We introduce a new definition of probabilistic forecasts’ coherence based on the divergence between forecasters’ expected utility and their own models’ likelihood function. When the divergence is zero, this utility is said to be local. A new micro-founded forecasting environment, the “Scoring Structure”, where the forecast users interact with forecasters, allows econometricians to build a formal test for the null hypothesis of locality. The test behaves consistently with the requirements of the theoretical literature. The locality is fundamental to set dating algorithms for the assessment of the probability of recession in U.S. business cycle and central banks’ “fan” charts.
Originalsprog
Engelsk
Udgivelsessted
Aarhus
Udgiver
Institut for Økonomi, Aarhus Universitet
Antal sider
37
Status
Udgivet - 21 sep. 2018
Serietitel
CREATES Research Papers
Nummer
2018-23
Forskningsområder
Business Cycle, Fan Charts, Locality Testing, Smooth Transition Auto-Regressions, Predictive Density, Scoring Rules and Structures