GDP growth forecasts and information flows: Is there evidence of overreactions?

• Published date: 01 June 2018
• Date: 01 June 2018
• Author: J. Daniel Aromí
• DOI: http://doi.org/10.1111/infi.12126

DOI: 10.1111/infi.12126

ORIGINAL ARTICLE

GDP growth forecasts and information flows:

Is there evidence of overreactions?

J. Daniel Aromí

Facultad de Ciencias Económicas,

Universidad de Buenos Aires, IIEP-

Baires, Córdoba, CABA, Argentina

Correspondence

J. Daniel Aromí, Facultad de Ciencias

Económicas, Universidad de Buenos

Aires, IIEP-Baires, Córdoba 2110 2nd,

CABA, Argentina.

Email:

jose.aromi@posgrado.economicas.uba.ar

Abstract

The association between GDP growth forecasts and past

information flows is evaluated for a sample of 49 countries

during the period 1990–2014. The analysis exploits an

extensive collection of forecasts available through IMF's

historical database. The empirical results indicate a robust

association between information arrival and subsequent

mean forecast errors (the average difference between

forecast and realization). Consistent with the overreaction

hypothesis, more positive information is followed by higher

mean forecast errors. The association is documented for

multiple metrics of past information flows: growth

performance, a novel metric of press sentiment, and lagged

forecast errors. When advanced and emerging economies

are differentiated, the regularity is detected for both groups

but is stronger in the case of emerging economies.

1

|

INTRODUCTION

This work evaluates the association between information flows and subsequent growth forecast

errors. The analysis is motivated by multiple episodes and opinions that point to instances in

which economic forecasts overreacted to incoming information. For example, some years before

the Asian crisis, Krugman (1994) warned against ‘popular enthusiasm about Asia's boom’.More

recently, Pritchett and Summers (2014) indicate that growth expectations regarding the Chinese

and Indian economies might suffer from excessive extrapolation of recent trajectories. In addition

to these warnings, further motivation is provided by macroeconomic episodes in which improved

economic prospects are followed by crises. For instance, several European economies, among

them Greece and Ireland, went through this type of trajectory. Another case is given by recent

events in Brazil, where prominent optimism regarding economic prospects was later proven

wrong in a stark manner.

122

|

© 2017 John Wiley & Sons Ltd wileyonlinelibrary.com/journal/infi International Finance. 2018;21:122–139.

The present study goes beyond the collection of informal evidence and implements a systematic

evaluation of the overreaction hypothesis. A large collection of medium-term economic forecasts is

accessed through IMF's World Economic Outlook Historical Forecasts database. The study covers 49

advanced and emerging economies for the period 1990–2014. Bearing in mind that information flows

can take different forms, they are approximated using three different indicators: lagged forecast errors,

changes in growth performance, and changes in sentiment as reflected by the economic press. The last

indicator is a novel metric that is computed by applying textual analysis techniques to a large corpus of

economic press articles.

The empirical analysis shows a significant association between mean forecast errors and earlier

information flows. The sign of the documented relationship is consistent with the overreaction

hypothesis. More positive information is followed, on average, by higher forecast errors, that is, by

increments in the mean difference between forecast growth and realized growth. This regularity is

documented for all metrics of information flows. When the information content of different metrics is

compared, press sentiment is shown to possess predictive ability for a larger set of model

specifications. This performance draws attention to the value of text analysis techniques in

macroeconomic contexts. At the same time, suggesting complementarity between the different

indicators, the economic significance of the results is greater when the information of the alternative

metrics is combined into a single indicator.

It is worth noting that the strongest evidence is documented for information flows and forecasts

errors that are between 4 and 8 years apart. In other words, the evidence indicates the presence of a

process that develops at a frequency that is lower than the usual business cycle frequency. In this way, it

can be linked to what has been identified as medium-term business cycles (Comin & Gertler, 2006;

Drehmann, Borio, & Tsatsaronis, 2012).

The documented association is robust to multiple variations in the specification of the empirical

model. While the pattern is verified for advanced economies and emerging economies, the estimated

association is more intense in the case of emerging economies. Additional robustness exercises

indicate that different sample periods are associated with similar relationships between information

flows and mean forecast errors.

Two related implications can be derived from the empirical results. First, they point to

inefficiencies in economic forecasts which might have significant macroeconomic consequences.

Economic fluctuations could be amplified by forecasts that overreact to incoming information. To

some degree, crises might be explained by overly optimistic assessments that deepen vulnerabilities. In

this way, these findings can complement the literature that analyses causes of economic and financial

crises.

1

Second, these findings suggest ways in which economic forecasts might be corrected to attain

higher precision. More specifically, according to the analysis, after times of atypical economic

expansion or manifest optimism, expectations should be corrected towards less optimistic assessments.

These corrections could improve decision-making of private actors and might help implement better

macroprudential policies.

2

The current analysis is connected to a literature that focuses on the role of uncertainty and learning

in macroeconomic contexts.

3

According to these diverse contributions, mistakes in judgements

regarding structural features of the economy can explain important characteristics of the business

cycle. The results here reported suggest that, in addition to shocks which cannot be anticipated, errors

in expectations include elements associated with the inefficient use of available information.

These inefficiencies have been the focus of the literature on bounded rationality. The literature

explores mechanisms that can account for errors in the process of expectation formation. The

systematic errors reported in this work are compatible with analyses that point to excessive

extrapolation of past outcomes (Barberis, Greenwood, Jin, & Shleifer, 2016; Fuster, Laibson, &

AROMÍ

|

123