False rape accusations are not rare: redux

About a decade ago I posted a refutation of the common claim that false rape accusations are rare – that they make up about 2% of the accusations. At the time, the literature seemed to support a false accusation rate of about 40%. Because of some technical issues, followed by a move to another state with another server, I took my blog down. I brought it back up about a year ago, but have not been posting much – my new job eats up a lot of time. However, after watching the Kavanaugh debacle, I thought I would review the literature again to see if it has changed.

First, a personal story of why this became an interest to me. Some details are changed from the actual stories, both for confidentiality and because memory is faulty after many years. I am a forensic pathologist, but I also have an interest in computer science and image processing. While I was in the military, one of my jobs was to do analysis of imagery with forensic and operational interest from a forensic pathology point of view.

One such case was that of a young male sailor who was charged with raping a young female sailor (who was serving on a different ship) while on liberty at a port in the middle east. The young man had met the young woman in a bar and, young service folk being the way they sometimes are, they ended up having sex. The next evening, the young man went to the woman’s ship to ask her out again. The NCO on deck said he’d check to see if she was available. Then, a few minutes later, the young man was surrounded by Shore Patrol and taken into custody.

The young woman had made the charge of rape. I was asked to do an evaluation of what appeared to be some patterned injuries on her neck, one of which she claimed was a “hickey” that the young man had forced upon her. After a bit of evaluation, I came to the conclusion that this was more likely something machined rather than organic, because of its regular geometry. I thought perhaps the assailant had pressed a piece of jewelry into her neck, leaving a mark – perhaps while throttling her with his hands.

So, I went to the NCIS agents and asked them if they could get the jewelry she had been wearing during the attack. They came back and said it was not available. The young lady claimed that she was so upset at the attack that she had thrown all her jewelry into the sea.

Then I looked at another mark on the victim’s neck. It looked like a series of vertical lines with a textile pattern – something that you can see with a tight knitted turtle neck sweater. So I went to the NCIS and asked them for photos of the clothing that the young lady had been wearing. Alas, the young lady indicated that she had destroyed all of her clothes as well.

So, I wrote my report saying that there were clearly marks on the neck, but they were not classic “hickeys,” and were likely made by machined parts or textiles. I signed off on the case and forgot about it, until I ran into one of the NCIS agents at a meeting. I asked whatever happened with that case, and was told that the young lady had recanted.

It turned out that she had gone out with the young man, had consensual sex, and that was that. Unfortunately for the young man, when he went to the woman’s ship to ask her out again, the NCO he talked to was the woman’s fiancee. As one might expect, the fiancee was concerned when a strange man told him that he dated his future bride. When he confronted the woman, she panicked and claimed rape.

Unfortunately, by the time the young man was exonerated, his career had been ruined. The entire process took about 18 months, during which time he was imprisoned, and had been denied training and duties. In the military, there is a rather strict timetable for getting various training done, getting tickets punched, etc. in order to be promoted. Worse, the military is in part a pyramid system, and if you don’t get promoted within a certain time period, you are released. This man was now so far behind in this process, there was no way he could save his career. So he left the military, which had, until then, been his career choice.

In a second event, another man I knew in the military was going through a difficult divorce. As part of that divorce, his wife claimed that he abused his infant. To prove this, she took the infant to the hospital and documented numerous injuries supposedly inflicted by the father. Again, the assumption was guilty until proven innocent, and he was looking at the destruction of his career. In this case, however, the military had tapped his phone as part of the investigation, and the wife called and bragged to the father about harming the child in order to set him up with a false complaint. He was exonerated, but again left the military not too long afterwards.

While I have been the victim of false accusation, I have never been the victim of serious criminal accusations such as these. These experiences early in my career made me a bit leery about accepting them uncritically. And, having now had about 33 years of experience in forensics, I have found demonstrably false statements in investigations, about all sorts of things, to be fairly common.

So, let’s look at a few things. First, where does this claim that only 2% of rape accusations are false come from? Well, it turns out that the literature on the subject has an extraordinarily wide range of estimates, basically from near zero to almost all. Why? Because in most cases nobody has a clue, since in the majority of cases a court finding is not achieved. In order to come up with a number, then it is necessary to subjectively assess (or subjectively create checklists of criteria that then are “objectively” applied). If you are skeptical, then the number of false accusations is high. If you “believe the woman” then the number of false accusations is low.

Here’s a list of references from a table in Rumney, written in 2006 (Rumney 2006). It’s as good as any:

  • Source %
  • Theilade1986                    1.5-10%
  • NY1974                                2%
  • Hursch1974                        2%
  • Kelly2005                             3-22%
  • Geis1978                               3-31%
  • Smith1989                            3.8%
  • Greenfield1997                   8-15%
  • Clark1977                              10.3%
  • Harris1999                           10.9-25%
  • Lea2003                                 11%
  • HMCPSI2002                        11.8%
  • McCahill1979                       18.2
  • Philadelphia1968                20%
  • Chambers1983                     22.4%
  • Grace1992                             24%
  • Jordan2004                          38-41%
  • Kanin1994                            41%
  • Gregory1996                        45%
  • MacLean                                 47%
  • Stewart1981                         90%

In addition, here’s a few more, taken largely from McGrath2018

  • Lisak2010                      5.9%
  • Lonsway2009              7.1%
  • Brown1997                   13%
  • MacDonald1973          18-25%
  • Kennedy2000               32%

You get the idea. Now, I have to be up front and admit that I’ve only found and read about half of these, but the pattern is clear. It’s all in the assumptions about the cases you just don’t know about. A bunch of these have documents online, though if you want to check them out – the links are in the references section.

A perhaps more important issue is that there is a conflation of cases where the entire situation is fabricated and where an innocent person is accused. In other words, people conflate “false accusations” with “false reports.” A “false report” means that nothing at all happened. A “false accusation” means that an innocent person is accused of a crime. In conversations about accusations, such as the recent Kavanaugh hearing, it seems that those who cite the 2% statistic almost invariably seem to be using the former to imply the latter. As an example of the difference, in the Home Office report Kelly2005, of 2643 cases reviewed, 83 were found to be “false reports,” 216 were found to be “false allegations., ” 386 were found to be of “insufficient evidence.,” and 662 did not proceed to trial because of “evidential difficulties.”

The other issue is that some of these articles look at rape and others at sexual assault. They are not the same, but this seems to be conflated in most discussions as well.

To show you how this works, I’m going to critically evaluate one of the “2%” articles. I’ve chosen this one because it’s the one that is currently referenced in the Wikipedia article that claims that there is a general consensus that the rate of false accusations is 2% (though when I looked today, it had been edited to the more correct 2-10% – Wikipedia is a moving target).

The article referenced is by Lisak, et al. 2010 (Lisak 2010), who looked at sexual assault (not just rape) allegations at a major university. They found that 5.9% of the accusations were proven false, and concluded that a reasonable range would be 2%-10%.

But let’s look at how they got this number. Lisak et al., looked at 136 cases of reported sexual assault accusations at a large university over a 10 year period. The categorized these as follows:

False report: These are demonstrably fabricated cases in which no attack occurred. This was 5.9%, or 8 of 136 cases.

Case did not proceed: These are cases did not result in “referral for prosecution or disciplinary action” because:
1) there was insufficient evidence of an attack, or
2) the victim withdrew from the process
3) no perpetrator was identified
4) because an incident occurred but it was not legally sexual assault

This was 61 cases or 44.9%

Case proceeded: After investigation, the case was referred for prosecution or there was some administrative action (e.g. the alleged perpetrator being barred from a building). This was 48 cases or 35.3%

Insufficient information: There was not enough information to come to any conclusion (e.g. the case file was incomplete or missing). This was 19 cases or 13.9% of cases.

Based on these criteria the conclusion of the authors was that there were 5.9% false reports, with a range of 2-10%.

So, let’s look at this. As someone who has done a few studies and published in peer reviewed literature in the field of forensic medicine rather than sociology, there is one problem that jumps to the front without any other investigation. The authors included no-data cases, which artificially lowers the rate of false reports. If I were to do an autopsy study of, say, whether or not people who are under the influence of methamphetamine at the time of death tend to have enlarged hearts, I would end up with three categories of people: 1) those who have methamphetamine in their blood on toxicologic analysis, 2) those who do not have methamphetamine in their blood, and 3) those who were not tested.

The correct thing to do is to exclude the cases where no test was done, because it is almost certain that some of these cases will, in fact have methamphetamine on board, and some will not. If I were to assume that simply not doing the test means that no drugs were present, it would be wrong and would bias the result.

That’s what happened here. Almost 14% of cases in the Lisak study should have been excluded, and the evaluation done only on the cases where a determination cold be made. The effect of this is to artificially lower the percentage of false reports. Only 117 of the 136 cases should have been used for statistical analysis. The no-data cases are exactly that – no-data. Thus even using the data of Lisak uncritically, the rate of false reports should have been 8/117= 6.8%, not 8/136=5.9%.

I’ll also point out that the range of 2-10% is not an actual statistical estimation of variance.  It’s just a guestimate based on reviewing some other articles in comparison with theirs.

but back to the data.  Using actual data instead of including no-data cases, the classification should have been:

False reports: 6.8%
Case did not proceed: 52%
Case proceeded: 41%

Without even critically looking at these categories, 6.8% doesn’t really seem all that much like 2% to me. But I’m not a social scientist.

But it gets worse. Let’s look at the category of “case proceeded.” Remember that this category simply means that the case was referred for prosecution or an administrative action was taken. It is assumed by the authors that none of these were false reports, much less false accusations. Using this categorization, the Duke Lacrosse case would not be a false report (since it was referred for prosecution). It is not clear whether the UVA case would be classified as “case proceeded” since there was administrative action (suspension of the accused fraternity). Further, of course, “administrative action” does not require due process, or in fact, much of an investigation, as numerous successful lawsuits and settlements by people wrongly accused have demonstrated. Clearly, a nontrivial number of these “case proceeded” cases are false reports, and more are false accusations. But in this study they are not

Let’s assume, though that not only are all of these “case proceeded” cases viable, but let’s go further and assume that they all either resulted in conviction, or would have had they gone to court. Well, there’s been a fair amount of work done on the rate of false convictions in these cases. The authors assume that false convictions do not occur, but they do. In one study of convictions in Virginia by The Urban Institute, funded by the National Institute of Justice in 2012, Roman et. al. (Roman 2012) found that in cases of conviction of sexual assault where DNA evidence was still available, evaluation of DNA exonerated the convicted defendant in approximately 15% of cases.

Now, of course, these exonerations occurred only because DNA evidence was still available, which itself is evidence of some sort of minimal quality of the investigation. In fact, of 634 cases with conviction that met criteria for study, only 230 had DNA. Obviously I can’t prove it, but my personal interactions with investigations over the past 32 years leads me to believe that those cases with less physical evidence are more likely to be false convictions, which would suggest that of the 404 cases without evidence, the false conviction rate would likely be even higher – but that’s speculation.

In a study performed by the Center for Prosecutor Integrity (CPI 2016), a review of the National Registry of Exonerations (NRE) was performed. The NRE tracks confirmed exonerations. They reviewed the database looking for all cases of sexual assault between 1989 to 2015. Of 1,575 total exonerations for all crimes, 269 were for sexual assault. Of these exonerations, 15..6% were for “no crime,” meaning that it turned out that no sexual assault had occurred at all.

There are a number of different estimates of the rate of false conviction, and the Roman article provides a bibilography and review for those interested. But let’s use their number for the Lisak article data as a lower bound.

The increasing plea bargaining rate and decreasing trial rate for all crimes has been an growing problem for years, but let’s assume a conviction/plea rate somewhere between half and all cases, and assume that in half of the non-convictions the defendant is guilty. That seems pretty generous. Let’s use that as an upper bound.

That would mean that in the most stringent view of Lisak’s data, of the 35.3% of cases that ‘proceeded”(using Lisak’s decisioin to include no-data case in the count), all were convictions or pleas, and of those about 15% were false convictions, or 35.3% * 15% = 5.3% of total cases being “false accusations” (with about 1% being “false reports”using the CPI data).

In the least stringent, of the 35.3% that “proceeded” , 17.7% were convictions and 17.6% were findings of not guilty, which would mean (using the same 15% false conviction rate), 2.7% of the total were false convictions. If 50% of those found not guilty were actually not guilty, then that’s another 8.8%, for a total of 11.5% false accusations, with 1.8% being “false reports” .

So, the “case proceeded” category almost certainly includes between 5.3%-11.4% false accusations, and 1%-1.8% false reports. So, now that 5.9% becomes 6.9% – 7.7% for false reports, and 11.2% – 17.4% false accusations. And that’s lowballing it.

Let’s move on to the next group, that of “cases did not proceed.” Remember that these are not just cases where people didn’t think they could get a conviction, but also includes cases where there was insufficient evidence to prove an assault happened – or didn’t happen. Thus it wasn’t proven or disproven, but it might be a “false report” in a real-world sense.

Clearly the rate of false accusations and false reports would be higher in these cases than in those cases where a decision to proceed had occurred. Thus, a low estimate would be to use Lisak’s own estimation that 5.9% are false reports. That would add another 2.6% false reports. If we use the 15% exoneration rate for false accusations, that would give an additional 6.7% false accusation rate.

That would raise the false report rate in Lisak’s data to 9.5%-10.3%, and the false accusation rate to 17.9%-24.1%.

But, really, using the 15% exoneration rate is lowballing it a lot. In cases where there is no evidence that a crime was committed, the probability that no crime was committed or of a false accusation is probably greater than when a conviction is achieved, right? You’d have to be a pretty committed feminist to believe otherwise. But how much? Who knows. Since this is an academic institution, and there’s not a lot of respect for due process, it seems, I personally suspect that it’s a lot greater. What to do? Double it? I’m not going to make up a number, but it means that the number above gets an asterisk.

And, finally, there were 13.9% with “insufficient information.” There’s no real reason to believe that these cases are all that different from the cases in which there was information, so you can assume the same breakdown. So, we can just divvy them up and distribute them as if they were of the same percentages as the cases with data (which is essentially the equivalent of removing them when it comes to percentages). That would give a total of :

11.0%-12% false reports
20.7%-28% fase accusations

with the “asterisk” that it’s probably at least 20% higher for the “case did not proceed” category.

And all you have to do to get this increase is assume that the Duke Lacrosse students were not guilty.

In order to claim this “2% to 10%” false accusation rate, it is necessary to pretend that there are no false convictions, that in all cases where there is insufficient evidence for conviction the accused is truly guilty, and that in all cases where there is no information to judge, the accused is guilty. Sounds like a progressive Democrat wrote this article.

This is particularly apropos with respect to the circus surrounding the Kavanaugh confirmation. The question is not whether or not Ford was assaulted in some way (a “false report”). The question is whether or not the accusation against Kavanaugh is false (a “false accusation”) . The 2% claim, even if it were not false on it’s statistical face, has nothing to do with this kind of issue – it has nothing to do with whether or not an innocent man is accused falsely.

As McGrath, et al, 2018 (McGrath2018) write:

There is no shortage of politicians, victim’s advocates, and news articles claiming that the nationwide false report rate for rape and sexual assault is almost nonexistent, presenting a figure of around 2%. This figure is not only inaccurate, but also it has no basis in reality. Reporting it publicly as a valid frequency rate with any empirical basis is either scientifically negligent or fraudulent.

I have disagreed with some of the authors of that chapter on a few things over the years – but they are spot on with this.


Brown1997: Brown, C., Crowley, S., Peck, R., & Slaughter, L. Patterns of genital injury in female sexual assault victims. American Journal of Obstetrics and Gynecology 1997 176:609–616.

Chambers1983: G. Chambers and A. Millar, Investigating Sexual Assault (Edinburgh 1983), 38–42. Citation copied from Rumney2006. I could not find this to download it.

Clark L, Lewis D. Rape: The price of coercive sexuality. 1977. Toronto, Ontario. Womens’ Press. I did not buy and read this book. Included from Rumney’s list. Haven’t read it myself.

CPI 2016: _, Wronful convictions of sexual assault: An untold story of systematic injustice. White paper 2016. http://www.prosecutorintegrity.org/wp-content/uploads/2016/01/Wrongful-Convictions-of-Sexual-Assault.pdf last accessed 28 SEP 2018

Geis1978: R. Geis et al., ‘‘Police Surgeons and Rape: A Questionnaire Survey’’ (1978) Police Surgeon 7,
cited in Taylor, note 4 above. Geis questioned police surgeons on how many false complaints
they believed they had dealt with in their careers. I was not able to find this article to read either.

Grace1992: S. Grace et al., Rape: From Recording to Conviction (London 1992), 6. http://library.college.police.uk/docs/horpu/rup071.pdf last accessed 1 OCT 2018

Greenfield1997: Greenfield, L. A. An analysis of data on rape and sexual assault: Sexual offenses and offenders 1997 . [NCJ-163392]. US DOJ, Bureau of Justice Statistics

Gregory1996: Gregory J, Lees S. Attrition and Rape in Sexual Assault Cases. British
Journal of Criminology 1996 36:4–5.

Harris1999: Harris J,. Grace S. A Question of Evidence? Investigating and Prosecuting Rape in the
1990s . 1999 Home Office Research Studies http://webarchive.nationalarchives.gov.uk/20110218144322/http://rds.homeoffice.gov.uk/rds/pdfs/hors196.pdf last accessed 2 OCT 2018

HMCP2002: Her Majesty’s Crown Prosecution Service Inspectorate/Her Majesty’s Inspectorate of Constabulary. A report on the joint inspection into the investigation and prosecution of cases involving https://www.justiceinspectorates.gov.uk/hmicfrs/media/joint-inspection-into-the-investigation-and-prosecution-of-cases-involving-allegations-of-rape-20020401.pdf last accessed 2 OCT 2018

Hursch1974 : C.J. Hursch and J. Selkin, Rape Prevention Research Project Mimeographed Annual Report of
the Violence Research Unit, Division of Psychiatric Service, Department of Health and
Hospitals, Denver, 1974, cited in S. Katz and M.A. Mazur, Understanding the Rape Victim: A
Synthesis of Research Findings (New York 1979) ch. 13. This is another citation from Rumney2006 that I could not find to read myself.

Jordan2004: Jordan, J. Beyond belief? Police, rape and women’s credibility. Criminal Justice: International Journal of Policy and Practice 2004 4 (1):29–59

Kanin 1994: Kanin EJ. False rape accusations. Archicves of sexual behavior. 1994 23:81-92 doi:10.1007/BF01541619

Kelly2005: Kelly et al., A gap or a chasm? Attrition in Reported Rape Cases, Home Office Research
Study 293 (London 2005), 46–47. http://webarchive.nationalarchives.gov.uk/20110218141141/http://rds.homeoffice.gov.uk/rds/pdfs05/hors293.pdf last accessed 2 OCT 2018

Kennedy2000: Kennedy, D., Witkowski, M. . False allegations of rape revisited: A replication of the Kanin study. Journal of Security Administration 2000 23:41–46.

Lea2003:Lea, S. J., Lanvers, U.,Shaw, S. Attrition in rape cases. British Journal of Criminology 2003 43:583–599.

Lisak2010: Lisak, D, Gardiner L, Nicksa SC, Cote A. “False allegations of sexual assault:An analysis of ten years of reported cases. Violence against women 2010 16(12):1318-1334.s

Lonsway2009: Lonsway, K., Archambault, J., Lisak, D. False reports: Moving beyond the issue to successfully investigate and prosecute non-stranger sexual assault. 2009 The Voice 3(1) :1–11.

MacLean 1979: MacLean NM, Rape and false accusations of rape. Police Surgeon 1979 Issue 15:29-40 (abstract only reviewed, could not find full paper).

MacDonald1973: MacDonald, J. False accusations of rape. 1973 Medical Aspects of Human Sexuality 7:170–194

McCahill1979: McCahill TW et al., The Aftermath of Rape (Lexington 1979), p. 115. From Rumney. I could not find and read it myself.

McGrath2018: McGrath M, Savino JO, Turvey BE. False allegations of sexual assault. False Allegations: Investigative and forensic issues in fraudulent reports. Chapter 9. Academic Press 2018 191-223

NY1974: ‘‘Remarks of Lawrence H. Cooke, Appellate Division Justice, Before the Association of the
Bar of the City of New York’’, 16 January 1974 (mimeo) p. 6, cited in S. Brownmiller, (this citation taken directly from Rumney2006. I can’t find it.)

Philadelphia1968: ‘‘Police Discretion and the Judgement that a Crime Has Been Committed—Rape in
Philadelphia’’ (1968) 117 University of Pennsylvania Law Review p. 277, 284. https://scholarship.law.upenn.edu/cgi/viewcontent.cgi?referer=https://duckduckgo.com/&httpsredir=1&article=6074&context=penn_law_review last accessed 1 OCT 2018

Roman 2012: Roman J, Walsh K, Lachman P, Yahner J. Post conviction DNA testing and wrongful conviction. 2012 Research report. NIJ contract 2008F-08165 https://www.urban.org/sites/default/files/publication/25506/412589-Post-Conviction-DNA-Testing-and-Wrongful-Conviction.PDF last accessed 28 SEP 2018

Rumney2006: Rumney, PNS. False allegations of rape. The Cambridge Law Journal 2006, 65(1):128 – 158

Smith1989: In L.J.F. Smith, Concerns About Rape, Home Office Research Study 106
(London 1989), 8. This is another citation from Rumney I couldn’t find. I could only find the abstract, which can be found at: https://www.ncjrs.gov/App/Publications/abstract.aspx?ID=119032

Stewart1981 Stewart CH. A Retrospective Survey of Alleged Sexual Assault Cases’. Police
Surgeon 1981 28, 32. (couldn’t find this one to read, either)

Theilade P, Thomsen JL. False Allegations of Rape Police Surgeon 1986 30:17-22. (could not find this one to read).

4 thoughts on “False rape accusations are not rare: redux”

  1. Excellent breakdown.

    I myself have witnessed two false accusations of child molestation levied against Sailors who were getting divorced. In both cases I knew the men involved and was personally convinced that they were innocent, largely because they were told if they allowed the divorce to proceed and did not fight for custody, lesser amounts of child support, lesser amounts of alimony, etc. that the charges would be dropped.

    I cannot imagine a case where a mother would allow the molestation of her children to go unpunished if it did in fact happen.

    1. Well, there are a number of documented cases of mothers prostituting their young children for drugs, money, or as part of a perverse relationship. So, mothers do indeed allow molestation of their kids.

  2. Hi Billo – glad to have discovered yout blog. Thughtful analysis. What impact do you think that “pleading down” of both true and false accusations of rape to lesser offenses influences the analysis. You allude to the plea problem, but I cannot figure a way to assess the likelihood that a person truly accused has pleaded down vs the odds that a person falsely occused – but with priors of various sorts, wrong ancestry, poor and represented by a public defender decides to plead to some lower degree offense to reduce the risks of false conviction for the higher-penalty rape.

    1. Traditionally, pleading increases the false conviction rate tremendously. The fact is that justice is too expensive for most defendants. Last year I testified in a case of a guy who was falsely accused of murdering his wife. He spent every penny he had on his defense in his original trial 8 years ago. It ended in a mistrial, and he spent 8 years in jail awaiting his retrial. The bottom line was that he simply could not afford the expert witnesses, etc. needed for an adequate defense. Eventually, his case was taken up by the Chicago innocence project folk, funded by an anonymous donor. He was acquitted in the second trial, but he is financially ruined.

      I see this as an expert. Most indigent defense services do not pay the going rate for expert consultation. Some of us will give a discount to these services, but many will not — and again, indigent defense doesn’t get good expertise.

      This has become an even bigger problem with the new focus on scientific evidence. A couple of studies have indicated that the Daubert decision has not affected the quality of “science” in courts at all. However, it has increased the expense of litigation to the point that the percentage of cases that make it to court have dropped preciptiously. Today, something like 97% of cases are plead out. See, for instance, https://www.prisonlegalnews.org/news/2013/jan/15/dramatic-increase-in-percentage-of-criminal-cases-being-plea-bargained/

      It’s a little like what Mueller did with Michael Flynn. Flynn would almost certainly prevail in court, but it would cost him everything he owns, and Mueller threatened to go after his son and ruin him as well — and it doesn’t matter whether anybody is guilty or innocent. The calculation is simple. Is a pyrrhic victory that partially saves your name worth financial ruin for yourself and the targeted destruction of your family by a federal government with unlimited resources? The answer is no.

      I remember giving a lecture at the FBI academy in Quantico many years ago. I sat down to dinner with one of my colleagues in law enforcement and we talked about it. He noted that today in the United States, there is no such thing as an “innocent” person. There are only “untargeted” people.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.