Organizing Team

Prof. Dr. Ulrich Woitek

Institut für empirische Wirtschaftsforschung

University of Zurich

Prof. Dr. Albrecht Ritschl

Institute of Economic History

Humboldt University

Date & Place

University of Zurich, 11-13 April 2007

Abstract

The aims of the workshop are to define unified methodological standards to be used within our priority area, to develop tools to be made available subsequently to a wider circle of participants, and to determine the use of common pool resources such as software routines and datasets. At the same time, results from pilot projects will be presented and common frames for subsequent studies established.

Scientific Summary

There are two approaches available for quantitative research of aggregate phenomena in economic history. The traditional one, to be labelled New Data Approach, has looked at reconstructed national accounts (RNA) and the information therein. An alternative is to employ large numbers of disaggregate data and to leave the aggregation to a statistical optimisation algorithm. The workshop is a first step toward making progress in the second field.

We draw on two different but related methodologies in recent statistical research on business cycle dating. One, based on frequency domain techniques, examines quasi-cycles in economic time series (GDP and components, prices, industrial production, employment, etc.) and analyzes interrelations between these by means coherence and phase spectrum to obtain information about co movement.

The second is based on statistical factor analysis and aims to aggregate the relevant information inherent in large numbers of disaggregate time series, extracting common factors that can themselves be analyzed using time series techniques.

Both approaches are being applied successfully in a currently emerging literature on the dating and analysis of historical business fluctuations. We aim to contribute to this research, developing the conceptual tools further and applying them to historical datasets. As the application of this class of statistical techniques to historical data is still quite new, we expect to have to adjust the methodology to meet the specific needs of the historical at hand. Naturally, much of the analysis is to be based on Bayesian techniques, which implies that suitable prior distributions will have to be experimented with.