One of the most basic tasks in the supply/demand analysis for grains at this time of year is the estimation of trend yields. You would think that something as simple as running a straight line through yield data and projecting one year ahead would be a no brainer. It turns out that a lot of ink has been spilled on the best way to estimate trend yields. I have contributed some of that spilled ink.
A question that has come up over and over in my experience is the impact of changing geographic patterns of acreage on trend yield estimation. For example, more corn acres have moved into the Dakotas and northern Minnesota in the last couple of decades. Since yields are lower in those areas than in the heart of the Corn Belt, wouldn’t that lower trend yields? Directionally, we know that has to be the case, but is the impact so small it can be safely ignored or do we have to figure out an adjustment? Important new evidence on this question was recently supplied by my colleagues in a FDD article last week: “Impacts of Corn Acreage Increases on National Corn Yields.” Impacts of Corn Acreage Increases on National Corn Yields – farmdoc daily (illinois.edu)
They find that the changing geographic composition of corn acreage in the US has an impact on corn trend yield estimation. Whether you think the impact is big or small depends on the length of time you use to estimate trend yield. If you use a sample that is 20 years or so, your trend coefficient could be as much as 0.2 bpa too low. For longer samples, the bias is less than 0.1 bpa. Since I focus on longer samples, like 1980 forward, I find their results reassuring. There is an impact but it is small enough that I don’t need to lose sleep over it.
Lastly, if you are concerned about shifting corn acreage patterns on trend yield estimates there is really only one solution. You have to estimate trend yields at the sub-national level and then take an acre-weighted average of the estimates. In other words, you could estimate a trend yield for say each of the largest 20 producing states, and then take an acre-weighted average of the 20 state trend yield estimates to estimate a national average trend yield.