Characterizing long term technological change Prof. Gregory Nemet November 2015
Technological change: surprise and stationarity Characteristics Ex ante ignorance Skewed outcomes Pervasive spillovers Combinatorial Depreciating knowledge Interaction w/ production Public early private later Drivers of smoothness long lifetimes risk aversion incremental improvement aggregation 2
80 years of PV prices
80 years of PV prices 4
Missing perspectives on technological change 1. incentives in bureaucracies 2. knowledge dissemination 3. international cooperation 4. credibility of policy targets 5. public attitudes to novel technologies 6. characterization of the social returns to innovation 7. near-term metrics to orient innovation toward longer term human needs. Nemet, G. F., A. Grubler and C. Wilson (2015). "Re-orienting energy innovation to address human needs requires new social science." Nature Energy In review. 5
Appendix 6
The closer you look To what extent does aggregation smooth surprises and disappointments? 7
x. Tech change: characteristics to address Skewed outcomes Diminishing returns Crowding out Technological opportunity Public early private later Interactions with production Depreciating knowledge Small $s vs deploy, taxes, damages 8
x. Tech change: characteristics to address The closer you look, the less predictable it seems Is there a benefit to not looking too closely? Does aggregation enable smootheness? 9
x. Tech change: drivers of smooth change S-shaped adoption Long lived assets Risk aversion in adoption 10
1. LEARNING CURVES 11
1. Learning curves: Appeal and problems Appeal: Makes tech dynamic Compact Data availability Some theory Historical evidence Goodness of fit Two broad problems: 1. Uncertainty poorly characterized 2. Omitted variable bias 12
1. Learning curves: LR variation across technologies LR median =0.185 ±1σ=0.064 0.306 A small change in learning rate makes a difference. Nemet, G. F. (2009). "Interim monitoring of cost dynamics for publicly supported energy technologies." Energy Policy 37(3): 825-835. 13
1. Learning curves: LR variation within technologies Learning rates for wind power (1981 2006) calculated for all periods >=10 years (n=153). Nemet, G. F. (2009). "Interim monitoring of cost dynamics for publicly supported energy technologies." Energy Policy 37(3): 825-835. 14
1. Learning curves: Other factors affect performance Other factors: 1. R&D-induced techδ 2. Scale effects Implications: - LR bias - error not random 3. Input costs 4. Diminishing returns - Forecasts biased 5. Knowledge depreciation - Incentives wrong 6. Serendipitous spillovers 7. Time 15
1. Learning curves: Two summary points 1. If we are to depend on learning curves, we need to characterize reliability of resulting forecasts. 2. Need to more fully represent the drivers of technological change. 16
4. Incentives If policy commitments are not fully credible Nemet question to venture capitalist: How do you value the benefits of policy in your decisions to invest in start-up companies? We ignore it. What the government giveth, it can taketh away. - Venture capitalist in energy sector 17
APPENDIX 18
2. R&D: Demand curves for coal CCS 19
Outcomes and conclusions PDF of costs Benefits to diversification Next: Use integrated assessment model Effect of policy changes on: emissions, concentrations, abatement costs 20
4. Incentives How did solar get cheap? 21
Summary of results Results: 1. Evidence that firms did learn from experience, 2. Firms learned from other firms experience 3. Clearer for operating performance than for performance of new installations...but... 4. rapid depreciation of knowledge from LbD 5. diminishing returns from LbD 6. Effects of policy varied: Capital cost incentive had negative effect Production incentives had a mixed effect 22
Interpretation for policy Policy implications: Spillovers exist so need demand-side policy...even in the presence of pollution pricing. Some evidence that performance incentives really do make a difference Codification to address knowledge depreciation? Open questions on policy implications: Geographic extent of spillovers LbD embodied in new technology? Systemic benefits? e.g. grid operators Firms are risk-averse? Unproven technology 23
5 premises for energy policy analysis: 1. Addressing energy problems requires changes in behavior and technology 2. Energy policy involves multiple objectives 3. Inertia in the energy system 4. Historical volatility in: policy, markets, and public interest 5. Technological change depends on expectations 24
purpose this workshop intends to strengthen current and future efforts to analyze technological change and to introduce endogenous technology dynamics in large scale energy or integrated assessment models. In your session on "taking stock", we envisage you will present insights into the current status of research on issues related to cost dynamics, learning curves, R&D and innovation, such as: factors explaining cost dynamics along the learning curve determinants of the pace of global innovation in energy technologies other relevant insights to date 25