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 J. Roberts & Wolfram Schlenker NC State University Columbia University CRC LCA Biofuel Workshop Argonne National Laboratory, IL October 18, 2011

2 Why are Commodity Prices High? Demand growth in Asia. Weather shocks. Ethanol. Climate Change? Goldman Sachs?

3 This Talk 1. Statistics on world agriculture and US role. 2. Supply, demand & the identification problem 3. Our econometric estimates

4 One Statistics on world agriculture and US role.

5 Four Key Crops (About 75% of world caloric base)

6 The United States Production 39% of corn 38% soybeans 9% of wheat 2% of rice Much larger shares of world exports

7 United States Ethanol 40% of US corn production 5% of world caloric base

8 US caloric share is about 23% World fluctuations shadow US fluctuations

9 World Crop Yields

10 Prices Fluctuate Together Prices fluctuations larger than quantity fluctuations Highly autocorrelated

11 Two Supply, Demand & the Identification Problem.

12 Food VS Fuel P S Higher prices P 1 P 0 Ethanol subsidies shift demand out D FOOD + FUEL D FOOD F 1 Fewer food calories Q 0 Q 1 More total calories Q

13 The Price Effect of Ethatnol

14 Identifying Supply and Demand P S u P BAD P 0 v D Q 0 Q

15 Weather Shocks Identify Demand P S BAD S EXPECTED S GOOD P BAD P EXPECTED P GOOD D Q BAD Q GOOD Q Q EXPECTED

16 Identification of Supply Traditional approach (Nerlove, 1958) Regress quantity on expected price 1. Autoregressive prediction 2. Futures price Problem: Prices still endogenous to marketanticipated supply shifts Consider what the error is in the supply equation Consider what drives variation in futures prices

17 Identification of Supply Storage buffers weather shocks. Quantity-consumed shock is smaller than quantity-produced shock. Q CONSUMED = Q SUPPLED + Inventories Transmits current weather shocks to future expected prices.

18 P Using weather shocks to identify supply S EXPECTED E[P BAD PAST] E[P GOOD PAST] Past weather variation shifts inventories, changing expected price Q G Q B Q

19 Estimated Equations Supply Demand

20 First Stage Equations Supply Demand

21 Three Our Econometric Estimates

22 Estimating Shocks Two approaches: 1. Yield shocks Sum jackknifed residuals from country-by-crop trends 2. Weather Good for United States Not so good for rest of world Large standard errors

23 Worldwide Caloric Yield Shocks Drive Price Fluctuations

24 Results for Demand -Basic two-stage least squares -Quadratic time trend -One yield-shock lag

25 Results for Demand Same except 3SLS

26 Results for Demand -2SLS -cubic time trend

27 Results for Demand -3SLS -cubic trend

28 Results for Demand -2SLS -cubic trend -two lags of shocks

29 Results for Supply

30 First Stage Results--Demand

31 First Stage Results--Supply

32 The Punchline (Food for about 120 million)

33 Source of Ethanol 2/3: New production 1/3: Less food

34 Growing Area Response to Price World

35 Growing Area Response to Price Brazil

36 Growing Area Response to Price United States

37 Growing Area Response to Price China

38 Growing Area Response to Price India

39 Agricultural Policy Drives US Land Use

40 Prices Drive Agricultural Policy (until recently )

41 Some Robustness Checks & Extensions 1. Flexibility of country-specific trend used for yield shock estimates. 2. Use trend harvested acres rather than actual harvested acres in yield shock estimates. 3. Separate shocks for different crops effects on aggregate price look similar. 4. Raw shocks and shocks relative to inventories 5. Different months for futures prices on the supply equation 6. Land area responses for major countries

42 FAQ Q: Why aggregate calories? A: (1) Simplicity. (2) Value-weighted averages give the same estimates. (3) Prices vary together so cross-price elasticities difficult to identify (but we are trying). Q: What if yields or weather are autocorrelated? A: We include current weather in the supply equation. Q: Are FAO inventory estimates any good? A: We think they are good for big countries and especially the United States. Errors do not have strong correlation with instruments. FAS numbers give similar results. Probably not good enough for countrylevel demand estimation. Q: Why not structural estimates? A: Good idea. But could the take home story be much different?

43 Some Extensions Underway 1. Replicate with USDA-FAS data rather than FAO data 2. Crop-specific estimates and cross-price elasticities 3. Model price transitions with calibrated storage model

44 Summary First-order approximation to food commodity supply and demand on a global scale. Prices are very sensitive to quantities. Supply somewhat more elastic than demand. About 15-60% higher world caloric price due to US ethanol expansion. About 1/3 of calories used in ethanol production come from food. Significant indirect land use effects.