The paper presents the properties of the rates of return distributions for Markowitz models and models with minimum semivariance. The special focus was placed on investigating the variation over time of the rates of return distributions for the studied portfolios. Non-parametric Kolmogorov-Smirnov tests and augmented Dickey-Fuller test were used for analysis of distributions over time. The studies showed that the distributions of rates of return for portfolios developed, particularly for high assumed rates of return were characterized by high variation. Considering selected distribution parameters SEM portfolios were more favorable than Markowitz portfolios although they showed a higher variation of distributions over time.