The Pace and Impact of Deep Decarbonization:�Ways to Make Model Simulations More Useful and Realistic����David G. Victor��Alphabet Modeling Talk Series��10 February 2026�

1

Are there ways to improve the IAMs by teaching them about political economy?

2

Integrated Assessment Models (IAMs):�Tremendous Output

3

Source: Ou et al Science (Nov ‘21)

Example: GCAM-USA

4

Socioeconomics

Demand

Supply

Emissions

Global Change Assessment Model (GCAM)

​

Essentially ~all IAMs decarbonize by…

• Policy makers impose caps or carbon taxes
• Political support for these policies is assumed to exist
• Most decarbonization comes from electrification
• Needed infrastructure is built when needed
• Where electricity is not available, “clean molecules” are essential
• The energy system becomes more capital-intensive
• Investors invest where needed to meet policy goals
• Investors vary in their ability to see the future
• New technologies come into service by assumption, and/or improve through “learning by doing”
• If climate goals are stringent then carbon removal is deployed on a massive scale.
• The “geography” of emission controls is ~global

​

5

A strategy for progress: ten analysts

6

One way to think about modeling strategy

7

After Peng et al (2021)

Essentially ~all IAMs decarbonize by…

• Policy makers impose caps or carbon taxes
• Political support for these policies is assumed to exist
• Most decarbonization comes from electrification
• Needed infrastructure is built when needed
• Where electricity is not available, “clean molecules” are essential
• The energy system becomes more capital-intensive
• Investors invest where needed to meet policy goals
• Investors vary in their ability to see the future
• New technologies come into service by assumption, and/or improve through “learning by doing”
• If climate goals are stringent then carbon removal is deployed on a massive scale.
• The “geography” of emission controls is ~global

​

8

9

Global marginal abatement costs

10

Source: Iyer et al 2015

11

Source: BNEF (this morning)

Costs of Second-Best Policies (aka life):�The Problem of Short Time Horizons

12

Note: The effect of anticipation on regulatory costs for developing countries (% deadweight loss of economic output from developing countries in our “second best” scenario). Calculated from WITCH and reported in Bosetti and Victor (2011)

Perfect foresight (15+ year anticipation)

Blind Response

(0-5 yr time horizon)

Muddy foresight (~8 year anticipation)

Essentially ~all IAMs decarbonize by…

• Policy makers impose caps or carbon taxes
• Political support for these policies is assumed to exist
• Most decarbonization comes from electrification
• Needed infrastructure is built when needed
• Where electricity is not available, “clean molecules” are essential
• The energy system becomes more capital-intensive
• Investors invest where needed to meet policy goals
• Investors vary in their ability to see the future
• New technologies come into service by assumption, and/or improve through “learning by doing”
• If climate goals are stringent then carbon removal is deployed on a massive scale.
• The “geography” of emission controls is ~global

​

13

Huge Variation in Public Concern and Support

14

Source: Peng et al Nature Climate Change (2021)

Implied: Huge Variation in Policy Strategies

15

Source: Peng et al Nature Climate Change (2021)

Mitigation cost

16

Key finding:

• Sub-nationally, Heterogeneous scenario shifts mitigation from low- to high-supporting states
• Nationally, Heterogeneous scenario is only modestly (~10%) more expensive than Uniform.

Source: Peng et al Nature Climate Change (2021)

17

The small cost increase in Heterogeneous scenario is mainly due to the flexibility in moving energy production activities across states with the help of trade

(e.g., electricity generation, bioliquids production)

Source: Peng et al Nature Climate Change (2021)

Essentially ~all IAMs decarbonize by…

• Policy makers impose caps or carbon taxes
• Political support for these policies is assumed to exist
• Most decarbonization comes from electrification
• Needed infrastructure is built when needed
• Where electricity is not available, “clean molecules” are essential
• The energy system becomes more capital-intensive
• Investors invest where needed to meet policy goals
• Investors vary in their ability to see the future
• New technologies come into service by assumption, and/or improve through “learning by doing”
• If climate goals are stringent then carbon removal is deployed on a massive scale.
• The “geography” of emission controls is ~global

​

18

19

A theory of politics:�endogenizing policy as a function of politics and industrial political power

20

Declining Cost of low-carbon technologies

Greater Market Share and Revenues

Revenues and Success Generate Political Power

Policies that create advantages for decarbonization

Deep decarbonization

Essentially ~all IAMs decarbonize by…

• Policy makers impose caps or carbon taxes
• Political support for these policies is assumed to exist
• Most decarbonization comes from electrification
• Needed infrastructure is built when needed
• Where electricity is not available, “clean molecules” are essential
• The energy system becomes more capital-intensive
• Investors invest where needed to meet policy goals
• Investors vary in their ability to see the future
• New technologies come into service by assumption, and/or improve through “learning by doing”
• If climate goals are stringent then carbon removal is deployed on a massive scale.
• The “geography” of emission controls is ~global

​

21

Integrated Assessment Models (IAMs):�Tremendous Output

22

Source: Ou et al Science (Nov ‘21)

Carbon removal technology is…technology

23

Source: Dias et al (2025) ERL

Carbon removal technology is…technology

24

Source: Dias et al (2025) ERL

WHAT ARE THE RIGHT HISTORICAL ANALOGS FOR OCEAN SEQUESTRATION?

HATCH & USGS

• Scaleup depends on 2 factors
• Achievable diffusion (growth) rates in the short run
• Depends on how fast a DAC industry and supply chains could coalesce and grow
• Committed financial resources in the long run

​

• These 2 constraints lead to classic S-shaped diffusion

​

• No hard ceiling on deployment (a unique feature of DAC)

26

How Rapidly could Direct Air Capture (DAC) scale up and remove CO₂?

Hanna et al. 2021, Nat Commun

Essentially ~all IAMs decarbonize by…

• Policy makers impose caps or carbon taxes
• Political support for these policies is assumed to exist
• Most decarbonization comes from electrification
• Needed infrastructure is built when needed
• Where electricity is not available, “clean molecules” are essential
• The energy system becomes more capital-intensive
• Investors invest where needed to meet policy goals
• Investors vary in their ability to see the future
• New technologies come into service by assumption, and/or improve through “learning by doing”
• If climate goals are stringent then carbon removal is deployed on a massive scale.
• The “geography” of emission controls is ~global

​

27

The Strategy of Cutting Emissions�(table 1, Keohane & Victor, 2016)

28

The Strategic Game�(table 1, Keohane & Victor, 2016)

29

High Leverage Via Soot and SLCPs

30

Logic and Approach:�Reported over a few papers

• Aakre, S., Kallbekken, S., van Dingenen, R. & Victor, D.G., Nature Climate Change (2017).
• Sand, M., Berntsen, T.K., von Salzen, K., Flanner, M.G., Langner, J. & Victor, D.G., Nature Climate Change 6, 286-289 (2016)
• Crippa, M. et al., Atmospheric Chemistry and Physics 16, 3825-3841 (2016)
• Keohane, R.O. & Victor, D.G., Nature Climate Change 6, 570-575 (2016)
• Victor, D.G., Kennel, C.F., & Ramanathan, V., Foreign Affairs 91(3), 112-121 (2012)

​

31

32

33

Source: UNEP/WMO (2011), based heavily on the IIASA GAINS approach and ECLIPSE databases

TM5-FASST 56 Regions

34

Source: TMG-FASST User’s Guide, JRC/ACU (2016)

Cross-Border Climate and Non-Climate Benefits�(two largest export dyads by country; UNEP/WMO 2011 policy scenario, million USD)

35

Source: Aakre et al., Nature Climate Change (2017)

The Potential for Small Clubs of Methane and Soot Emitters:�Gains from Cooperation as a Function of Club Size

36

Final Observations

37

1. Capital is Key

38

Source: BNEF (this morning)

39

Source: BNEF (this morning)

2. Can we Model Different “varieties” of clean industry?

40

Source: Victor, 26 January 2026, Brookings

Climate Tech Thriving (?) Despite Chaos�The US Case

41

Source: BNEF (this morning)

Costs of Second-Best Policies (aka life):�The Problem of Short Time Horizons

42

Note: The effect of anticipation on regulatory costs for developing countries (% deadweight loss of economic output from developing countries in our “second best” scenario). Calculated from WITCH and reported in Bosetti and Victor (2011)

Perfect foresight (15+ year anticipation)

Blind Response

(0-5 yr time horizon)

Muddy foresight (~8 year anticipation)

Markets versus Regulation

• Theory says broad-based market strategies (cap & trade, carbon taxes) are best
• Real world: Regulation strongly favored over ”markets”
• Theory says “sticks” are important; reality reveals that “carrots” are more widely used
• Policy is fragmented into political jurisdictions

​

​

43

3. Analyst Confidence Is the Enemy�(IPCC WG3 Approved SPM)

44

Source: Victor (2016) Nature

Thank You

45