In a forthcoming article in Tax Notes, we propose a method for Treasury and IRS to calculate the social costs and benefits of tax regulations that affect revenue collections. This method, which we call the “marginal revenue rule,” builds upon recent work by economists Martin Feldstein, Raj Chetty, Michael Keen, and Joel Slemrod. We argue that revenues resulting from behavioral changes should be counted toward the social benefits of a tax regulation, whereas other revenues (what economists refer to as “mechanical transfers”) should not. For non-revenue costs and benefits, cost-benefit analysis of tax rules should proceed along the same lines as that for non-tax regulations.

We flesh out this argument in greater detail and address potential objections and complications in the paper, a draft of which is available upon request. For now, a simple example to illustrate the approach: Imagine two sectors of the economy — a taxed sector and an untaxed sector. Individuals can work in either. Let’s say that the tax rate in the taxed sector is 30%. Workers will seek to allocate their labor between the two sectors so that they are indifferent between working an additional hour in the taxed sector and an additional hour in the untaxed sector. At the margin, an individual will be indifferent between, say, working an additional hour in the taxed sector (and thereby earning a $100 pre-tax wage) or working an additional hour in the untaxed sector (and thereby earning $70 that goes untaxed).

Now imagine that a new tax rule makes working an additional hour in the taxed sector ever-so-slightly more attractive than it was before (e.g., by slightly lowering the rate, or by slightly reducing compliance costs). An individual who responds by reallocating an hour of labor from the untaxed sector to the taxed sector is, to a first approximation, no worse off than before, because she previously was indifferent between the two options. But the total resources available to society increase by $30, because the individual is now paying $30 more in taxes than she otherwise would have. Because real resources have increased, this additional revenue resulting from a behavioral change should count as a social benefit. Symmetrically, any loss in revenue resulting from a behavioral change should count in the social welfare calculus as a cost.

Read more at Whatever Source Derived