The NBER Ag. workshop begins
So I'm going to present the latest version of our paper on world supply and demand of food commodities at the NBER Friday morning. The workshop starts tomorrow and I'm flying out at the wee hour of 6am.
Here's a link to the schedule and papers that will be presented:
http://www.nber.org/~confer/2010/AGs10/program.html
My paper with Wolfram Schlenker is here.
For the little world of agricultural economics, I think a lot of the big players are represented. I feel lucky to be talking to this audience, especially on a big-picture ag. econ. topic.
So Wolfram Schlenker and I have presented this paper a number of times at most of the top ag. econ departments and keep revising to answer questions we get. But I remain a little surprised by some of the largest sources of confusion.
In a nutshell, here's are basic ideas (a bit wonkish):
Weather is a good instrument for identifying both demand and supply of agricultural commodities. The story for demand is well known and goes back to the birth of instrumental variables estimation: weather causes exogenous shifts in supply that facilitate identification of demand. Surprisingly, however, agricultural economists simply haven't made use of this rather obvious instrument. Because much of the weather shock is buffered by storage, to estimate demand we need to regress *quantity consumed* on price, which is instrumented with weather shocks.
For supply it is a little more subtle: *past* weather shocks cause an exogenous shift in inventories, which cause prices to rise in subsequent years, and thus a supply response to anticipated higher prices. We therefore regress quantity produced on the futures prices, instrumented with *past* weather shocks.
I think the basic idea is solid. Of course, the devil is in the details, and we need to make some compromises when it comes to implementing the basic idea. These compromises do raise some understandable questions and we try to address these as best we can.
But the big challenge in presenting this paper is getting the basic ideas across. This has kind of surprised me. Then again, sometimes when you're really close to something you forget that what's become obvious to you may not be so obvious to others the first time they see it.
Here are the key challenges:
(1) A lot of agricultural economists do not understand why futures prices need to be instrumented in the supply equation. These economists don't seem to realize that anticipated future supply influences anticipated future price. I wonder what these economists think is in the error of their supply equations. To make inference valid, they would have to assume that the error in all their supply regressions is 100% measurement error in the dependent variable. This seems implausible to me. Anyway, at least since early Clark medalist Marc Nerlove, it seems that no one in the huge literature on agricultural supply response has worried about the fundamental endogeneity of futures prices.
An example can illustrate this idea. Consider what happened in the 2005 growing season, just after the arrival of soybean rust to the United States. There was worry about what kind of damage or additional costs would stem from the new pest. So farmers planted more corn and less soybeans. At the same time, futures prices for corn went down and soybean prices went up, all long before planting time. The price and quantity movements were along the demand curves, not the supply curves. If the econometrician then regresses quantity on futures price one gets negative bias. Futures prices are still endogenous to supply.
In fact, I think what we're doing is a natural extension of Nerlove. I suspect Nerlove saw that much of the short term price-quantity fluctuations were weather (i.e., supply shifts) which created movement along the demand curve. To estimate supply he wanted to purge that variation along the demand curve, so he instead examined planted acreage in relation to expected price. This didn't fully solve the problem of price endogeneity, but it went a long ways in that direction. I was sad Nerlove wasn't able to make it to my seminar when I presented this at Maryland... so it goes.
(2) A lot of agricultural economists do not understand why past weather shocks affect current prices. This connection comes from any of a number of classic storage models: storage smoothes production shocks over time, causing prices to fluctuate less than they would without storage, but also causing transmission of a shock in one year onto prices in future years. So if we have a really bad year, we will draw down inventories to buffer much of the shock. But prices will stay high for a few years until inventories are replenished. Thus, past weather shocks (which are exogenous) affect future expected prices, allowing us to identify supply separate from demand.
Anyway, that's the basic idea. If you're interested in the gory details check out the paper.
Here's a link to the schedule and papers that will be presented:
http://www.nber.org/~confer/2010/AGs10/program.html
My paper with Wolfram Schlenker is here.
For the little world of agricultural economics, I think a lot of the big players are represented. I feel lucky to be talking to this audience, especially on a big-picture ag. econ. topic.
So Wolfram Schlenker and I have presented this paper a number of times at most of the top ag. econ departments and keep revising to answer questions we get. But I remain a little surprised by some of the largest sources of confusion.
In a nutshell, here's are basic ideas (a bit wonkish):
Weather is a good instrument for identifying both demand and supply of agricultural commodities. The story for demand is well known and goes back to the birth of instrumental variables estimation: weather causes exogenous shifts in supply that facilitate identification of demand. Surprisingly, however, agricultural economists simply haven't made use of this rather obvious instrument. Because much of the weather shock is buffered by storage, to estimate demand we need to regress *quantity consumed* on price, which is instrumented with weather shocks.
For supply it is a little more subtle: *past* weather shocks cause an exogenous shift in inventories, which cause prices to rise in subsequent years, and thus a supply response to anticipated higher prices. We therefore regress quantity produced on the futures prices, instrumented with *past* weather shocks.
I think the basic idea is solid. Of course, the devil is in the details, and we need to make some compromises when it comes to implementing the basic idea. These compromises do raise some understandable questions and we try to address these as best we can.
But the big challenge in presenting this paper is getting the basic ideas across. This has kind of surprised me. Then again, sometimes when you're really close to something you forget that what's become obvious to you may not be so obvious to others the first time they see it.
Here are the key challenges:
(1) A lot of agricultural economists do not understand why futures prices need to be instrumented in the supply equation. These economists don't seem to realize that anticipated future supply influences anticipated future price. I wonder what these economists think is in the error of their supply equations. To make inference valid, they would have to assume that the error in all their supply regressions is 100% measurement error in the dependent variable. This seems implausible to me. Anyway, at least since early Clark medalist Marc Nerlove, it seems that no one in the huge literature on agricultural supply response has worried about the fundamental endogeneity of futures prices.
An example can illustrate this idea. Consider what happened in the 2005 growing season, just after the arrival of soybean rust to the United States. There was worry about what kind of damage or additional costs would stem from the new pest. So farmers planted more corn and less soybeans. At the same time, futures prices for corn went down and soybean prices went up, all long before planting time. The price and quantity movements were along the demand curves, not the supply curves. If the econometrician then regresses quantity on futures price one gets negative bias. Futures prices are still endogenous to supply.
In fact, I think what we're doing is a natural extension of Nerlove. I suspect Nerlove saw that much of the short term price-quantity fluctuations were weather (i.e., supply shifts) which created movement along the demand curve. To estimate supply he wanted to purge that variation along the demand curve, so he instead examined planted acreage in relation to expected price. This didn't fully solve the problem of price endogeneity, but it went a long ways in that direction. I was sad Nerlove wasn't able to make it to my seminar when I presented this at Maryland... so it goes.
(2) A lot of agricultural economists do not understand why past weather shocks affect current prices. This connection comes from any of a number of classic storage models: storage smoothes production shocks over time, causing prices to fluctuate less than they would without storage, but also causing transmission of a shock in one year onto prices in future years. So if we have a really bad year, we will draw down inventories to buffer much of the shock. But prices will stay high for a few years until inventories are replenished. Thus, past weather shocks (which are exogenous) affect future expected prices, allowing us to identify supply separate from demand.
Anyway, that's the basic idea. If you're interested in the gory details check out the paper.
Great post and congrats on presenting at the NBER conference. A critical comment, however: I don't know what sorts of questions you've received at seminars, but I'm surprised to hear that ag. economists aren't aware of the usefulness of yield variation as a tool to identify supply and demand from price and quantity data. At least, ag. economists who do price analysis certainly realize this. You don't have to go all the way back to Nerlove to see this idea productively employed. A basic reference: Tomek and Robinson "Agricultural Product Prices" (chapter 14, 4th edition).
ReplyDeleteFurthermore, ag. economists have also taken advantage of yield variation to calibrate and validate a computable general equilibrium model. An example from my own work:
Valenzuela, Hertel, Keeney, and Reimer. “Assessing Global Computable General Equilibrium Model Validity Using Agricultural Price Volatility” AJAE 2007.
Jeff Reimer
Jeff,
ReplyDeleteThanks for the reference--that seems like a good way to calibrate a general equilibrium model. But it's also an area of the literature where I wasn't looking. I'll definitely have a look.