Beyond accuracy: Structure and evolution of the quality of economic forecasts

Aim. To substantiate a methodological approach to analyzing the evolution of economic forecasting based on a multidimensional assessment of the quality of forecast models and the identification of stable configurations of requirements for forecast validity. Objectives. To develop a formalized system of predicates for the quality of economic forecasting that reflects the key methodologically independent dimensions of forecast validity of models; to define the universe of forecast models and to form a standardized model passport for their comparable analysis in the quality space; to verify the predicate assessment of forecast models from different methodological directions and historical periods using the proposed quality system; to identify stable types of forecast models and characteristic configurations of predicates for each stage of development; to propose an original periodization of the evolution of economic forecasting and to identify the main methodological conflicts that determine the dynamics of forecast model development. Methods. The research employs systemic and comparative analysis of economic forecast models. Methods of multi-criteria quality assessment, a predicate approach to formalizing model characteristics, entropy analysis, quality frontier analysis, assessment of the hypervolume of the configuration space, and cluster analysis (k-means) are used. The empirical basis consists of 96 economic forecasting models from the early 20th century to the present. Results. A formalized system for assessing the quality of forecast models is presented, comprising six methodologically independent predicates: Accuracy, Robustness, Reproducibility, Falsifiability, Structural Consistency, and Uncertainty Calibration. It is shown that the evolution of economic forecasting is cumulative in nature and is accompanied by an expansion of the space of quality configurations while preserving a limited number of stable architectures. Four stable regimes of forecast model quality are identified, and a five-stage periodization of the development of economic forecasting is proposed that reflects the shift in methodological priorities. Conclusion. The quality of economic forecasts cannot be reduced merely to accuracy indicators. The development of forecast models is driven by the formation of stable combinations of different quality characteristics that reflect trade-offs between accuracy, robustness, reproducibility, interpretability, and the ability to operate under uncertainty. The proposed methodology creates a unified formal approach to the comparative assessment of forecast models, and it can be applied both to studying the evolution of forecasting methods and to substantiating the choice of models in applied practice.