Whenever the Trump administration proposes to eliminate a regulation, many people are tempted to give it a standing ovation. Many others are tempted to boo and hiss. Sometimes it’s right to do one or the other. But with respect to the Environmental Protection Agency’s recent decision to rethink its controversial mercury regulation — well, it’s complicated and unusually interesting.

The story begins in 2012, when the EPA finalized that regulation under President Barack Obama. (I was administrator of the White House Office of Information and Regulatory Affairs at the time, and involved in the discussions that led to the final rule.) The EPA’s analysis proceeded in two stages.

In the first stage, it concluded that regulating mercury was, in the words of the Clean Air Act, “appropriate and necessary” in light of mercury’s adverse effects, especially on children (who, among other things, can lose IQ points as a result of exposure to mercury in fish, ingested by their mothers). In the second stage, it concluded that its final rule would have benefits far in excess of costs.

To be sure, compliance with the rule would be unusually expensive, with annual costs of $9.6 billion. But it also theoretically could prevent up to 11,000 deaths annually. In monetary terms, the benefit of doing that would be $37 billion or more a year — far in excess of the $9.6 billion annual expenditure.

Importantly, the EPA emphasized that the actual benefit of smaller mercury emissions would be modest — just $4 million to $6 million a year. The overwhelming majority of the estimated public health benefits, including those 11,000 prevented deaths, would not involve mercury at all. They would be “co-benefits.” The technologies companies were likely to use to reduce mercury would be expected to simultaneously reduce emissions of particulate matter, another air pollutant, which produces serious public health problems.

States and private companies immediately challenged the EPA’s regulation in federal court. The Supreme Court sided with them on an important but relatively narrow ground. When the EPA decided, at the first stage, that it was “appropriate and necessary” to regulate mercury, it looked at benefits only and failed to consider costs, or to undertake any kind of balancing. The justices sent the regulation back to the EPA to do exactly that.

Under Obama in 2016, the EPA responded with two arguments. First, it said that the provisions of the Clean Air Act that govern hazardous air pollutants, such as mercury, are primarily concerned with public health and not cost. For that reason, the EPA said that it would go forward with its regulation even if the expected health benefits were low and the monetary costs were very high.

Second, the EPA said that even it was required to balance benefits and costs, it would go forward anyway. Because of the sheer magnitude of the expected co-benefits, it was “appropriate and necessary.”

Last month, the EPA proposed to reject both of the previous administration’s conclusions, and sought public comments on a radically different approach.

First, it denied that it would be “appropriate and necessary” to list a pollutant if the costs were much higher than the benefits. Second, it announced that it would not consider co-benefits at all. It emphasized that separate provisions of the Clean Air Act are specifically designed to control particulate matter.

But the EPA did not propose to eliminate its mercury regulation. A federal court has ruled that to do that, the EPA would have to go through a specific process. With the court’s ruling in mind, the EPA said that it did not intend to take its regulation off the books.

The agency almost certainly had a pragmatic reason in mind. Coal companies have already taken steps to comply with the mercury regulation. Having adjusted to it, they are not asking for it to be taken away.

On that count, the EPA deserves applause for its pragmatism. If a regulation is expected to save thousands of lives annually, and if its industry has already incurred many or most of the costs, there’s not much point in eliminating it.

In saying that it would insist on balancing costs and benefits, the EPA also got it right. Suppose that a regulation would cost billions of dollars and do little to protect public health. That would be a pretty silly rule.

It is true that the agency might not be able to quantify the benefits of some environmental regulations. But the EPA did not deny that proposition. It said, more simply, that if the costs dwarf the benefits, it will not go forward. That makes a lot of sense.

But the EPA blundered in saying that it would not consider co-benefits. The Clean Air Act can be read to require that conclusion, but that’s not the best reading of the law. It’s ambiguous, which means that the EPA gets to take account of co-benefits if it wants. In principle, all costs and all benefits however distantly related should be counted. A decision to disregard some of them is arbitrary and unreasonable.

There’s a larger point here. The Trump administration is giving a lot of attention to the costs of regulation, and it deserves real credit for that. But it is sometimes neglecting possible benefits of regulation, even in cases when public health is on the line. That’s a big mistake, a form of regulatory malpractice.

Cass Sunstein is a Bloomberg Opinion columnist.

Load comments