The Laffer Curve Explained: What It Claims and What the Evidence Shows

The Laffer curve is the claim that tax revenue does not rise forever as the tax rate rises: it climbs, reaches a peak, and then falls back toward zero. The logic behind it is short. A tax rate of 0% collects nothing, because nothing is taken from anything. A tax rate of 100% on the taxed activity also collects nothing, because nobody undertakes an activity whose entire proceeds are taken. If revenue is positive at any rate in between, and if revenue moves smoothly as the rate moves, then somewhere between the two endpoints there is a rate that collects the most.

That much is close to arithmetic, and it is not controversial among economists. What is controversial, sharply and persistently, is where the peak sits. The endpoint argument fixes the two ends of the curve and says nothing at all about the middle. Locating the peak is an empirical question about how much the taxed activity shrinks when the rate rises, and that answer differs by tax, by country, by decade, and by which slice of the rate schedule you are moving. This guide works through the mechanics, builds the curve from an explicit model so you can see exactly what determines the peak, sets out the evidence and where it is contested, and then covers the ways the curve is misused in public argument.

The two effects behind the curve

Arthur Laffer described the curve as the interaction of two forces that pull in opposite directions.

The arithmetic effect. Revenue equals the rate multiplied by the tax base, the quantity of income, sales, or profit the rate is charged on. Holding the base fixed, raising the rate raises revenue in direct proportion. Cut the rate by a fifth and you collect a fifth less on every dollar of base. This effect is mechanical and its direction is never in doubt.

The economic effect. Taxes change behavior. A higher rate on the next dollar earned lowers the return to earning it, so some of the taxed activity stops, gets postponed, moves into a form the tax does not reach, or is not reported. The base shrinks. A lower rate does the reverse. This effect works against the arithmetic effect whenever the base responds negatively to the rate, which is what every credible estimate finds, and its size is what nobody can pin down from theory.

Write revenue as R = t x B, where t is the rate and B is the base. Because the base depends on the rate, raising t raises the first term and lowers the second. Near a rate of zero the arithmetic effect dominates, because there is barely any tax to escape. Near a rate of 100% the economic effect dominates, because the base has collapsed. A peak is a rate where the two exactly cancel. Everything interesting about the Laffer curve is a question about the relative strength of these two effects at the rate a country actually charges.

Building the curve from a model

The clearest way to see what fixes the peak is to write down a base that responds to the rate and solve for the maximum.

Let the reported base respond to the net-of-tax rate, the share of the next dollar a taxpayer keeps, which is (1 minus t). Suppose the base has a constant elasticity e with respect to that share:

B = B(0) x (1 - t)^e

so revenue is R = t x B(0) x (1 - t)^e. Maximizing this with respect to t gives a clean result:

Revenue-maximizing rate = 1 / (1 + e)

The peak depends on one number and nothing else. Every other feature of the economy has dropped out. Put values in:

Check one row by hand. With e = 0.5 and a starting base of 1,000, a rate of 50% collects 0.50 x 1,000 x (0.50)^0.5 = 354. A rate of 67% collects 0.67 x 1,000 x (0.33)^0.5 = 385. A rate of 80% collects 0.80 x 1,000 x (0.20)^0.5 = 358. Revenue rises to a peak near 67% and falls after it, exactly as 1/(1 + 0.5) predicts.

Notice what the table shows and what it does not. It shows that the curve has a peak for any positive elasticity, which is the uncontroversial part. It also shows that the peak can sit anywhere from a third to nearly nine tenths depending on a parameter the model does not supply. Nothing in the theory tells you which row you are in. That is the whole of the disagreement, compressed into one column.

Why the endpoints prove less than people think

Two qualifications belong on the endpoint argument, because both are routinely skipped.

The 0% endpoint is an identity. Zero times anything is zero, and no assumption about behavior is needed.

The 100% endpoint is a behavioral claim, not an identity. It says that if a taxpayer keeps none of the next dollar, the dollar is not earned, or not reported. That is highly plausible for a rate applied to an entire base, but it is a statement about the rate on the marginal dollar across the whole base, not about any schedule that contains a 100% band. A system with a 100% marginal tax rate on income above some very high threshold would still collect a great deal of revenue from income below the threshold, because the average tax rate most people pay would be far under 100%. Mixing up the marginal rate the curve is drawn against with the average rate revenue is collected at is one of the most common errors in this topic, and it is worked through in detail on the marginal tax rate vs Laffer curve comparison.

The endpoint argument also does not guarantee the tidy symmetric hump that gets drawn on whiteboards. It establishes that revenue rises then falls. It does not fix the peak's height, its position, its symmetry, or even that there is only one peak. The smooth arch is an illustration, not a derived shape. Whether the taxed quantity is even bounded by a 100% rate depends on how the rate is quoted: Mathias Trabandt and Harald Uhlig, modelling Laffer curves for the United States and Europe, found that in their framework the consumption-tax curve has no peak at all. Because a consumption tax rate quoted on the pre-tax price has no ceiling at 100%, and because the extra revenue is returned as transfers that are then spent and taxed again, revenue rises throughout and converges to a finite level rather than turning back down.

Elasticity of taxable income: the parameter that decides everything

In the modern literature the parameter e has a name: the elasticity of taxable income, usually shortened to ETI. It measures the percentage change in reported taxable income when the net-of-tax share rises by one percent. It is deliberately broader than a labor supply elasticity, because it captures every margin by which reported income responds to the rate:

• Real responses. Hours worked, effort, career choice, saving, investment, entrepreneurship, and where people choose to live and work.
• Timing responses. Moving income between tax years, most visibly by choosing when to realize a capital gain or when to take a bonus.
• Shifting responses. Recharacterizing income into a more lightly taxed form, such as retained corporate profit, fringe benefits, deferred compensation, or a different legal entity.
• Avoidance and evasion. Deductions, sheltering, and unreported income.

Revenue depends on what is reported, so all four count for the Laffer question. They do not all count the same way for welfare, which matters later.

For a rate that applies only to the top bracket, the formula changes shape slightly. Peter Diamond and Emmanuel Saez derive the revenue-maximizing top rate as

Revenue-maximizing top rate = 1 / (1 + a x e)

where a is the Pareto parameter of the upper tail of the income distribution, which they estimate at roughly 1.5 for US tax return data. The extra term appears because a top-bracket rate applies only to income above a threshold, so the revenue at stake is smaller relative to the base that can respond. Their headline calculation uses e = 0.25, which they describe as a mid-range estimate from the empirical literature, giving

Revenue-maximizing top rate = 1 / (1 + 1.5 x 0.25) = 73%

against a combined US top marginal rate on earnings they put at about 42.5% when they wrote in 2011, counting federal income tax, Medicare, and average state income and sales taxes. They also note the mirror image of that calculation: the rate then in force would itself be the revenue-maximizing rate only if the elasticity were around 0.9, far above anything in the mainstream range. Run the same formula across the plausible range and you get:

What the evidence actually says

Here the honest summary is narrower than either side of the public argument usually admits.

On the existence of the curve there is broad agreement. The endpoint logic is sound, behavioral responses to tax rates are large and well documented, and few economists dispute that some rate maximizes revenue.

On the location of the peak there is real disagreement, but it is bounded. In the most cited survey of this literature, Emmanuel Saez, Joel Slemrod, and Seth Giertz reviewed the ETI estimates and concluded that while there are no convincing estimates of the long-run elasticity, the best available estimates fall in a range of roughly 0.12 to 0.40, with the midpoint around 0.25. Their assessment of what that implies for the United States is direct: even at the top of that range, they write, the US marginal top rate is far from the top of the Laffer curve. They also note that the ETI appears to be higher for high-income taxpayers, who have more access to avoidance opportunities, so the top-rate peak sits lower than an economy-wide peak would.

Cross-country modelling points the same way, with exceptions. Trabandt and Uhlig estimated that under their benchmark parameters the United States could raise labor tax revenue by around 30% and capital income tax revenue by around 6% by raising those rates, with smaller figures of about 8% and 1% for the EU-14, placing both on the upward-sloping side of their curves. They did find individual cases on the far side: in their results Denmark and Sweden sat beyond the peak for capital income taxation. Different models with different assumptions about how strongly labor and capital respond produce different answers, which is exactly the point.

Surveys of academic economists show the same asymmetry. When the University of Chicago Booth School's panel of academic economists was polled in June 2012 on whether a cut in US federal income tax rates at that time would raise total annual tax revenue within five years, not one panelist agreed. Thirteen strongly disagreed, nine disagreed, and three were uncertain. On a companion question about whether the same cut would raise GDP, the same panel split, with many agreeing and many uncertain. The pattern is worth reading carefully: a rate cut plausibly raising output is a different and much weaker claim than a rate cut paying for itself.

The fair summary is therefore two-sided. Anyone claiming the Laffer curve is a fiction is wrong about the logic. Anyone claiming it shows that a tax cut in a typical developed economy would pay for itself is asserting an elasticity several times larger than the published estimates. The measured responses do mean a rate cut usually costs somewhat less than a static calculation predicts, and that is a real and useful finding, but "costs less" is not "costs nothing."

Working the break-even condition

The claim that a cut pays for itself has a precise arithmetic test, and running it is the fastest way to see how large a response it demands.

Suppose a rate is cut from 40% to 35%. That is a 12.5% cut in the rate, since 5 divided by 40 is 0.125. For revenue to be unchanged, the base must rise by enough to offset it: 1 divided by 0.875 is 1.1429, so the base must grow by 14.3%.

Now ask what elasticity that requires. The net-of-tax share rises from 0.60 to 0.65, an increase of 8.33%. To get a 14.3% base response out of an 8.33% change in the net-of-tax share needs an ETI of 14.3 / 8.33, which is about 1.7. That is four to fourteen times the range the survey literature reports.

Run the same cut with e = 0.25 instead. The base rises by 0.25 x 8.33% = 2.1%. Revenue goes from 0.40 x B to 0.35 x 1.021 x B, which is 89.3% of what it was. A static calculation would have predicted 87.5%. So the behavioral response offset roughly one seventh of the cut's static cost, and the remaining six sevenths shows up as lost revenue. That is the shape of the mainstream result in one line of arithmetic, and it is worth noticing that this stylized calculation is an illustration of the mechanism rather than a revenue score of any actual policy.

Short run versus long run

Two rate responses can look identical in the first year's data and mean completely different things.

Timing responses are fast, large, and temporary. When a rate change is announced in advance, taxpayers move income across the boundary. Realized capital gains are the classic case: US realizations surged in the year before an announced increase in the rate on realizations took effect. This is real behavior, but it moves the same income between years rather than creating or destroying it. Measure the response in the year of the change and you overstate the lasting effect badly. The United Kingdom's introduction of a 50% additional rate produced a well-documented episode of the same kind, with income brought forward into the previous tax year, and the argument over how much of the observed drop in reported top income was forestalling rather than lasting behavior is precisely why the official assessment of that rate was contested rather than settled.

Shifting responses are fast and can be permanent, without changing output at all. When one rate falls below another, income migrates between the two bases. After the US top individual rate was cut below the corporate rate in the mid-1980s, a large amount of business activity moved into pass-through entities that are not subject to corporate tax. Reported individual income rose sharply. Much of that was relabeling, not new production.

Real responses are slow and hard to identify. Changes in saving, capital accumulation, career choice, and firm formation play out over years, and separating them from ordinary growth, inflation, demographic change, other tax changes in the same period, and the business cycle is genuinely difficult. Saez, Slemrod, and Giertz are explicit that empirical methods are most convincing for short-term responses, that they regard estimates of the long-run elasticity as unconvincing, and that at the top of the income distribution the well-identified responses fall into the timing and avoidance categories rather than the category of real economic response.

The direction of the resulting bias is not settled, which is why honest summaries stay uncertain. Short-run estimates inflated by forestalling would push the estimated peak too low. If long-run capital accumulation effects are large and unmeasured, the true long-run response could be larger than the short-run one, pushing the peak the other way.

There is one further implication that deserves more attention than it gets. Because avoidance and shifting are part of the measured elasticity, the peak's location is partly a policy choice rather than a fact of nature. Saez, Slemrod, and Giertz make the point directly: if the behavioral response in the current system is large, the response that follows is not necessarily to lower rates, but to broaden the base and close the avoidance channels, which shrinks the elasticity and moves the peak higher. The ETI is not a constant of the universe. It depends on how the tax code is written.

Why the curve gets misapplied

The Laffer curve is unusually easy to invoke and unusually hard to use correctly. The recurring errors are worth naming.

Treating the curve as evidence about which side you are on. The curve establishes that a peak exists. It contains no information whatsoever about whether a given country's current rate is left or right of it. Sketching the hump and pointing at the downhill slope is an assumption dressed as a diagram. Answering the question requires the elasticity, and the elasticity requires data.

Confusing a bigger base with more revenue. Cutting a rate almost always widens the base to some degree, and this is often reported as vindication. It is not. Revenue is rate times base, and the base growing is compatible with revenue falling, which is what the mainstream estimates imply for rates in the ranges most developed economies charge.

Reading raw revenue totals before and after a rate change. Revenue moves with inflation, population, income growth, other tax changes enacted in the same window, and the business cycle. Comparing the total collected two years before a cut with the total two years after confounds all of these with the rate. It is why researchers identify the elasticity from differences across taxpayer groups facing different rate changes rather than from national revenue totals.

Generalizing from one tax to another. The curve is specific to a tax, a base, a country, and a period. A narrow excise on a good with close substitutes peaks at a low rate, because buyers escape by switching, which is the same logic as the demand side of a tax incidence problem and depends on the same elasticity reasoning. A mobile base such as corporate profit peaks lower than an immobile one. A broad tax on wages, where escape routes are limited, peaks high. A result about one tells you very little about another.

Reading the peak as a target. The revenue-maximizing rate answers exactly one question: which rate collects the most money. Most tax arguments are also about the distribution of the burden, the deadweight loss the tax creates, growth, and administrative cost. A tax system sitting exactly at the peak would be maximizing revenue at the point where the cost of collecting one more dollar becomes infinite, since at the peak no additional revenue can be raised at all. The peak is best understood as a ceiling that any revenue-raising argument has to respect, not as a destination.

Using it as a general theory of tax policy. The curve is one relationship between one rate and one revenue stream. It says nothing about who bears the burden, which is the subject of tax incidence, nothing about the fairness of the rate structure, which is the subject of progressive versus regressive taxes, and nothing about whether the resulting deficit matters, which runs through national debt versus deficit and crowding out.

Where the curve came from

The mechanism is much older than the name. Ibn Khaldun, writing in the fourteenth century, observed in the Muqaddimah that at the beginning of a dynasty taxation yields a large revenue from small assessments and at the end yields a small revenue from large assessments. John Maynard Keynes made a similar observation in the twentieth century. Laffer himself has credited both in his own account of the idea's history.

The modern version dates to a Washington dinner in 1974, at which Arthur Laffer, then an economics professor, sketched the curve for Donald Rumsfeld and Dick Cheney, both then officials in the Ford administration, and the journalist Jude Wanniski. Wanniski gave the curve its name in a 1978 article in The Public Interest titled "Taxes, Revenues, and the 'Laffer Curve'." The cloth napkin bearing a version of the sketch is now in the Smithsonian's National Museum of American History, though the provenance is genuinely disputed: the date inscribed on it does not match Wanniski's account of the dinner, and Laffer has said he does not believe the museum's napkin is the original.

The idea became one plank of supply-side economics, a program built on shifting long-run aggregate supply outward rather than managing aggregate demand. In the United States, the Economic Recovery Tax Act of 1981 cut the top federal marginal rate on ordinary income from 70% to 50%, and the Tax Reform Act of 1986 cut it in two steps to 28%. Those cuts started from historically high rates: the top federal rate on ordinary income had been above 90% from the mid-1940s through 1963, at 91% for most of that stretch and 92% in 1952 and 1953, before being cut in stages to 70% by the mid-1960s. That starting point matters for reading the historical debate, because a rate of 70% or 91% is far closer to the range where the published elasticity estimates would place a peak than any rate in force in a large developed economy today. An argument that applied at 70% does not automatically transfer to 35%.

Common questions

Does the Laffer curve prove that tax cuts pay for themselves?

No. It proves that a revenue-maximizing rate exists somewhere between 0% and 100%. A tax cut raises revenue only if the current rate is above that peak. Below the peak, cutting the rate loses revenue in the ordinary way, though usually somewhat less than a static calculation predicts because the base widens a little. Whether any particular cut pays for itself depends entirely on the elasticity of the base, which the curve itself does not supply.

Where is the peak of the Laffer curve?

There is no single answer, and that is the substantive point rather than a dodge. It differs by tax, by country, and by period. For a top income tax rate, applying the standard formula to the published range of elasticity estimates puts the peak somewhere in the region of 63% to 85% for the United States, well above the combined top rate actually charged. For a narrow excise on a good with close substitutes, or for a highly mobile base, the peak is much lower.

Why does a 100% tax rate raise no revenue?

Because the taxpayer keeps none of the next dollar, so the taxed activity is not undertaken, or not reported. This applies to a marginal rate charged across an entire base. A 100% band applied only above a very high threshold would still collect substantial revenue from income below the threshold, so the endpoint is a statement about the rate on the marginal dollar, not about any schedule containing a 100% rate.

Do economists accept the Laffer curve?

Broadly yes as a logical proposition, and broadly no as a claim about current rates in developed economies. The existence of a peak follows from the endpoints and is not seriously contested. The available estimates of how strongly reported income responds to rates place developed economies on the upward-sloping side, meaning higher rates would raise more revenue and lower rates would raise less, with the strength of that conclusion varying by tax and by country.

Is the Laffer curve on the AP exam?

Not by name. The Laffer curve does not appear in the AP Macroeconomics Course and Exam Description, which covers supply-side fiscal policy (LO POL-4.A) without naming the curve, so it will not be asked for as a labelled diagram. It is named explicitly in A-level economics specifications, where it is usually a diagram to label plus an explanation of why revenue falls past the peak. An answer that earns full marks states the two endpoints, identifies the revenue-maximizing rate between them, explains the arithmetic and economic effects, and then says that the peak's location is an empirical matter rather than something the diagram determines.

Practice and connect

The skill worth drilling is the break-even calculation: take a proposed rate cut, compute the percentage fall in the rate, compute the percentage rise in the base needed to offset it, and then ask whether any credible elasticity delivers that response. Run it on the marginal tax rate calculator and check the wedge mechanics on the tax incidence calculator.

For the surrounding theory, the fiscal policy module places tax changes inside the aggregate demand and aggregate supply model, fiscal policy versus monetary policy separates the two levers, and what is deadweight loss explains the efficiency cost that sits behind the economic effect. Lock in the definitions in the Laffer curve, tax base, and fiscal policy glossary entries, and browse the full glossary for the surrounding vocabulary. Once you can state what the curve establishes, what it leaves open, and what number would have to be true for a cut to fund itself, you can handle the topic in an exam and read the public argument about it with a clear head.