Stylized facts of asset returns are widely established. Among these are e.g. non-normality,
volatility clustering and high persistence. Another important recurring aspect is the existence of
financial exuberance, often interpreted as an explosive price bubble. Exuberance periods consist
of two parts, (i) the temporary explosive price period and (ii) the mean-reverting reverse period
of market correction. We provide a comprehensive analysis of exuberance periods by analysing 30
markets from different categories over a time period of fifty years. We cover international stock
markets, the US housing market, Gold, Silver and Oil as well as the Bitcoin prices. Overall,
we find 143 exuberance phases and document evidence on important characteristics like (i)
durations of explosive phases, (ii) collapse duration and behaviour during market correction
phases, (iii) magnitude of autoregressive parameters during exuberance and market correction
and (iv) distributional characteristics like fat tails and shifts in the innovation variance. We
classify the cross-sectional results on 143 explosive phases into relatively low, middle and high
values. We test a number of common beliefs in the literature and provide new insights into
typical empirical properties of explosive prices and their collapse. Our results indicate significant
discrepancies with typical settings in the literature. Empirical explosiveness is much milder and
collapse phases are in most cases smooth rather than abrupt. Moreover, prices do not revert
back to the initial value, but stay significantly above. The simplified view that prices are
strongly exploding with a full collapse in short time is not supported by our results. Duration
dependence modelling reveals that the length of the explosive phase is positively affected by
economic growth, while the collapse duration is only driven by the length of the preceeding
explosive phase in a positive way. Finally, we offer empirically relevant parametrizations for
data generating processes and study the consequences for the empirical performance of popular
bubble detection and date-stamping procedures.