How to estimate changes in seed purchase behaviour of soybean farmers in Brazil
Kleffmann Group developed a forecast model of high accuracy based on own panel data
As a leading provider of agricultural market research information, Kleffmann Group has a great treasure of data at its disposal. These form the basis for the development of meaningful models that make it possible to predict changes in farmers’ purchasing behaviour.
A farmer’s decision on which type of seeds to plant is usually driven by several distinct input factors, such as seed prices, herbicide, insecticide and fungicide costs, local commodity prices, labour costs or machinery costs. Kleffmann Group has used these explanatory variables to estimate their impact on the selection of seed types and to predict the share of different seed types with respect to the planted area.
Kleffmann used a discrete choice model to analyse the probability with which a soybean farmer chooses one particular seed type over the others, based on the aforementioned input variables. As one would, expect the probability for selecting one specific seed type is negatively correlated to its price and particularly genetically modified traits show a high elasticity in this regard. If one raises the prices for stacked herbicide tolerant & insect-resistant traits, for instance, and keeps all other prices fixed, farmers tend to mainly select herbicide tolerant seeds as an alternative.
Commodity prices have a high effect on the choice of seed type
The impact of crop protection costs on the choice of seed type is not so clear, but estimations show that higher herbicide and fungicide costs increase the probability that farmers plant stacked herbicide tolerant & insect-resistant traits. Another major driving factor that affects the choice of seed type are commodity prices. Higher local soybean prices favour the selection of more expensive genetically modified traits and also farm saved seeds, while the probability of choosing commercial seeds with no traits decreases significantly.
Approach allows to predict the share of planted areas by seed
The discrete choice approach allows not only to estimate the probability that an individual farmer plants a distinct seed type according to his input characteristics, but also to predict the share of planted areas by seed type on a state level.
For example, Figure 1 shows the actual shares of soybean seed types used in Mato Grosso. The dark shaded bars are based on Kleffmann end-user panel information and the light shaded bars on the estimated shares were derived by aggregating individual information to the state level. The chart illustrates the accuracy of estimations with respect to historical data. Forecasts for the development of seed type usage and adoption of genetically modified traits can easily be generated out of this, when information or certain expectations about the evolution of explanatory variables are available and the estimated model is applied to this new data.
Nowadays, genetically modified seeds play a dominant role in Brazilian soybean production in general and its largest soybean producing state of Mato Grosso in particular. After the commercial introduction of the first herbicide tolerant soybeans in 2005, genetically modified traits, which include both herbicide tolerant traits and stacked herbicide tolerant & insect-resistant traits, have continuously gained market share and account for more than 75 % of planted soybean area in Mato Grosso today. At the same time, the share of commercial soybean seeds with no traits declined and dropped from 90 % in 2005 to less than 10 % in 2019. Surprisingly, the importance of farm saved seeds has increased in recent years as well and now represents almost 20 % of the planted area.
Dr. Stefan Kersting, Data Analyst & Project Manager Forecasting – amis® AgriGlobe®