Thread by @ClaireBerlinski

Okay, so let's look a bit more closely at this: September 5, 2014 U.S. Department of HHS Centers for Disease Control and Prevention Morbidity and Mortality Weekly, Report Prevalence and Characteristics of Sexual Violence, Stalking, and Intimate Partner Violence Victimization --

Crime victimization surveys like this are suppose to elucidate the figures that are truer to reality that police reported crime.

To understand why the dark figure of crime escapes exact measurement, realize that for a crime to be officially recorded, three things must happen:

1) someone must be aware that a crime has been committed;
2) someone must report that crime; and
3 and the police must accept that a law has been violated.

But each link in the chain is easily broken.

People may be unaware that a crime has been committed because they view it as normal or trivial behavior: in some neighborhoods, it would seem perfectly natural to settle a dispute with a good brawl, while in others, this would be seen as assault.

Other crimes may go unrecognized because the victims are unaware that they have been victimized—either because of the nature of the crime, such as fraud, or because the victims are drunk, mentally ill, or otherwise incapable of understanding what has happened.

Even when victims recognize that a crime has been committed, they may not report it. Think of children, or of immigrants who don’t speak the language well enough to explain what happened to them. Rapes often go unreported because the victims are ashamed.

So-called victimless crimes involving sex and drugs also go unreported, of course, because the criminals have no motivation to inform the police that they are hiring prostitutes or shooting up. Crimes can also go unreported because victims fear reprisals.

Above all, crimes can go unreported because victims feel no confidence in the police and see reporting a crime as pointless.

Even if a crime is reported, it will not necessarily be added to the official statistics. The police may conclude, for example, that there is ---

-- insufficient evidence to believe the report. Moreover, poorly performing police departments have an incentive to stop recording crimes: it makes them look more successful than they are. For these reasons and many more,

criminologists commonly posit that the dark figure of crime is far larger than the official figure—perhaps by as much as an order of magnitude. That's why police reports are supplemented by crime victimization studies. These are very hard o conduct well. So this 9.5.2014 study--

--is a good faith start, but a long way from providing irrefragable evidence.
METHODOLOGY: Surveillance SummariesMMWR/September 5,

Prevalence and Characteristics of Sexual Violence, Stalking, and Intimate Partner Violence Victimization...CDCAbstractProblem/Condition:

"Sexual violence, stalking, and intimate partner violence are public health problems known to have a negative impact on millions of persons in the US each year, not only by way of immediate harm but also through negative long-term health impacts.

"Before implementation of the National Intimate Partner and Sexual Violence Survey (NISVS) in 2010, the most recent detailed national data on the public health burden from these forms of violence were obtained from the National Violence against Women Survey conducted in 95–96.

This report [used] data from 2011. [It] describes the overall prevalence of sexual violence, stalking, and intimate partner violence victimization; racial/ethnic variation in prevalence; how types of perpetrators vary by violence type;

and the age at which victimization typically begins. For intimate partner violence, the report also examines a range of negative impacts experienced as a result of victimization, including the need for services.

ISVS is a national random-digit–dial telephone survey of the non-institutionalized English- and Spanish-speaking U.S. population aged ≥18 years. NISVS gathers data on experiences of sexual violence, stalking, and intimate partner violence among adult women and men ....

by using a dual-frame sampling strategy that includes both landline and cellular telephones. The survey was conducted in 50 states and the District of Columbia; in 2011, the second year of NISVS data collection,12,727 interviews were completed, and 1,428 interviews were

partially completed."

Okay, you can immediately see the biggest methodological problem, can't you? I'lll wait a bit so you can think about it.

No trick questions--just what's the obvious problem? Hint: It's a long, complex, anatomically-detailed questionnaire asking asked over the phone, by a stranger who could be anyone, about your most person life;

That's right: You can't be sure you're getting a truly sample, and it's possible you're getting a sample of the views of people who are cool with sitting down and discussing these things with some stranger on the phone. That probably biases the survey group--

---it may be, for example, that people without strong personal boundaries are both more likely to answer these questions when the government asks, and are more likely to put themselves in situation where other women might say,
"This feels uncomfortable, I don't want to be here."

But maybe not. Nontheless: A question mark. The report examines the four subtypes of intimate partner violence that comprise CDC’s definition of being a victim of intimate partner violence: sexual violence, physical violence, stalking, and psychological aggression.

And though it breaks these down in its tables, its final figure uses all four to come up with the figure representing "the percentage of women who have been victims of intimate partner violence." Another methodological mistake.

A mistake that leads to the claim, "an estimated 43.9% of [American] women [have] experienced sexual violence." Sounds terrifying, and it would be. They "estimated" respondents' race and ethnicity. Made no formal statistical comparisons of demographic subgroups were made.

Statistical inference for prevalence and population estimates were made on the basis of weighted analyses, in which complex sample design features (including stratified sampling, weighting for unequal sample selection probabilities, and nonresponse adjustments) were used --

-- to produce nationally representative estimate. But not necessarily an estimate that's relevant to you. if you're not in the circumstances where rape is unusually common. So, what did they find:

Completed forced penetration was experienced by an estimated 11.5% of women.

An estimated 43.9% of women experienced *sexual violence other than rape" during their lifetimes,

27.3% of women were estimated to have experienced "some form of unwanted sexual contact: during their lifetimes.

32.1% of women were estimated to experienced to have experienced "some type of noncontact unwanted sexual experience" during their lifetimes,

Now note how many event classes they put together, suggesting that they were essentially the same kind of event. All of these things are of course awful. Unacceptable. But should they be added to the numbers of women who have been "victims of sexual violence?"

I'd say they're sure, "Signs you need to leave this guy TODAY, because violence is the next step," but I wouldn't say they're violence, per se. Anyway ... just be careful with these statistics. Most people don't read the full reports.