Evaluating survey-based forecasts of interest rates and macroeconomic variables

Abstract

PurposeThis study examines the forecasting performance of the professional analysts participating in the Blue Chip Economic Indicators Survey using an alternative methodological research design.Design/methodology/approachThis work employs two methodologies, namely a panel specification, with the cross-section being the forecast horizon (from 1-month to 18-months ahead forecasts) and the time period being the time that the forecast was made and a quantile regression technique, which evaluates the hidden nonmonotonic relations between the forecasts and the target variables being forecasted.FindingsThe empirical findings of this study show that survey-based forecasts of certain key macroeconomic variables are generally biased but still efficient predictors of target variables. In particular, we find that survey participants are more efficient in predicting long-term interest rates in the long-run and short-term interest rates in the short run, while the predictability of medium-term interest rates is the least accurate. Finally, our empirical analysis suggests that currency fluctuations are very hard to predict in the short run, while we show that survey-based forecasts are among the most accurate predictors of GDP deflator and growth.Practical implicationsEvaluating the accuracy of economic forecasts is critical since market participants and policymakers utilize such data (as one of several inputs) for making investment, financial and policy decisions. Therefore, the quality of a decision depends, in part, on the quality of the forecast. Our empirical results should have immediate implications for asset pricing models that use interest rates and inflation forecasts as variables.Originality/valueThe present study marks a methodological departure from existing empirical attempts as it proposes a simpler yet powerful approach in order to investigate the efficiency of professional forecasts. The employed empirical specifications enable market participants to investigate the information content of forecasts over different forecast horizons and the temporal evolution of forecast quality.

Interest rate forecasts in Latin America

“…Even though the unbiasedness test is one of the most traditional tools for assessing forecast quality (cf. Friedman, 1980), it is still widely used today (Fassas et al , 2021; Filiz et al , 2019; Miah et al , 2016; Baghestani et al , 2015). In the context of the unbiasedness test, many different procedures can be applied depending on the research question and the structure of the database.…”

“…For example, Davies and Lahiri (1995) recommend pooling forecasts in order to get a clearer picture of rationality along the entire forecast path. Following their recommendation, Fassas et al (2021) examine the unbiasedness of forecasts with a panel regression, taking into consideration 18 different forecasts with horizons from 1 to 18 months for each actual datapoint. While this approach can produce insightful results, it is most efficient when the database contains various different forecasts horizons.…”

US unemployment rate: Federal Reserve versus private information

“…With the three-and four-quarter-ahead forecasts free of systematic bias, in line with Fassas et al (2022), we further examine whether these forecasts are unbiased by estimating the following test equation, where the forecast is unbiased when we cannot reject the null hypothesis that α 5 0 and β 5 1.…”

Sticky Stock Market Analysts

“…The actual success of stock market forecasts is thus best checked against real ex-ante forecasts. In the area of interest rate forecasts, the evaluation of continuously published forecasts has a long tradition (Filiz et al 2021;Fassas et al 2021;Filiz et al 2019;Kunze et al 2017;Miah et al 2016;Pierdzioch 2015;Baghestani et al 2015;Oliver and Pasaogullari 2015;Spiwoks et al 2015). In the area of stock market forecasting, however, there are only a small number of studies that check continuously published stock market forecasts for their reliability (see the synoptic overview in Table 1).…”