The Forecasting and Policy System: stochastic simulations of the core model

Uncertainty in applied macroeconomic policy analysis arises from three distinct sources. The first, often referred to as model uncertainty, arises because the models used for policy analysis are simple abstractions of the complex behavioural interactions that occur in an economy. The second source, denoted shock uncertainty, arises from unforeseen events that the analysis cannot explicitly factor in ex ante. Finally, starting-point uncertainty reflects the fact that given data lags and revisions, often it is difficult to assess the current state of the economy. This paper discusses the approach the Reserve Bank has taken to enable its Forecasting and Policy System (FPS) to quantify the implications that the typical level of shock uncertainty might be expected have on the analysis of alternative policy actions designed to achieve the objectives of monetary policy. The technique uses the impulses from an estimated vector autoregression (VAR) model of the New Zealand economy to represent the typical level of shock uncertainty policy must deal with. These impulses are mapped into shocks to key behavioural relationships in the calibrated structural macroeconomic model that lies at the heart of FPS. This approach leads to random disturbances that capture both the serial and cross correlations in the data. The structural model is then simulated while being randomly subjected to the representative impulses. This exercise builds-up statistical distributions for key macroeconomic variables. The model generated macroeconomic properties are compared to the historical properties of the New Zealand economy. Given the amount of structural change that the New Zealand economy has undergone over the historical sample period, and the functional form of the FPS policy reaction function, a-priori one might not expect model generated moments to perfectly replicate historical moments. Nevertheless, the model generated properties appear to be reasonably close to the historical properties of the New Zealand economy. Sensitivity analysis is also performed to examine the stochastic behaviour of the core model under different characterisations of the shock generating process. Changing the informational structure that the monetary authority has about the shocks, incorporating exogenous disturbances to the policy instrument, and altering the causal ordering of the VAR model, have little impact on the behaviour of the model economy relative to history. However, ignoring the cross-correlations in the data leads to stochastic behaviour of the model economy that is implausible relative to history. The views expressed in this paper are those of the authors and do not necessarily represent the views of the Reserve Bank of New Zealand.