Investment portfolios of insurance companies usually consist of traditional investments, the largest share of which is dominated by bonds. Factors that affect the returns of these financial instruments refer primarily to the macroeconomic environment and the level of interest rates on the financial markets. Most of the analyzed period of this research is characterized by an environment of historically low interest rates, which led to a significant decrease in the return of insurers' investments. The main goal of this doctoral dissertation was to analyze the impact of including low-correlated alternative assets in an investment portfolio, which exclusively consists of traditional assets, as a solution to the problem of low interest rates in the analyzed period. Alternative investments, due to their low correlation with the returns of traditional asset classes, can have a major positive effect on risk diversification and attaining higher investment returns. Traditional assets refer to Croatian bonds, stocks and money market instruments, while alternative assets are represented by hedge funds and managed futures. Hypothesis about the benefits of including alternative investments were tested using the Markowitz and multicriteria optimization models. The test results in the analyzed period, according to both models, confirm the hypotheses about the positive impact of the inclusion of alternative assets on return, risk, and the efficiency of the investment portfolio measured by the Sharpe, Sortino and Omega ratio. In Hypothesis 1, the aim was to test whether the inclusion of alternative assets in the investment portfolio, which only contains traditional assets on the Croatian financial market (bonds, shares, money market instruments), has a stronger positive impact than the inclusion of additional foreign traditional assets (e.g., EU, USA, emerging markets bonds and stocks). The results of testing the hypothesis H1 in the analyzed period indicated that alternative assets have a stronger positive impact on the efficiency of the portfolio measured by the Sharpe and Sortino ratio in almost all tested combinations of weights of the observed assets. Hypothesis H2 tested whether the use of a multicriteria optimization model would give a more efficient portfolio compared to the use of the Markowitz model. With the aim of representative comparison of simulated and optimized portfolios, equal weights restrictions were used for the included assets classes. The Omega ratio was used as an efficiency measure for portfolio evaluation since, in addition to expected return and variance, it considers higher portfolio moments related to the coefficients of asymmetry and kurtosis, which are also optimized in a multicriteria model. A portfolio with a higher Omega measure is considered more efficient. The results of testing H2 also confirm the hypothesis, which enables better creation and selection of available assets in the investment portfolio. Findings show the benefit of the multicriteria model, which is coded in the R programming language, and that further facilitates the application to interested investors and other stakeholders. As previously mentioned, this scientific paper also presents the characteristics of the multicriteria optimization model where conflict of objective function occurs. In this paper, the model is used to optimize central portfolio moments, respectively expected return, variance and coefficients of asymmetry and kurtosis which are often referred to as higher central moments. In other words, the return distribution first and third moment (return and asymmetry) are maximized and the second and fourth moment (variance and kurtosis) are minimized simultaneously. The use of the multicriteria model brings several advantages. First, investors can arbitrarily choose preferences regarding certain central moments in the optimization itself, which then affects the higher or lower value of these moments depending on the preference. Multicriteria optimization is suitable for use in situations where time series of the observed variables do not follow the shape of a normal distribution, which makes it possible to overcome the normality assumption from the Markowitz optimization model. Numerous empirical papers have already pointed out that time series return distributions of many traditional and alternative assets do not hold the normality assumption. In addition to previously mentioned, using the unrestricted vector autoregression model (VAR model), the hypothesis H3 about the strength of interest rate changes impact on the returns of domestic stocks and bonds, as well as hedge funds and managed futures, was tested. The theoretical assumption that forms the basis of H3 is that the change in the level of interest rates, regardless of whether they are rising or falling, has an expected stronger effect on the change in bond and stock returns than on the returns of hedge funds and managed futures. Accordingly, investors whose investment portfolios include bonds and stocks can reduce exposure to the risk of interest rate changes through the inclusion of hedge funds and futures, i.e., reducing the share of bonds and stocks. Despite the theoretical assumption of a stronger impact of the change in interest rates on the change in the return of bonds and stocks, it was not possible to fully confirm the hypothesis, nor to reject it. All calculations and optimizations in hypothesis testing were performed in R programming language.