The paper discusses the theory and practice of optimal decision-making, i.e. the choice of a macroeconomic policy in the circumstances of uncertainty.

The first part of the text contains an overview of literature. The characteristic feature of all the quoted cases is the integration of two aspects: the macroeconomic (model of economy) and the theory-based decision-making one. The classical model of economic policy is reviewed, where the components of the model of economy (distortion, parameters, data) are considered to be uncertain (random) and optimal decisions are taken in accordance with the principle of equivalence, Brainard's principle of conservatism, etc. Doubts about the model components lead to questioning of the whole economy model. Therefore, different ways to deal with the model uncertainty are presented: the principles of robustness by Levin, Williams and Wieland and a comprehensively Bayesian approach to the process of decision-making. Where uncertainty cannot be quantified by the probability distribution (as, in Knight's understanding, it is an uncertainty, not a risk), techniques based on the minimax function of the goal are proposed. In this group of cases, the robust control theory by Hansen and Sargent is described in more detail.

The second part presents methods applied in practice. The author draws from academic research, opinions of practitioners and decision-makers, as well as unofficial information from analysts of central banks. It may be observed that where theoreticians suggest refined, one-stage techniques referring to the uniform model of economic policy, practitioners prefer multi-stage procedures based on intuition and expertise. Simultaneous operation of multiple models, development of variant projections, designing pie charts to describe subjective risk and defining ranges of policy objectives have become virtually standard procedures. Thus a gap between the theory and practice is clearly visible. However, the size of the gap is only partially explained by the fact that many academic proposals are at an initial stage of development and that institutional constraints are not taken into account.