How Many Sexual Assaults are Reported to Police?

“I mean, the Department of Justice own numbers say that 22 percent of rape victims ever come forward.” - Mark Shields, PBS Newshour, Sept 21, 2018.

“Some studies suggest that only about one-third of rapes and sexual assaults are reported to law enforcement officials.” - John Fritze, USA Today, Sept 21, 2018.

These are just two of the hundreds of news stories and commentaries last week that featured statistics about how many sexual assaults are reported to police, following reports of a sexual assault that was alleged to have taken place in the 1980s but not reported to the police.

Where do these statistics come from, and why are they different?

Both statistics come from the same source - the National Crime Victimization Survey (NCVS), a large national survey administered by the Bureau of Justice Statistics within the US Department of Justice.

Statistics collected by police departments cannot tell us how many crimes are not reported to the police. You have to ask people about crimes that happened to them, and then ask whether they reported those crimes to the police. That is what the NCVS has done, every year since 1973.

The first statistic - 22 percent - comes from the 2016 NCVS. More than 200,000 US residents were interviewed for the survey, but most of them were not crime victims and thus contribute no information about police reporting.

The 2016 NCVS data contained only 167 incidents of rape and sexual assault. The statistic of 22 percent (actually, it’s 22.9 percent in the report) reported to police was calculated from those 167 incidents. Because it is based on such a small sample size, it has a large margin of error: plus-or-minus 10 percentage points.

The other statistic - 1/3 of rapes and sexual assaults are reported to police - is calculated from NCVS data between 2006 and 2010. Combining the data for several years gives a larger sample size from which to calculate a percentage. The report said that 48% of violent crimes, and 35% of rapes and sexual assaults, are reported to police. Even with combining several years of data, the margin of error for the statistic 35% is plus-or-minus 6.6 percentage points.

Why are the numbers different? First, the statistics are for different sets of years: one is for 2016, the other is for the period from 2006 to 2010. Second, both are calculated from small numbers of incidents: the difference in percentage points, 13 (= 35 - 22) is smaller than the margin of error for the difference, which implies that the two estimates are, from a statistical point of view, consistent with each other.

Both statistics have relatively large margins of error, but the margin of error does not include all the sources of uncertainty about the estimates. The margin of error reports only the uncertainty from taking a sample instead of interviewing every person in the US. It does not include uncertainty from how the questions are asked, from persons who refuse to participate in the survey, or from other aspects of how the survey is taken.

The NCVS statistics assume that every sexual assault victim who participates in the survey tells the interviewer about the assault and recounts the details accurately. A recent National Academies of Sciences panel on Estimating the Incidence of Rape and Sexual Assault argues that the NCVS questions fail to capture some incidents that are often considered to be sexual assault, such as incidents that occur when the victim is incapacitated by drugs or alcohol. If the NCVS misses a substantial number of incidents, its estimates of the percentage of incidents reported to the police may be too high (assuming that persons who do not report an incident to the survey also would not report it to the police).

But the question for the news stories is not how many people report rapes and sexual assaults to the police now. The question is how many reported them in the 1980s.

The NCVS cannot answer that question. It did not ask specific questions about rape and sexual assault until 1992. A Bureau of Justice Statistics report explained:

At the inception of the NCVS in the early 1970's, it was deemed inappropriate for a government-sponsored survey to ask respondents directly about rape. Reports of rape and attempted rape were obtained only if the respondent volunteered this information in response to questions about assault and attacks.

What about other surveys in the 1980s? There were some, but most were not nationally representative and did not have the same kind of statistical soundness as the NCVS.

The only nationally representative survey from that era asking about rape and sexual assault was the National Women’s Survey in 1990. It estimated that 16% of rape victims reported the crime to the police. But this estimate is also imprecise - the telephone survey interviewed only 4,000 women in all, of whom 500 said they had experienced sexual assault, and had a low response rate (34%) compared with other telephone surveys of the time.

So we do not know exactly how many sexual assault victims report the crime to the police, now or in the 1980s. But we never know anything exactly - that’s why statisticians report margins of error and look at other sources of uncertainty about the estimates. Even with uncertainty about the individual estimates, all the evidence that we have indicates that the majority of sexual assault victims do not report the crime to the police.

Sexual assault statistics - indeed, all crime statistics - are challenging to measure and interpret. Crime statistics depend on how crime is defined, who provides data (and who refuses to provide data), what questions are asked, and how estimates are calculated. My forthcoming book Measuring Crime: Behind the Statistics tells you where crime statistics come from, and how you can assess their trustworthiness.

Copyright (c) 2018 Sharon L. Lohr