Modeling the Revenue and Distributional Implications of a Value Added Tax

When incomes change—as is the case in response to indirect taxes—so do taxes that are based on income. Consider the introduction of a new indirect tax. If the Federal Reserve holds the price level fixed and nominal incomes fall, the bases for taxes such as individual and corporate income taxes will fall. If instead prices are allowed to rise, the real value of tax receipts—the quantity of goods and services the government can purchase per dollar of revenue—is lower. In either case, an increase in indirect taxes causes an offsetting decrease in revenues elsewhere, meaning that the government raises less than one dollar per new dollar of indirect taxes.

Official scorekeepers account for this budgetary “offset” by estimating the effective marginal tax rate on different factor income categories then scaling any first-order indirect tax revenue changes by one minus the total rate. The Joint Committee on Taxation (JCT) currently estimates that the offset is generally about 24 cents—that is, one dollar of new indirect tax revenue leads to about 76 cents of net government revenue.⁵ This approach can be interpreted as reflecting the pass-backward scenario.

Indirect taxes affect different types of income—and thus taxpayers—differently. In the case of a new broad-based value added tax, the extent to which a taxpayer is burdened depends on several factors:

• The economic character of their market income. Labor income falls in real terms regardless of whether prices rise or nominal incomes fall. Capital income attributable to investments made prior to the tax increase is generally burdened, as income available to pay creditors and owners is either lower (pass-backward) or the real value of these nominal income streams is lower (pass-forward). For investments made after the introduction of the tax, when nominal rates of return are renegotiated to account for the higher price level, only above-market (“supernormal”) returns are burdened.
• How much wealth they own. If prices rise, the amount of goods and services that can be consumed by spending down existing net worth falls in real terms; if prices are held fixed, then nominal asset prices would fall to reflect lower nominal future cash flows.  Some of this burden is avoided if only a fraction of wealth is eventually consumed and the rest is bequeathed to heirs or charity.
• To what extent taxes and transfers are indexed to inflation. If taxes and transfers are indexed to inflation, then indirect taxes have no effect on the real value of net transfers (either net transfer income rises with the price level or the nominal value of transfers and consumption are unchanged). If taxes and transfers are instead not indexed to inflation, then a pass-forward scenario leaves transfer recipients worse off in real terms. Social Security is a special case due to its pension-like qualities. In the pass-forward scenario, Social Security benefits are unchanged for current retirees due to the cost-of-living adjustment, but future benefits for current workers are not scaled up to offset the higher price level because nominal wages are unchanged. The pass-backward scenario is economically equivalent: current retirees are unaffected, but nominal wages cuts generate smaller future benefits for current workers.
• The progressivity of the taxes and transfers. As per the discussion of the revenue offset above, a real decline in incomes implies a real decline in taxes. This lower tax burden is a real benefit which partially offsets the loss of income caused by the indirect tax hike. The size of this offset and its distribution by income depends on the current-law distribution of taxes.

A complete theoretical account of these dynamics can be found in Toder et al. (2011), Burman et al. (2006), and this 1992 CBO study.

To estimate the gross revenue raised under a VAT, we start with total personal consumption expenditures (PCE) and adjust the tax base to match that of a specific proposal. Common adjustments to PCE to arrive at a proposed VAT base include the removal of imputed consumption (such as certain financial services), the adjustment of housing to be on a gross investment basis rather than a consumption-flow basis, and other exclusions for certain categories, like health or education spending. The size of a given adjustment is based on recent historical averages of category shares of total PCE, which itself is assumed to grow with projected GDP.

For the exercise below, which is not concerned with a specific VAT base but rather how other revenues change in response to indirect tax changes, we simply assume that a non-specified 50 percent of PCE is subject to the VAT, taxed at a 10 percent rate.

Indirect taxes create a wedge between factor prices and market prices. In the pass-forward scenario, market prices rise by the amount of the tax. In practice, however, there is no single “price level” but rather several indexes which measure the price of different bundles of goods and services.

For our purposes, we care about two broad categories of price indexes. The first are indexes which measure the price of personal consumption, including the CPI, chained CPI, and the PCE deflator. These indexes serve two purposes: first, because they are a suitable measure for understanding how household living standards change, we can use them to calculate real income changes for distribution tables; and second, chained CPI is the current-law tax code indexation metric. To calculate the level shift attributable to the introduction of a VAT, the projected VAT liability is divided by projected PCE. In our example, the price level effect for a 10 percent VAT on 50 percent of PCE generates a price level increase of 5 percent.

The other relevant index is GDP deflator, a measure of the price of all components of GDP, not just consumption. When expressing VAT-scenario revenues in baseline dollars, we require an answer to the question: what is the relevant price level from the government’s perspective? While not an especially well-defined question (especially when compared with the price level for consumers), one reasonable answer is that the government should principally care about GDP as a whole. As such, when calculating the real revenue raised by a VAT in the pass-forward scenario, we deflate counterfactual scenario revenues by the change in GDP deflator, calculated as VAT liability over GDP. In our example, the GDP deflator price level effect is about 3.4 percent.

In the tax microsimulation model, the new consumer price level feeds through into the module which projects the value of inflation-indexed tax parameters. In the case of a VAT, for example, tax brackets would be higher, the standard deduction would be larger, and so on—changes which reflect the tax code’s attempt to hold real tax liability constant across different price levels. Note that the maximum taxable earnings limit for Social Security payroll taxes would not be updated in this scenario, as only the consumer price level, not the wage level, would rise. 

As discussed above, some forms of income would rise with the overall price level in the pass-forward scenario. The first such example is Social Security benefits, which would rise for current retirees but stay unchanged for future retirees. In Tax-Simulator, we first impute benefit claiming age for all records with Social Security benefits. Then, we scale up benefits by the cumulative excess consumer price inflation in the years since retirement. The effect of this logic is that, in the year following the introduction of a VAT, all Social Security income is scaled up such that there is no real decline in benefits. But as time goes on, the share of total benefits receiving a cost-of-living adjustment falls, as more and more current-period recipients retired after the VAT was enacted and thus receive no adjustment.

Investment income is the other major category of income that requires adjustment for VAT-driven price level changes. The size of the adjustment varies along two dimensions: new capital vs old capital, and normal returns vs supernormal returns. Nominal returns to new capital, which are not burdened by the VAT, rise with the price level increase; returns to old capital do not. Of returns to new capital, some fraction—assumed to be 50 percent in our model—represents the normal return, which rises with the price level; the remaining supernormal portion, burdened by the VAT, does not rise.

How investment income is allocated between new/old and normal/supernormal depends on the financing structure. For net interest income (i.e. returns to debt finance), we impute the share of returns accruing from debt issued after each year—the debt’s “vintage”—based on the maturity structure of marketable Treasury security in 2024.⁶ All debt is assumed to earn the normal return. Thus, in the context of a new VAT, the adjustment factor for net interest income for a given year is the cumulative excess consumer price inflation multiplied by the share of income attributable to debt issued after VAT enactment.

For capital gains and dividends (i.e. returns to equity finance), we impute the share of returns attributable to new investment by assuming a 5.7 percent economic depreciation rate. The equity return adjustment, then, is 50 percent (the share of income assumed to reflect supernormal returns) of the cumulative excess consumer price inflation multiplied by the share of income attributable to capital installed after VAT enactment. For pass-through income, we assume that 20 percent of net income represents the returns to capital (the other 80 percent being labor) and scale this component by the equity return adjustment.

The code implementing these adjustments can be found in economy.R. 

After making all adjustments for price level changes, taxes are calculated at the micro level then aggregated to produce revenue projections. Three sets of revenue estimates are generated: nominal, real (i.e. deflated by the increase in GDP deflator), and share-of-GDP. The table below shows these metrics for fiscal years 2025 and 2026 under the VAT experiment previously described (i.e. a 10 percent VAT on 50 percent of PCE). 

Tax-Simulator calculates various distribution metrics by comparing tax liability under baseline with tax liability under the reform scenario, then calculating averages across some dimension like income or ages. In the case of a VAT, however, tax burden takes the form of lower real incomes, not changes in liability. To calculate the distributional effects of a VAT, we simply compare after-tax incomes under the baseline with “real” after-tax incomes under the reform—that is, figures expressed in baseline dollars. This adjustment allows us to capture both the diminished purchasing power of incomes and any reduction in real tax liability.

The impact of a VAT varies by composition of income and by time since enactment. The figure below decomposes our estimated distributional effects of the 10 percent VAT, both in the year following enactment and in the 10th and 30th years of enactment.