Defensive gun usage (DGU) is about 30 times as common as the commonly cited NCVS figure. You’ll often hear that there are 70,000 DGU per year. Gary Kleck challenged that claim as far back as twenty years ago. Now we see that the CDC had major survey data dating back to 1996 from which we can infer that Kleck was too modest. 2.5 million DGU per year is more likely.
How can a study by a reputable organization be that far off? The linked article explains it well enough: it was never intended to be a study of DGU. It lays out three or four reasons but the one I find most memorable goes like this: the majority of DGU (and the best ones) are where the intended victim pulls a gun and the attacker leaves the scene with no violence and no retribution. But because the only crime is too minor to report, it never enters the statistics.
“[DGU] is rare”! Criminals kill people with guns 37 times as often as an innocent DGU goes horribly wrong and the intended victim actually kills the attacker.
Left unremarked is every interaction where no-one dies (and a good number of scenarios where someone does).
(longer, fuller fact-filled gun policy article here)