Macroeconomic forecast uncertainty in the euro area

Abstract

This paper estimates aggregate measures of macroeconomic uncertainty from individual density forecasts by professional forecasters. The methodology used in the paper improves on the existing literature along two dimensions. Firstly, it controls for changes to the composition of the panel of respondents to the survey. And secondly, it assigns more weight to the information submitted by forecasters with better forecasting performance. Using data from the European Central Bank’s Survey of Professional Forecasters from 1999 Q1 to 2014 Q3, the paper finds that the effects of changes in the composition of the panel on aggregate uncertainty can be large in a statistical and economic sense. It also finds that the estimates of aggregate uncertainty that use performance-based weights differ significantly from the simple averages used in the literature and their dynamics are more consistent with the dynamics displayed by the estimates of uncertainty computed from financial indicators.

Does Uncertainty Affect Non-response to the European Central Bank’s Survey of Professional Forecasters?

“…3 Despite the growing interest in the SPF, there is a surprisingly scarce amount of research on the factors that affect non-response to this survey. Engelberg et al (2011) and López-Pérez (2016) explored the effects of changes in the composition of the panel of participants on aggregate results from the survey, but the decision to reply is not investigated. Furthermore, Engelberg et al (2011Engelberg et al ( : 1076 concluded that,…”

“…If increases in uncertainty augment the cost of forecasting, the removal of these observations would bias the results presented in this paper against any negative effect of uncertainty on response. 13 The interested reader may find in López-Pérez (2016) all the details about the preliminary treatment of the data.…”

Does Uncertainty Affect Participation in the European Central Bank's Survey of Professional Forecasters?

“…4 Despite the growing interest in the SPF, there is a surprisingly scarce amount of research on the factors that affect non-response to this survey. Engelberg et al (2011) and López-Pérez (2016) explored the effects of changes in the composition of the panel of participants on aggregate results from the survey, but the decision to reply is not investigated. Furthermore, Engelberg et al (2011Engelberg et al ( : 1076 concluded that, "We observed in the Introduction that, in the absence of knowledge of the forecaster recruitment and participation process, the assumption that data are missing completely at random is not refutable.…”

“…Turning now to the measure of uncertainty, the data is obtained from López-Pérez (2016) where several measures of individual uncertainty are computed from SPF density forecasts. One of these measures, the Gini index of the individual density forecast, is used in this paper.…”

“…when the forecaster faces maximum uncertainty and all the intervals look equally likely. Note that the standard deviation of a density forecast reaches its maximum when 13 The interested reader may find in López-Pérez (2016) all the details about the preliminary treatment of the data. In particular, open-ended intervals are assumed to have the same width than regular intervals, and the individual density forecasts with too much probability in open-ended intervals are excluded from the analysis.…”