IEEE - Aerospace and Electronic Systems - August 2023 - 10

Airbus A32x Versus Boeing 737 Safety Occurrences
Table 1.
Mean (m), Standard Deviations (s), and One-Tailed Two Sample t-Test p Values for Annual Reduction in Rates of Hull
Loss (HL) and Fatal Hull Loss (F-HL), in Terms of Relative Annual Reduction (Dr/r) and CAGR
Metric
Rate
HL
Dr/r
F-HL
HL
CAGR
F-HL
Manu ms pKS
Boeing
Airbus
Boeing
Airbus
Boeing
Airbus
Boeing
Airbus
Kolmogorov-Smirnov testfor normality critical value 0.26.
0.012
0.051
0.025
0.073
0.004
0.036
0.016
0.046
0.020
0.067
0.020
0.081
0.008
0.016
0.009
0.027
0.023
0.021
<0.001
<0.001
0.10
0.16
0.17
0.19
The conclusion that suggests Airbus is performing better
than Boeing in terms of safety mirrors previous studies
[14], [15].
A reader may ask the question, what would happen ifthe
" anomalous " data points of the Boeing 737MAX were
removed? The findings for fatalities relative to all safety
occurrences (see Figure 8) remain statistically significant in
Airbus' favor, while fatalities relative to accidents
(Figure 13) becomes statistically insignificant, still in Airbus'
favor. So, at best, one could say that on some metrics
Airbus and Boeing are level, while in other Airbus is outperforming
Boeing. However, the case to exclude these as so
called " black swan events " [30] is not justifiable. In fact,
while the total of this data set represents the reported occurrences
for Boeing and Airbus end users, the Boeing
737MAX occurrences are events, which were more associated
with the manufacturer than the end user. To yield unbiased
comparison between the Airbus and Boeing data
requires all of it to be considered. Furthermore, prior to
2018, Boeing 737 aircraft were experiencing 1.55 fatal hull
loss accidents per year, while up to 2020 this only increased
to 1.65 fatal hull loss accidents per year. This variation is
well within a traditional uncertainty. As such, claiming that
the two Boeing 737MAX accidents are " black swan " events,
and hence should not be factored in is not justifiable scientifically.
In fact, this would constitute " cherry picking, " a common
technique ofscience misinformation [31].
The important context of these numbers is the population.
To be clear, the total number of Boeing aircraft
in operation during all years of this study was
significantly greater than the number of Airbus aircraft.
As noted, Boeing 737 aircraft performed 4.2 times as
many departures compared to Airbus A32x aircraft in
2006, which reduced to 1.7 times by 2019. The
10
purpose of using the rate per million departures is to
remove this population skew. Just looking at the rates
from 2006 to 2008 shows that Airbus and Boeing were
on par. The average age of the Boeing fleet might play
a role in this. However, the age of the fleet is an
objective feature of that fleet. Furthermore, it should
be noted that for technical issues, such as mechanical
fatigue, issues are associated more with a period up to
the first D check than they are for older airframes [22].
A core assumption in the work is that the ICAO occurrence
dataset as a sample is representative of the population.
There is no reason to doubt the validity of this
assumption. This is supported by the fact that Annex 13 to
the Chicago Convention requires that Contracting States
report to ICAO information on all aircraft accidents,
which involve aircraft of a maximum certified takeoff
mass of over 2250 kg [32]. Further to this, all serious incidents
involving aircraft with a maximum certified takeoff
mass ofover 5700 kg are also collected.
The reliability of the data, in terms of completeness
and correctness, as available from the ICAO database is
assumed. There is no independent way to verify this, and
a study by the European Spreadsheet Risks Interest Group
reported that less than one in ten spreadsheets are error
free [33]. Given most errors are computational errors, a
database as presented is taken as complete and correct.
Furthermore, there is no reason to suspect that errors
would favor Boeing or Airbus over the other.
The results presented here in terms of statistical inferences
are the result of the aggregate of the data over the
period from 2008 to 2019. Those statistical inferences can
only describe if observed differences in the proportions
between the samples are statistically significantly different.
Why any confirmed independence between Airbus
IEEE A&E SYSTEMS MAGAZINE
AUGUST 2023

IEEE - Aerospace and Electronic Systems - August 2023

Table of Contents for the Digital Edition of IEEE - Aerospace and Electronic Systems - August 2023