HIGH-FREQUENCY FORECASTING MODELS OF NATIONAL ACCOUNT VARIABLES IN TRANSITION ECONOMIES (RUSSIA, ARMENIA, AND THE CZECH REPUBLIC)

Open Access
Author:
Roudoi, Andrei
Graduate Program:
Economics
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
June 27, 2005
Committee Members:
  • Barry William Ickes, Committee Chair
  • Michael Bernhard, Committee Member
  • N Edward Coulson, Committee Member
  • Nezih Guner, Committee Member
  • Lawrence R Klein, Committee Member
Keywords:
  • Transition Economies
  • Forecasting
Abstract:
This study shows that despite the shortness of time-series and significant structural change the economies of three transition countries – Russia, Armenia, and the Czech Republic – have started to demonstrate regularities that can be used in some types of econometric forecasting. The high-frequency models developed here generate relatively accurate short-term forecasts of quarterly national account variables, overall 53 for the three countries, and ultimately real GDP. They can serve as a foundation for real-time short-term forecasting models and can provide initial conditions for medium-term or long-term structural forecasting models. Government agencies and business analysts worldwide monitor high-frequency information that is released on a monthly and or even more frequent basis to estimate lower frequency data, such as quarterly national account data, before reports on these data are released. Such estimates are important, for example, in the preparation of government budgets. For example, the U.S. Treasury Department routinely uses high-frequency modeling to project GDP and tax receipts. The timeliness of these estimates may be critical near business-cycle peaks and troughs, since the government needs time to prepare counter-cyclical measures. The models that I have developed may help the governments of transition countries in economic policy decision-making. The two main methodological features of the models – the forecasting of real GDP directly and through its elements and the method of principal components – allow for the incorporation of a large number of explanatory variables. The selected complex forecasting regressions satisfy a series of absolute and comparative forecast accuracy criteria, and this study shows that forecasting accuracy improves if the model takes into account signals from a large variety of indicators. Chapter 1 describes the model methodology, forecast accuracy testing, and statistical problems. Chapter 2 provides a review of relevant literature. It focuses on the place of the model among forecasting approaches, compares the costs and benefits of the methodology applied here with those of other methods, and discusses practical applications. Chapter 3 describes economic growth trends in the three countries and the indicators that are generally considered growth determinants. It provides an overview of the selection of forecasting regressions and the satisfaction of forecast accuracy criteria.