Modeling Uncertainty: Predictive Accuracy as a Proxy for Predictive Confidence

Abstract

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may AbstractThis paper evaluates current strategies for the empirical modeling of forecast behavior. In particular, we focus on the reliability of using proxies from time series models of heteroskedasticity to describe changes in predictive confidence. We address this issue by examining the relationship between ex post forecast errors and ex ante measures of forecast uncertainty from data on inflation forecasts from the Survey of Professional Forecasters. The results provide little evidence of a strong link between observed heteroskedasticity in the consensus forecast errors and forecast uncertainty. Instead, the findings indicate a significant link between observed heteroskedasticity in the consensus forecast errors and forecast dispersion. We conclude that conventional model-based measures of uncertainty may be capturing not the degree of confidence that individuals attach to their forecasts but rather the degree of disagreement across individuals in their forecasts. 1The most popular example of this modeling approach is the Autoregressive Conditional Heteroskedasticity (ARCH) model of Engle (1982) and its various extensions.It is widely recognized that macroeconomic outcomes depend critically on both peoples' expectations and the confidence attached to those expectations. Because these magnitudes are largely unobservable, a considerable amount of work has focused on the empirical modeling of forecast behavior. As a result of this effort, there is now general agreement among applied researchers on this issue. Following the tenets of the rational expectations hypothesis, subjective predictions are assumed to be optimal forecasts given all available information and are equated to the objective conditional expectation from the specific model under consideration. With regard to modeling forecast uncertainty, the prevailing approach relies on time series models of heteroskedasticity in which the variance surrounding a prediction is allowed to change over time.1 Temporal variation in subjective uncertainty is equated to the objective conditional variance of a series, with heightened (diminished) uncertainty associated with episodes of decreased (increased) predictability.In spite of the important role played by uncertainty in economic behavior, there is a very limited understanding about the nature of this process. Consequently, the use of time series models of heteroskedasticity to generate estimates of uncertainty has la...

Real-Time Price Discovery in Stock, Bond and Foreign Exchange Markets

“…However, the fit was generally the best for the model in equation (5.3), although the corresponding estimates for the mean parameters in (5.2) essentially remained the same. This is consistent with our earlier empirical results for the spot foreign exchange market in ABDV (2003) that the mere presence of an announcement, quite apart for the size of the corresponding surprise, tend to boost volatility; see also the discussion in Rich and Tracy (2003). 23 Details concerning all of the parameter estimates, including the coefficients in the volatility equation, are available upon request.…”

The Measurement and Behavior of Uncertainty: Evidence from the ECB Survey of Professional Forecasters

“…The availability of matched point and density forecasts from the ECB-SPF, however, provides a unique opportunity to formally address this issue. Moreover, we can draw upon the work of Rich and Tracy (2003) and use simple moment conditions for the data to conduct the analysis.…”

Bias in Federal Reserve Inflation Forecasts: Is the Federal Reserve Irrational or Just Cautious?

“…49 If the loss function is known, condition (19) can be used to evaluate the optimality of a particular sequence of forecasts. The test consists on finding whether L (π t+l − f t+l,t , φ) is uncorrelated to v t , and power against alternative hypotheses is achieved by selecting the appropriate v t .…”