Methodology: Our research is based on a combination of methods applied in the contemporary game theory: axiomatic, algorithmic and strategical, on methods of probability theory such as stochastic geometry, random matrix theory and concentration of measure tools. We apply also stochastic optimization methods and combinatorics.
Empirical base of research: The research has a theoretical character.
Results of research:
For the class of divisible fair allocation problems some previous results of computer modelling have being obtained. An algorithm finding the Nash maxproduct solution was realized. With its help several hypothesis on its properties for an enough general case, when there are big number of objects to be allocated, and individual utilities are independent idential distributed random values with continuous density, were proposed. One of the hypothesis is an assumption then the solution allocations are envy free with a great probability.
For matrix games of big size with random Gauss distributed entries the hypothesis of Johansson that the supports of optimal strategies for such matrices approximately equal one half of the number of all strategies have been checked.
For the optimal strategies we obtained the joint distribution of its support sizes and weights.
A new consistency property has been defined. The SD-prenucleolus satisfies this property. Also, an axiomatic characterization of the SD-prenucleolus has been defined. This characterization uses three properties: consistency, anonymity and covariance.
With the help of a new invariance axiom - self-covariance - that is weaker than the popular covariance axiom, the class of cost sharing methods for two agents (Moulin 2000) has been extended to the solutions for the whole class of two-person cooperative games. In particular, the solution for superadditive games gives the extension of Moulin's methods to the class of two-person surplus sharing problems.
For a class of big optimization problems the absence of "efficiently calculable" strategies has been proved. It was shown that when the size of a big problem tends to infinity, the equiprobable mixed strategy approximates the optimal one. The absence of efficiently calculable strategy giving a result for every optimization problem has been proved as well.
The nonlinear models with two main types of sticky prices are estimated for set of the countries. The first type of price rigidity is Calvo pricing (that suggests heterogeneous agents). The second one is Rotemberg pricing (that suggests homogenous firms). The Calvo pricing produces significant better performance for large share of countries in the sample. However, the opposite is true for some countries. Thus, we generate the list of countries which are the more favorable for models with heterogeneous agents.