Information Rigidities: Comparing Average and Individual Forecasts for a Large International Panel

Abstract

We study forecasts for real GDP growth using a large panel of individual forecasts from 36 advanced and emerging economies during 1989-2010. We show that the degree of information rigidity in average forecasts is substantially higher than that in individual forecasts. Individual level forecasts are updated quite frequently, a behavior more in line "noisy" information models (Woodford, 2002; Sims, 2003) than with the assumptions of the sticky information model (Mankiw and Reis, 2002). While there are crosscountry variations in information rigidity, there is no systematic difference between advanced and emerging economies.

High-Frequency Credit Spread Information and Macroeconomic Forecast Revision

“…Recent studies have focused on the linkages between forecast smoothing and models of information rigidities. Dovern et al (2015), for example, examine information rigidities in GDP growth forecasts and show that forecast revision dynamics are in line with the predictions of noisy information models (Woodford 2001;Sims 2003). A similar conclusion is reached by Coibion and Gorodnichenko (2012Gorodnichenko ( , 2015a.…”

“…Matching monthly revisions with daily credit spread data, we take forecast revisions from 7th to 18th updates, as we expect the relationship between output forecasts and credit spread information should be most relevant. 8) Dovern et al (2015) find that the average size of revision is larger at mid-horizons than at very long or short forecast horizons. Sheng and Wallen (2014) also document that the professional forecasters included in Consensus Forecasts data revise their medium-term forecasts (10 to 17 months ahead) most frequently.…”

“…19) This result indicates that, although 18) Our estimate of the rigidity parameter () is larger than those from previous studies with different sample periods. For example, Ager et al (2009)'s estimated rigidity coefficient is 0.28 based on the data from 1996 to 2006, andDovern et al (2015) report a coefficient of 0.33 using a sample from 1989 to 2010. 19) The survey deadlines do not necessarily match the dates when surveys are actually submitted or the dates when the respondents form their predictions.…”

“…Our model specification is in the class of ADL-MIDAS regressions, introduced byAndreou et al (2013), offering the structure of augmented distributed lag regression with mixedfrequency data. The inclusion of an autoregressive term of order one is from the literature testing the weak form of forecasting efficiency(Nordhaus 1987; Deschamps and Ioannidis 2013;Dovern et al 2015).Several previous studies proposed diverse functional forms of MIDAS polynomial weights aiming at parsimonious but flexible specifications.15) 13) For an overview of the MIDAS regression framework, seeArmesto et al (2010),Andreou et al (2011), and Foroni and Marcellino (2013). 14) Notations for the revision horizons () are abstracted as each forecast time () matches to one corresponding consensus revision.…”

Information rigidities and the news-adjusted output gap

“…The survey responses of professional forecasters have been used recently to examine the nature and extent of information rigidities by Carroll (2003), Andrade and Le Bihan (2013), Gorodnichenko (2011, 2012) and Dovern et al (2014), inter alia. In those papers, the analysis of forecasters' responses at the individual level and at an aggregated level provides evidence in favour of both sticky information models and relatively simple noisy information models.…”