If a monetary authority desires to test for optimal policy solution before applying policy to the economy, it will use, either explicitly or implicitly, a macroeconomic model, which projects the effect hypothetic policy action to the dynamics of the key indicators of economic growth. The dominant approach here is the use of dynamic stochastic general equilibrium models (DSGE) and policy rules, representing the generalized principle of policy reaction. Within this modeling framework the policy rule is an equation, usually linear linking the basic economy indicators, such as output gap, price inflation and lagged central bank interest rate in the classical Taylor rule to the present value of monetary instrument.
However, in case of the transitional economies this approach faces some major conceptual issues. The DSGE methodology represents the behavior of the economy around its known steady state, so the long-term path of the model is basically predetermined and the policy task is simplified to shock anticipation. In case of economy in transition, the continuous structural transformations lead to the situation when distinguishing temporary shocks from long-term structural shifts become a complicated task. Moreover, the monetary authority might need to react to major structural breaks not only using general policy rule approach, but also by asymmetrical policy measure, like switching to another regime of monetary policy. In this case, the single linear policy rule methodology will be inapplicable for optimizing long-term policy principles.
This paper introduces nonlinear policy rule as a way to simulate and interpret both the shock anticipation and regime shifting components of the monetary policy in the Ukrainian economy during the years of 1998-2008. In order to meet this objective, we estimate multi-regime neural coefficient smooth transition regression (NCSTR) to fit the nonlinear policy rule to historical data. Furthermore, we analyze the interaction of NCSTR-form policy rule with a vector-autoregressive model (VAR), which represents the economy, building up a flexible model with good fit capabilities. Due to the abovementioned flexibility, ability to operate on comparatively small samples and possibility to omit extensive set of assumptions, this approach is conceptually suitable for analysis in terms of transitional economy. Based on this modeling framework we perform scenario simulation and discuss ways to optimize monetary regime transition conditional on the state of the developing economy.