The economic analysis of climate change represents one of the most complex challenges in modern resource economics. It is not merely a positive undertaking of prediction but is inextricably linked with normative frameworks that dictate how future welfare is valued and how existential risks are managed.

The assessment of climate policy measures depends fundamentally on three distinct yet interdependent modeling decisions—what we call the “Archimedean point” of any cost-benefit analysis:

“Climate change and technological progress are closely intertwined through a dichotomy of economic externalities. Pollution represents a negative externality—social costs exceed private costs. Conversely, knowledge and innovation stand as a positive externality—firms cannot fully internalize the returns on R&D investments.”

Consequently, the unregulated market provides too much pollution and too little innovation. Understanding this dichotomy is essential for designing effective climate policy that addresses both market failures simultaneously.

The literature is characterized by a dichotomy between “bottom-up” and “top-down” modeling approaches, each offering distinct advantages and limitations in capturing the nuances of technical progress (Löschel, 2002).

Bottom-Up Models: The Engineering Perspective

Bottom-up models (MARKAL, MESSAGE, POLES) are rooted in detailed technological description of the energy system. They emphasize specific engineering characteristics and treat the broader economic structure rather rudimentarily. Their strength lies in explicitly modeling learning-by-doing—since specific technologies exist as distinct entities, costs can be dynamically linked to cumulative installed capacity.

Top-Down Approaches: The Macroeconomic Perspective

Top-down models emphasize aggregate economic behavior and can be subdivided into:

• Macroeconometric Models (DGEM): Based on time-series data with neo-Keynesian theoretical foundations, suitable for short- to medium-term forecasting (Carraro, 2002).
• Computable General Equilibrium Models (PACE, GEM-E3, MIT-EPPA): Arrow-Debreu frameworks mapping the economy using CES production functions, analyzing structural adjustments and welfare effects across sectors.
• Integrated Assessment Models (DICE, RICE, MERGE, WITCH, FUND): These close the loop between economy and climate system, integrating economic modules with geophysical climate modules and damage functions—the primary tool for cost-benefit analysis of climate policy.

For decades, the standard practice was to treat technological change as exogenous. In these frameworks, energy efficiency improvements are modeled as a pure function of time—essentially as “manna from heaven” (Grubb et al., 2002).

Autonomous Energy Efficiency Improvement (AEEI)

The most common representation is the AEEI parameter—capturing non-price-driven decline in energy intensity over time. While computationally tractable, this approach implies technological progress is immutable and proceeds at a fixed rate, regardless of carbon prices, R&D subsidies, or environmental regulations.

Richels and Blanford (2008) demonstrated the sensitivity of baseline emissions to AEEI assumptions:

Backstop Technologies

The second pillar is the “backstop” technology (Hotelling, 1931; Dasgupta & Heal, 1974; Nordhaus, 1973)—a carbon-free energy source available in unlimited quantities at constant marginal costs. While deterministic in most models, Sue Wing (2006) introduced “semi-endogenous” features where backstop costs decline over time.

The transition from exogenous to endogenous technological change represents a paradigm shift in climate-economic modeling. This shift acknowledges that innovation is a response to economic incentives—not an autonomous process.

Price-Induced Technical Change

The “induced innovation hypothesis” dates back to Hicks (1932): changes in relative factor prices induce innovations that economize on the factor that has become relatively more expensive. In climate economics, this implies that higher carbon prices should induce energy-saving innovations.

Empirical evidence is substantial: Popp (2002) found that a 10% increase in U.S. energy prices leads to a 3.5% increase in energy-related patents. Newell, Jaffe, and Stavins (1999) demonstrated that rising electricity prices significantly affected the rate and direction of innovation in air conditioners and water heaters.

Learning-by-Doing (LBD)

Arrow's (1962) seminal contribution established that productivity improvements arise as a byproduct of production itself—what he called “learning by experience.” Wright (1936) had earlier documented this empirically in aircraft production, giving rise to Wright's Law:

The implications for climate policy are profound: early deployment subsidies can “buy down” the learning curve, making clean technologies cost-competitive faster than market forces alone would achieve.

The most recent theoretical advance comes from Directed Technical Change (DTC)—primarily associated with the work of Daron Acemoglu (2002) and applied to climate by Acemoglu et al. (2012, 2016).

DTC explicitly models the allocation of scientific resources between clean and dirty technologies. Unlike earlier frameworks, DTC captures not just the rate but the direction of innovation. Key insights include:

• Path Dependence: Innovation creates its own momentum. If dirty technologies have a productivity advantage, market forces will continue directing R&D toward them—unless policy intervenes.
• Policy Timing: Early intervention can redirect the trajectory permanently. Delay allows dirty technology lock-in to deepen.
• Policy Mix: Carbon taxes alone may be insufficient. The optimal policy combines pricing (Pigouvian taxes) with technology policy (R&D subsidies) to address both externalities.

“If the initial productivity gap between dirty and clean technologies is large and both types of research exhibit diminishing returns, then an immediate switch to clean research becomes optimal. The carbon tax alone may be unable to redirect research activities when dirty technologies are significantly more advanced.”— Acemoglu et al. (2012)

This framework has been implemented in the WITCH model and has profound implications: it suggests that delay is costly not just in emissions terms, but in terms of technological lock-in that may require even more aggressive intervention later.

The evolution from exogenous to endogenous technological change in climate-economy models represents a major advance in our ability to assess climate policy. Yet significant challenges remain:

Despite these challenges, the central message is clear: technological change is not manna from heaven. It responds to prices, policies, and expectations. Climate policy is not merely about reducing emissions today—it is about redirecting the entire trajectory of technological development toward a sustainable path. This insight has profound implications for the design, timing, and ambition of climate policy.