Professor within the Department of Finance at the Faculty of Law, Economics and Finance, Roman Kräussl, explores the way strategic bias in the predictions of economic forecasters can act as a harbinger of economic surprises (actual economic data relative to consensus estimates).
In a 2020 paper entitled “Strategic bias and popularity effect in the prediction of economic surprises,” Prof. Kräussl and coauthors Luiz Félix (APG Asset Management) and Philip Stork (VU Amsterdam, Tinbergen Institute Amsterdam) study how professional economic forecasters make decisions using publicly and privately available information, and considering different types of incentives (i.e., wage schemes) or risks (i.e., reputational risks). The researchers look at what “hidden” information may become apparent when these forecasts are converted to plot points on a graph.
The main hypothesis of the paper is that the skewness in the distribution of forecasts (meaning how much the actual data curve differs from an expected, normal distribution) contains information and is therefore able to forecast economic surprises. The researchers empirically establish that skewness in the distribution of forecasts present in a wide and global data set of economic expectations is able to significantly predict economic surprises. Forecasters behave strategically by making off-consensus forecasts as they possess superior private information, which is revealed via the skewness of the distribution of forecasts.
The researchers find that the importance of the skewness in the distribution of forecasts in predicting economic surprises increases steadily through time and versus the anchor bias (a cognitive bias whereby decision makers rely too much on pre-existing information). They also show that the prevalence of biases is related to the number of forecasters posting estimates per indicator, with more popular economic indicators experiencing higher levels of bias.
Their findings are useful for regulators, policy-makers and market participants who now have a model for understanding the meaning of skewness in the distribution of forecasts. Their research also may contribute to improving economic surprise indexes and their interpretation. Since the research has shown that anchoring bias is only one of many different biases in economic forecasting, benchmarks for the assessment of economic surprise models should be complemented with the suggested skewness measure. As regards the popularity effect identified by the researchers, it supports the usage of a weighted scheme versus an unweighted one in the construction of economic surprise indexes.
The paper was published in the Journal of Forecasting in January 2021 and may be downloaded from the Wiley Open Library.