Because participation decisions are reflected by discrete outcomes in two stages (e.g., use versus nonuse of insurance; and then choice among products in the case of use), a binomial logit model is used to evaluate the participation decision and an unordered multinomial logitmodel is estimated sequentially in the second stage across those purchasing insurance. In the second stage, the dependent variable is composed of three discrete choice hail insurance, yield insurance, or revenue insurance. In a multinomial logit model, one case is designated as the base case and the influences of the factors are expressed relative to their influence in the base case (Greene).
Use of hail only was designated as the base case in stage 2 of the model. Significant coefficients indicate the extent to which a variable contributes to the probability of selecting that choice relative to its influence in the base case.
The probability that a farmer’s product choice, y, results in product j,is defined as: