A politician takes to a podium. Behind them, a graphic shows a bar chart: pit bulls responsible for the majority of fatal dog attacks. The audience gasps. The journalist covers the press conference. The statistic circulates on social media. Within days, a city council has a BSL proposal on its agenda.
What never makes it into the news cycle is this: that bar chart is built on data so compromised that the Centers for Disease Control stopped collecting it. The methodology behind it would not survive peer review. And the researchers who actually study dog bites for a living believe those numbers tell us almost nothing useful about preventing attacks.
Understanding dog bite statistics — what they measure, what they miss, and how they are manipulated — is essential to evaluating any breed-based legislation. Here is a rigorous look at what we actually know.
The Scale of Dog Bites in America
Dog bites are a genuine public health concern. The American Veterinary Medical Association estimates that approximately 4.5 million Americans are bitten by dogs each year. About 800,000 of those bites require medical attention. Roughly 30 to 50 people die from dog bite injuries annually, a number that has remained relatively stable for decades despite population growth.
These numbers sound alarming until you consider that there are approximately 90 million pet dogs in the United States. The vast majority of dogs never bite anyone seriously. The average dog bite fatality rate is roughly 0.00005% per year — meaning for every dog alive in America, the probability of it killing a person in a given year is about one in two million.
Dog bites are tragic. Dog bite fatalities are rare. Preventing them requires understanding what actually causes them, which breed statistics consistently fail to capture.
Why Breed Bite Data is Unreliable
The fundamental problem with breed-specific bite statistics is breed identification. Every dog bite report that attributes an attack to a specific breed relies on a human visually identifying that breed. Research has repeatedly shown this identification is wrong at alarming rates.
A landmark study published in the Journal of Applied Animal Welfare Science tested shelter workers, veterinarians, rescue volunteers, and other animal professionals. These were people who worked with dogs daily and considered themselves knowledgeable about breeds. When asked to identify pit bull type dogs from photographs, they agreed with DNA testing only 36% of the time. Two-thirds of their identifications were wrong.
This is not a marginal problem. If every bite attributed to a pit bull has a two-thirds chance of actually involving a different breed, pit bull bite statistics are fundamentally meaningless. The data on breed misidentification is unambiguous: visual identification is too unreliable to form the basis of any policy.
The problem compounds with mixed-breed dogs, which make up a substantial portion of the dog population. A dog that is one-quarter pit bull and three-quarters Labrador might be visually identified as a pit bull, a Lab mix, or something else entirely depending on which physical features happen to be most prominent. That same dog would appear in bite statistics as whatever the observer decided it looked like.
The CDC's Decision
The Centers for Disease Control stopped collecting breed-specific bite data in the 1990s because they concluded the data was too unreliable to be useful for public health purposes. The CDC determined that breed identification errors and missing population data made breed statistics inadequate for risk assessment.
The Missing Denominator
Even if breed identification were reliable, bite statistics would still be misleading without population data. Raw bite numbers tell you how many bites were attributed to a breed. They do not tell you what proportion of that breed's population was involved, which is the actual risk measure.
Consider a hypothetical: if large dogs account for 60% of fatal bites, but large dogs also account for 60% of the dog population, their rate is exactly average. They are neither safer nor more dangerous than the dog population as a whole. Only per-capita rates — bites per population unit of each breed — allow meaningful comparisons, and those calculations are impossible without reliable population data.
No reliable census of dog breeds exists in the United States. Ownership statistics come from voluntary surveys, AKC registrations (which capture a small fraction of dogs), and shelter records. None of these sources provide the population denominators needed to calculate true per-capita bite rates by breed.
What Actually Predicts Bite Risk
When researchers control for the variables that breed statistics ignore, a consistent picture emerges of what actually predicts dangerous dog behavior. Breed is not on the list of significant factors.
A 2013 study in the Journal of Veterinary Behavior examined the circumstances of fatal dog attacks and found that 80% involved at least one of the following factors: no able-bodied adult present, the victim had no familiar relationship with the dog, the dog was not neutered, the dog was kept chained, or the dog was obtained for guarding or fighting purposes. These factors cut across every breed.
Owner behavior is consistently identified as the strongest predictor of dog bite risk. Dogs that are poorly socialized, kept chained or isolated, trained using aversive methods, or owned by irresponsible handlers are dangerous regardless of breed. This is why breed-neutral dangerous dog laws that focus on owner accountability consistently outperform BSL.
Dog size matters independently of breed. Larger dogs cause more severe injuries when they bite, which is physics rather than genetics. This creates an apparent severity bias in bite statistics: the bites that make headlines and appear in fatality data are disproportionately from large dogs simply because large dogs cause more damage. This does not mean large dogs bite more frequently — only that their bites are more likely to require medical attention or cause fatalities when they occur.
The Media Amplification Effect
Dog bite statistics are further distorted by media coverage patterns. Research examining news coverage of dog bites found that attacks attributed to pit bulls received significantly more coverage than attacks attributed to other breeds, even when injuries were comparable in severity.
A 2018 study analyzed three years of dog bite news coverage and found that pit bull type attacks received four times as many stories as Lab attacks of similar severity. This coverage asymmetry shapes public perception, makes pit bull attacks more salient in memory, and causes people to dramatically overestimate the frequency of pit bull attacks relative to attacks by other breeds.
When politicians cite the public's fear of certain breeds as justification for BSL, they are often responding to a perception shaped by media coverage rather than actual relative risk. The data on child safety and dog bites makes this particularly clear: the breeds most commonly involved in serious incidents with children are popular family dogs, not the breeds targeted by BSL.
Reading Statistics Critically
When you encounter dog bite statistics, several questions will help you evaluate their reliability. Who collected the data and how were breeds identified? Were DNA tests used, or visual identification by non-experts? What was the population size for each breed studied? Are raw numbers presented without per-capita context? Is the data from a source that has a stake in the outcome, such as a breed-specific advocacy group?
Reliable bite data comes from peer-reviewed veterinary and public health journals, uses methodology that accounts for breed identification limitations, acknowledges the absence of reliable population denominators, and focuses on behavioral and situational risk factors rather than breed-based conclusions.
The next time a politician cites breed-specific bite statistics to justify a ban, you now have the tools to evaluate that claim. Ask for the methodology. Ask where the population data comes from. Ask why the CDC stopped collecting breed bite data. The answers will tell you whether the proposed policy is based on evidence or fear.
Good policy requires good data. The scientific consensus is clear that current breed bite data is not good data — and that the policies built on it are not good policy.