.08)	0.02 (0.01, 0.03)	0.03 (0.02, 0.05)	0.07 (0.04, 0.1)
50% price pass-through to consumers (95% UI)
Reduction in no. MI events, per million adults	2 391 (1 385, 3 333)	2 564 (1 598, 3 527)	981 (598, 1 361)	3 994 (2 353, 5 654)	3 452 (2 145, 4 700)	2 543 (1 553, 3 544)	4 397 (2 505, 6 183)
Reduction in no. stroke events,b per million adults	74 (7, 134)	230 (167, 304)	13 (−19, 46)	126 (0, 251)	−90 (−207, 28)	−372 (−442, −316)	−291 (−422, −193)
Reduction in total stroke deaths,b per million adults	−35 (−15, −60)	−119 (94, 144)	−125 (112, 142)	212 (156, 266)	−209 (−237, −185)	−49 (−95, 3)	−226 (−291, −177)
Reduction in total IHD deaths, per million adults	913 (637, 1 190)	542 (239, 860)	329 (214, 442)	739 (313, 1 131)	500 (329, 681)	743 (455, 1 039)	1 136 (702, 1 617)
Increase in life expectancy, y	0.02 (0.01, 0.03)	0.02 (0.01, 0.03)	0.01 (0.00, 0.01)	0.03 (0.01, 0.04)	0.01 (0.01, 0.02)	0.02 (0.01, 0.02)	0.04 (0.02, 0.05)

Notes. IHD = ischemic heart disease; MI = myocardial infarction; SSB = sugar-sweetened beverage; UI = uncertainty interval. Microsimulation results from the CVD PREDICT model. See Table D, available as a supplement to the online version of this article at http://www.ajph.org, for pre- and posttax health outcomes and sensitivity analyses showing corresponding results.

^aConsumers categorized by insurance status: private = private + (single service plan) + (private + other government) + (other coverage); Medicare = Medicare + (Medicare + other government) + (private + Medicare); dual-eligible = Medigap + (Medicare + Medicaid); other government = other government + (state sponsored). See Appendix E and Table B, available as supplements to the online version of this article at http://www.ajph.org, for details.

^bIn some cases, impacts on stroke were negative in a competing risk framework, with reductions in MI events leading to a corresponding increase in the number of subsequent strokes.

Cost-Effectiveness Perspectives

Health care sector.

The mean survival in the no-tax scenario was 83.1 years. With full price pass-through, the average adult would gain 0.020 lifetime QALYs (Table 3 for reductions, and Table E, available as a supplement to the online version of this article at http://www.ajph.org, for complete counts with and without tax), corresponding to 3.4 million QALYs across the population. Total lifetime health care costs would be reduced by $270 per person (a savings of approximately 1.2% of all CVD medical costs, relative to medical costs without an SSB tax), or $45.00 billion nationally, largely attributable to reduced costs for acute events and chronic disease (Table 3). When stratified by insurance status, the per-adult discounted lifetime savings in health care costs were highest for Americans with no coverage ($499) followed by Medicaid recipients ($447), and lowest for those receiving Medicare ($117). With 50% price pass-through, both the health gains and reductions in cost were approximately half as large. The intervention was cost-saving from the overall health care sector perspective and each insurance category of the health care sector.

TABLE 3—

Lifetime Cost-Effectiveness Among US Adults of a National Sugar-Sweetened Beverage Tax, From the Health Care Perspective

		6 Consumer Categories by Insurance Statusa
	Health Care Sector Overall	Private	Medicare	Medicaid	Dual-Eligible	Other Government	No Coverage
100% price pass-through to consumers (95% UI)
Total health care cost saving, ($) per personb	270 (158, 383)	257 (144, 371)	117 (65, 170)	447 (258, 637)	315 (184, 448)	282 (164, 398)	499 (286, 713)
QALYs gained, per person	0.0201 (0.012, 0.0286)	0.0190 (0.0104, 0.0282)	0.0081 (0.0048, 0.0114)	0.0273 (0.0150, 0.0401)	0.0127 (0.0072, 0.0187)	0.0182 (0.0098, 0.0270)	0.0371 (0.0220, 0.0533)
ICER	Cost-saving	Cost-saving	Cost-saving	Cost-saving	Cost-saving	Cost-saving	Cost-saving
50% price pass-through to consumers (95% UI)
Total health care cost saving, ($) per personb	145 (86, 204)	130 (73, 190)	59 (34, 86)	236 (145, 335)	170 (103, 237)	149 (90, 213)	262 (153, 374)
QALYs gained, per person	0.0110 (0.0066, 0.0153)	0.0094 (0.0048, 0.0138)	0.0044 (0.0028, 0.0061)	0.0138 (0.0076, 0.0198)	0.0064 (0.0035, 0.0093)	0.0089 (0.0045, 0.0131)	0.0201 (0.0126, 0.0277)
ICER	Cost-saving	Cost-saving	Cost-saving	Cost-saving	Cost-saving	Cost-saving	Cost-saving

Notes. ICER = incremental cost-effectiveness ratio; QALY = quality-adjusted life year; UI = uncertainty interval. All values are per person, with results discounted at 3% annually and presented in 2013 dollars. Microsimulation results from the CVD PREDICT model. See Table E (available as a supplement to the online version of this article at http://www.ajph.org) for pre- and posttax health care costs, and sensitivity analyses showing corresponding results. See Figure D (available as a supplement to the online version of this article at http://www.ajph.org) for per-person health care cost saving by consumer category and ultimate-cost bearers. Health care saving is scaled to total population in Table H (available as a supplement to the online version of this article at http://www.ajph.org).

^aConsumers categorized by insurance status: private = private + (single service plan) + (private + other government) + (other coverage); Medicare = Medicare + (Medicare + other government) + (private + Medicare); dual-eligible = Medigap + (Medicare + Medicaid); other government = other government + (state sponsored). See Appendix E and Table B, available as supplements to the online version of this article at http://www.ajph.org, for details.

^bTotal health care cost savings regardless of ultimate cost bearer. Costs include drug costs, acute costs (costs related to the acute hospitalization event), and chronic costs (costs not related to the acute event or drugs).

Societal.

From the societal perspective, the discounted lifetime implementation costs of $1.84 billion (divided between government tax collection costs and beverage industry compliance costs) were dwarfed by the health care cost savings of $45.00 billion, making the intervention highly cost-saving (Table 4). When evaluated over time, the SSB tax was cost-saving after just 1 year of implementation (not shown). In the societal perspective, the $91.90 billion in tax revenue was considered a transfer and not incorporated. With 50% pass-through, health care cost savings were approximately half as large, corresponding to approximately half the reduction in SSB intakes.

TABLE 4—

Lifetime Cost-Effectiveness Among US Adults of a National Sugar-Sweetened Beverage Tax, Overall Societal Perspective, and for 9 Distinct Stakeholder Groups

		6 Consumer Categories by Insurance Statusa							Private Sector
	Societal Perspective Overall	Private	Medicare	Medicaid	Dual-Eligible	Other Government	No Coverage	Government	Private Sectorb	Beverage Industry
Population, millions	166.52	95.65	32.54	4.76	2.16	7.73	23.68
Population, %		57.44	19.54	2.86	1.30	4.64	14.22
100% price pass-through to consumers (95% UI)
Costs
Tax payments, $ billionc	0.00 (0.00, 0.00)	53.37 (50.29, 56.18)	7.97 (7.51, 8.39)	3.97 (3.74, 4.18)	1.20 (1.13, 1.26)	4.46 (4.21, 4.70)	20.92 (19.71, 22.02)	−91.90 (−96,73, 86.60)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)
Implementation costs, $ billiond	1.84 (1.73, 1.93)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	0.92 (0.87, 0.97)	0.00 (0.00, 0.00)	0.92 (0.87, 0.97)
Health care cost saving, $ billione	45.00 (25.43, 65.04)	9.39 (5.26, 13.56)	0.68 (0.38, 1.00)	0.11 (0.06, 0.15)	0.02 (0.01, 0.03)	0.83 (0.48, 1.17)	2.86 (1.64, 4.08)	15.58 (8.86, 22.59)	15.60 (8.77, 22.46)	0.00f (0.00, 0.00)
Net monetary cost, $ billiong	−43.16 (−63.32, 23.51)	43.98 (36.66, 50.97)	7.29 (6.53, 8.01)	3.86 (3.59, 4.12)	1.18 (1.11, 1.25)	3.63 (3.02, 4.21)	18.06 (15.59, 20.37)	−106.56 (−109.10, 104.50)	−15.60 (−8.77, 22.46)	0.92 (0.87, 0.97)
Aggregate QALYs saved, millionh	3.40 (1.85, 4.77)	1.82 (0.99, 2.70)	0.26 (0.16, 0.37)	0.13 (0.07, 0.19)	0.03 (0.02, 0.04)	0.14 (0.08, 0.21)	0.88 (0.52, 1.26)	NA	NA	NA
Net monetary cost/QALY ratio, $/QALY	Saving	26 265 (13 627, 51 642)	29 431 (17 470, 51 279)	31 878 (18 710, 57 379)	46 133 (27 451, 80 471)	28 124 (14 464, 55 723)	21 955 (12 347, 39 385)	NA	NA	NA
50% price pass-through to consumers (95% UI)
Costs
Tax payments, $ billionc	0.00 (0.00, 0.00)	28.31 (27.61, 29.02)	4.23 (4.12, 4.34)	2.10 (2.05, 2.16)	0.63 (0.37, 0.90)	2.37 (2.31, 2.43)	11.09 (10.82, 11.37)	−97.48 (−99.94, 95.08)	0.00	48.74 (47.54, 49.97)
Implementation costs, $ billiond	1.95 (1.90, 2.00)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	0.00 (0.00)	0.97 (0.95, 1.00)	0.00	0.97 (0.95, 1.00)
Health care cost saving, $ billione	23.05 (13.10, 33.59)	4.75 (2.66, 6.95)	0.35 (0.20, 0.50)	0.06 (0.03, 0.08)	0.01 (0.01, 0.01)	0.44 (0.27, 0.63)	1.50 (0.88, 2.14)	7.99 (4.67, 11.70)	8.01 (4.48, 11.59)	0.00f (0.00, 0.00)
Net monetary cost, $ billiong	−21.10 (−31.69, −11.11)	23.56 (20.65, 26.38)	3.88 (3.62, 4.14)	2.05 (1.97, 2.12)	0.63 (0.61, 0.65)	1.93 (1.68, 2.16)	9.59 (8.69, 10.49)	−104.50 (−106.04, −103.46)	−8.01 (−11.59, 4.48)	49.72i (48.49, 50.97)
Aggregate QALYs saved, millionh	1.81 (0.93, 2.40)	0.90 (0.46, 1.32)	0.14 (0.09, 0.20)	0.07 (0.04, 0.09)	0.01 (0.01, 0.01)	0.07 (0.03, 0.10)	0.48 (0.30, 0.66)	NA	NA	NA
Net monetary cost/QALY ratio, $/QALY	Saving	29 071 (15 759, 57211)	28 204 (18 403, 45 375)	33 357 (20 915, 58 459)	48 570 (30 258, 86 012)	31 070 (16 653, 62 057)	21 221 (13 332, 35 368)	NA	NA	NA

Notes. QALY = quality-adjusted life year; UI = uncertainty interval. Lifetime cost-effectiveness results are discounted at 3% annually and presented in 2013 dollars. Microsimulation results from the CVD PREDICT model. See introduction and Methods texts about rationale and methods to categorize 9 stakeholder groups. See Table J (available as a supplement to the online version of this article at http://www.ajph.org) for per-person results.

^aConsumers categorized by insurance status: private = private + (single service plan) + (private + other government) + (other coverage); Medicare = Medicare + (Medicare + other government) + (private + Medicare); dual-eligible = Medigap + (Medicare + Medicaid); other government = other government + (state sponsored). See Appendix E and Table B, available as supplements to the online version of this article at http://www.ajph.org, for details.

^bIncludes employers (e.g., employer-sponsored health insurance, worksite health, workers’ compensation), hospitals, nonprofits, and health-related philanthropic support, and excludes beverage industry private sector, which is evaluated separately.

^cTax payments equal per-person lifetime sugar-sweetened beverage intake multiplied by the tax rate ($0.01/oz) and by the population. For 50% price pass-through, industry pays half the tax. For the government, the negative sign indicates revenue rather than payments.

^dEquals 2% of tax payments, divided equally between government (tax collection cost) and beverage industry (compliance cost).

^eFor each of the 9 stakeholder groups, these are the direct health care cost savings borne by each group. For example, for the consumer categories, these include personal premiums plus out-of-pocket costs; for the government, these include costs of Medicare and Medicaid, and for the private sector, these include the costs of company-sponsored premiums. See Table H (available as a supplement to the online version of this article at http://www.ajph.org) for more details on separation of health care costs by each stakeholder group.

^fEmployment in the beverage industry (North America Industry Classification System 3121) is a tiny fraction of total US nonfarm employment (less than one fifth of 1% in 2016); therefore, its health care saving is small enough to be negligible.

^gFor net monetary cost, a negative sign indicates positive net monetary savings. In this cost-effectiveness analysis, these costs included only actual monetary costs (not assigning an assumed dollar value to QALYs as in cost–benefit analysis).

^hFor each of 6 consumer categories, aggregate QALYs equal per-person QALYs gained (Table 3) multiplied by the population. Aggregate QALYs are attributed to consumer categories only as the groups most rationally interested in these gains. Thus, we did not assign QALYs or compute a traditional incremental cost-effectiveness ratio for the government, other private sector, or beverage industry perspectives.

ⁱThe total loss of producer surplus attributable to the tax includes this $48.76 billion in effective tax payments plus approximately $1.41 billion in producer surplus losses attributable to the smaller quantity sold (see Appendix B, available as a supplement to the online version of this article at http://www.ajph.org, for further details).

Cost-Effectiveness by 9 Stakeholder Groups

From the perspective of different stakeholder groups, a national SSB tax would have variable effects on costs, health, and health care savings (Table 4).

Six consumer categories.

Among the different consumer categories, tax payments, health gains, and health care cost savings were distributed according to baseline SSB intakes, other risk factors, disease rates, and the proportion of health care costs paid by consumers themselves. We estimated ICERs in net dollars of cost per QALY saved. The intervention was not cost-saving for individual consumer groups (as in the societal perspective) but was cost-effective for each consumer group, with ICERs less than $50 000 per QALY in all groups. The ICERs ranged from a low of $21 955 per QALY for the no-coverage group to $31 878 per QALY for the Medicaid group and $46 133 per QALY for the dual-eligible group. The cost per QALY for consumers was generally similar with either 100% or 50% price pass-through, as these 2 scenarios produced proportional tax costs and QALY gains, but the overall health gains were about twice as high with 100% pass-through.

Government.

The SSB tax generated substantial tax revenue and health care cost savings for the government, far exceeding tax-collection costs, with a total net savings of $106.56 billion (Table 4). The government paid varying percentages of total medical costs, ranging from 23.8% for the privately insured to 95.0% for the Medicaid category (Table G, available as a supplement to the online version of this article at http://www.ajph.org). Although the government perspective could incorporate QALYs saved in the population, we did not include QALYs gained as a government benefit nor report net costs per QALY under the government perspective to avoid double-counting QALYs, as health gains were already attributed to the 6 categories of consumers in our multistakeholder perspective. With 50% price pass-through, the government’s tax revenue was slightly larger (because the tax base of SSB spending was slightly larger), while health care cost savings were approximately half as large, with a net result of similar overall cost savings to the government of $104.50 billon (Table 4).

Private sector.

From the perspective of the private sector (principally employers, excluding beverage industry), the main impact of the SSB tax was health care cost savings (including the employer portion of health insurance premiums) equal to $15.60 billion with 100% pass-through.

Beverage industry.

For the beverage industry, costs varied substantially depending on the extent to which the excise tax was passed on to consumers in the form of higher prices. With 100% pass-through, the lifetime cost for the beverage industry was $0.92 billion (reflecting the industry’s burden of complying with the tax). With 50% pass-through, the producers also experienced costs attributable to their partial absorbing of the tax payments. In this scenario, the net lifetime costs of these tax payments were $48.74 billion, far larger than the $0.97 billion in implementation costs. In the 50% pass-through scenario, the total loss of producer surplus attributable to the tax includes this $48.74 billion in effective tax payments plus approximately $1.41 billion in producer surplus losses because of the smaller quantity sold (see Appendix B, available as a supplement to the online version of this article at http://www.ajph.org, for further details).

Sensitivity Analysis

We reported the results of probabilistic sensitivity analysis in the 95% UI in Tables 2 through 4. In addition, assuming a differential price elasticity showing a greater price responsiveness (–1.025) among lower-income adults and lower price responsiveness (–0.505) among the higher-income adults led to moderately higher health care cost savings for the Medicaid and dual-eligible populations (Tables E and I, available as supplements to the online version of this article at http://www.ajph.org). As a consequence, in this sensitivity analysis, these 2 consumer categories had ICERs that were similar to the other categories.

DISCUSSION

From both health care and societal perspectives, the policy resulted in substantial health gains and was also cost-saving. In the societal perspective, the health care savings were more than 24 times the tax-implementation costs. These results are consistent with previous US and international modeling studies in finding SSB taxes to be cost-saving.^{6,13–16,26,27} For instance, Long et al.¹³ estimated that a national SSB tax of $0.01 per ounce would cost the United States $51 million to implement in the first year, but generate $23.6 billion in 10-year obesity-related health care savings. Wang et al.¹⁴ reported $20.1 billion in 10-year potential medical savings when assuming no substitution effect and a slightly smaller $17 billion in savings when assuming that 40% of calories reduced from tax would be compensated by other beverages. A microsimulation analysis by Basu et al. estimated that a tax on SSB purchases with Supplemental Nutrition Assistance Program benefits would save 31 000 deaths over 10 years, and would be cost-saving from a government revenue perspective (counting the tax payments as positive revenue).²⁸

Our findings demonstrated that conventional cost-effectiveness analysis perspectives ignore large intrasocietal transfers because of the tax itself: $91.90 billion in taxes (or $3 billion per year), which matter substantially for political decision-making in a democracy that must respond to the views of different stakeholders. Consumers must pay the SSB tax, yet they also save in health care costs and enjoy the better health and lower mortality reflected in QALYs saved. From their stakeholder-specific perspective, the SSB tax policy was not cost-saving, but was still shown to have favorable ICERs for all consumer groups, indicating that the costs of an SSB tax for cardiovascular health gains are comparable to other medical “best buys” that consumers currently pay through individual premium and out-of-pocket health care costs. In addition, we found that although lower-income consumers (e.g., Medicaid, no-coverage) had moderately higher tax payments partly because of their higher consumption, their health gains and overall health care cost reductions were correspondingly higher. Notably, the downstream savings to some lower-income consumers were proportionally higher (e.g., no-coverage) or lower (e.g., Medicaid), depending on the fraction of health care costs borne by these consumers themselves. These findings echo recent evidence that SSB taxes are not necessarily regressive.²⁹

We identified substantial cost savings for both the government and overall US private sector, with financial gains of $106.56 billion and $15.60 billion, respectively. In an era of constrained financial resources, these estimated gains from both tax revenue and health care savings represent important findings. In addition, if all or some of this revenue to government is earmarked for public health programs or returned to consumers in the form of lower taxes or greater government services, then the benefits to consumers could exceed the health care cost savings alone.

Finally, we demonstrated substantial differences in costs to the beverage industry depending on the extent to which the excise tax is passed on to consumers. Net industry costs for implementation were relatively modest ($0.92 billion lifetime), yet jumped to $49.72 billion if industry elected to absorb half of the tax themselves. Empirical evidence on the SSB tax’s pass-through rate is inconsistent, and the rate varies by timing, store, and beverage type. Whereas Cawley and Frisvold³⁰ detected 43.1% pass-through in the Berkeley tax 6 months after implementation, others reported more complete tax pass-through in Berkeley 1 year after implementation,²⁵ Mexico,³¹ and France.³²

Limitations

Our analysis has several potential limitations. We were conservative in our estimates of the potential policy benefits. We modeled health gains for CVD, but not other conditions that may be attributable to SSB intake (e.g., dental caries, obesity-related cancers, gallstones). Furthermore, we did not include savings attributable to increased productivity and other indirect costs of CVD events. According to other analyses, such additional savings may be 2-fold higher than the direct health care cost savings.³³ When medical costs are attributed to stakeholders, the percentage distribution of expenses for all health services may differ from the percentage distribution of expenses for services related to cardiovascular events. We modeled the tax’s effects on adults, not children and adolescents whose health outcomes may occur decades later.

We assumed that changes to SSB intake altered relative risks within 1 year, and time lags could be longer. We incorporated the effect on industry of tax compliance and partial tax pass-through, but did not attempt to model the industry profit function or count the reduction in quantity sold per se as a loss to the beverage industry. Silver et al.²⁵ found that while SSB sales declined by 9.6% in Berkeley stores 1 year after tax implementation, overall store revenue did not fall more in Berkeley than in control cities. Powell et al.³⁴ demonstrated possible beverage industry job losses from an SSB tax yet highlighted potentially offsetting employment increase in the nonbeverage sector and government. Lastly, we did not model specific replacement scenarios; yet, by choosing an elasticity estimate closely matching the empirical evidence emerging from Berkeley and other SSB tax evaluations and modeling a 50% pass-through scenario, it is unlikely that we overestimated potential benefits.

Public Health Implications

An SSB tax is one policy strategy for reducing consumption of caloric beverages associated with increased risk of obesity and chronic disease. As noted by the 2015 Dietary Guidelines Advisory Committee, other such strategies include nutrition education, changes to food marketing practices, and improvements in the healthfulness of the food retail environment, taking account of the social and cultural settings in which people live.³⁵ We replicated previous research finding that a national penny-per-ounce SSB tax is highly cost-effective and developed a new method for separately analyzing these costs and effects by stakeholder group. We found that tax payments and health care cost savings varied across consumer categories and that tax payments were central to the overall economic effects of the tax for the government, the beverage industry, and other private-sector entities. For decision-making by a benevolent sovereign maximizing the interests of society at large, it would be understandable to exclude taxes as mere transfers. By contrast, for analysis of public health policy under democratic decision-making, it is essential to understand costs and effects for distinct stakeholder groups that have influence over legislative decisions.