The two equations can each be estimated consistently with single-equation probit method. However, such a commonly used approach is inefficient because it ignores the correlation between the error terms εdv and εmv of the underlying stochastic utility function of a diversified cropping system and modern rice monoculture, respectively (Greene, 2003). We apply the bivariate probit model in order to circumvent this limitation. The bivariate probit model is based on the joint distribution of the two normally distributed variables and is specified as (Greene, 2003):