Probabilistic Approach to Assessing Macroeconomic Uncertainties

PRAMU

ESRC/ORA Project

Aims - Summary & Objectives


Summary

The aim of the project is to develop new methods of forecasting of macroeconomic indicators, like inflation, interest rates and output, where extreme events (hyperinflations, rapid devaluations, etc) are not treated as anomalies, but as intrinsic parts of economic processes.

It is motivated by the finding that the currently used methods of assessing uncertainties of main macroeconomic indicators and instruments (inflation, exchange rates, and interest rates) do not account properly for extreme events like substantial and sudden devaluations, deflation or the developments of high inflation.

The essence of the methods which are to be developed within the project is in applying novel theoretical distributions for explaining macroeconomic uncertainties, where the drastic events are treated as intrinsic characteristics of these distributions. This new class of distributions (often called the tempered stable distributions) allows for flexible modelling of various types of extreme scenarios, asymmetries and odd events.

 Objectives

There are three groups of tasks here. The first group aims at developing technical procedures for the analysis of such tempered stable distributions which are particularly well suited for describing macroeconomic uncertainties.

The second group deals with the analysis of time series of data of macroeconomic indicators with the use of tempered stable distributions. Within the third group of tasks new ways of forecasting uncertainties for long horizons will be investigated.

  • Static analysis: develop new technique suitable for macroeconomic research, of parameters' estimation of tempered stable distributions, fitting these distributions to data and hypothesis testing.
  • Dynamic analysis: extend the static analysis to the analysis of time series of data.

Multivariate analysis: propose a new forecasting methodology consisting of creation of multistep forecast by merging together tempered stable distributions with different parameters in such way that they will constitute a multivariate distribution with pre-determined dependencies.