Log in to access the full article.
Abstract Crashes, measured as strong price decreases, are sometimes difficult to reconcile with historical events. This can be explained by the fact that a price variation will have a greater negative impact in a stable financial context than a similar variation during a highly volatile period. For example, French stocks decreased painlessly by 16 per cent in August 2002, whereas a similar fall in January 1882 led to the failure of several brokers. Market volatility was very low at the end of the nineteenth century, whereas investors are now used to dealing with large price movements. A fall of 16 per cent was much more of a shock in 1882 than it would be today. To control for the instability of the volatility, a new method for identifying crashes is proposed. Each price variation is measured in numbers of standard deviations of the preceding period. These adjusted variations can then be ranked to identify the worst market crashes. This method is tested on four long-term series. A better match between crashes and historical events is achieved than with pure price variations. This improved matching brings new insights to several historical debates.