Macroeconomic Implications for Decarbonization Policies and Actions: Proceedings of a Workshop (2025)
Chapter: Incorporating Modeling Insights into Policy Design
Incorporating Modeling Insights into Policy Design
Hafstead and Känzig moderated the workshop’s third panel discussion, focused on incorporating modeling insights into policy design. The panelists, who included Wei Peng, Princeton University; John Bistline, Electric Power Research Institute (EPRI); David Hémous, University of Zurich; and Stephane Hallegatte, World Bank Group, considered current and emerging methodologies for incorporating insights from various modeling disciplines, such as energy systems, financial systems, and global perspectives, to inform actionable, granular decarbonization policy design. Participants also discussed opportunities to improve existing models, develop complementary approaches, or create new approaches to better inform policy design.
USING INTEGRATED ASSESSMENT MODELS TO IMPROVE POLICY REALISM
Wei Peng, Princeton University, discussed bringing political economy insights into integrated assessment models (IAMs). She noted that a persistent challenge in decarbonization modeling is the fact that models cannot always accurately account for real-world complexities, because policy results depend on adoption, implementation, and behavioral response (Stern et al. 2023). To close this gap and provide information relevant to the feasibility of climate mitigation actions, IAMs can be a helpful approach. With IAMs, researchers can input extensive demographic and policy assumptions into a core model simulating the interactions among the economy and energy, land, and climate systems to generate outputs across multiple dimensions including economic outcomes, emissions, energy pathways, and land use (Figure 5). IAMs can improve policy realism if modelers prioritize advancements that are both politically important and computationally tractable and encode the process to integrate social science research, Peng said.
SOURCE: Presented by Peng on September 13, 2024 (adapted from Carbon Brief 2018).
Policy realism can also be improved by an exploration of model comparisons and coupling opportunities, which Peng said will require deeper collaboration between modelers and political economists. Toward this end, she described how a multidisciplinary team of modelers compared IAMs, engineering-economic optimization modeling, and agent-based modeling to explore how each modeling approach accounts for the role of institutions and different actors, such as households, firms, and national-level decision makers (Davidson et al. 2024). The insights resulting from this work can help to guide the selection of modeling approaches for a given situation based on the strengths of each kind of model in simulating the impact of institutional heterogeneity on sustainability outcomes, she noted.
CAPABILITIES AND TRADE-OFFS OF ADVANCED MODELING APPROACHES
John Bistline, EPRI, highlighted insights and perspectives from EPRI’s U.S. Regional Economy, Greenhouse Gas, and Energy (REGEN) Model. Given the complex changes and interconnections across the energy landscape (Figure 6), he described how advanced models like REGEN can go beyond analysis to answer questions that policymakers and decision makers have. For example, specialized tools can link decision support and policy design by modeling macroeconomic feedbacks; coupling sectoral and supply-side models; and performing affordability, distributional, and reliability analyses.
With their highly detailed analyses, advanced models can provide greater accuracy and more actionable insights, making them a potent tool for capturing
SOURCE: Blanford et al. 2022.
underlying dynamics and informing decision making (Bistline et al. 2024). In selecting models and interpreting their results, Bistline stressed that it is important to recognize that different models have quantifiable trade-offs depending on what is modeled, how much detail is represented, and how the model is applied. All models can provide insights on potential changes and highlight systems interactions, but the appropriate model to use depends on the question being asked (Figure 7), he explained.
Moving forward, Bistline suggested that more research is needed to advance linking and comparison tools; navigate trade-offs between model detail and tractability in modular models to better facilitate uncertainty analyses; align empirical and structural modeling and linking across different model scopes; address novel policy proposals that stretch current analytical capabilities; and perform multi-model comparisons to interpret variance across models, address stakeholder questions, and aid policy development.
INTEGRATING ENDOGENOUS INNOVATION INTO CLIMATE MACROECONOMIC MODELS
David Hémous, University of Zurich, described approaches to more realistically integrate innovation into climate macroeconomic models. While most climate models consider technological progress as exogenous—changes that happen based on externalities, such as a carbon tax resulting in lower emissions—he explained that his work focuses on technological progress as endogenous, reflecting the ways in which innovation responds to economic incentives and policies. Models that integrate endogenous innovation can bring new insights to climate
SOURCE: Presented by Bistline on September 13, 2024.
policy—for example, by projecting which efforts may generate the largest profits and then modeling how that might influence further technology development, he noted.
Highlighting some of the insights that have emerged from such models, Hémous said that studies have shown that innovation is path-dependent when two inputs are substitutes for each other, revealed the high costs of delaying intervention, and suggested that optimal policies would couple carbon pricing with subsidies for clean energy research (as opposed to providing subsidies for research into energy-saving behavior) (Acemoglu et al. 2012; Hassler, Krusell, and Olovsson 2021; Hémous and Olsen 2021). For countries that are willing to work alone, he said that this work supports creating a national green industry policy over unilateral carbon pricing (Hémous 2016). The research findings also suggest that focusing on intermediate energy sources, such as natural gas, can delay the energy transition if they are not accompanied by appropriate economic policies (Acemoglu et al. 2024). Finally, the models suggest that the energy transition may stall without adequate support for storage technologies.
Moving forward, Hémous suggested that endogenous innovation models could be integrated into quantitative frameworks, used to answer political economy questions, and used to study how to “greenify” whole supply chains, ultimately strengthening the economic rationale for government subsidies for green technologies and carbon pricing.
LESSONS FROM COUNTRY-BASED MODELING OF LOW-EMISSION PATHS
Stephane Hallegatte, World Bank Group, described the World Bank Group’s approach to modeling climate and development trajectories in countries around
the world, which is captured in its country climate and development reports. Recognizing the profound impacts climate change has across all countries, the World Bank has worked to integrate climate action into its development and finance work to model and support resilient low-emission paths. Interpreted by experts and tailored to individual countries, Hallegatte stressed that these paths are bottom-up and country-specific, in contrast to global models that cannot fully capture the complex development and socioeconomic contexts of each country, including domestic policy priorities and political constraints.
To model these pathways, sector experts integrate projected investments, costs, feasibility, and benefits to create sectoral roadmaps, Hallegatte explained. He described how these roadmaps are then combined into a larger macroeconomic model that reflects what a country can afford and interplay between sectors to outline different paths to achieve both development goals and climate objectives. The result is not one “optimal” pathway but a series of building blocks reflecting feasible emission scenarios and the investment requirements and trade-offs, Gross Domestic Product growth potential, labor market implications, and household consumption changes that would be involved in each, he added. To aid in the interpretation and use of these models, Hallegatte noted that a multidisciplinary team of experts adds nuance and qualitative analysis to models’ typically quantitative output. From a political messaging perspective, he also added that it is important to frame the different pathways in light of the problems countries are facing with the changing climate, rather than comparing options against an imaginary baseline scenario in which climate change does not exist.
DISCUSSION
Following the panelists’ opening remarks, participants discussed challenges related to modeling approaches and data, as well as approaches for supporting policy feedbacks and understanding uncertainties.
Modeling Challenges
Känzig asked panelists about strategies to determine how much detail is needed as modelers consider how to balance increasingly extensive models with the level of detail and transparency necessary for decision making. Bistline replied that a modular modeling approach, informed by empirical best practices and econometric studies and incorporating multiple variations, enables modelers to run numerous scenarios (e.g., for technology, markets, or consumer preferences) that create detailed, transparent projections to inform decision making. He cautioned, however, that too much detail or too many parameters can compound uncertainty.
Peng agreed and emphasized that making models more complicated should be a means, not an end goal in itself. She said it is important to carefully consider what elements are included and excluded, such as physical systems or public opinion, and how the output will represent those factors. She highlighted the value of a sequential approach, where simpler models with strong empirical foundations are coupled with other models to improve representations. Hémous agreed and added that it is important for modeling design to be informed by the ultimate goal of the modeling and how it will be used.
Replying to a question from Carley, Hallegatte said that it is important for modelers and policymakers to interact via a two-way flow of information. For example, the World Bank learned that external reports would be ineffective and unpopular unless they were created in close collaboration with each country’s government to understand their climate and development goals, embed them into the larger developmental narrative, and enable a feeling of ownership, all of which reduce dependency on one person, party, or election cycle. Getting the framing right is essential, he said, noting that focusing on energy access or energy security is more palatable for countries than focusing on climate policy, and he emphasized the importance of allowing feedback while avoiding politically motivated negotiation. In his view, the most successful reports are those that reveal new questions and opportunities and provide an on-the-ground, bottom-up approach that complements—and could be integrated with—global models. Hafstead noted that Resources for the Future follows a similar approach, working early on to understand a group’s needs before facilitating a two-way conversation to design policy solutions.
Boushey asked how models can incorporate and engage behavioral and socioeconomic research, and Hémous suggested balancing modeling with case studies. He added that case studies could even be used to inform models, to better illustrate the role of politics and behavior, capture the unquantifiable, and identify root causes when policies are not working.
Data Challenges
Hafstead asked participants to discuss the biggest data challenges they encounter in the area of decarbonization modeling. Hallegatte replied that precise data collection is a problem in many countries, including the United States, but especially in countries and sectors with more informal economies. For example, he suggested that many firms will find it challenging to meet the strict data reporting requirements, such as emissions measurements in the E.U. Carbon Border Adjustment Mechanism (CBAM), creating not just data gaps but institutional or cultural gaps as well. This risks creating metrics that exclude some exporters that are not necessarily energy intensive but are unable to report and measure their emissions in accordance with the requirements. He posited that this data gap can
be surmounted by substituting other data or simplifying implementation, even at the cost of efficiency, suggesting that modelers should “fix the policies for the data gaps and not the other way around.”
Bistline agreed that data gaps are a challenge, and he suggested fixing them through empirical work and improved data gathering. For example, data on local ordinances for wind and solar buildout were once considered impossible to quantify and collect, but researchers were eventually able to do so and incorporate them into models to aid regional decarbonization policies. He also suggested that data gaps in other countries could be addressed by replicating U.S. open-source communities that make energy data available. He added that countries would benefit from interdisciplinary collaborations that expand the data used in climate models, both on the supply and demand sides, as well as from new policies that expand data access.
Hémous suggested closing data gaps by studying methods to more widely incorporate substitutions or elasticities into models to better understand how they may react to policies. Peng added that the timing of modeling and data collection is important, as modeling during policy windows can provide important decision-making tools and identify what additional data or model structures would be needed in the future.
Modeling for Policy Feedbacks and Uncertainties
In reply to a question from Lenton, Hallegatte confirmed that policy feedbacks, where new policies make further changes possible, are included in models. Policies are influenced by past implementations, he continued, and modeling can project new policy success. For example, countries that have energy audits are more likely to also have energy efficiency standards, and vice versa, he explained. In fact, he noted, the process is often sequential; for example, a long-term goal that seems out of reach, such as a carbon tax, can eventually be met through smaller, less politically charged intermediate steps, such as implementing emissions monitoring. Hémous agreed, noting that a pathway toward a carbon tax could start with voluntary emissions reporting and low tax rates that, over time, reveal misreporting and ultimately reduce opposition. Peng noted that IAMs also use policy feedback and sequencing. She added that while IAMs are currently better at processing policy feedbacks related to technologies than human behavior or qualitative factors, they should be able to incorporate those complexities eventually.
Stock asked how models incorporate technology or policy uncertainties. Bistline replied that the challenges include quantifying uncertainty, especially for consumer behavior and policy windows; and communicating modeling results, especially if the models are highly complex. While some insights can be replicated from simpler models, he said that best practices are needed to incorporate stochastic elements like uncertainty into new models. Peng noted that to incorpo-
rate or determine uncertainties in modeling, she uses exploratory modeling with multiple large-scale future scenarios—for example, to view both the beneficial and detrimental health effects of carbon pricing. She said that the most effective models are those that combine multiple elements to arrive at an understanding of how society and physical systems may evolve.
As an example of uncertainty and in reply to a question from Wendy Edelberg (Brookings Institution) regarding the impact of government subsidies for climate innovation on a country’s GDP, Hémous noted that subsidies for new technologies are risky because their potential is uncertain, and they can initially lower GDP. However, if successful, they can help countries avoid climate damage and inspire knowledge spillover. He added that more research is needed to understand uncertainty when technological innovation is endogenous. Hallegatte agreed, noting that large-scale research is challenging given the multitude of uncertain, non-linear scenarios. More granularity of each scenario is needed to determine combined effects of different technologies and policies, which will require more resources, he noted. He added that another key challenge is communicating more effectively with policymakers to improve their understanding of modeling outputs and forecasts.
