Who actually pays a tax
A parliament can decide who hands the money to the state. It cannot decide who ends up poorer. The first is a line in a statute. The second is settled by how easily each side of the market can walk away, and you can watch it being settled below.
Every tax law names a payer. The shop remits the value added tax. The employer remits half of the social contribution. The importer remits the customs duty. Read the statute and you would think the question of who bears the cost has been answered in the text.
It has not. The statute fixes the statutory incidence, which is the name on the transfer. What we usually want to know is the economic incidence, which is whose consumption falls. These are different quantities, and the market decides the second one without consulting the first.
The argument is short enough to state in a sentence, and that is exactly why it is worth running rather than reading. Below is a market. You set how sensitive each side is to price by dragging its curve, you set a tax, and you choose which side of the market the law tells to pay. Two of those three controls change the answer. One of them does nothing at all.
01The market, and a tax placed on it
Quantity runs along the bottom, price up the side. The downward line is demand: how much buyers want at each price. The upward line is supply: how much sellers will bring at each price. Where they cross, the market clears. Drag either line to make it flatter or steeper. A flat line means that side responds strongly to price and can walk away. A steep line means it cannot.
Buyer bears 00 borne by seller
02Change who writes the cheque
The last control in the rack is the one to distrust. It moves the tax from the seller to the buyer, which in a real system would mean rewriting the law, retraining a finance department and reprinting the forms. On the picture, one line moves instead of the other. The supply curve shifts up by the tax, or the demand curve shifts down by it.
Watch the six figures in the readout while you switch. The price the buyer pays does not move. The price the seller keeps does not move. The quantity does not move. The revenue does not move. The share each side bears does not move. The tax opens a gap of a fixed size between what the buyer pays and what the seller receives, and a gap does not care which end of it you nail down first.
This is the result that surprises people, and it is worth being precise about what it claims. It does not say that the law is irrelevant to anything. It says that in this model, holding everything else equal, the legal side of the transfer does not enter the arithmetic that determines the two prices. Section five is where I set out the evidence that this clean claim is not the whole story in the world.
03The rule, in one line
If the statute does not decide the split, something has to. It is elasticity: the percentage change in quantity that follows a one per cent change in price. A buyer who can substitute easily has elastic demand and can refuse to pay more. A seller who cannot cut production has inelastic supply and cannot refuse to accept less. Whoever is less able to leave is the one who stays and pays.
Written out, with the elasticity of supply as e of s and the magnitude of the elasticity of demand as e of d, the buyer's share of a small tax is this and nothing else.
buyer's share of the tax = e of s / (e of s + e of d) The whole of section two is contained in the fact that the side named by the statute does not appear anywhere in that expression.
The curve above is that expression and nothing else. The marker is where the sliders currently put you. Move them and the marker slides along a fixed line, because the market picture and this curve are two drawings of one equation. When supply and demand are equally elastic the marker sits at half. When one side is ten times as elastic as the other, the split is roughly ten to one against the side that cannot move.
04The triangle nobody collects
Raise the tax and a shaded wedge opens between the curves to the right of the new quantity. Those are trades that would have happened and now do not. The buyer valued the unit above what the seller would have accepted, so there was a gain to be had, and the tax made the two of them unable to reach each other. Nobody receives that value. It is not revenue. It is simply gone.
The shape matters. Double the tax and the revenue rectangle roughly doubles, while the triangle roughly quadruples, because both of its sides grow with the tax. Set the tax to five and note the loss. Set it to ten and note it again. That is the arithmetic behind the standard advice to raise a small amount from many bases rather than a large amount from one, and it is also why the loss is small enough to ignore at low rates and impossible to ignore at high ones.
Notice too that the triangle grows with elasticity. Make both curves steep and a large tax destroys almost no trades, because nobody was going to change their behaviour anyway. That is the same property that made the inelastic side bear the burden, seen from the other direction: a tax on something nobody can avoid raises money efficiently and falls on people who have no way out. The efficiency and the distribution are the same fact reported twice.
05Where the model stops
The picture above is a partial equilibrium model of one market with straight lines in it. That is a strong set of assumptions, and every one of them is doing work. Here is what the model is not entitled to tell you.
Assumptions and limits
- One marketOne market is drawn and everything else in the economy is held still. In truth a tax on fuel moves the market for cars, for freight and for labour. Those effects can be larger than the one shown here, and they are entirely outside the frame.
- No income effectThe demand curve is treated as a schedule of willingness to pay. It does not account for the fact that a tax makes buyers poorer overall, which shifts what they want to buy of everything. The model measures a substitution and calls it the whole response.
- StaticThere is a before and an after and nothing in between. Real adjustment takes time, and elasticity is almost always larger over years than over months, so a tax that lands on producers in the first quarter can land on consumers by the third year. A single pair of elasticities cannot express that.
- Straight linesLinear curves make elasticity vary along the curve, so the elasticity you set is the elasticity at the point where the market clears without the tax and nowhere else. The share formula in section three is exact for a small tax and an approximation for a large one.
- Competition assumedBoth sides are price takers. A monopolist, or a market with a handful of sellers watching each other, can pass on more than the tax or less than the tax, and this model has no way of representing that.
- The statutory side really can matterThe clean result in section two is a property of the model, not a law of nature. Saez, Matsaganis and Tsakloglou studied a Greek reform that raised both employer and employee contributions for workers hired after a cutoff date, and found that employers compensated those workers for the higher employer contribution but not for the higher employee contribution. The nominal split changed the outcome. Benzarti collects further cases where it does. Both are cited below.
- Numbers are illustrativeThe prices, quantities and elasticities on this page are made up so that the mechanism is visible. Nothing here is an estimate of any real market, and no figure from this page belongs in an argument about a real tax.
06Sources
The model is standard partial equilibrium price theory and is not attributable to anyone in particular. The empirical claims made above are these three, and only these three.
- Employers in a Greek natural experiment compensated new workers for higher employer contributions but not for higher employee contributions, so the statutory split affected the economic outcome. Emmanuel Saez, Manos Matsaganis and Panos Tsakloglou, Earnings Determination and Taxes: Evidence From a Cohort-Based Payroll Tax Reform in Greece, Quarterly Journal of Economics 127(1), 2012, pages 493 to 533. academic.oup.com/qje/article-abstract/127/1/493/1832830 Working version read at eml.berkeley.edu. Checked 2026 08 21.
- Value added tax pass through is asymmetric: increases are passed to consumers several times more fully than cuts are, and the gap persists for years. The same survey collects evidence that who remits a tax can affect its economic incidence. Youssef Benzarti, Tax Incidence Anomalies, NBER Working Paper 32819, August 2024. nber.org/system/files/working_papers/w32819/w32819.pdf Checked 2026 08 21.
- The Congressional Budget Office estimates that about 58 per cent of the additional burden of a broad increase in the payroll tax rate falls on employees in the short run, with the rest borne by capital and other factors. Dorian Carloni, Revisiting the Extent to Which Payroll Taxes Are Passed Through to Employees, CBO Working Paper 2021 06, June 2021. cbo.gov/publication/57089 Checked 2026 08 21.