An introduction to building models forecasting the prices of grain and grain products

January 2020 | Przegląd Zbożowo-Młynarski (Grain and Milling Review)

I dedicate this article to entrepreneurs engaged in the purchase and storage of grain. This activity rests to an enormous degree on predictions of the future prices of grain and of grain-processing products. Writing about grain, I have in mind above all wheat, barley and rye, as well as triticale and oats.

In the following articles I will show how to build simple forecasting models which will make it possible to achieve a competitive advantage in the area of trading in agricultural raw materials. These models will be built thanks to free tools available on the internet. These free programs and scripts will be among the most modern, fully professional tools used by Data Science experts.

It should be emphasised that raw material markets differ significantly from currency markets or capital markets. At the beginning let us present the most important features of the market in grain and grain products.

W-MOSZCZYNSKI pzm 1-20

The difference between raw material markets and capital markets

1. Raw material markets, including the grain market, are only to a small degree connected with indicators of the state of the economy such as the level of inflation, GDP growth or the level of debt. It is different in the case of currency and capital exchanges. There the state of the economy influences the level of quotations dynamically. For speculative investors, the limited connections of the grain market with the economy and with capital market indices are an opportunity to diversify their investment portfolios in order to reduce the level of risk of the investments they conduct.

2. The greatest influence on raw material markets is exerted by the entities which use the raw materials quoted on those markets. Grain and grain raw materials are directly used in food processing. The price is therefore to an enormous extent a reflection of the economy’s demand for particular grain products. Of course speculative investors also operate on the grain market, but their influence on the raw material market is far smaller than on other exchanges.

3. A further significant element which differentiates agricultural raw material markets from capital markets is the absence of the influence of confidential information. According to the theory of the perfectly competitive market, prices are shaped by the participants of the market game, who possess full market information. Capital markets are never perfectly competitive because of the existence of „insiders”, that is, persons possessing particular knowledge inaccessible to others by reason of their employment in the companies whose securities are the subject of trading. Employees working in companies know better than others what the true financial condition of their company is and what opportunities and threats to development exist. There is no such phenomenon on the agricultural raw materials market. Here every participant in the market game has access to the same information. The appearance of a drought or of a plague will obviously increase grain prices, but this will not be a phenomenon difficult to explain.

4. Capital markets are to a greater degree subject to the phenomenon of the „investment bubble”. There prices are shaped by the often irrational, emotional behaviour of investors. The speed of transactions and the sensitivity to information make the capital market much more difficult to forecast compared with stable raw material markets. Grain markets, compared with currency markets and capital markets, react much more slowly to changes in the economy. A crash on the agricultural raw materials exchange (understood as a fall of several dozen per cent in the value of the raw materials) is practically impossible to occur. A sudden, abrupt rise in the prices of grain raw materials may happen in the case of some sudden climatic cataclysm or a war. The slow reactions of the market are a factor facilitating the building of effective forecasting models.

5. Raw material markets differ from capital markets in the regularity and durability of their cycles and trends. The regular cycles and trends of the market in grain and grain raw materials are closely linked with the calendar of the seasons and of the harvests. An additional element which evens out sudden seasonal fluctuations in grain prices is the storage sector (grain elevators). Clear cyclicality makes it possible to work out precise and more effective forecasting models.

What is a forecasting model?

A forecasting model is an equation which takes the sought value as its result. When we build a model forecasting the price of rye, the result of the equation will be the future price of rye.

For the model, that is, the forecasting equation, to work, it must have explanatory variables. For example, a model calculating the volume of bread from flour will have a describing variable in the form of the number of cm³ of flour, and the result will be the number of cm³ of bread. It is not hard to guess that in this case the forecasting model will be the recipe for baking bread.

Now we will get to know examples of commonly known variables which influence the price of grain. Market players, building forecasting models, use a variety of variables, often seemingly unconnected with the outcome variable. They also use various indicators which are combinations of the proportions of different variables. Which describing variables should be introduced into the model depends on experience, on the goal and on the scope of the model.

Below I present example variables which we will use in the following publications when building a simple linear regression model.

Describing variables for a forecasting model of grain prices

The monthly report of the United States Department of Agriculture (USDA). The publication of this department has a great influence on world grain prices, including grain prices in Poland. It verifies the forecasts of the volume of production. When the report indicates high grain harvests, higher than those predicted by the market, prices on the market will fall. It is not hard to notice that those who created the most accurate forecast of the size of the harvest will gain.

The cycles of the seasons. In the case of agricultural raw materials there is a clear, constant cyclicality of prices resulting from the seasons. In the case of wheat, over the last dozen or so years a regular fall in prices has been observed beginning at the start of March. This trend is maintained until the harvest, when prices are at their lowest in the year. From August the price of wheat rises until autumn, when a further fall in prices occurs again. The very strong and at the same time repeating cyclicality on the wheat market, recurring for over a dozen years, may be the only explanatory variable in the simplest linear regression model. Thanks to the repeatability of the cycles this model may turn out to be accurate in its forecasts.

Weather forecasts. It is commonly known that the best weather models are held by American government centres. That is why their long-term weather forecasts have the greatest influence on raw material markets, including grain markets. The size of the yields depends above all on the amount of precipitation in April and May. The forecast period of the beginning of the growing season, closely linked with temperature, has, in the opinion of experts, secondary importance. If in the published weather forecast for April and May the threat of drought appears, world wheat prices will rise and will possibly change after the publication of the department’s report on the current size of the harvest.

Political information. Significant, but not the most important, factors shaping the prices of wheat and grain are announcements of political changes. Unfortunately it is difficult to convert into numerical values an announcement of the introduction of an embargo on grain supplies to some region. Forecasting models use exclusively numerical variables.

There are a great many various variables which may turn out to be significant when building forecasting models. The appropriate selection of variables may contribute to building a model which will contribute to generating high profits on the raw materials market. In the process of creating forecasting models the greatest importance is held by knowledge and feel for the market. The technological and statistical skills of the people creating the models are not of as great importance as acquired experience and knowledge of the market.

Wojciech Moszczyński

Wojciech Moszczyński — graduate of the Department of Econometrics and Statistics of Nicolaus Copernicus University in Toruń; specialist in econometrics, finance, data science, and management accounting. He specializes in the optimization of production and logistics processes. He conducts research in the area of the development and application of artificial intelligence. For years he has been engaged in the popularization of machine learning and data science in business environments.