Read. Think. Then Post.

An easy three steps that would (still!) keep a lot of people out of trouble.

Instapundit links this, apparently approvingly. "Drunk driving" kills, so why do cops mainly give out speeding tickets? Slogans and fallacious logic driving spending priorities, the linked editorial suggests (while Instapundit thinks it's all about revenue).

Many, many methodological fallacies permeate its five short paragraphs. Again, let us count the ways.

1) "Drunken drivers". Since these are NHTSA stats, numbers of "drunken drivers" are actually going to be "those whose blood-alcohol content levels are at or above .08%". Sorry, but for most people, the level of impairment after one or two beers is nowhere near "drunken". Bzzzt for unsubstantiated hyperbole.

2) "Distracted drivers". This category comes to us also from the NHTSA with some unknown contribution from VA Tech. While in the main I sympathize with attributing influence to this conceptual factor, as an empirical, measurable phenomenon it leaves quite a bit to be desired. Presumably data must be self-reported in many cases, and mixing self-reported with other data sources (passengers; post-accident circumstantial evidence) increases unreliability across the data pool.

All normal people multitask when driving, in part because driving requires multitasking in the first place: You must be able to check your speed and scan other dials and indicators, flick the turn signal, start/stop the wipers, hit the lights/brights, etc., without coming to a screeching halt every two minutes to adjust one of the functional driving controls. Thus, if you are driving along and fooling with the radio, are you a "distracted driver"? Maybe, maybe not. If you are driving along obsessing over the fight you just had with a random store clerk or loved one, are you distracted from driving? Maybe, maybe not.

Taking it one step further: There are bad drivers and good drivers. I'll posit that good drivers are ones who have a substantial foundation of automatic multitasking and learned reactions to smaller or more peripheral signals. Good drivers can take on significantly more distraction -- or apparent distraction -- than a less-experienced or a less naturally alert person, and/or can selectively tune in and out with respect to safety signals versus other 'noise' in the environment.

So, who are the "distracted drivers"? Those who report afterwards that they were talking to others in the car or changing the radio station, or those who report afterwards that they were distracted by talking to others in the car or changing the radio station? Or those whose passengers report the baby in the back seat was screaming (when we all know that parents can be astoundingly oblivious to the horrific noises their offspring make any time, any place)? Or the cop who finds a cell phone on the seat of a car after an accident? Is it methodologically defensible to use any of these criteria? Which? How has the validity of whatever measure was used been tested?

Philosophically, what traffic accident isn't caused at some level by a distracted driver? Is the theory that there is a significant class of accidents caused by people who are paying attention? Sure, occasionally a tire may blow out or cars can slide on snow. However, since those types of incidents don't invariably result in accidents, and in fact the skill, experience/knowledge, and attention level of the driver (and sometimes sheer luck) often avert damage to that car or others, wouldn't all drivers who have accidents also be drivers who were distracted, if only by their own thoughts or perhaps simple disinterest in the task at hand?

3) Unfortunately, the editorial links nowhere, so there's no way to assess whether or how the numbers presented in conjunction with these conceptually and empirically murky were constructed.

3a) In the days of Usenet, I researched official (USG) figures on drunk driving and drunk driving related statistics. It turned out at that time that accidents "caused" by drunk driving are, for the purposes of USG slogans and MADD proselytizing, any accident that includes a driver whose BAC is over the legal limit is deemed to have been caused by the "drunk" behind the wheel. Now, this is nutbar territory on the face of it. This means I could be in my lane, traveling at the safe and posted speed limit, with a BAC of .080%, and some idiot could run a stop sign or red light and ram into the side of my car, and the statistics would show that that accident was my fault. As I say, nutbar territory.

NB: It's a little jangly, but more dissection along these lines here, including one precious bit I'd forgotten from that long-ago Usenet debate: If a sober driver hits a pedestrian who has been drinking, that counts as an accident caused by "drunk driving". Another item I don't recall from back when: An accident involving a designated driver counts as one caused by drunk driving -- passenger BACs count as much as anyone behind the wheel. Well, that's one way to hit 40% "caused" by "drunks".

3b) Awareness of this feature of the statistical landscape raises my suspicions about the claim that "fully 80%" of accidents are (Warning!) "caused" by distracted drivers.

4) ObHyperbole Alert: "The latest data demonstrates that such drivers kill and maim far more people than speeders." Not in this editorial the latest data do not. "Fully 80%" of crashes does not exceed the unknown number of persons killed, which presumably is over twice the "nearly half" of all fatalities from crashes. If there are other data, the editorialist should have produced them, since otherwise this looks like a specious conflating of unrelated statistics (which are, see above, highly dubious and unreliable in themselves).

*****
And that's the first two paragraphs...
*****

5) I'll just note in passing the emotive language throughout the 'policy' grafs. A classic tool of demagoguery that editorials typically use, it's a dangerous and foolish way to make public policy. Its mixture with the bad numbers used badly does add to the bad odor of this piece.

6) Whether or not speeding is "more" important than drinking and distracting might be an interesting question, but not as this editorialist frames it. Public policy requires juggling interests and priorities along a number of dimensions, and in this editorial we are in the flat land of If It Saves One Life mob-raising talk instead. Not a single number is presented that distracted driving "causes" traffic deaths, even in the distinctly loose and ambiguous way "drunks" do. And surely no one can claim that people aren't upset enough about drunk driving, when the Constitution has already been impaired by legalization of checkpoints and current talk is of pushing legal limits down to .05%?

It appears that the editorialist has jammed together the largely unrelated issues of distraction with drinking in order to drive (heh) up the Deadly! Numbers! factor lacking on the distraction side and the Let's Get Serious About This Serious Problem which is clearly not lacking on the drunk-driving side. A little semantic sleight of hand and, boy, it really looks like the cops have everything all backwards!

But until we don't have speed limits, it just makes sense for traffic cops to give tickets for speeding. That seems at least more useful than parking tickets.

And what's the justification for not giving out speeding tickets? The editorial presents two: Speeding tickets are expensive, and giving speeding tickets at this time of year is scrooge-like. As to the first: Yes, dear, that is the point. The expense is part of the deterring structure of laws against speeding. The second: Totally unserious.

7) Furthermore, there's an eentsy problem with the idea that cops should stop writing so many speeding tickets and start "focusing" on drunks and distracted drivers. How?

Sure, we could set up more checkpoints. I'm against them in principle, but chasing speeders isn't preventing checkpoints. And there's all the extra manpower freed up by red-light cameras, too. Personally, I would consider Scrooge a pretty jolly, well-intentioned fellow by comparison with cops and cop administrators who feel increased intrusion into my private life and lawful business is justified on the theory that Something (More) Must Be Done about the non-zero probability that someone, somewhere, might be in hypothetical danger in the vicinity of a vehicle and BAC>.0-whatever.

Beyond that point, how the heck are cops supposed to focus in on distracted drivers? There are already laws against driving-and-talking in many (most?) places [NB: I don't have a cell phone, so I don't pay attention to those laws], and all sorts of things distract (bad) drivers. Food, non-alcoholic beverages, conversation, life, love, the universe, and all that. Distracted-while-driving status can't be determined by a test, so even checkpoints won't help. Rip out radios and CD players, criminalize Ipods and crying babies -- or all passengers, require No-Doz before trips over 15 minutes? But most accidents occur close to home! And society wants folks to carpool! Wow, this social engineering stuff is Hard.

8) Wrt revenue: A red herring to the editorial's central point, that we should care more about nabbing, prosecuting, and penalizing drinkers and distractables (whence could cometh, presumably, a whole lot more revenue) than we do about poor ol' speeders. Revenue is an outcome of nabbing whomever. Which miscreants public policy targets for the revenue extraction (as penalty or deterrent) is the question. Sure, other questions can be raised, but ultimately that's not "why" we do it.

In sum: Drunks, cell-phone users, people with kids, people who get sleepy (or tired or enraged [and whatever happened to activism drumming up Concern about road rage?]), and just plain bad drivers -- heck yeah, get 'em all off the road. And I admit, speeders don't bother me at all unless they're tailgating me, which is a different problem. I'll even admit that I recently got a speeding ticket, which was pretty much a big PITA. So, I suppose I should be on the editorialist's side.

But I'm not. I'm never on the side of the sloppy pseudo-factoid argument. I don't want cops' priorities determined by the loudest whiner -- not MADD, true; but not this writer either. I don't want public policy determined by the Nanny-est of the Nanny State attitudes. People die. People get killed unfairly. Speed kills, drunk driving kills, cell phones kill (for all I know). Even if I disagree with the mania against doing anything but sitting still in a chair until every .000001% of alcohol content has left your blood, I see no advantage in that attitude being displaced by some other irrational phobic mania.

Policy should be informed by facts, and still will involve hard choices and trade-offs on principles and/or resources. Even if that's not how it works, it's how we should try to make it work. The fanatic fighting the shrill demonizing the hysterical only means we'll all end up disrespecting everything.

How far are we from that now?

Interpretation

I was looking here, and as usual had serious qualms not only about the validity (are we measuring what we intend to measure?) but the interpretations.

Most serious, however, are the validity concerns. I will be absolutely frank and say that if I were asked, "3. Generally speaking, in your opinion is the United States headed pretty much in the right direction these days or is it headed off on the wrong track?, I would say right track, without equivocation. In reality, in my heart of hearts, I have serious misgivings about the idiocies and vacant wank fodder swirling through the countrysides -- or universities, pseudoprofessional classes, and city-sides, to be more accurate. There's no rule that the Supreme Court must be "balanced" and that Sandra Day O'Connor should be replaced by a swing voter (not to be confused with swing dancers). There's no incontrovertible imperative that US tax dollars have to be handed over to a bunch of terrorists just because they won an election, and the fact that the guys before were pretty much terrorists too doesn't make it binding precedent for the next gang of sociopaths. "Pretty much" may be a nod and a wink from "really", but it's still a far cry from "hysterical commitment to violent annihilation of a neighboring country". People who believe in God are not lunatics, not even if they have faith in a loving God in the face of hatemongering nutbar humankind -- or even merely crippling incompetence, obliviousness, and profoundly self-centered insincerity -- all around us.

But I digress. Let's imagine for a moment that I answer the poll honestly, and 50 percent of the respondents agree exactly with my take on the subject, and also answer honestly. We have serious concerns about the shrieking asshattitude that's seemingly taken Very Seriously in Serious Institutions such as Papers of Record, Halls of Decision-Making, and Ivied Towers of Higher Learning, for instance, although the examples above are a trivial listing.

That response will be interpreted as a resounding rejection of everything Bush is doing, has done, or ever will do. Doubt me? Read the commentary at the link. It's all about what's negative versus favorable for Bush, despite the fact that only one question mentions the man, complete with "scare quotes" for anything that can be mocked for having a possibly slightly positive connotation. They congratulate themselves on bending over backward not to be partisan by saying

Our "wrong track" direction of the country numbers are much more favorable to the Bush administration viewpoint than those of other polling firms.

By "favorable" they mean fewer respondents claim the country's on the wrong track. Polls from the other side of course do the same thing.

So, politically astute respondents will not respond honestly to that question, or most others. They will, if things are going really well all around them and as far as they can see or imagine in the great wide world, say "wrong track" if they hate Chimpretzler McHitlerburton, and vice versa as I have indicated above.

Methodologically speaking, in a self-aware world, any potential for validity in broadly generalized impression questions is terminally gutted where interpretation demonstrates bias in data use.