Slog News & Arts

Line Out

Music & Nightlife

« County Politics Get Interestin... | And While We Were Arguing Abou... »

Thursday, June 19, 2008

More Data for the Great Slog Pit Bull Debate, ‘08

posted by on June 19 at 17:42 PM

I put this figure together, based on data from a May 1997 MMWR.

Undeniably, more people have been killed by “pit bull” breeds, or dogs with some “pit bull” mixture, than any other breed. This does not prove the breed is bad. Fatality is a combination of a whole variety of factors—age of the victim, size of the dog, the training of the dog and the breed.

Also of note? Nearly two-thirds of fatal attacks involve unrestrained dogs, at the owners house or running wild in the streets.

Fatalities from dog attacks are very, very rare. Vastly more people are bitten by dogs each year, often with enough injury to require medical treatment. It’s much easier to study what leads to a dog bite, accounting for breed, training and so on. A 1994 study in the journal Pediatrics did just that. The money table:

(Click for a larger version.)

“Pit bulls” aren’t included in this study. This study is based on data from Denver, where pit bulls were already banned.

If the matched odds ratio 95% confidence interval is entirely below 1.0, this factor makes a bite less likely to happen. If the interval is completely above 1.0, this factor increases the risk of a bite. If it spans 1.0, the factor doesn’t necessarily increase or decrease the risk of a bite.

Big factors increasing the risk of a bite? Breed. Male dogs. Non-neutered dogs. Big dogs (> 50 pounds). Young dogs (<5 years old). Dogs chained in yards. Dogs that don’t get regular rabies vaccines. Having a child in the house. The owner not bothering to license the dog in the past year.

Reducing the risk? Dog that were ever disciplined by a takedown (holding a dog to the ground, on its back, while holding its neck) or stringup (lifting a dog by its neck chain). That’s it.

Training and discipline didn’t seem to significantly reduce the risk of biting, at least in this study.

Have at it!

RSS icon Comments


My cat is following me through the house. Someone, please come help me. Eli knows where I live. But he's on vacation! Oh god...

Please. I'm not kidding. My cat is going to kill me...she's chewing through the closet door right now...

Posted by Jubilation T. Cornball | June 19, 2008 5:51 PM

Cool, but Jonathan, where were you when this study was released? Grade school? I see your point, but surely there is better data out there that isn't more than a decade old.

Posted by hohoho green giant | June 19, 2008 5:55 PM


Unfortunately, this is such a small study and the CI's are, generally, ridiculously wide. Also, (not such a big deal in an observational study like this) there's a bit of a multiple comparison issue: if they adjusted the CI's for OR's for every test they conducted they'd probably all contain 1.

I'm also guessing the data is self-reported and thus of questionable quality.

Posted by umvue | June 19, 2008 5:57 PM

According to your first graph only 70 people were killed by pits over 20 years. I can live with that, even if hundreds were injured over that time period.

When folks start talking about wiping out a breed, I get fucking nervous.

Posted by rutro | June 19, 2008 5:58 PM

So is it statistically incorrect to try to say that the opposite of factors that increase do, in fact, reduce? As in "neutering your dog reduces the risk"? I suspect that maneuver is something that true statisticians would condemn.

I still want to know if it's true that there's never been a fatal attack by a neutered dog!

Posted by leek | June 19, 2008 6:04 PM

@1 Perhaps if you stopped wearing tuna cologne your cat would back off.

umvue, please explain your relationship with statistics, your knowledge is vast and impressive...

Posted by PopTart | June 19, 2008 6:05 PM

Oh, and...

In re the top graph, we need to know the denominator. This plot could be indicating that Pits are the gentlest of breeds - it really depends on what proportion of the dog population they represent.

Posted by umvue | June 19, 2008 6:06 PM


Fair points, every one. It is what it is. I tried, at least, to explain to people how to properly interpret an odds ratio.

hohoho green giant --
I liked this study the best of those available to me. Just because data is old doesn't automatically make it invalid. The best argument is newer training techniques might change the results.

A newer study does exist, in the Journal of the American Veterinary Association. I don't have access, and am unwilling to pay $10 to help settle a slog debate. ;p

If anyone has a copy, and wants to e-mail me, I'd be glad to read it.

Posted by Jonathan Golob | June 19, 2008 6:07 PM

The top graph is out of 129.

Posted by Jonathan Golob | June 19, 2008 6:09 PM

I trust Amy Kate Horn.

Posted by elenchos | June 19, 2008 6:11 PM

@4: Why? Are you secretly part-pitbull?

Posted by banjoboy | June 19, 2008 6:14 PM

If the non-neutered/fatal attack link is really as strong as AKH posted, I think that a requirement for all pit bull and pit bull mix dogs to be spayed or neutered unless the owner was specially registered with the county, with x, y, and z requirements met, would be totally defensible, if diifficult to enforce.

(But then, a "breed ban" would also be difficult to enforce, given the number of pit-type mixes out there.)

Posted by leek | June 19, 2008 6:15 PM

@9 do you mean those fatalities are rates out 129 for each breed? How can it be a rate spanning such a long period? (where does 129 come from?) Maybe I don't understand, but I agree with umvue that the fatality count can't be interpreted without knowing how many dogs of each breed are around...

Posted by Jude Fawley | June 19, 2008 6:16 PM


If we lived in a world where bear attacks were a daily occurrence then I could see a reason to have a pit. But since we don't so they're 100% expendable.

Posted by karst | June 19, 2008 6:19 PM

Leek -- It is statistically and scientifically and logically OK to look at and state things sideways sorta. If the factor under consideration is binary (e.g. neutered vs. not neutered) then if being neutered is associated with a decrease in the odds of a bite it follows that not being neutered is associated with an increase in the odds of a bite.

PopTart -- I AM a statistic (and so are you). I got statistics initials (M.S., PhC). My role in life, as my advisor once told me, is to go fourth into the light and spread darkness.

Jonathan -- Just spreading darkness. That is, trying my best to raise doubt.

Posted by umvue | June 19, 2008 6:19 PM

Jude Fawley--

129 total fatalities in the time period, for which breed-identifying information was available. There could be a bias, that pit bull attacks are more likely to come with breed-identifying information than other fatal attacks.

Read the MMWR report. It does an admirable job of breaking down the methodology and complexities in this study.

Posted by Jonathan Golob | June 19, 2008 6:20 PM

@6 -- Kus ima shelcha with your tuna. I find a little mackerel juice on my pulse points keeps me alluringly provocative all evening long. Why, when I walk through The Cuff, the crowd parts like the Red Sea...

Posted by Jubilation T. Cornball | June 19, 2008 6:21 PM

Dan Savage give pit bull can of Rize, pit bull drink Rize, mellow out, Snoopy out, PIT BULL BECOME SNOOPY!


And no, I don't feel like letting it go.

Posted by The Incredible Sulk | June 19, 2008 6:23 PM

A correction. 199 total for the top figure.

Umvue: Keep spreading the doubt. I'll keep introducing empiric evidence as a way of settling questions. We'll do some good for the world.

Posted by Jonathan Golob | June 19, 2008 6:25 PM

I'm completely shocked that St. Bernards killed anyone. A friend of the family had one when I was growing up and I always thought it was the sweetest, mellowest most tolerant dog.

Guess I gotta read up some on the breed, because I want one some day.

Posted by K X One | June 19, 2008 6:28 PM


Posted by Pit Bull | June 19, 2008 6:43 PM

The authors of the MMWR report recognize (some of) its limitations. They write, in part:

" definitively determine whether certain breeds are disproportionately represented, breed-specific fatality rates should be calculated. The numerator for such rates requires complete ascertainment of deaths and an accurate determination of the breed involved, and the denominator requires reliable breed-specific population data (i.e., number of deaths involving a given breed divided by number of dogs of that breed). However, such denominator data are not available, and official registration or licensing data cannot be used because owners of certain breeds may be less likely than those owning other breeds to register or license their animals (3)."

So, we have an estimate of how many deaths there were due to a breed but we don't know the number of dogs of that breed. This could mean, for example that if there is only one St. Bernard in the US it is a particularly vicious one. If there are millions of Pit Bulls then they are generally as gentle as kittens. I'm exaggerating but that's the limitation of the report/graph.

Posted by umvue | June 19, 2008 6:46 PM

Not on the subject of pit bulls, rather on the subject of randomness and I assume some statistics as well, I was wondering if any of you had read The Drunkard's Walk yet? I hear it isn't anything new but that it is well written so non-science and math geeks like me can understand the subject.

Posted by PopTart | June 19, 2008 6:48 PM

It's 11 year old data and it doesn't adjust for the number of each breed. But it proves Pit Bulls are killahs! We got graphs, man.

Posted by Bob | June 19, 2008 6:54 PM

Oh man, there's totally an unrestrained pit bull wandering around the parking across the street right now. Windows up! Secure your children!

Posted by Garth | June 19, 2008 6:57 PM

@24 -- Well, OK. But what in the past 11 years to your knowledge has occurred to meaningfully change the behavioral modeling of the breed? Anything? If not, then the data stands.

George W. Bush doesn't count.

Posted by Jubilation T. Cornball | June 19, 2008 7:08 PM

Yeeahh man, right on. Pit bulls have totally, like, mellowed out over the past 11 years. Must have been inhaling all that good stuff from those inner city black kids right duuuuuuude?

Posted by karst | June 19, 2008 7:13 PM

Maybe it *IS* still stupid owners, but stupid owners are more liklely to own those breeds?

Posted by The Baron | June 19, 2008 7:13 PM

So, I believe that the question about the denominator in the graph you made is not "Out of how many dog-induced fatalities were pitbulls responsible for 70" - we could derive an approximation for that number by adding all the different breeds' number of fatalities together.

No, the question is, "Out of how many total pitbulls existing at the time were 70 responsible for fatalities?"

Example One: four St Bernard-caused fatalities out of a total national population of 2 million St Bernards, and 70 pitbull-caused fatalities out of a national population of 1 million pitbulls.

Example Two: 4 St Bernard-caused fatalities out of a total national population of 1 million St Bernards, and 70 pitbull-caused fatalities out of a nation population of 80 million pitbulls.

You see how, without knowing the denominator, it is impossible to know what significance (if any) your very attractive graph might have.

Posted by Thisbe | June 19, 2008 7:18 PM

@26, I don't know, and the point is neither do you or anyone else. Maybe it's gotten worse, maybe better or no change. It's all speculation without the data. Maybe other breeds are making a run for the money. How about Presa Canarios? They've killed a few folks in a relatively short period of time and they are a pretty rare breed.

Gotta have the data.

Posted by Bob | June 19, 2008 7:28 PM

Indeed, it would be better if we knew the exact denominator for the time period of the study. The lack of this denominator doesn't make the data useless, just imperfect.

We can take this information, and perform a case-control study. We have 199 dogs that we know caused fatalities. We can go out into the population and find 199 random dogs that have not killed someone.

Seventy of the 199 killer dogs were pit bulls. If being a pit bull doesn't matter, than we'd expect about 70 of the non-killer dogs to also be pit bulls. If being a bit bull increases the chance of being a killer dog, we'd expect quite a bit less than seventy of the non-killer dogs to be pit bulls. If less likely, more than seventy of the non-killer dogs would be pit bulls.

I'm grossly simplifying here, but this is pretty much how the pediatrics study from which I grabbed the table did their work.

Posted by Jonathan Golob | June 19, 2008 7:35 PM

JTC -- One thing that's changed since this study was conducted, interestingly enough, is that Denver (the municipality wherein the study was conducted) has apparently repealed its Pit Bull ban.

PopTart -- Haven't read The Drunkard's Walk yet I heartily recommend it based on a quick glance, reviews, and the sense that it can only help. And OUCH! "...randomness and I assume statistics as well." Some of us like to think that with probability theory and statistics we're kinda dealing with the randomness thing on the front lines.

Jonathan -- UW may not give you direct access to the JAVA article but you can request it via interlibrary loan. I did. We'll see what happens. You know, we're trained to think about what the ideal data would be and so we imagine the randomized clinical trial... Johnny gets sent to the pen with the pit bull and Mary gets sent to the pen with the poodle... Probably can't do that trial.

Posted by umvue | June 19, 2008 7:36 PM

@30 -- Rube. You mean to say, it's all speculation until YOU see data that supports your view.

For your sake, we'll just suspend mammalian behavior, as recorded by science for, oh, 1000 years plus, and assume that pit bulls, of all god's creatures, can totally change their behavioral model in 11 years.

Sure. Why not? Happens all the time.

Eat me.

Posted by Jubilation T. Cornball | June 19, 2008 7:38 PM

@32 Sorry, I didn't mean to offend, see I need to read the book so I can understand!

Posted by PopTart | June 19, 2008 7:48 PM

@33, you're an idiot. You're on stupid ignore until you show otherwise. Bitch.

Posted by Bob | June 19, 2008 7:53 PM

@35 -- Ahh, yes, the mighty "ignore." It's what fat gamer trolls have instead of god.

Enjoy the silence.

Posted by Jubilation T. Cornball | June 19, 2008 7:59 PM

Jonathan -- I like the idea of a case-control study. I don't imagine ever trusting the results of such a study though. First of all there's questions re ascertainment; not all dog bite fatalities were captured. Second, there's the changing fashion and popularity of dog breeds (weren't none of those deadly pocketbook things back in '79).

@32 -- No offense taken. Randomness is a beautiful thing - whether it exists or not.

Posted by umvue | June 19, 2008 8:16 PM

Tasty, tasty huumaans!!! Keep debating your silly statistics -- it means more huumaan faces for me to eat!! Yummm!!!

Posted by Hungry pit bull | June 19, 2008 8:16 PM

No, useless is actually pretty accurate here. You have a small number of events spread out over 17 years and 10 million square miles. The relative populations of these breeds have presumably changed as breeds have gone in or out of fashion, and there's also the geographical distribution of the breeds to consider. So your case-control idea is a complete bust.

Absent even the vaguest sense of the denominator, we could just as easily conclude that all of these breeds are equally likely to kill and that Pit Bulls just happen to constitute 37% of the total dog population. Or, given that 2/3rds of the fatal attacks involved unrestrained dogs, maybe all of these breeds are equally represented in the population but Pits are just that much more likely to be unrestrained.

We don't need an "exact denominator" to get something useful out of these data, but we do need more than you've given us.

Posted by Bison | June 19, 2008 8:21 PM

Uh, your first table is useless without taking into account the total number of each breed. It's a bit like the news when they announce that Honda Civics are the most stolen car.

Posted by bob | June 19, 2008 9:58 PM

Mr. Cornball, do be careful with Bob the Internet Tough Guy. Were he to type a post to the effect that in a bar fight he would defeat both you and perhaps even you and one or more of your comrades, and then post this statement, you would be unable to respond. Much like the powerful "YOUR A FAG", the classic "bar fight" post has no defense.

I've seen Bob in action and he is not afraid to say he could win in a bar fight. Not afraid to say it at all, Mr. Cornball...

Posted by elenchos | June 19, 2008 10:13 PM

Ouch. I was actually for banning the breed before this article. 70 deaths in 17 years? For an average of 4.1 deaths/year? In the entire country?

More people are killed by slipping in their bathtubs. More people are killed by mislabeled prescriptions. More people are killed, I will speculate, from heart attacks brought on by reading hyperventilating crusades on the internet.

Please, Stranger, turn your attention to something more important. Like Iraq. Or gay rights. Or gas prices. Or Seattle's stupid "no parking north of here" signs. Or the uneven application of asphalt at the edges of our gutters.

Because this crusade is venal and pointless and masturbatory.

Posted by also | June 20, 2008 12:19 AM

How many of the 70 deaths were of innocents? That is, if someone trains their dog to be mean and the dog turns and kills them, it hardly seems fair to count that death against the breed.

Posted by bob | June 20, 2008 12:37 AM

The single quote in front of 08 should be an apostrophe.

Posted by Eric F | June 20, 2008 1:04 AM


Exactly. Pits and Rotts are a lot more common than the other breeds listed.

Look at the huskies and malamutes on that list. Those dogs aren't particularly common. Most of the dogs on that list aren't particularly common; I'd say all but maaaaybe the Shepherd are less common than the Pitbull.

So of course Pits top the list.

Posted by violet_dagrinder | June 20, 2008 7:48 AM

I hate it when my Slog posts magically disappear.

Yeah, Pits are vastly more common than most of the other breeds on that list. So of course they top it.

I think the thing about the takedown/stringup discipline being effective is really telling. Obedience classes do not teach people how to manage big, strong dogs. FUCK clicker training. If a big dog shows any behavior that is even remotely dominant, they need to be shown, in dog language, that they are not the boss of anything, ever. And the methods described are exactly how that is done (among lots of other, more subtle things; the dog never eats first or eats human food, or goes through doorways first, or sleeps on the bed. Those sorts of things confer status to dogs. It's silly, but true).

I think it's really telling that much less common dogs (Huskies/Malamutes, Chows) are so well represented on that list. Those dogs are NOT breeds that tend to be encouraged to be nasty. They just ARE nasty. If any breeds should be banned, those should.

Posted by violet_dagrinder | June 20, 2008 7:56 AM

I guess the first post was just delayed. Whoops. Anyway.

Posted by violet_dagrinder | June 20, 2008 7:57 AM

Holy shit, who warps a Great Dane to the point it kills someone? Those people are the real meance.

@40 and 46 have good points: what is the rate of fatal attacks to overall population numbers for that breed?

Posted by Jessica | June 20, 2008 8:04 AM

@40 and 46 bring up a good point: the data is more telling when you look at the fatal attacks per population of the breed.

And who the fuck warps a Great Dane into a violent dog? Those assholes should be dragged into traffic and shot. Danes are, obviously, the sweetest dogs ever.

Posted by Jessica | June 20, 2008 8:07 AM


Danes being on that list surprised me a little bit too. But all breeds have some crazies.

They DO tend to be kinda neurotic. And those breeds, the ones that can be sort of shy and skittish, will then tend to be unpredictable. Big, unpredictable dogs are dangerous.

But it just goes to show, it's not about the breed.

Posted by violet_dagrinder | June 20, 2008 8:52 AM

I wonder if the handful of St. Bernard fatalities were clustered around 1983 release of Cujo.

Posted by lostboy | June 20, 2008 9:12 AM

Who cares if pit bulls are the most common? If pit bulls are banned, it's not like every former pit bull owner is going to buy a dangerous(?) Chow to replace it.

That kind of logic is like saying if handguns are banned, every gun owner is going to go buy poison blowdarts to replace them. No, both pitbulls and guns are popular and dangerous because assholes think they're cool. Those asshole don't think poison blowdarts and chows are cool.

Posted by poppy | June 20, 2008 9:27 AM

in high school, i worked at a vet clinic / kennel, where we had a couple of great danes as boarders. put a 130lb dog that is already skittish/neurotic in a small enclosure without adequate exercise and socialization, and it turns aggressive and, frankly, kind of scary.

Posted by jon c | June 20, 2008 9:31 AM

Akitas, huskies, and Alaska malamutes, are probably more closely related than all of the dogs lumped into the pit bull category. If you lump them together they win.

Posted by hillpagan | June 20, 2008 9:38 AM


Jonothan Golob said:

A newer study does exist, in the Journal of the American Veterinary Association. I don't have access, and am unwilling to pay $10 to help settle a slog debate. ;p

Check your email.

Posted by elenchos | June 20, 2008 9:54 AM

At least where I come (Chicago) from pits are way more common than any of these other breeds. I believe I heard that they make up about 30% of all the dogs and obviously a much higher percentage of dogs who are large enough to do damage. Was this taken into account at all?

Also pits are used as fighters here, so they are severely abused and trained to fight- what those dogs do can't be used against well cared for pets.

Posted by Allie | June 20, 2008 10:25 AM

One big risk factor you didn't mention was for a female dog that had had at least one litter. Presumably, more unpredictable when they have puppies, and this was an even larger risk than for unaltered male dogs. Moral of the story seems to be: Be a good owner - people who fix their dogs, care to license them, take them to the vet and to obedience classes will have fewer problems than trailer trash looking for yet another extension to their flabby members.

Posted by Tlazolteotl | June 20, 2008 10:53 AM

Interesting that Wolf mixes, which are already banned in most places, account for so few deaths in comparison. I suppose that has a lot to do with the fact that there are fewer of them.

That actually makes me wonder. That first chart doesn't seem to take into account how many of each dog there are. There are certainly a lot of German Shepherds out there, they're one of the most popular breeds, but there are few deaths by them. While I've noticed more Pit Bulls in the last few years it's typically been a pretty rare breed. I think it would be helpful to know what percentage of the population is Pit Bull. If only 2% of all of the dogs are Pit Bulls but they kill almost as many people as all other breeds combined then I think you've got a pretty good argument to ban them.

Posted by Colin | June 20, 2008 1:19 PM

Akitas, huskies, and Alaska malamutes, are probably more closely related than all of the dogs lumped into the pit bull category. If you lump them together they win.

That would work well if they were related, but aside from being part of the AKC "working group" they're not. Just because they have similar appearance doesn't mean they have the same lineage. It'd be like throwing chihuahuas, dachsunds and pomeranians into one category because they're "small."

Where are the psychotic Samoyeds in your assessment? They look similar to huskies, akitas and malamutes but are much more prone to brain-crazies.

Posted by Buffalo Bill | June 20, 2008 4:02 PM

Of which I speak re: samoyeds, above: I've had numerous huskies and malamutes over the years and have had very few issues with any of them. With a couple of samoyeds, however, although they're generally thought of as a "happy" breed, there were a couple of times we had to take a baseball bat to them to keep them back, and it didn't slow them down at all. They were fucking NUTS.

Posted by Buffalo Bill | June 20, 2008 4:07 PM

Bad enough, but those numbers are dwarfed by the number of brain cells killed every time someone takes home a papillon.

Posted by butterw | June 20, 2008 4:36 PM

What's a papillon? Isn't that a beef cut?

Posted by Buffalo Bill | June 20, 2008 5:15 PM

"While I've noticed more Pit Bulls in the last few years it's typically been a pretty rare breed. I think it would be helpful to know what percentage of the population is Pit Bull. If only 2% of all of the dogs are Pit Bulls but they kill almost as many people as all other breeds combined then I think you've got a pretty good argument to ban them."

As far as I can tell there aren't good numbers to prove this one way or the other, but I think you're wrong about how common Pits are. My wife used to be a manager at the local Humane Society, and she said that Pits were one of the most common breeds in their shelter population, possibly the most common. And if you watch any of the Animal Cops shows on Animal Planet, you'll see that Pits are extremely common in other shelters as well.

Posted by Bison | June 20, 2008 8:08 PM

Comments Closed

Comments are closed on this post.