Five questions to ask About a Farm Bill Decision Aid

We have been modeling crop insurance and farm policy for years.  Tremendous advances have been made in quantifying agricultural risk. As farmers face decisions regarding their participation in federal farm programs and crop insurance various decision aides have been developed to evaluate alternatives.  Based on our experience, here are five questions to ask anyone who tells you they have a decision aide for evaluating the ARC/PLC choice.

  • How does the decision aide account for uncertain prices and yields over the life of the bill?

Most spreadsheet aides are simply calculators, meaning they are ‘deterministic’ in that they calculate a payment based on the exact yields and prices provided. The problem, of course, is that one can’t possibly know with certainty what yields and prices will occur. How does the decision aid account for the likelihood of various prices and yields over the next 5 years when estimating payments?

  • If the decision aid accounts for risk, what risks are modeled?

There are five major unknown variables that must be accounted for in any  crop insurance, ARC, and/or PLC decision aide.  These are: three prices – cash prices, futures market prices, market year average prices, and two yields – farm and county yield.  Does the decision aid account for the likelihood of different outcomes for all of these unknown variables?

  • If the decision aide accounts for risk, then how is the correlation of random variables handled?

These five unknown variables are not necessarily independent, meaning there is a relationship (or correlation) between them.  In fact, there is good reason to believe that many of them are related.  For example, farm and county yield are most likely positively correlated.  In the Midwest, yield and price for corn likely have a negative relationship (as yield declines, corn price would increase).  Cash, futures, and MYA price are likely positively correlated.  Prices and yields across years are also often positively correlated (trends develop over time).  There are more relationships, for example: a farm considering individual ARC with three crops potentially needs to account for 120 correlations.  Modelling correlation is difficult, but very important and shouldn’t be avoided to accurately assess the farm program and crop insurance options.

  • Does the model ask you for a lot of farm yield data?

Nobel Prize winner Daniel Kahneman points out the problem of using only a few years of data to form expectations often provides faulty outcomes.  Our research suggests that evaluations of farm-level crop insurance and farm program outcomes with less than ten years of farm yield data will be highly inaccurate.

  • Does the decision aid help you understand risk protections as well as expected returns?

The new programs offered from the 2014 farm bill are intended to help farms reduce exposure to the risks of low price, low yield, or low revenue.  Simply reporting the ‘deterministic’ expected payments – payments that come from one price and one yield – from the programs ignores the question of whether the payments help mitigate risk exposure. In other words, how does the farm program and crop insurance decision fit into the entire operation’s business portfolio?

In summary, predicting the future is extremely difficult. However, methods to provide guidance with respect to the uncertainty and correlation amongst the multitude of possible outcomes do exist but are often difficult to apply. Some of these are built into the current offering of decision aids provided by Texas A&M and Illinois, but few are available in simple spreadsheets built by others. For example, the two spreadsheets we have provided (CLICK HERE) only give the base reallocation calculation and the calculation of how generic acres will be distributed based on a given number of planted acres, both of which are simple calculators. While these types of “decision aids” can be very useful, keep the questions we pose here in mind as you evaluate the results generated from them.