Another question is where did the 50% success rate for guns number come from? I think Dean has calculated a much higher number for the success rate for gun uses against bears, more like 98 or 99%.
It is explained in the article. It is another case of misleading by using statistics.
Here is how it was arrived at. The authors of the study searched records for cases of bear attacks where there were firearms. They subjectively chose 269 cases. Fifteen percent of the cases they chose were incidents where firearms were not fired, for several different reasons. They rejected hundreds of cases where firearms were fired.
In the supplemental information for the Efficacy of firearms paper, they say injuries occurred in 151 of the 269 events. They also say of the 444 people involved, 122 were injured. The numbers and math in the paper are difficult to interpret. The 50% figure comes from claiming injuries occurred in 151 incidents of 269. The authors also say if they had included all the data they could, the success rate for firearms would be higher.
If you subtract the 15% of events where a firearm was not fired, the percent successful becomes 89% for long guns and 99% for handguns.
A big problem with the paper is the authors will not release their data. A simple table showing the dates of the incidents, the number of people involved, the species of bear involved, the firearms involved, and whether the firearms were shot or not, would clear up much of the confusion.
If you deliberately pick incidents by looking for incidents which have been recorded as bear attacks, you are going to have a much higher rate of injuries.
It is called selection bias, and the authors of the paper acknowledge it in the paper.
In my and colleagues study of how effective handguns are when fired against bears, we deliberately chose to include every case we could document, with a description of what happened. This was done to prevent selection bias. So far, 170 cases, 98% success rate of stopping the attack. We have not done a percent of injuries.