Multiple herbicide use in cropland: A discrete-continuous model for stated choice data

Using a random utility consistent discrete-continuous model we analyze decisions by a sample of Australian farmers on the mix of herbicides and their intensity of use. The proposed model is flexible and accommodates even a single herbicide type discrete-continuous decision. We use structural estimates to forecast how the demand for herbicide brands changes as their prices change. To do so, we develop a forecasting algorithm that forecasts optimal herbicide types and allocations via a Sequential Quadratic Programming approach. This novel forecasting approach allows for calibration of alternative specific constants so as to reproduce brand market shares.