The importance of time‐varying volatility and country interactions in forecasting economic activity

• Published date: 01 September 2017
• Date: 01 September 2017
• DOI: http://doi.org/10.1002/for.2457
• Author: Steven Trypsteen

Received: 26 February 2015 Revised: 3 December 2016 Accepted: 5 December 2016

DOI: 10.1002/for.2457

RESEARCH ARTICLE

The importance of time-varying volatility and country

interactions in forecasting economic activity

Steven Trypsteen

ING Belgium, Brussels, Belgium

Correspondence

Steven Trypsteen, INGBelgium, B-1000 Br ussels,

Belgium.

Email: steven.trypsteen@ing.be

Abstract

This paper examines the relative importance of allowing for time-varying volatility

and country interactions in a forecast model of economic activity.Allowing for these

issues is done by augmenting autoregressive models of growth with cross-country

weighted averages of growth and the generalized autoregressive conditional het-

eroskedasticity framework. The forecasts are evaluated using statistical criteria

through point and density forecasts, and an economic criterion based on forecasting

recessions. The results show that, compared to an autoregressive model, both com-

ponents improve forecast ability in terms of point and density forecasts, especially

one-period-ahead forecasts, but that the forecast ability is not stable over time. The

random walk model, however, still dominates in terms of forecasting recessions.

KEYWORDS

density forecast, forecasting recessions, GARCH models, global VAR, probability

forecast

1INTRODUCTION

This paper investigates the relative importance of allowing

for country linkages and time-varying volatility in forecast-

ing economic activity for the G7 countries, and tests whether

the forecast ability changes over time. As output growth data

are highly correlated across countries because of trade and

financial links (Canova, Ciccarelli, & Ortega, 2007; Kose,

Otrok, & Whiteman, 2003, 2008), accounting for these links

in a forecast model could lead to improved forecasts. The

financial crisis of 2008–2009 and the European sovereign

debt crisis are important examples of the power of coun-

try linkages. These two crises also highlight the importance

of time-varying volatility or uncertainty. A forecast model

that takes increased uncertainty during a crisis into account

could lead to better forecasts as it can capture the spreading

out of distributions of future events. Moreover, as shown by

Hamilton (2010), estimates of the conditional mean can be

substantially more efficient if observed heteroskedasticity is

modeled and so could.lead to improved forecasts.

Allowing for a time-varying variance could also be impor-

tant for another reason. Timely information on economic

activity is an important input in economic decision mak-

ing, ranging from investment decisions to the formulation

of monetary policy. High-frequency data, which is released

in a timely manner, is therefore particularly useful to fore-

cast macroeconomic series of a lower frequency. Golinelli

and Parigi (2007), for example, show in the context of bridge

equations that monthly industrial production is a crucial indi-

cator to forecast quarterly gross domestic product (GDP) for

the G7 countries. High-frequency data, however, are poten-

tially more volatile compared to lower-frequencydat a; indus-

trial production growth is more volatile compared to GDP

growth for example, and so modeling the variance of such

series explicitly could improve forecasts.

To investigate the relative importance of country link-

ages and time-varying volatility, the paper compares the

forecast performance of a range of models that aim to iso-

late the contribution of the two factors. The most general

model considered is an autoregressive model of output growth

augmented with two components. I include cross-country

weighted averagesof growth to allow for country interactions.

Using cross-country weighted averages in this way is a tech-

nique developed in Pesaran, Schuermann, and Weiner (2004)

and Dées, Di Mauro, Pesaran, and Smith (2007) in the con-

text of vector autoregressive (VAR) modeling, the so-called

Global VAR, and in Pesaran, Schuermann, and Smith (2009)

in the context of forecasting. Pesaran, Schuermann, and

Journal of Forecasting.2017;36:615–628. wileyonlinelibrary.com/journal/for Copyright © 2017 John Wiley & Sons, Ltd. 615