The objective of this research was to compare the method proposed by Markowitz (mean-variance) and the method proposed by Estrada (mean-semivariance), in the choice of an agricultural portfolio. The data were the returns of five agricultural products for the period 1980-2009; both the covariance matrix and semicovariance matrix were estimated to be used in either method. Later, a size n = 30 simulation of 100 replications was performed, to obtain the yields of each product, thus 100 solutions per method. For comparative purposes, a histogram was constructed; this was completed with the t test concluding that the average portfolio is the same under both methods.

....