3/2006 Agostino Consolo. Forecasting measures of inflation for the Estonian economy

Working Papers of Eesti Pank. No 3, 2006

The aim of this paper is to forecast some of the most important measures of inflation of the Estonian economy by making use of linear and non-linear models. Results from comparing classes of optimal models are similar to those in the forecasting literature. In particular, there are gains from using more sophisticated methods such as factor analysis and time-varying parameters methods.
Model discrimination is based on evaluation criteria which are computed by a real-time dynamic estimation procedure. Moreover, forecasts uncertainty is appropriately taken into account: Fan Charts can exhaustively describe the final output for what concerns out-of-sample forecasting.
JEL Code: C22, C32, C53, E31
Key words: Estonian Economy, forecasting, inflation modelling

*  I am grateful to Marco Aiolfi, Andrea Carriero, Martin Lindpere and Massimiliano Marcellino for several conversations on some of the issues discussed in this paper. I acknowledge the financial support of Eesti Pank. Any remaining errors are my own responsibility.

Author's e-mail address: email [at] agostino.it

The views expressed are those of the author and do not necessarily represent the official views of Eesti Pank.


1. Introduction
2. Forecasting Methodology
2.1. Forecasting by Dynamic Estimation
2.2. Forecasts Evaluation
2.3. Forecasting with Uncertainty
3. Methods Specification
3.1. Linear Methods
3.2. Autoregressive Time-Varying Parameters
3.3. Smooth Transition Regression
3.4. Phillips' Curve
4. Data Source and Description
5. Forecasting Inflation Measures
5.1. Model Selection
5.2. Model Specification
5.2.1. AR Models
5.2.2. VAR Models
5.2.3. Logistic Smooth Transition
5.2.4. Time-Varying Parameters
5.2.4. Phillips' Curve: A Factor Analysis
5.3. Forecasts Evaluation
5.4. Out-of-Sample Forecasting
6. Conclusions
Appendix 1. Absolute MAFE and MSFE

Forecasting Measures of Inflation for the Estonian Economy, Working Papers of Eesti Pank No 3, 2006 (PDF*)

* To read PDF file, you need Adobe® Acrobat® Reader™ freeware, it may be downloaded from Adobe homepage.