Improving Airline Risk Estimates


The AirSafe Journal
vol. 1 num. 15 - 1 January 2001
(Revised 11 January 2001)


In spite of the amount of media coverage and industry attention given to aviation accidents and aviation risk, there is no single standard for summarizing aviation safety events. Some of the most common ways of doing so involve the use if risk figures that express a combination of the frequency of an event and the severity of that event.

Until now, the summary statistic most widely used by AirSafe.com has been the fatal event rate, defined as the number of events involving at least one passenger fatality divided by the number of underlying airline flights. While this statistic provided a way to consistently compare airline or aircraft model risks, it had the disadvantage of weighing all events equally, regardless of the proportion of people killed. An event killing one out of 300 counted as much as one that killed 300 of 300 passengers.

This new method of computing fatal event rates, based on the work of Prof. Arnold Barnett of MIT, sums the proportion of passengers killed and divide that by the underlying number of flights. The sum of these proportions would be the full-loss equivalents among the underlying flights, where the full-loss equivalent for a given flight is the proportion of passengers who did not survive the flight. For example, as of the end of 2000 the 757 had a total of five fatal events. In three of the five events, some passengers survived. The sum of the proportions of passengers killed was 3.4. Given the total of 8.7 million flights, that implied a fatal event rate of 0.39 per million flights. The previous method would have give a rate of 0.57 per million flights.

This new risk measurement provides the passenger mortality risk per randomly chosen flight. By weighing individual accidents by the proportion of passengers killed, this provides somewhat more information than the previous method. The previous method, while easy to compute and to understand, provided no insight into the likelihood of passenger deaths in a group of fatal events.

The page providing the fatal event rates per aircraft model and the one providing fatal event rates by airline has been revised to use this new risk measurement. Other pages in the site that provide fatal event rates will be adjusted to incorporate this new method.


Related Article
Passenger-mortality Risk Estimates Provide Perspectives About Airline Safety
Arnold Barnett and Alexander Wang; Flight Safety Digest; April 2000

Definitions
Full Loss Equivalent: This is the sum of the fractions of passengers killed for each fatal event. For example, 50 out of 100 passengers killed is an FLE of 0.50, 1 of 100 would be a FLE of 0.01. The rate is computed as the FLE divided by the number of flights in millions.

Fatal Event: Any circumstance where one or more passengers die during the flight from causes that are directly related to a civilian airline flight. These events include sabotage, hijacking, or military action and exclude cases where the only passenger deaths were to hijackers, saboteurs, or stowaways. Every airliner included in the list have at least one million airline flights.

AirSafe Journal - Issue 15
http://airsafe.com/journal/issue15.htm -- Revised: 24 May 2015