Specific features of modelling rules of monetary policy on the basis of hybrid regression models with a neural component

• Main Authors: Lukianenko Iryna H., Zhuk Vasyl M.
• Format: Article
• Language: English
• Published: PH "INZHEK" 2014-01-01
• Series: Problemi Ekonomiki
• Subjects: information asymmetry; monetary policy; monetary rule; macro-economic stability; hybrid econometric models; neural network
• Online Access: http://www.problecon.com/pdf/2014/1_0/323_329.pdf

The article considers possibilities and specific features of modelling economic phenomena with the help of the category of models that unite elements of econometric regressions and artificial neural networks. This category of models contains auto-regression neural networks (AR-NN), regressions of smooth transition (STR/STAR), multi-mode regressions of smooth transition (MRSTR/MRSTAR) and smooth transition regressions with neural coefficients (NCSTR/NCSTAR). Availability of the neural network component allows models of this category achievement of a high empirical authenticity, including reproduction of complex non-linear interrelations. On the other hand, the regression mechanism expands possibilities of interpretation of the obtained results. An example of multi-mode monetary rule is used to show one of the cases of specification and interpretation of this model. In particular, the article models and interprets principles of management of the UAH exchange rate that come into force when economy passes from a relatively stable into a crisis state.