Investigating Airplane Accidents Over the Past 110+ Years

Meena Aier
Data Girl
Published in
7 min readMar 17, 2019

Airplanes have gotten safer, haven’t they? Yes, consistent improvements in technology have made flying safer on the whole. However, airline business models increasingly focus on achieving efficiencies of scale by transporting a greater number of passengers per flight. This has dramatically increased the cost of every crash — not just in $ terms, but more importantly, when we consider the number of human lives lost.

The Ethiopian Airlines crash last week brought a renewed focus on aircraft manufacturers’ as well as regulators’ responsibilities in maintaining the highest standards of aviation safety. The days following the crash saw all major regulators across the globe grounding Boeing’s 737 Max 8 jets, which as per analyst estimates, could potentially cost all stakeholders involved billions of dollars over the next few months. The cost of 157 lives lost is incalculable, and far more acutely felt.

The fact that air travel has gotten safer is commonly known — but I wanted to take a quick look at the underlying data, and try and determine how some key trends have evolved over the past century. Thankfully, there are some great open source datasets available, which make it possible to have a fantastic data frame that spans over 111 years. Towards the end of this post, I note all of the resources I have used to collect, clean, analyze and visualize this data.

First things first. Let’s take a look at some aggregate numbers.

How many plane accidents have occurred since 1908? Nearly 5,800. This includes commercial, military and private flights.

How many lives have been lost in these accidents? Over 115,000.

What was the deadliest year for plane crashes? 1972. This year witnessed 105 plane crashes, which resulted in nearly 3,000 deaths.

The graph below visualizes the number of plane crashes and fatalities over the past 111 years, with some annotations.

The left axis plots the number of accidents annually (grey area chart). The right axis charts fatalities every year (red line graph).

I will return to the question of safety again, towards the end of this post. But from an overall number of accidents and fatalities perspective, air travel safety seems to have a markedly improved. Now, turning to specific trends — are there certain operators, and aircraft manufacturers that experience a greater number of accidents? In order to answer these three questions, I plotted accidents and fatalities across top 25 operators and manufacturers.

For operators, Aeroflot (i.e. Russian Airlines) has clearly witnessed the most number of accidents (260 in total), and fatalities (over 9,000). The 70s were a particularly rough decade, where 86 Aeroflot planes were involved in crashes. This is followed by U.S. Air Force planes (178 crashes and over 3,700 deaths).

The left axis plots the total number of accidents per airline operator (Grey bars). The right axis charts fatalities associated with these accidents (red plane icons).

In terms of manufacturers’ models, the legendary Douglas DC-3 planes have been involved in over 300 crashes till date. Most of these crashes were concentrated in the 15–20 years following World War II. However, I will note that DC-3 truly revolutionized commercial flying, is considered to be a very rugged plane, and dominated passenger flights in the post World War II years. So, on a more nuanced level, I would take these factors into account and accordingly adjust my interpretation of this datapoint, to indicate that this model was the most advanced of its time, and was built decades before technological progress brought about greater safety to air travel.

The left axis plots the total number of accidents per aircraft manufacturer (Grey bars). The right axis charts fatalities associated with these accidents (red plane icons).

For those of you that might be looking for additional details around air crashes, airline operators and aircraft manufacturers, I’ve designed an interactive visualization, which can be accessed here:

This visualization takes into account the top 100 Airline operators by the number of crashes over the past 110+ years, and further breaks them down into aircraft manufacturer. The innermost circle aggregates all the crashes across top 100 airlines (2,124 accidents in total). The second circle breaks them down by airline operator. For instance, hovering over the light blue section will indicate that the Aeroflot planes were involved in the most number of crashes. The outermost circle further segments these crashes by plane manufacturer for each airline operator. So if I were to hover over the green section, I can tell that about 1% of Aeroflot crashes can be attributed to the Yakolev Yak-40 plane models. Here’s a screenshot of what this looks like in the interactive visualization:

A screenshot of the interactive visualization, which can be accessed at http://rpubs.com/datagirl/476810

Now, returning to the key question: Has air travel become safer? I looked at the overall number of incidents and fatalities, and could tell that from an overall numbers perspective, it does look like the number of incidents have been steadily falling over the past century. This holds true even when I examine a relative measure — i.e. percentage of all flights and all air passengers that resulted in crashes and fatalities. Unfortunately, uniform data for global air traffic isn’t freely available for years prior to 1970, and has been updated only until 2017. However, this still gives us data for the past 45+ years. The graph below clearly shows two things — in 1970s, the probability of crashes and fatalities were fairly low to begin with (~0.00020% of air traffic resulted in accidents and fatalities). Over the past 45 odd years, this number has further dropped by nearly 100%. As of 2017, the fatalities/air traffic proportion had dropped to 0.0000006%. In other words, there were 3 fatalities for every 5 million people that travelled by planes. Similarly, the accidents/air traffic proportion has decreased exponentially to 0.000005%. Simply put, on average, for every 200,000 planes flow, 1 end ups in an accident.

The left axis plots percentage of flights that are involved in accidents per year, from 1970 t0 2017 (Grey line chart). The right axis charts percentage of all passengers that died on account of plane accidents per year, from 1970 to 2017 (Red step line chart).

These are considerable improvements. Like I noted, consistent improvements in aircraft technology have made our skies safer. However, technology improvements and rising standards of living have been associated with a greater demand for air travel. Airplane operators (and manufacturers) have responded with business models that seek to achieve ever increasing efficiencies of scale. This has translated amongst other things, into a demand for aircrafts that are highly fuel-efficient, while accommodating greater number of passengers per flight. As a result, while the overall number of accidents have decreased, the cost of each accident has steadily increased. When I specify costs — I mean costs of human lives rather than monetary costs. This is not to minimize monetary losses for various stakeholders, but rather to clarify that monetary costs are not a focus of this analysis.

The graph below charts a very simple measure — the average number of fatalities per accident, on an annual basis from 1908 until 2019. The grey dotted trend line is a simple linear measure, which indicates that the average number of fatalities has been steadily increasing these past 111 years. Of course, the story is a little more complicated than that. If we were to look a little more closely, we can tell that around 1990s, the average number of fatalities appears to have been gradually dropping. However, in the past 5 years or so, this trend seems to have reversed, with steep spikes in 2014 and 2015, followed by another steeper spike in 2018 and 2019 (until March that is). In fact, assuming we don’t have other crashes this year (hopefully), 2019 will have been the deadliest over the past 111 years for average fatalities associated with air traffic accidents.

While it is too early to tell how this trend might develop over the next five years, it is nevertheless concerning. It raises important questions around how air travel is regulated, and what passes for an acceptable degree of risk when it comes to air traffic (in an ideal world, we would have zero risks — an ideal state that should form the basis of our regulators’ and aircraft manufacturers’ objectives). As the Boeing 737 Max-8 investigation develops, it will be interesting to watch how regulators (especially major ones such as FAA) respond, and what (if any) measures they take to ensure that air travel is afforded the highest levels of safety.

Notes:

First, I must send out a massive thank you to planecrashinfo.com — an open source website that has carefully curated airplane accidents from 1908, and continues to be updated regularly. I will also thank hocnx from Kaggle, for providing the basic building blocks of Python code to scrape, clean and organize this data in a machine readable format. I used World Bank’s open datasets to gather data around total air traffic from 1970s until 2017. I used Python to collect this data, and a combination of R and Tableau for analysis and visualization. I designed the interactive visualization in R using the fantastic SunburstR package built by Kent Russell.

A quick disclaimer: I am not an air traffic expert, and most of my plane crash knowledge comes from analyzing this dataset and reading articles while analyzing this data. As such, if I have overlooked something important, please do let me know via comments or email.

--

--