Hacker News Reader

Comment by ergocoder | Hacker News Reader

This is off topic. But I've been skeptical of the commercial aviation accident risk compared to cars.

I'm not saying cars are safer. But using accident count per mileage seems a bit odd, especially when the most risky parts of flying are taking off and landing.

A plane travelling 400 more miles doesn't incur much more risk. But a car travelling 400 more miles incur a much bigger risk (e.g. fatigue, car malfunctioning). So, of course, the stats for planes is going to look much better.

I wonder if anyone has an insight into why we are using this metrics, and what is a better metrics?

View on HN

The relevance of the metric depends on what you're using it for. If you decide whether to drive or fly to a destination based on how safe it is, risk/distance is a perfectly reasonable metric. Since flying is a means of transportation, it seems reasonable to judge the risk relative to the amount of transportation you get out of it.

That doesn't discount the fact that most of the risk is in take-off/landing, so if you use fleet-average statistics you'll bias the risk of short flights low and long flight high. Air safety studies do also use additional metrics like accidents/hour and accidents/segment, but those are mostly useful for comparing between different types of aviation.

I'm not sure how your comment addresses the core problem that risk/distance is not a reasonable metric if long-distance driving is less risky than short-distance driving, and you apply the short-distance driving risk to long-distance driving.

As a passenger on either it doesn't matter where the risk is concentrated . I am only interested in the outcome of how safe my journey is going to be.

As a technology such differentiated metrics helps to understand aspects of flying which is why broken metrics are used by the industry in more focused forums.

But that's exactly my point. As a passenger, it matters if the calculation of how safe your journey is is wrong. And the calculation is wrong if it doesn't take into account how long the drive is.

Wikipedia has a pretty handy table of metrics here: https://en.wikipedia.org/wiki/Aviation_safety#Transport_comp...

Ignoring the silly ones (Space Shuttle, Skidiving and Paragliding), we see that Motorcycles are by far the most dangerous mode of transport, across all metrics. Buses do very well, only losing out to Aircraft on distance.

When you look at Aircraft by number of journeys, the risk is significantly higher - nearly 3 times higher than car journeys!

That said, there are othe factors to consider here: these metrics only track _deaths_. So non-fatal incidents (of which the majority involving cars are) aren't considered.

Thanks for sharing it.

It seems, if the risk is based on the distance, then air is 60x safer. But, for per journeys, air travel is worse. For per hours, the air travel is only 3x safer.

The popularized concept of air travel's risk gave me the impression that air travel was sooo much safer than car. smh. If you asked me a month ago, I would say air travel should be, at least, 1000x safer than car.

Read the paragraphs below the table.

I read it, but I don't know what you are trying to convey. Please elaborate. Thank you.

The same sort of problem occurs on the other end when you try to compare the risk of driving vs walking. Cars are 5 times faster in the city and 20 times faster on the highway than walking. The trips are almost always shorter on foot, you drive to the far away big box store but would walk to the neighbourhood store. Better metrics than risk per unit of distance would be risk per unit of time spent or risk per trip.

I would point out that in a similar way few people drive between continents but flight is common.

It's a useful metric because, when comparing auto and plane safety, consumers are usually deciding between taking a plane and a car for the same distance. We could give them risk as a function of distance by taking the increased takeoff/landing risk into account, but generally people only compare the two for fairly long trips.

For the context of driving long distance, that makes sense.

Do the car incident statistics only take into account long distance trips?

For cars they count all passenger-miles equally. Car accidents are much more strongly correlated with distance than aircraft accidents, but I don't know the actual risk factors - these are all heuristics for much more detailed models that I'm sure some academic is building.

What is eye-opening here is that: Aviation industry insurers base their calculations on the deaths per journey statistic while the aviation industry itself generally uses the deaths per kilometre statistic in press releases

from https://en.wikipedia.org/wiki/Aviation_safety#Transport_comp...

So, insurers don't wanna use this death per distance.

Perhaps a better metric would be accidents per unit of time.

Or accidents per departure and arrival. Since take-off and landing are the two most dangerous times, gauging accidents based on how many times one leaves and arrives could provide a more direct way of measuring how dangerous any one departure/arrival actually is?

I'd say it's accidents per person/hour. And I've read somewhere that the figures used to be similar for cars and airplanes (though aircraft safety is constantly improving, cars probably less).

Per passenger might be off because a passenger doesn't affect the risk. The number of flights vs. the number of trips might be more fair.

Accidents? Or fatalities? Cars today are certainly safer than cars 50 years ago.

Cars have been improving safety significantly over time as well. However, the fatality rates seem to be significantly better for airlines.

https://www.bts.gov/content/fatality-rates-mode

For the sake of argument, assume cars travel an average of only 10 mph, and air carriers carry just 50 passengers (low estimates tilt the comparison in favor of cars). That would mean cars had 1.1 fatality for every 10 million hours in 2015. The corresponding figure has been between 0.0 and 0.6 for the last decade, and airlines come out way ahead even after amortizing the 5.8 from 2001 over the following years. And this is with very conservative assumptions.

On the other hand if you add in non-occupant deaths, the numbers might work out better for cars, even after using less conservative assumptions, once you amortize the non-occupant deaths from 9/11.

We use it because there is really no good metric to compare the two.

Maybe just # of accidents in vehicle/# of trips taken in vehicle ?

Not to mention you have a fairly good chance of surviving a car accident whereas a plane falling out of the sky is almost ways bad news for passengers.

I have always thought the only relevant statistics is accidents per travel. But good luck getting this data.

Seems to be here: https://en.wikipedia.org/wiki/Aviation_safety#Transport_comp...

And traveling by air seems to be 3x more deadly than by car, if measuring by journey.

Reply on news.ycombinator.com