The Off Topic Topic

[Josquius]

Yeah. Driving more leading to more chance of accidents and more chance of deaths is kind of the main point.

[DGuller]

I was about to say the same thing.  People travel a lot more by car in the US, so all else being equal of course you would expect more deaths by auto.  Ideally you want to split it even finer if you want to identify a specific factors like country, because a mile of travel on a federal highway is way safer than a mile of travel on a two-lane road.

Wouldn't "deaths per passenger mile" implicitly remove "encourage other modes of transportation and reduce driving" from the potential solution set?

First step is to correctly analyze the situation.  Second step is to identify the problem.  Bad things happen if you start with step 2 and then tailor step 1 to fit it.

If your objective is to reduce death from travel, you first want to understand if the problem is exposure or the problem is with risk factors (i.e. road safety).  If the extra deaths in the US are all down to greater number of miles traveled, you're not prohibited from asking whether these miles are necessary.

[Jacob]

First step is to correctly analyze the situation.  Second step is to identify the problem.  Bad things happen if you start with step 2 and then tailor step 1 to fit it.

If your objective is to reduce death from travel, you first want to understand if the problem is exposure or the problem is with risk factors (i.e. road safety).  If the extra deaths in the US are all down to greater number of miles traveled, you're not prohibited from asking whether these miles are necessary.

No disagreement from me there.

Also, going back to look at the graphic, the "modes of transportation" is not as applicable as it includes deaths on bicycles and in buses as well - so it's not just about cars.

Went looking for Deaths Per Mile and found US statistic here: https://www.iihs.org/topics/fatality-statistics/detail/state-by-state

South Carolina, Mississipi, and Arkansas are the worst at 1.97, 1.90, and 1.88 deaths per 100K miles travelled respectively, while Massachussets, Minnesota, and New Hampshire are doing the best at 0.67, 0.76, and 0.87 respectively.

Though it looks like there may be a pretty strong correlation between deaths per 100k miles travelled and per capita deaths. If we look at Deaths per 100K population the three worst are Mississipi, Wyoming, and Arkansas at 25.4, 22.0, and 21.2 respectively (and South Carolina comes fourth at 20.7) while the three best are Massachusetts, DC, and NY at 4.9, 5.2, and 5.2 respectively. Minnesota sits at #7 (6.9) and New Hampshire at #9 at (8.2).

So maybe road deaths by population isn't that bad a measure after all, even if it's a proxy?

Haven't looked at EU stats yet though...

[CountDeMoney]

In related news, posted on NPR today:

https://www.npr.org/2023/04/06/1167980495/americas-roads-are-more-dangerous-as-police-pull-over-fewer-drivers


The reduction in enforcement isn't helping.  Less prioritization on traffic enforcement + policing by petulance = MOAR DEAFS

[Jacob]

So I went looking for "deaths per mile travelled" stats for Europe, and I couldn't find any. My intention was to see if there's a correlation between deaths per mile travelled and deaths by population in Europe as well, and how that compares to the US. But it seems the stats are all recorded and reported by population, not per mile travelled - including by agencies that presumably have data scientists working for them.

Going back to the US statistics and looking at the top 10 and bottom 10 in deaths/mile travelled and deaths/100K population the membership has significant overlap (8 out of 10 on the best 10 are on both lists, 7 out of 10 of the worst are on both lists).

I did come across the argument in a Bloomberg article:

Another common American response to road safety critiques is to challenge deaths per capita as a metric. Some argue that deaths per mile driven is a better comparison, since it takes into account the added risks of driving more miles, as Americans are wont to do. But this flunks the test of common sense. Consider: If traffic deaths are flat, but everyone drives twice as far, is society safer? Furthermore, rural interstate driving is significantly less dangerous per mile than driving on urban arterials, so a country could grow "safer" on a deaths/VMT basis simply by moving urban residents into the countryside.

... but from my relatively superficial comparison of the two metrics in the US context it seems there'sa reasonably strong correlation, so maybe arguing about the merits of the two types of measurement isn't as relevant as it might seem at first blush?

[DGuller]

South Carolina, Mississipi, and Arkansas are the worst at 1.97, 1.90, and 1.88 deaths per 100K miles travelled respectively, while Massachussets, Minnesota, and New Hampshire are doing the best at 0.67, 0.76, and 0.87 respectively.

Though it looks like there may be a pretty strong correlation between deaths per 100k miles travelled and per capita deaths. If we look at Deaths per 100K population the three worst are Mississipi, Wyoming, and Arkansas at 25.4, 22.0, and 21.2 respectively (and South Carolina comes fourth at 20.7) while the three best are Massachusetts, DC, and NY at 4.9, 5.2, and 5.2 respectively. Minnesota sits at #7 (6.9) and New Hampshire at #9 at (8.2).

So maybe road deaths by population isn't that bad a measure after all, even if it's a proxy?

Haven't looked at EU stats yet though...

If death per capita is a good proxy for death per mile, then the implication is that miles traveled per capita is roughly equivalent.  It may be a reasonable assumption between different states in the US (excluding states with huge cities with good public transit, like NY), but it's not a tenable assumption when you're comparing US with a European country.

EDIT:  My guess was correct, most of the states are very close in miles traveled per capita at around 10,000, with only a few material exceptions.  https://www.fhwa.dot.gov/policyinformation/statistics/2019/ps1.cfm

[Jacob]

I guess that leads to the conclusion that if you want to reduce traffic deaths, you need to reduce driving... which I guess is not super shocking.

In which case I suppose the explanation that for the US rate of deaths having gone up while the European on has gone down is that the US has seen an increase in driving and Europe a decrease?

So if the answer to "much larger number of Americans die from driving than Europeans" is "well of course, they drive that much more" then I guess the question is what's the next conclusion? To shrug and say "since Americans need to drive this much, there's nothing we can do" or to say "therefore we should work to provide alternatives to driving, so fewer people die." I guess that's where it gets political.

...

On a separate tangent - if the number of miles traveled per capita is roughly 10,000 across all states - and if deaths per mile is a reasonable proxy for deaths per capital - what factors explain the differences between US states?

The state with the most deaths per per mile (South Carolina, 1.97 per 100K) is more than three times higher that of the lowest (Massachusetts, .63 per 100K). The state with the most deaths per capita (Mississippi, 25.4 per 100K) has more than 5 times those of the state with the least (Massachusetts, 4.9 per 100K).

What is Massachusetts (and states in the same bracket like Hawaii, Minnesota, New Jersey) doing differently than South Carolina and Mississippi (and states in the same bracket like Louisiana, Montana, and Kentucky)?

[Admiral Yi]

I think Wyoming is where there is no speed limit on some stretches of Interstate.

[DGuller]

I guess that leads to the conclusion that if you want to reduce traffic deaths, you need to reduce driving... which I guess is not super shocking.

I think the whole point is that this one conclusion is not something you're led to, this one conclusion is something you know to be true.  You know for a fact that driving more will expose you to more chances to get killed, what is of interest when it comes to discussing road safety is whether something else on top of it will also get you killed more (or less).  You factor out exposure because the effect of exposure is already known.

In which case I suppose the explanation that for the US rate of deaths having gone up while the European on has gone down is that the US has seen an increase in driving and Europe a decrease?

No, not necessarily.  Miles driven is a factor, but it's not necessarily the only factor.  US fatalities per mile driven could have gone up for the last few years, but we wouldn't know unless we actually look at the numbers that way.  Is is super helpful to know whether Americans are just driving more, or whether something happens like cops throwing a hissy fit and refusing to enforce traffic laws that make a mile traveled more dangerous.

On a separate tangent - if the number of miles traveled per capita is roughly 10,000 across all states - and if deaths per mile is a reasonable proxy for deaths per capital - what factors explain the differences between US states?

The state with the most deaths per per mile (South Carolina, 1.97 per 100K) is more than three times higher that of the lowest (Massachusetts, .63 per 100K). The state with the most deaths per capita (Mississippi, 25.4 per 100K) has more than 5 times those of the state with the least (Massachusetts, 4.9 per 100K).

What is Massachusetts (and states in the same bracket like Hawaii, Minnesota, New Jersey) doing differently than South Carolina and Mississippi (and states in the same bracket like Louisiana, Montana, and Kentucky)?

Road type is another major factor that is controlled for even in numbers I've seen made public (which I'm sure are just at surface level).  Different road types have very different deaths per mile, which is why death per mile are calculated separately for several types of road.  As with general mileage, you know that a mile of a two-lane road where you can get a head-on crash at any point is much more deadly than a mile of interstate highway, so there is no point in letting those differences conflate your results.  It doesn't mean that you shouldn't consider replacing two-lane roads with limited access highways, but you're looking for signs of novel road safety factors, you should control for ones you're already well aware of.  Before looking into reasons why some states are more dangerous, you would want to know what the breakdown of their road types is by miles traveled.

Once you do account for road types, another thing I can think of is differences in fleet types.  We know that in some states you're going to have a lot more pickup trucks than in other states, for example, or the general age of cars could be higher in lower income states.  Cars in general are dramatically safer the newer they are, which is why it drives me nuts when parents say "he's a young driver, he's probably going to crash his first car, so it shouldn't be an expensive one".  If your son is going to crash, wouldn't you want him to crash in a car with more modern safety features?

[Zanza]

Here is an OECD Report that shows deaths per billion vehicle kilometers:

https://www.itf-oecd.org/sites/default/files/docs/irtad-road-safety-annual-report-2018_2.pdf

I guess that's not as precise as lookimg at deaths per passenger distance, but better than just looking at deaths per population. 

[PDH]

I think Wyoming is where there is no speed limit on some stretches of Interstate.

Max is 80 on those stretches.

The real killer in Wyoming is single car rollovers.  Drivers on empty 2-lane highways going too fast, drifting off the pavement with two tires, and overcorrecting.  Of course, it is lack of seatbelt use in these cases that kills, but the legislature says you can't ticket for no seatbelt, only as an add-on for another driving offense.

[Josquius]

I think the whole point is that this one conclusion is not something you're led to, this one conclusion is something you know to be true.  You know for a fact that driving more will expose you to more chances to get killed, what is of interest when it comes to discussing road safety is whether something else on top of it will also get you killed more (or less).  You factor out exposure because the effect of exposure is already known.
 

... So you factor out one of the big changes that can be done to improve things because it's already known that improving that would help a lot?

Isn't this kind of like factoring out eating a lot when it comes to advising people on losing weight?

[grumbler]

I think the whole point is that this one conclusion is not something you're led to, this one conclusion is something you know to be true.  You know for a fact that driving more will expose you to more chances to get killed, what is of interest when it comes to discussing road safety is whether something else on top of it will also get you killed more (or less).  You factor out exposure because the effect of exposure is already known.
 

... So you factor out one of the big changes that can be done to improve things because it's already known that improving that would help a lot?

Isn't this kind of like factoring out eating a lot when it comes to advising people on losing weight?

It's kind of like factoring what people eat and not just how much they eat.  Losing weight in a healthy way isn't just about eating less of the same things one has always eaten.

[DGuller]

I think the whole point is that this one conclusion is not something you're led to, this one conclusion is something you know to be true.  You know for a fact that driving more will expose you to more chances to get killed, what is of interest when it comes to discussing road safety is whether something else on top of it will also get you killed more (or less).  You factor out exposure because the effect of exposure is already known.
 

... So you factor out one of the big changes that can be done to improve things because it's already known that improving that would help a lot?

Yes?  What is the purpose of learning something you already know?

[Josquius]

I think the whole point is that this one conclusion is not something you're led to, this one conclusion is something you know to be true.  You know for a fact that driving more will expose you to more chances to get killed, what is of interest when it comes to discussing road safety is whether something else on top of it will also get you killed more (or less).  You factor out exposure because the effect of exposure is already known.
 

... So you factor out one of the big changes that can be done to improve things because it's already known that improving that would help a lot?

Yes?  What is the purpose of learning something you already know?

To provide further evidence to decision makers that action is required.