Brand Inertia in Seed Demand: State Dependence and Heterogeneity

A commonly observed feature of differentiated product markets is brand inertia, the tendency of consumers to purchase brands they have purchased in the past. In this paper, we develop and estimate a micro-level random coefficients logit model to study two competing explanations of brand inertia, state dependence and heterogeneity, in the U.S. soybean seed industry. Specifically, heterogeneity is captured by brand-specific random coefficients and state dependence is incorporated through a brand purchase history variable. We further deal with two important issues in the identification: we apply a correction to the initial conditions problem similar to the procedure outlined in Wooldridge (2005); and to deal with price endogeneity, we use the control function approach in our nonlinear regression environment. The model is estimated using a large dataset of more than 200,000 seed purchase decisions by roughly 28,000 farmers over the period 1996-2016. We find that state dependence and heterogeneity are both important features of seed demand. On average, farmers are willing to pay (WTP) an additional $6.77/unit for a brand if it was purchased in the previous period, equivalent to about 15% of the average retail price. We also find that farmers are willing to pay large premiums for brand labels and the glyphosate tolerance (GT) technology, however there is considerable heterogeneity in this willingness. To investigate the implications of state dependence, especially as it relates to the introduction and diffusion of the GT innovation, we simulate several counterfactual scenarios (with/without state dependence and/or the GT technology). Our simulations show that state dependence has little effect on the diffusion of the GT technology, but it functions as a cushion for the structural benefits brought about by the innovation—it reduces the gains/losses in brands’ market shares. We also show that there is an “early adoption” advantage associated with the marketing of the GT trait, which is reinforced by state dependence.