The Air Safety Institute is in the process of updating and revising its report on instructional accidents, last issued in 2004. That’s the good news. The bad news is that, well, the news may not all be good.
For years we’ve believed that flight instruction enjoys an accident rate lower than that of general aviation as a whole, and less than half of that afflicting personal flights. That assessment is based on a pattern repeated throughout years of analysis for our annual Nall Report, which consistently found that less than 15 percent of noncommercial fixed-wing accidents happened on instructional flights.
Of course, a rate is the number of events per some amount of time, and there’s a potential disconnect whenever the event count and the time estimate don’t arise from the same source. In our case, the number of accidents comes from the NTSB, usually relying on their classification of the purpose of the flight to distinguish training accidents from those occurring in other pursuits. The amount of time devoted to flight instruction is estimated by the FAA as part of its annual General Aviation and Part 135 Activity Survey, which solicits samples of both individual owners and fleet operators to complete a questionnaire detailing how and why their aircraft were flown during the preceding year. Sophisticated but well-understood statistical wizardry is then used to adjust the results to help cancel out both deliberate imbalances in the proportion of eligible owners of a particular type who are asked to respond, and uncontrollable variation in the proportions who actually do. For example, the survey is sent out to 100 percent of Alaskan operators and owners of aircraft less than five years old, while other segments of the community may be sampled at less than 20 percent. As for response rates, in the past few years the owners of amateur-built aircraft have been among the most consistent respondents, with more than half returning their surveys.
Anyway, after said wizardry we usually find that more than 15 percent of fixed-wing time has been attributed to flight instruction. Since its share of accidents is lower than its share of exposure, it would follow that the accident rate for flight instruction should be less than the overall average. Personal flights, on the other hand, usually account for around one-third of fixed-wing flight time but more than two-thirds of the accidents, suggesting a substantial excess risk.
As it turns out, though, there is (you should pardon the expression) a fly in the ointment. Of course, the FAA’s estimates of flight time aren’t perfectly accurate, and the agency acknowledges that; each is accompanied by an estimate of the associated margin of error. As far as anyone can tell, though, those estimates are unbiased: They’re just as likely to be too high as too low.
That’s not true of the way the NTSB classifies accident flights. In the course of parsing the data to identify instructional accidents, Air Safety Institute staff noticed something odd: More than a quarter of all accidents on student solos had been classed as “personal” flights. That was hard to swallow; we’re in the habit of thinking that any authorized student solo is part of flight instruction, so we took a closer look.
As it turned out, some of those flights were not authorized; indeed, some of the worst accidents occurred on flights their instructors had specifically told the students not to attempt. Others weren’t actually solos; the accident narrative identified an instructor or another pilot even though that pilot wasn’t represented in the standard fields of the database. Even so, this sweep identified 275 accidents over a 10-year period that should have been counted as instructional (either dual or solo) but hadn’t been. That’s a 16-percent increase over the number officially reported by the board.
And that’s not all. Looking at accidents on two-pilot flights in which at least one was a CFI, we find another handful that weren’t listed as training flights but should have been. That cross-check is still in progress, but the final tally will probably be somewhere on the order of 40 to 60 during the same 10-year period. Many of these were either flight reviews or make-and-model transition training, two types of training in which there’s reason to suspect that accidents are particularly likely to be fatal.
Other types of training accidents are almost impossible to identify if they’re not labeled correctly the first time. If a private pilot flying solo loses control while practicing eights-on-pylons for his commercial checkride, for example, the true purpose of the flight will almost certainly never be known unless the pilot survives to tell investigators—and even then, there’s no guarantee this information will make it into the database. Checkride accidents pose an additional problem. The practical test is an essential step in earning that certificate or rating, but pass or fail, the examiner won’t sign the flight off as “dual given.” Should any accidents that result be put on the training industry’s tab? If not, where else would you count them?