Fundamentally Speaking

By Joel Karlin
DTN Contributing Analyst

At the USDA Ag Outlook conference last month, their economists elaborated on their new weather adjusted yield model.

This has generated quite a bit of buzz given their 2013 U.S. yield forecasts of 163.6 bushels per acre (bpa) for corn, the second highest ever next to the 164.7 bpa record in 2009 and an all-time high for soybeans at 44.5 bpa.

For corn, the USDA uses a weather adjusted trend yield where the dependent variable yields is derived by taking a number of explanatory variables and running them through a multiple regression model.

Using data from 1988 to 2012 the USDA includes a trend variable to account for yields trending higher over time due to improvements in technology such as new seed hybrids, better pest management, more accurate seed dispersal, and enhanced fertilizer application that have helped support greater plant populations per acre.

In addition to the trend variable, the USDA uses a number of weather related variables including mid-May planting progress, June precipitation only in years when it is quite dry, and July temperatures and precipitation.

What is interesting is that while the impact of mid-May plantings, June precipitation, and July temperatures are linear or each unit of change has a constant effect on yields, the effect of July precipitation has a non-linear impact.

There is an asymmetric response of corn yields to different amounts of precipitation above and below its average.

What this means is the reduction in corn yields when rainfall is below average are much larger than gains in corn yields when rainfall is above average.

A note within the paper says that the model uses a squared term for July precipitation to represent this asymmetric effect.

The accompanying graphic tries to replicate this effect by plotting the percent that final Illinois corn yields deviated from the 1960-2012 trend plotted vs. the amount of July rainfall that season as measured by the difference from the 50-year average of 3.97 inches.

As an example, the worst year saw the yield 41.9% below trend with July rainfall that season 2.60 inches or 1.37 below the long-term average.

The greatest moisture deficiency was a deficit of 2.49 inches below average resulting in the second largest negative deviation from trend of 37.5%.

Note that positive yield deviations do not exceed 17/% yet there are times when negative yield deviations top 20%, 30%, and even 40%.

We have used a quadratic trend-line for the percent deviation from yields to account for this asymmetric effect.

(KA)