Harvard law professor Cass Sunstein writes a lot of books, but he says his latest, The Cost-Benefit Revolution, is special.
It’s the culmination of decades of writing and research Sunstein has done in administrative law, with particular attention to the way that agencies like the Environmental Protection Agency or the Food and Drug Administration translate laws into rules and regulation. Sunstein has been a vocal advocate of having agencies quantitatively compare the benefits of those rules and regulations to their costs, junking rules that don’t pass the bar and speeding through ones that do.
It’s an ür-technocratic approach to governance, one that, as Sunstein describes, has become been adopted by much of the federal government due to pushes by the Reagan, Clinton, and Obama administrations. Sunstein himself served as administrator of the Office of Information and Regulatory Affairs (OIRA), the White House’s hub for cost-benefit analysis, under President Barack Obama from 2009 to 2012. He was perhaps the avatar for Obama’s overall approach to domestic policy: careful, cautious, and deferential to experts and quantitative research.
That approach earned Sunstein considerable scorn from some progressives, who thought the process wound up junking or delaying invaluable life-saving regulations. More broadly, Sunstein’s philosophy represents a commitment to technocracy that a growing number of liberals and leftists are increasingly rejecting as politically naive at best, and suicidal at worst given the realities of asymmetric polarization and Republican interest in minority rule. Why strand your own policies in a procedural quagmire of your own making when the other side just does what it wants when its party is in power?
With that critique in mind, I talked to Sunstein about whether adding up the costs and benefits of regulations is a workable way to run a government, whether the politics can ever work, and if the Trump administration is undermining the “revolution” his book documents. The transcript has been lightly edited for length and clarity.
Cost-benefit analysis attempts, in a way, to turn fights over policy into fights over concrete facts. In the book’s preface, you write, “the issues that most divide us are fundamentally about facts rather than values.”
That is a bold claim, and also one in tension with a lot of recent psychology and political science work. It sure looks like a lot of people pick a political team and adapt their beliefs on factual claims to better conform to the values of their group. Is that not the case?
Let me spell out my claim a little bit. If you could show that a certain approach to, let’s say, motor vehicle safety would save 700 lives annually and cost $8,000, it wouldn’t matter what your values are, if you’re sane. That’s a pretty good thing to do.
Suppose you could show that an approach favored by environmental groups would cost $60 billion and only modestly contribute to public health. It would be very hard, even if you’re a very fervent progressive, to think that’s a good idea.
On many of the issues that divide us, whether that’s clean water or endangered species or workplace safety, if we get clear on the facts, the value disagreement starts to seem uninteresting. I saw in Washington in my job, time and again, immersion in the facts made people’s political convictions look a little like background noise.
Working with Republicans on food-safety issues, if the consequence of a regulation was to save a lot of lives and not cost much, at least behind closed doors they’d say, “Go for that.” Among people on the left, if it turned out that a regulation on occupational safety would hammer the construction industry and have modest effects on safety, people committed to workplace safety would, behind closed doors, say, “Let’s not do that one.”
Often, immersion in the facts often makes value disagreements feel much less relevant.
It’s true that people often use value cues as a basis for judgment, especially when they don’t know about the facts. You might think, “I’m a Republican and if the EPA wants to go hard after greenhouse gases, I don’t want that.” You might think, “I’m a Democrat, if this is going to regulate banks, I’m for that.” That use of value cues and identity cues as a mental shortcut for dealing with hard problems is real, and it’s important.
But it is less real once people spend the time to think about — this bank regulation is going to have these beneficial effects, and these harmful effects, and often the value disagreement will dissipate. On some issues, like abortion or gun control, it might be that even once people are clear on the facts, they weigh the effects differently, based on their competing values. But across many different terrains, value disagreement precedes engagement with the facts.
You write that the “revolution” you’re talking about “does not involve the actions of the national legislature.” If we’re limiting it to executive agencies then, operating under relatively limited legal authorization from Congress, is there a risk that this approach produces inherently conservative outcomes?
I don’t think it has a conservative bias. If it ends up going in a conservative direction, that tells us something important. What it tells us is that the conservative view is correct.
If something you think about clean water is in apparent tension with the fact that what you favor would impose massive costs and do very little for the environment, it may turn out that the so-called conservative view is correct. On many issues, the cost-benefit analysis will turn out to support the conservative view. Whether you support Democrats or Republicans, you should learn something from that.
The idea that there’s a conservative bias, I’m not sure what that means. Republican administrations have gone forward with regulations that pass cost-benefit analysis and Democratic presidents, if the benefits are high or the costs are low, the process has a pro-regulatory bias.
I think the way of approaching the world that looks through conservative or liberal bias is destructive. We want to have a welfare bias. What is helping people? If it’s helping people, let’s all march together on its behalf.
Doing cost-benefit analysis well relies on having good evidence, and often we don’t. Social science, even medical and so-called “hard” science, often doesn’t replicate. Things we thought were true five years ago turn out to be false, or uncertain.
And funding mechanisms behind research — maybe the research comes from ideological think tanks, which biases the results, or it comes from industry-funded researchers and is likely to, say, low-ball the costs of automobile emissions or the health effects of cigarettes — can make the research an even worse proxy for the truth.
Is there a chance the epistemic case could be so severe that cost-benefit analysis ceases to be helpful, on net?
There’s a really important point there, and I’ve gotten more taken with it in the past three years.
The point is that cost-benefit analysis often suffers from what followers of Hayek call the knowledge problem. This is emphasized by conservatives who love Hayek, but also emphasized by people who aren’t conservatives and don’t love [20th-century economist and philosopher Friedrich] Hayek, and it’s really important. Cost-benefit analysis is based on a series of predictions and they might be wrong. The question is what you do about that.
The naive view is “use the best available evidence, conscious that it might turn out to be wrong.” The slightly less naive view is “build up processes for epistemic updating.” President Obama was very clear on the importance of updating evidence, in his charmingly named Executive Order 13563, and to his credit President Donald Trump is also on that train. He’s using epistemic updating to eliminate regulations that aren’t justified, rather than both eliminating and extending or expanding regulations where greater stringency is called for.
I think some people on the left think we underestimate the benefits of regulation and overestimate the costs, so cost-benefit analysis is inadequate or it needs to be accompanied by a discounting of the costs we see or a boosting of the benefits we see. The data on the past 30 years doesn’t justify that broad conclusion. Such data as we have suggests that agencies do err — sometimes in understating and overstating both costs and benefits — without a systematic skew toward benefits understatement.
One example you give for cost-benefit analysis leading sometimes to more regulation is Reagan’s embrace of ozone regulation. That seems like another era of Republican: a recent OIRA [Office of Information and Regulatory Affairs] report estimates that the benefits of Obama’s regulations wildly outstripped the costs, but Trump is still trying to undermine fuel economy standards, the Clean Power Plan, and so forth. Could a cost-benefit analysis ever persuade a skeptical Republican administration again?
If you have a president who’s, let’s say, on the left, and determined to do stuff whatever the cost-benefit stuff shows, that’s a problem, and the only thing that can be done is for people in the executive branch to take a stand. If you have a Republican administration determined to take things away or not to do things, that’s formally the same problem.
My observation from afar says that the Trump administration in some ways has been admirably committed to cost-benefit analysis, which is why the assault on the administrative estate has been less severe than widely reported.
At the same time, there are areas where the Trump administration has been less careful in consideration of costs and benefits than it ought to be. The greenhouse gas area is one. Another appears to be the coming alteration of the mercury regulations, which at least according to published reports is not based on objective cost-benefit analysis.
If that’s happening, then we have a problem. The Bush administration did approximately 2,500 final regulations and on cost-benefit grounds, it’s an impressive record of environmental protection and cost reduction. I’m compromised here, because I was there, but the Obama administration’s record was really good.
Saying Bush had an impressive record of environmental protection is a pretty bold claim! And especially so given that he only regulated greenhouse gasses after he was sued by environmentalists and liberal states and the Supreme Court forced him to do it. It seems somewhat glib to just add up the number of regulations as a metric. Some regulations could mean the difference between a habitable planet and an uninhabitable one.
On greenhouse gases, I meant something specific: the social cost of carbon under Obama went through a very rigorous and technical process that was based on the best available science and economics and came up with $36.
The Trump administration has changed it to $1 or $6, and it’s very, very hard to defend the slashing of the social cost of carbon in non-political terms. It’s possible [to defend it]. Part of the reason for the slashing is the Trump administration is only costing domestic costs of carbon whereas Obama counted worldwide, but the Trump administration hasn’t adequately explained its decision not to use the global number and the Obama administration offered a detailed account of why they used the global number. [The Trump administration] just haven’t answered that. Its failure to provide [a reason] is highly unusual and disturbing. It suggests that this is a decision that might not fit within the bounds of reasonableness.
The Bush administration didn’t proceed against climate change, and the Supreme Court ruled against them, but that they failed to do that does not fall outside the bounds of reasonableness, in my view.
One area where smart lefties I know are particularly skeptical of cost-benefit analysis is on financial regulation. They see it as fundamentally a mechanism for banks and other interest groups to poison the regulatory process and weaken safeguards (and as inappropriate when financial rules are, in a sense, creating the market they’re in turn regulating).
Even if you don’t see it that way, it raises a big question about evidence. Climate science is a real science that is in some sense making progress. I don’t know that that’s true at all about macroeconomics; this year’s Nobel laureate Paul Romer has argued the field actually moved backward in the last three decades. How do you do good, effective cost-benefit analysis when the evidence in question comes from a field with a pretty poor predictive track record about the effects of regulation?
So for something that’s trying to reduce the risk of a meltdown, sometimes the best you can do is specify the cost of the rule, and figure out what the benefits would have to be to realize those costs. One simple exercise is: Does this produce a 1/x (where x is the cost) reduction to the risk of a meltdown? If it’s 1/10,000, that’s not nothing, given the magnitude of a possible meltdown. You can do lower and upper bound projections and do basic arithmetic.
It might seem arbitrary or a little ridiculous. But to do this is better than do nothing. To say a capital requirement is going to impose an arbitrary cost, but it reduces the odds by 1/5,000, then it’s well worth it, given the cost of a meltdown. That kind of exercise can help you be assured that something is reasonable, or be clear that something isn’t reasonable.
This book comes out at a time when I think a lot of people, myself included, are rethinking the value of a technocratic approach to politics. Your Harvard law colleague and past co-author, Adrian Vermeule, who’s abandoned liberalism entirely in recent years for a kind of far-right political Catholicism, has a pinned tweet that reads, “When the liberals are in power, we ask them for freedom because that’s their principle; and when we are in power we take theirs away, because that’s our principle.”
I hear that sentiment a lot from both sides — that liberals and leftists are being hampered by the generosity of their principles, that they need to learn to play dirtier and can’t afford to hamstring themselves by, say, subjecting their policies to cost-benefit analysis when faced with a figure like Trump who’s not hamstrung by anything. Why should liberals hold onto technocracy in that situation?
Let’s suppose there’s a Republican administration that is not doing real analysis, fair analysis. It’s just, let’s stipulate, being foolish, or it’s refusing to attend to costs and benefits, or it’s in the grips of interest groups. I think even if that’s the case, a Democratic administration should still embrace cost-benefit analysis, at least as the presumptive rule.
You don’t want food-safety regulations that are hammering the economy but not benefiting people much. You don’t want an environmental regulation that has adverse effects on businesses but does little to help public health.
The answer is that the right stuff is what cost-benefit analysis says is the right stuff.
The book is not an unambivalent celebration of the revolution. We’ve referred to distributional factors, which matter, and the knowledge factor, which matters. Cost-benefit analysis is the best way of capturing human welfare right now. But we’re increasingly learning that something might have increasingly high costs and not as high benefits, but it might make people’s lives better regardless. Something Trump just endorsed, which is a tribute to this, is mandatory cameras in cars so you can see behind you. That saves lives, and driving is a lot easier. You can see if you’re about to hit something.
A we/they attitude toward policy is not helpful. The right attitude toward policy is “what’s this policy going to do.”
Sign up for the Future Perfect newsletter. Twice a week, you’ll get a roundup of ideas and solutions for tackling our biggest challenges: improving public health, decreasing human and animal suffering, easing catastrophic risks, and — to put it simply — getting better at doing good.