# 13 Dr. Reed Walker – Estimating the Marginal Cost and Benefits of U.S Regulations
In this edited transcript, Berkeley professor Reed Walker discusses his American Economic Review paper with Joe Shapiro on the costs and benefits of U.S. air-pollution regulation, using Clean Air Act offset markets to infer marginal abatement costs, and why the results suggest regulation is often too lenient on the margin. We also touch on his Journal of Political Economy paper on the long-run consequences of cleaner air for children’s adult earnings.
Arvid Viaene: Air pollution has fallen dramatically in the United States over the past half-century—but a harder question remains: are we regulating it efficiently today, at the margin? In other words, when we push pollution one step lower, do the added health and welfare benefits outweigh the added costs to firms and the economy—or have we reached the point where extra cleanup is no longer worth it?
This episode revisits that question through one of the most clever data settings in modern environmental economics: Clean Air Act offset markets, where firms buying the right to expand emissions must pay other firms to reduce them. Those transactions reveal real-world, marginal cleanup costs—market-by-market, year-by-year—allowing a direct comparison to leading estimates of marginal damages. The headline finding from the research Reed co-authored with Joe Shapiro: in most markets, marginal benefits still far exceed marginal costs, implying regulation is often too lenient on the margin—with notable exceptions like Houston.
If you listened to the earlier episode with Joe Shapiro, you’ll recognize the core idea and the same paper. But this conversation is different in emphasis. With Reed Walker, we go deeper into the economic intuition of revealed-preference cost measurement, why cost curves can still look surprisingly “low” even after decades of cleanup, and why certain places—especially Houston—break the pattern. We also broaden out beyond this one paper: Reed explains how cleaner air can shape outcomes decades later, drawing on his research linking Clean Air Act-driven improvements around birth to higher adult earnings.
My guest is Reed Walker, professor of economics and public policy at UC Berkeley and faculty co-director of the Opportunity Lab’s Climate and Environment Initiative. Reed’s work sits right at the intersection of environmental regulation, health, and economic opportunity. With that, let’s dive in.
Arvid Viaene: I’m very excited because today Reed Walker is joining me. Reed is a professor of business and public policy and economics at UC Berkeley. His research explores the social cost of environmental externalities—such as air pollution—and how regulations to limit these externalities contribute to gains and/or losses to the economy.
He is the faculty co-director of the UC Berkeley Opportunity Lab’s Climate and Environment Initiative. He’s also a research associate at the Energy Institute at Berkeley, a faculty research fellow at the National Bureau of Economic Research, and a research fellow at IZA. He received his PhD in economics from Columbia University.
So Reed, welcome to the podcast.
Reed Walker: Thanks for having me.
Arvid Viaene: I’m very excited to talk about air pollution because it’s one of my favorite topics, and I think it doesn’t quite get enough attention anymore. For those who didn’t listen to the episode with Joe Shapiro—or who can’t quite remember—could you give us a quick summary of your paper and what you found?
Reed Walker: Sure. Joe Shapiro and I have a recent paper in the American Economic Review exploring the costs and benefits of environmental policy in the United States—specifically, the efficiency of regulation.
When we think about efficiency from an economic perspective, we’re interested in whether the cost of regulation on the margin—the incremental cost of an additional unit of regulation—is equal to, larger than, or smaller than the benefits of that same unit of pollution reduction to society.
Quite naturally, you could imagine a situation where a regulation costs a billion dollars per ton to reduce air pollution. If the social benefits of that same unit of reduction were only a hundred dollars per ton, we might say that’s not really worth it from a social perspective. It’s way too costly for the benefits. So we’re trying to understand: on the margin, are the social benefits of better air equal to the private and social costs of regulating air pollution? That’s hard to answer because costs are often unobserved, and benefits are hard to measure too.
We explore a specific program in the United States under the Clean Air Act that requires firms that want to locate in certain areas to purchase permanent rights to emit pollution in those areas. Those purchases are useful because they let us get closer to understanding compliance costs. The intuition is similar to cap-and-trade markets: transaction prices can, through revealed preference, tell us something about how firms perceive the cost of compliance.
If the market price for pollution rights is extremely expensive—say, a billion dollars per ton—then a firm that wants to locate in the region will look for other technologies or ways to reduce its own pollution, so it doesn’t have to buy those expensive rights. But at some point, the firm’s own abatement becomes too expensive. If marginal abatement costs are increasing, then once they get high enough, and the firm still wants to locate there, it will buy the remaining amount it hasn’t abated on the open market. That tells us something about marginal incremental cost.
On the supply side, these permits come from incumbent firms. If an incumbent sees a market price of, say, a billion dollars per ton, and it can abate for a hundred dollars per ton, it might say: “I can make money here.” It can permanently reduce emissions, certify those reductions with the regulator, and sell them into the market—earning a profit.
So these supply and demand forces reveal something about costs on the margin. We then connect that to the best available estimates in the literature of the benefits of pollution reductions—on the margin—for specific pollutants.
One last point: we’re studying local pollutants, not global pollutants like greenhouse gases. The Clean Air Act regulates “criteria pollutants,” where damages are fairly localized. When we emit a ton of particulate matter, damages disperse, but they’re still relatively local. That means damages can be very heterogeneous depending on where emissions occur. A ton of pollution in New York City has much higher social costs because of population density than a ton emitted in, say, the Nevada desert.
So efficiency for criteria air pollutants is local: where damages are high, regulation should be more stringent. These offset programs exist in many cities across the U.S. and allow us to observe abatement costs on the margin in different places—LA versus Houston versus New York—and compare them to leading estimates of marginal benefits. That’s the paper in a nutshell.
Arvid Viaene: Awesome—thank you for that. I really like that explanation. The title of your paper is Is Air Pollution Regulation Too Lenient? Could you talk about what you find in terms of results? You compare the marginal costs to the marginal benefits—could you give an overview of the general findings?
Reed Walker: Big picture: I had a very diffuse prior going into this study. Air pollution in the United States has improved dramatically over the past 50 years—we’ve seen something like 90% improvements in ambient air quality relative to 1970 levels.
If you think abatement costs increase with the amount of abatement, you might wonder whether costs on the margin are now extraordinarily high relative to the benefits—whether that last 97–98% reduction is “worth it,” so to speak. But what we find is kind of the opposite. The marginal abatement cost identified in these offset markets is almost 10 times lower than our leading estimates of marginal social benefits, on average.
What’s cool is we can look across different markets and ask whether any places look over-regulated or under-regulated. By “over-regulated,” we mean a place where marginal abatement costs are much higher than marginal benefits. “Under-regulated” is the opposite: benefits are enormous relative to costs.
For the vast majority of markets, regulation appears too lenient—we could ratchet down further and move up the marginal abatement cost curve until marginal costs and marginal benefits are equal. But there are some markets—Houston in particular—where, on the margin, compliance costs look higher than benefits. There are some unique reasons for that, related to both costs and benefits.
Arvid Viaene: I want to come back to Houston because it’s a really interesting example. Overall, you find the marginal benefits of reducing air pollution are still substantially higher than the marginal cost of abatement. When you presented this to academics—or policymakers, or people in industry—were they surprised?
Reed Walker: I don’t know that I got very strong reactions, but I think people found it interesting that we now have a proxy for costs on the margin, and that we can estimate them in different places and at different points in time.
In the literature, it’s hard to point to a paper that estimates marginal abatement costs for criteria pollutants broadly. If it exists, it’s often for a specific industry in a specific location. So the ability to look across markets and over time is useful. It also raised questions for me about long-run versus short-run abatement cost curves. How can it be that we’re at roughly 90% reductions in pollution and marginal abatement costs still don’t look that high?
The answer has to involve technological change, induced innovation, or other ways to squeeze out further reductions at lower cost. But there hasn’t been a lot of compelling research on that, in my view—and it’s quite interesting.
Arvid Viaene: In the paper you also mention engineering estimates, like the ones the EPA has used. Did you find a noticeable difference versus those engineering estimates?
Reed Walker: There are engineering estimates for compliance costs that the EPA has developed internally. We looked at the software, the inputs, and talked to people at the EPA who use and develop these tools.
Ultimately, we came away dissatisfied with the quality of the estimates and the inputs being used. And if you ask EPA officials directly, they’re also unhappy with their ability to estimate these costs and keep databases updated. It’s incredibly complex. Information is scarce, resources are scarce. They’re trying to estimate the cost for a specific facility to install a specific technology—yet every facility is different.
There may be additional re-engineering costs that are hard to observe. And there are very few people trying to quantify engineering costs for essentially every facility in the United States. It’s just not practically feasible.
Arvid Viaene: Let’s come back to Houston. You mentioned it as a unique case where, for parts of the sample period, marginal costs can exceed marginal benefits. What’s going on there?
Reed Walker: Houston is a non-attainment area. In non-attainment areas—places out of compliance with the Clean Air Act—there’s a hard cap on stationary-source emissions of these pollutants. That means there can be no net increase in emissions. If a new facility wants to locate there, it needs a willing counterparty in Houston to offset emissions so the new facility doesn’t lead to a net increase, because Houston is out of compliance.
Houston is also a major petrochemical and refining hub. It’s close to the West Texas shale boom and can refine and process natural gas and oil from the Permian Basin.
With the advent of hydraulic fracturing—fracking—and strong production from West Texas shale, demand for expanded refining capacity skyrocketed in Houston, under a fixed emissions cap. When you get a huge increase in demand under a fixed cap and a fairly inelastic supply of offsets, prices spike. You see that in the data.
When prices spike, suppliers may get interested—thinking they can make money by supplying offsets—and you can see prices fall later, potentially reflecting supply responses. But the spike is very clear.
The benefit side is also unusual. The prevailing wind pattern is generally west to east. And east of Houston is largely the Gulf of Mexico (or Gulf of America, if you want to call it that). Fewer people are exposed downwind compared to other dense urban corridors, so damages can be somewhat lower than you might expect for a major city—because a lot of emissions get blown out over the ocean.
So Houston is a case where costs were high because of demand pressure from expanded petrochemical activity, while benefits were comparatively smaller because of where emissions tend to travel.
Arvid Viaene: Do you think Houston is unique, or could we see this elsewhere?
Reed Walker: I think it’s possible elsewhere. You need (1) a big expansion in demand and (2) fairly inelastic supply of offsets/permits. The inelastic supply is also related to how long an area has been in non-attainment. If it’s been in non-attainment for decades, a lot of the low-hanging fruit on the supply side has already been picked.
But if an area falls into non-attainment more recently, there may be a lot of low-cost abatement options—so more potential supply. So you could imagine similar scenarios in other locations if demand rises sharply under a fixed cap and supply is limited.
Arvid Viaene: Let’s talk about the benefit side. How did you estimate the benefits of air-pollution reductions?
Reed Walker: Joe and I did something deliberate: we didn’t estimate benefits ourselves in this paper. Estimating the benefits of air-pollution reductions is difficult, context-dependent, and sometimes controversial. Instead, we took “off-the-shelf” estimates from the frontier and compared them to our compliance cost estimates.
By “the frontier,” I mean integrated assessment models—most notably Nick Muller and collaborators’ AP3 model. These models combine inputs from economics and atmospheric science and then monetize damages.
The basic idea is a kind of source–receptor framework: if you emit a ton of pollution from a smokestack of a certain height in a certain place (say LA), the model uses information about wind and atmospheric processes to predict where it travels and where it lands.
Then it uses concentration-response functions to map changes in pollution exposure into health outcomes—morbidity and mortality—and then uses values like the value of a statistical life to monetize damages.
It’s a lot. But the technology portable: you can run simulations like “What’s the marginal damage of emitting a ton of PM in Seattle?” The model can include mortality only, mortality plus morbidity, and in principle could incorporate other channels like crop yields—depending on what’s included.
Arvid Viaene: That bridges nicely to your other work on health impacts. I wanted to ask about your 2017 Journal of Political Economy paper, “Every Breath You Take, Every Dollar You’ll Make: The Long-Term Consequences of the Clean Air Act of 1970.” What were you trying to do there?
Reed Walker: This started back in graduate school with fellow graduate students, when we were interested in an emerging literature—originating in epidemiology and picked up by economists—around what’s called the fetal origins hypothesis.
The idea is that the in utero environment can be particularly important and predictive for long-run wellbeing and later-life outcomes. There’s evidence from different contexts—famines, malnutrition, and other shocks—suggesting that cohorts exposed in utero can look different decades later in outcomes like IQ and other measures.
We were interested in exploring this in the context of air pollution, partly because there was work suggesting air pollution affects infant health—like foundational studies on the Clean Air Act and infant mortality. So we asked: if exposure to cleaner air around birth improves infant health, do those cohorts look different 30 or 40 years later?
The key for progress was data. We found data infrastructure developed at the Census Bureau, with linkages to the Social Security Administration, that allowed us to observe where people were born (county of birth), their exact date of birth, and link that to adult earnings from tax-related forms. That linkage let us take an existing research design used in earlier Clean Air Act work and connect it to outcomes measured 20–40 years later.
Arvid Viaene: What magnitude of impact did you find?
Reed Walker: From what I remember, it was on the order of $200 to $400 per year in earnings at around age 30. Cohorts born when air cleaned up substantially—relative to similar cohorts born in the same year but in places where air quality didn’t improve as much—had adult earnings that were a bit higher in that range. There’s a lot more in the paper about other outcomes and potential mechanisms, but that’s the basic magnitude.
I don’t think it’s an average effect for everyone, though. It’s more like an average of some people being severely affected and others not affected at all. But when you aggregate across the population, the total impacts can be large.
Arvid Viaene: One of the things that really struck me when reading this is, and I’ll even just quote from it: “Nevertheless, our estimates suggest that the long-run welfare cost of exposure to environmental toxins as measured by lifetime earnings losses may be as large or larger than the monetized costs of death based on short-run impacts of infant mortality, examined previous research.” So I imagine it’s just purely because of the scale of people being exposed to this, that it adds up to just this big amount.
Reed Walker: Yeah, I think that’s right. And I think that’s true more generally. There’s some interesting review work by people like Matt Neidell or Josh Kraft-Ziven, thinking about like the non-health cost of environmental externalities, maybe in terms of labor productivity or or other outcomes that are not directly, that may you know, tied to health, but are not direct health outcomes. And these are kind of typically on the margin, like a little bit smaller effects. But given the scale and the breadth of these impacts on society, when you start kind of wrap rolling these things up and and aggregating, they they become quite substantial.
Arvid Viaene: Joe told me you were the one who reached out to him about offsets—like, “Have you heard of offsets?” Finding the dataset is often a huge part of answering the question. Do you have any advice for researchers on finding the right data?
Reed Walker: I’ll give one piece of advice that’s a little flippant. Someone once told me: doing research is easy—you just need to know more about a topic than anybody else. And there’s a lot of truth in that. It’s hard for me to know more about returns to schooling than someone like David Card. But in energy and environmental economics, these can be relatively nascent literatures. You can read every paper on a topic quickly, and if you’re stubborn enough, you can dive into the institutional details and learn a lot.
Then questions start to pop up. Another approach is to think about what existing work couldn’t do—what held it back. In the fetal origins literature, for instance, some work used Scandinavian register data, and in the U.S. there were attempts using smaller datasets. The limited data held things back. So when we saw a county-of-birth variable linked to long-run outcomes, we realized it could open up a lot of questions people had been working on. There was also some luck: at Columbia at the time, people were actively developing the frontier of that literature, so there were ideas in the air.
One more thing: institutional detail really matters. But there’s an art to it. You can go so deep into the institutional weeds that your project becomes interesting only to one other person. The art is figuring out which part of the institution or research design is broadly interesting—something that transcends the niche and can support compelling causal inference.
Arvid Viaene: On that note—anything else you’d like to add, either on offsets or the JPE paper? Any takeaway we haven’t emphasized?
Reed Walker: Just that research begets research. These projects raise lots of questions that don’t necessarily fit in the original paper. When people ask where research ideas come from, a lot of the time it’s from other research projects. The hardest part is getting started. Once you start building an agenda, ideas begin to “fall out of the sky,” because you’re deep in the weeds and learning a lot—about tools, institutions, and the topic. It doesn’t have to be perfect at the beginning—just get started. Sometimes lightning strikes in a good way, and new ideas emerge.
Arvid Viaene: Amazing. Thank you so much for coming on the podcast.
Reed Walker: Yeah—thanks for having me. I look forward to listening and learning.