How Much Will the Inflation Reduction Act Reduce Emissions?

This interview was originally released on August 8, 2023. The transcript of the conversation has been edited for length and clarity.

Daniel Raimi: Let’s talk about the study that the two of you coauthored, which was published in the journal Science. Listeners of our show will probably have a good idea of the main provisions of the Inflation Reduction Act (IRA), so we don’t need to give a lot of background on the of bill itself. Can you tell us about the new study?

Nicholas Roy: When the IRA was released last year, a lot of different research teams came out with modeling studies to represent what they believed the bill would do. What our study has achieved is seeing the commonalities among all these different studies. Usually, with policy analysis, we don’t really get the opportunity to compare such a broad set of studies.

Our paper has nine different teams involved, four of which were releasing studies on the IRA right after the bill came out. We were one of those teams—as well as the REPEAT project at Princeton, Rhodium Group, and Energy Innovation—that were trying to quantify the emissions impacts of the bill. We also included five other groups from national labs, government agencies, and universities.

This was all organized by John Bistline at the Electric Power Research Institute. He made this heroic effort of coordinating all our different teams with all our different assumptions, all our different data inputs, and all our different modeling frameworks to see where you can get these common outputs and compare direct apples to apples across the different models.

Maya Domeshek: I would say that what makes this new study unique is the ability to look across so many studies and try to figure out what we can learn that is or is not robust across studies.

What are some of the headlines that have come out of this work in terms of energy and emissions outcomes?

MD: The big takeaway from the paper is that the IRA is likely to reduce US emissions. That’s robust across all the models; they all found that there would be a fall in emissions. I think the range that we state in the paper is something like 33–40 percent below 2005 levels by 2030, which doesn’t quite reach the US goals in the Paris Agreement—which is 50–52 percent by 2030—but it’s on the way there and much better than in the absence of the IRA.

I think the second main takeaway is that most of those emissions reductions are coming from the power sector. We see pretty dramatic decreases in emissions in the power sector across the models: something like 47–83 percent below 2005 levels by 2030.

We also got more detailed results across models, looking at how many renewables are built out, how much consumption increases, what happens to coal and gas capacity, and more.

I think the last thing that’s particularly interesting is the question of electrification, because some of the models were representing the entire US energy sector, and they tried to look at what the uptake of the vehicle tax credits or other credits would mean for electrification. Some of the models projected what consumption might be in the future, whereas other models, like ours, made an assumption about what the consumption will be. We got this assumption about projected energy consumption from the Energy Information Administration’s Annual Energy Outlook that projected US demand for electricity back in 2021.

That’s great—those are the key outcomes for the energy system and for carbon dioxide emissions. How about costs, which is a key talking point that many people care a lot about? What happens to energy costs under the IRA? When we compare those costs to the benefits that the bill gives us, what does the cost-benefit ratio look like?

NR: We can think about costs in a bunch of different ways. One method, which we use in the paper, is to look at the total energy system and sum up all the costs from the modeling exercise and see what those costs look like prior to the bill and after the implementation of the bill.

When you have something like the IRA, which mainly uses subsidies to drive decarbonization, you subtract out those subsidies from the costs on the grid. That’s a big reason why we see a reduction in costs for retail prices, for example.

The prices of electricity are different from how much money is being spent by the government. The government is spending money on these climate provisions, and it’s raising money from somewhere else.

We’re concerned about this cost on the energy system. When we want to compare those costs on the energy system to the benefits from reducing these emissions, we have to find some similar or comparable metric. What we do here at Resources for the Future—as well as in environmental economics more generally and in benefit-cost analysis in the government—is consider what’s called the social cost of carbon.

Currently, the most up-to-date research says that the social cost of carbon is somewhere near $185 per ton of carbon dioxide that’s added to the atmosphere. The idea is that, with each additional ton of carbon dioxide emissions emitted into the atmosphere, society as a whole would incur a cost of around $185. If the cost per ton of carbon dioxide removed from the atmosphere, or not emitted into the atmosphere, due to this bill are lower than that value on average, then you could say that the benefits outweigh the costs.

What we find in this study is that, when we consider the cost in terms of dollars per ton of carbon dioxide that’s not put into the atmosphere, we find a metric of $27–$102 per ton as the range across the studies. That $27–$102 is a lot lower than the $185. We would say that the climate benefits of this bill far outweigh the costs on the energy system.

Older social cost of carbon models, such as the Obama administration’s, lands within the range of our costs, at just about the same as the benefits. If you want to go back to the Trump administration and look at their estimate of $7 per ton for the social cost of carbon, you would see that the climate benefits are not worth the costs of this bill.

But in summary, across all these modeling studies, we do see that the benefits far outweigh the costs of implementing this bill when using the most up-to-date studies on the cost of climate damages.

Just to clarify, we’re not going to go into the details on the social cost of carbon. Many parts of the economy will be damaged by climate change, but those costs are not accounted for in the current best estimates of the social cost of carbon. We’re also not talking about the social costs of other greenhouse gases associated with the energy system, like methane and nitrogen oxides.

MD: We’re also not including the other benefits of the bill, like the health benefits that we might expect from reduced fossil fuel usage.

How will the IRA affect electricity bills?

MD: The paper does not talk very much about electricity-price impacts, but almost all of the individual studies that contributed to the paper did look at price impacts, and so did we.

In another paper that we published in October last year, we found that the IRA is likely to decrease the price of electricity generation relative to what the price would’ve been in the absence of the IRA. The law also likely will decrease the volatility of the price of electricity, because the electricity sector as a whole is relying more on renewables and less on fossil fuels. Fossil fuels have notoriously volatile prices. For example, last year, when the price of gas went up due to the war in Ukraine, electricity prices all over the world also went up. A grid that’s more reliant on renewables would see less of that kind of impact.

We found a decrease in volatility and decrease in electricity-generation prices. That’s all happening because the government is subsidizing electricity effectively and moving us to an overall—in the long term—cheaper and cleaner grid. Whether that means cheaper or more expensive household bills is a separate question, because, first of all, electricity-generation price is not the same as the electricity price that households are paying, due to transmission and distribution costs.

Second of all, your bill depends on how much electricity you’re consuming. In fact, one of the things the IRA is trying to achieve is getting people to consume more electricity, because we’re trying to electrify the whole economy. Reducing the price of electricity makes it easier for people to consume more units of electricity. What this means for your bill, I don’t know; that remains to be seen. But even if bills go up due to electrification, the idea is that people’s higher electricity consumption will be counterbalanced by lower consumption of fossil fuels, ideally leading to a net reduction in expenditures. We do know that the implementation of the IRA is likely to decrease the price of electricity generation relative to what it would’ve been in the absence of the policy.

Electricity prices are set by the marginal cost of electricity generation in whatever region you’re in. Do you think that the marginal cost is more likely to be set by renewables in the future, or is the marginal cost still going to be set by gas? I’m imagining that renewables are going to generate whenever they can generate and supply energy at that margin. Can you talk more about how the IRA might affect the marginal price of electricity generation?

MD: You’re absolutely right that electricity prices, especially in regions of the country with a deregulated electricity sector, are set by the marginal unit that’s generating. The price that you’re paying over the course of the year, on average, is reflecting the marginal price in a bunch of different hours. The more hours that are shifted away from having gas as the marginal unit, the cheaper your overall average annual price is going to be. One would hope that lower demand for gas and other fossil fuels also means those prices are lower, so the marginal gas unit also is not as expensive.

NR: Something that I really liked that you pointed out, Maya, was the temporal aspect of the cost of generation. Something that economists, especially energy economists, have been interested in—especially those at Berkeley’s Energy Institute—is the idea of dynamic pricing.

The IRA subsidizes those renewable generating units, which means that, if you’re going to implement something like dynamic pricing in the future, where people basically can get different prices at different hours (which is being implemented already in some aspects by some utilities), you’d be able to get an even cheaper price during certain hours than you would if you’re averaging out the price across all hours. Dynamic pricing also enables more benefits for the climate, more benefits to people’s electricity bills, and a more efficient system.

MD: I also want to return to this original question about the impact of the IRA on household electricity bills. Because again, in our earlier study, we looked at the distributional impact of reducing electricity prices. We find that reducing electricity prices by subsidizing them with government funds is effectively a progressive (in the technical sense) policy, because you’re reducing the amount that households have to spend on a crucial good, and you’re paying for it with the tax system, which is somewhat more progressive than the flat quantity of electricity that most households are consuming.

That’s an important aspect and an important goal of the IRA: to keep costs down for households.

You and your colleagues in this analysis carry out excellent modeling work—but models inherently are limited representations of the real world, because you have to leave some considerations out of the analysis. What are some of the most important things you had to leave out, or you can’t model for one reason or another, and how do you think those omissions might affect the outcomes?

NR: I appreciate you talking about the limitations of modeling, because I think every modeler would say that they’re some of the last people to trust the results of models as something that you can guarantee. As most researchers will tell you: all models are wrong; some are useful.

I think that’s really important to keep in mind. A lot of things in this paper, and in the broader discussion on the IRA, aren’t being captured by models. These models represent a version of a world under textbook market conditions that’s heavily simplified so we can analyze these policies in a quick but also interesting and in-depth analytical manner.

Because of that simplification, certain things don’t fit into the framework we’ve used, and we haven’t been able to implement certain things directly. Interconnection delays for renewables is one example. Some of the electricity markets have issues with getting their renewables online, even after the project has been planned and the capital for that project is ready to go. It’s difficult to actually get renewable generation set up and ready to connect to the grid. You see the same thing with transmission. All this money is flowing to build renewables, and the projects are ready to be done, but actually implementing them is difficult from the perspective of interconnection and transmission. All sorts of institutions need to improve their efficiency to be able to handle this level of build-out. That’s one thing that we just simply don’t assess in our models, because it’s an institutional question, rather than an inherently economic, analytical question.

Other things could get in the way of the IRA reaching the emissions targets. For instance, supply-chain backlogs or critical mineral shortages. During the COVID-19 pandemic, a lot of people saw prices going up because of supply-chain issues. That’s probably a big part of the reason this bill was named the “Inflation Reduction Act” and not the “Emissions Reduction Act.” Some resource constraints can get in the way of the process. Interest rates that have been raised to address inflation have also slowed the flow of capital, which can also slow clean energy build-out.

Another consideration is some aspects of human behavior that economic models have never really quite tried to implement. Labor shortages already are being talked about as big friction points in the implementation of the IRA. Siting and permitting is another big issue that’s being discussed federally and the state level, which sometimes comes from local opposition. Sometimes people just don’t want wind turbines in their backyard. These models don’t consider every backyard and every potential wind turbine. The models just say where the costs make it possible and most cost-effective to build these renewables.

Another example is related to facilitating hydrogen fuel, which depends on the US Department of the Treasury releasing guidance on exactly how these tax credits are understood to be in the law and how the US Internal Revenue Service makes the credits available. We could end up generating a little bit of hydrogen, or the Treasury rules could allow for a lot of hydrogen production, which could lead to a lot more electricity demand, which in turn actually would undo a lot of the emissions reductions that the IRA could bring about, because you could increase demand more than the increase in renewable energy generation. How Treasury ends up deciding all these rules is a big uncertainty that these models are not trying to capture.

Maybe because I’m a pessimist, I usually think about the downsides related to the risks from these uncertainties: problems with local opposition, interconnection queues, labor supply, and getting materials. Do you think it’s true that most of these unmodeled aspects would tend to limit the benefits of the IRA, or could some sources of uncertainty go in the other direction and produce benefits?

NR: Some uncertainties definitely go in the other direction. I was tempted to bring an old output sheet from modeling done by the same team that we’re on now, back in 2008, because that modeling projected that emissions would be a lot higher today than they are and that electricity demand would be a whole 1,000 terawatt-hours greater in the United States.

I thought it was interesting that they also under-predicted how much solar and wind would get deployed, because the capital costs were so high back then for those technologies—but the costs have come down a lot in the past decade. The same thing could happen for a lot of advanced technologies that are in development right now.

Those are the kinds of things that modelers don’t like to make bets on. You don’t want to make a bet on an optimistic outcome when it comes to costs.

MD: The same modeling team that wrote this recent paper also looked at the sensitivities like how fast the models think it’s possible to build out renewables.

I’d also add that I think it’s useful to pay attention to the downside risks, because you do more in advance if you’re preparing for downside risks than if you’re just waiting for something great to happen.

Definitely. If we know what the downside risks are, then we can try to deal with them before they happen and prevent those bad things from happening.