26 February 2009

Do Plane Crashes Happen in Threes? - Sometimes, Yes

Note: updated version of article published 24 July 2014

Since AirSafe.com was launched in 1996, the site has tracked fatal events and other significant events involving airline passengers. When these events occur, especially if two occur a just a few days apart, I sometimes get the "Do bad things like plane crashes always happen in threes?" question asked by visitors to the site, by members of the media, and by others. I used to just dismiss the question out of hand because events like plane crashes, especially those involving passenger airliners, are very rare, and the circumstances are usually very different for each crash, often involving different airlines, different aircraft types, and even different countries.

Although it's easy to reject the original question, it is quite legitimate to ask a related question about how frequently groups of rare events occur over a relatively short period of time. One day, just for fun, I turned the "things happening in threes" question into something that could be analyzed systematically using the information within AirSafe.com. I changed the general question into the following specific question: "How frequent are sequences of three or more fatal or significant aviation safety and security events where the time between events is ten days or less?"

For the purposes of answering this question, I limited the data to those events that are regularly tracked by AirSafe.com. These would be plane crashes or other airline events that kill at least one passenger, or other events that AirSafe.com considers to be significant with respect to aviation safety or aviation security (see http://www.airsafe.com/events/method.htm for a detailed definition of a fatal or significant event). Multiple events due to the same cause (for example, the four crashes associated with 9/11) were treated as one event.

Significant events that don't kill anyone sometimes attract more media attention than the average plane crash. The January 2008 ditching of a US Airways A320 in the Hudson River in New York was one example. It was very dramatic, it got a huge amount of media exposure, and no one was killed.

A review of the AirSafe.com records from 1996, the year AirSafe.com was launched, to February 2009 revealed some interesting facts:

* With the exception of 2007 and 2009, every year since 1996 included at least one sequence of three fatal or significant events that were separated by no more than ten days In fact, there was a sequence of five significant (but nonfatal) events in January 2008.

* There were seventeen sequences of three or more events that were separated no more than ten days. Three were sequences of five events, three sequences had four events, and the other eleven consisted of three events each. Correction: Since this article was published, a previously overlooked fatal accident from 24 August 1999 was added, and this led to an additional sequence of three for that year.

* Most of the fatal and significant events tracked from 1996 to the present were not part of any sequence of three or more events.

* Well known fatal events that were a part of one of these sequences include the Swissair MD-11 crash in 1998, the Concorde crash in 2000, the August 2006 crash of a Comair jet, in Lexington, KY, and the August 2008 crash of a Spanair MD83 in Madrid.

* Well known events that were not a part of one of these sequences include the ValuJet and TWA Flight 800 crashes in 1996, the Alaska Airlines crash in 2000, and the four crashes associated with 9/11.

* There have been five fatal or significant events in a 70-day period between 20 December 2008 and 25 February 2009, but only two of them have been less than ten days apart.

For twelve of the past 13 calendar years (1996 to 2008), there has been at least one grouping of three or more fatal or significant events that occurred over a relatively short period. At the same time, no information has come about in the investigations of any of those events that indicates that there was any sort of connection among them, or any evidence that the earlier events in the sequence made a later event more likely.

After reviewing the facts, I no longer say that plane crashes don't happen in threes. Since at least 1996, they have happened in threes, in fours, and in fives. Let us all hope that there are no sixes and sevens in our future.

Review AirSafe.com Events by Year
1996, 1997, 1998, 1999, 2000, 2001, 2002
2003, 2004, 2005, 2006, 2007, 2008

4 comments:

  1. You only listed the events, you also need a probability model so that you can actually compare whether these events in three etc. are statistically significant sequences, one model would be a hazard rate model which tests the time between events. Can you do that? Before I see such an analysis, I still think the events of three etc. are random events.
    - cheers, arent greve

    ReplyDelete
  2. There's a posibiltiy that airline accidents and incidents are more common than reported. But when there is a huge collision that makes the news, other incidents come to light and are reported.

    ReplyDelete
  3. I need to see more of a correlation. Random accidents happen all the time.

    ReplyDelete