Donald Trump promises to impose, soon after his inauguration, a new requirement on federal agencies: If they want to issue a new regulation, they have to rescind two regulations that are now on the books. The idea of “one in, two out” has rhetorical appeal, but it’s going to be extremely hard to pull off.

In the abstract, of course, it sounds like a gimmick, and it’s a pretty dumb idea. As presidents from Ronald Reagan to Barack Obama have recognized, the real question is whether regulations, whether new or old, are justified. That requires a careful analysis of their costs and their benefits.

For some agencies, the right approach might be “zero in, 10 out,” because there’s no justification for anything new and a lot has to go. For other agencies, the right approach could be “10 in, zero out,” because all 10 have benefits well in excess of costs, and there’s really nothing to eliminate.

It follows that the right approach is not “one in, two out” but a careful check on issuing new rules, with the help of cost-benefit analysis — accompanied by an ambitious program to scrutinize rules on the books to see if they should be scrapped. The Trump administration doesn’t need a gimmick to make progress on both fronts.

But life isn’t lived in the abstract. It’s reasonable for the Trump administration to reduce regulatory activity, certainly in some domains, and the “one in, two out” idea is likely to deter new rulemaking - which may be the main goal. It’s also reasonable to create a strong incentive for agencies to get rid of unjustified rules. On that count, “one in, two out” — a variation on an idea that has been tried in Canada and the U.K. — might be a lot less random than it seems.

Even if that’s so, Trump’s appointees are going to face some serious implementation problems. Four categories of regulations should probably be exempted — and if that happens, the “one in, two out” rule will have a pretty narrow scope.

1: Some regulations are deregulations; they reduce existing burdens. For example, the Environmental Protection Agency might exempt small businesses from an air pollution rule. By the very logic of the “one in, two out” rule, deregulation should not be covered.

2: Agencies regularly issue “technical amendments,” responding to a request for clarification, correcting an inadvertent error, or allowing a bit more flexibility (say, by allowing less frequent filing of reports). It would be a big mistake to deter technical amendments by requiring them to be accompanied by “two out.” (True, the line between a technical amendment and a substantive change is sometimes elusive.)

3: Numerous regulations impose no significant costs on the private sector or anybody else — for example, they might provide valuable information or combat fraud, waste, and abuse. A recent one would allow the Agricultural Marketing Service to remove “board members and staff who fail to perform their duties or who engage in dishonest actions or willful misconduct.” Another recent one “establishes the opening date for all Atlantic shark fisheries.” Yet another proposes “to amend procedural regulations governing the filing and delivery of documents to allow for electronic filing of documents.” Do we really want to deter agencies from issuing rules like those?

4: Numerous regulations implement federal spending programs. They might provide hurricane disaster assistance, establish rules for disability insurance for soldiers, or make changes to the program that offers payments for end-stage renal disease. Sure, the Trump administration might want to scale back some of these programs. But so long as that spending is required by law, the rules that make such spending possible seem a poor fit for “one in, two out.”

So what’s a good fit? The best answer points to a subcategory of what the executive branch calls “economically significant” rules — those that impose $100 million or more in costs on the private sector. Most administrations issue dozens of such rules every year. This is the most workable version of Trump’s proposal: Whenever a regulator is imposing significant costs on people (either $100 million or some lower threshold), it must take two regulations away at the same time.

But that raises further questions. Which regulations? Would any do? Surely not. If you’re imposing $1 billion in costs but eliminating two regulations whose total cost is $10,000, you get “one massive rule in, two tiny rules out” — hardly what Trump has in mind. What most matters is the cost of rules, not the number. For the private sector, a dozen inexpensive rules may not be a big deal, while two that cost over $1 billion might be a horror show.

So it’s tempting to say that whenever a regulator is issuing an expensive new rule, it must eliminate two that cost at least as much as it does. The problem is that within the next year, one agency or another (headed by a Trump appointee, after all) is going to want to issue a rule that costs more than $100 million (say, to reduce the risk of illegal immigration, to protect against violent crime, or to promote safety on the highways), and it just won’t be able to come up with two expensive rules that it makes sense to eliminate.

The best way to solve that problem is for Trump to give someone in the White House — most likely, the head of the Office of Information and Regulatory Affairs — the authority to waive the “one in, two out” idea after a compelling showing of need. Or perhaps the idea of “one in, two out” could be understood to apply government-wide, and not be imposed on particular agencies.

The conclusion? In theory, “one in, two out” is silly, and in practice it’s likely to be a bit of a mess. It’s hardly the most sensible approach to regulatory reform. But with a little flexibility, and a lot of determination, executive branch officials might be able to make it work.


Sunstein is director of the Harvard Law School’s program on behavioral economics and public policy.