Innovating Climate-Economy Modelling: Dealing with Uncertainty, Risk and Complexity

1 POLICY BRIEF Innovating Climate-Economy Modelling: Dealing with Uncertainty, Risk and Complexity Authors: Kostas Fragkiadakis, Stelios Tsiaras, Giannis Charalampidis (E3M) Jonas Haas, Sebastiano Bacca (Global Climate Forum) Ugne Keliauskaite (Bruegel) Reinhard Mechler (IIASA) Femke Nijsse, Ian Burton (University of Exeter) Jamie Pirie, Chris Thoung (Cambridge Econometrics) This policy brief was developed within the framework of the DECIPHER project, funded by the European Union’s Horizon Europe research and innovation programme (Grant Agreement No. 101056898). 2 1. Motivation Designing effective climate, biodiversity and energy policy requires making decisions today about uncertain futures. Policymakers face complex risks that unfold over decades, shaped by technological innovation, evolving financial conditions, and the macroeconomic impacts of extreme weather. To navigate these challenges, models of the climate– economy system are indispensable for exploring alternative futures and stress-testing policy choices. The value of these models depends on how well they capture uncertainty, incorporate behavioural dynamics, and reflect real-world constraints such as financing costs, systemic shocks, and physical damages. Without these elements, even sophisticated analyses can miss critical vulnerabilities or opportunities. The DECIPHER (Decision-making framework and processes for holistic evaluation of environmental and climate policies) project, funded under Horizon Europe, aimed to overcome limitations in conventional policy appraisal by creating a new, holistic decisionmaking framework that better captures the complex nexus of climate change, biodiversity, and the economy. The project included an iterative knowledge co-creation process for more transparent, inclusive and representative policy design and evaluation. Moreover, it developed new generation of state-of-the-art economic models featuring feedback loops with physical system models and embedded systemic risks and uncertainty and capture behavioural and knowledge dynamics allowing it to improve the representation of the economy-climate-biodiversity nexus. Compared to mainstream economic models that often neglect uncertainty and resilience, DECIPHER integrates advanced economic and biophysical modelling, empirical methods, and stakeholder co-creation to assess the feasibility, resilience, risk, and opportunity dimensions of climate and environmental policy options. The framework is designed to be operational under real-world conditions and is being applied to key EU policy domains such as the “Fit for 55” package, LULUCF regulation, and national recovery plans. Through this approach, DECIPHER equips policymakers with tools that do more than forecast: they reveal trade-offs, highlight robust strategies, and strengthen the legitimacy and adaptability of decisions in an uncertain future. DECIPHER research draws upon the comprehensive framework of IRGC (2006) 1 to identify discourses that science, policy, and application must address in the ‘uncertainty’ space. These discourses are based on different classes of risk and uncertainty, broadly aligned with the Knightian risk and uncertainty definitions of simple and complex, high-uncertainty, and high-ambiguity risk (see Figure 1). • Instrumental discourse is suitable for addressing clearly defined risk problems, employing well-tested decision support methods like cost-benefit analysis and focusing on economic incentives and technical solutions. 1 IRGC-International Risk Governance Council (2006). Risk governance: Towards an integrative approach. White paper no. 1. IRGC, Geneva. 3 • Strong participatory discourse is crucial, particularly (but not exclusively) for addressing issues of ambiguity where values are contested, aiming at conflict resolution and broad-based stakeholder engagement. • Reflective discourse becomes essential when dealing with high levels of uncertainty, emphasizing precautionary principles and the need for careful consideration of 'danger' and 'adaptation limits.’ • Epistemological discourse is particularly important for characterizing the available evidence for understanding risk across the entire risk spectrum, especially in the face of complexity and ignorance. Figure 1: The risk and uncertainty space. Source: Mechler et al., 2025. 2 Based on Knight, 1921 This policy brief presents five recent modelling advances that strengthen the evidence base for climate and biodiversity action and economic decision-making amidst uncertainty based on DECIPHER research: 1. Emulation for Uncertainty Quantification – enabling thousands of simulations to identify which policies are robust across many futures. 2. Rational Expectations in Computable General Equilibrium (CGE) Models – showing how forward-looking investment behavior can lower transition costs and smooth shocks. 3. Cost of Financing in Technology Diffusion – reflecting how interest rates and capital costs affect the uptake of clean technologies. 4. Flood and Coastal Damage Assessment – linking detailed coastal risk assessment with macroeconomic models to reveal the economic value of adaptation. 5. Multiple Resilience Dividend – capturing avoided losses, development co-benefits and inequality reductions from risk-management investments, strengthening the case for sustained adaptation and mitigation. 2 Mechler, R., Żebrowski, P., Clercq-Roques, R., Patil, P., Stefan Hochrainer-Stigler (2025).The role of extreme event and systemic risk - assessment and guidance. DECIPHER project, Deliverable 5.3 Instrumental Risk Ambiguity Uncertainty Complexity Epistemological Reflective Instrumental Participatory 4 These innovations share a common goal: to assist decision-makers in moving beyond static forecasts toward more resilient policy development. They identify targeted measures, such as investing in adaptation, improving access to affordable finance for renewable energy, or considering investor expectations, which might help contain costs and deliver long-term economic and social benefits. Incorporating these approaches into climate and energy planning and impact assessment tools can help policymakers make credible choices across various future scenarios, increasing confidence that policies will safeguard citizens and economies while supporting a well-managed transition. The following sections present each modelling innovation by setting out the problem it addresses and why it matters for policy, the solution with key modelling features, illustrative figures showing main results, guidance on application, including when the method is most suitable and its limits, and references to academic work by project partners for readers seeking further detail. 2. Incorporating uncertainty in impact assessment tools 2.1. Emulation for Uncertainty Quantification Problem Traditional climate and macroeconomic scenario models often fail to capture uncertainty in policy inputs and assumptions systematically, making it challenging for policymakers to assess policy robustness, defined as the ability to achieve policy goals under uncertainty and shocks, across various possible futures. Robustness differs from resilience, which emphasises ‘returning to a stable equilibrium point after a shock’. Modelling solution A machine-learning-based emulator was developed to serve as a computationally inexpensive surrogate for complex simulation models, such as FTT:Power. The emulator allows for: • thousands of scenario evaluations at minimal cost, • simultaneous variation in uncertain inputs (for example, techno-economic parameters such as learning rates 3 , as well as policy ambition). The emulator systematically explores uncertainty in 15 techno-economic parameters, such as build and connection speed, learning rates, energy demand growth, cost of capital, and technology lifetimes. In addition, the emulator explores the effect of different policy instruments within the country that implements them and cross-boundary. Illustrative example Key uncertainties in the speed of transition are identified in Figure 2. Over the last decade, the pace of building out solar and wind energy has accelerated, and grid expansion has struggled to keep pace with the growth of renewables in some countries. These two 3 A measure of reduction in costs of energy technologies for each doubling of cumulative production or capacity. 5 uncertainties lead to the highest variation in FTT:Power. Learning rates for onshore wind come next; its mean learning rate is much lower, so that slight variations in the learning rate can have a large effect on cost-competitiveness. Finally, Chinese policy plays a key role in the global cost of certain technologies, as their large market has the strongest ability to induce innovation. US subsidies, and their potential rollback, have a limited direct effect via induced innovation. Solar PV is no longer sensitive to these dynamics of regional policy. Figure 2: One-at-a-time analysis (oaat), for the top 19 variables in the analysis, (a) in terms of global emissions, capacity of onshore wind and capacity of offshore wind. Panel (b) shows the average sensitivities across all three outputs. Note, the figure was produced using code adapted from McNeall et al. (2024). Figure 3 shows the robustness of policy combination against key uncertainties. None of the policies are very robust against grid delays and worsening of relative build times for solar and wind (bottom row). In terms of other uncertainties, such as the cost of finance and high demand growth, the combination of subsidies and phase-outs is the most robust. Figure 3. Share of emulator runs meeting India’s 2030 targets. The bottom row shows the policy combinations tested (current policy, and combinations of upfront subsidies (Sub), carbon pricing (CP), and phase-outs (Phase). 6 Application Use when: a high number of model runs is required to assess the sensitivity or robustness of policy outcomes. Avoid when: the target model has rapidly shifting dynamics that are too complex to be accurately captured by emulators without a large amount of training data. 2.2. Rational Expectations in Computable General Equilibrium (CGE) Models Problem Expectations shape economic decisions as they reflect how agents believe key economic fundamentals will evolve. In economic modelling, the way expectations are formed strongly affects model results and policy implications, determining not only the eventual equilibrium but also the path the economy takes to reach the equilibrium. Typically, in economic models, expectations refer to the trajectory of prices or costs, and there are two mainstream approaches: • Myopic expectations, where agents lack foresight and base their decisions solely on current conditions, • Rational expectations, where agents have perfect foresight, meaning they possess complete information on how, for example, prices will change in the future, and their behaviour is influenced not only by current conditions but also by anticipated future conditions. Myopic expectations, when coupled with restrictions on capital mobility, may lead to an overestimation of costs related to the clean energy transition. The literature has highlighted how myopic expectations can lead to stronger responses to shocks and how rational expectations can lead to milder responses. However, since these are two extreme cases of expectation formation, researchers have also explored alternative methods, such as incorporating savings into the intertemporal problem. 7 Modelling solution A short version of CGE model, the GEM-E3 model, was developed that introduces two enhancements: • Rational expectations, allowing investors to form forward-looking views about future returns, • Capital mobility constraints, which can be set as partial or full, to capture limits on how easily capital flows between sectors or regions. These features enable a more realistic treatment of policy impacts and capital reallocation dynamics. Additionally, rational expectations were introduced to the full version of the GEME3 model, and the economic implications of the green energy transition were examined for Germany and Italy. Illustrative example To illustrate the influence of expectation formation and capital mobility, a permanent demand shock for photovoltaic (PV) equipment is simulated in the short version of the GEME3 model within the three-region model. Regions R1 and R2 act as net importers of PV equipment, while Region R3 serves as the exporter. The shock originates in R1. Scenarios compare full capital mobility (ALL) with partial mobility (PART). Figure 4: Unit cost of capital under myopic vs rational expectations. Myopic expectations lead to no anticipatory adjustment, while rational expectations trigger earlier investment, moderating cost spikes. The assumptions on capital mobility greatly influence the magnitude of the impacts, as capital costs change under partial mobility, peaking at a range of 10% to 18% compared to 0.7% under the assumption of full capital mobility. Figure 5: Investment patterns under myopic vs rational expectations. 8 Under rational expectations, investment begins earlier (2025) in anticipation of 2030 demand. Myopic agents react only at the time of the shock, resulting in higher capital costs and inefficient allocation. These implications are clearer under the assumption of limited capital mobility, as the shock in capital prices is significantly higher, hence the adjustment of investments under perfect foresight begins much earlier. Then, rational expectations were incorporated into the fully-fledged GEM-E3 model version. To achieve this, a first-order approximation was performed due to the model's large scale and complexity. Figure 6: Change in the unit cost of capital (A) and investments (B), (C) in % from the reference. -0.40% -0.30% -0.20% -0.10% 0.00% 0.10% 0.20% 2025 2026 2027 2028 2029 2030 2035 2040 2045 2050 (A) Germany Italy -2.0% -1.0% 0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 8.0% 2025 2026 2027 2028 2029 2030 2035 2040 2045 2050 (B) Germany (rational) Germany (myopic) -4.0% -3.0% -2.0% -1.0% 0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 2025 2026 2027 2028 2029 2030 2035 2040 2045 2050 (C) Italy (rational) Italy (myopic) 9 Changing the way expectations are formed may also affect the sectoral structure of investment, shifting it away from traditional manufacturing sectors and towards the production of clean energy and the manufacturing of clean energy equipment. Furthermore, this shift may put pressure on the current account balance due to higher investments in the short to medium term, which leads to an increase in imports of investment goods. Overall, the rational expectations assumptions lessen the impacts of the transition on the economy and lead to a higher GDP compared to myopic expectations. This, along with other structural changes and shifts in macro drivers during the transition, highlights the need to carefully consider the formation of expectations when assessing the impacts of the clean energy transition. Application Use when trying to assess the temporal effects of shocks or policy announcements with long lead times. Avoid when focusing on very short-term impacts or for situations where reliable data on expectations are lacking. 2.3. Cost of Financing in Technology Diffusion Problem The transition requires high levels of investment in low-carbon technologies. These technologies are more capital-intensive than fossil-fuel technologies, making the question of financing more important, both in itself and when comparing the relative attractiveness of the technology options. An understanding of the investment environment and the relative merits of different technology options is vital to inform effective policy. Conversely, failing to account for such specificities risks undermining the reliability of projected technology transitions, especially in scenarios in which wider macroeconomic or policy conditions might change. Under conventional modelling treatments, a fixed discount rate (for example, 10%) is more common, across all technologies (failing to identify technology-specific features) and regions (failing to ignore more local financing conditions). This simplification thus ignores differentials that might be consequential in determining the pace and global distribution of the transition, or the role of differential changes in such conditions e.g. policy-induced uncertainty and rising interest rates. Modelling solution The FTT: Power model was enhanced to incorporate weighted average cost of capital (WACC), which takes into account the cost of debt and cost of equity that vary by: • Technology type (for example, solar PV, onshore/offshore wind, gas, coal), • Region. The implementation of the WACC rate is done at a granular level to allow for the modelling of how financing costs could vary under different macroeconomic and policy conditions (for example, interest rate hikes, green financing options). This enriches the FTT-Power’s 16 Application Use when: evaluating development—oriented disaster risk reduction and adaptation policies across scales. Avoid when: doing technically—minded disaster risk reduction and adaptation policies across scales. Reference Mechler, R. , Żebrowski, P. , Clercq-Roques, R., Patil, P. & Hochrainer-Stigler, S. (2025). Positive Externalities in the Polycrisis: Effectively Addressing Disaster and Climate Risks for Generating Multiple Resilience Dividends. International Journal of Disaster Risk Science 10.1007/s13753-025-00661-2. 3. Conclusion The DECIPHER project shows that advances in climate–economy modelling can provide more policy-relevant insights when they explicitly address uncertainty, behavioural dynamics, and real-world constraints. Across the methods presented, ranging from machine-learning emulators to high-resolution damage assessments and financial risk models, the common theme is a stronger foundation for assessing policy impact against a wide range of potential futures. These innovations do not replace traditional modelling; instead, they improve it by identifying where policy outcomes are most sensitive to external shocks, where targeted interventions can reduce costs, and how adaptation and mitigation measures can produce co-benefits beyond emissions reduction. Incorporating these approaches into national and 17 regional planning can assist governments in devising strategies that stay credible and effective even as economic and climatic conditions evolve. Key Policy Takeaways • Integrate uncertainty analysis into policy formulation. Employ emulation and other methods to assess whether policies meet their objectives across a range of probable scenarios, not just a central forecast. This is particularly relevant for the 2025–2026 NECP revisions. • Consider expectations and financing conditions. Forward-looking behaviour and the cost of capital greatly influence investment patterns and the speed of technology diffusion. Policies should take these dynamics into account to prevent underestimating transition costs. Policies should account for these dynamics to avoid underestimating transition costs that is a key consideration for the Green Deal Industrial Plan and monitoring investment flows under NextGenerationEU. • Integrate flooding and coastal damage risk into macroeconomic planning. Highresolution flood and coastal damage assessments, as well as financial risk models that link climate impacts to asset values, provide vital evidence for adaptation and financial stability strategies. These insights can inform the EU Strategy on Adaptation to Climate Change and the design of resilience components in Cohesion Policy funds. • Recognise potential broader benefits of climate action. Policies that lower risk can also promote economic growth, social fairness, and development advantages. Valuing these “multiple resilience dividends” reinforces the economic argument for early and sustained investment, in line with the European Green Deal’s Just Transition Mechanism, which combines emissions reduction with social fairness. • Encourage collaboration between modellers and policymakers. Ongoing dialogue helps guarantee that modelling innovations meet policy needs and that results are interpreted and used appropriately. By applying these lessons, policymakers can design climate and energy strategies that are more resilient, economically sound, and capable of protecting citizens and economies in an uncertain future.