We Need Some References, STATS!

Published 20 years, 1 day past

During a recent (somewhat contentious) debate, a friend tossed out the statistic that every nine seconds, a woman is beaten in the United States.  Later on, I did some math, and determined even if we assume that every one of those beatings is suffered by a different woman—that is, no woman is beaten more than once in a given year, which is most certainly not the case, but we’ll take it as a premise anyway—that means just over 3.5 million women are beaten every year.  That’s fairly shocking, if it’s true, since that’s about 2.5% of all females in the country (of all ages; there were approximately 144 million females in the U.S. as of March 2002, according to the Census Bureau document Women and Men in the United States: March 2002).

But is it true; or more appropriately, is it an accurate reflection of what’s really happening?  I started to wonder about this, because I have a tendency to question premises pretty closely.  What’s meant by “beaten”—does it include incidents where a single punch is thrown in anger, and instantly regretted?  Does it refer only to reported incidents, or is it based on both reported and estimates of unreported incidents?  Does medical attention have to be sought?  Does it include beatings of women by women, or is it only concerned with times when a man beats a woman?

So I turned to Google to do a little basic research.  The search “woman beaten every seconds” immediately turned up claims that varied from every nine seconds to every fifteen seconds.  That latter interval would mean that about 2.1 million women are beaten every year, again assuming every incident involves a newly beaten woman, which is quite a drop from 3.5 million.  I also found STATS.org, a site that claims to “check out the facts and figures behind the news.”  They claim, in a list of what they term “commonly accepted fallacious statistics,” that the actual interval is every two minutes twenty seconds, based on a figure of “220,000 serious violent incidents” for calendar year 1999.  Which works out to every two minutes 23.34 seconds—if it’s true.

But is it?  Our friends at STATS aren’t much help, because they provide no direct reference for the figure, so there’s no easy way to check up on their methodology either.  Further obscuring the picture is that they don’t define a “serious violent incident.”  To reiterate some earlier questions, does it refer only to reported incidents, or is it based on both reported and estimates of unreported incidents?  Does medical attention have to be sought in order to count as a “serious” incident?  They aren’t saying, nor do they provide any links to more detailed information.

So I started digging a little more deeply, again through Google.  Eventually I found a document on “Intimate Partner Violence” (and more on that in a moment) at the Department of Justice that reports:

  • The number of female victims of intimate violence declined from 1993 to 1998.  In 1998 women experienced about 900,000 violent offenses at the hands of an intimate, down from 1.1 million in 1993.
  • In both 1993 and 1998, men were victims of about 160,000 violent crimes by an intimate partner.
  • Considered by age category, 1993-98, women ages 16 to 24 experienced the highest per capita rates of intimate violence (19.6 per 1,000 women).
  • About half the intimate partner violence against women, 1993-98, was reported to the police; black women were more likely than other women to report such violence.
  • About 4 of 10 female victims of intimate partner violence lived in households with children under age 12.  Population estimates suggest that 27% of U.S. households were home to children under 12.
  • Half of female victims of intimate partner violence reported a physical injury.  About 4 in 10 of these victims sought professional medical treatment.

So that gives us some actual numbers into which we can sink our calculators.  I’ll take one million as an average for the period 1993-1998, which is a crude but convenient measure.  That works out to a beating every 31.536 seconds; we’ll round down to every 31 seconds.

There are three things to note here.  One, this is “intimate partner violence,” which includes spouses, ex-spouses, boy/girlfriends, and ex-boy/-girlfriends.  It therefore doesn’t count random attacks like violent muggings, rapes by strangers, and so on.  Two, there were a million violent offenses, which does not necessarily mean a million different women, but there’s no way to measure that so we’ll continue to assume that every incident involves a different woman.  Three, it’s stated that about half of such violent incidents are actually reported to the police, which means there’s potential uncertainty in the data set.  The description of methodology restores some confidence; in the period 1993-1998, they interviewed “approximately 293,400 households.”  That’s a pretty good data set.

The report also states:

Women were more likely to be victimized by a nonstranger, which includes a friend, family member, or intimate partner, while men were more likely to be victimized by a stranger.

“More likely” doesn’t give us much of a handle on the proportion of intimate-to-stranger violence, unfortunately.  If intimate partner violence constitutes 55% of all violent crimes against women, that’s a much different story than if it’s 90%.  Furthermore:

Sixty-five percent of all intimate partner violence against women and 68% of intimate partner violence against men involved a simple assault, the least serious form of violence studied.

“Least serious” is a bit of a misnomer, in my opinion, since “simple assault” is later defined as:

Simple assault is an attack without a weapon resulting either in no injury, minor injury (such as bruises, black eyes, cuts, scratches, or swelling) or an undetermined injury requiring less than 2 days of hospitalization.  Simple assaults also include attempted assaults without a weapon. 

So a slap in the face is lumped in with a beating that leaves marks or requires up to two days in the hospital.  We’re also including incidents where a person (male or female) tried to attack a woman without making use of a weapon, but failed.  Or succeeded.  That’s a very, very wide range of incidents and types of violence.

Since the next step up in assault severity is aggravated assault, which includes incidents in which “the victim is seriously injured,” we could decide to count all non-simple assaults as serious violent incidents.  That would mean around 350,000 such incidents in a year, which is obviously higher than STATS’ figure of 220,000 for 1999.  Now, I suppose it’s possible that the figure dropped that much between 1998 and 1999, but I give such an occurence a probability somewhere around that of my being made the first astronaut to Mars.  So we’re still left wondering what “actual figure as estimated by the Justice Department” they’re using, since the actual figures I got from the Justice Department seem to have nothing to do with their figures.  (But at least I pointed you to my source, so you can check up on my assertions; see the Bureau of Justice Statistics’ main page on Intimate Partner Violence to get links to PDFs, Excel spreadsheets, the document I’ve been referencing, and more.)

At the end of all this, we seem to have arrived at an answer between the commonly-repeated figure of “every nine seconds” and STATS’ claim of every two-and-a-third minutes (which in turn leads me to harbor deep skepticism about their other claims, since their domestic violence number seems to be, well, fallacious).  As noted, the numbers I’ve been using all cover intimate partner violence.  I didn’t find similar information on other kinds of violence, although I’m sure it exists somewhere.  Once those incidents were added in, they would lower the average interval between beatings, although they couldn’t lower the interval all the way to nine seconds, or even fifteen.  To do that, there would have to be more stranger-perpetrated violence than intimate partner violence, which the report says isn’t the case.

So what’s my point?  I have three, as it happens.

My first point is that obtaining an accurate picture of the world is a messy, complicated business, and simple unattributed figures don’t help at all.  I’m not trying to say that violence against women isn’t a big deal: it is.  I personally think violence of any kind, no matter who is the victim and whom the attacker, is a big deal, and we should work to lessen such incidents.  I am saying that it may or may not be as bad as we think—and, in fact, the document I used states that intimate partner violence and homicides dropped over the covered period, despite the fact the national population was rising.  That says to me that we should work harder to figure out the causes driving that decrease, and exert more efforts along the same lines.  I’m not so naïve as to think we can ever totally eliminate violence, but we can and should do our best to get as close to that goal as we can.

My second point is that the news media don’t help at all in clarifying this stuff—no great shock there, I suppose, but it’s something that been bothering me more and more of late (as I wrote yesterday).  I previously linked to an article about the gross inaccuracies in reporting about the cost of the proposed missions to the Moon and Mars, and this is another example of how convenient, unexamined “facts” become common conversational currency.  I know I’ve heard the “every nine seconds” figure on the news, or at least a figure very much like it.

The third and perhaps most important point, and the one I found most personally fascinating, is that I was able to do some in-depth fact checking of my own in less than an hour, using nothing but Google and some well-chosen search terms, and obtain a more accurate picture of the world than I’d had before.  I believe that this ability to self-inform is one of the most important and often underappreciated benefits of the Web.  If nothing else, I’m glad I went on this particular search because it reminded me that the Web really is something worth fighting for, and that improving the Web is always an effort worth undertaking.

Comments (23)

  1. I was in a very similar position yesterday, and wrote about it.

    My problem was with the news stories about gasoline prices. Most said that gas prices are at a record high in terms of absolute dollars, but not for inflation.

    But I wondered how the prices related to today’s average income, or how much more the average American has to drive.

    I didn’t take my research as far as you, but I agree with point 3: the web is an invaluable tool for dissecting the easy statistics we are given by mainstream media.

  2. I really liked your article. Especially the last article. It was well written and interesting. It really goes to show the power of numbers and what a little research can get you.

  3. There was a report (and if I could find it, I would… yes I realize the irony) that talked about the amount of violence against women that takes place on Super Bowl Sunday as being overstated to the point where the TV station gave away a free PSA about domestic abuse… when it turns out that, when the numbers were looked at carefully, there was no increase over any other Sunday. So basically they used fudged statistics to get a free ad worth several million dollars.

    The end doesn’t justify the means. I agree that violence against “X” is bad and would like to find ways of reducing it, but it seems there’s no interest in reporting when it has been reduced, perhaps through the (very likely) result that people will think it is no longer a big problem. “Oh, well we’re down to 1,000,000 women abused from 2,000,000…. so we don’t really have to worry about that any more,” strikes me as being a bit absurd, but it also wouldn’t surprise me to hear that someone (and by “someone” I include funding agencies) thought that way.

  4. I wish we could get some good browser statistics. The stats I use at TheCounter.com don’t quite cut the mustard. I wish Google provided a more detailed analysis than the image shown on their Zeitgeist.

  5. I find it interesting how, now that you’ve spent that hour on this subject and written a seemingly comprehensive assessment of the situation, those who follow you would merely have to do a google search and reach this page to obtain that same picture even faster — further tribute to the potential of the Web.

    Unfortunately for those future researchers, though possibly beneficial for you professionally, this site’s page rank for “women beaten every seconds” is rather low.

  6. I think your points regarding information and stats gathering are excellent but you chose a very poor example with which to make them. The topic is too volatile for some of us. Reading this got me almost as upset as I got during the initial conversation! Not because that was ever your intention, Eric, which it wasn’t in the first place I’m certain – not your style, so to speak.

    But I do believe you have little or no subjective experiences regarding the types of domestic violence you are discussing here. You have nothing clouding your ability to look at stats just as stats and draw some general conclusions without a personal bias.

    I’m not criticizing, I’m saying you’re lucky. You grew up to the best of my understanding the child of parents who wanted and loved you, and had a structured environment to help you along in your growth. In turn, you are an amazing husband and father, creating a stable and loving environment along with your wonderful spouse to provide the absolute best for your child, and children to come.

    Other people’s experiences, including my own, are vastly different. As a result, even if we wish to be able to jump over the emotional damage that violence can cause (and I am not limiting this by gender or relationship) it’s not easy or perhaps even possible. To see the stats beneath the issue becomes clouded in bias. Wrong? Possibly, but that’s human frailty for you. Or maybe just my own.

    I’m with Simon Jessey. Let’s focus on browser statistics – I’m far more able to look at that without such a deep bias. When it comes to the statistical numbers of people harmed, I personally can’t see the statistics through the proverbial trees, and in the end, what good does that do?

  7. Simon, the Google stats are much cited as a potentially really accurate source, but I tend to believe that they’re no more reliable than any other source because:

    We don’t know how they’re collected (so, we have no way of knowing if there are errors)

    The type of errors that affect browser stats are systematic rather than random, so just increasing the sample size doesn’t help.

    See, for example this thread at MozillaZine.

  8. Even though I liked reading the article, I couldn’t help but wonder what the/your point was. As it happens, you made three very valid ones, of which I find the last one particularly inspiring. Great effort. Thanks.

  9. Excellent work. Although the topic may be a bit controversial, these type of statements need to be scrutinized. Many times they are unsubstantiated to begin with, but because of the nature of the subject matter, it’s never challenged.

    “Facts” that someone pulls out of their butt are commonly used to further political goals/agendas. It is one of the things that truly irritates me about the media and politicians. And it can give a poor impression of a particular group of people – men in particular, in this case.

    Don’t get me wrong, to my way of thinking, there is no excuse for striking anyone unless it is in defense of your person or your loved ones. Furthermore, one person beaten by their significant other is one too many.

    However, you must take a cold, hard, look at what you are reading and seeing – and determine if the information is being skewed and if so why. That is often the bigger story, frankly.

  10. Molly, but that _precisely_ his point. The more likely there is bias in anything, the more you have to essentially distrust the report.

    Browser stats? Unless you work for MS or AOL, who really cares? What does someone gain by skewing stats? Nothing.

    Domestic or gender abuse? Much different. If you can’t read Eric’s words without anger, then – I’m very sorry to say – anyting YOU have to say I will listen with caution as to YOU and YOUR agenda. And trust mee, everyting being reported in the media today is part of somebody’s agenda.

  11. I’ve only just read the first article about domestic violence (I intend to read the others). You make some excellent points about how deceptive statistics alone can be. Taken from context by people writing reports; skewing the data only slightly to make a point; using one set of figures on a particular subject instead all of the available information – all of these, as well as many others, are used daily by people either intentionally, or foolishly, misleading the public.

    In the case of domestic violence, there are several sources that need to be used to get a full picture. Once is the Uniform Crime Reports (UCR) done by the FBI. This is the first report generally used any time crime statistics are sought. However it has many limitations including: dividing all crime in two categories; relying almost solely on crimes reported to, and recorded by, police (thereby excluding any crime not reported or recorded); limiting the count to the perceived most serious crimes; and having little detailed information about the crimes, with the exception of homicide.

    The Justice Department publishes the National Crime Victimization Survey (NCVS) which looks at crime from a different perspective. The survey is given to large numbers of households (about 60,000 at any given time, representing about 135,000 individuals), it covers crime not reported or acted upon by the police, and gives more detailed information about the crime victims than does the UCR.

    Even though using both of these reports will give a clearer picture than either alone, there are still nuanses of victim identification and explanations of specific crime that are omitted in both reports.

    For domestic violence (as well as other crimes, including crimes against children which are rarely officially reported anywhere) the official numbers are both too conservative and too broad, depending on the aspect they are attempting to address.

    The truth is that crime and violence, no matter who is on the receiving end, impacts the whole of society in some fashion. That friends and family are much more likely to harm other friends and/or family, makes the statistics – whatever they may be – hard to digest. That a large number of women are killed by their loved one every year, is only out-paced by the number of women assaulted again by loved ones. Included in these figures are unknown numbers of children who are exposed to violent assault each year, either as a witness or a victim.

    I applaud your point that we must be cautious about what we believe when it comes to statistics. At the same time, I hope that people understand that the fault is not usually with the numbers themselves, but with the individuals interpreting them.

  12. …every nine seconds, a woman is beaten in the United States.

    And by god, we’ve got to find that woman and help her!

  13. Excellent piece Eric.

    As you say, this ability to be a conduit for self-education is one of the most amazing aspects of the web. That it allows smart, thoughtful people such as yourself to easily reach people all over the planet is another.

    I would comment to Molly that a great value in examining such statistics is that it makes it possible to spot bs more easily and helps us to understand the motives of those who would profit from it. (For instance it becomes hard to justify a no-tolerance attitude on crime where people are locked up and the key is thrown away if it turns out that the crime rate is independent of the incarceration rate.) Also it lets us see that we are becoming a better society, even though we may still have a long way to go.

  14. A well thought out and structured article exploring the perils of lies, damned lies, and statistics.

    It seems, particularly with Molly’s response, that what was being measured by these particular statistics spoiled some readers’ enjoyment of your post. I must admit though I paid no attention to that as I was far too interested in how close you’d get to concretely confirming or denying a particular quoted statistic.

    The problem with statistics is that even when they are accurately quoted they require context and caveats so that we all know what we are talking about. I find that it is generally better to stick to principals and discuss those, utilitarians be damned.

  15. The thing that most amazes me is that on the web, a format expressely designed for linking information, we haven’t managed to make a dent in the amount of supporting and contextual information we provide with pieces (my blog’s probably as guilty as any).

    While resources like Google and online databases are worthwhile, I think it’s unfortunate that even scholarly works often don’t provide easy links to sources & context.

  16. Perhaps this is why development of the Semantic Web is so important. It will surely make information like this easier to gather and process – a logical step forward from what Google already does.

    Frankly, I’m disgusted by the amount of violence in society. And the awful thing about it is that for every physically-battered woman out there, there is probably another who suffers emotional or mental abuse as well.

  17. Raw data can be really friendly. I have tasked myself with figuring out why users aren’t using the website we’ve created, (well, where they’re going/not-going, etc.) and needed raw logs, not the single-figure “you saw X users last month” stats the server people give us.

    I got access to the raw IIS logs, and being familiar with The Webalizer, I munged them into something it could handle and came up with some scary but fascinating numbers apart from what I’m looking for.

    For example, better than 50% of all requests result in a 304 Not Modified. All of the 304s are on the same small set of files. 30% of requests result in 200 OK, and 15% of requests result in 302 Found.

    Getting rid of all the 304s and the 302s would take somebody about an hour’s worth of work apiece and save perhaps 5G of traffic/month, probably more, (since I’m not sure how big a request/response pair is) or 20% of the web server’s traffic.

  18. I don’t quite understand Molly’s objections to it, other than what seems to be an emotional reaction independent of the points container herein. I thought this was well done, and a point well made.

    I also gotta say, though, that I wish there were some better, widely available, comprehensive statistics on browser usage. Perhaps the W3C should start a yearly browser census.. hehe..

  19. Nice article. Your comments echo very much the writing of Joel best in his book Damned Lies and Statistics: Untangling Numbers from the Media, Politicians, and Activists (ISBN 0520219783). These include such concepts as ‘mutant’ statistics – numbers which are misquoted or unattributed which obtain a life of their own outside their original context, and of course the propensity of media and activists to choose the extreme of a statistical range to push their point of view. You should read this book if you’re interested in this sort of thing.

  20. Just playing the devil’s advocate (not normally something I do), but doesn’t the Yin/Yang symbol represent the balance between good and evil? The article seemed to represent getting rid of the evil of violence.

    OK, I’m done with the nit-pick (not normally something I do).

    The breadth of your interests Eric truly amazes me. One question I have is what can we do societally or culturally that will slow the frequency or quantity of these types of violence? Can sharper punishment fix things? We’re facing full jail cells right now in many parts of the US, is it time we export them to Iraq? Should Iraq become the new Australia? How many people once perpetrating such a thing and then punished legally never do it again? This entry brings up more questions than it answers ;)

  21. Trackback ::

    Reagan's Blog

    Stats Crunch
    Eric Meyer has an interesting posting that looks at the statistics we are given by the media and pretty much proves, at least to me that the media just pull numbers from thin air….

  22. interesting article, and being able to skilfully source and check statistics is something that everyone should be able to do and understand the importance of.

    but some statements made me feel a bit uncomfortable, and even though your point was about statistics, you’re dealing with a very touchy and sometimes lethal issue here…

    “I am saying that it may or may not be as bad as we think”

    for instance, what’s your point here, exactly? why point out that a few hundred thousand beatings is technically better than a million, unless you want to imply that this means we shouldn’t be as concerned as we were before?

    “and, in fact, the document I used states that intimate partner violence and homicides dropped over the covered period, despite the fact the national population was rising. That says to me that we should work harder to figure out the causes driving that decrease, and exert more efforts along the same lines.”

    i could be going out on a limb, but maybe one of the causes is those irresponsible people going out there quoting inaccurate figures in a genuine attempt to raise awareness of systemic violence against women.

    and one of the reasons that a slap in the face is counted as violence as well as a punch in the face is that in some cases if you don’t break up with the guy who slapped you, he may punch you next month…

    as i say, an interesting article, but it’s dangerous to feign neutrality…

  23. Did you know that 73% of all statistics are made up on the spot?