Estimating Input Cost Shares for Agriculture Using a Multinomial Logit Framework

Many econometric analyses include dependent variables constrained to the interval between zero and 1 Under such conditions, simple regression procedures break down Several alternative stochastic models which avoid this problem can be defined depending on the assumed error structure Two alternative forms of the logit model are treated here The multivariate logit approach assumes that the share specification is an accurate representation of the underlying input demand structure The multinomial logit approach treats the dependent variable as a probability with a multinomial density