A problem with these types of statistics is the use of headline-catching parameters, together with the implication that these represent the future.
The bean-counters become body-counters; non fatal accidents are tolerated until there is a death.
This approach doesn’t relate to the overall risk involved. Past risk is historical where data includes both frequency of event and outcome severity (risk definition), but the future at best might only project the past frequency without any idea about outcome severity; accidents with the same aircraft type could range the loss of max pax load, to the crew only in a freighter. Thus future safety is not the same as historical views, however measured. EE #10
It more important to consider what is being done and will continue to be done to address the hazards identified by accidents, and projected hazards of the future with forethought.
A classic example is where safety initiatives - LOC training - have been based on body count, whereas the projected exposure of a much larger number of overruns should be of greater concern - the likelihood of future harm vs a low historical value based on fatalities.
The Airbus approach to Statistical Summaries provides a more balance view of the state of safety and relates to different technologies.
https://www.airbus.com/content/dam/c...-1958-2017.pdf
There could be an interesting debate re accident rate vs aircraft generation when placing the Lion 737 MAX accident in comparison to ‘new’ generations. Airbus define this version as level 3, but is the design philosophy still level 2 ?