Expectations regarding future oil prices are of great importance for a variety of
economic and financial applications. The state-of-the-art methodology based on
a Gaussian affine term structure model provides monthly risk-adjusted financial
market expectations (Hamilton and Wu, 2012, 2014). The term structure model
outperforms a number of competitors, see Baumeister (2023). We investigate the
potentially time-varying performance of these oil price expectations and those from
the Energy Information Administration (see Garratt, Vahey, and Zhang, 2019).
It turns out that the superiority of the term structure model over a simple nochange
benchmark is not only time-varying, but also characterized by phases in
which the ranking among the competing expectations are reversed. Importantly,
these phases are characterized by locally explosive oil price behavior and subsequent
crashes (see Pavlidis, Paya, and Peel, 2018). To this end, we study the role of
established real-time oil price shock measures (see Kilian and Vigfusson, 2011, 2013)
alongside a newly proposed indicator which is obtained from a recursive monitoring
test statistic against explosiveness (see Phillips and Shi (2018)). From a perspective
of conditional expected performance (see e.g. Granziera and Sekhposyan, 2019),
we investigate the predictability of the loss differential by different shock series.
Our results underline the importance of the new indicator reflecting temporary
exuberance and subsequently collapsing prices. Our findings have consequences for
the literature on oil price speculative bubble testing and beyond.