Commodity price variability and supply/demand responsiveness (wonkish)

The other day I wrote about how a classic puzzle in the academic literature on commodity prices has apparently been solved in a nice paper by Cafiero et al.  The puzzle concerned an apparent excess autocorrelation in commodity prices and the solution was rather technical.  It turned out that approximation errors in the statistical calibration of that model, not the theory itself, led to the puzzle.  A better calibration by Cafiero and coauthors led to a model with much more inelastic demand that also fit the price data well.

But on a deeper level I think some puzzles remain. The model traditionally used in the commodity price/storage literature is one with a steep consumption demand curve and a vertical, perfectly inelastic supply curve that shifts horizontally with the weather.  Production shocks are buffered by a competitive storage market that buys and stores commodities when prices are low and sells when prices are high.

For explaining price behavior,  it is somewhat arbitrary whether it is demand, supply or both that respond to price.  The point is that both supply and demand must respond in a very inelastic manner else prices would not be so variable and not so autocorrelated.  Or at least that's my strong hunch.

This brings me to a totally separate literature on commodity supply response.  At least for agricultural commodities this literature is quite substantial and most supply elasticity estimates center around one, which means production doubles if prices double, all else the same.  If you look more closely at the literature, it's clear that supply elasticities bounce all over the place and in the most careful studies supply is far less elastic.  So maybe we shouldn't take that literature too seriously. (Sometimes I wonder if the elasticities center around one because that seems like a safe, referee-pleasing number.  Anyway...)

But what I haven't seen anyone articulate very clearly is a huge apparent disconnect between what the commodity price/storage literature has to say and what the supply literature has to say.  If the only way to get commodity prices to fluctuate as much as they do and to be as autocorrelated as they are is to have *both* supply and demand be very inelastic, then most of the supply response literature must be very wrong.  Or, if it's not wrong, and supply is even somewhat responsive to price, then the price puzzles of too much volatility and too much autocorrelation creep back into the picture.  Alternatively, one might make some strange assumptions about the costs of storage. But those assumptions would be so strange, I believe, that they would be broadly deemed implausible.

My own number crunching with Wolfram Schlenker suggests supply is in fact quite inelastic, but still about twice as responsive as demand.  Taken together, I still find the elasticities large enough that I have a hard time seeing how and why prices could be so volatile and autocorrelated in a well-functioning competitive market with storage.  But I haven't done a full-blown analysis to check.

The recent activity in wheat prices, and the crazy behavior of rice in 2008, has me thinking that uncertainty about policy--like export bans--is an important part of the picture.  These could build in a market expectation of a more inelastic supply and demand responses than we might otherwise expect.

Update: An important piece of the underlying intuition here is that the physical size of the shocks--physical quantity of stuff that actually pushes prices up or down--tends to be very small relative to the size the related price spikes. Take Russian wheat shortage, for example--it's really not that big of a shock relative to the world wheat market.  With storage buffering these tiny shocks, we just shouldn't be getting that much in the way of price movements.  But we are.  And I still don't think we in academia have a very clear explanation for why this is the case.