CHICAGO—The only nice thing about being stuck on an airplane (aside from free soda) is the chance to catch up on one’s reading. On this trip, I (finally) turned to Jason Furman’s article “Our Unhealthy Tax Code” from the premiere issue of Democracy. I address Furman’s objections to health savings accounts in a recent paper. (Refreshingly, we actually agree on one or two things.) But Furman also commits what I think is an important error when discussing tax deductions for health care.

Congress exempts employer-provided health benefits from income and payroll taxes, which costs the federal government an estimated $200 billion per year in lost tax revenue. Furman describes this as the federal government “spending approximately $200 billion annually in subsidizing employer-provided insurance.” But that is flat incorrect. A tax deduction allows workers to keep more of their own income, provided they engage in a desired behavior — in this case, obtaining employer-sponsored health insurance. It is not government spending because the government cannot spend money that it never possesses. Nor is it a subsidy, because to subsidize means to transfer resources, and again the government never possesses those resources. The tax deduction may have the same effect on government revenues (and the economy) as government spending for the same activity. But that still doesn’t make it spending.

Suppose I robbed Peter and gave his money to Paul. Suppose alternatively that Peter gave his money to Paul because I threatened to punch him in the nose unless he did. In the first scenario, I would be spending Peter’s money on Paul. But in the second, Peter would be spending his money on Paul. I wouldn’t be spending anything. I merely would be threatening to assault Peter unless he did what I said.

A more precise way of describing such tax deductions would be to say that they create price distortions between favored and non-favored activities. If a health insurance policy and a plasma TV each have a nominal price tag of $5,000, the fact that health insurance is tax-deductible reduces its price relative to the TV. (If the price distortion is greater than my preference for the TV, the tax deduction will induce me to consume the less-valued option, creating economic losses.)

The distinction might seem semantic, but it has important normative implications. Furman despises the tax deduction for employer-sponsored health insurance on equity grounds: The wealthy get larger tax deductions than lower-income workers. (I despise it too, for different reasons.) A government spending program that disproportionately subsidizes upper-income workers would be offensive to more people than a policy that merely lets some workers keep more of their own money. The former description suggests (assumes?) that the money belongs to the state, and that the state has the right to spend it on something else at its discretion. The latter suggests (correctly) that the money belongs to the people who earned it, and for the state to spend that money would require a $200 billion tax increase.

To anticipate a predictable objection, I am aware that the federal government itself prefers the former description, and the term tax expenditures. It is hardly surprising that the federal establishment chooses the description that presumptively increases its claim on society’s resources. That it is a convention makes it no less incorrect.