Recently developed techniques are adapted and combined for the modeling and simulation of crop yields and prices that can be mutually correlated, exhibit heteroskedasticity or autocorrelation, and follow nonnormal probability density functions. The techniques are applied to the modeling and simulation of probability distribution functions for the returns of three tropical agroforestry systems for coffee production. The importance of using distribution functions that can more closely reflect the statistical behavior of yields and prices for risk analysis is discussed and illustrated.