I believe that framing people for possession of child pornography is a widespread practice, and that this accounts for almost all convictions on that charge. I base this on the evidence that is typically used in such cases, all of which comes from computers which may have been compromised; and in fact, trials usually mention evidence that the computers in question were compromised (although it's possible for an attacker to remove all evidence of that fact), and that hasn't been a successful defense. If a person were to actually want child pornography, there are simple technical measures which could create a nearly iron-clad guarantee against being caught; and conversely, similar measures with a similar guarantee protect people from being caught planting evidence. Finally, the societal irrationality surrounding child pornography means that successfully getting someone accused of having it will not only get them jailed, but thoroughly destroy their reputation and shame them as well.
The traditional murder motives apply: revenge, and eliminating rivals. Revenge seems like it would be the most likely motive.
There have been cases in which prosecutions were based solely on the use of a credit card number which the owner claims must have been stolen; those cases most likely involve card numbers which really were stolen, but from a random victim, not to frame someone in particular. However, those cases are publicly visible evidence that framing someone is easy, and at least some malefactors must have noticed.
People convicted would tend to maintain their innocence regardless of whether they were innocent or guity, so that can't be used to determine how many people were really innocent. Computer forensics aren't very useful, because a sophisticated attacker can modify all the evidence. One way to find out would be to pose as a black market buyer wishing to frame a (fictional) person, and ask anyone who offers to perform the service whether they have experience doing that sort of thing. However, that would provide only qualitative data, nothing quantitative.