State and Local Sales Taxes (5 Percentage Point Increase in Sales Tax Rate)
Introduction
Sales taxes are the second largest source of tax revenue for state and local governments in the United States. Forty-five states and the District of Columbia levy a statewide sales tax on most goods and some services, and counties, cities and special districts frequently add their own rate on top.
Gaarder and Henry de Frahan (2025) estimate the welfare costs of various changes to state sales tax rates. The analysis combines NielsenIQ Retail Scanner Data (weekly sales, prices and volumes from approximately 38,000 stores in 48 states over 2006–2016) with a county-level panel of sales tax rates built from Thomson Reuters OneSource. The estimation sample keeps Food, Drug and Mass Merchandiser stores, where NielsenIQ coverage is highest, and the 263 best-selling product modules (the top quartile of 2008 sales), giving a balanced panel of module-store pairs across 22,626 stores in 2,180 counties observed at six-month frequency. Because sales tax bases are set state by state, with extensive exemptions, the paper hand-collected the taxability status of each of the 263 modules in each of the 48 states for every year of the sample.
Identification comes from staggered changes in state, county and city sales tax rates: 2,974 rate changes between 2008 and 2014. The average increase was 0.54 percentage points and the average decrease 0.66 percentage points. The paper finds that pass-through of the tax to consumer prices is essentially complete, suggesting that consumers bear the full incidence of these tax changes. A one percent increase in the tax-inclusive price reduced quantity sold by about 0.5 percent, with the adjustment complete 18 to 24 months after the change. Instrumenting the consumer price with the tax change gives a demand elasticity of 0.453, the causal input to the MVPF estimates.
The paper estimates the MVPF of a 5 percentage point increase in the sales tax rate. The paper separately estimates the MVPF of a marginal increase in the sales tax rate and the MVPF of moving from no tax to the current sales tax rates (equivalent to eliminating state sales taxes).
MVPF = 1.0
Net Cost
Net cost is the revenue the 5 percentage point increase actually raises. The paper expresses net costs as per $1 of revenue raised after the increase.
That dollar is the net of two components. The mechanical component is the five extra points of tax collected on the goods consumers were buying at the old rate. The behavioral component is negative: the higher tax-inclusive price reduces the quantity sold, so the base the tax is levied on shrinks. Formally, per dollar of pre-reform sales the additional revenue is
with the quantity ratio traced out by the estimated demand curve between the two tax-inclusive prices. With a demand elasticity of 0.453, adding five points to an average total rate of about 7 percent contracts the base by roughly 2 percent, and collections come in about 5 percent short of the mechanical amount.
The behavioral loss is large because the fiscal externality depends on the product of the tax wedge and the elasticity, and this reform roughly doubles the wedge.
Willingness to Pay
Willingness to pay is the combined amount consumers and firms would pay to avoid the 5 percentage point increase: $1.039 for every dollar of revenue those taxes raise.
Because this is a non-marginal reform and the quantity response is first-order, the envelope theorem does not apply. Willingness to pay is then the true equivalent variation: the entire area under the demand curve between the current tax-inclusive price and the higher one:
This includes the Harberger triangle rather than approximating it away, and it is integrated along the estimated demand curve rather than under the conventional assumption that demand is locally linear.
Willingness to pay exceeds the revenue raised because consumers lose more than they hand over: on top of the extra tax paid on what they still buy, they lose the surplus on the purchases they forgo.
The split between consumers and firms is governed by the pass-through rate. The paper estimates pass-through of 1.034 (SE 0.028), which suggests consumers bear the entire burden: in this case, every purchaser of taxable retail goods in the affected jurisdictions.
MVPF Summary
Dividing willingness to pay by net cost gives $1.039 / $1.000 = 1.039.
Every additional dollar of revenue raised by a 5 percentage point increase in sales taxes would impose about $1.039 of welfare loss on consumers and firms.
The result is robust to relaxing the model assumptions:
- This is the point-identified estimate under iso-elastic demand. Under the most flexible specification considered (an eighth-order polynomial), the MVPF is bounded between 1.009 and 1.057.
- Allowing imperfect competition, the paper backs out a conduct parameter of 0.002 from the observed pass-through and the bounds are essentially unchanged.
- Under imperfect salience the envelope theorem no longer holds, because consumers are buying more of the taxed good than they would if they attended fully to the tax. The paper reports that at a salience parameter of 0.25 the MVPF of a 5 percentage point increase falls to between 1.000 and 1.012 under the most flexible demand shape.
The paper also computes the MVPF state by state. Bounds for an individual state are somewhat wider than for the average state, but for 40 of the 44 states in the sample the value of a 5 percentage point increase can be bounded below 1.08 even under the most flexible demand shape; for the remaining four the upper bound runs between 1.09 and 1.13. Increases are larger in states with higher initial rates and higher price levels.
References
Gaarder, Ingvil and Lancelot Henry de Frahan (2025). “The Welfare Effect of Marginal and Nonmarginal Changes in Sales Taxes in the U.S.” Becker Friedman Institute Working Paper No. 2025-83. https://bfi.uchicago.edu/working-papers/the-welfare-effect-of-marginal-and-nonmarginal-changes-in-sales-taxes-in-the-u-s/
Policy Features
- Category
- Taxes
- Sub-Category
- Local Taxes/Subsidies
- Beneficiary Type(s)
- Adults
- Country of Implementation
- United States
- Year of Implementation
- 2006-2016
- Empirical Method
- Difference in Differences
- Research Type
- Primary
- Peer Reviewed
- No