Its a YOLO mobile for a reason.
2 Likes
Can anybody pull the statistics on how many husbands were talked into a minivan by their wives who then get into a crash and then blame the husband for getting an unsafe car? 
1 Like
Disclaimer: not a statistician.
“Objective metrics” can be spun all sorts of ways. It would actually be helpful to hear from someone who works in the insurance industry about what metrics they use.
Using crashes per x miles drive makes sense when you’re looking at aggregrate (that’s not the right word, but, like I said, I’m not a statistician) data based on demographics. The table you linked to shows crash types and by age per miles driven. Miles driven makes sense in this situation b/c one of the variables, the driver’s age, doesn’t changed based on the miles driven.
I didn’t fully read the article I linked (whoops), but IIHS was looking specifically at small overlap crash results. So number of miles driven, in this situation, doesn’t really make sense.
So the metrics might be, of total crashes, how many crashes are small overlap? Or, of significant or fatal injuries in a crash, what percent of that involves small overlap?
The IIHS was also specifically looking at how a child (or a small woman, I guess) would fare in a small overlap crash while sitting in the 2nd row, right? One could be total callous and say, who gives a f*ck about anyone sitting in the 2nd row of a minivan since minivans only constitute 5% of car sales (https://www.caranddriver.com/news/a40133555/minivans-sales-increase/).
But one would assume that, when minivans are in use, the 2nd row is often occupied and is often occupied by a child or young teenager. So, in that case, even though the type of car being tested (and thus the type of passenger and seating position) are low in an absolute sense, the way the data being generated is probably very useful for that type of car. And, if you’re a parent, you probably don’t care about absolute numbers when it’s your kid sitting back there.
The other thing is that the data being gathered is also about risk stratification for insurance purposes, right? So, again, the number of fatal crashes per miles driven is not the only metric. If I’m an insurance carrier, you ask me to provide insurance for the minivan you purchased, a team of actuaries is presumably crunching the numbers based on actual crash statistics but also perhaps cross referenced w/ IIHS data to spit out a concrete number.
So, while “objective data” is certainly better than no data, it’s important to first determine what someone is trying to measure and if the metrics chosen actually provide that data.
At any rate, the headlight data provided by @Stonksonlygoup does not seem trivial at all to me.
1 Like
Pretty important line from the article, as most minivan drivers (myself included) have them because of children with car seats and booster seats:
“Even with these developments, the back seat remains the safest place for children, who can be injured by an inflating front airbag, and the rating does not apply to children secured properly in child safety seats.” (emphasis added).
3 Likes
I’m open to any data-driven analysis.
But it all starts with the probability of a negative outcome (not a dent in the door, but a serious injury, and separately but far less likely, death).
Most of the safety discussion tends to be about how people feel.
“We can’t let anything happen to Baby Calliope,” without regard to actual risk.
I mean, it makes no sense to take a kid in the car in the first place; it’s dangerous.
If you’re serious, you’ll leave the kid at home with a medical doctor and a baby sitter. Make sure you’ve done background checks, have backup sitters and MDs on call, and lock all of the steak knives in a safe. And cameras in every room, with redundant backups.
Too expensive, logistically difficult, and inconvenient?
Doesn’t matter. IT’S FOR SAFETY.
I don’t know about that, actually. At least, not in this country. I think it actually starts w/ the likelihood of monetary compensation for a significant adverse outcome. And, in terms of studies are done (at least in my field), it begins w/ a hypothesis. The hypothesis could actually be that there is no difference btw the study or control groups.
::shrug::
When there is an expectation of a concrete monetary amount associated with an event, then there will of course be an expectation of data to support that payout.
Do other countries, that are less litigious, track the same data?
I find the use of the term “data-driven” to be a bit curious b/c data can always be manipulated n a variety of ways. I mean, if the table you posted show the absolute numbers of fatal crashes per 100 milion miles drive, my thought is that the absolute numbers are quite low. So why should we care about fatal auto accidents at all.
But that’s not what the statistical analysis was designed to measure. The analysis was done to show that difference age groups differ significantly in the rate of crashes (further sub-divided into all, injury, an fatal). It can be a very fuzzy thing to use data designed for one analysis and then use it for a different purpose…
Sorry, I wasn’t very clear.
I was suggesting that many consumers buy “safety” without regard to the size of the actual risk.
Especially moms.
I’m not arguing, for example, that 300 headlights on a car aren’t technically safer than 2, I’m saying that the cost of engineering 300 headlights isn’t worth the cost. You certainly wouldn’t pay $20,000 extra for the option.
Not every risk is worth mitigating. I mean, look at the armored vehicles that transport heads of state. You don’t need one of those either, nor could most of us afford one. 
Yes, I agree.
I think most people in most countries do most things w/o much thought… 
and 
1 Like
Fun anecdote, when we did a market study and surveyed folks about our vehicles (back in the early 2000’s), a lot of women believed sunroofs were unsafe. Because in the event of a rollover/crash they thought passengers/kids could be ejected through the sunroof.
Are you allowed to say which manufacturer???
I am more risk adverse than @trism yet I am not at all concerned about this risk. But the women were not wrong. https://www.nytimes.com/2018/02/22/business/car-sunroof-safety.html?smid=nytcore-ios-share&referringSource=articleShare
Although if you are concerned about being ejected from the car, maybe just….wear a seatbelt.
2 Likes
It is not trivial but it is also almost certainly not stasitically significant. The study only looked at single car collisions. There are two main issues, one, there is just no way to know for sure whether a driver was sober. People don’t volunteer that info. In a single car collision where the police aren’t called and the driver walks away, we would almost never know if someone was impaired. Even if the driver was injured and went in an ambulance, it is likely the BAC test doesn’t occur for hours after the accident and probably doesn’t get reported to the police anyhow.
And then what about distracted driving. Absent litigation that results in pulling phone records, we rarely know if an accident was caused by distracted driving and almost never know in single car accidents since there is rarely litigation afterward and no one self-reports that they were texting at the time of the crash.
You need a huge sample size to counteract this unresolvable variable and the study had a noted small sample size.
This is basically why any sort of “data driven analysis” on crash prevention technology is so expensive. You need a huge sample size to account for the fact that a lot of the accidents are due to intoxication or distracted driving.
I think it isn’t just the absurd people at 70 not being tested though. 
1 Like
Hopefully next year, IIHS ratings might change due to action from the MFG.
Also with headlights there is a cost in terms of blinding the opposing car. As bright as possible isn’t always for the best. It’s quite annoying and frankly dangerous driving into some modern light systems which are way too slow to turn off auto high beams
1 Like
Honda and Acura the leaders in blinding my eyes since the 2016 Acura MDX.
1 Like
This where the trained statistician would be helpful (I’m sure there must be one here!). I think a result can be statistically significant (which simply means that the result was not by chance) but that there can be multiple confounding variables (so it’s unclear if the variable being studied is what actually caused the result).
But I think your main points were that it’s hard to gather data on all the variables and that, in the case of car crashes, the data will thus always be incomplete (thus resulting in an asterisk whenever “data” is presented). And I think you said earlier that incremental improvements across multiple (safety-related) domains can have a potentially large impact. And I would agree w/ all of that.
On an only slightly related note, the headlight thing really sticks w/ me b/c I find the low beam headlights on my own car (VW Golf Sportwagen w/o halogen lights) to be utterly dreadful.
The US revised the archaic headlight rules that prevent auto manufacturers (mainly European ones, I think) from bringing over the cool tech that allows the headlight pattern to re-map itself, if it detects another car ahead, right? So hopefully being blinded will change.
Of course it can be statistically significant. You just need a big enough sample size. And any study needs to acknowledge that they cannot determine if the driver of any accident was caused by driving under influence or distracted driving. Those are always going to be unresolved confounding variables.
1 Like
And Musk ignored your survey anyways.
1 Like
Suv are stronger I rather flip then get crushed.