Identifying multiple outliers in heavy-tailed distributions with an application to market crashes

Heavy-tailed distributions, such as the distribution of stock returns, are prone to generate large values. This renders difficult the detection of outliers. We propose a new outward testing procedure to identify multiple outliers in these distributions. A major virtue of the test is its simplicity. The performance of the test is investigated in several simulation studies. As a substantive empirical contribution we apply the test to Dow Jones Industrial Average return data and find that the Black Monday market crash was not a structurally unusual event.