Ages ago, in a throwaway post called "More Pit Bulls, Less Crime?" I drew upon the sociometric debate over the deterrent effect of right-to-carry gun policies to speculate that notorious dog breeds such as Rottweilers and "Pit Bulls" were probably getting a bad rap. My reasoning started with the easily observed disconnect between sensational media accounts of pooch-wrought carnage and the statistically infinitesimal real-world risk of death by dog. When snarling canines maul youngsters and and urbanite lesbians, the public response is characterized by deep-rooted fear and morbid fascination — pretty much what we should expect, considering that for most of our evolutionary history people had good reason to fear being eaten alive by animal predators. You get grisly headlines and red-font Drudge links and tragic newsmagazine close-ups of bawling moms and dads and soon the public is stirred to Do Something about another Very Serious Problem. Municipalities hastily impose breed-specific bans, and no one considers that there might be another side that can't be dismissed as  mere PETA-hearted sentimentalism. 

So you drill through the panic-mongering litigation-bait in your first round of Googling until eventually you discover the stats and it turns out that in the United States maybe just over a dozen people on average are killed by dogs (of all breeds) in a given year. Maybe an even twenty, depending on your source. Of course, most of the victims are wee tots, but before before you hit back with your "one child's death is too many" applause line, keep in mind that children face greater mortal risk from cribs, swimming pools, buckets of water, bicycles and their own parents. Relative to any other calculable mortal risk, the dogstats are just tiny. In the scheme of public-spirited epidemiology, they barely register. And to the extent that the yet tinier subsets of  breed-specific fatal attacks  may be considered (here is a recent report from Dogbite Law, and her is an  older one from the CDC), the question is far more complicated than it seems on first pass, in part because those tinier numbers keep breaking down into even tinier subgroups; unneutered male Pits and Rotts account for proportionately more canine mayhem than their female and neuter littermates. And then there is the problem of negligent breeding subcultures, which leads to the sort of inconvenient sociology that really shouldn't be overlooked.  

But there's no need to get bogged down in an interesting subject. Because even if we grant the lawyers their scariest stats and assume the worst about stigmatized breeds, the relevant hook is clear. People are afraid of dogs. The curiously unexplored empirical question then becomes: does this  fear translate into a crime deterrent, to the benefit of dog owners? — with the ancillary question of special concern following: if such a deterrent can be demonstrated, does the effect differ by breed? That is, do Rottweilers and Staffordshires and other feared pedigrees confer a greater crime-reducing benefit to their owners than less notorious breeds? And finally, if there is a deterrent that tracks by breed, is the effect great enough to offset the supposed threat posed by liberal dog-owning policies? Do Pit-Bulls save lives?

When I first floated this question, I assumed it would be a short matter of time before some Levitt-styled quant-nerd would begin crunching the stats. I figured that the faddish move toward breed-specific bans in major cities (most notably in Denver) would prompt at least a few econometricians to wonder what I wondered, and to investigate what I have neither the time nor smarts to investigate. I'm told that such naivety is common among nonacademic admirers of academia; if you stumble upon a notion,  you assume that someone has already worked it out, or soon will. But that's just dumb cause there ain't no research been done. Not of the kind I have in mind.

It is interesting about the gun thing. It's similar. And it's different.

With
the empirical research on guns and crime, you started out with a lot
of overconfident sociology cultivated in the cultural fear of crime and motivated by frank anti-gun
prejudice. Eventually, matters were complicated by a number of
population surveys (perhaps most notably those conducted by
criminologist Gary Kleck and later written up in his book, Point Blank),
where reports by convicted criminals and civilian gun-owners converged
to provide evidence that guns were being used in self-defense far more
frequently than the old-school sociologists had allowed, and usually without a shot fired. A one-dimensional and largely political debate thus gave
way to complex matrix of possibilities.

When the concealed-carry
controversy was heating up anew, there was John Lott with his
data-laden county-comparative charts and graphs that seemed to show
that liberalized permit laws, contrary to dire predictions, had the
effect of reducing crime rates to some significant degree. Lott's
critics were left with plenty of nits to pick, but mostly they just
kept shuffling the variables until they could seize upon some strange
set of controls to massage the regressions into something nearer to a wash. People like to make fun of Lott, and he set himself up for a lot of it, but his work really did change the terms of the debate. I
don't think there are many serious field scholars who still hold to the
simplistic "more guns cause more crime" thesis, even if that was the
original default. The center-stage debate
now focuses not on the question of  how much crime is attributable to
firearms, but on the extent to which gun-crime is mitigated by
the countermanding deterrent benefit of gun ownership. The null hypothesis
isn't dead, but the one-tailed conceptualization of the matter has been
traded in favor of a more nuanced cost-benefit analysis.

As
with the presently stalled move to ban notorious dog breeds, the old gun control movement was based on a superficial reading of reality. In the early rounds, gun policy research suffered from a value-bound failure of
imagination and I think this is what we're seeing now with the dog issue. When you look in only one direction, the view is
limited, and perhaps skewed by intrusive images from the id. Demon-guns
and demon-dogs.  Oh my.

I think the dog question differs from the gun question in a couple
of important respects. First, there is the simple matter of numbers,
already mentioned. If you exclude gun suicides (as I must insist), the 
CDC's mortality calculator
sets the latest gun homicide rate in the United States (for 2005) to
be somewhere around 4.5 per 100,000, representing  roughly 12,400
criminally motivated kills. With at least a couple of hundred million
firearms in private possession, any deterrent effect must be considered
against this relatively high measure of lethal harm. But when it comes
to dogs, the mortality stats are so small as to nearly defy stable
expression in standard per capita terms that can be adjusted over time. Some years there are fewer
than 10 deaths attributed to dog attacks. Other years there are just around 30, and if you believe the National Canine Research Council,
the numbers may be on the decline. It's hard to know, really, since subtle shifts in the relevant populations combine with stat-noise to keep things slippery. But with lethal dog attacks occurring so infrequently, it would seem that even the
slightest crime-reducing counter-effect could be enough to overturn the
prevailing assumption.

On the other hand, the
theorized mechanism of gun-facilitated crime reduction needs to be
taken into account. Especially with the concealed-carry angle, where
the effect is usually conceptualized as a kind of positive
externality that benefits members of the general population whether
they own guns or not. Here and again, the idea owes to rational
criminal behavior. A stick-up fiend is thwarted when one or
two would-be victims  brandish heat and he revises his worldview to
account for the newly increased risk. Since he doesn't know who's
packing and who's not, the standard MO becomes too dangerous, or too
costly. Until everyone is a little safer. With dogs, the situation is different. Any deterrent effect owing to
dog ownership would seem to confer more narrowly to the dog owning population, without necessarily spilling over into
anything  more generalized. On a leash, behind a fence, or barking
behind the front door, dogs remain conspicuous. You don't tuck them
under your belt, except on very special occasions.

If you want some graspable grounds for breed-specific crime deterrence, here's a decently referenced snip from a police home security manual:

Dogs have proven to be an effective deterrent to burglars. Researchers Paul  Cromwell, James Olson and D'Aunn Avary write in their book, Breaking and Entering: An Ethnographic Analysis of Burglary (Sage, 1991) "When asked what were considered absolute "no go" factors, most burglars responded that dogs were second only to occupancy. However, approximately 30% of the informants initially discounted the presence of dogs as a deterrent. Yet, during "ridealongs" the sight or sound of a dog at a potential target site almost invariably esulted in a "no go" decision. 

And:

Professional dog handlers suggest some breeds are better at "watchdog" duties than other breeds. Dr. Stanley Coren in his book, The Intelligence of Dogs: A Guide to the Thoughts, Emotions and Inner Lives of Our Canine Companions (Bantam, 1995) consulted experts and found the following breeds to be good "guard dogs": Bull Mastiff, Rottweiler, Doberman Pinscher, Komondor, Puli, Giant Schnauzer, German Shepard, Rhodesian Ridgeback, and Kuvasz. Good "watchdogs" are the Rottweiler, German Shepard, West Highlander White Terrier, Yorkshire Terrier, Cairn Terrier, Airedale Terrier, Poodle, and Miniature  Schnauzer. Breeds such as the Rhodesian Ridgeback and Rottweiler are good guard and watch dogs, but do require close supervision and obedience The worst watch dogs identified by Dr. Coren are: Bloodhound, Newfoundland, English Bulldog, Pug or Scottish Deerhound.

The manual focuses on burglary and home invasion, but it seems reasonable to assume that publicly exposed dog-walkers would also make less than optimal targets for robbery or rape. I figure a leashed Mastiff  signals a louder "no-go" to a would-be aggressor than would a cute bugeyed pug. After all, we know that criminals are rational actors. And the same breed-phobic bias that reads as edging moral panic when expressed in opinion polls makes perfect rational sense when expressed by a crook sizing up his mark.

If you want another starting point, there is the U.S. Pet Ownership & Demographics Sourcebook, published by the AVMA.  I haven't read it, but according to the TOC, it contains statistics on dog ownership by breed and region. Could be useful. So could the regional data gathered at the Breed Specific Legislation repository.     

I understand that real-world-factored regression analysis is a byzantine bitch, but it sure seems like you could get things rolling simply by taking before and after criminological snapshots of  municipalities where breed bans have been imposed. If there's a difference that deviates from general trendlines, then let me know. Of course, the question of ordinance enforcement may turn out to be crucial, but it shouldn't be too hard to scour the magistrate records to determine where dog-snatching cops are playing by the letter and where they aren't. So maybe you need to control for enforcement, too. OK then. Do that, too. Then get back with me. Matters may become more complicated later, but just start with before and after, and let's not be distracted by those boring lawyer-pimped dog-bite stats. Just see if there's a possible effect in overall crime trends, in whichever direction. I'll be curious to see what you come up with. And keep in mind the exceptional microtude of the  mortality stats to be trumped, if my precious pits are to be vindicated.  Even a blip could be significant.

You have your work cut out for you. Expect false starts. Now get to work.

I'll be waiting in front of the TV. Surrounded by cats.

Memento mori.