Where tax cuts pay for themselves and where they honestly do not

Here is a sentence guaranteed to start a fight at any dinner table with an economist and a politician seated side by side: "Tax cuts pay for themselves." The politician nods. The economist winces. And the rest of us are left holding the gravy, wondering who is right.

The honest answer is that both are right about different taxes. A cut on one kind of tax, in one kind of circumstance, can come close to paying for itself or even raise money. A cut on another kind does not come close. The conservative case for lower taxes is strong enough that it does not need the slogan. It needs precision. Let me show you where the slogan holds, where it breaks, and why the case for cutting survives the break.

Define the terms before the shouting starts

Three terms do most of the work in this argument, and the fight is usually a fight about which one is being used.

Revenue is what the Treasury collects. Rate is the share of each additional dollar the government takes. Base is the thing being taxed: wages, profits, capital gains, sales. Revenue is, roughly, rate multiplied by base. That is not a slogan; it is arithmetic.

When Congress cuts a rate, the multiplication has two moving parts. The rate goes down, which lowers revenue if the base stays put. But the base can move, because people respond to what things cost. They work more or less, save more or less, move income across years, restructure a business, relocate a company, or simply decide to report a gain now instead of never. The question "does a tax cut pay for itself?" is really the question "does the base grow enough to cancel the lower rate?"

A fourth term belongs here. A revenue-maximizing rate is the rate at which the government collects the most from a given tax. Raise the rate past it and the base shrinks so much that revenue falls. This is the idea behind the Laffer curve, popularized in the 1970s: a zero percent rate collects nothing, a 100 percent rate collects nothing because no one would bother earning taxable income, and somewhere between lies a peak. Almost no one disputes that the curve exists in theory. The dispute is where the peak sits for each tax. That location is an empirical question, and it is different for different taxes.

The arithmetic of "paying for itself"

Take a hypothetical, labelled as one. Suppose a tax of 30 percent applies to a base of $100, so the Treasury collects $30. Congress cuts the rate to 27 percent, a ten percent reduction in the rate.

If the base does not change, revenue falls to $27. To collect the original $30 at the new rate, the base would have to grow from $100 to about $111.11, because $30 divided by 0.27 is 1.111. So a one-tenth cut in the rate needs the base to grow by about eleven percent, and it needs to do so because of the cut, not because of the ordinary growth that would have happened anyway.

Now run the same exercise on a larger cut. A rate of 28 percent cut to 20 percent needs the base to grow by 40 percent, because 28 divided by 20 is 1.4. Forty percent more of something people were already earning, in response to a tax change alone, is a heroic ask. For some taxes it has happened. For most it has not.

This is the number conservatives should carry in their heads. The deeper the cut and the less responsive the base, the less of it comes back. A slogan hides that. A table does not.

Where the base is highly responsive

Some taxes sit on a base that moves a great deal when the rate moves. The clearest case is the tax on capital gains, the profit when you sell an asset for more than you paid.

A capital gains tax is unusual, because the taxpayer chooses when to pay it. You owe the tax only when you sell. A high rate gives an owner a reason to hold on, sometimes for decades, sometimes until death, when the tax on the gain can disappear for heirs under the rule that resets the cost basis. Economists call this the lock-in effect, and I have written about it at length in capital gains taxes and the lock-in effect that keeps money frozen in place. When the rate falls, owners who were waiting sell, the gains get realized, and the Treasury collects tax on gains that would otherwise never have been taxed at all.

The record bears out the direction of this effect. In the late 1990s Congress lowered the top federal capital gains rate from 28 percent to 20 percent, and the gains that taxpayers realized grew a good deal in the years that followed, alongside a booming stock market. Disentangling the cut from the boom is hard, and honest economists argue about the size of the response. But few deny that the response was large enough to recover a substantial share of the cost, and some estimates say all of it. Check the Joint Committee on Taxation and the Congressional Budget Office on how each of them models realizations; both treat the timing response of capital gains as real.

Other highly responsive bases include corporate profits that can be shifted across borders, and the incomes of a small number of mobile, high-earning people who can change where and when they are paid. I will come back to the corporate case in the corporate rate cut to 21 percent and the inversions it brought to an end.

Where the base barely moves

Now the other end. A broad tax on wages is a different animal. Most workers cannot change their hours much in response to a few points of tax. They have a job, a schedule and a mortgage. Some respond at the edges: a second earner in the family deciding whether the paycheck is worth the day care, a worker choosing overtime, an owner of a small business deciding to expand. Those responses are real, and they are the reason the marginal rate matters so much; see the marginal rate is the one that decides whether you work the extra shift. But in aggregate the wage base moves much less than the capital gains base.

So when the top rate on wages falls, or when the middle rates fall, the base does not leap to fill the hole. A broad cut in income tax rates, on its own and with nothing else changing, loses revenue in the first years. Many conservative economists have said exactly this, including those who worked in the administrations that cut rates. The Reagan cuts of 1981 were followed by revenue increases in nominal terms, but also by a 1982 bill that raised taxes to stem the deficit, and by years of growth that would have lifted revenue under almost any rate. The lesson of that decade is not that cuts paid for themselves. It is that growth plus restraint can outpace a loss of revenue, and that the loss was not what drove the deficits.

A parallel happened in the 1960s. The Revenue Act of 1964, signed by Lyndon Johnson after John Kennedy had proposed it, lowered the top rate from 91 percent to 77 percent, then to 70 percent in 1965. Revenue rose in the years after, and so did the economy. But the economy was also being pushed by other forces, and later estimates generally say the cuts recovered a portion of their cost, not the whole.

How Washington counts: conventional against dynamic scoring

Now the machinery. When Congress considers a tax bill, two official scorekeepers estimate what it costs: the Congressional Budget Office and the Joint Committee on Taxation. Their method is the center of the fight.

Conventional scoring accounts for behaviour, but holds the size of the economy fixed. A scorekeeper using it will assume that people shift when they realize a gain, change how they take compensation, and move income between tax categories. What the scorekeeper will not assume is that the cut raises total output, or total hours worked, or total investment. The macroeconomic path is the same with or without the bill.

Dynamic scoring relaxes that last assumption. It models how a policy changes labour supply, saving and investment, and so changes the size of the economy and the revenue it generates. The Joint Committee on Taxation produces dynamic estimates for major bills, and the House of Representatives adopted a rule at the start of 2015 requiring that major legislation be scored dynamically. The results for large tax bills have a consistent shape: the growth feedback offsets part of the static cost, and not all of it.

Here is where conservatives talk themselves into trouble. Dynamic scoring is a gain for our side of the argument, because it concedes that taxes affect growth, which the old static approach implicitly denied. But it is not a magic wand. The feedback is real and it is partial. The right way to say it is that dynamic scoring recovers a meaningful share of the cost of a growth-oriented tax cut, and a smaller share of the cost of a cut that does not change the incentives to work, save or invest. Anyone who claims the full cost disappears is overselling. Anyone who claims there is no feedback at all is out of date.

If you want to learn to read these estimates yourself, and you should, see Read a CBO cost estimate yourself before you believe the headline.

The steelman: "cuts just blow a hole in the budget"

Let me state the best case against tax cuts the way its defenders would.

The argument goes: One, the federal government runs a deficit every year, and has for most of the last half century. Two, a tax cut that does not fully pay for itself adds to the deficit unless spending falls by an equal amount. Three, spending has not fallen. Four, therefore a cut raises the debt, and the debt must eventually be paid, either by higher taxes later or by inflation or by crowding out private investment. Five, therefore, the honest way to cut taxes is to cut spending first or at the same time.

That is a serious argument, and I do not think it is wrong on its facts. I think it is wrong about the order of the conclusion.

The answer: rates are not the problem

Here is the reply, in numbered points.

One: spending is the cause of the deficit, not the tax cut. A deficit is spending minus revenue. When spending outruns revenue, there are two ways to close the gap: raise revenue or lower spending. The tax-cut critique treats the first as the default and the second as optional. The Founders saw it the other way. The government is a servant that should ask for what it needs, not a landlord that decides what it will collect. If spending is the thing that grew, then a policy that restrains the government's income is a legitimate way to force the spending argument, even if it works imperfectly.

Two: the revenue number is only one of the costs of a tax. A tax takes money, and it also changes behaviour. Economists call the second cost the deadweight loss: the work not done, the investment not made, the business not started, because of the tax. That cost does not show up in the revenue line, and it grows faster than the rate. Cutting a high marginal rate is worth doing even when revenue falls, because the loss to the economy from the tax shrinks by more than the revenue loss to the Treasury.

Three: the principle is not the Laffer curve. A person's income is theirs before it is the government's. The argument for taking less of it does not depend on whether the Treasury ends up with more. It rests on the claim that a free people should keep what they earn unless there is a compelling reason to take it, and a deficit is a reason to examine spending, not a reason to examine whether citizens have been taxed enough. That is a value claim, not a fact claim, and conservatives should say so out loud and stand on it.

Four: the fairness of the thing. The federal income tax already falls heavily on a narrow slice of earners, which I lay out in Who actually pays the federal income tax, read straight from the IRS tables. A system that takes a rising share of each extra dollar from the people who produce the most is a system that penalizes success at the very point where we want more of it.

Five: complexity. Many of the "costs" of tax cuts come from the structure of the code, not from the rates. Credits, phase-outs and carve-outs make the revenue effect of a rate change unpredictable and give lobbyists a reason to buy the exceptions, a subject for tax complexity is a subsidy for lobbyists, and the code shows how it works. A simpler base with lower rates tends to be sturdier than a complicated base with higher ones.

What the Founders would have asked

The Founders did not have a Laffer curve, but they had a good sense for the thing it describes. The Constitution gives Congress in Article I, Section 8 the power "To lay and collect Taxes, Duties, Imposts and Excises, to pay the Debts and provide for the common Defence and general Welfare of the United States." The purpose is stated in the clause: to pay debts and provide for defence and welfare. It is not a power to collect as much as possible. And the people who wrote it had just fought a war over taxation by a distant legislature, a fight that began with taxes Parliament laid on the colonies.

Their instinct was that every tax is a cost borne by a person, and that a government should justify each one. Read the text of the Constitution at the Constitution page and notice how much of it concerns limits on what government may take and how it may spend.

A verdict, with the evidence attached

So where do tax cuts pay for themselves, and where do they honestly not?

• Highly responsive bases, such as capital gains and mobile corporate income: a cut can recover most of its cost, and in some circumstances all of it.
• Broad wage and income rates: a cut loses revenue, even after counting growth. How much it loses depends on how much it changes work, saving and investment.
• Anything that lowers the cost of a new investment or a new hire at the margin: these tend to recover more than a general rate cut, because they change the decision.

The conservative position is not weakened by that list. It is sharpened. We say: cut where the response is large and the revenue comes back; cut elsewhere too, because the economy grows faster and people keep more of what they earn; and pair every cut with a plan on the spending side, because a country that keeps borrowing will eventually be taxed one way or another. That is the part of the argument where critics are right, and we should say so.

The case in brief

1. Revenue equals rate times base, and a cut pays for itself only if the base grows enough to offset the lower rate. A cut from 28 to 20 percent needs a base 40 percent larger.
2. The base is highly responsive for capital gains and mobile income, where cuts recover much of their cost, and much less responsive for broad wages, where they usually do not.
3. Conventional scoring holds the economy fixed and dynamic scoring lets it move; the dynamic feedback is real but partial for large tax bills.
4. The case for lower rates does not rest on the slogan. It rests on growth, on the smaller deadweight loss from taxes, and on the principle that a free people should keep what they earn.
5. Cuts should be paired with spending restraint, because the deficit is a spending problem that no rate can solve alone.

What you can do: pull up the next tax bill's score on the Joint Committee on Taxation's website, find the line that says whether it uses dynamic estimates, and send that line to your representative with one question about the spending side of the ledger.