Identifying Supply and Demand Elasticities of Agricultural Commodities: Implications for the US Ethanol Mandate

Transcription

1 Identifying Supply and Demand Elasticities of Agricultural Commodities: Implications for the US Ethanol Mandate Michael Roberts 1 Wolfram Schlenker 2 1 North Carolina State University 2 Columbia University and NBER Michigan State University - April 26, 2010

2 Outline 1 Motivation 2 Methodology 3 Data 4 Empirical Results 5 Conclusions

3 Outline 1 Motivation 2 Methodology 3 Data 4 Empirical Results 5 Conclusions

4 Setting the Stage Background - World Food Prices Recent almost fourfold increase in price of maize Between Summer 2006 and Summer 2008 Prices for wheat, soybeans, and rice increased as well Food is basic commodity Rising hunger and malnutrition New York Times: consumers in developing countries, not U.S. cut back on food consumption Potential for increased conflict Miguel et el.(2004): weather induced income shocks lead to civil conflict in Africa Most African countries are net food importers Increase in price implies reduction in real income

5 Setting the Stage Background - World Food Prices Recent almost fourfold increase in price of maize Between Summer 2006 and Summer 2008 Prices for wheat, soybeans, and rice increased as well Food is basic commodity Rising hunger and malnutrition New York Times: consumers in developing countries, not U.S. cut back on food consumption Potential for increased conflict Miguel et el.(2004): weather induced income shocks lead to civil conflict in Africa Most African countries are net food importers Increase in price implies reduction in real income

6 Setting the Stage Background - World Food Prices Recent almost fourfold increase in price of maize Between Summer 2006 and Summer 2008 Prices for wheat, soybeans, and rice increased as well Food is basic commodity Rising hunger and malnutrition New York Times: consumers in developing countries, not U.S. cut back on food consumption Potential for increased conflict Miguel et el.(2004): weather induced income shocks lead to civil conflict in Africa Most African countries are net food importers Increase in price implies reduction in real income

7 Setting the Stage Goal of this Paper Estimate supply / demand for calories Calories: aggregate maize, wheat, rice, and soybeans Traditional Approach Regress supply on expected price (futures price) Our concern: Futures price is endogenous Example: Soybean Rust (Pest) Farmers anticipate soybeans rust (error term u) Response: decrease planting area Futures market: increase in futures price Lower quantity and higher price? Movement along demand curve Will bias supply elasticity towards zero

8 Setting the Stage Goal of this Paper

9 Setting the Stage Goal of this Paper

10 Setting the Stage Goal of this Paper

11 Setting the Stage Goal of this Paper

12 Setting the Stage Goal of this Paper

13 Setting the Stage Goal of this Paper Estimate supply / demand for calories Calories: aggregate maize, wheat, rice, and soybeans Traditional Approach Regress supply on expected price (futures price) Our concern: Futures price is endogenous Example: Soybean Rust (Pest) Farmers anticipate soybeans rust (error term u) Response: decrease planting area Futures market: increase in futures price Lower quantity and higher price? Movement along demand curve Will bias supply elasticity towards zero

14 Setting the Stage Goal of this Paper Estimate supply / demand for calories Calories: aggregate maize, wheat, rice, and soybeans Traditional Approach Regress supply on expected price (futures price) Our concern: Futures price is endogenous Example: Soybean Rust (Pest) Farmers anticipate soybeans rust (error term u) Response: decrease planting area Futures market: increase in futures price Lower quantity and higher price? Movement along demand curve Will bias supply elasticity towards zero

15 Setting the Stage Goal of this Paper Estimate supply / demand for calories Calories: aggregate maize, wheat, rice, and soybeans Traditional Approach Regress supply on expected price (futures price) Our concern: Futures price is endogenous Example: Soybean Rust (Pest) Farmers anticipate soybeans rust (error term u) Response: decrease planting area Futures market: increase in futures price Lower quantity and higher price? Movement along demand curve Will bias supply elasticity towards zero

16 Setting the Stage Goal of this Paper

17 Setting the Stage Goal of this Paper Estimate supply / demand for calories Instrument: Yield Shock (deviations from trend) Identification of Demand Current yield shock shifts supply-curve Used since P.G. Wright introduced IV (1928) Identification of Supply Past yield shocks shift expected price Instrument futures price in supply equation New extension: Previous estimates find inelastic supply, yet simulations use positive elasticity. Assess U.S. ethanol mandate

18 Setting the Stage Goal of this Paper Estimate supply / demand for calories Instrument: Yield Shock (deviations from trend) Identification of Demand Current yield shock shifts supply-curve Used since P.G. Wright introduced IV (1928) Identification of Supply Past yield shocks shift expected price Instrument futures price in supply equation New extension: Previous estimates find inelastic supply, yet simulations use positive elasticity. Assess U.S. ethanol mandate

19 Setting the Stage Goal of this Paper Estimate supply / demand for calories Instrument: Yield Shock (deviations from trend) Identification of Demand Current yield shock shifts supply-curve Used since P.G. Wright introduced IV (1928) Identification of Supply Past yield shocks shift expected price Instrument futures price in supply equation New extension: Previous estimates find inelastic supply, yet simulations use positive elasticity. Assess U.S. ethanol mandate

20 Setting the Stage Goal of this Paper Estimate supply / demand for calories Instrument: Yield Shock (deviations from trend) Identification of Demand Current yield shock shifts supply-curve Used since P.G. Wright introduced IV (1928) Identification of Supply Past yield shocks shift expected price Instrument futures price in supply equation New extension: Previous estimates find inelastic supply, yet simulations use positive elasticity. Assess U.S. ethanol mandate

21 Setting the Stage Goal of this Paper

22 Setting the Stage Goal of this Paper

23 Setting the Stage Goal of this Paper

24 Background: Agricultural Production World Caloric Production and Prices Commodity crops form basis of food chain Cassman (1999) attributes two thirds of caloric production to maize, wheat, and rice Adding in soybeans brings ratio to 75% Conversion of production quantities into calories Williamson and Williamson (1942) Green revolution Caloric Production (4 commodities): +249% ( ). Growth on intensive margin (output per area) Limited expansion in area Increase by 40%

25 Background: Agricultural Production World Caloric Production and Prices Commodity crops form basis of food chain Cassman (1999) attributes two thirds of caloric production to maize, wheat, and rice Adding in soybeans brings ratio to 75% Conversion of production quantities into calories Williamson and Williamson (1942) Green revolution Caloric Production (4 commodities): +249% ( ). Growth on intensive margin (output per area) Limited expansion in area Increase by 40%

26 Background: Agricultural Production World Caloric Production and Prices Commodity crops form basis of food chain Cassman (1999) attributes two thirds of caloric production to maize, wheat, and rice Adding in soybeans brings ratio to 75% Conversion of production quantities into calories Williamson and Williamson (1942) Green revolution Caloric Production (4 commodities): +249% ( ). Growth on intensive margin (output per area) Limited expansion in area Increase by 40%

27 Background: Agricultural Production World Caloric Production and Prices

28 Background: Agricultural Production World Caloric Production and Prices

29 Background: Agricultural Production U.S. Share of Caloric Production Country Market Share Maize United States of America China Brazil 5.21 USSR 3.52 Mexico 3.01 Wheat USSR China United States of America India 8.53 Russian Federation 6.86 Soybeans United States of America Brazil China Argentina 6.62 India 1.63

30 Background: Agricultural Production U.S. Share of Caloric Production

31 Ethanol History of Biofuels Long history of ethanol as fuel Ford s Model-T designed to run on ethanol Slow phase-out of ethanol as petroleum became cheaper Renewed interest in ethanol to combat CO 2 emissions 2005 U.S. Energy Bill: 7.5 billion gallons by U.S. Energy Bill: 36 billion gallons by Renewable Fuels Standard: 11 billion gallons in percent of world caloric production Gasoline use: 0.39 billion gallons per day 11 billion gallons: 28 days (8% of yearly demand)

32 Ethanol History of Biofuels Long history of ethanol as fuel Ford s Model-T designed to run on ethanol Slow phase-out of ethanol as petroleum became cheaper Renewed interest in ethanol to combat CO 2 emissions 2005 U.S. Energy Bill: 7.5 billion gallons by U.S. Energy Bill: 36 billion gallons by Renewable Fuels Standard: 11 billion gallons in percent of world caloric production Gasoline use: 0.39 billion gallons per day 11 billion gallons: 28 days (8% of yearly demand)

33 Ethanol History of Biofuels

34 Ethanol History of Biofuels Long history of ethanol as fuel Ford s Model-T designed to run on ethanol Slow phase-out of ethanol as petroleum became cheaper Renewed interest in ethanol to combat CO 2 emissions 2005 U.S. Energy Bill: 7.5 billion gallons by U.S. Energy Bill: 36 billion gallons by Renewable Fuels Standard: 11 billion gallons in percent of world caloric production Gasoline use: 0.39 billion gallons per day 11 billion gallons: 28 days (8% of yearly demand)

35 Outline 1 Motivation 2 Methodology 3 Data 4 Empirical Results 5 Conclusions

36 Theory Storage Literature Competitive agricultural producers make two decisions Food availability is z t How much to store x t Cost of storage φ(x) convex How much effort to put into new production λ t Cost of effort g(λ) convex Production subject to random weather shock s t+1 = λ t ω t+1 ω t+1 is random weather draw Equation of motion z t+1 = x t + λ t ω t+1

37 Theory Storage Literature Competitive agricultural producers make two decisions Food availability is z t How much to store x t Cost of storage φ(x) convex How much effort to put into new production λ t Cost of effort g(λ) convex Production subject to random weather shock s t+1 = λ t ω t+1 ω t+1 is random weather draw Equation of motion z t+1 = x t + λ t ω t+1

38 Theory Storage Literature Competitive agricultural producers make two decisions Food availability is z t How much to store x t Cost of storage φ(x) convex How much effort to put into new production λ t Cost of effort g(λ) convex Production subject to random weather shock s t+1 = λ t ω t+1 ω t+1 is random weather draw Equation of motion z t+1 = x t + λ t ω t+1

39 Theory Storage Literature Competitive agricultural producers make two decisions Food availability is z t How much to store x t Cost of storage φ(x) convex How much effort to put into new production λ t Cost of effort g(λ) convex Production subject to random weather shock s t+1 = λ t ω t+1 ω t+1 is random weather draw Equation of motion z t+1 = x t + λ t ω t+1

40 Theory Storage Literature Bellman equation v(z t) = max x t λ t {u(z t x t) φ(x t) g(λ t) + δe [v(z t+1 )]} subject to z t+1 = x t + λ tω t+1 x t 0, z t x t 0, λ t 0 Solved by Scheinkman and Schechtman (1983) and Bobenrieth et al. (2002) (i) consumption c t = z t x t is strictly increasing in z t (ii) storage x t is weakly increasing in z t (iii) effort λ t is weakly decreasing in z t

41 Empirical Implementation Model Estimated equations log(s t) = α s + β slog (E[p t t 1 ]) + γ sω t + f (t) + u t log(z t x t) = α d + β d log(p t) + g(t) + v t s t : production of calories at time t z t x t : demand for calories at time t p t : price of calories at time t og ( E(p t t 1 ) ) : futures price (delivery in t, traded in t 1) ω t : Yield shocks (weather induced) at time t f (t), g(t): time trend (technological change, population growth) u, v: error terms

42 Outline 1 Motivation 2 Methodology 3 Data 4 Empirical Results 5 Conclusions

43 Agricultural Data Data Sources FAO series (country-level) Crops used: maize, wheat, rice, and soybeans Production, area, and yield ( ) Change in inventories ( ) Common unit: calories Conversion using calories per unit of production USDA Inventory levels (1961): corn, wheat, and rice

44 Agricultural Data Data Sources FAO series (country-level) Crops used: maize, wheat, rice, and soybeans Production, area, and yield ( ) Change in inventories ( ) Common unit: calories Conversion using calories per unit of production USDA Inventory levels (1961): corn, wheat, and rice

45 Agricultural Data Data Sources FAO series (country-level) Crops used: maize, wheat, rice, and soybeans Production, area, and yield ( ) Change in inventories ( ) Common unit: calories Conversion using calories per unit of production USDA Inventory levels (1961): corn, wheat, and rice

46 Yield Shocks Data Sources - Caloric Yield Shocks Yield shocks Baseline model: country-specific deviations from quadratic yield trends for each crop Countries used Countries with more than 1% of world production of crop Remaining countries lumped together as Rest of World Caloric Shock Sum of country and crop-specific shocks Normalized by quadratic production trend

47 Yield Shocks Data Sources - Caloric Yield Shocks Yield shocks Baseline model: country-specific deviations from quadratic yield trends for each crop Countries used Countries with more than 1% of world production of crop Remaining countries lumped together as Rest of World Caloric Shock Sum of country and crop-specific shocks Normalized by quadratic production trend

48 Yield Shocks Data Sources - Caloric Yield Shocks Yield shocks Baseline model: country-specific deviations from quadratic yield trends for each crop Countries used Countries with more than 1% of world production of crop Remaining countries lumped together as Rest of World Caloric Shock Sum of country and crop-specific shocks Normalized by quadratic production trend

49 Yield Shocks Jackknifed Residuals

50 Yield Shocks Jackknifed Residuals

51 Price Data Data Sources - Prices Long time series Crop prices in US in December of each year Futures prices Chicago Board of trade Delivery: September (wheat), November (soybeans), December (corn) p t : average price at delivery log (E[p t t 1 ]): one year before Only available for maize, soybeans, and wheat (not rice) Price per calory Converted using calory per unit of production Deflated using CPI

52 Price Data Data Sources - Prices Long time series Crop prices in US in December of each year Futures prices Chicago Board of trade Delivery: September (wheat), November (soybeans), December (corn) p t : average price at delivery log (E[p t t 1 ]): one year before Only available for maize, soybeans, and wheat (not rice) Price per calory Converted using calory per unit of production Deflated using CPI

53 Price Data Data Sources - Prices Long time series Crop prices in US in December of each year Futures prices Chicago Board of trade Delivery: September (wheat), November (soybeans), December (corn) p t : average price at delivery log (E[p t t 1 ]): one year before Only available for maize, soybeans, and wheat (not rice) Price per calory Converted using calory per unit of production Deflated using CPI

54 Outline 1 Motivation 2 Methodology 3 Data 4 Empirical Results 5 Conclusions

55 Supply - Demand Model Regression Results - Demand Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS β d (s.e.) (0.0190) β s (s.e.) p (95%) p t t t 2 t e-02 (1.90e-02) 4.26e-02 (8.32e-04) -4.18e-04 (2.34e-05) Demand N 42 I 2 K 1

56 Supply - Demand Model Regression Results - Demand Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS β d (s.e.) (0.0190) (0.0167) β s (s.e.) p (95%) p t -5.05e e-02 (1.90e-02) (1.67e-02) t 4.26e e-02 (8.32e-04) (8.57e-04) t e e-04 (2.34e-05) (2.28e-05) t 3 Demand N I 2 2 K 1 1

57 Supply - Demand Model Regression Results - Demand Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) β s (s.e.) p (95%) Demand p t -5.05e e e e-02 (1.90e-02) (1.67e-02) (2.43e-02) (2.15e-02) t 4.26e e e e-02 (8.32e-04) (8.57e-04) (2.50e-03) (2.81e-03) t e e e e-04 (2.34e-05) (2.28e-05) (1.53e-04) (1.63e-04) t e e-06 (2.26e-06) (2.37e-06) N I K

58 Supply - Demand Model Regression Results - Demand Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) p (95%) Demand p t -5.05e e e e e e-02 (1.90e-02) (1.67e-02) (2.43e-02) (2.15e-02) (2.41e-02) (2.26e-02) t 4.26e e e e e e-02 (8.32e-04) (8.57e-04) (2.50e-03) (2.81e-03) (3.03e-03) (3.44e-03) t e e e e e e-04 (2.34e-05) (2.28e-05) (1.53e-04) (1.63e-04) (1.77e-04) (1.93e-04) t e e e e-06 (2.26e-06) (2.37e-06) (2.57e-06) (2.74e-06) N I K

59 Supply - Demand Model Regression Results - Supply Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) p (95%) E[p t t 1 ] ω t t t 2 t e-01 (2.86e-02) 2.46e-01 (3.37e-02) 4.46e-02 (9.34e-04) -3.54e-04 (2.66e-05) Supply N 42 I 2 K 1

60 Supply - Demand Model Regression Results - Supply Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) p (95%) (21.32,50.14) E[p t t 1 ] ω t t t 2 t e-01 (2.86e-02) 2.46e-01 (3.37e-02) 4.46e-02 (9.34e-04) -3.54e-04 (2.66e-05) Supply N 42 I 2 K 1

61 Supply - Demand Model Regression Results - Supply Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) (0.0241) p (95%) (21.32,50.14) (20.69,36.62) E[p t t 1 ] 1.17e e-01 (2.86e-02) (2.41e-02) ω t 2.46e e-01 (3.37e-02) (2.94e-02) t 4.46e e-02 (9.34e-04) (8.74e-04) t e e-04 (2.66e-05) (2.40e-05) t 3 Supply N I 2 2 K 1 1

62 Supply - Demand Model Regression Results - Supply Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) (0.0241) (0.0217) (0.0189) p (95%) (21.32,50.14) (20.69,36.62) (23.75,60.31) (22.01,40.80) Supply E[p t t 1 ] 1.17e e e e-02 (2.86e-02) (2.41e-02) (2.17e-02) (1.89e-02) ω t 2.46e e e e-01 (3.37e-02) (2.94e-02) (2.65e-02) (2.38e-02) t 4.46e e e e-02 (9.34e-04) (8.74e-04) (2.04e-03) (1.89e-03) t e e e e-04 (2.66e-05) (2.40e-05) (1.12e-04) (1.04e-04) t e e-06 (1.68e-06) (1.55e-06) N I K

63 Supply - Demand Model Regression Results - Supply Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) (0.0241) (0.0217) (0.0189) (0.0208) (0.0189) p (95%) (21.32,50.14) (20.69,36.62) (23.75,60.31) (22.01,40.80) (22.23,50.00) (22.79,48.40) Supply E[p t t 1 ] 1.17e e e e e e-02 (2.86e-02) (2.41e-02) (2.17e-02) (1.89e-02) (2.08e-02) (1.89e-02) ω t 2.46e e e e e e-01 (3.37e-02) (2.94e-02) (2.65e-02) (2.38e-02) (2.56e-02) (2.35e-02) t 4.46e e e e e e-02 (9.34e-04) (8.74e-04) (2.04e-03) (1.89e-03) (2.32e-03) (2.14e-03) t e e e e e e-04 (2.66e-05) (2.40e-05) (1.12e-04) (1.04e-04) (1.26e-04) (1.16e-04) t e e e e-06 (1.68e-06) (1.55e-06) (1.81e-06) (1.68e-06) N I K

64 Supply - Demand Model Regression Results - First Stage Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS Demand ω t -1.19e e e e e e+00 (2.62e-01) (2.47e-01) (2.93e-01) (2.65e-01) (2.97e-01) (2.57e-01) ω t e e-01 (2.95e-01) (2.02e-01) t -8.43e e e e e e-02 (9.73e-03) (1.01e-02) (2.64e-02) (2.84e-02) (3.22e-02) (3.28e-02) t e e e e e e-03 (2.24e-04) (2.28e-04) (1.47e-03) (1.53e-03) (1.72e-03) (1.71e-03) t e e e e-05 (2.27e-05) (2.33e-05) (2.60e-05) (2.54e-05) N I K

65 Supply - Demand Model Regression Results - First Stage Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS Supply ω t e e e e e e-01 (2.14e-01) (1.91e-01) (2.26e-01) (1.98e-01) (2.20e-01) (1.96e-01) ω t e e-01 (2.21e-01) (1.89e-01) ω t -6.10e e e e e e-01 (2.10e-01) (1.97e-01) (2.27e-01) (2.09e-01) (2.20e-01) (1.99e-01) t -1.04e e e e e e-02 (8.15e-03) (7.64e-03) (2.46e-02) (2.26e-02) (2.77e-02) (2.51e-02) t e e e e e e-04 (1.85e-04) (1.73e-04) (1.32e-03) (1.21e-03) (1.43e-03) (1.30e-03) t e e e e-05 (1.99e-05) (1.83e-05) (2.10e-05) (1.91e-05) N I K

66 Contrast to Other Approaches Contrast to Other Approaches SUR - Price Not Instrumented Demand Instrumented / Supply Not Instrumented (1) (2) (3) (4) (5) (6) β d (s.e.) (0.0094) (0.0098) β s (s.e.) (0.0182) (0.0162) p (95%) (-694,1148) (-647,1145) I 2 3 K n.a. n.a. L n.a. n.a.

67 Contrast to Other Approaches Contrast to Other Approaches SUR - Price Not Instrumented Demand Instrumented / Supply Not Instrumented (1) (2) (3) (4) (5) (6) β d (s.e.) (0.0094) (0.0098) (0.0180) (0.0180) (0.0180) (0.0243) β s (s.e.) (0.0182) (0.0162) (0.0239) (0.0251) (0.0274) (0.0239) p (95%) (-694,1148) (-647,1145) (37,300) (36,295) (34,344) (32,209) I K n.a. n.a L n.a. n.a

68 Area Responses Explaining World Production Area World (1) (2) (3) (4) (5) (6) ω t (0.0147) (0.0186) E[p t t 1 ] N I 2 3 K n.a. n.a.

69 Area Responses Explaining World Production Area World (1) (2) (3) (4) (5) (6) ω t (0.0147) (0.0186) E[p t t 1 ] (0.0146) (0.0148) (0.0130) (0.0140) N I K n.a. n.a

70 Area Responses Explaining World Production Area United States (1) (2) (3) (4) (5) (6) ω t (0.0654) (0.0826) E[p t t 1 ] (0.0562) (0.0566) (0.0504) (0.0527) N I K n.a. n.a

71 Area Responses Explaining World Production Area Brazil (1) (2) (3) (4) (5) (6) ω t (0.0731) (0.0897) E[p t t 1 ] (0.1096) (0.0947) (0.0986) (0.0877) N I K n.a. n.a

72 Area Responses Explaining World Production Area China (1) (2) (3) (4) (5) (6) ω t (0.0272) (0.0340) E[p t t 1 ] (0.0299) (0.0311) (0.0265) (0.0277) N I K n.a. n.a

73 Area Responses Implications Ethanol mandate: 5% of world caloric production Food prices increase 30 percent (no distiller grain recycling) Loss in consumer surplus: 155 billion annually But: offsetting increase in producer surplus Potential consumer surplus from lower fuel prices Rajagopal (2007) Need demand / supply elasticity of fuels Elastic supply Lower price increase Larger expansion in area / yield 2percent increase or 30 million acres Land use constitutes 20% of CO 2 emissions

74 Area Responses Implications

75 Area Responses Implications Ethanol mandate: 5% of world caloric production Food prices increase 30 percent (no distiller grain recycling) Loss in consumer surplus: 155 billion annually But: offsetting increase in producer surplus Potential consumer surplus from lower fuel prices Rajagopal (2007) Need demand / supply elasticity of fuels Elastic supply Lower price increase Larger expansion in area / yield 2percent increase or 30 million acres Land use constitutes 20% of CO 2 emissions

76 Area Responses Implications Ethanol mandate: 5% of world caloric production Food prices increase 30 percent (no distiller grain recycling) Loss in consumer surplus: 155 billion annually But: offsetting increase in producer surplus Potential consumer surplus from lower fuel prices Rajagopal (2007) Need demand / supply elasticity of fuels Elastic supply Lower price increase Larger expansion in area / yield 2percent increase or 30 million acres Land use constitutes 20% of CO 2 emissions

77 Outline 1 Motivation 2 Methodology 3 Data 4 Empirical Results 5 Conclusions

78 Summary Conclusions Demand and supply model of commodity calories What s new? Aggregation of crops by caloric content New supply instrument (instrumented lagged price) Major results Significant supply and demand elasticities Previous literature found insignificant supply elasticities Implications for U.S. ethanol mandate Predicted to raise world prices by 30 percent Scales linearly in recycling (distiller grains) Loss of consumer surplus 150 billion annually Expansion in area by 2 percent (30 million acres)

79 Summary Conclusions Demand and supply model of commodity calories What s new? Aggregation of crops by caloric content New supply instrument (instrumented lagged price) Major results Significant supply and demand elasticities Previous literature found insignificant supply elasticities Implications for U.S. ethanol mandate Predicted to raise world prices by 30 percent Scales linearly in recycling (distiller grains) Loss of consumer surplus 150 billion annually Expansion in area by 2 percent (30 million acres)

80 Summary Conclusions Demand and supply model of commodity calories What s new? Aggregation of crops by caloric content New supply instrument (instrumented lagged price) Major results Significant supply and demand elasticities Previous literature found insignificant supply elasticities Implications for U.S. ethanol mandate Predicted to raise world prices by 30 percent Scales linearly in recycling (distiller grains) Loss of consumer surplus 150 billion annually Expansion in area by 2 percent (30 million acres)

81 Summary Conclusions Demand and supply model of commodity calories What s new? Aggregation of crops by caloric content New supply instrument (instrumented lagged price) Major results Significant supply and demand elasticities Previous literature found insignificant supply elasticities Implications for U.S. ethanol mandate Predicted to raise world prices by 30 percent Scales linearly in recycling (distiller grains) Loss of consumer surplus 150 billion annually Expansion in area by 2 percent (30 million acres)

82 Background Methodology Data Results Outline 6 Agriculture and Ethanol 7 Methodology 8 Data 9 Empirical Results

83 Background Methodology Data Results World Food Prices Background - World Food Prices Possible explanations for threefold price increase Detrimental weather Prolonged drought in Australia Rising oil price Increased demand for meat (China and India) China: 33fold increase in per-capita meat consumption ( ) Meat requires 5-10 times the area per calorie 20% reduction in U.S. meat consumption equivalent to switching from Camry to Prius Currency Fluctuations (Dollar) U.S. ethanol subsidies / mandate

84 Background Methodology Data Results World Food Prices Background - World Food Prices Possible explanations for threefold price increase Detrimental weather Prolonged drought in Australia Rising oil price Increased demand for meat (China and India) China: 33fold increase in per-capita meat consumption ( ) Meat requires 5-10 times the area per calorie 20% reduction in U.S. meat consumption equivalent to switching from Camry to Prius Currency Fluctuations (Dollar) U.S. ethanol subsidies / mandate

85 Background Methodology Data Results World Food Prices Background - World Food Prices Possible explanations for threefold price increase Detrimental weather Prolonged drought in Australia Rising oil price Increased demand for meat (China and India) China: 33fold increase in per-capita meat consumption ( ) Meat requires 5-10 times the area per calorie 20% reduction in U.S. meat consumption equivalent to switching from Camry to Prius Currency Fluctuations (Dollar) U.S. ethanol subsidies / mandate

86 Background Methodology Data Results World Food Prices Background - World Food Prices Possible explanations for threefold price increase Detrimental weather Prolonged drought in Australia Rising oil price Increased demand for meat (China and India) China: 33fold increase in per-capita meat consumption ( ) Meat requires 5-10 times the area per calorie 20% reduction in U.S. meat consumption equivalent to switching from Camry to Prius Currency Fluctuations (Dollar) U.S. ethanol subsidies / mandate

87 Background Methodology Data Results World Food Prices Background - World Food Prices Possible explanations for threefold price increase Detrimental weather Prolonged drought in Australia Rising oil price Increased demand for meat (China and India) China: 33fold increase in per-capita meat consumption ( ) Meat requires 5-10 times the area per calorie 20% reduction in U.S. meat consumption equivalent to switching from Camry to Prius Currency Fluctuations (Dollar) U.S. ethanol subsidies / mandate

88 Background Methodology Data Results Ethanol Mandate Background - Ethanol Mandate Environmental Protection Agency Recent Report (February 4th) Net CO 2 reduction Advocates increased use of biofuels Driving forces behind analysis Increased yield per acre Little area expansion required Searchinger et al. Critical of analysis Land expansion results in big CO 2 increase

89 Background Methodology Data Results Ethanol Mandate Background - Ethanol Mandate Environmental Protection Agency Recent Report (February 4th) Net CO 2 reduction Advocates increased use of biofuels Driving forces behind analysis Increased yield per acre Little area expansion required Searchinger et al. Critical of analysis Land expansion results in big CO 2 increase

90 Background Methodology Data Results Ethanol Mandate Background - Ethanol Mandate Environmental Protection Agency Recent Report (February 4th) Net CO 2 reduction Advocates increased use of biofuels Driving forces behind analysis Increased yield per acre Little area expansion required Searchinger et al. Critical of analysis Land expansion results in big CO 2 increase

91 Background Methodology Data Results Ethanol Mandate Background - Ethanol Mandate Environmental Protection Agency Recent Report (February 4th) Net CO 2 reduction Advocates increased use of biofuels Driving forces behind analysis Increased yield per acre Little area expansion required Searchinger et al. Critical of analysis Land expansion results in big CO 2 increase

92 Background Methodology Data Results US Share U.S. Effect on World Markets U.S. market share calories from maize, wheat, rice, soybeans roughly 23 percent Any policy that changes U.S. production has potential to influence world prices What is the influence of ethanol mandates?

93 Background Methodology Data Results US Share U.S. Effect on World Markets U.S. market share calories from maize, wheat, rice, soybeans roughly 23 percent Any policy that changes U.S. production has potential to influence world prices What is the influence of ethanol mandates?

94 Background Methodology Data Results US Share Implications for Commodity Markets U.S.: Ethanol predominantly produced from maize 11 billion gallons Require 4.23 billion bushels of corn Using 2.6 gallons per bushel average conversion Total U.S. maize production 13 billion bushels Ethanol Mandate One third of U.S. corn production 13 percent of world maize production 5 percent of world caloric production

95 Background Methodology Data Results US Share Implications for Commodity Markets U.S.: Ethanol predominantly produced from maize 11 billion gallons Require 4.23 billion bushels of corn Using 2.6 gallons per bushel average conversion Total U.S. maize production 13 billion bushels Ethanol Mandate One third of U.S. corn production 13 percent of world maize production 5 percent of world caloric production

96 Background Methodology Data Results US Share Implications for Commodity Markets U.S.: Ethanol predominantly produced from maize 11 billion gallons Require 4.23 billion bushels of corn Using 2.6 gallons per bushel average conversion Total U.S. maize production 13 billion bushels Ethanol Mandate One third of U.S. corn production 13 percent of world maize production 5 percent of world caloric production

97 Background Methodology Data Results Outline 6 Agriculture and Ethanol 7 Methodology 8 Data 9 Empirical Results

98 Background Methodology Data Results Storage Theory Implications of Storage Literature Negative weather shock in current period Reduces consumption c t Increases price p t Increases price p t+1 (prices linked through storage) Increased effort in t + 1 (supply response) Past yield shocks can be used to identify supply Storage links periods

99 Background Methodology Data Results Storage Theory Implications of Storage Literature Negative weather shock in current period Reduces consumption c t Increases price p t Increases price p t+1 (prices linked through storage) Increased effort in t + 1 (supply response) Past yield shocks can be used to identify supply Storage links periods

100 Background Methodology Data Results Storage Theory Supply / Demand System First-stage regressions: K 1 log(p t ) = π d0 + µ dk ω t k + log (E[p t t 1 ]) = π s0 + k=0 K µ sk ω t k + k=0 I ρ di t i + ɛ dt i=1 I ρ si t i + ɛ st i=1

101 Background Methodology Data Results Storage Theory Supply / Demand System Second-stage supply: log(s t ) = α s + β slog (E[pt t 1 ]) + λ s0 ω t + I τ si t i +u t i=1 } {{ } f (t) Stage-one variable excluded from the stage-two: ω t k,k=1...k

102 Background Methodology Data Results Storage Theory Supply / Demand System Second-stage demand: log(s t x t ) = α d + β d log(pt ) + I τ di t i +v t i=1 } {{ } g(t) Stage-one variable excluded from the stage-two: ω t k,k=0...k 1

103 Background Methodology Data Results Outline 6 Agriculture and Ethanol 7 Methodology 8 Data 9 Empirical Results

104 Background Methodology Data Results Country Display Major Agricultural Producers Maize: Production Share greater than 1 Percent

105 Background Methodology Data Results Country Display Major Agricultural Producers Wheat: Production Share greater than 1 Percent

106 Background Methodology Data Results Country Display Major Agricultural Producers Rice: Production Share greater than 1 Percent

107 Background Methodology Data Results Country Display Major Agricultural Producers Soybeans: Production Share greater than 1 Percent

108 Background Methodology Data Results Top Producers Production Shares Country Share Country Share Wheat Maize USSR United States of America China China United States of America Brazil 5.21 India 8.53 USSR 3.52 Russian Federation 6.86 Mexico 3.01 France 5.33 Yugoslav SFR 2.47 Canada 4.81 Argentina 2.35 Turkey 3.48 France 2.32 Australia 3.13 Romania 2.15 Germany 2.89 South Africa 2.01 Ukraine 2.69 India 1.91 Pakistan 2.49 Italy 1.54 Argentina 2.23 Hungary 1.41 Italy 2.06 Indonesia 1.26 United Kingdom 2.01 Canada 1.15 Kazakhstan 1.87 Rest of World Iran, Islamic Republic of 1.54 Poland 1.38 Yugoslav SFR 1.29 Romania 1.27 Spain 1.16 Czechoslovakia 1.05 Rest of World 12.12

109 Background Methodology Data Results Top Producers Production Shares Country Share Country Share Rice Soybeans China United States of America India Brazil Indonesia 7.50 China Bangladesh 5.48 Argentina 6.62 Thailand 4.27 India 1.63 Vietnam 3.97 Canada 1.04 Japan 3.67 Rest of World 6.49 Myanmar 3.12 Brazil 2.08 Philippines 1.87 Korea, Republic of 1.59 United States of America 1.44 Pakistan 1.07 Rest of World 8.86

110 Background Methodology Data Results Yield Shocks Jackknifed Residuals

111 Background Methodology Data Results Yield Shocks Jackknifed Residuals

112 Background Methodology Data Results Yield Shocks Jackknifed Residuals

113 Background Methodology Data Results Yield Shocks Jackknifed Residuals

114 Background Methodology Data Results Yield Shocks Jackknifed Residuals

115 Background Methodology Data Results Yield Shocks Jackknifed Residuals

116 Background Methodology Data Results Yield Shocks Jackknifed Residuals

117 Background Methodology Data Results Yield Shocks Jackknifed Residuals

118 Background Methodology Data Results Weather Data Data Sources - Weather Data Potential concern Are yields endogenous to price? Higher price could lead to higher sowing density Higher price could imply shift to marginal land Lack of autocorrelation in yields suggests no Lack of correlation between years suggests no Sensitivity check Country-and-crop specific yield regressions US data (Schlenker and Roberts, 2009) World data: CRU (monthly averages on 0.5 degree grid) Average over agricultural area Caloric Shock Attributable to deviations from average weather

119 Background Methodology Data Results Weather Data Data Sources - Weather Data

120 Background Methodology Data Results Weather Data Data Sources - Weather Data Potential concern Are yields endogenous to price? Higher price could lead to higher sowing density Higher price could imply shift to marginal land Lack of autocorrelation in yields suggests no Lack of correlation between years suggests no Sensitivity check Country-and-crop specific yield regressions US data (Schlenker and Roberts, 2009) World data: CRU (monthly averages on 0.5 degree grid) Average over agricultural area Caloric Shock Attributable to deviations from average weather

121 Background Methodology Data Results Weather Data Data Sources - Weather Data Potential concern Are yields endogenous to price? Higher price could lead to higher sowing density Higher price could imply shift to marginal land Lack of autocorrelation in yields suggests no Lack of correlation between years suggests no Sensitivity check Country-and-crop specific yield regressions US data (Schlenker and Roberts, 2009) World data: CRU (monthly averages on 0.5 degree grid) Average over agricultural area Caloric Shock Attributable to deviations from average weather

122 Background Methodology Data Results Weather Data Growing Areas CRU grid system

123 Background Methodology Data Results Weather Data Growing Areas Maize: Growing Area (Fraction of Grid)

124 Background Methodology Data Results Descriptive Statistics Descriptive Statistics Variable Unit Mean Std. Dev. Min Max Year Caloric Production billion people Caloric Storage million people Caloric Stock million people Caloric Shock (Linear Trend) million people Caloric Shock (Quadratic Trend) million people Caloric Shock (Weather Linear) million people Caloric Shock (Weather Quadratic) million people Caloric Price (Futures Delivery) US$2007 per year Caloric Price (Futures Prev. Year) US$2007 per year Caloric Price (Dec. USDA Prices) US$2007 per year

125 Background Methodology Data Results Outline 6 Agriculture and Ethanol 7 Methodology 8 Data 9 Empirical Results

126 Background Methodology Data Results Sensitivity Checks Sensitivity Check - First Stage Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS Demand ω t,ms -1.14e e e e e e+00 (4.71e-01) (4.03e-01) (4.79e-01) (3.87e-01) (4.83e-01) (3.74e-01) ω t,rw -1.22e e e e e e+00 (3.51e-01) (3.10e-01) (4.41e-01) (3.59e-01) (4.58e-01) (3.54e-01) ω t 1,MS 1.95e e-02 (4.90e-01) (3.38e-01) ω t 1,RW -9.29e e-01 (4.35e-01) (3.34e-01) t -8.75e e e e e e-03 (1.02e-02) (1.03e-02) (3.11e-02) (3.17e-02) (4.25e-02) (3.96e-02) t e e e e e e-04 (2.37e-04) (2.34e-04) (1.71e-03) (1.69e-03) (2.23e-03) (2.04e-03) t e e e e-06 (2.60e-05) (2.54e-05) (3.26e-05) (2.97e-05) F-stat χ 2 -stat p-value

127 Background Methodology Data Results Sensitivity Checks Sensitivity Check - First Stage Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS Supply ω t 1,MS -6.88e e e e e e-01 (3.67e-01) (3.15e-01) (3.66e-01) (3.03e-01) (3.64e-01) (3.11e-01) ω t 1,RW -8.56e e e e e e-01 (3.00e-01) (2.59e-01) (3.25e-01) (2.76e-01) (3.37e-01) (2.88e-01) ω t 2,MS 7.62e e-02 (3.61e-01) (2.88e-01) ω t 2,RW -5.75e e-01 (3.28e-01) (2.64e-01) ω t,ms -2.63e e e e e e-01 (3.64e-01) (3.32e-01) (3.61e-01) (3.20e-01) (3.50e-01) (3.01e-01) ω t,rw -8.22e e e e e e+00 (2.94e-01) (2.66e-01) (3.42e-01) (3.03e-01) (3.45e-01) (2.96e-01) t -1.31e e e e e e-02 (8.75e-03) (7.94e-03) (3.18e-02) (2.78e-02) (3.77e-02) (3.22e-02) t e e e e e e-03 (2.00e-04) (1.81e-04) (1.66e-03) (1.46e-03) (1.90e-03) (1.63e-03) t e e e e-05 (2.43e-05) (2.14e-05) (2.72e-05) (2.33e-05) F-stat χ 2 -stat p-value

128 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations Jackknifed Yield Residuals Check: Linear instead of quadratic trend Production trend Check: Linear instead of quadratic trend

129 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS Panel A: Baseline β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) (0.0241) (0.0217) (0.0189) (0.0208) (0.0189) p (95%) (21.32,50.14) (20.69,36.62) (23.75,60.31) (22.01,40.80) (22.23,50.00) (22.79,48.40) Panel B: Caloric Shock Derived using Linear Time Trend β d (s.e.) (0.0192) (0.0169) (0.0234) (0.0211) (0.0234) (0.0220) β s (s.e.) (0.0261) (0.0219) (0.0230) (0.0194) (0.0229) (0.0206) p (95%) (22.88,54.59) (22.32,39.68) (23.80,61.39) (22.33,41.26) (22.07,50.88) (22.67,48.94) N I K

130 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations Caloric shock is product of Jackknifed yield residuals Area harvested Caloric conversion (calories per unit of output) Check: Predicted area (quadratic trend) instead of actual

131 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS Panel A: Baseline β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) (0.0241) (0.0217) (0.0189) (0.0208) (0.0189) p (95%) (21.32,50.14) (20.69,36.62) (23.75,60.31) (22.01,40.80) (22.23,50.00) (22.79,48.40) Panel C: Caloric Shock Derived using Quadratic Area Trend β d (s.e.) (0.0185) (0.0165) (0.0233) (0.0211) (0.0233) (0.0216) β s (s.e.) (0.0274) (0.0230) (0.0206) (0.0178) (0.0199) (0.0180) p (95%) (21.66,49.45) (21.23,37.39) (24.04,58.22) (22.59,41.46) (22.44,48.72) (23.15,47.34) N I K

132 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations Caloric conversion factors Given in Williamson and Williamson (1942) Check: Ratio of caloric conversion factors equals ratio of averages prices

133 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations

134 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations

135 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS Panel A: Baseline β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) (0.0241) (0.0217) (0.0189) (0.0208) (0.0189) p (95%) (21.32,50.14) (20.69,36.62) (23.75,60.31) (22.01,40.80) (22.23,50.00) (22.79,48.40) Panel D: Rescaled Caloric Conversion Factors to Equalize Average Prices β d (s.e.) (0.0184) (0.0151) (0.0237) (0.0167) (0.0200) (0.0199) β s (s.e.) (0.0362) (0.0289) (0.0165) (0.0142) (0.0154) (0.0139) p (95%) (18.71,46.27) (19.92,37.27) (23.97,52.50) (25.63,41.52) (25.44,51.32) (25.87,51.01) N I K

136 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations Shock ω t Ratio of relative caloric shock to relative inventory level Check: Do not normalize by inventory level

137 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS Panel A: Baseline β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) (0.0241) (0.0217) (0.0189) (0.0208) (0.0189) p (95%) (21.32,50.14) (20.69,36.62) (23.75,60.31) (22.01,40.80) (22.23,50.00) (22.79,48.40) Panel E: Caloric Shock not Divided by Inventory β d (s.e.) (0.0180) (0.0158) (0.0225) (0.0198) (0.0218) (0.0205) β s (s.e.) (0.0285) (0.0230) (0.0208) (0.0172) (0.0193) (0.0169) p (95%) (21.55,50.13) (21.46,36.99) (24.67,60.68) (23.65,41.84) (23.51,50.83) (24.11,47.24) N I K

138 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations Futures prices Traded one year in advance Check: Northern hemisphere has knowledge of outcome in South at planting Futures prices at planting in North (March)

139 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Yield Deviations Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS Panel A: Baseline β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) (0.0241) (0.0217) (0.0189) (0.0208) (0.0189) p (95%) (21.32,50.14) (20.69,36.62) (23.75,60.31) (22.01,40.80) (22.23,50.00) (22.79,48.40) Panel F: Futures Price for Maize and Soybeans Traded in March β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0323) (0.0268) (0.0232) (0.0199) (0.0218) (0.0197) p (95%) (20.20,49.76) (19.40,34.71) (23.18,59.42) (21.35,39.31) (21.87,49.45) (22.33,47.74) N I K

140 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Weather Shocks Caloric shocks Deviations from yield trend Check Yield shocks that are attributable to weather shocks

141 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Weather Shocks Model 2SLS 3SLS 2SLS 3SLS 2SLS 3SLS Panel A: Baseline β d (s.e.) (0.0190) (0.0167) (0.0243) (0.0215) (0.0241) (0.0226) β s (s.e.) (0.0286) (0.0241) (0.0217) (0.0189) (0.0208) (0.0189) p (95%) (21.32,50.14) (20.69,36.62) (23.75,60.31) (22.01,40.80) (22.23,50.00) (22.79,48.40) Panel B: Production Shock Derived using Observed Weather β d (s.e.) (0.1144) (0.0494) (0.1197) (0.0539) (0.0621) (0.0347) β s (s.e.) ( ) (0.0388) ( ) (0.0461) (0.4016) (0.3542) p (95%) (-1.65,1.66) (14.88,55.91) (-2.43,2.47) (13.75,58.30) ( ,170.22) ( ,131.75) N I K

142 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Weather Shocks Weak instruments and decrease significance levels Likely due to data problems Correlation between two yield shocks Deviations from trend Attributable to weather US (good daily data): 0.71 Rest of world not as good

143 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Weather Shocks

144 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Weather Shocks

145 Background Methodology Data Results Sensitivity Checks Sensitivity Check - Weather Shocks