1. Rare Earth Substitutability in Clean Energy Technology: Using Expert Elicitation to Estimate Elasticities The Case of Permanent Magnets Braeton J. Smith and Roderick Eggert Division of Economics and Business Colorado School of Mines 33 rd USAEE/IAEE North American Conference Pittsburgh, PA October 25-28, 2015

2. Clean Energy and Rare Earth Elements Clean energy technology Wind Solar EVs Fluorescent lighting/LEDs High efficiency technology Direct drive motors in wind and EVs Direct drive motors Permanent magnets Permanent magnets (PMs) NdFeB SmCo Rare earth elements (REEs) Neodymium Dysprosium Praseodymium Samarium 2

3. Rare Earth Price Spike 3

4. Purpose Survey of magnet producer and user responses to rare earth price changes Two Purposes: 1.Historical - Document and quantify responses during and following the rare earth (RE) price shock of 2010/2011 2.Forward Looking - Assess possible future substitution and price responsiveness Estimate price elasticities of demand and substitution of REEs in PMs Apply expert elicitation methods to estimate forward-looking elasticities 4

5. Motivation Substitutability is a key indicator of material criticality Need to know what to expect if price shock happens again – not necessarily the same (non-constant elasticity) – Estimate elasticities not based on historical data Substitutability is an important consideration in simulation modeling 5

6. Substitution Taxonomy 1.No substitution - “do nothing” (low cost share, technically difficult) a)Cost pass-through b)Cost absorption 2.Material Substitution – general characteristics of end product are maintained a)Element for Element Substitute or reduce material content (e.g., reduce Dy content of NdFeB magnet) b)Alloy for Alloy Use different, but comparable, grade (e.g., N42SH instead of N45SH) 3.System Substitution – requires reengineering of overall system a)Magnet for Magnet Use different magnet type altogether (e.g., SmCo instead of NdFeB) b)System for System Use different system altogether (e.g., gearbox vs. direct drive) 4.Increased Manufacturing Efficiency – use materials more efficiently 6

7. Price elasticity of demand: Elasticity of substitution: – Expenditure on good 2 relative to good 1: – When E 21 < 1, c2/c1 falls by less than p2/p1 increases – When E 21 > 1, c2/c1 falls by more than p2/p1 increases – When E 21 = 1, c2/c1 is independent of relative prices 7 Price Responsiveness Quantification

8. Variables Target Variables – Price elasticity of demand (of Dy in a NdFeB magnet) – Elasticity of substitution (between Dy and Nd in a NdFeB magnet) Query Variables – Price and quantity (% composition) combinations – Composition and cost share for given price changes 8

9. Survey Structure Survey divided into 2 parts – Part 1: Historical Substitutability and Price Responsiveness Historical questions focus on events during and post RE price shock – Part 2: Future Substitutability and Price Responsiveness Probability elicitation of potential future responses Four categories provide framework for survey 9

10. Historical Questions Approach 1.No Substitution: – Sales price change during and after price shock – Cost shares before, during, and after price shock 2.Material Substitution (element for element): – Material composition of magnet before and after price shock 3.Material Substitution (alloy for alloy): – Change in sales and production shares of different grades of NdFeB magnets during and after price shock 4.Increased Mfg. Efficiency: – Reduction in material purchase requirement during and after price shock 10

11. Historical Anecdotes No substitution – Some evidence of cost pass-through in short run, but difficult for non- Chinese producers – Some evidence of cost absorption Material substitution – Evidence of element for element subst. over time for Dy, less so for Nd – Japanese companies reduced Dy content 6-8% (Roskill, 2015) System Substitution – No evidence of magnet for magnet substitution for wind technology, although much research into alternatives to NdFeB – Evidence of System for System subst. in wind technology Increased Mfg. Efficiency – Some evidence of recovery and reuse (Roskill, 2015) 11

12. Quantifying Uncertainty Probability Encoding and Sources of Bias – “Process of extracting and quantifying individual judgment about uncertain quantities” - Spetzler and Von Holstein (1975) Fixed probability (FP) vs. fixed value (FV) (Abbas et al., 2008) Many applications in natural sciences, risk analysis, and decision analysis Applications in economics mainly focus on R&D and technical breakthrough: – Baker et al. (2009), Baker and Keisler (2011), Anadon et al. (2012), Abdulla et al. (2013), Catenacci et al. (2013), Nemet et al. (2013), many others Quantifying uncertainty around future substitutability is new application 12

13. Define prototype magnet and market conditions now Elicit probabilities using mix of FV and FP methods for each potential response Generate probability distribution for each expert and response 13 Forward Looking Questions Approach

14. Example - Prototype Consider a sintered N45SH neo magnet that is used in a generator for a 2.5 MW wind turbine. 14 MaterialNd (%)Dy (%)Pr (%) N45SH23%7%3.3% PropertiesBrHcBHcJ(BH)maxTW Material Typical mT Typical Gauss Min kA/m Min Oersteds Min kA/m Min Oersteds Typical kJ/m3 Typical MGOe Max o C N45SH1,35013,50097912,3001,59220,00035444.0150

15. FV Example – Material Substitution If the price of the element in Column 1 were to change by the amount in Column 2, what is the probability that the material composition of the element in Column 5 would change by the amount in Column 6? E.g.: If the Dy price were to increase by 50 percent, what is the probability that the Dy content of the N45SH magnet would remain the same? 15 (1)(2)(3)(4)(5)(6)(7)(8)(9) Element Price Change Current Price ($/kg) Future Price ($/kg) Change Element Content Change Current Content Future Content Probability Estimate Dy+50%$500$750Dy0%7%7.00% Dy+50%$500$750Dy-5%7%6.65% Dy+50%$500$750Dy-10%7%6.30% Dy+50%$500$750Dy-25%7%5.25% Dy+50%$500$750Dy-50%7%3.50%

16. FP Example – Material Substitution Please give your 10 th, 25 th, 50 th, 75 th, and 90 th percentile estimates for the amount that the material content of the material in Column 1 would change for the price change indicated in Column 2. 16 Material Price Change 10 th 25 th 50 th 75 th 90 th Dy+50% Dy+100% Dy+500% Dy+1000%

17. Thanks! Questions? 17

18. Hypothetical Result 18