What incident records can and cannot say about a dog

Few claims about this breed travel further than the statistic: bites per year, attacks by breed, the number that ends the argument at a dinner table. The honest position this magazine takes is that the number is the least interesting part of the document. What matters is how the record was compiled, what it was counting, and what it was never designed to say. A household that learns to read an incident record stops being gullible in both directions at once.

How does an incident become a statistic?

Through a pipeline of decisions, each one a place where the data can bend. Somebody reports an event to an authority that keeps records; somebody codes it into categories chosen in advance; somebody aggregates the coded forms into a table; somebody quotes the table, usually without the categories. At every joint the figure drifts from the event. Breed identification in incident reports is the famous weak joint: the person filling the form names a breed by eye, and visual identification of mixed dogs is measurably unreliable. The reader who wants to see how a serious incident database handles these problems can look at a pipeline incident data site, which publishes American pipeline safety records with the methodology, the cause categories and the limits of comparison stated beside every table.

What can the record actually support?

Three legitimate uses survive scrutiny. Orders of magnitude: a record can say whether events are counted in tens or in thousands, even when the exact number is soft. Trend within a consistent method: a database that changes nothing about how it counts can show whether the count is rising, even when the count itself is imperfect. And the distribution of causes as recorded, read with the caveat that the categories were chosen before the events arrived. What the record cannot support is the comparison everyone reaches for first, which breed is dangerous, because the denominator is missing: a count of incidents without a count of dogs is a numerator looking for its fraction.

Why do the same figures support opposite conclusions?

Because a figure without its method is a Rorschach test. The reader who wants to see a dangerous breed reads the numerator; the reader who wants to see a maligned one reads the missing denominator; the honest reader reads the compilation method and concludes less than either. This is not indecision, it is calibration, and it is the same calibration the legal pages apply to the dangerous dog procedure, where the state judges the individual animal precisely because the aggregate cannot do it justly.

What is the household version of this discipline?

Smaller and more useful. When a number arrives attached to this breed, the questions are mechanical: who counted, what were they counting, over what period, in what population, and who published the count with what interest? A figure that survives those questions is worth a place in the file; one that does not is worth a place in the conversation you do not have. The federal agencies that publish transport and incident statistics show what transparent methodology looks like at full scale; the National Transportation Safety Board’s public record at ntsb.gov is the model of an incident database that states its categories beside its findings.

Reading less, believing better

• The incident report is compiled before it is collected: the categories a form offers decide what the statistics can later say, and a breed field filled by guesswork is the dataset’s first weakness.
• Severity and frequency answer different questions, and a table that mixes them is a table that has stopped describing dogs and started describing reporting.
• The denominator is the figure every alarming statistic lacks: a count of incidents means little without the count of dogs, and the honest tables say so.
• A local ordinance debate that cites a national figure is using the wrong scale twice: incidents are local events, and the national dataset was compiled for other questions.
• The reader’s discipline is the compiler’s mirror: ask who coded the record, under which definitions, for which purpose, before quoting the number it produced.

The discipline this page argues for is not skepticism as a posture but method as a habit: the figure is always the last step of a process, and the process is the part worth reading. A household that holds that habit through the adoption paperwork, the veterinary file and the dinner table statistic is the household this magazine writes for. The household that has its paperwork in order faces a quieter question next, the daily rhythm that actually settles the dog, which is the subject of the household rhythm.