Tuesday, April 10, 2012
I'm not really sure where I'm going with this, but it took a long time to make so I'm inflicting it on readers regardless!
The above shows, for a sample of thirty counties in Illinois, the departure of the corn yield from its eleven year moving average in that county. (Data: NASS) Obviously corn yields, in Illinois as elsewhere, have been increasing as a result of technology - but we can use the departure from the moving average to measure the degree to which the circumstances of a particular year were markedly better or worse than the average. I picked Illinois for no better reason than being a fairly representative midwestern state that I happened to visit recently.
You can see that there is substantial covariance - a bad year tends to be bad for most counties, while vice versa in a good year. However, there is not complete covariance - for example 1954 was a very bad year, but if we look at the cumulative fraction of all 102 counties with a particular yield fraction (relative to the moving average) we get the following (blue line):
It was pretty bad in about half the state - some counties had less than a quarter of the normal yield - but the other half was within 10% either way of the moving average. 1934 and 1983 were more consistently bad - but even in 1983 the fraction of the moving average crop yield ranged from in the 25% range in the worst hit counties to over 80% in the least hit counties. In short, the scale of even the most consistent crop failure in recorded history is not so much larger than the state of Illinois that it pretty much hits the whole state equally. And most crop failures left some portion of Illinois untouched.
If I get the time to do some more heavy-duty number crunching I'm interested in understanding the spatial structure of yield failures like this, and whether there's any way to detect any signature of recent changes in the weather patterns causing more extensive or more correlated yield failures.