and he picked up the 98% from "everybody knows".
No, he claims that this was the result of a telephone survey (since replicated), that he conducted with a handful of undergraduates for interviewers. Much was made of this on a couple of anti-gun blogs (thence in the media) in February of this year, amid claims that such a survey would entail a huge paper trail, including grant money, phone banks, etc. In point of fact, Lott always claimed that the survey was a shoestring effort, and involved only a couple of thousand subjects. Subsequently, a subject of that survey has come forward to offer corroboration. It would be better if this were more clear, but the evidence to date, IMHO, supports Lott, not his critics.
I first heard it three years ago, when it was being attributed to Kleck.
Kleck is the source of the statistic that 98% of defensive gun uses result in no injury to anyone -- perp or defender. The key difference is that Kleck's results include 'warning shots' and missed shots, as well as brandishing a gun. This is not necessarily inconsistent with Lott's results, since the basic design of the two studies has not been compared in the public literature (as far as I know): ie, it's possible that Lott defined defensive gun uses more liberally than Kleck did -- in either case, the results are not so far apart that study design can't account for the discrepancy.
Lott's work is particularly important, since his results have stood up to the most determined scrutiny. We can't allow a campaign of ad-hominem attacks to discredit the work itself, particularly when that campaign has substantially overstated the serious, and unproven charge of academic dishonesty, and made much of the proven, if embarassing charge, of making anonymous postings on a web site. (Yes, it's an example of poor judgement, but casts no light on his research.)