Toyota acceleration: why I’m skeptical

by Ted Frank on March 11, 2010

Dating back to 1992 models, LA Times reporters found 56 deaths reported to NHTSA over the course of 19 model-years. If Toyota is suffering from electronic problems, these electronic problems should affect all drivers equally. If Toyota sudden acceleration is caused by driver pedal misapplication, then we should expect to see a disproportionate number of elderly and short drivers. Unfortunately, we don’t have driver heights, and in only 24 of the 56 cases, did the Times list the age of the driver.

The ages: 18, 21, 22*, 32, 34, 44, 45, 47, 56, 57, 58, 60, 61, 63, 66, 68, 71**, 72, 72, 77, 79, 83, 85, 89.
*Passenger victim was 22 and “friend” of driver.
**Passenger victim was 71 and married to husband-driver for 46 years.

The median age is 60.5; the majority of drivers are 60 or older; a third are older than 70. And I left out the case of a driver who was the son of a 94-year-old victim rather than guesstimate his age to be 65. That looks suspiciously like the makeup of Audi sudden acceleration cases, and a lot like driver error to me. Color me skeptical. Very very skeptical.

Update, March 12: Megan McArdle has done some very impressive journalism following up on my work to fill in the gaps that the LA Times left out. Here’s her spreadsheet. (McArdle also has the guts to mention the disproportionate number of immigrants in the sample, which I didn’t.) Her report makes me realize I made a mistake in the sequence above: I confused an 89-year-old passenger with a 71-year-old driver. In addition, the driver I conservatively estimated to be 71 above turns out to have been 75. And McArdle says that a driver I listed as 61 is 60. Here’s McArdle’s more complete and more accurate sequence; I’ve estimated three of the ages where they were not listed:

18, 21, 21*, 20s**, 32, 34, 36, 44, 45, 47, 56, 56, 57, 58, 60, 60, 63, 60s***, 66, 68, 71, 72, 72, 75, 75, 77, 77, 79, 83, 87
*Driver was with 21-year-old friend
**Driver had girlfriend and young daughter
***Driver was picking up 67-year-old friend for church.

This actually strengthens my case: the median age is 60, 16 out of 30 (or 15 out of 29) are 60 or older (as compared to 16% of drivers in all automotive fatalities)—that’s a relative risk of over 6. We’ve gone from a small sample size of 24 to a slightly less small sample size of 29-30, improving statistical significance.

Separately, reader G. writes:

Hey Ted: one more data point on why Mr. Prius Acceleration is likely a fake — the stretch of I8 where the incident occurred. If you were to pick the one stretch of highway in San Diego County where you could go 94 MPH with almost no traffic and almost no curves, that is the stretch. At about 15 miles east of San Diego that road becomes deserted at all hours — it runs out into the Imperial Valley and then into Arizona. I have driven it tens of times, at all times of day, and never hit traffic unless there was a Border Patrol checkpoint. It is also almost straight– with some very moderate curves and some hills. Counter-data points: (a) About 60 miles east of San Diego (heading East) you hit some severe curves and a steep downhill grade as the road heads out of the mountains and onto the desert floor. I wouldn’t want to head into that at 94 MPH, even if I was faking the acceleration; and (b) dude is from Jacumba, which is on that highway (he didn’t drive from another part of the area just to drive on the road.

{ 14 trackbacks }

“I am not afraid of my Toyota Prius”
03.11.10 at 2:01 pm
Toyota (i.e., the Car; Not the Company) Discriminates against the Elderly | The Barstool Economists
03.12.10 at 8:29 am
Theodore H. Frank: I am not afraid of my Toyota Prius « Thoughts Of A Conservative Christian
03.12.10 at 3:08 pm
Average age of runaway Toyota drivers is over 60. | The Daily Caller - Breaking News, Opinion, Research, and Entertainment
03.12.10 at 3:24 pm
Waldo Jaquith - links for 2010-03-12
03.12.10 at 6:03 pm
The age of people who experience “sudden acceleration” in their Toyotas | Life, Liberty, and Property
03.12.10 at 11:01 pm
Variations on the theme of beige « Blunt Object
03.13.10 at 1:28 am
03.13.10 at 11:48 pm
London The » ‘Sudden Acceleration’ Evidence Against Toyota Weakens Under Examination
03.14.10 at 12:35 am
Quite possibly the last car you’ll ever own! | Akkam's Razor
03.14.10 at 9:25 am
Walter Olson: Toyota is Audi all over again | The Daily Caller - Breaking News, Opinion, Research, and Entertainment
03.15.10 at 10:12 am
The Toyota witch hunt « Benighted Comment
03.15.10 at 3:18 pm
Prius Frenzy: If you’re not elderly, phantom acceleration probably won’t kill you
03.15.10 at 9:24 pm
(Jalopnik) America, You Brought The Toyota Hoax On Yourself « Interesting finds
03.16.10 at 4:49 pm


1 steve 03.11.10 at 2:35 am

I’m wondering where all the news reports of people trapped at high speed are for the rest of the world?

I’m sure its no coincidence that every previous report of out of control cars was followed up later by reports of mental illness, and large amounts of anti -psychotic drugs. Floor mat entrapment does happen, but mats have much better anti creep features than they did some years ago. Electrical problems, where the gear, brake and off button don’t work….I think not

2 Clark 03.11.10 at 6:04 am

While in general I am suspicious of cause other than driver error, I think that intense computerization of cars can cause this problem. The reason in my opinion is no one can accurately predict how the million of lines of code controlling the entire vehicle will react under every circumstance. It is entirely possible to have rare events occur as a result of a combination of sensor inputs interacting with code to create sudden acceleration. These rare events would be virtually impossible to identify and replicate. This is the Achilles heel of computer programming. Programing requires that all conditions and variables are known in advance. This is simply impossible with our current technology.

3 Ted Frank 03.11.10 at 6:31 am

But why would the computers pick on the elderly? Surely if they were going to discriminate, they’d want to take out the youthful, who are most likely to take up arms against the rise of the machines.

4 Bob Lipton 03.11.10 at 7:22 am

I’d like to see the average age of the Toyota owner on that page.


5 RainerK 03.11.10 at 8:03 am

Back to 1992? What do we make of the fact that electronic throttle wasn’t used before 2003? Did the LAT mention that fact? of course not. Good journalism, as usual!
The floormat issue strikes me as a driver irresponsibility issue, plain and simple. Or do we now make the manufacturer responsible for the safe condition of vehicles for ever?

6 Bob F. 03.11.10 at 8:38 am

You may be right but should compare your findings against the demographics of the owners of the vehicles at issue -does it trend older? That would weaken your theory.

7 meerdahl 03.11.10 at 9:33 am

“Another telling tale about buying age from CNW was that the average age of shoppers choosing a domestic vehicle was 49.4 years old in 2007 — older than the average 42.5-year-old buyer of Asian cars but younger than the 50.6-year-olds choosing European nameplates.

The oldest average shoppers industry-wide were looking at Ford or its Mercury and Lincoln brands, at 54.3 years. GM shoppers averaged 48 years old, while Chrysler shoppers came in at 44.

On the foreign-manufacturer front, an interesting auto rivalry also had interesting statistics: The average age of Toyota shoppers was 46.6 years old, while the average age for Honda shoppers was 51.2.

And while Buick is typically the butt of jokes about buyers who are somewhere between retired and deceased, the average age of a Buick shopper last year was 55.2 years old, considerably younger than the average 63.6-year-old Mercedes-Benz shopper. ”

8 delurking 03.11.10 at 9:40 am

When did Toyota switch from a throttle cable to an electronic sensor-computer-actuator? I would be very surprised to find it was before 1992.

9 A.W. 03.11.10 at 9:55 am

the other problem here is this. The government is regulating an industry it is participating in. That is a very obvious conflict of interest. its like going to a ball game between the yankees and red sox, and you learn that the umpire is also the coach for the yankees.

Or to get deeper, way back in the first exercise of judicial review, Dr. Bonham’s case, the court enunciated a simple principle. No person can be a judge in their own case. this principle was cited by our founding fathers. one can very reasonably say that it is embedded in the idea of due process.

And when the government takes over a car company, and attempts to regulate the auto industry, it becomes a judge in its own case.

That doesn’t mean i currently believe it is all hype, but in that situation you have to be suspicious. you are an idiot if you aren’t.

10 TonyG 03.11.10 at 10:16 am

Clark – It is possible to “prove” a software program is correct, it is hard to do and I have no idea whether Toyota took the expense and effort to do so here. As systems get more complex they get much harder to prove and this is generally why software development companies don’t go to the expense of doing this proof for everything but good software engineering processes can ensure that you deliver a high quality product.

There will still be corner cases where unusual combinations of inputs cause problems and this may be why the group experiencing the issue may not be normalized over the entire user base. That group may have a reason for leaning towards a set of inputs that cause the issue whereas other groups don’t.

That said, I find it hard to believe that there is truly a fault in the system until someone can explain what the characteristic is that causes it to primarily involve older drivers. Sure there could be some combination of driving techniques that these people were taught that create greater susceptibility to this issue but even in software the concept of Occam’s Razor applies.

Rather than spending the money on “legal theories”, they should be spending the money to find the common characteristics in the group and determining if the system really fails based on that.

Personally I lean towards accidental pedal push and panic as the culprit but that’s just my opinion.

11 Ted Frank 03.11.10 at 10:28 am

The average age of a Prius owner is older than the average age of an average car-0wner (in part because of the premium expense of the car), and is somewhere between 45 and 50. The average age of the average car owner is about 40.

NB that younger drivers drive much more often than older drivers, however, so the median age of the driver-mile would be lower.

12 stan 03.11.10 at 11:23 am

If you are looking at deaths, rather than numbers of incidents, would it not be possible that the numbers are skewed by age-related abilities to react and escape from an incident, too? At the bottom end, lack of driving experience, and at the upper end, mental inability to deal with an unexpected problem… Therefore, I don’t think the age distribution of deaths is necessarily indicative of the nature of the underlying problem.
I think that there probably is some sort of problem, and possibly multiple problems, but am utterly unqualified to say what it might be. I was very skeptical of the floor mat theory and am agnostic about the physically sticking throttle theory. My wife and I own two Toyotas, and overall, I don’t really have any concerns. I figure either of us could shut down our car if the throttle stuck. I had to do it once, many, many years ago, when a throttle cable froze in the open position while driving in a blizzard.

The only thing so far that disturbs me about the drive-by-wire systems is ‘traction control”. I put this in quotes, because I find it removes control from the driver and reduces our ability to make the car do what we want it to do. I find traction control a major hinderance for driving in snow, and would like to find ways to disable it.

One solution to all of these throttle concerns would be to have application of the emergency brake (not the regular brake) override and cut off the throttle. If it could re-boot the software, too, this might even be better.

13 Pete Krawczyk 03.11.10 at 11:57 am

The problem I have with this analysis is that it assumes each acceleration event leads to death. I suspect what you’ve discovered is that the elderly are less likely to identify and react in time to an acceleration event, leading to death.

I would further expect that “younger” drivers would be more likely to be able to apply as many techniques as possible (stand on the brake while shifting into neutral while attempting to steer safely).

If the margin of error between life and death in these instances is slim, it would makes sense that those who are not able to react well would be the ones affected. The fact that there are only 2 drivers in your sample between 25 and 40, and another 3 between 40 and 50, speaks to that.

14 CarLitGuy 03.11.10 at 12:02 pm

while programming for the electronics in a modern auto might be thousands, or tens of thousands of lines of code, spread across numerous modules, its quite likely that the suspect portion of the code used by the manufacturer here is likely only a few thousand lines – one of the issues people have made much about is the lack of an electronic throttle cut off when the brake is applied, suggesting that those sensors aren’t monitored.

Assuming the peddle uses some form of potentiometer to determine position, we can surmise that the code in the vehicle need only monitor a reference voltage, and comparative voltage from the potentiometer. Again surmising, the manufacturer likely compares the outputs its computer is receiving from those sensors to an acceptable range, and to one another. The lack of codes in other systems suggests that the WOT “Wide Open Throttle” position being reported from the accelerator pedal to the other portions of the engine control software is occuring normally – so only the code reading the sensor input and testing for fault would be highly suspect. In the grand scheme of things, that’s far easier to isolate and debug than the software and formulas used to maximize fuel economy based on spark timing, perception of “knock” in combustion, barometric pressure, outside temperature, fuel pressure, engine load, engine temperature, combustion efficiency (as measured by various 02 sensors) and the like.

If we, as a public, knew more about how the “expert” rewired the Toyota to command WOT w/o setting codes (possibly by shorting or shunting voltage from one sensor to the other, creating the appearance in system of a fully depressed throttle), we’d have more to work with – but what he did does not appear to be a software error, it appears to be a mechanical fault manufactured by introduction of an external contaminate the software doesn’t test for, because, by design, the vehicle is not supposed to be mechanically altered in that manner. Toyota’s ability to replicate the symptoms with other manufacturer’s vehicles suggests just such an approach is being used.

Windows OS, for example, does not provide the dreaded “Blue Screen of Death” when you unplug your desktop by mechanically pulling the power cord – not because the software is faulty (at least, not in that fashion), but because the software wasn’t written to test for that eventuality, and the hardware it runs on was not designed to maintain power while it did so. Neither does the OS on MACs, or any of the common Linux platforms I am aware of. The expert’s demonstration may turn out to be just as simply explained.

15 L Nettles 03.11.10 at 12:20 pm

Popular Mechanics has two articles on the Toyota Accelerator

This one is on the design

This is on the Gilbert demo modification

16 Tom T. 03.11.10 at 12:25 pm

Regarding reports in other countries, Japan is going to investigate thirty-some cases of Toyota acceleration reported in that country. No way of knowing if there’s any more substance to those than the US ones.

17 CarLitGuy 03.11.10 at 12:53 pm

L Nettles, thank you. The Popular Mechanics articles describe just what I was trying to outline, but with far greater skill, as to the mechanics, and provide greater illustration of the capacity for error detection using only a little code and very few inputs than I had room to list above. Wish I had thought to look for similar articles earlier.

18 kimsch 03.11.10 at 1:09 pm

I wonder a bit about these newer cases. The man in San Diego did not put his vehicle into neutral until after the 911 operator told him to and I heard that he said afterward that he hadn’t done so because he was afraid…

The woman in New York with “sudden acceleration” while leaving her driveway and hitting a stone wall… that one sounds to me like wanting a new car and perhaps help with the mortgage or the latest in plastic surgery a butt enhancement/lift…

call me cynical.

19 MF 03.11.10 at 6:27 pm

It is possible to “prove” a software program is correct, it is hard to do and I have no idea whether Toyota took the expense and effort to do so here. As systems get more complex they get much harder to prove and this is generally why software development companies don’t go to the expense of doing this proof for everything but good software engineering processes can ensure that you deliver a high quality product.

I am a software engineer and have over 25 years of experience at the very highest of the high tech companies. I can tell you that the statement that it is possible to prove a software program is ridiculous in many instances. When you’re dealing with a finite set of inputs and outputs, I’ll agree. But when you’re dealing with real time programming (in other words, events that happen in real time, and the programming is event-driven, and one operation can interrupt the execution of another based on these events), there are an infinite number of possibilities. One cannot possibly “prove” correctness. It is very difficult even to prove near-correctness.

20 Charles Platt 03.11.10 at 8:09 pm

Always remember the overflow condition of the utterly primitive computer on the lunar lander, during the first attempt to land on the moon. And how many times had that system been tested, in every possible failure mode?

I would bet that car systems are potentially vulnerable to heat, cold, electromagnetic radiation, vibration, G forces, and numerous other external variables, even if the code is absolutely totally bug free.

I would also bet that younger drivers may respond with more initiative to sudden uncontrolled acceleration than older drivers, and thus are less likely to have accidents as a result.

I note also that if you pump the brakes while the throttle is wide open, you deplete the vacuum reservoir within about three brake applications, at which point, you have no power assist anymore. I can certainly imagine someone braking, then releasing the brakes, at which point the car accelerates again, so, the driver brakes again–and so on.

Add it all up and I would guess maybe 20 percent of the reported instances could be real, another 50 percent driver error, and the rest consisting of people angling for damages. And I would guess many more cases that were never reported and never became law suits. Indeed I know of one such case personally.

21 marco73 03.12.10 at 9:27 am

One thing that really astounds me, from this entry and other reports I’ve seen in the media, is that there is no mention of a student driver ever encountering SUA. At any given time, there must be hundreds of thousands of student drivers, typically teenagers, driving Toyotas and other cars with all the electronic nannys. Anyone who has ever spent any time with a student driver (I’m on my third), knows that they make all kinds of silly, inconsistent inputs to the gas pedal, brake, shifter, etc. Certainly if there was a an electronic defect, student drivers would hit on the magical combination necessary for SUA more often than fully licensed drivers. Probably the saving grace is that the student driver is sitting next to an experienced driver who can calmly talk them through their mistakes, hopefully without causing any damage. Teenagers crash their cars all the time, and we are conditioned to believe that it is driver error because of their inexperience. But when elderly drivers crash because they don’t react correctly or quickly enough, its the car’s fault? Yeah, right.

22 John Sequeira 03.12.10 at 10:11 am

The sudden acceleration thing happened to my wife and I when we rented a Highlander ~ 6 years ago.

It happened to her, and she freaked out of course but managed to slow it down. I suspected driver error, just like everyone else. Then when driving back to the airport, it happened to me *3 times* – each time I had to put the damn thing in neutral and pump the accelerator to unstick it.

Based on our experience, I think elderly people are less able to respond to their car going psycho on them… is that driver error? (correlation, causation, all that)

23 Paul 03.12.10 at 11:01 am

Fortunately there is a software quality assurance methodology that’s used in mission critical applications (such as civilian aerospace) that is used to assure very high confidence – take a look at DO-178b It is not mandated for vehicles, but more and more companies are using this approach (and the related DO-254 for complex electronic hardware).

24 Will 03.12.10 at 11:41 am

These are deaths, not people who experience. Younger drivers are more likely to have the reflexese to avoid a fatal collision if the acceleration were to occur. This evidence does not say that the sudden acceleration happens to more old people, but rather, when it happens, old people are less able to cope. (No surprise here)

25 WL 03.12.10 at 1:24 pm

If this is the case why do we not see a similar set of incidents across all car manufacturers?

26 Joshua David 03.12.10 at 1:26 pm

Let me point out this: There are lies, there are damn lies, and then there are statistics. It would not even be a challenge to cook up some statistics to refute these statistics. Here’s one: 100% of the lawyers who wrote this article are well known lawyers. How about statistics on who is more likely to die in a car wreck, a frail old person over 60 years of age, or a healthy young person in their 30s? How about a statistic on what age group is more able to control a vehicle at unexpected high speeds? How about a statistic on the overall number of crashes, instead of just fatalities? How about a statistic on how these ages of fatal driving accidents compare against other makes of cars? It’s easy to be a skeptic. It’s easy to cast doubt on something which is confusing or isn’t easily explained, but don’t try to present yourself as an unbiased skeptic when you only present a tiny piece of the statistical evidence required to draw a conclusion, and it only consists of the statistics that favor that obvious bias.

27 Ted Frank 03.12.10 at 2:19 pm

Sure: in 2008, 16% of automotive accidents involving fatalities involved drivers 60 or over. For Toyota sudden acceleration, the number is 54%—which is consistent with the disproportionate rate of elderly drivers in Audi and GM sudden acceleration.

Comments on this entry are closed.