Ebook Leading Indicators In A Globalised World
Monetary policy decisions affect the economy with long and varying lags. It is therefore crucial to have an educated judgement about the economic conditions and outlook prevailing at the time. Leading indicators constitute an important tool in applied business cycle analysis to form such a judgement. A main potential shortcoming of country-specific leading indicators is, however, that their components are commonly domestic variables. The ability of such leading indicators to predict economic activity might have diminished due to the rapid advances in globalisation, as reflected in the deepening of international financial and trade linkages. Notwithstanding the significant structural changes that have taken place at the global level over recent years, to our knowledge the possible implications for the leading indicator properties have not been addressed in a systematic fashion as yet.
This paper aims at filling this gap. It looks into this issue based on the OECD composite leading indicators (CLI) across 11 countries and uses, in line with common practice, industrial production as the reference series. The OECD CLI has the advantage (1) to be widely monitored by practitioners and (2) to be available for a wide variety of countries, on a monthly basis and over a long time span (see also Camba-Mendez et al., 1999).
The objective of the paper is two-fold. Firstly, we assess the properties of the OECD CLI and check whether its accuracy has declined over time as the process of globalisation has evolved. In line with standard practice, the assessment is made on the basis of root mean squared forecast errors calculated – as suggested in Marcellino, Stock and Watson (2005) – based on iterated multi-step forecasts up to one year ahead. The analysis is based on a data sample between January 1975 and April 2008, using the first fifteen years for estimation purposes and the remainder of the sample for the recursive estimation and pseudo out-of-sample forecasting, based on information available at the time the forecast is carried out. We find clear evidence that the forecast properties of a model making use of the domestic CLI are significantly better than naïve autoregressive benchmark forecast. We also document evidence that taking cointegration relationships between the CLI and industrial production into account is useful as it further improves the accuracy of the forecasts of industrial production. However, regarding the evolution of the forecast accuracy over time we find indications that it indeed declined over time for most of the countries.
Secondly, we analyse whether adding information on the international business cycle improves the forecast accuracy of the OECD indicator. To this end, we augment the basic forecast equation, which includes the country-specific leading indicators (and lagged values of domestic industrial production) with a leading indicator for the external environment, constructed from the CLI of the other OECD countries. The results appear quite promising. In spite of differences in the results across models and countries, we find that the inclusion of external leading indicators can improve the forecast performance quite substantially. This is particularly the case if forecast combination techniques are employed. For some countries, we also find evidence that the projections stemming from models augmented with international information are improving over time relative to the model focusing purely on domestic information. However, these conclusions are subject to the caveat that the variation of relative forecast performance is substantial over time.
CONTENTS
Abstract
Non-technical summary
1 Introduction
2 Data and stylised facts
3 Econometric analysis
4 Empirical strategy and results
- 4.1 Leading indicator properties of the domestic CLI
4.2 The temporal dimension: forecasting performance of CLI models over time
5 The external dimension: can international leading indicators improve domestic forecasts?
6 Conclusions
Appendix
References
European Central Bank Working Paper Series
Posted in :
