Rethinking Human Error Statistics in Aircraft Accidents 

By Robert Baron, Ph.D

The Aviation Consulting Group

March 2010

 

It seems like every time I read an aviation safety-related research study there is a ubiquitous statistic that always grabs my attention. I am referring to the “70%-80% of airplane accidents are attributable to human error” statistic. This stat grabs my attention not because of its seemingly high estimate; in fact, it surprises me that it is unrealistically low. It might be time to start acknowledging that almost every airplane accident has a human error component. I posit that 95%-99% of airplane accidents are attributable to human error. My statistical estimate is based on facts which will be discussed shortly.

Airplanes are so reliable and safe these days that in almost every case it is human error that begins, contributes to, or ends an accident sequence. The reason why I do not go as far as to say that 100% of accidents are caused by human error is because there are a few exceptions. The most recent example of this would be the Airbus A-320 that was ditched successfully in New York’s Hudson River. The ditching was initiated due to a bird strike which caused both engines to flame out. In this particular case there was no human error. In fact, through Captain Sullenberger’s heroic acts and superior airmanship, there were no lives lost and only minor injuries. This accident, however, is the exception and not the norm. Most other accidents are the concatenation of human errors.  

To exemplify the point, I set out to conduct an informal study to investigate the contributing factors in a sample of airplane accidents. While not a true scientific study, the sample drawn was a fairly good representation of overall accident causes and provides enough meaningful data for the reader to draw his or her own conclusions. The study used the following data and methodology:

1.   The National Transportation Safety Board (NTSB) aviation accident database was used to provide factual data.

2.   The sample consisted of all Major Investigations conducted by the NTSB for the 10 year period 2000-2009.

3.   All Major Investigations in the sample had at least one fatality and ranged from private aircraft to large air carriers.

4.   There were a total of 29 accidents.               

5.   Three of the most recent accidents did not have a Probable Cause issued as of the date of this writing. However, preliminary data, where available, were used to determine whether human error may have caused, or contributed to, the accident. Four accidents were acts of terrorism and were excluded. 

6.   Twenty-three accidents were usable for this study.

7.   The accidents were additionally categorized as;

a.      No Anomalies (nothing technically wrong with the aircraft), or

b.     XXXX Malfunction (something technically wrong with the aircraft).

 

Results and Discussion

The results appear to support my soft hypothesis that the majority of aircraft accidents are due to human error. Out of the total of 23 usable accidents, each one had a human error component (or components). In fact, none of the accidents were caused by an aircraft malfunction due to reliability or airworthiness issues not propagated by an erroneous human intervention. Refer to the table below.

For clarity I am going to provide some additional information regarding the way the data were categorized in the table. Accidents that were excluded were done so due to the fact that those accidents did not yet have a Probable Cause issued by the NTSB or they were acts of terrorism. The remaining accidents were decomposed into two broad categories; (a) those accidents with no aircraft malfunctions, and (b) those accidents with aircraft malfunctions. The distinction between these categories is important because, as the results indicate, the majority of accidents occurred when there was absolutely nothing wrong with the aircraft. On the other hand it should be pointed out that in the accidents that did occur where there was a malfunction onboard the aircraft, the pilots should have either been able to regain control of the aircraft—or—the pilots were unable to regain control of the aircraft due to a maintenance error. If this is the case, then although technically there was nothing the pilots could have done to recover the aircraft, there were still human errors propagated by the maintenance crew which eventually manifested into an unrecoverable airborne malfunction.

The results indicated that in each of the usable accidents human error was a primary factor. Thus 100% of this sample of accidents could be considered to have been caused by human error, whether by maintenance activities or in the operation of the aircraft. Maintenance errors tend to be more distal (a delay between the error and the actual accident) whereas operational errors tend to be more proximal (happen very quickly as a result of poor judgment or decisions by pilots). In either case there was some erroneous human action that precipitated the events leading up to the accident.

There was one major limitation to this informal study; there was a very small sample drawn from the NTSB database. Time constraints did not allow for a larger sample. Thus if other clusters of accidents were analyzed (say all accidents for the year 2005) the results may differ somewhat. However, in this particular study it was shown that 100% of the accidents had a human error component. In another, larger sample, it is likely that the rate would be somewhat less than this figure but probably not by much. The figure would most likely remain in the 95%-99% range that I proposed at the beginning of this article.


Summary and Recommendations

With some relevant data to support the theory that almost 100% of aircraft accidents are attributable to human error the next logical question would be what can be done to reduce this statistic? Indeed, this is a tricky question. Consider the fact that in the last century of powered flight we have been able to significantly reduce the amount of aircraft accidents attributable to pure mechanical/technical causes to nearly zero. On the other hand, accidents attributed purely to human error have grown exponentially and continue to be one of the last remaining impediments to realizing a near-zero accident rate. Speaking of accident rates, one should understand the potential limitations of the statistics used in this article. Let’s say that in a certain year there were 10 commercial airline accidents and each was caused by human error. Twenty years later a study is conducted that shows that there was a significant reduction in commercial airline accidents (just one for the whole year compared to the previous ten). This would indicate a 90% decrease in accidents over time; however the one accident that did occur was due to human error. Thus, even though the rate decreased to just one accident, that one accident was caused by human error and therefore, for that year, each accident (1) was attributable to human error (100%).   

The hard truth is that human error can never be completely eradicated and statistics will continue to point towards human performance deficiencies in the events that lead up to aircraft accidents. A maxim that I like to profess is, “To err is human but to recognize and recover is divine.” This acknowledges that as long as there are humans involved in the operation of an aircraft there will always be plenty of opportunities for error. It should not be construed that this makes it acceptable to commit an error. It means that as long as we acknowledge that erroneous actions are inherently human we should take steps to mitigate the impact of errors should they occur. The difference between an accident that happened and one that did not is not just luck. In the accident that happened there was a chain of events that managed to breach the system’s built-in defenses. These defenses included such things as regulations, policies, training transfer, operating specifications, task cards, and the use of standard operating procedures.

Countermeasures are being deployed and so far they seem to be having a positive effect. From an organizational standpoint we are seeing a considerable increase in “soft skills” training courses worldwide (depending on your location some courses may be mandatory while others are highly recommended by your regulator). These soft skills awareness courses specifically focus on human performance issues and error reduction strategies. Human factors (HF), crew resource management (CRM), and threat and error management (TEM) are examples of these types of courses. Additionally, a safety management system (SMS) will help to provide a “central safety hub” for human performance issues with the added benefit of high-level management accountability.

On an individual level I believe that we are making good progress in increasing our awareness of the fact that we, as humans, can and do commit errors. Hopefully, with this knowledge, there will be that little extra attention to detail, that one extra look before closing it up, or the prudent decision to go-around when confronted with an unstabilized approach. On the other hand we also need to understand that there will always be organizational and cultural influences that may cause an otherwise rational aircraft maintenance technician or pilot to do something he or she would not normally do, usually due to the confluence of pressure, unrealistic deadlines, and stress. Needless to say, there can be a very fine line between being a company person doing whatever it takes to get the job done and a being a rational, expert decision maker who is fully aware of the inherent risks of pushing the safety envelope. 

In closing, we do need to understand that although most aircraft accidents are due to some form of human error, aviation is still by far the safest mode of transportation. Yet, each time we read an accident report we tend to see the same types of recurring causal factors; unprofessional behavior, deviations from standard operating procedures or maintenance manuals, skipped steps, improper decisions, poor judgment, etc. These factors are typically compounded by fatigue (which is a complex issue and may be propagated at both the regulatory and organizational levels as well as at the individual level).

Remember that although human error may be inevitable, it still does not make it acceptable. You may be the last line of defense in deflecting the error trajectory that results in a serious accident. Always maintain situation awareness (know your surroundings and status of tasks). Think about your actions and their corresponding safety implications (will what I do compromise the safety of the aircraft and passengers?). Remember that just because everyone else is doing it that doesn’t always make it right (i.e., skipping functional or operational checks or not using the checklist). Always be a true professional and abide by the regulations (i.e., no non-essential chatter when the sterile cockpit rule is in effect). I could go on and on but I think you get the point.    

Modern day aircraft are highly reliable machines that come from the factory ready to give many years of safe, reliable transportation. It is only when the human gets involved with the maintenance or operation of the aircraft do we start to see the types of recurring accidents that are posted every day to the NTSB website. With a concerted effort we can continue to diminish the accidents attributable to human error and make flying even safer, both now and in the future. Since you are the human, you can make the difference!

The NTSB data that were used for this paper can be viewed by clicking here