Technology cost
forecasting and the
energy transition
1
Dr Rupert Way
Affiliated researcher in the Complexity
Economics Group at the Institute for New
Economic Thinking
University of Oxford
OECD IFI Climate working group
8 December 2022

2
Rupert Way Matt Ives
Penny Mealy Doyne Farmer

Motivation
3
• All energy-economic models require cost forecasts for all energy technologies
• Where do these forecasts come from?
• How do they model technological change?
• How reliable have their forecasts been?
• How do technology cost forecasts affect model output?

Motivation
4
Whydowebelievewhatwebelieveaboutmitigationcosts?
Technology cost forecasts
• All energy-economic models require cost forecasts for all energy technologies
• Where do these forecasts come from?
• How do they model technological change?
• How reliable have their forecasts been?
• How do technology cost forecasts affect model output?

5
• “Stopping climate change
will be very slow or very
expensive”
• “Lower economic growth,
lower energy consumption”
• “Higher energy prices, lots of
Carbon Capture and
Storage”
• Why?… Because Integrated
Assessment Models (IAMs)
say it will cost 1 - 5% of GDP
The existing narrative

6
Howwellhavemajormodels
capturedtechnological
progress?
10000
1000
100
1980 2000 2020 2040
$(2020)/kW
Actual PV cost
2014 IAMs
2018 IAMs (SR15)
2019 IAMs
2022 IAMs (AR6)
PV system costs in IAMs 2010-2022
• Systematically overestimated future
costs of key green technologies
Why?
• Floor cost assumptions
• Deployment rate limits
• Intermittency constraints
• Cost-production feedback loop

7
It’snotjustsolar
• Different technologies progress at
very different rates
• Fossil fuels, nuclear, CCS, biofuels
(pipes, pistons, fluids, combustion)
• High knowledge technologies
(electricity / electronics /
computing) progress rapidly
• PV, wind, batteries, EVs,
electrolyzers, P2X fuels, heat
pump heating
100 years of energy technology costs

Ourstrategy
8
• Collect as much tech data as possible
• Backtest different cost forecasting
models to evaluate their performance
• Choose the “best” forecasting model
• Apply to the energy system
• Empirical laws work well (Moore/Wright)
• For progressing techs, we use Wright’s
law (aka learning curve / experience
curve / endogenous tech change etc.)
• For others, we use a simple AR1 model Source: Lafond et al. 2018

Probabilisticlearningcurveforecasts
9
• Increasing experience leads to
higher probability of progress
along the experience curve
• i.e. if we invest in certain techs
then we expect innovation
• BUT… different techs progress at
different rates, so we need to
focus on high learning techs
Cost
($/MWh)
Experience (TWh)
Solar PV Cost vs. Experience

Threescenarioswithidenticalusefulenergy
10
• Combine exogenous deployment trajectories to form full scenarios
• Generate cost forecasts, sum costs, apply discounting to calculate Net Present
Cost of each scenario
Fast Transition Slow Transition No Transition
2020 2040 2060 2020 2040 2060 2020 2040 2060

Results:example-solarforecasts
11
Cost
($/MWh)
PV costs and experience
Cumulative
generation
(TWh)
Year
Fast transition
No transition

12
• Faster deployment of key
green techs pushes cost
distributions down
• At 1.4% discount rate, the
Expected NPC saving is
12 $TN
• Avoided climate damages
lead to 30-700 $TN extra
savings
Conclusion:afasttransitionislikelycheaperatalldiscountrates

Summary
13
• Fossil fuel costs have barely changed over the last century
• Key green technology costs have been falling consistently for decades
• These trends are likely to continue
• We can accelerate these cost reductions by accelerating experience (in specific
technologies only)
• Major models do not capture these trends well, have systematically overestimated
key green tech costs, and hence transition costs
• The chance of all key green tech costs failing to fall is very small, yet this is the
space of scenarios considered by major models and the IPCC
• Our model does capture these trends and shows net savings of many trillions
• High FF prices just make the learning dynamic more obvious, increase savings

14
• We should move beyond the “all of the above” approach, e.g. hydrogen cars,
nuclear, CCS have made no progress despite sustained effort
• We will need new factories and supply chains, upgraded grid infrastructure, EV
charging etc, plus… skills to get all of this equipment installed and maintained
• There will be huge opportunities in the transition
• We should ensure each sector and industry is prepared to use cheap
renewables as soon as they can. This will unlock the largest savings
• We must overturn the notion that transition is expensive, it’s not
• But also… we can make it even cheaper, by coordinating action and going
faster
Moresummary

Thank you
15

16
References
Pindyck 1999 - The long-run evolution of energy prices
Nagy et al. 2013 - Statistical basis for predicting technological progress
Trancik et al. 2015 - Technology improvement and emissions reduction as mutually reinforcing
efforts
Farmer & Lafond 2016 - How predictable is technological progress?
Lafond et al 2018 - How well do experience curves predict technological progress- A method for
making distributional forecasts
Meng et al. 2021 - Comparing expert elicitation and model-based probabilistic technology cost
forecasts for the energy transition
Way et al. 2022 - Empirically grounded technology forecasts and the energy transition