German DUI Study

A study of 4,615 accidents examining the effects of drinking and driving in Germany, where speeds over 140 mph on the autobahns are common, has all the earmarks of a bureaucrat trying to justify his position.  But it contains much useful data, particularly considering that we have 45.8% more traffic fatalities per 100,000 population than Germany.  1,968 drivers were determined to have been responsible for some portion of these accidents, and 330 drivers, or 7.2% of them, had a bac greater than 0.01%.   This group was compared against a Roadside Survey of 9,269 Germans in Bavaria who were stopped at random and analyzed for their blood alcohol content, and it was reported that 5.5% of German drivers have a bac greater than 0.01%.

But--5.2% or 501 of them refused the test.

Now.  Stop right there.  This study proceeds as if though only 5.5% of German drivers (in a country which consumes 50% more alcohol per capita than Americans) have a bac greater than 0.01%, which by itself should sound all kinds of alarms.  But the fact that the study ignored the *reason* that most of the 5.2% of them refused the test, namely that they had been drinking and did not want police with guns who stopped them at random in the middle of the night to know that, completely invalidates this control group.   No data derived from such a grievous manipulation of the facts can be trusted from this point forward, no matter how well done the study may be. 

ARE DRINKING DRIVERS 32% MORE DANGEROUS THAN NONDRINKING DRIVERS?

If only 5.5% of all German drivers are drinking drivers [read: have a bac > 0 at any one time], but 7.15% of them are involved in accidents, then nondrinking [read: have a bac = 0] drivers are 24.4% less likely to have an accident than drinking drivers. 

OR--ARE NONDRINKING DRIVERS 55% MORE DANGEROUS THAN DRINKING DRIVERS?

If 10.7% of all German drivers have a bac > 0, and only 7.15% of them are involved in accidents, then the nondrinking drivers are 55.6% MORE likely to have an accident than drinking drivers.  Why did this study ignore this point?  And why is this significant?  Well, it invalidates the entire premise of the study, namely that drinking drivers are more dangerous than non-drinking drivers.  The authors didn't want you to know that because it was funded by the German government who wants to crack down on drinking drivers, and it's difficult to justify the draconian measures they have in mind if they're cracking down on the wrong group. Germans are very good at paying attention to detail, which is why it's surprising that they would permit a "detail" which would have completely reversed their conclusions to slip by. 

If the original faulty premise doesn't reflect on the rest of the study, then it does contain some other interesting facts.  Let's get equally as sloppy as the authors and proceed with the assumption that 100% of those who refused this DWI test had been drinking alcohol, which would mean that 10.7% of the drivers in the control group had a bac greater than 0.01.  Let's assume that they are representative of all German drivers--that 10.7% of German drivers have a bac greater than 0.01% at any one time.  Their discovery that drivers with a bac between 0.01 and 0.06 have a safer driving record than drivers with a bac of zero and that "heavy drinkers" and "excess drinkers" who adjust to drinking and driving improve their safety record quite a bit are points that cannot be ignored, particularly if the amount by which drivers with a bac between 0.01 and 0.06 are safer than drivers with a bac = 0 is greater than the amount by which drivers with a bac of .12 are more dangerous than drivers with a bac between 0.01 and 0.06.  If you want to crack down on drivers with a bac = .12, you must also crack down on drivers with a bac = 0.  After all, it  would be "discrimination" to ignore them. 

NONDRINKING DRIVERS 43% MORE DANGEROUS THAN 'THE GOLD STANDARD'

Can you even imagine the irony that the German government funded this study because it wanted to prove that drinking and driving causes accidents, but instead confirmed numerous other studies (i.e., the "Grand Rapid Study", and a study of airline pilots) that nondrinkers have 43% more accidents than moderate drinkers with a bac = .04%?  What kind of government policy should be derived from such a revelation, especially since "liberals" believe that "if it costs a trillion dollars just to save one life, then it's worth it"? The difference is not small. It shows that if all nondrinking drivers had a propensity to have accidents that was equivalent to drivers with a bac = .04, that they would have 30% fewer accidents. Needless to say, this report doesn't make this point directly, but anyone who reads it can easily determine this.

Why should sober drivers be permitted to drive around having 95% of all the accidents just because they're 43% more likely than moderate drinkers to have accidents? Should they be outlawed, driven out of town on a rail, and tarred and feathered?  Should yellow arm bands be wrapped around their tail pipes?  Should they all be required to wear Dewey Buttons?

The least we can do is award drivers with a bac = .04% (those who have proven time and time again, around the world, in numerous studies of various hand/eye coordination tasks, that they have 30% fewer accidents than those with a bac = 0) a title.  Let's award them the title of the "Gold Standard".  If all drivers and pilots had their safety record, the lives saved would be worth far more than gold.

WHAT ABOUT WOMEN DRIVERS?

What about women drivers?  It's well established that women drivers are 70% more likely per mile to have an accident than men drivers.  But compared to the Gold Standard who are 30% less likely to have an accident than the group of men drivers to whom women drivers are compared, the difference changes drastically, to 143%.  In other words, the ordinary woman driver is 2.4 times more of a danger to society than the Gold Standard.  What about women drivers who drive 30% of the miles in this country, who are 2.4 times more likely to have an accident than the Gold Standard, and who increase the risk of an accident for all drivers, including the Gold Standard?

HOW TO SAVE 12,450 LIVES PER YEAR WITH 'THE GOLD STANDARD'

Consistent with the pattern discovered in the US, this report revealed two significant dips in the curve, one with a bac = .12% and another with a bac =.20%. These dips were so large that drivers who had these two bac levels were not that much more likely to have an accident than women drivers with a bac = 0. It pales in comparison to how much more dangerous the sober driver in this study was than the Gold Standard, mainly because there were 6,651 nondrinking drivers who were 43% more likely than the Gold Standard to have an accident, whereas there were only 18 drivers with a bac = .12% and only 28 drivers with a bac = .20%.

What would the perfect totalitarian state be thinking right now?  With citizens having already accepted the notion that the state should monitor our bodily fluids to attempt to predict beforehand who will and who will not have an accident, it's not that much of a stretch to simply equip every car with an intravenous system which maintains our blood alcohol content at precisely .04%.   After all, human life is at stake here, and intravenous systems are cheap. A 30% reduction in our traffic fatality rate would save 12,450 American lives per year. This is not a serious proposal.

PLAYING THE DEVIL'S ADVOCATE!

If the distribution of bac of German drivers who refused the test was equivalent to drivers who did take it, then the Gold Standard (the driver with a bac = .04%), rather than being only 30% less likely to have an accident than the sober driver, would now  be two thirds less likely, or one third as likely, to have an accident.  Where it would have been estimated that there should have been 59 drivers with a bac = .04 involved in these accidents, there were actually only 21 (64% fewer).  Where there should have been 107 with a bac = .02, there were actually only 53 (50% fewer).  Where there should have been 24 with a bac = .06, there were actually only 13 (46% fewer).

Even though this and other flaws render the study completely and totally worthless, the data is invaluable.  The mere fact that they insisted that they don't even know with certainty whether or not 5.5%, or 10.7%, of German drivers have a bac > 0 at any one time is very sloppy scholarship for Germans, which illustrates how strict their orders are to NOT release the actual data.  Of course they know exactly what these figures are but just don't want the taxpayer asking too many questions.

The creative way they got the word out that nondrinking drivers are 43% more likely than The Gold Standard to have an accident confirms an otherwise inexplicable feature of the NHTSA data: that nondrinking drivers are more dangerous than drinking drivers.  This isn't necessarily because they have a higher accident rate than drivers with a bac > .20, but because there are 25 times as many of them. Of the 1,638 nondrinking drivers who were at fault for some portion of these 4,615 accidents, 491 of them would not have had an accident if they had had an accident rate equivalent to The Gold Standard.  This is more than twice as many drivers as this study claims had an accident because of alcohol, using even their most liberal assumptions.  If all the nondrinking drivers in this study had had a safety record equivalent to The Gold Standard, the overall accident rate in this sample would have been 10.6% lower and the number of drivers causing accidents would have been 25% lower. 

METHODOLOGICAL ERRORS

1. While the body of the study reports that 5.5% of German drivers in the Roadside Survey had a bac > 0, their Table 1 used to compare this data against the drivers involved in accidents shows that 605 of 9,043 drivers, or 6.7% of them, had a bac > 0.  Their own Table 1 is in serious conflict with their own conclusions. 
2. It's a serious problem that all of the drivers from the Roadside Survey were compared to only 1,968, or 28.2% of the 6,981 drivers involved in these 4,165 accidents.  You MUST compare all of both groups to each other to get a valid representation of the data.  You would never compare the subset of one group against all of another group--unless you had a particular goal in mind [read: unless you wanted to "prove statistically" that the drinking driver needs to be closely regulated]. 
3. Another bit of evidence that they had such an agenda is that they never even mentioned how many drivers were involved in these 4,615 accidents.  We must estimate it based on the average of 1.51 drivers per accident in the US. 
4. Where they estimated that 330 drinking drivers represented 16.7% of their subset of 1,968 drivers who were responsible for some subset of the accidents, we now see that drinking drivers were only 4.7% of all the drivers involved in these accidents.

Had this subset of 1,968 German drivers who were involved in 4,681 accidents been a randomly selected subset of all 6,981 drivers who were involved in these accidents, this might have been an acceptable statistical analysis.   But to compound the error, they admitted that this subset was not randomly selected. The subset included only the German drivers who the police themselves, not courts, and not the drivers, determined were "responsible" for the accident.   Note the wording "responsible", and ask yourself how 5,013 or 71.8% of these German drivers could not have been "responsible" for an accident they were in.  To compound the error even further, they made the presumption that all drivers who had a bac > 0 were "responsible" for the accident, while 71.8% of the drivers were not.  This is great advocacy, but it's terrible science.  It's worse than "guilty unless you prove your innocence", because drivers who were never even proven to have been "responsible" for the accident, who never even had a chance to present their case to a court of justice, were used as an excuse to terrorize other German drivers.

In other words, they started with a faulty premise, and when the data didn't support that premise, they changed the statistical rules to make the data fit the faulty premise.  If you think drinking drivers are the criminals--then you don't understand how serious the problems are that scientists like this have caused to societies all around the world.

If you've followed this so far, at least you now know why they wanted to compare the Roadside Survey against only a subset of the accident data--it made the drinking driver appear to be involved in 3 1/2 times as accidents as he was.  By comparison, police reports of American accidents show that only 4% are "alcohol involved".   There are several reasons that this report may have concluded that .7% more German than American accidents are "alcohol involved":

1. Germans actually consume 50% more alcohol per capita than Americans.
2. It could be that there are more drivers per accident in German accidents.
3. They may have intentionally inflated the data.

The conclusions of this report would have been drastically different had the statistical analysis been done properly.   Where it appears that drinking drivers were only 5.5% of all drivers, but were 7.2% of all drivers involved in accidents (suggesting that alcohol was a factor in those accidents), we now know that drinking drivers are 6.7% of all the drivers, but are involved in only 4.7% of all of the accidents.  Where the odds ratio favored nondrinking drivers as 34.4% less likely than the drinking driver to have an accident, the odds ratio now favors the drinking driver to be 45.7% less likely than nondrinking drivers to have an accident.  Let's put this another way.  The procedure used by the German statisticians "disclosed" that drinking drivers as a group were 32.7% more likely than nondrinking drivers as a group to have an accident,  whereas the correct procedure would have  shown that nondrinking drivers as a group are 45.7% more likely than the drinking drivers as a group to have an accident.

Why did these government sponsored researchers resort to such a skewed approach?  

BECAUSE IT MEANS THAT THE ENTIRE NON-DRINKING GROUP IS MORE DANGEROUS THAN THE DRINKING GROUP, which is a conclusion which is contrary to their assigned task.  Their task was to justify bigger government--not to seek truth and justice.  The two are usually mutually exclusive of each other.

  SUMMARY OF THE DATA

Copyright @ 2007 by Fathers' Manifesto & Christian Party