Drivers of the World Grain Price Crisis in the Short- and Long-Run: A Spatial-Temporal Rational Expectations Equilibrium Approach Randall Romero-Aguilar Mario J. Miranda Seminar at Center for Development Research University of Bonn July 8-9, 2014 Outline 1 Introduction 2 The Model The model equations Solving the model 3 Results Long-term results Short-term results 4 Conclusions Introduction Introduction We build a model to examine some proposed drivers of the 2007-2009 World Food Price Crisis: low grain stock levels; trade restrictions by wheat exporters; public storage by wheat importers; and diversion of corn production to biofuels. Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 1 / 18 Outline 1 Introduction 2 The Model The model equations Solving the model 3 Results Long-term results Short-term results 4 Conclusions The Model The model equations The Model Two commodities: wheat and corn One fixed input: land International trade Storage World markets x Corn Exporters 1−λ w Corn Corn Importers Romero-Aguilar, Miranda x́ λ ẃ Wheat ywm m Wheat Exporters wheat land corn land yẃḿ Storage Storage Drivers of the World Grain Price Crisis Wheat Importers ḿ 2 / 18 The Model The model equations Supply, demand, production, consumption Supply and demand: Q̃i + Zi,−1 ≡ Ai production storage availability = Ci + Zi + Yi consumpt. storage net exports Production Q̃i = qi,−1 · ˜i acreage yield Consumption demand Ci = αi Pi −βi price Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 3 / 18 The Model The model equations Supply, demand, production, consumption Supply and demand: Q̃i + Zi,−1 ≡ Ai production storage availability = Ci + Zi + Yi consumpt. storage net exports Production Q̃i = qi,−1 · ˜i acreage yield Consumption demand Ci = αi Pi −βi price Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 3 / 18 The Model The model equations Supply, demand, production, consumption Supply and demand: Q̃i + Zi,−1 ≡ Ai production storage availability = Ci + Zi + Yi consumpt. storage net exports Production Q̃i = qi,−1 · ˜i acreage yield Consumption demand Ci = αi Pi −βi price Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 3 / 18 The Model The model equations Supply, demand, production, consumption Supply and demand: Q̃i + Zi,−1 ≡ Ai production storage availability = Ci + Zi + Yi consumpt. storage net exports Production Q̃i = qi,−1 · ˜i acreage yield Consumption demand Ci = αi Pi −βi price Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 3 / 18 The Model The model equations Supply, demand, production, consumption Supply and demand: Q̃i + Zi,−1 ≡ Ai production storage availability = Ci + Zi + Yi consumpt. storage net exports Production Q̃i = qi,−1 · ˜i acreage yield Consumption demand Ci = αi Pi −βi price Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 3 / 18 The Model The model equations Trade and private storage Trade: 0≤ yjk ≤ ȳjk export j to k capacity ⊥ Pk − τjk − Pj transport cost unrestricted min[Pk − τjk , P̄j ] − Pj price ceiling Private storage 0 ≤ Zi ≤ Z̄i capacity Romero-Aguilar, Miranda ⊥ δ E Pi0 − Pi − K expected price Drivers of the World Grain Price Crisis storage cost 4 / 18 The Model The model equations Trade and private storage Trade: 0≤ yjk ≤ ȳjk export j to k capacity ⊥ Pk − τjk − Pj transport cost unrestricted min[Pk − τjk , P̄j ] − Pj price ceiling restricted Private storage 0 ≤ Zi ≤ Z̄i capacity Romero-Aguilar, Miranda ⊥ δ E Pi0 − Pi − K expected price Drivers of the World Grain Price Crisis storage cost 4 / 18 The Model The model equations Trade and private storage Trade: 0≤ yjk ≤ ȳjk export j to k capacity ⊥ Pk − τjk − Pj transport cost unrestricted min[Pk − τjk , P̄j ] − Pj price ceiling restricted Private storage 0 ≤ Zi ≤ Z̄i capacity Romero-Aguilar, Miranda ⊥ δ E Pi0 − Pi − K expected price Drivers of the World Grain Price Crisis storage cost 4 / 18 The Model The model equations Public storage Public storage 0 ≤ Zi ≤ Z̄i ⊥ P̄i − interv. price Pi Zḿ Z̄ḿ alternative: Zi = Z̄i 1 + exp [−s P̄i ] 1 + exp [−s(P̄i − Pi )] 0 ≤ Zḿ ≤ Z̄ḿ ⊥ P̄ḿ − Pḿ Zḿ = Z̄ḿ 1+exp −sP̄ḿ 1+exp −s(P̄ḿ −Pḿ ) P̄ḿ Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis Pḿ 5 / 18 The Model The model equations Public storage Public storage 0 ≤ Zi ≤ Z̄i ⊥ P̄i − interv. price Pi Zḿ Z̄ḿ alternative: Zi = Z̄i 1 + exp [−s P̄i ] 1 + exp [−s(P̄i − Pi )] 0 ≤ Zḿ ≤ Z̄ḿ ⊥ P̄ḿ − Pḿ Zḿ = Z̄ḿ 1+exp −sP̄ḿ 1+exp −s(P̄ḿ −Pḿ ) P̄ḿ Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis Pḿ 5 / 18 The Model The model equations Acreage Share of land cultivated with wheat λmin ≤ ≤ λmax λ E Pẃ0 − ϕ E Pw0 ⊥ share wheat wheat corn Acreage qw = (1 − λ) L Expected corn production Production possibilities frontier land available λmin λmax qẃ = λL Expected wheat production Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 6 / 18 The Model The model equations Acreage Share of land cultivated with wheat λmin ≤ ≤ λmax λ E Pẃ0 − ϕ E Pw0 ⊥ share wheat wheat corn Acreage qw = (1 − λ) L Expected corn production Production possibilities frontier land available λmin λmax qẃ = λL Expected wheat production Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 6 / 18 The Model Solving the model Numerical solution strategy Dynamic model, with state variable A. Given Ai , model can be reduced to a mixed-complementarity problem with unknowns P, Z, λ, y ; 17 or 18 variables. P E Pi0 is unknown ⇒ collocation methods: pi ≈ p̂i = h ch φh (A) Inner loop iteration to solve MCP, using A0 and numerical integration to evaluate E Pi0 ≈E H X h=1 ch φh (A0 ) ≈ X ωj X j ch φh (Z + q0j ) h Outer loop iteration to solve for collocation coeffs ch Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 7 / 18 The Model Solving the model Numerical solution strategy Dynamic model, with state variable A. Given Ai , model can be reduced to a mixed-complementarity problem with unknowns P, Z, λ, y ; 17 or 18 variables. P E Pi0 is unknown ⇒ collocation methods: pi ≈ p̂i = h ch φh (A) Inner loop iteration to solve MCP, using A0 and numerical integration to evaluate E Pi0 ≈E H X h=1 ch φh (A0 ) ≈ X ωj X j ch φh (Z + q0j ) h Outer loop iteration to solve for collocation coeffs ch Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 7 / 18 The Model Solving the model Numerical solution strategy Dynamic model, with state variable A. Given Ai , model can be reduced to a mixed-complementarity problem with unknowns P, Z, λ, y ; 17 or 18 variables. P E Pi0 is unknown ⇒ collocation methods: pi ≈ p̂i = h ch φh (A) Inner loop iteration to solve MCP, using A0 and numerical integration to evaluate E Pi0 ≈E H X h=1 ch φh (A0 ) ≈ X ωj X j ch φh (Z + q0j ) h Outer loop iteration to solve for collocation coeffs ch Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 7 / 18 The Model Solving the model Numerical solution strategy Dynamic model, with state variable A. Given Ai , model can be reduced to a mixed-complementarity problem with unknowns P, Z, λ, y ; 17 or 18 variables. P E Pi0 is unknown ⇒ collocation methods: pi ≈ p̂i = h ch φh (A) Inner loop iteration to solve MCP, using A0 and numerical integration to evaluate E Pi0 ≈E H X h=1 ch φh (A0 ) ≈ X ωj X j ch φh (Z + q0j ) h Outer loop iteration to solve for collocation coeffs ch Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 7 / 18 The Model Solving the model Numerical solution strategy Dynamic model, with state variable A. Given Ai , model can be reduced to a mixed-complementarity problem with unknowns P, Z, λ, y ; 17 or 18 variables. P E Pi0 is unknown ⇒ collocation methods: pi ≈ p̂i = h ch φh (A) Inner loop iteration to solve MCP, using A0 and numerical integration to evaluate E Pi0 ≈E H X h=1 ch φh (A0 ) ≈ X ωj X j ch φh (Z + q0j ) h Outer loop iteration to solve for collocation coeffs ch Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 7 / 18 The Model Solving the model Dealing with the “curse of dimensionality” Collocation with Chebyshev polynomials, 9 nodes per dimension. A has 6 dimensions ⇒ 96 = 531 441 nodes and basis functions if using tensor product. 1 2 1 q √ 2+ 2 1 √2 2 Included 1 2 q √ 2− 2 Use Smolyak's method to choose nodes and bases. -1 q √ 2− 2 2 Drivers of the World Grain Price Crisis 1 √ 2 2 + √ 2 2 1 q 2 − 1 2 √ 2 − 1 q 2 2 − √ 2 − 1 q 2 2 + √ 2 Romero-Aguilar, Miranda -1 − √ 2 √ 2 −1 2 q √ −1 2 + 2 2 1 q −1 2 1 2 Result: only 1409 nodes and 389 polynomials. Not included 8 / 18 The Model Solving the model Data sources and regions Most parameters calibrated with historical data from PSD database (USDA). PSD region North America Former Soviet Union Oceania South America East Asia Southeast Asia Middle East North Africa Sub-Saharan Africa European Union Romero-Aguilar, Miranda World Corn Exporter Importer X World Wheat Exporter Importer X X X X X X X X X X Drivers of the World Grain Price Crisis X X X X X X 9 / 18 Outline 1 Introduction 2 The Model The model equations Solving the model 3 Results Long-term results Short-term results 4 Conclusions Results The Policy Scenarios We consider these scenarios 0: baseline 1: wheat exporters set price ceiling at baseline average 2: wheat importers set public storage, price around baseline average 3: wheat exporters and importers apply policies simultaneously 4: ethanol production increases corn demand by 20% Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 10 / 18 Results Long-term results Long-term prices: mean and standard deviation Scenario World Corn Exporter Importer World Wheat Exporter Importer Mean 0: Baseline 1: Price ceiling 2: Public storage 3: Ceiling + storage 4: High demand 100.00 101.68 101.63 103.22 118.18 100.00 101.68 101.63 103.22 118.18 108.91 110.50 110.41 111.87 126.22 100.00 102.45 98.28 99.17 115.19 100.05 93.80 98.34 94.69 115.23 112.22 114.66 110.51 111.43 127.41 Normalized prices: World baseline mean=100 Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 11 / 18 Results Long-term results Long-term prices: mean and standard deviation Scenario World Corn Exporter Importer World Wheat Exporter Importer Mean 0: Baseline 1: Price ceiling 2: Public storage 3: Ceiling + storage 4: High demand 100.00 101.68 101.63 103.22 118.18 100.00 101.68 101.63 103.22 118.18 108.91 110.50 110.41 111.87 126.22 100.00 102.45 98.28 99.17 115.19 100.05 93.80 98.34 94.69 115.23 112.22 114.66 110.51 111.43 127.41 Standard Deviation 0: Baseline 1: Price ceiling 2: Public storage 3: Ceiling + storage 4: High demand 14.72 16.07 17.23 19.26 21.31 14.72 16.07 17.23 19.26 21.31 13.98 14.93 16.11 17.50 16.88 18.02 24.54 11.45 14.64 18.09 18.08 6.92 11.54 4.73 18.13 18.02 24.54 11.44 14.53 18.09 Normalized prices: World baseline mean=100 Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 11 / 18 Results Long-term results Long-term prices: mean and standard deviation Scenario World Corn Exporter Importer World Wheat Exporter Importer Mean 0: Baseline 1: Price ceiling 2: Public storage 3: Ceiling + storage 4: High demand 100.00 101.68 101.63 103.22 118.18 100.00 101.68 101.63 103.22 118.18 108.91 110.50 110.41 111.87 126.22 100.00 102.45 98.28 99.17 115.19 100.05 93.80 98.34 94.69 115.23 112.22 114.66 110.51 111.43 127.41 Standard Deviation 0: Baseline 1: Price ceiling 2: Public storage 3: Ceiling + storage 4: High demand 14.72 16.07 17.23 19.26 21.31 14.72 16.07 17.23 19.26 21.31 13.98 14.93 16.11 17.50 16.88 18.02 24.54 11.45 14.64 18.09 18.08 6.92 11.54 4.73 18.13 18.02 24.54 11.44 14.53 18.09 Normalized prices: World baseline mean=100 Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 11 / 18 Results Long-term results Long-term World prices: conditional on initial stock World Price (baseline average = 100) Corn Wheat 200 100 5 15 25 35 45 55 65 75 85 5 15 25 35 45 55 Global Initial Storage (units of grain) Baseline scenario Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 12 / 18 Results Long-term results Probability of crisis: conditional on initial stock Corn Wheat Probality of Price > 150 (percent) 12.5 10.0 7.5 5.0 2.5 0.0 0 25 50 75 100 0 25 50 75 100 Global Initial Storage (units of grain) 0: Baseline Romero-Aguilar, Miranda 1: Price ceiling 2: Public storage Drivers of the World Grain Price Crisis 3: Ceiling + storage 4: High demand 13 / 18 Results Short-term results Short-term prices: Impulse response function We next consider the short-term adjustment to 3 different shocks 1 change of policy regime; 2 change of regime while 20% drop in Exporter wheat production; and 3 change of regime, production shock, when initial wheat stocks are low. Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 14 / 18 Results Short-term results 1: Regimen Change 2: [1]+Low Yields 3: [2]+Low Stocks 90 Corn 60 30 0 90 Wheat Deviation with respect to baseline (percent) Short-term adjustment 60 30 0 0 2 4 6 0 2 4 6 0 2 4 6 Periods after shock 0: Baseline Romero-Aguilar, Miranda 1: Price ceiling 2: Public storage 3: Ceiling + storage Drivers of the World Grain Price Crisis 4: High demand 15 / 18 Results Short-term results 1: Regimen Change 2: [1]+Low Yields 3: [2]+Low Stocks 90 Corn 60 30 0 90 Wheat Deviation with respect to baseline (percent) Short-term adjustment 60 30 0 0 2 4 6 0 2 4 6 0 2 4 6 Periods after shock 0: Baseline Romero-Aguilar, Miranda 1: Price ceiling 2: Public storage 3: Ceiling + storage Drivers of the World Grain Price Crisis 4: High demand 15 / 18 Results Short-term results 1: Regimen Change 2: [1]+Low Yields 3: [2]+Low Stocks 90 Corn 60 30 0 90 Wheat Deviation with respect to baseline (percent) Short-term adjustment 60 30 0 0 2 4 6 0 2 4 6 0 2 4 6 Periods after shock 0: Baseline Romero-Aguilar, Miranda 1: Price ceiling 2: Public storage 3: Ceiling + storage Drivers of the World Grain Price Crisis 4: High demand 15 / 18 Outline 1 Introduction 2 The Model The model equations Solving the model 3 Results Long-term results Short-term results 4 Conclusions Conclusions Conclusions: Long-term prices Ethanol mandate: only policy with large effect on long-term price (both grains) greatly increases price volatility of both grains, in all regions Wheat Exporter price ceiling increases wheat price volatility in other regions Public storage in Wheat Importer, despite displacing private storage, reduces wheat price volatility in all regions Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 16 / 18 Conclusions Conclusions: Short-term prices By itself, introducing price ceiling or public storage have small impact in short-term prices Initial grain storage is a key determinant of likelihood of crisis An export price ceiling can worsen a crisis originated in production shock & low stocks Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 17 / 18 Conclusions Some limitations of the model = Opportunities to improve No disincentive to farmers from price ceiling Stationary model: no productivity growth, no population growth No “panic” purchases from importers Ethanol shock as one-time permanent demand increase on corn demand Romero-Aguilar, Miranda Drivers of the World Grain Price Crisis 18 / 18