The Euro‐Sting Revisited: The Usefulness of Financial Indicators to Obtain Euro Area GDP Forecasts

• Author: Agustin Garcia‐Serrador,Maximo Camacho
• Date: 01 April 2014
• DOI: http://doi.org/10.1002/for.2284
• Published date: 01 April 2014

The Euro-Sting Revisited: The Usefulness of Financial Indicators to

Obtain Euro Area GDP Forecasts

MAXIMO CAMACHO

1

*AND AGUSTIN GARCIA-SERRADOR

2

1

Departamento de Métodos Cuantitativos para la Economía, Facultad de Economía y Empresa,

Universidad de Murcia, Spain

2

BBVA Research, Madrid, Spain

ABSTRACT

This paper uses an extension of the Euro-Sting single-index dynamic factor model to construct short-term forecasts of

quarterly GDP growth for the euro area by accounting for financial variables as leading indicators. From a simulated

real-time exercise, the model is used to investigate the forecasting accuracy across the different phases of the business

cycle. Our extension is also used to evaluate the relative forecasting ability of the two most reliable business cycle

surveys for the euro area: the PMI and the ESI. We show that the latter produces more accurate GDP forecasts than

the former. Finally, the proposed model is also characterized by its great ability to capture the European business cycle,

as well as the probabilities of expansion and/or contraction periods. Copyright © 2014 John Wiley & Sons, Ltd.

key words real-time forecasting; eurozone GDP; business cycles

INTRODUCTION

The financial crisis that erupted in 2007 called into question the ability of traditional forecasting methods which were

unable to anticipate the turning point in time. This forecasting failure triggers some proposals that tried to improve

upon the early detection of the unforeseen downturn. Focusing on euro area forecasting, the small scale dynamic

factor model suggested by Camacho and Perez Quiros (2010) has been one of the most successful in computing

accurate and timely assessments of short-term GDP developments. The good forecasting performance of their model

comes from the fact that it mixes quarterly and shorter frequencies and that it deals with asynchronously published

economic indicators which allow the model to compute real-time forecasts on the basis of timely updated data.

Although their model mixes the information contained in hard and soft data, it does not include financial indicators.

However, according to Wheelock and Wohar (2009), financial variables (such as the slope of the yield curve) can be

helpful as leading indicators in forecasting growth. In this context, the main contribution of this paper is the analysis of

financial time series as leading indicators of output growth, in a dynamic factor model that accounts for asynchronous

co-movements between the financial and the real activity indicators.

A further drawback of the analysis developed in Camacho and Perez Quiros (2010) is that their sample dated from

the early 1990s to 2007, which includes only one recession. In addition, their real-time forecasting analy sis started in

2003 and ended in 2007, which became a period of relatively stable high growth. To overcome this potential

shortcoming, we use enlarged historical time series of euro area GDP growth rate, obtained from the AWM Database

elaborated for the Euro Area Business Cycle Network (€ABCN).

1

According to its methodology, the historical data

since the first quarter of 1970 are based on the aggregation of available country data. The main source is Eurostat,

complemented by the OECD National Accounts, the OECD Main indicators, the BIS and the AMECO databases.

The new dataset used in this paper is now dated back to the early 1980s, which allows us to develop the following

twofold exercise.First, the common factor is now available fordata back to the 1980s. Hence, we use recent techniques

to support the view that the factor can become a reliable economic indicator of the euro area business cycle.

Second, we develop a pseudo real-time exercise to evaluate the performance of the model to compute short-term

forecasts of euro area GDP growth rates. The main feature of this forecasting analysis is that the data vintages are

constructed by taking into account the lag of synchronicity in data publication that characterizes the real-time data

flow. In addition, according to the standard literature on forecasting, the forecasts are carried out in a recursive

way and with every new vintage, as the model is re-estimated and the forecasts for different horizons are computed.

In the empirical analysis, we show that our model would have accurately forecast GDP over the past 20years. The

model yields significant forecasting improvements over benchmark predictions computed from models that are only

based on standard autoregressive specifications.

*Correspondence to: Maximo Camacho, Departamento de Métodos Cuantitativos para la Economía, Facultad de Economía y Empresa,

Universidad de Murcia, 30100, Murcia, Spain.

E-mail: mcamacho@um.es

1

The AWM Database and its methodology are available on the website of €ABCN (http://www.eabcn.org/area-wide-model) .

Journal of Forecasting,J. Forecast. 33, 186–197 (2014)

Published online 24 January 2014 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/for.2284

Copyright © 2014 John Wiley & Sons, Ltd.