A research and development investment strategy to achieve the Paris climate agreement

Achieving the Paris Agreement goal of limiting global warming to well below 2 °C requires rapid peaking and reduction of GHG emissions through phasing out fossil-based energy and scaling low-carbon and negative-emission technologies (renewables and CCS). Many necessary technologies are not yet fully market-ready, implying a critical role for public R&D to accelerate innovation and lower costs. Decisions on the technology portfolio, timing of R&D, and financing mechanisms are crucial, yet most integrated assessments neglect research-driven innovation and how to fund it. This study addresses these gaps by linking two integrated assessment models to determine optimal R&D investment pathways aligned with climate stabilization and to assess a financing mechanism via carbon revenue recycling. The analysis focuses on five key low-carbon technologies (batteries for vehicles, advanced biofuels, solar, wind, CCS) and energy efficiency, aiming to inform policy on technology prioritization, intertemporal optimization of R&D, and macroeconomic implications of financing strategies.

The paper builds on extensive literature indicating that public RD&D provides patient capital that helps overcome uncertainty and fosters non-incremental energy innovation. Empirical studies document learning-by-research and learning-by-doing effects that reduce technology costs, with particularly strong benefits for less mature technologies. Historically, coherent policy frameworks combining supply- and demand-side measures, supported by public R&D, enabled cost-competitive low-carbon technologies such as wind, solar, LEDs, and EV batteries. The literature indicates that public R&D investments contribute significantly to lowering energy technology costs and that portfolio choices among technologies at different maturity levels are essential, with regional resource endowments and system interactions influencing optimal allocations. Despite this, most IAM-based climate stabilization assessments overlook explicit research-driven innovation and its financing, underscoring the need for the present integrated approach.

The study soft-links two integrated assessment models to derive optimal R&D pathways and evaluate financing and macroeconomic effects: (1) WITCH (intertemporal optimization) identifies cost-effective, region-specific optimal R&D investment trajectories for five low-carbon technologies plus energy efficiency under different climate targets, accounting for technology substitution and complementarity over a full century and incorporating two-factor learning (learning-by-research via knowledge stocks and learning-by-doing via cumulative capacity), five-year lags between R&D and knowledge accrual, a 5% annual knowledge depreciation rate, and international knowledge spillovers. (2) GEM-E3 (computable general equilibrium) assesses the financing of additional R&D via carbon revenue recycling and macroeconomic outcomes (GDP, employment, trade, competitiveness) with detailed sectoral resolution (67 sectors; 10 low-carbon manufacturing). Learning rates for each technology are harmonized between models using literature values; floor costs and initial knowledge stocks are aligned. Sensitivity analyses on learning rates (Section 4 SI) test robustness; results are generally robust, with some sensitivity for advanced biofuels (notably in 1.5 °C) and wind. Scenarios: REF (SSP2, no additional climate policy; temperature ≈3.5 °C by 2100 via MAGICC 6.0), 2 °C and 1.5 °C carbon-budget scenarios with cumulative CO₂ budgets of 1460 and 710 GtCO₂ for 2011–2100, respectively. For each climate target, two variants are simulated: OPT (optimal R&D) and FIX (R&D fixed at REF levels). In WITCH, a global carbon tax is iteratively determined to meet the budget; in GEM-E3, carbon prices are endogenously set to match WITCH emissions pathways. Financing in GEM-E3: carbon revenues first finance the additional R&D needs beyond REF; remaining revenues reduce payroll taxes (double dividend). Energy efficiency R&D is optimized in WITCH but, due to linkage constraints, not embedded in GEM-E3 runs; post-processed checks confirm carbon revenues suffice to finance these as well. Battery R&D estimates include significant private expenditures; other technology R&D relates to public spending. Data and code availability are provided (zenodo DOI 10.5281/zenodo.7755725; WITCH and MAGICC sources).

- Portfolio and timing: Across scenarios, the largest shares of R&D go to batteries for EDVs and energy efficiency. In REF and 2 °C, wind and solar follow; in 1.5 °C, advanced biofuels becomes third, ahead of wind and solar. For mature technologies (solar, wind), stringent climate policy shifts R&D earlier (not necessarily larger cumulatively), e.g., solar R&D is advanced toward 2030, leveraging two-factor learning (earlier R&D lowers costs, accelerates deployment, and further reduces costs via learning-by-doing). - Less mature technologies: CCS exhibits the strongest increase under climate policy; R&D in 2030 is about 2× (2 °C) and 2.6× (1.5 °C) REF levels, rapidly reducing costs of fossil with CCS and partially competing with renewables for deployment shares. Batteries show modest R&D increases under policy, but high learning rates yield large cost reductions from small additional investments. - Global R&D requirements: Achieving 2 °C (1.5 °C) requires an 18% (64%) increase in global cumulative low-carbon R&D investment by mid-century relative to REF. In the 1.5 °C scenario, the global average R&D investment (2020–2050) is 18.8 billion 2005$. - Regional heterogeneity: The USA, EU, and Japan–South Korea are top contributors across scenarios. Reaching 1.5 °C requires deeper changes than 2 °C, with several regions (China, Japan–South Korea, Southeast Asia, Latin America, Sub-Saharan Africa) needing to at least double R&D relative to REF, notably for advanced biofuels. CCS R&D increases especially in Latin America (excluding Brazil), Canada, USA, China, and reforming economies. - Financing with carbon revenues: Carbon revenues in 2050 reach 2.5% (2 °C) and 4.1% (1.5 °C) of global GDP (higher in carbon-intensive producers, e.g., ~14% in Saudi Arabia and Russia). Additional R&D funding beyond REF requires only about 2% of global carbon revenues; regionally it spans ~0.5% (Argentina) to 21% (Sweden) of carbon revenues in 1.5 °C. Financing the entire low-carbon R&D (including energy efficiency) would require up to 18% (22% in 1.5 °C) of regional carbon revenues in 2050. In many oil-exporting and fossil-intensive regions, required R&D shares are below current fossil fuel subsidies. - Macroeconomic impacts: Optimal R&D lowers technology and mitigation costs, reducing required carbon prices and global policy costs by about 7–19% by mid-century, with more pronounced effects under 1.5 °C. GEM-E3 indicates that each 2005$ of R&D investment raises GDP by 1.64 (2 °C) and 8.01 (1.5 °C) 2005$ in 2050; per 2005M$ of R&D, employment increases by 7 (2 °C) and 96 (1.5 °C) persons in 2050. Regional competitiveness effects drive outcomes: e.g., in 1.5 °C, China’s GDP is +0.4% vs FIX in 2050 (battery and EV exports), EU28 +0.25% (advanced biofuels exports and higher consumption); Argentina +1.2%, Indonesia +0.7%, Brazil +0.13% (biofuels). Employment effects are broadly positive even where GDP may fall relative to FIX, reflecting labor-intensive R&D. - Feasibility and alignment: The optimal R&D trajectory aligns with higher early carbon revenues, supporting front-loaded R&D; as decarbonization progresses and revenues fall, R&D needs also decline.

The study demonstrates that explicitly optimizing and financing R&D within IAMs enhances the feasibility of Paris-consistent pathways. Early, well-timed R&D in mature technologies (solar and wind) delivers substantial cost reductions without necessarily increasing cumulative R&D, while less mature technologies (CCS) require higher R&D to meet stringent targets. Technology competition is evident: investing in CCS can reduce the need for additional renewable R&D and deployment. By recycling carbon revenues to fund R&D and reduce payroll taxes, policy makers can implement the polluter-pays principle while achieving double dividends—lower mitigation costs, improved competitiveness, and higher employment. The multi-model approach shows robust reductions in carbon prices and global policy costs, particularly under 1.5 °C, and identifies region-specific macroeconomic winners linked to comparative advantages in clean technology manufacturing and biofuels. Overall, optimizing R&D portfolios and timing is a crucial enabler for cost-effective and equitable decarbonization.

This paper introduces a consistent, multi-model framework linking WITCH and GEM-E3 to identify optimal low-carbon R&D strategies aligned with Paris targets and to assess a practical financing mechanism via carbon revenue recycling. Key contributions include: (i) quantifying technology-specific and region-specific R&D needs and their timing; (ii) showing that early R&D in mature technologies and scaled R&D in less mature ones reduce mitigation costs and carbon prices; (iii) demonstrating that additional R&D requirements for 2 °C and 1.5 °C can be financed with a small share of carbon revenues while yielding macroeconomic and employment gains; and (iv) highlighting regional competitiveness dynamics in batteries and advanced biofuels. Future research should explore broader financing strategies (including international transfers and mechanisms for countries with limited fiscal capacity), integrate additional emerging technologies (e.g., DAC, hydrogen) as modeling capabilities and data improve, and refine representations of learning and spillovers to further assess uncertainty and policy design.

- Technology scope: The analysis excludes some potentially relevant technologies (e.g., Direct Air Capture, hydrogen) due to current modeling and data limitations, potentially understating portfolio substitution effects. - Financing scope: GEM-E3 finances only the additional R&D beyond REF with carbon revenues; results do not assume full R&D funding via carbon revenues, which would be infeasible for some countries, especially under 2 °C. International financial transfers are not considered. - Model linkage: Energy efficiency R&D is optimized in WITCH but not embedded in GEM-E3 due to methodological linkage constraints; feasibility of financing efficiency R&D is assessed via post-processing. - Learning parameters: Learning-by-doing and learning-by-research rates are literature-based and may involve partial double counting; while sensitivity tests show robustness overall, advanced biofuels (in 1.5 °C) and wind are more sensitive. - Data and attribution: Battery R&D includes significant private investments, whereas other technology R&D pertains mainly to public spending; differences in public/private shares may affect comparability with observed public R&D data. - Scenario stylization: Global carbon budgets and tax pathways are stylized; outcomes may vary with alternative policy designs, regional policy heterogeneity, and real-world frictions.