## wpiea2026072-source-pdf

## Source details

**Canonical URL:** [wpiea2026072-source-pdf](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026072-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026072-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026072-source-pdf.pdf.json)

---

### Market structure, key features, and concentration
- REEs comprise 17 metals, grouped into light (LREEs) and heavy (HREEs).
- HREEs are less abundant and typically occur in lower concentrations than LREEs.
- Four elements—neodymium (Nd), praseodymium (Pr), dysprosium (Dy), and terbium (Tb)—account for 96 percent of total rare earth oxide (REO) market value while representing only 23 percent of production by weight.
- Permanent magnets:
  - Account for roughly 75–85 percent of total REE demand by value.
  - Rely predominantly on Nd and Pr (approximately 85 percent of REE inputs in magnet production), with smaller but performance-critical contributions from Dy and Tb.
- Global supply-chain stages: ore extraction → benefaction → solvent extraction (oxide separation) → metal refining/alloy production.
- Geographic concentration (approximate shares):
  - Oxide separation capacity: China ≈ 85–90 percent.
  - Metal refining: China ≈ 90–95 percent.
- HREEs concentration (China + Myanmar combined):
  - Mining: 98 percent
  - Oxide separation: 96.5 percent
  - Metal refining: 95 percent
  - Permanent magnet manufacturing: 89.5 percent
- LREE mining diversification notes:
  - China’s share of global LREE mining ≈ 97 percent in 2010 and ≈ 58 percent in 2024.
  - Mine production of LREEs outside China and Myanmar reached 98,000 tons in 2025, accounting for 25 percent of global output.
- Global reserves outside China exceed 41 million tons of rare earth oxide equivalent distributed across Brazil, Australia, Russia, Vietnam, United States, and several other jurisdictions.

### Persistence of concentration, costs, and project lead times
- Causes of persistent concentration:
  - Technological complexity and tacit knowledge in processing and separation.
  - High capital intensity and environmentally damaging, energy-intensive separation processes.
  - Regulatory burdens and specialized expertise.
- Project lead times:
  - New REE projects outside China typically require 10–15 years or more from discovery to commercial production.
- Project economics (21 non-Chinese projects assessed):
  - Substantial dispersion in capital and operating costs; many projects on the upper end of the global cost curve.
  - Only a limited number of projects profitable at current price levels; many require substantially higher prices.
- Historical price dynamics discouraged irreversible investments (example: 2010–2011 price spike followed by prolonged subdued prices).

### Two-stage market model (miners and refiners) — structure and first-order conditions
- Two sequential stages: upstream extraction (miners) and downstream refining (refiners).
- Prices:
  - Upstream price: p^x_t (raw REEs).
  - Downstream price: p^y_t (refined REEs).
- Miners:
  - Extraction dynamics: x_jt = (1 − η_j) x_j,t−1 + i^x_jt, with exponential decline rate η_j > 0.
  - Objective (discount factor β_j ∈ (0, 1)):
    - max_{x_jt,i^x_jt} Σ_{t=0}^∞ β_j^t [ p^x_t (1 + s^x_jt) x_jt − (1 + s^ix_jt) c^ix_j(i^x_jt) − c^x_j(x_jt) ]
  - First-order condition (net marginal CAPEX):
    - (1 + s^ix_jt) ∂c^ix_j/∂i^x_jt = β_j (1 − η_j) (1 + s^ix_jt+1) ∂c^ix_j/∂i^x_jt+1 + p^x_t (1 + s^x_jt) − ∂c^x_j/∂x_jt
  - Aggregate mined supply: R^s_t ≡ Σ_{j=1}^J x_jt
- Refiners:
  - Capacity dynamics: y_jt = (1 − δ_j) y_j,t−1 + i^y_jt, with decline rate δ_j > 0 and y_jt = z_jt (linear technology).
  - Objective:
    - max_{z_jt,y_jt,i^y_jt} Σ_{t=0}^∞ β_j^t [ p^y_t (1 + s^y_jt) y_jt − p^x_t z_jt − (1 + s^iy_jt) c^iy_j(i^y_jt) − c^y_j(y_jt) ]
  - First-order condition (net marginal CAPEX):
    - (1 + s^iy_jt) ∂c^iy_j/∂i^y_jt = [ p^y_t (1 + s^y_jt) − p^x_t ] − ∂c^y_j/∂y_jt + β_j (1 − δ_j) (1 + s^iy_jt+1) ∂c^iy_j/∂i^y_jt+1
    - Gross refining margin = [ p^y_t (1 + s^y_jt) − p^x_t ]
  - Aggregate raw REE demand: R^d_t ≡ Σ_{j=1}^J z_jt = Σ_{j=1}^J y_jt

### Final-good demand, equilibrium, and steady-state characterization
- Country demand: q_jt = a_jt (p_y_t)^−ε with elasticity ε ∈ R_+ common across countries.
- World aggregate demand: Q_t = A_t (p_y_t)^−ε where A_t ≡ Σ_{j=1}^J a_jt.
- Competitive equilibrium defined by miners’ and refiners’ optimality and market clearing:
  - Raw REE market: Σ_{j∈J} x_jt = Σ_{j∈J} y_jt
  - Refined REE market: Σ_{j∈J} y_jt = Σ_{j∈J} q_jt
- Functional marginal-cost forms:
  - Miners:
    - ∂c_ix_j/∂i_x_jt = α_ix_j + γ_ix_j (x_jt − (1−η_j) x_jt−1)
    - ∂c_x_j/∂x_jt = α_x_j
  - Refiners:
    - ∂c_iy_j/∂i_y_jt = α_iy_j + γ_iy_j (y_jt − (1−δ_j) y_jt−1)
    - ∂c_y_j/∂y_jt = α_y_j
- Steady-state (no industrial policies s_· = 0, β ≡ 1/(1 + r)):
  - Mining output: x_j = p_x − α_x_j − β_x α_ix_j B_x_j, where β_x ≡ 1 − β(1−η) = (r + η)/(1 + r) and B_x_j ≡ γ_ix_j η β_x.
  - Processing output: y_j = p_y − p_x − α_y_j − β_y α_iy_j B_y_j, where β_y ≡ 1 − β(1−δ) = (r + δ)/(1 + r) and B_y_j ≡ γ_iy_j δ β_y.
  - Aggregate demand in steady state: Q = A (p_y)^ε.
  - Equilibrium prices satisfy:
    - p_x = (p_y Γ_y − Ψ_y + Θ_x) / (Γ_y + Γ_x)
    - A (p_y)^ε (Γ_y + Γ_x) + Θ_x Γ_y + Ψ_y Γ_x = p_y Γ_y Γ_x
  - Parameter bundles:
    - Γ_y ≡ Σ_j 1/B_y_j = Σ_j 1/(γ_iy_j δ β_y)
    - Ψ_y ≡ Σ_j α_y_j + β_y α_iy_j B_y_j
    - Γ_x ≡ Σ_j 1/B_x_j = Σ_j 1/(γ_ix_j η β_x)
    - Θ_x ≡ Σ_j α_x_j + β_x α_ix_j B_x_j

### Calibration: key choices, parameters, and numeric anchors
- Calibration base year: 2024 using feasibility studies, industry reports, academic literature.
- Demand and growth:
  - Short-run price elasticity ε = −0.15 (unweighted average for Nd, Pr, Dy, Tb).
  - China: 50 percent of global refined REE consumption in base year.
  - Demand growth rates (IEA Global Critical Minerals Outlook (2025)):
    - annual growth 2025–2029 = 4.7 percent
    - annual growth 2030–2034 = 1.47 percent
    - ≥ 2035 = 0 percent
- Breakeven representative outside China:
  - Representative breakeven price (2024 USD per kg NdPr) = $77/kg = $43/kg operating + $34/kg annualized capital charge.
- User-cost identity and parameter computation:
  - r = 0.08; η = 0.035; (r + η)/(1 + r) = (0.08 + 0.035)/1.08 ≈ 0.1065.
  - α_i (outside China) = 34 / 0.1065 ≈ 318 USD/kg of annual NdPr capacity.
- Mining–refining decomposition (outside China):
  - Operating costs: 70% mining, 30% refining → α_x ≈ 30 USD/kg; α_y ≈ 13 USD/kg (0.70 × 43 ≈ 30; 0.30 × 43 ≈ 13).
  - Annualized capital: 40% mining, 60% refining:
    - Mining capital annuity ≈ 13.6 USD/kg ⇒ α_ix = 13.6 / 0.1065 ≈ 127 USD/kg.
    - Refining capital annuity ≈ 20.4 USD/kg ⇒ α_iy = 20.4 / 0.1065 ≈ 191 USD/kg.
  - Resulting outside-China parameters: α_x = 30; α_ix = 127; α_y = 13; α_iy = 191 USD/kg.
  - Normalization by α_iy = 191 yields ̄α_x = 30/191, ̄α_ix = 127/191, ̄α_y = 13/191, ̄α_iy = 191/191.
- Chinese calibration anchors:
  - Historical median NdPr price = $42.5/kg (2006–2010 inference).
  - Decomposition: 75 percent operating ($31.9/kg) and 25 percent capital ($10.6/kg).
  - α_i (China) ≈ 10.6 / 0.1065 ≈ 99 USD/kg.
  - Stage allocation: 69 percent mining and 31 percent refining → stage totals ≈ $29.3/kg (mining) and $13.2/kg (refining).
  - Capital allocation: 30 percent mining, 70 percent refining:
    - α_ix (China) ≈ 3.2 / 0.1065 ≈ 30 USD/kg.
    - α_iy (China) ≈ 7.4 / 0.1065 ≈ 69 USD/kg.
  - Residual operating intercepts: α_x = 26 USD/kg; α_y = 6 USD/kg.
  - Normalized Chinese intercepts by 191: ̄α_x = 26/191, ̄α_ix = 30/191, ̄α_y = 6/191, ̄α_iy = 69/191.
  - Notable: approximately 58 percent of the total cost gap between Chinese producers and those outside China originates in the refining stage.
- Cost uniformity across non-Chinese regions: identical cost parameters outside China (motivated by limited evidence of systematic cost ranking).

### Investment slope parameters (γ) calibration and identification
- Decomposition: γ_is_j ≡ κ_s ̄γ_j_is for s ∈ {x, y} with κ_s common across countries.
- Calibration targets: match observed 2024 mining and refining production shares; U.S. normalized so ̄γ_ix_US = ̄γ_iy_US = 1.
- Identification outside China:
  - x_j = (p_x − α_x − β_x α_ix)/(γ_ix_j η β_x) ⇒ x_j / x_m = γ_ix_m / γ_ix_j
  - Hence ̄γ_ix_j = x_US / x_j for regions outside China (analogous for refining).
- China requires joint computation with equilibrium p_x due to different cost intercepts.
- Regions with zero production assigned large γ to keep them inactive at initial prices.

### Price-level calibration, scaling, and steady-state outcomes
- Two scaling parameters determined: κix and κiy.
- Identification targets:
  - Refined price normalized: p_y = 1 in initial steady state.
  - Mining-to-refining price ratio target: p_x/p_y = 0.55.
- Calibrated scaling parameters:
  - κix = 600
  - κiy = 5,000
- Reported steady-state outcomes (Table 2 reported pairs):
  - p_y_SS target/value = 1.00 / 0.96
  - p_x_SS/p_y_SS target/value = 0.55 / 0.55

### Baseline simulation inputs and horizon
- Simulation horizon: T = 150 years starting in 2025.
- Initial (2024) base-year production (NdPr-equivalent, kt/yr):
  - Mining output: U.S. 43.0; China+ 285.8; EU 0.0; Japan 0.0; Anglophone 18.0; RoW 6.5.
  - Refining output: U.S. 1.13; China+ 86.72; EU 0.67; Japan 0.67; Anglophone 0.00; RoW 5.42.
- Discount and depreciation: r = 0.08; η = 0.035.
- Selected cost intercepts ($/kg):
  - αxj: U.S. 30; China+ 26; EU 30; Japan 30; Anglophone 30; RoW 30.
  - αyj: U.S. 13; China+ 6; EU 13; Japan 13; Anglophone 13; RoW 13.
  - αixj: U.S. 127; China+ 30; EU 127; Japan 127; Anglophone 127; RoW 127.
  - αiyj: U.S. 191; China+ 69; EU 191; Japan 191; Anglophone 191; RoW 191.

### Elasticities and interpretation (short-run and long-run)
- Reported elasticities (Table 3 values):
  - Mining supply elasticity:
    - Net-REE Importers: SR 0.0037; LR 1.79
    - China: SR 0.0031; LR 1.41
  - Mining demand elasticity:
    - Net-REE Importers: SR −0.0044; LR −2.09
    - China: SR −0.0031; LR −1.46
  - Refining supply elasticity:
    - Net-REE Importers: SR 0.0083; LR 3.81
    - China: SR 0.0056; LR 2.66
  - Refining demand elasticity (refined product): SR and LR = −0.15
- Interpretation:
  - Short-run elasticities near zero imply limited short-run quantity adjustments.
  - Long-run elasticities are substantially higher, especially for net-REE importers.

### Supply disruption / market fragmentation scenario (two-block fragmentation)
- First-pass percent changes (Table 4, relative to pre-shock quantities = 100):
  - Mining:
    - China+ supply of mined product: −80.88
    - China+ demand of mined product: −91.67
    - Excess supply in mining markets outside China: +10.78
  - Refining:
    - China+ supply of refined product: −91.67
    - China+ demand of refined product: −50.00
    - Excess supply in refining markets outside China: −41.67
- Dynamic model impact on prices and quantities (key numeric outcomes):
  - Impact price spikes:
    - Refined REE price spikes to almost 80 times baseline.
    - Mined REE price increases to around 7.4 times baseline.
  - Quantity and investment responses on impact:
    - Consumption of refined products falls by around 48% on impact.
    - Investment in refining capacity increases, leading to more than threefold increase in supply of refined products.
    - Demand for mined inputs increases to about 1.3 times baseline on impact.
  - Long-run adjustments:
    - Mining output rises to about 2.1 times pre-shock level.
    - Refining output rises to about 4.5 times pre-shock level.
    - Upstream equilibrium price falls from 7.4 to 1.7 times baseline (remaining about 70% above baseline).
    - Downstream refined-product price declines from 80 to 2.5 times baseline (about 150% above baseline).
    - Final consumers converge to around 89% of baseline level (long-run loss ≈ 11%).
- Interpretation:
  - Low short-run elasticities produce large short-run price spikes and demand collapses; long-run CAPEX-driven expansion mitigates price spikes and partially restores consumption.

### Industrial policy instruments, targets, and quantified scenarios
- Policy instruments modeled:
  - CAPEX subsidies to refiners: s^iy_jt ∈ [−1, 0].
  - Stylized revenue-support (proportional sales subsidy / price floor): s^y_jt ≡ f^y_j − 1 with f^y_j ≥ 1.
- Implementation scenarios:
  1. Unilateral U.S. action (policies apply only to U.S.-based producers).
  2. Simultaneous action (U.S., EU, Japan, Australia, and other economies apply instruments uniformly).
- Targets assessed: 25% de-risking by 2035 and 50% de-risking by 2035.
- Parameter values (Table 5 reproduced):
  - CAPEX Subsidy, unilateral (j = U.S.): −74.41% (25% target), −88.72% (50% target)
  - CAPEX Subsidy, simultaneous (j̸= China): −73.97% (25% target), −87.91% (50% target)
  - Price Floor, unilateral (j = U.S.): 2.19 (25% target), 4.08 (50% target)
  - Price Floor, simultaneous (j̸= China): 2.00 (25% target), 3.05 (50% target)
- De-risking definition:
  - De-risking is the share of refined REE consumption produced at home or from countries other than the market leader (analysis focuses on China only).
  - Construction: exporter shares xs_{j,t}, allocated net imports m_{ij,t}, degree of de-risking σ_{i,t} = 1 − max_{j∈J} m_{ij,t} / q_{i,t}.

### Key simulation findings on de-risking achievement and dynamics
- All policies achieve their target exactly in 2035 by construction.
- Pre-2035 dynamics:
  - Unilateral policies deliver higher pre-2035 de-risking for the U.S. than simultaneous action.
  - In the long run, simultaneous policies outperform unilateral ones.
  - Conditional on simultaneity, CAPEX subsidies and revenue-support policies are broadly equivalent in delivering de-risking over the first 13–15 years.

### Fiscal costs: expressions and quantitative results
- Fiscal cost formulations (relative to baseline consumption):
  - If CAPEX subsidy:
    - Φ_{jt} = s^iy_{jt} c^iy_j(i^y_{jt}) p^y_{t,BAS.} q_{jt,BAS.}
  - If revenue support:
    - Φ_{jt} = p^y_t(f^y_j − 1) y_{jt} / (p^y_{t,BAS.} q_{jt,BAS.})
- Main fiscal-cost results (selected numeric outcomes):
  - Simultaneous action reduces fiscal cost for the U.S. and participating countries relative to unilateral action.
  - CAPEX subsidies frontload costs and are initially more expensive than revenue support, but are substantially less costly in the long run.
  - Revenue-support (price-floor) costs backload and remain elevated as the economy converges.
  - Example unilateral revenue-support (relative to baseline consumption in 2025):
    - Cost for 25% de-risking by 2035: about 13.4% of baseline consumption in 2025.
    - Cost for 50% de-risking by 2035: exceeds 44.5% of baseline consumption in 2025.
    - Long-run stabilization of unilateral revenue-support fiscal cost: around 41.6% (25% target) and over 269% (50% target).
  - Aggregate fiscal-cost magnitudes (total fiscal cost 2025–2035 for U.S. under 25% target):
    - CAPEX subsidies: 134% of the annual U.S. REE market under unilateral action and 90% under simultaneous action.
    - Stylized revenue-support policies: 248% under unilateral policies and 172% under simultaneous policies.
  - Fiscal costs are substantially higher for 50% de-risking targets.

### Spillovers across mining, refining, consumption, and prices (selected values)
- Mining output changes:
  - 25% target, unilateral: China +0.26%; net importing countries +0.3%.
  - 50% target, unilateral: China +0.66%; net importing countries +0.79%.
  - Simultaneous action (larger increases): China +1.44% (25%) and +2.8% (50%); net importing countries +1.72% (25%) and +3.3% (50%).
- Refining output changes:
  - China declines:
    - 25% target: −2.5% (unilateral) and −14.8% (simultaneous).
    - 50% target: −6.8% (unilateral) and −27.7% (simultaneous).
  - Rest of world increases substantially:
    - 25% target: +25.9% (unilateral) and +154.7% (simultaneous).
    - 50% target: +70.8% (unilateral) and +290% (simultaneous).
- Consumption:
  - Global consumption increases modestly:
    - 25% target: +0.27% (unilateral) and +1.5% (simultaneous).
    - 50% target: +0.69% (unilateral) and +2.91% (simultaneous).
- Prices:
  - Unilateral subsidies, mining prices on impact: +1.48% (25% target) and +3.82% (50% target); long-run +0.17% and +0.49% respectively.
  - Simultaneous subsidies, mining prices on impact: +8.66% (25% target) and +17.66% (50% target); long-run +1.13% and +2.22% respectively.
  - Refining prices decline in response to policy-driven refining expansions.

### Distributional effects and fiscal-cost allocation
- Unilateral U.S. policies:
  - Increase U.S. de-risking while leaving other net importers broadly unchanged or slightly worse off.
- Simultaneous policies:
  - Enable higher de-risking for U.S., EU, and Japan.
  - Baseline de-risking: EU 6%; Japan 8.9%.
  - Under 25% U.S. target with simultaneous action: EU reaches 17.1%; Japan reaches 23.1%.
  - Fiscal-cost distribution: CAPEX subsidy costs are nearly as high in Japan as in the U.S. for the 25% target and significantly lower in Europe; costs concentrate in countries with larger refining sectors.

### Nonlinearities, subsidy elasticities, and policy implications
- Nonlinear cost dynamics:
  - Subsidy rates below 60% allow at most about 18% de-risking for the U.S.
  - Increasing subsidy from 80% to 90% raises de-risking by over 20 percentage points.
  - Moving from 85% to 95% raises de-risking by approximately 45 percentage points.
- Fiscal-cost elasticity with respect to de-risking > 1 across considered subsidy ranges (fiscal cost rises more than proportionally with de-risking).
- Comparative policy assessment:
  - De-risking is fiscally costly but less costly when pursued simultaneously by multiple importers.
  - CAPEX subsidies perform better in the long run (target expansion of production capacity).
  - Revenue-support (price-floor) mechanisms are less costly initially but become more expensive over time.
  - Ambitious de-risking targets (e.g., 50%) are associated with exponentially rising fiscal costs.

### Dataset construction and economic supply-curve methodology (project-level)
- Dataset:
  - Mine-by-mine assessment covering 21 non-Chinese projects in Australia, Sub-Saharan Africa, Brazil, Canada, Greenland, Sweden, and the United States; projects included if sufficient public documentation available.
  - Primary data sources: S&P Capital IQ Pro database and supporting disclosures.
- Manual reconstruction:
  - Extracted TREO volumes, oxide distributions, annual production quantities.
  - Non-REE by-product revenues included when reported; incentive price excludes non-REE revenues.
- Inflation and discounting:
  - Constant annual inflation rate = 2%.
  - Uniform real discount rate = 8%.
- Project classification realized price factors:
  - Mixed rare earth chemical products: realized price factor = 62.5%.
  - Separated rare earth oxides: realized price factor = 90%.
- NdPr-equivalent production metric:
  - q_NdPr ≡ ∑_i p_o,i q_o,i / p_NdPr_o
- Incentive price (p*), solving NPV = 0:
  - p* = [ CAPEX + Ω · (OPEX − Rev_nREE) ] / (Ω · q_NdPr), where Ω ≡ ∑_{t=1}^T 1/(1 + r)^t.
- Economic supply curve construction:
  - Projects ranked by break-even (incentive) price and aggregated cumulatively (horizontal axis: cumulative NdPr-equivalent production capacity in tpa; vertical axis: incentive price in US$/kg NdPr).
  - Benchmark 2024 global market Nd price reported in Figure 13 note: US$55/kg.
- Key dataset numeric parameters preserved:
  - Number of non-Chinese projects: 21.
  - Inflation rate: 2%.
  - Discount rate: 8%.
  - Realized price factors: 62.5% (mixed products), 90% (separated oxides).
  - Benchmark cost basis: 2024–25.

*Source: Excerpt from IMF Working Paper — "The Concentrated Economics of Rare Earths" and associated sections in wpiea2026072-source-pdf.*

### 2.1    The Concentrated Economics of Rare Earths

### 2.1    The Concentrated Economics of Rare Earths

### Overview of REEs and Key Market Features
- REEs are a group of 17 chemically similar metals, typically grouped into light (LREEs) and heavy (HREEs) based on atomic weight.
- HREEs are significantly less abundant in the earth’s crust than LREEs and typically occur in lower concentrations, making them more difficult to extract at scale.
- Rare earths are employed in small quantities in various manufacturing sectors including automotive production, wind energy, military applications, semiconductors, and consumer electronics.
- They provide unique properties that are difficult to replicate cheaply, notably in applications requiring strong magnetic performance, thermal stability, or high-power density.

### Market Concentration and Value Distribution
- Four elements—neodymium (Nd) and praseodymium (Pr) (both LREEs), and dysprosium (Dy) and terbium (Tb) (both HREEs)—account for 96 percent of total rare earth oxide (REO) market value despite representing only 23 percent of production by weight.
- Permanent magnets are the fastest-growing application of rare earths and account for roughly 75–85 percent of total REE demand by value.
- Permanent magnets rely predominantly on Nd and Pr, which together account for approximately 85 percent of REE inputs in magnet production, with smaller but performance-critical contributions from Dy and, to a lesser extent, Tb.

### Global Supply Chain Structure and Geographic Concentration by Stage
- The supply chain consists of four sequential stages: ore extraction, benefaction (concentration into mixed rare earth compound), solvent extraction (oxide separation), and metal refining/alloy production (e.g., NdFeB magnets).
- China grew to become the primary source of global rare earth supply in the 2000s due to accumulated technical expertise, a large domestic market, established infrastructure, policy support, and relatively flexible regulations.
- LREE mining diversification:
  - China’s share of global LREE mining peaked at roughly 97 percent in 2010 and declined to around 58 percent in 2024.
  - New production from Australia, the United States, and Myanmar contributed to this shift.
- Downstream processing concentration (global shares, approximate):
  - China accounts for approximately 85–90 percent of global oxide separation capacity.
  - China accounts for 90–95 percent of metal refining.
- For HREEs, concentration is more extreme across stages (reported combining China and Myanmar):
  - 98 percent of mining,
  - 96.5 percent of oxide separation,
  - 95 percent of metal refining,
  - 89.5 percent of permanent magnet manufacturing.
- Key distinction:
  - For LREEs the critical vulnerability lies in processing (separation/refining), not mining.
  - For HREEs both mining and processing exhibit a high degree of geographic concentration.

### Persistence of Concentration: Technology, Costs, and Lead Times
- Diversification in REE supply chains has remained slow despite rising demand, temporary supply shortages, and a wave of industrial policy interventions since 2025.
- Causes of persistence:
  - Technological complexity and tacit knowledge accumulated over decades in processing and separation.
  - High capital intensity and environmentally damaging, energy-intensive separation processes requiring specialized expertise and intellectual property.
  - Regulatory burden and specialized expertise required for processing.
- Project lead times:
  - New REE projects outside China typically require 10–15 years or more to move from discovery to commercial production due to permitting, environmental regulation, financing, and process development.
- Project economics:
  - A bottom-up assessment of 21 REE projects outside China reveals substantial dispersion in capital and operating costs, with many projects on the upper end of the global cost curve.
  - Only a limited number of projects are profitable at current price levels; many require substantially higher prices to become viable.
  - Historical price dynamics (e.g., the 2010–2011 price spike followed by prolonged subdued prices) and price volatility have discouraged large irreversible investments.
- Resource and reserve notes:
  - Global REE reserves outside China exceed 41 million tons of rare earth oxide equivalent, distributed across Brazil, Australia, Russia, Vietnam, United States, and several other jurisdictions.
  - Mine production of LREEs outside China and Myanmar reached 98,000 tons in 2025, accounting for 25 percent of global output.
  - For HREEs, commercially viable deposits are concentrated in southern China and Myanmar; current production and proven reserves in other jurisdictions are insufficient to support comparable self-sufficiency targets.

### Industrial Policy Responses and Market Impacts
- China introduced REE export licensing requirements in April 2025, triggering an immediate market slowdown:
  - Permanent magnet exports fell by roughly 70 percent year-on-year at the trough.
  - Dy and Tb prices in European markets nearly tripled by mid-2025.
- Equity market response:
  - Equity prices of REE producers outside China rallied in July 2025, coinciding with major U.S. industrial policy announcements, suggesting improved market expectations about supply-chain buildup outside China.
- Three policy approaches deployed to de-risk supply chains:
  1. Price floors and offtake agreements to provide investment certainty in volatile markets.
     - Example: Many REE projects are not commercially viable at current neodymium prices (about $55/kg).
     - July 2025 agreement between the U.S. government and MP Materials introduced a price-protection mechanism for key REE products.
     - Price floors featured prominently in high-level G7 critical minerals discussions.
     - In the chapter’s quantitative analysis, a stylized revenue-support instrument for refiners is used as a proxy for downside protection.
  2. Direct financial support (equity stakes, loans, grants) to supply capital and signal long-term commitment.
  3. International agreements and coordinated procurement:
     - Agreements signed in October 2025 between the United States and Japan, Malaysia, Thailand, and Australia, alongside the G7 Critical Minerals Action Plan, mobilized an estimated $6.4 billion in public and private funding to de-risk REE supply chains.
- Observations on manufacturability:
  - Permanent magnet manufacturing presents comparatively lower barriers to capacity expansion; established producers already operate in the United States, Japan, and Europe, though at smaller scale than in China.

### Model Overview: Two-Stage Market Structure
- The market is modeled with two sequential stages: upstream extraction (miners) and downstream refining (refiners).
- Global prices:
  - Upstream price: p^x_t (global market for raw REEs).
  - Downstream price: p^y_t (global market for refined REEs).
- There are J countries/blocks with a representative miner and a representative refiner in each country j ∈ {0, . . . , J}.

### Miners: Dynamics, Costs, and Optimization
- Extraction follows an exponential decline with rate η_j > 0.
- Investment in new extraction capacity is i^x_jt; extraction rate x_jt evolves as:
  - x_jt = (1 − η_j) x_j,t−1 + i^x_jt
- Miners can face subsidies to extraction capacity and investment: sequences {s^x_jt} and {s^ix_jt}.
- Miners take the sequence of upstream REE prices {p^x_t} as given.
- Cost components:
  - Investment (CAPEX): c^ix_j(i^x_jt)
  - Extraction (OPEX): c^x_j(x_jt)
- Miners maximize discounted net revenues with discount factor β_j ∈ (0, 1):
  - max_{x_jt,i^x_jt} Σ_{t=0}^∞ β_j^t [ p^x_t (1 + s^x_jt) x_jt − (1 + s^ix_jt) c^ix_j(i^x_jt) − c^x_j(x_jt) ]
  - subject to x_jt = (1 − η_j) x_j,t−1 + i^x_jt and x_jt ≥ 0
- First-order conditions yield a net marginal CAPEX condition equating net current marginal CAPEX to discounted future marginal CAPEX plus the net extraction margin:
  - (1 + s^ix_jt) ∂c^ix_j/∂i^x_jt = β_j (1 − η_j) (1 + s^ix_jt+1) ∂c^ix_j/∂i^x_jt+1 + p^x_t (1 + s^x_jt) − ∂c^x_j/∂x_jt
- Aggregate mined supply to refiners at time t:
  - R^s_t ≡ Σ_{j=1}^J x_jt

### Refiners: Capacity, Margins, and Optimization
- Refining capacity y_jt follows an exponential decline at rate δ_j > 0 with new investment i^y_jt:
  - y_jt = (1 − δ_j) y_j,t−1 + i^y_jt
- Refining technology is linear in input and satisfies y_jt = z_jt (refiners’ output equals raw REE input).
- Refiners can receive subsidies to refining capacity and investment: sequences {s^y_jt} and {s^iy_jt}.
- Refiners take sequences {p^x_t} and {p^y_t} as given.
- Cost components:
  - Investment (CAPEX): c^iy_j(i^y_jt)
  - Refining (OPEX): c^y_j(y_jt)
- Refiners maximize discounted profits with discount factor β_j ∈ (0, 1):
  - max_{z_jt,y_jt,i^y_jt} Σ_{t=0}^∞ β_j^t [ p^y_t (1 + s^y_jt) y_jt − p^x_t z_jt − (1 + s^iy_jt) c^iy_j(i^y_jt) − c^y_j(y_jt) ]
  - subject to y_jt = z_jt and y_jt = (1 − δ_j) y_j,t−1 + i^y_jt
- After substitution, first-order conditions give a net marginal CAPEX condition:
  - (1 + s^iy_jt) ∂c^iy_j/∂i^y_jt = [ p^y_t (1 + s^y_jt) − p^x_t ] − ∂c^y_j/∂y_jt + β_j (1 − δ_j) (1 + s^iy_jt+1) ∂c^iy_j/∂i^y_jt+1
  - where [ p^y_t (1 + s^y_jt) − p^x_t ] is the gross refining margin.
- Aggregate raw REE demanded in the global market at time t:
  - R^d_t ≡ Σ_{j=1}^J z_jt = Σ_{j=1}^J y_jt

*Source: Excerpt from "2.1 The Concentrated Economics of Rare Earths" in the provided IMF content unit.*

### 3.3    Final Good Demands

### 3.3    Final Good Demands

### Demand specification
- Country j demands the processed/refined REE material according to q_jt = f(a_jt, p_y_t), where a_jt ∈ R_+ is a country-specific and time-specific demand shifter.
- In quantitative analysis, demand functions grow over time to reflect anticipated demand growth for refined REEs.
- Individual country demand for the refined REE product is isoelastic and static:
  - q_jt = a_jt (p_y_t)^−ε
  - ε ∈ R_+ is the elasticity of substitution (common across countries).
- World aggregate demand:
  - Q_t = Σ_{j=1}^J q_jt = A_t (p_y_t)^−ε, where A_t ≡ Σ_{j=1}^J a_jt.

### Competitive equilibrium (definition and market clearing)
- Given initial conditions (x_j0, y_j0), demand shifters {a_jt}, and subsidy rate sequences {s_x_jt, s_y_jt, s_ix_jt, s_iy_jt} for each country j ∈ J, a competitive equilibrium is allocations {x_jt, y_jt, q_jt}_{j∈J,t>0} and prices {p_x_t, p_y_t}_{t>0} such that:
  - (i) allocations solve miners’ problem (2), refiners’ problem (6), and are consistent with the demand specification;
  - (ii) raw REE market clears: Σ_{j∈J} x_jt = Σ_{j∈J} y_jt;
  - (iii) refined REE market clears: Σ_{j∈J} y_jt = Σ_{j∈J} q_jt.

### Functional forms: miners and refiners (marginal costs and steady-state supply)
- Miners’ marginal CAPEX and OPEX:
  - ∂c_ix_j/∂i_x_jt = α_ix_j + γ_ix_j (x_jt − (1−η_j) x_jt−1)
  - ∂c_x_j/∂x_jt = α_x_j
  - Country j’s supply of raw REEs solves a second-order difference equation (equation (9)) with Λ_jt ≡ γ_ix_j(1 + s_ix_jt + β_j(1−η_j)^2(1 + s_ix_jt+1)).
- Refiners’ marginal CAPEX and OPEX:
  - ∂c_iy_j/∂i_y_jt = α_iy_j + γ_iy_j (y_jt − (1−δ_j) y_jt−1)
  - ∂c_y_j/∂y_jt = α_y_j
  - Country j’s demand of raw REEs is given by equation (10) with ζ_jt ≡ γ_iy_j(1 + s_iy_jt + β_j(1−δ_j)^2(1 + s_iy_jt+1)).

### Steady state equilibrium (no industrial policies)
- Industrial policy rates set to zero: s_x_j = s_ix_j = s_y_j = s_iy_j = 0 for all j ∈ J.
- Discount factor and depreciation rates common across countries. Let β ≡ 1/(1+r), where r is the discount rate.
- Steady-state supply and processing outputs:
  - Mining output by producer j:
    - x_j = p_x − α_x_j − β_x α_ix_j B_x_j  (equation (11))
    - β_x ≡ 1 − β(1−η) = (r + η)/(1 + r)
    - B_x_j ≡ γ_ix_j η β_x
  - Processing output:
    - y_j = p_y − p_x − α_y_j − β_y α_iy_j B_y_j  (equation (12))
    - β_y ≡ 1 − β(1−δ) = (r + δ)/(1 + r)
    - B_y_j ≡ γ_iy_j δ β_y
- Aggregate demand in steady state:
  - Q = A (p_y)^ε  (equation (13)), where A ≡ Σ_j a_j.
- Market clearing requires Q = Y and Y = X, where Y ≡ Σ_j y_j and X ≡ Σ_j x_j.
- Equilibrium prices:
  - Mining price as function of processed material price:
    - p_x = (p_y Γ_y − Ψ_y + Θ_x) / (Γ_y + Γ_x)  (equation (14))
  - Processed material price satisfies nonlinear condition:
    - A (p_y)^ε (Γ_y + Γ_x) + Θ_x Γ_y + Ψ_y Γ_x = p_y Γ_y Γ_x  (equation (15))
    - Alternatively: A (p_y)^ε = ((p_y Γ_y − Ψ_y) Γ_x − Θ_x Γ_y) / (Γ_y + Γ_x)  (equation (16))
- Parameter bundles:
  - Γ_y ≡ Σ_j 1/B_y_j = Σ_j 1/(γ_iy_j δ β_y) (aggregate processing supply responsiveness)
  - Ψ_y ≡ Σ_j α_y_j + β_y α_iy_j B_y_j (weighted processing costs)
  - Γ_x ≡ Σ_j 1/B_x_j = Σ_j 1/(γ_ix_j η β_x) (aggregate mining supply responsiveness)
  - Θ_x ≡ Σ_j α_x_j + β_x α_ix_j B_x_j (weighted mining costs)

### Calibration overview
- Calibrated to a 2024 base year using feasibility studies, industry reports, and academic literature.
- Four-step calibration:
  1. Fix demand elasticity and country-level demand shifters using published elasticity estimates and observed consumption shares.
  2. Set cost intercept parameters (α_x_j, α_y_j, α_ix_j, α_iy_j) using a breakeven analysis of rare earth projects.
  3. Decompose investment cost slope parameters into common element κ_is and country-specific element ̄γ_j_is such that γ_is_j = κ_is ̄γ_j_is; calibrate ̄γ to match observed production shares.
  4. Pin down scale parameters (κ_ix, κ_iy) that govern absolute equilibrium price levels.
- Table 1 (in source) summarizes parameter values.

### Demand parameters (calibration specifics)
- Short-run price elasticity set to ε = −0.15, based on the unweighted average for Nd, Pr, Dy, Tb from Shojaeddini et al. (2025).
- Country-level demand shares:
  - China accounts for 50 percent of global refined REE consumption in the base year.
  - Remaining 50 percent allocated across other regions in proportion to their shares of Chinese REE exports in 2019.
- Demand shifters a_j are backed out to rationalize observed 2024 consumption at the equilibrium refined price; standardized demand shifters ̄a_j ≡ a_j / Σ_j a_j express each country’s demand as a share when p_y = 1.
- Demand growth rates from IEA Global Critical Minerals Outlook (2025):
  - annual growth of 4.7 percent during 2025–2029,
  - 1.47 percent during 2030–2034,
  - and zero thereafter.
- These growth rates enter as exogenous shifts in a_j over time, preserving relative country shares.

### Cost intercept parameters (calibration specifics and numeric values)
- Four cost intercepts per region: α_x_j, α_y_j (operating costs), α_ix_j, α_iy_j (investment cost intercepts).
- Breakeven methodology for producers outside China:
  - Evaluated 21 rare earth projects with published feasibility studies.
  - Representative breakeven price outside China (2024 USD per kg NdPr) is $77/kg, of which $43/kg reflects operating costs and $34/kg reflects the annualized capital charge.
- User-cost identity linking CAPEX annuity to investment cost intercept:
  - CAPEX annuity = (r + η)/(1 + r) α_i_j  (equation (17))
  - r = 0.08 and η = 0.035. Under baseline, (r + η)/(1 + r) = (0.08 + 0.035)/1.08 ≈ 0.1065.
  - Inverting gives α_i_j = CAPEX annuity / [(r + η)/(1 + r)].
  - Applied to integrated capital charge $34/kg gives α_i = 34/0.1065 ≈ 318 USD/kg of annual NdPr capacity.
- Mining–refining decomposition (outside China):
  - Allocate 70 percent of operating costs to mining and 30 percent to refining:
    - α_x ≈ 30 USD/kg and α_y ≈ 13 USD/kg (computed as 0.70 × 43 ≈ 30 and 0.30 × 43 ≈ 13 USD/kg).
  - Allocate 40 percent of annualized capital costs to mining and 60 percent to refining:
    - Mining capital annuity ≈ 0.40 × 34 ≈ 13.6 USD/kg ⇒ α_ix = 13.6 / 0.1065 ≈ 127 USD/kg.
    - Refining capital annuity ≈ 0.60 × 34 ≈ 20.4 USD/kg ⇒ α_iy = 20.4 / 0.1065 ≈ 191 USD/kg.
  - Under this decomposition, mining accounts for approximately 56 percent of the integrated breakeven cost outside China, refining 44 percent.
  - Resulting parameter values for all regions outside China:
    - α_x = 30, α_ix = 127, α_y = 13, α_iy = 191 USD/kg.
  - Normalization by α_iy (marginal investment cost of refiners outside China) yields transformed variables denoted with a bar: ̄α_x = 30/191, ̄α_ix = 127/191, ̄α_y = 13/191, ̄α_iy = 191/191.

- Chinese cost parameters:
  - Anchor to historical median NdPr price $42.5/kg (2024 USD), inferred from 2006–2010.
  - Decomposition assumptions:
    - 75 percent of breakeven reflects operating costs and 25 percent capital costs → total operating ≈ $31.9/kg and annualized capital ≈ $10.6/kg (computed as 0.75 × 42.5 ≈ 31.9 and 0.25 × 42.5 ≈ 10.6).
    - Total investment cost intercept from equation (17): α_i ≈ 10.6 / 0.1065 ≈ 99 USD/kg.
    - Allocate 69 percent of China’s total breakeven to mining and 31 percent to refining → stage-level total costs ≈ $29.3/kg for mining and $13.2/kg for refining (computed as 0.69 × 42.5 ≈ 29.3 and 0.31 × 42.5 ≈ 13.2).
    - Allocate 30 percent of annualized capital charge to mining and 70 percent to refining:
      - Mining capital annuity ≈ 0.30 × 10.6 ≈ 3.2 USD/kg ⇒ α_ix = 3.2 / 0.1065 ≈ 30 USD/kg.
      - Refining capital annuity ≈ 0.70 × 10.6 ≈ 7.4 USD/kg ⇒ α_iy = 7.4 / 0.1065 ≈ 69 USD/kg.
    - Residual operating cost intercepts:
      - α_x = 29.3 − 3.2 ≈ 26 USD/kg
      - α_y = 13.2 − 7.4 ≈ 6 USD/kg
    - Normalized Chinese cost intercepts (by 191):
      - ̄α_x = 26/191, ̄α_ix = 30/191, ̄α_y = 6/191, ̄α_iy = 69/191.
  - Notable calibration feature: majority of total cost gap between Chinese producers and those outside China—approximately 58 percent—originates in the refining stage rather than mining.

- Cost uniformity assumption:
  - Identical cost parameters across all regions outside China, motivated by limited evidence of systematic cost ranking across non-Chinese producers in the 21-project sample.

### Investment slope parameters (γ calibration and identification)
- Investment cost slope parameters γ_ix_j and γ_iy_j govern convexity of investment costs and influence equilibrium production scale.
- Decomposition: γ_is_j ≡ κ_s ̄γ_j_is for s = {x, y}, where κ_s > 0 common across countries and ̄γ_j_is > 0 country-specific.
- Calibration targets: match observed 2024 mining and refining production shares; U.S. normalized so ̄γ_ix_US = ̄γ_iy_US = 1.
- Identification outside China (identical cost intercepts across regions):
  - From steady-state mining expression x_j = (p_x − α_x − β_x α_ix)/(γ_ix_j η β_x), relative outputs satisfy:
    - x_j / x_m = γ_ix_m / γ_ix_j  (equation (18))
  - Hence ̄γ_ix_j = x_US / x_j  (equation (19)) for regions outside China; analogous relationship for refining.
- China (different cost intercepts):
  - ̄γ_ix_CHN = (x_US / x_CHN) · (p_x − α_x_CHN − β_x α_ix_CHN) / (p_x − α_x_US − β_x α_ix_US)  (equation (20))
  - Evaluating requires equilibrium mining price p_x; γ_ix_CHN computed jointly with price-level calibration.
- Zero-production regions (identification limits):
  - Regions with zero production in a segment (e.g., EU and Japan no mining; Anglophone region no refining) have γ not identified by equation (18).
  - Assign large values to these γ parameters so that zero-production regions remain inactive at initial price vector; market clearing ensures X = Y.

*Source: IMF working paper chapter section 3.3–4.4 (content unit: wpiea2026072-source-pdf - 3.3    Final Good Demands).*

### 4.5    Price-Level Calibration

### 4.5    Price-Level Calibration

### Price-level calibration procedure
- The relative γ parameters across regions are pinned down by the procedure in Section 4.4, but their absolute scale — and hence the level of equilibrium prices — is undetermined.
- Two scaling parameters, κix and κiy, multiply all mining and refining slope parameters, respectively. Formally, the effective investment cost slope for region j in mining is κix ̄γixj, and analogously κiy ̄γiyj for refining.
- Identification uses two equilibrium price targets:
  - Refined price normalized to py = 1 in the initial steady state (serves as unit of account).
  - Mining-to-refining price ratio targeted at px/py = 0.55, so the mining stage receives 55 percent of integrated revenues in steady state.

### Calibrated scaling parameters and steady-state price outcomes
- Given all other parameters, κix and κiy are jointly determined by the steady-state equilibrium price conditions.
- Calibrated values:
  - κix = 600
  - κiy = 5,000
- Normalizations and steady-state outcomes (as reported):
  - p y SS normalized to 1 (target); reported value in calibration exercise: p y SS = 1.00 (target) and value 0.96 (Table 2 reports target/value pair).
  - p x SS / p y SS target = 0.55; reported value = 0.55 (Table 2).

### Numerical algorithm (Section 4.6)
- Compute deterministic steady state absent industrial policy:
  - Start with initial guess for final good price; solve optimal steady-state mining and refining capacity over grid of intermediate good prices px.
  - For each grid point construct aggregate demand and supply and compute equilibrium price where excess supply of raw REEs is zero.
  - Update final good price and iterate to convergence.
- Solve model dynamics using open-loop equilibrium approach (paths determined at t = 0):
  - Start from initial conditions {xj,−1, yj,−1}Jj=1 and initial guesses for time path of final good price {p0,y t}Tt=0 and state variables {x0 j,t}j,t and {y0 j,t}j,t.
  - For each intermediate good price grid point, solve system of difference equations forward, compute equilibrium price at each date t, update guesses, and iterate until convergence.

### Summary of calibration choices and implications (Section 4.7)
- Key calibration features:
  - Cost intercepts embed a large competitive gap between producers in China and those in other regions, with roughly half of this gap concentrated in the refining stage.
  - China’s low γ values reflect dominant production shares; the investment cost function is relatively flat, allowing China to scale capacity at low marginal adjustment cost.
  - High γ values for regions with zero production imply that entry would require substantial policy intervention to overcome steep investment costs.
  - The model is solved using standardized parameters only (parameters denoted with a bar).
- Matched calibration moments (Table 2, target and reported values):
  - ̄γix CHINA: Target (China’s mining share, steady state) 0.80; Value 0.81.
  - ̄γiy CHINA: Target (China’s refining share, steady state) 0.91; Value 0.92.
  - p y SS: Price of refined product, steady state: Target 1.00; Value 0.96.
  - p x SS / p y SS: Refined-to-mined product price ratio, steady state: Target 0.55; Value 0.55.

### Baseline simulation setup and key numeric inputs
- Simulation horizon: T = 150 years starting in 2025.
- Initial conditions use observed levels of mining and refining quantities in 2024 (Table 1 base-year production, NdPr-equivalent units):
  - Mining output (kt/yr): U.S. 43.0; China+ 285.8; EU 0.0; Japan 0.0; Anglophone 18.0; RoW 6.5.
  - Refining output (kt/yr): U.S. 1.13; China+ 86.72; EU 0.67; Japan 0.67; Anglophone 0.00; RoW 5.42.
- Demand growth: Annual growth 2025–29, 2030–34, ≥ 2035 = 4.7, 1.47, 0 (% per yr) (Table 1).
- Discount and depreciation: r = 0.08; η = 0.035 (Table 1).
- Selected cost parameters (Table 1 excerpts, monetary values in 2024 USD):
  - Mining operating cost intercept αxj ($/kg): U.S. 30; China+ 26; EU 30; Japan 30; Anglophone 30; RoW 30.
  - Refining operating cost intercept αyj ($/kg): U.S. 13; China+ 6; EU 13; Japan 13; Anglophone 13; RoW 13.
  - Investment cost intercepts αixj ($/kg/yr): U.S. 127; China+ 30; EU 127; Japan 127; Anglophone 127; RoW 127.
  - Investment cost intercepts αiyj ($/kg/yr): U.S. 191; China+ 69; EU 191; Japan 191; Anglophone 191; RoW 191.
  - Reported (pre-scaling) investment cost slopes ̄γixj and ̄γiyj (Table 1 entries retained as reported).

### Industry elasticities (short-run and long-run) and interpretation
- Short-run (SR) and long-run (LR) supply and demand elasticities are derived analytically (expressions in text) and reported in Table 3 (exact values):
  - Mining supply elasticity:
    - Net-REE Importers: SR 0.0037; LR 1.79
    - China: SR 0.0031; LR 1.41
  - Mining demand elasticity:
    - Net-REE Importers: SR -0.0044; LR -2.09
    - China: SR -0.0031; LR -1.46
  - Refining supply elasticity:
    - Net-REE Importers: SR 0.0083; LR 3.81
    - China: SR 0.0056; LR 2.66
  - Refining demand elasticity (refined product): SR and LR = -0.15 (constant across time and countries; equals exogenous ε)
- Interpretation:
  - Short-run elasticities of demand and supply are close to zero, indicating limited ability to adjust quantities for both miners and refiners in the short run.
  - Supply and demand are more rigid in China relative to other countries.
  - In the long run, both mining and refining become significantly more flexible, with greater responsiveness for net-REE importers.

### Effects of a potential supply disruption (market fragmentation scenario)
- First-pass impacts of market fragmentation into two blocks (Table 4, percent changes relative to pre-shock quantities=100):
  - Mining:
    - China+’s supply of mined product: -80.88
    - China+’s demand of mined product: -91.67
    - Excess supply in mining markets outside China: +10.78
  - Refining:
    - China+’s supply of refined product: -91.67
    - China+’s demand of refined product: -50.00
    - Excess supply in refining markets outside China: -41.67
- Dynamic model results (key numeric outcomes reported in text and Figure 5 description):
  - On impact, prices spike:
    - Refined REE price spikes to almost 80 times the baseline price.
    - Mined REE price increases to around 7.4 times the baseline price.
  - Quantity and investment responses on impact:
    - Consumption of refined products falls by around 48% on impact.
    - Investment in refining capacity increases, leading to more than threefold increase in supply of refined products.
    - Demand for mined inputs increases to about 1.3 times baseline on impact.
  - Long-run adjustments:
    - Mining output rises to about 2.1 times its pre-shock level.
    - Refining output rises to about 4.5 times its pre-shock level.
    - Upstream equilibrium price falls from 7.4 to 1.7 times baseline, remaining about 70% above baseline.
    - Downstream refined-product price declines from 80 to 2.5 times baseline, i.e., about 150% above baseline.
    - Final consumers recover part of lost demand and converge to around 89% of baseline level, implying a long-run loss of about 11% of baseline demand.
  - Interpretation:
    - Short-run low elasticities cause large short-run price spikes and large initial demand declines.
    - Long-run CAPEX-driven expansion in both upstream and downstream capacity reduces prices and partially restores consumption, consistent with much higher long-run elasticities.

*Source: Excerpt from IMF working paper section 4.5–5.2 (calibration, numerical algorithm, summary, baseline results, elasticities, and supply disruption scenario).*

### 5.3    Industrial Policies

### 5.3    Industrial Policies

### Context and objectives
- Rising vulnerabilities in REE supply have prompted adaptation strategies: stockpiling, recycling, reshoring, and import diversification. Stockpiling is a short-term buffer; recycling has longer-term potential but is not yet a primary supply source. Reshoring and import diversification face long development time-lines, coordination challenges, and potential shortages of skilled labor.
- Optimal de-risking is framed as an insurance problem: the premium (efficiency losses in normal times) should be commensurate with expected disruption losses in crisis times. Security benefits show diminishing returns; the first 10–20 percent increase in de-risking delivers the largest reduction in vulnerability.
- De-risking is defined as the total consumption of refined REEs that is either produced at home or from other countries other than the market leader. The analysis focuses on refiners because a global supply shock primarily affects downstream segments.

### Policy instruments, scenarios, and targets
- Instruments modeled:
  - CAPEX subsidies to refiners, with s^iy_jt ∈ [−1, 0].
  - Stylized revenue-support (implemented as a proportional sales subsidy), defined s^y_jt ≡ f^y_j − 1, with f^y_j ≥ 1.
- Implementation scenarios:
  1. Unilateral U.S. action: policy instruments apply only to U.S.-based producers.
  2. Simultaneous action: U.S., EU, Japan, Australia, and other economies apply instruments uniformly.
- Targets assessed: 25% de-risking by 2035 and 50% de-risking by 2035.
- Parameter values used (Table 5 reproduced in content):
  - CAPEX Subsidy, unilateral (j = U.S.): −74.41% (25% target), −88.72% (50% target)
  - CAPEX Subsidy, simultaneous (j̸= China): −73.97% (25% target), −87.91% (50% target)
  - Price Floor, unilateral (j = U.S.): 2.19 (25% target), 4.08 (50% target)
  - Price Floor, simultaneous (j̸= China): 2.00 (25% target), 3.05 (50% target)

### De-risking metric and computation
- For each country j and period t, construct matrix of net import flows M_t from net exporter j to importer i.
- Exporter share:
  xs_{j,t} = max(y_{j,t} − q_{j,t}, 0) / ∑_i max(y_{i,t} − q_{i,t}, 0)
- Allocate net imports proportionally:
  m_{ij,t} = xs_{j,t} · max(q_{i,t} − y_{i,t}, 0)
- Degree of de-risking for country i at time t:
  σ_{i,t} = 1 − max_{j∈J} m_{ij,t} / q_{i,t}
- Interpretation: σ_{i,t} close to 1 indicates high de-risking; values close to 0 indicate strong dependence on a single supplier. The analysis restricts de-risking attention to China only.

### Key simulation findings: achievement of de-risking targets
- By construction, all policies achieve their target exactly in 2035.
- For both 25% and 50% targets:
  - Unilateral policies deliver higher pre-2035 de-risking for the U.S. than simultaneous action.
  - In the long run, simultaneous policies outperform unilateral ones.
  - Conditional on simultaneity, CAPEX subsidies and revenue-support policies are broadly equivalent in delivering desired de-risking over the first 13–15 years.

### Fiscal cost outcomes and dynamics
- Fiscal cost expression (relative to baseline consumption) used:
  - If CAPEX subsidy:
    Φ_{jt} = s^iy_{jt} c^iy_j(i^y_{jt}) p^y_{t,BAS.} q_{jt,BAS.}
  - If revenue support:
    Φ_{jt} = p^y_t(f^y_j − 1) y_{jt} / (p^y_{t,BAS.} q_{jt,BAS.})
- Three main fiscal-cost results:
  (i) Simultaneous action reduces the fiscal cost to achieve a given level of de-risking for the U.S. and all participating countries, because the U.S. can rely on imports from other participating countries and global REE output rises more under simultaneous action.
  (ii) CAPEX subsidies frontload costs (investment precedes production) and are initially more expensive than revenue support, but are substantially less costly in the long run. Revenue-support (price-floor) costs backload and remain elevated as the economy converges to a new steady state.
  (iii) Fiscal cost increases faster than achieved de-risking. Example (unilateral revenue support reference):
    - Cost of achieving 25% de-risking by 2035: about 13.4% of baseline consumption in 2025.
    - Cost of achieving 50% de-risking by 2035: exceeds 44.5% of baseline consumption in 2025.
    - Long-run stabilization of unilateral revenue-support policy fiscal cost: around 41.6% of baseline consumption for a 25% target, but rises to over 269% for the 50% target.
    - Doubling the de-risking target more than doubles the fiscal cost.

- Aggregate fiscal-cost magnitudes (total fiscal cost over 2025–2035 for U.S. under 25% target):
  - CAPEX subsidies: 134% of the annual U.S. REE market under unilateral action and 90% under simultaneous action.
  - Stylized revenue-support policies: 248% under unilateral policies and 172% under simultaneous policies.
- Fiscal costs are even higher when targeting 50% de-risking.

### Spillovers: impacts on mining, refining, consumption, and prices
- Mining:
  - 25% target, unilateral: mining output increases by +0.26% in China and +0.3% in net importing countries.
  - 50% target, unilateral: China +0.66%; net importing countries +0.79%.
  - Simultaneous action yields larger increases: China +1.44% (25% target) and +2.8% (50% target); net importing countries +1.72% (25%) and +3.3% (50%).
- Refining:
  - China: refining output declines under policies targeting refiners:
    - 25% target: −2.5% (unilateral) and −14.8% (simultaneous).
    - 50% target: −6.8% (unilateral) and −27.7% (simultaneous).
  - Rest of world: refining output increases substantially:
    - 25% target: +25.9% (unilateral) and +154.7% (simultaneous).
    - 50% target: +70.8% (unilateral) and +290% (simultaneous).
- Consumption:
  - Relatively small effects on consumption; global consumption increases:
    - 25% target: +0.27% (unilateral) and +1.5% (simultaneous).
    - 50% target: +0.69% (unilateral) and +2.91% (simultaneous).
- Prices:
  - Under unilateral subsidies, mining prices initially rise by +1.48% (25% target) and +3.82% (50% target), settling at +0.17% and +0.49% in the long run respectively.
  - Under simultaneous subsidies, mining prices rise on impact by +8.66% (25% target) and +17.66% (50% target), converging to +1.13% and +2.22% in the long run respectively.
  - Refining prices decline (reflecting changes in mining and refining dynamics).

### Distributional effects across importers
- Unilateral U.S. policies increase U.S. de-risking but leave other net importers broadly unchanged or slightly worse off.
- Simultaneous policies allow the U.S. and other importing countries (EU, Japan) to increase their degree of de-risking:
  - Baseline de-risking: EU 6%, Japan 8.9% (values reported in text).
  - Under a 25% U.S. target achieved via simultaneous action: EU reaches 17.1% and Japan 23.1%.
- Fiscal-cost distribution under simultaneous policies (25% target):
  - Cost of CAPEX subsidies is nearly as high in Japan as in the U.S., and significantly lower in Europe. Fiscal costs are higher in countries with larger refining sectors.

### Nonlinear cost dynamics and elasticity
- Achieving higher de-risking levels requires disproportionately higher subsidy rates and fiscal costs:
  - Subsidy rates below 60% allow at most about 18% de-risking for the U.S.
  - Increasing subsidy from 80% to 90% raises de-risking by over 20 percentage points.
  - Moving from 85% to 95% raises de-risking by approximately 45 percentage points.
- Fiscal-cost elasticity with respect to de-risking is above 1 for all subsidy rates considered: fiscal cost rises more than proportionally with attained de-risking rates over the displayed range.

### Policy implications and comparative assessment
- De-risking supply chains via targeted industrial policies is fiscally costly, but costs are:
  - Lower when de-risking is pursued simultaneously by various importers.
  - Lower when policy instruments directly target expansion of new production capacity (CAPEX subsidies perform better in the long run).
- Revenue-support (price-floor) mechanisms are less costly initially but become more expensive over time as production and prices rise.
- Ambitious de-risking targets (e.g., 50%) become increasingly costly, with fiscal costs rising exponentially as targets increase.

### Suggested extensions for future research
- Develop a more granular model distinguishing among the 17 rare earth elements, differentiating heavy (HREE) and light (LREE) elements, and embedding these distinctions in supply-constraint modeling.
- Analyze additional policies such as stockpiling or reshoring.
- Enlarge the set of producer countries beyond China for richer counterfactuals.

*Source: IMF Working Paper — Section 5.3 Industrial Policies (chapter content provided).*

### References

### wpiea2026072-source-pdf - References

### Dataset scope and project coverage
- Dataset constructed from a mine-by-mine assessment of rare earth projects located in: Australia, Sub-Saharan Africa, Brazil, Canada, Greenland, Sweden, and the United States.
- Projects included only where sufficient public documentation is available (technical reports, economic assessments, prefeasibility or feasibility studies).
- Primary data sources: S&P Capital IQ Pro database and supporting project disclosures, occasionally supplemented with more recent publicly available information.
- Resulting dataset covers 21 non-Chinese projects with sufficient cost, production, and technical information to reconstruct project-level economic metrics.

### Manual reconstruction of production, costs, and revenues
- For each project, extracted: total rare earth oxide (TREO) production volumes, reported oxide distributions, and annual production quantities for individual rare earth oxides.
- Non-REE by-product revenues:
  - When detailed production data available: annual volumes × corresponding commodity prices to estimate revenue streams.
  - When only nominal annual revenue figures reported: incorporated directly into revenue calculations.
  - Incentive price used for the supply curve excludes non-REE revenues so break-even is determined solely by rare earth project economics.
- Capital expenditure (CAPEX):
  - Initial CAPEX taken directly from project documentation.
  - Where reported, sustaining capital expenditures included.
- Operating expenditure (OPEX):
  - Approach 1: where total life-of-mine operating costs reported → average annual OPEX = total life-of-mine operating costs ÷ expected mine lifespan.
  - Approach 2: where studies directly reported average annual operating expense → used as stated.
  - Approach 3: where operating costs reported as unit cost (US$ per kilogram of TREO) → annual OPEX = reported unit cost × average annual TREO production.

### Inflation adjustment, discounting, and uniform parameters
- Monetary values converted to a common 2024–25 cost basis using a constant annual inflation rate of 2%.
- Uniform real discount rate of 8% applied across all projects.
- Adjustments applied consistently to CAPEX, OPEX, and revenue components for comparability across projects with feasibility studies from different years.

### Project classification and realized price factors
- Projects grouped into two categories:
  - Mixed rare earth chemical products: realized price factor of 62.5%.
  - Separated rare earth oxides: realized price factor of 90%.
- Purpose: capture difference in realized value between partially processed intermediate products and more refined separated oxides and ensure processing-depth differences are reflected in revenues.

### NdPr-equivalent production metric and incentive price definition
- NdPr-equivalent production metric definition (as used to normalize multi-element revenue streams into NdPr terms):
  - q_NdPr ≡ ∑_i p_o,i q_o,i / p_NdPr_o
  - This normalization converts a project’s multi-element revenue stream into a single NdPr-equivalent production metric.
- Net present value representation:
  - NPV(p) = ∑_{t=1}^T [ p q_NdPr + Rev_nREE − OPEX ] / (1 + r)^t − CAPEX
  - Where: OPEX = total cash operating costs; CAPEX = initial capital expenditures; Rev_nREE = annual revenues from non-REE minerals; r = discount rate; T = project life. Prices and quantities are assumed constant over time.
- Incentive price solving NPV = 0:
  - p* = [ CAPEX + Ω · (OPEX − Rev_nREE) ] / (Ω · q_NdPr)
  - Where Ω ≡ ∑_{t=1}^T 1/(1 + r)^t is the annuity factor.
  - Interpretation: p* is the minimum NdPr oxide price at which projected discounted cash flows recover initial capital outlay, making the project economically viable.
- Implicit assumption: relative price ratios across elements remain fixed at their 2024 benchmark levels so a change in the NdPr price is representative of a proportional change across the rare earth price complex.

### Construction of the economic supply curve
- Projects ranked in ascending order of break-even (incentive) price and aggregated cumulatively according to their production capacity.
- Horizontal axis of Figure 13: cumulative NdPr-equivalent production capacity in tons per annum (tpa), constructed by converting total project revenues into an equivalent quantity expressed in NdPr terms.
- Vertical axis of Figure 13: corresponding incentive price in US$/kg of NdPr oxide.
- Both quantities and prices expressed in NdPr terms.
- Output: stepwise economic supply curve preserving project-level resolution and illustrating cost distribution of potential non-Chinese rare earth supply.
- Project labels identify regions: AFR (Africa), LA (Latin America), NA (North America), OC (Oceania), RoW (rest of the world).
- NdPr-equivalent production capacity identity used:
  - ∑_i p_i q_i = P_M2 · Q_M2  ⇒  Q_M2 = ∑_i p_i q_i / P_M2
  - Where ∑_i p_i q_i denotes total project revenues across all reported rare earth elements and P_M2 is the NdPr oxide price.

### Key numeric parameters and reference values preserved from the analysis
- Number of non-Chinese projects with sufficient data: 21.
- Constant annual inflation rate used for adjustments: 2%.
- Uniform real discount rate applied across all projects: 8%.
- Realized price factor for mixed rare earth chemical products: 62.5%.
- Realized price factor for separated rare earth oxides: 90%.
- Benchmark year cost basis: 2024–25.
- 2024 global market price of Nd reported in Figure 13 note: US$55/kg.

### Figures and data sources cited
- Figure 13 sources: LSEG Refinitiv Workspace (2024); S&P Global Market Intelligence (2026b); and IMF staff calculations.
- Other referenced datasets and reports cited in the References section include (selection preserved as in source):
  - LSEG Refinitiv Workspace (2024): Rare earth oxide prices for specified oxides and grade ranges.
  - S&P Global Market Intelligence (2026b): Lanthanides development studies dataset; project-level data on lanthanide mining and processing.
  - U.S. Geological Survey (2025) and (2026): Mineral commodity summaries 2025; Rare earths statistics and information.
  - International Energy Agency (2025): Global critical minerals outlook 2025.
  - Additional academic and policy working papers listed in the References.

*Industrial Policy and Supply Chain Bottlenecks in Critical Minerals: Evidence from Rare Earth Elements — Working Paper No. WP/2026/072*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026072-source-pdf.pdf_
