3 ms·
Unpredictability in Economic Analysis, Econometric Modeling and Forecasting
- jruohonen 3y agoA good primer to Hendry's work: "The first entails that conditioning information does not alter uncertainty, so that the unconditional distribution is the best basis for all three activities, which are therefore on an equal footing of ‘uninformativeness’. The second adds the possibility that even when the distributional form is known, either unconditionally or conditionally (so otherwise there would be some predictability), an outcome can be highly discrepant at an unexpected time, and thereby impose substantial costs. Such outliers do not persist, but are usually in dfferenced variables (like rates of return) so can permanently change a level. The third concerns unanticipated shifts in distributions, which do persist, as with location shifts."