Elevation

ATI had one in Japan back in 2011. I think it was a passenger injury or something like that which crossed the line and was rated as an accident. ABX did have one more than ten years ago. The difference is that ABX has logged many more departures than ATI over the same timeframe.

And, yes, there is a huge difference between accidents and incidents. That's why I'm referring to accidents. Accidents meet NTSB 830 standards and normalize for company to company differences in reporting, etc.

ABX hasn't had an accident in the last ten years. In 2018 they had a rolling 10 year average of 245k departures per accident. Those numbers extrapolated would currently have them averaging about 210k departures/accident over the last ten years. (This is due to the fact that ABX is flying a lot less than they used to so they are demonstrating fewer reliable departures as their busy history moves more than a decade into the past.) ATI is averaging about about 97k departures/accident. Atlas is around 84k departures per accident with 4 accidents in the last ten years.

For what it's worth, Omni has worse numbers in terms of departures per accident. They have a rolling average of about 54k departures/accident.

ATSG is probably making some pretty smart moves by moving safety personnel from ABX to areas that are experiencing rapid growth right now. That's the type of move that positions them for long-term sustainable growth. They're preempting problems which they can anticipate as street-captains get hired at Omni and the upgrade at ATI keeps getting shorter.

So, I mean, good work! There's a lot going wrong at all of our carriers, but this actually looks smart.

The blazing, red, neon-light problem at Atlas is the combination of relatively low departures per accident and a relatively high operational tempo. Omni has problems, but they fly very few departures. So their statistical mean tim to next accident is about 113 months (just under ten years). Atlas flies a lot more. Our mean time to our next accident is sitting around 19 months, assuming no growth in departures.

I checked these numbers against what is actually happening by comparing these methods to what very busy carriers fly. They're reasonably accurate with increasing accuracy as mean time to next accident decreases.

What I couldn't do was apply better modeling since I was comparing carriers that fly a lot with many data points to carriers that fly very few departures. So censoring the data was tough, and I couldn't appropriately apply any decent modeling to the numbers. Simple averages and trends are all that I trust from the publicly available data.