FBI statistics
FBI reports consistently put the number of "unfounded" rape accusations around 8%. The average rate of unfounded reports for Index crimes is 2%.[2] However, "unfounded" is not synonymous with false allegation.[3] Bruce Gross of the Forensic Examiner's says that:
This statistic is almost meaningless, as many of the jurisdictions from which the FBI collects data on crime use different definitions of, or criteria for, "unfounded." That is, a report of rape might be classified as unfounded (rather than as forcible rape) if the alleged victim did not try to fight off the suspect, if the alleged perpetrator did not use physical force or a weapon of some sort, if the alleged victim did not sustain any physical injuries, or if the alleged victim and the accused had a prior sexual relationship. Similarly, a report might be deemed unfounded if there is no physical evidence or too many inconsistencies between the accuser's statement and what evidence does exist. As such, although some unfounded cases of rape may be false or fabricated, not all unfounded cases are false.[4]
[edit]British Home Office study (2005)
A 2005 study, "A gap or a chasm? Attrition in reported rape cases" was the largest and most rigorous study to date commissioned by the British Home Office on UK rape crime, from the initial reporting of a rape through to legal prosecutions. The study was based on 2,643 sexual assault cases (Kelly, Lovett, and Regan, 2005). Of these, 8% were classified by the police department as false reports.[5]
The researchers noted that some of these classifications were based simply on the personal judgments of the police investigators and were made in violation of official criteria for establishing a false allegation. Closer analysis of this category applying the Home Office counting rules for establishing a false allegation and excluding cases where the application of the cases where confirmation of the designation was uncertain reduced the percentage of false reports to 3%. The researchers concluded that "one cannot take all police designations at face value" and that "[t]here is an over-estimation of the scale of false allegations by both police officers and prosecutors." Moreover, they added:
The interviews with police officers and complainants’ responses show that despite the focus on victim care, a culture of suspicion remains within the police, even amongst some of those who are specialists in rape investigations. There is also a tendency to conflate false allegations with retractions and withdrawals, as if in all such cases no sexual assault occurred. This reproduces an investigative culture in which elements that might permit a designation of a false complaint are emphasised (later sections reveal how this also feeds into withdrawals and designation of ‘insufficient evidence’), at the expense of a careful investigation, in which the evidence collected is evaluated.[6][7][8]

Survey data are less biased than self-reports, but they’re hardly airtight. A study published in 2009 found that changes in survey question wording led to a tenfold difference in reports of past-year rape victimization (as opposed to lifetime victimization), from about two per 1,000 to about 20 per 1,000 among U.S. college women. There are about 7 million full-time female college students in this country, so that’s the difference between 14,000 rapes and 140,000. (Police data, by contrast, showed 459 forcible rape cases at U.S. universities in 2009, a number I got by downloading the FBI’s Uniform Crime Report data and summing across all universities.)

On surveys and via police, hospitals, and other sources, female victims report more frequently than male victims; the rich report more frequently than the poor. Better educated people are more likely to report, as are victims who suffer more serious physical injury from their attacks. Increasing reports of rape may represent increased rape—or they could simply represent a change in reporting dynamics. Maybe there’s a new female police officer. Maybe this year’s survey enumerators are better trained. Frustratingly, it’s almost never a good idea to make strong inferences from data on patterns of rape, whether those patterns are temporal, demographic, or geographic.

Why does this matter? Obviously, getting it right is key to prevention and accountability efforts. Ultimately, the goal is less rape, not just more awareness, and the way we approach evidence affects the policies we put in place. On a more fundamental level, though, I am convinced that getting it right is a respect issue. Looking at rape data and asking the wrong questions (or overgeneralizing, or ignoring complexity) is akin to looking at a rape survivor and doing the same. I want to be sure we do better for the women and men who have already suffered enough.