Numerous approaches have been developed to model and forecast basis. These range from relatively straightforward
methods, such as moving averages of previous basis, to complex statistical models. Regardless of the approach, the
foundation for nearly every modeling strategy is information. And better information leads to better basis predictions.
In addition to having good information, predicting basis accurately also requires that only the relevant information
is used. Behavior of wheat basis in the Northern Great Plains (NGP) region is unique. For example, much of the commodity
is shipped to ports in the Pacific Northwest for export, protein content considerations are critical, and harvests
occur several weeks after wheat is harvested in the U.S. Midwest. These and other factors play important roles in
affecting local basis.
In modeling basis in NGP markets, the following factors were considered:
- Wheat class (hard red spring, HRS; hard red winter, HRW)
- Wheat quality (protein content)
- Geographic differences
- Futures prices
- Futures price volatility
- Spread between HRS and HRW wheat futures prices
- Basis in previous periods
- Seasonal fluctuations
To most accurately use this information, several criteria were used to evaluate the ability of a particular
statistical model to predict basis. First, each model was assessed based on its capacity to predict basis
values that already occurred (in-sample predictions). Second, models were graded on their accuracy of forecasting
future harvest-time basis (out-of-sample predictions). The statistical models that minimized the prediction
errors were then chosen. Greater emphasis was placed on selecting models that made the fewest errors in forecasting
future basis values.
The full methodology and description of the model was recently published in a peer-reviewed academic outlet, the Journal of Agricultural and Resource Economics.
You can access the paper here: "Forecasting a Moving Target: The Roles of Quality and Timing for Determining Northern U.S. Wheat Basis."
Although the selected models are those that currently provide the best basis forecast, it is likely that as market
conditions change and new information becomes available, these models will become less effective. Consequently,
NGP basis models will periodically be re-evaluated to determine how the information has impacted basis forecasting
accuracy. The outdated models will then be updated.
The map below provides a view of the grain handling facilities for which basis forecasts can be obtained.
Note: Locations shown are approximate and represent the centroid of the ZIP
code where a facility is located.
Use for reference purposes only.