Predict Carry in Futures Contracts With Gro's Supply, Demand Data

Talk to our our team about Gro's offering
Talk to our team

A widening or narrowing in the price spread between futures contracts can be caused by a host of variables, as we described in a related Insight article.

This so-called calendar spread, which can be calculated using the Gro Intelligence API (see the related Insight to learn how), is used regularly by grain handlers and food companies to manage risk in their grain sales and purchases. Calendar spreads in which the price of the later contract is greater than that of the earlier contract is called a carry, while the opposite is called an inverse. Predicting the value of a calendar spread using available data and forecast supply and demand allows a market participant to take advantage of the current carry structure of the futures market by making a direct bet on the spread in the contracts or by shifting the timing of physical grain trading.

In this Insight we walk through an example analysis of soybean supply and demand metrics to predict the spread between the July and November CME futures contracts. Users can adapt this analysis to other commodities and time periods that comprise the carry structure of a futures market to actively bet on a calendar spread or plan out business decisions over the coming months.

The July soybean futures contract is closely associated with the Brazilian crop. Brazil harvests the majority of its crop in March and export volume peaks from April to June. By contrast, the November soybean contract is the reference contract for the US crop. It is used to set the harvest price for revenue insurance and coincides with the peak of US exports. The spread between the two contracts is often driven by the relative prospects of soybean supplies from these two major exporters.

This analysis will look at the average price spread over the month of June for the last 13 years. In June, the July and November contracts are the most actively traded.

Three independent variables need to be constructed from data available to users of either the Gro API or the Gro web app: the US soybean ending stocks-to-use ratio, the Brazil ending stocks-to-use ratio, and the Brazilian soybean yield as a deviation from trend. Since we are trying to predict the value of the calendar spread in the month of June it is important that we use the variables as they were in June of each year in the backtest.

To determine these variables, we use Gro’s “Show data revisions” feature in the Edit Chart tool, found under “Configuration” for a line chart. Create charts for the following metrics: Stocks, Ending Quantity (mass); Domestic Consumption (mass); and Export Volume (mass). “Soybeans” should be selected as the item in the US and “Soybeans, local” in Brazil. For each chart, choose USDA PS&D as the source and make sure “Show data revisions” is selected under “Configuration.” Saving those options will visualize the monthly revisions available on Gro.

The June data point for each year back to 2007 is data from the June WASDE reports, which represent the USDA’s forecast for each of those metrics at the time. To calculate the stocks-to-use variable you divide Stocks, Ending Quantity (mass) by the sum of Domestic Consumption (mass) and Export Volume (mass) for both countries. Stocks-to-use is a frequently used measure of available supply and thus an important factor in futures prices.

For the Brazil soybean yield variable, take Yield (mass/area) for Soybeans provided by IBGE for 1974 to 2019. A linear trend over that yield history can be used to calculate the deviation from trend in each of the years from 2007 to 2019 as a ratio of actual yield over trend minus 1. This variable measures how the Brazil soybean crop performed compared to expectations in a normal weather scenario. The size of the Brazil crop has a direct impact on global trade flows and therefore the forecast of supply and demand balance for the US.

These three variables together are then regressed against the average price of the July-November calendar spread in June. The result is an R-squared of 0.43 or, in other words, 43% of the variation of the price is explained by these three variables.

This exercise demonstrates the ability of supply and demand data to predict carry in the futures market. Gro’s Brazil soybean yield model provides daily in-season forecasts, and previous Gro Insights have outlined ways to create a supply and demand balance sheet. Other variables can be applied to improve predictive power, and this analysis can be performed for different commodities and across the spectrum of calendar spreads.

Above are two of the key variables that enable Gro users to predict the spread between futures contract prices, an important risk-management practice. The chart on the left shows Stocks, Ending Quantity (mass) for soybeans in the US with “Show data revisions” enabled. Each line represents the monthly progression of forecasts for a given crop year from the USDA’s WASDE report. The forecast changes significantly over the season. Looking at prior revisions offers a snapshot of the market perception for a given supply or demand metric. The chart on the right is Brazil’s soybean yield history from IBGE including the 2020 forecast. IBGE provides yield, production, and area down to the district level across many commodities. Click on the image to interact with the charts on the Gro web app.

This insight was powered by the Gro platform, which enables better and faster decisions about factors affecting the entire global agricultural ecosystem. Gro organizes over 40,000 datasets from sources around the world into a unified ontology, which allows users to derive valuable insights such as this one. You can explore the data available on Gro with a free account, or please get in touch if you would like to learn more about a specific crop, region, or business issue.

Get a demo of Gro
Talk to our enterprise sales team or walk through our platform