Title: A Bayesian ECM for forecasting real ruble exchange rate under structural change in the monetary policy
Authors: Nikita Fokin - Russian Presidential Academy of National Economy and Public Administration (Russia) [presenting]
Andrey Polbin - Russian Presidential Academy of National Economy and Public Administration (Russia)
Abstract: At the end of 2014, the Central Bank of Russia changed the monetary policy regime. The goal is to construct the model, which can accurately forecast the real exchange rat immediately after changing the monetary policy regime. Following the assumption of long-term neutrality of money, we assume that the transition to a new regime has changed the mechanism of adapting of the real exchange rate to long-term equilibrium in response to the shock of oil prices, but the long-term parameters of the model have not changed a lot. First, we estimate the frequentist ECM on the period of the old monetary policy regime. The estimates of parameters of the cointegration relation and their confidence intervals are set as priors for long-term parameters of the equation of the new regime. For short-term parameters, uninformative priors are set. Bayesian specification of the model, in contrast to the frequentist model, has two advantages. The first one is that it allows us to predict the real exchange rate immediately after the monetary policy regime was changed, when for the frequentist model it is necessary to collect a minimum set of observations to estimate the parameters of the new regime. The second advantage is that we allow long-term parameters to change slightly due to a change in the monetary policy regime, when in the frequentist model we would have to impose a strict restriction on the equality of the parameters of the cointegration relation in different regimes.