Will The Inflation Reduction Act Beat Business as Usual?

Comparing IRA Models with a Historic Decarbonization Baseline

Will The Inflation Reduction Act Beat Business as Usual?

It has been less than two years since the passage of the Inflation Reduction Act (IRA). Early indications suggest that IRA’s tax incentives, together with other key macroeconomic advantages enjoyed by US manufacturers, have stimulated impressive flows of new capital investments in clean technology manufacturing and production. Trends in deployment of clean technology and infrastructure, however, have not yet shown similar upticks. Recent analysis by MIT, Rhodium Group, and Princeton’s REPEAT Project concludes that wind and solar deployment in the power sector, where much of the emissions benefits of IRA over the next decade were anticipated to come from, is already lagging projections produced by leading energy system models at the time that IRA was passed.

In this analysis, we revisit early modeling of IRA emissions impacts, comparing projected nationwide and sectoral greenhouse gas reductions across three leading models produced by the Rhodium Group, Energy Innovation LLC, and Princeton’s REPEAT Lab. In part, these models sought to assess how much of a difference IRA would make in helping the U.S. achieve the Biden White House’s goal of reducing nationwide greenhouse gas emissions to half of 2005 levels by 2030. At the time, these studies concluded IRA’s policies did not place the country on track for the Biden administration's emissions commitment, but would at least put national emissions on a trajectory putting the goal within better reach.

Ultimately, contextualizing forward-looking talk about projections and goals requires a comparison with the past and present trajectory of U.S. climate progress. Our analysis compares normalized reference and MidCase scenarios from those three models against a naive projection based upon the historic rate of improvement in emissions per unit of U.S. GDP from 1990-2022. We then use normalized sectoral greenhouse gas emissions projections from the three models to construct best and worst case estimates by combining the highest and lowest sectoral emissions estimates across the three models.

Our analysis finds that reference case emissions from all three models underperform the historic US decarbonization rate over the last three decades. Two of the models, from Energy Innovation and Rhodium, respectively, project reference case emissions that are substantially higher than what a naive projection of historic decarbonization would suggest. Normalizing emissions scenarios also suggests that there is somewhat less convergence across these leading models than initially appeared to be the case. At the sectoral level, the three leading models see more substantial variance, with significant differences in projected emissions from the power, transportation, and industrial sectors. Overall, the span of possible best-case to worst-case emissions outcomes through 2035, ranging from outcomes well above the historic decarbonization rate to outcomes well below MidCase estimates of post-IRA emissions, is far more uncertain than initial IRA modeling perhaps suggested.

What is certain is that the key limitations to the pace of U.S. decarbonization are now deployment constraints like permitting obstacles and transmission capacity—which not only existed before IRA but noticeably persist after its passage. Focus on regulating power plant and car emissions, increasing subsidies, watering down domestic content credit rules, or introducing new climate targets is tertiary, if not even counterproductive altogether. Clean technology progress nationwide is now a question, principally, of alleviating deployment constraints.

Shifting Baselines

The greenhouse gas emissions intensity of the American economy fell by 4.48% year-on-year on average over 1990-2022 (this is the period for which the U.S. Environmental Protection Agency constructed their 2023 update to the national greenhouse gas emissions inventory on an economy-wide basis). A simplistic crude extrapolation of this trend would suggest that U.S. emissions would fall to 4822 MMT by 2035. We normalized modeled emissions projections to the REPEAT analysis’s future projected GDP to compare all the models against this same extrapolated trends in CO2-equivalents per unit GDP.

The resulting normalized reference scenario emissions from all three models end up trailing behind this historic US decarbonization rate. These counterfactual reference case emissions either imply that US decarbonization over the next decade would have slowed down without IRA investments, or these reference cases lean conservative, and may overestimate the additional decarbonization impact of IRA investments.

IRA Figure 1 final
Figure 1: Comparison of GDP-normalized emissions projections in post-IRA modeling from REPEAT, Energy Innovation, and Rhodium Group’s Reference and MidCase scenarios against the historic 1990-2022 trend of improving U.S. CO2e/GDP (-4.48% year-on-year) extrapolated to 2035, as well as the Biden administration’s 2030 goal for U.S. emissions cuts. GDP assumptions are normalized to REPEAT’s future GDP projections. Emissions shown do not include land use, land use change, and forestry sources and sinks.

We note that the effect of this GDP normalization is minimal, as all three models assume a relatively similar future U.S. GDP growth to 2030 and 2035 with modest differences. As shown in Figure 2, the envelope of future economy-wide greenhouse gas emissions projections looks relatively similar for both GDP-normalized and non-normalized results, although the range of the normalized 2035 values is slightly wider by around 151 million metric tons CO2e.

IRA normalized vs reported vf
Figure 2: Side-by-side presentation of original non-normalized U.S. economy-wide emissions projections (left) versus GDP-normalized economy-wide emissions projections (right) for the MidCase and Reference pathways in each of the three modeling studies examined.


  • Emissions intensity: metric tons of CO2 equivalents, as presented by EPA national inventories (2023 ver.), emitted across the entire U.S. economy and divided by U.S. GDP.
  • Historic trend of emissions intensity: From 1990 to 2022, U.S. economy-wide emissions intensity fell at an average rate of -4.48% each year. Assuming that this exact rate of emissions intensity improvement persists through 2035 yields a projected emissions intensity for each future year (falling from 295 grams CO2e per 2020$USD in 2022 to 162 grams CO2e per 2020$USD in 2035). Combining these annual emissions intensity values per unit GDP with projected U.S. GDP—in this case REPEAT’s assumed future GDP values—produces a time series of future economy-wide emissions based on the historic rate of improvement in U.S. emissions intensity.
  • Sectoral CO2 emissions: This includes only energy-based CO2 emissions across the power, transport, and industry sectors, as reported in the Bistline, et al paper. Our team combined the separated non-energy CO2 emissions from these sectors and Building sector CO2 emissions into one category - Other emissions.

Projected Emissions Reductions Vary Across and Within Sectors

The broad range of possible outcomes covered by these post-IRA, top-line, economy-wide emissions projections, contains even greater sector-specific differences in projections for the power, transportation, and industrial sectors. Using the best-case and worst-case normalized emissions projections for each individual sector from the three modeling studies to visualize the full envelope of possible economy-wide emissions implied by sector-level uncertainties, the range of potential emissions outcomes through 2035 is wide and uncertain.

IRA figure 3 final
Figure 3: Range of “best case” and “worst case” MidCase and Reference scenarios constructed using the sum of the lowest and highest sector-specific emissions projections, respectively, from each of the three modeling efforts.

Relative to reference case emissions, the MidCase IRA scenarios deliver additional reductions in annual power sector emissions of between 317-862 MMT CO2 in 2035, while the modelers expect only 83-236 MMT CO2 of reductions from the industrial sector and 39-207 MMT CO2 of reductions from transportation.

Figure 4: Range of power sector, transportation sector, and industrial sector CO2 emissions reduction outcomes in 2035 across all MidCase scenarios, relative to emissions for those same sectors in the Reference scenarios.

Below, we can see the three models’ assumed additional emissions reductions post-IRA relative to reference scenario assumptions in the power, transport, and other sectors. The largest absolute difference in sectoral emissions between IRA and non-IRA comes from the power sector, whereas additional emissions reductions from IRA in the transportation, industrial, and other categories are far more modest.

Figure 5. Range of power sector CO2 emissions in 2035 as modeled in the Energy Innovation, Rhodium Group, and REPEAT studies, for both the Reference and MidCase scenarios.
Figure 6. Range of transportation sector CO2 emissions in 2035 as modeled in the Energy Innovation, Rhodium Group, and REPEAT studies, for both the Reference and MidCase scenarios.
Figure 7. Range of industrial sector, non-CO2, and other greenhouse gas emissions in 2035 as modeled in the Energy Innovation, Rhodium Group, and REPEAT studies, for both the Reference and MidCase scenarios.

Mixed Real-World Outcomes

With some time having elapsed since IRA’s passage, analysts are now able to compare actual real-world deployment of clean energy and transportation technologies against projected deployment rates. There are positive trends in the real-world data. Researchers at the REPEAT Lab, Rhodium, and MIT have emphasized that in 2023, electric vehicle “actual sales came in at the top end of the range of post-IRA projections.”

Figure 8: Zero-emissions vehicle share of total U.S. light-duty vehicle sales, with real data shown in blue and compared against the range of 2023 projections from the three models of interest in red. Figure adapted from the Clean Investment in 2023 report.

But notably, the longer-term range of the modelers’ EV projections becomes much wider after the mid-2020s. According to Energy Innovation LLC’s and Rhodium’s models, the EV share of new nationwide road vehicle sales in 2035 would be between 30-40%. But REPEAT’s model expects that figure to exceed 80%. In order to keep up with REPEAT’s forecast, EV sales would have to more than double from 2023 levels by 2025 and keep rising from there.

Figure 9: Electric vehicle sales share of total U.S. new light-duty cars and trucks, with the red dot indicating progress as of 2023 at 9%. Figure adapted from the 2023 John Bistline et al. research paper.

As noted above, the transportation sector is responsible for a relatively small portion of the IRA’s expected emissions reductions. In the power sector, where the bulk of projected future reductions from IRA occur, real-world deployment trends of clean electricity generation and grid batteries is somewhat more disappointing. 2023 clean generation capacity additions (32.3 GW) were just around half of the range of 46 to 79 GW deployed on average in 2023-2024 in the three models, and falls well short of the 70 to 125.9 GW/yr of average modeled deployment from 2025-2030. This gap in achievement is all the more important when one considers that the important factor for decarbonization is not really deployment, but rather retiring existing fossil-fueled power plants and vehicles and meeting new demand with cleaner alternatives.

Figure 10: Annual U.S. clean electricity capacity additions, with real data shown in blue and compared against the projected ranges of annual average capacity additions from 2023-2024 and 2025-2030 from the three models examined (red). Figure adapted from the Clean Investment in 2023 report.

This sluggish clean power deployment can be mainly attributed to declines in wind capacity additions over the last several years, which fell from a recent peak of 15 GW added in 2020 to just 6 GW added in 2023. The modelers projected that wind would make up much of the new low-carbon generation over the next decade, so this real-world trend represents one of the starkest contrasts with their forecasts. And while 2023 saw an uptick, US solar deployments have also yet to massively expand. Higher relative US module prices (heavily affected by trade policy related to anti-dumping and forced labor import restrictions despite a general trend of falling module prices), interest rates, and higher non-financing project/installation costs limited solar’s expansion.

Figure 11: Clean energy technology capacity additions per year, with real data shown as colored lines and square data points, compared against the range of average annual capacity additions from 2024-2035 across various post-IRA modeling analyses as shown by the shaded region and circled points. Figure credit: John Bistline.

Subsidies Aren’t Everything

These are approximately the trends one would expect from the Inflation Reduction Act’s expansion of deployment subsidies to a mix of more and less mature technologies. Solar photovoltaics and onshore wind are quite clearly large commercial-scale industries at this point, with their combined generation approaching that of nuclear energy and more than double that of hydropower in the United States. It has been obvious for some time that what constrains further wind and solar expansion is increasingly a question of enabling infrastructure and permitting policy. Renewable energy developers are regularly faced with local and state siting restrictions, lengthy environmental review processes, and sporadic interest group opposition, as well as years of wait time spent studying the implications of connecting each project to the electricity grid.

That contrasts with the case of electric vehicles, a less mature technology that benefits proportionally more from public deployment support. Nevertheless, fluctuating growth rates of sales suggest that EVs are encountering similar infrastructure and policy obstacles as solar panels and wind turbines, in which federal subsidies matter less than the availability of charging infrastructure and other bottlenecks, not to mention fickle consumer behavior. Even if these bottlenecks are cleared, transportation and industrial emissions will likely prove much more stubborn to mitigate due to the immaturity of low-carbon alternatives outside electric power and light-duty vehicles.

The recurring infrastructural and regulatory barriers to rapidly building out clean energy infrastructure must be addressed to successfully expand renewable deployment, while also helping pave the way for deployment of newer technologies. Such less-mature low-carbon technology categories—new nuclear reactor designs, enhanced geothermal, non-conventional solar, offshore wind, green hydrogen, carbon removal, and electric vehicles—are in turn the best targets for deployment subsidies. These technologies have yet to realize the substantial consumer demand or cost curve effects that solar and wind have. To wit, the IRA models disagree substantially on the size and composition of the projected future markets for new technologies like green hydrogen, underscoring the immaturity of such climate solutions and the value of well-designed early-stage deployment support.

Figure 11: Hydrogen production by method/technology in 2030, showing widespread variance across the modeling groups’ expectations. Figure adapted from John Bistline et al. 2023.

Federal deployment tax credits for solar and wind are arguably the big clean energy success story of the last decade or two. Solar panels and wind turbines became radically cheaper over this period and scaled not just in the United States but around the world. But tax credits were neither the sole policy instrument acting in solar and wind’s favor, nor are they a one-size-fits-all tool for driving technological innovation and deployment. Solar and wind in the mid-2000s also benefited from heavy German, Japanese, and especially Chinese industrial policy—including worrying Chinese environmental and labor practices in the case of solar photovoltaics. More recently, intermittent energy from solar and wind has also benefited to a significant degree from the legacy of the shale revolution itself, with abundant and affordable natural gas helping balance the electric grid (alongside batteries) when the sun isn’t shining or the wind isn’t blowing.

As the Rhodium/REPEAT/MIT analysis concludes, “The biggest barriers to deployment between now and 2030 are non-cost in nature—like siting and permitting delays, backlogged grid interconnect queues, and supply chain challenges.” This was likely true for wind and solar even before the passage of the IRA, but it is certainly the case in a persistently high interest rate environment as wind and solar strain existing transmission infrastructure and start to face value deflation headwinds in some regions.

Other low-carbon technologies such as electric vehicles, enhanced geothermal drilling, green hydrogen, advanced nuclear reactors, and more, still have substantial learning, economies of scale, infrastructural and behavioral barriers, and regulatory hurdles separating them from fully fledged commercial viability and mass adoption. Federal tax subsidies under the IRA will be essential for overcoming many of these challenges, and in scaling these nascent low-carbon industries. But as the wind and solar industries are now discovering, tax credits aren’t everything.


The magnitude of “additional” emissions reductions achieved by Inflation Reduction Act policies are ultimately dependent on the baseline, counterfactual future emissions trends that energy system modelers project. As our analysis shows, those reference case baselines vary among the three major IRA modeling efforts, but all of them are outpaced by a simple continuation of the historic US decarbonization rate. If the IRA was necessary to keep emissions on track with the historic rate, then early indications of the law’s impact are not encouraging.

The electric power sector, in which the largest share of emissions reductions is projected to occur, is experiencing slower-than-anticipated deployment of clean energy. Electric vehicle sales, while keeping up with modeled expectations through 2023, may be falling behind pace, instead of accelerating as projected.

These trends in the most mature of low-carbon technology industries in turn portend difficult conditions for less mature sectors like hydrogen, carbon capture, advanced nuclear, geothermal, and more.

It’s encouraging to see the modelers themselves acknowledging the gaps between their forecasts and real-world experience, and drawing more attention to the “non-cost” barriers to deployment that went largely unaddressed in initial attempts to model the IRA’s impacts on decarbonization efforts. But we must convert these acknowledgements and this attention into policy reforms if we hope to realize anything close to these models’ emissions reductions expectations.