This is going to be my last look for a while at the decline in rape debate because I’ve researched it quite a bit and frankly, I’m tired of looking at numbers from the various reports. This isn’t to say that I think rape isn’t important but my examining of the rape studies isn’t. I’ve read more reports that Nenena suggested, which she said would show that there has not been a dramatic decrease in the number of rapes in America since the 1970s. Nenena wanted me to look at a report by Mary Koss and to compare it to one by Fisher, Cullen, and Turner. Both examined rapes in college. Nenena suggested that the Koss study from 1982 and first published in 1988 in Ms. I found what I think is the study entitled “The Scope of Rape: Incidence and Prevalence of Sexual Aggression and Victimization in a National Sample of Higher Education.” It was first published in The Journal of Consulting and Clinical Psychology, and I think republished in Ms. I hope that is the same study because I had to pay about $12 to buy it. The Fisher study is called “The Sexual Victimization of College Women,” and it’s free, easier to read, and has better graphics. Thank you, taxpayers!
Before we look at those studies, let’s remind ourselves why we are looking at them instead of “Criminal Victimization 1999: Changes 1998-00 with Trends 1993-99" by Callie M. Rennison of the Bureau of Justice Statistics (BJS), which is what I was mainly looking at last time. As you may recall, the reason we were looking at it is because Nenena said that 1 in 4 women would be raped in their lifetime and supported that statement by writing:
The 1 in 4 number is from the Rennison study which I mistakenly attributed to F/C/T. Full cite is: Rennison, Callie M. "Criminal Victimization 1999: Changes 1998-00 with Trends 1993-99." Bureau of Justice Statistics, U.S. Department of Justice, August 2000. I can't find a copy of the paper online, but several trustworthy sources cite it, including RAINN, the sexual assault wiki, and several "safe campus" and rape crisis centers.
Nenena gave the study a ringing endorsement ... until she found out what it said. When I pointed out that table 7 on page 10 of the report said that in the six years from 1993 to 1999, instances of rape/ sexual assault had declined by 32%, which was more than Nenena said they had dropped, she wrote:
Scott, you should know better. That ain't how statistics work. A drop from 1993 to 1999 is statistically insignificant when you're talking about a rate of change over 35-some years.
She is correct that we can’t really tell from just 6 consecutive years what has happened over the whole 35 years (although, it is statistically significant if I’m remembering my statistics class correctly.) However:
-
I wasn’t suggesting that we could tell from that what happened over the entire 35 years. I was saying that even those 6 years had a bigger drop than any she would admit to.
-
How is it possible that Nenena can believe both that a study that only covered 6 years can not be used to figure out what happens over 35 years, but feels it is able to tell how many times a woman will be raped over her lifetime of about 80 years? How can the study be both insignificant for predicting 35 years but significant for 80? Especially since it doesn’t say that 1 in 4 women will be raped in their lifetimes???
-
The study didn’t end after 6 years. The National Crime Survey (NCS) continued on, and we have the data on rape all the way from 1993 to 2005, which is 12 years.
Twelve years still presents some problems, but let’s note a few things. First, Nenena is suggesting that it is all right to compare the Koss and Fisher studies to figure out if a decline in rape has happened. It looks like Koss’s data was collected in 1982 and Fisher’s was collected in 1996, that’s only 14 years different, which also presents problems for a real understanding of the whole 35. Second, the NCS interviewed about 134,000 people each year for all 12 years for a total of about 1,608,000 respondents, but Koss only interviewed 6,159 and Fisher 4,446 for a total of 10,605. The NCS interviewed 15,163% more respondents than Koss and Fisher. Also because the NCS is done every year, it is easier to see a trend. When comparing Koss and Fisher, you are only seeing two years. You can’t tell if what you are looking at that year is spike, a valley, or something average in a trend that you could see if you look at the every year. For instance, if you looked only at the NCS rape figures for 1976 and 1991, you think that there was no trend up or down in the numbers of rapes in the US because both found 2.2 out of a 1000 people had been raped that year; however, if you look at the numbers around them, you can see that 1976 had an uncharacteristic dip in the number of rapes and 1991 had a spike, but the trend was moving down. The average including the 2 years before and after 1976 was 2.42, but the average including the 2 years before and after 1991 was 1.82, a 25% drop. You, gentle reader, can decide for yourself if you think a study conducted 12 times over 12 years with 1,608,000 respondents is more compelling than 2 studies spaced over 14 years with only 10,605 respondents. I trust your judgment.
If you went with the NCS, Nenena would be disappointed with you. She has a number of issues with it. For instance, the NCS did not include rape information for people under 12 years of age. Nenena wrote concerning that:
Synder 2000 - over 1/3 of rape and sexual assault victims are under the age of 12, hence the omission renders the NCVS results nearly worthless.
NiJ report - 48% of sexual assault victims are under the age of 12, so the NCVS statistic is *really* worthless.
Hmm, studies that exclude people under 12 are useless. Instead, Nenena says you should use studies of college students because then you get everyone of every age! Or rather, you don’t. Unless there were a lot more “very special episodes” of Doogie Houser, MD than I remember, I’m going to suggest that the number of people under 12 reporting their rapes in the Koss and Fisher studies was exactly zero. In fact, Koss’s study specifically asked respondents how many times they’d been victimized since the age of 14, and (let me do some math ... multiply times pi ... carry the 1 ... and ... yes) 14 is older than the 12, so not only did Koss not interview people under 12, she didn’t interview people who were 13 or 14, and probably no one 15 or 16. Maybe some 17 year olds snuck in. Of course, Koss and Fisher also didn’t get many older people either. Very few of their respondents were over 30. So how is it possible that the NCS is “worthless” ... excuse me, I mean “*really* worthless” when it recorded more people of greater ages than Koss and Fisher? Nenena explains:
Comparing similar demographics is a MUCH more accurate measure of change than looking at overall numbers, because of the myriad other factors that can come into play for these types of surveys.
Aaaaah! Now, excluding people makes the study more accurate ... excuse me, I mean “MUCH more accurate.” When she thinks a study disproves her argument, leaving out some groups makes the study “*really* worthless,” but when she thinks a study proves argument, leaving out groups makes it “MUCH more accurate.”
Not only is this new explanation hypocritical, it is also false. While it is true that limiting demographics helps to see changes more clearly in that specific demographic, it is not true that looking at a narrow demographic lets you see more clearly what is happening to the people as a whole. The data suggests that college women are more likely to be raped than just about any other demographic, so they aren’t representative of the whole population. Things that might affect colleges (e.g. changes in drinking patterns, ratio of male to female students – both of which have changed) might affect college students in ways that do not effect the population as a whole. Nenena’s theory would have us believe that we could better tell what was happening generally with bird populations if we looked exclusively at dodos instead of the population of birds as a whole. OMG! Dodos are extinct! And because looking at changes in one demographic is a MUCH more accurate measure of change than looking at overall numbers, we can conclude that all birds are extinct! Bullshit. Things that might affect the rate of death or rape in one distinct element of the whole would not necessarily affect the population as a whole. It is precisely because there are a myriad of factors that affect some groups disproportionately that to understand what’s happening over all, it is better to use larger samples from broader demographics. Specifically in the case of college students, the sample comes slanted in the areas of race and class among other things. I would frankly be astonished to find that Nenena believed that changes in one race or class meant that the same change was going to be felt in the others. For instance, would she ever suggest that a study that found that crime was lowering for rich whites meant that we could conclude that crime was lowering for everyone, including poor blacks? I hope not. Further, can anyone see Nenena saying that a report on a decline in the rapes of men meant that women were being raped less too? The NCS has a practical concern of not calling kids and asking them about rape, but it has a broader sample than either Koss or Fisher. If the NCS is “*really* worthless” because it excludes 1/3 to 48% of rapes, one can only imagine what Koss and Fisher are, which must exclude something like 90%.
Nenena’s flip-flop on whether excluding groups from the study makes the study “*really* worthless” or “MUCH more accurate” is not the only flip-flop she’s done. When Nenena first thought that "Criminal Victimization 1999: Changes 1998-00 with Trends 1993-99" supported her theory (you know, before she read it), she wrote:
I can't find a copy of the paper online, but several trustworthy sources cite it, including RAINN, the sexual assault wiki, and several "safe campus" and rape crisis centers.
After Nenena found out that it and the subsequent editions of the report, which have also been cited by RAINN, did not support her theory, she dismissed them by writing:
Those reports rely on NCVS data, which we've discussed the flaws of already.
My, my, my. They went from sounding so trustworthy to being fatally flawed when the only thing that changed was that Nenena read their findings.
But she didn’t stop there. She said that we couldn’t compare the pre-1993 NCS data with the post-1993 data because there had been some changes in the methodology. And I agree. Sure, the people at the Bureau of Justice Statistics revised the pre-1993 data based on their later and more accurate studies, but still, we can’t tell for sure if the earlier stuff really matches up with the later stuff. Even the BJS warns against comparing the two. However, Nenena was perfectly happy to prove that rape didn’t decline much by comparing Koss with Fisher, two studies with different methods! When comparing studies with different methods disproves Nenena’s theory, the very idea is ridiculous. When doing so proves her theory, it’s the very essence of logic.
And make no mistake about it; Koss and Fisher used different methods. Their questions were very much alike, at least for the data we will be looking at, but there was one glaring difference in their methods that stood out to me immediately. There may have been other significant differences, but the one I saw was that Koss asked about sexual victimization that occurred over a whole year, from September to September, while Fisher asked about victimization that occurred for 6.91 months during the school year. There are 2 things about this difference that are important.
First, there is significant evidence to suggest that college students are less likely to be raped in the summer when they are away from school than they are during the months they are in school. For instance, Fisher said:
It is noteworthy that large concentrations of young women come into contact with young men in a variety of public and private settings at various times on college campuses. Previous research suggests that these women are at greater risk for rape and other forms of sexual assault than women in the general population or in a comparable age group.
What that means is that research suggests that women in the same age group as the college students appear to be less likely to be raped than women who are college students. This suggests that there is something about being at college that increases the risk. Fisher’s own study bears out this suggestion. Fisher found that women who lived off campus were less likely to be raped than women who lived on campus. If being off campus lessens the probability of rape, then a study that included the whole year, including the summer, should have a deflated number of rapes compared to a study the excluded the summer months. Koss included the summer month which would tend to deflate her numbers relative to Fisher who excluded summer months.
The second problem with comparing a 12 month study with a 6.91 month study that extrapolates the data for the whole 12 months is a mathematical problem. Both studies found that students could be raped more than once in a year. Fisher specifically noted groups for which the risk of rape was greater, and one of those groups was people who had been sexually assaulted before. This means that as year progressed, the people who were raped near the end of the year had a significantly greater chance of already having been raped earlier in the year. That in turn means that the number of new people being raped decreases as the year continues even if the number of rapes stays constant, so a study that covers the whole year will have a deflated number of rape victims relative to a study that finds the number of victims for 6.91 months of the year and multiples that number to get a total of victims for the whole year.
To illustrate this effect -- and please don’t take this illustration as me saying that people being raped multiple times is OK with me as this is only being used to show the problem with comparing the 2 studies -- I’m going to do a demonstration but simplify the math by doing everything I can using multiples of ten, and I’m going to assume that no one group is more likely to be raped than any other. The rapes happen purely at random. So we are going to use a school with 10,000 students, where 10% of the student body is raped each month (do not go to this school!), and there are only 10 months in a year (36.5 days a month so every other day the day ends at noon instead of midnight – and I repeat, do not go to this school!) OK, so in the first month of our 10,000 students, 1000 have been raped, but it too earlier in the year for anyone to have been raped more than once yet. That means that for the first month, 10% of the total student body have been raped. The next month there are another 1000 rapes, but this time 10% of the student body have already been raped, so of those 1000 rapes, 10% of them (i.e. 100) were committed against people who have already been raped. That means that only 900 new people have been raped this month, so the total raped on campus is 1900 or 19%. If we look at the first 5 months of our imaginary 10 month year, it looks like this:
1. 1000 -- 10%
2. 1900 – 19%
3. 2710 – 27.1%
4. 3439 – 34.39%
5. 4095 --40.95%
If we use Fisher’s method, we could stop here and guess that if after 5 months 40.95% of students have been raped, then after 10 months, twice that many would be raped, and we’d figure that 81.9% would be raped. But if we use Koss’s method of actually looking at the whole 10 months, we find the following results:
6. 4686 – 46.86%
7. 5217 – 52.17%
8. 5694 – 56.94%
9. 6125 – 61.25%
10. 6513 – 65.13% (not the 81.9% that the Fisher method came up with.)
Because the number of students who have been raped gets bigger each month, the percentage of students raped that month who have been raped before becomes larger, and the number of new victims is less every month even if the number of rapes stays constant. In this example, we can see that the Fisher method came up with a percentage that was 25.8% larger than the actual number of people raped for a whole year, which we would find using Koss’s method.
My example doesn’t use the same percentages as Fisher and Koss nor did it take into account that some groups of students are more at risk for rape than others, which actually increases the percentage of rapes that occur against people who have already been raped and increases the effect on the math, so my demonstration can’t be used to tell exactly how the different methods used by Koss and Fisher affected their numbers, but we can look at their data to see if this mathematical effect was happening in our comparison of Koss and Fisher. If it is in effect, we should find that when Koss looked at rapes over the whole year, she found that a larger percentage of the rapes were committed against women who had already been raped that year than the Fisher study. And sure enough, 41.4% of the rapes in Koss study were committed against women who had already been victimized in its 12 month period, while in Fisher, only 21.5% had over its 6.91 month period. Koss’s instances were nearly twice Fisher’s, and that had to have an effect on the math.
Fisher’s method of calculating rape victims apparently artificially bumps up the number of rape victims relative to Koss’s by excluding summer months from the year and multiplying the number of victims from part of the year to get the total for the year. This would make any real decrease in the number of college rapes appear to be less than it really was when comparing the 2 studies. Keep this in mind as we look at Nenena’s claims, which look like this:
In the 1970's and early 1980's, 1 in 4 college-aged women had been raped.
In 2000, between 1 in 4 and 1 in 5 college-aged women had been raped.
The numbers are either not moving down, or moving down AT MOST by 20%.
And later, like this:
Fishen/Cullen/Turner.
One of many studies that I cited for your benefit, which I notice that you haven't addressed in your post here. The conclusion there was that between 1 in 4 and 1 in 5 female college students will be raped (or be the victim of attempted rape). ...
Why is the demographic qualifier "college-aged women" important? Because that's the demographic that we can directly compare to the Koss survey from which the old 1 in 4 statistic comes.
So comparing similar demographics displaced in time, you'll see that if there is a drop in rape, it's a drop of 20% at most.
The first thing to note about Koss and Fisher is that Koss never says that 1 in 4 college women will be raped in college and Fisher does not conclude that between 1 in 4 and 1 in 5 female college students will be raped. You see, unlike Nenena, who tells us what is in studies that she hasn’t read, I actually read them. (And if they are respected peer reviewed studies, I tend to believe them even if they run contrary to my expectations, which is another difference between Nenena and me.) Koss discovered that about 27% of women in college had been raped between their 14th birthday and the time they took the survey, which is pretty much a 1 in 4 number, but it looks at a lot of time outside of college. For the freshmen who took the survey, those rapes would mainly be from outside of their college experience, so that doesn’t seem to be the 1 in 4 figure we should be looking at. The figure we should be looking at is that Koss found that over a 6 month span, about 3.8% of the female college students will be raped. Multiply that times 8 or so to represent a woman’s total time in school and take into account that a number of rapes will happen to people who have been raped before and you might get a figure approximating 25% or 1 in 4. But Koss never does that math. She doesn’t say how long women tend to stay in college or what percentage of the rapes will occur against women who have already been raped as the years go on or any of those factors that we’d need to know to get the 1 in 4 figure. That ratio was created by people who read Koss’s report but not by Koss herself as far as I can tell. I’m not saying 1 in 4 is wrong, but I think if we are going to look at Koss’s findings, we ought to look at Koss’s actual findings and not a rough estimate expressed in a simplistic ratio that is derived from an unknown equation using Koss’s figures. Why not look at Koss’s actual figure of 3.8% over 6 months?
Fisher does include a ratio in her work, but it isn’t a conclusion of hers. Below is what she wrote (emphasis mine):
At first glance, one might conclude that the risk of rape victimization for college women is not high; “only” about 1 in 36 college women (2.8 percent) experience a completed rape or attempted rape in an academic year. Such a conclusion, however, misses critical, and potentially disquieting, implications. The figures measure victimization for slightly more than half a year (6.91 months). Projecting results beyond this reference period is problematic for a number of reasons, such as assuming that the risk of victimization is the same during summer months and remains stable over a person’s time in college. However, if the 2.8 percent victimization figure is calculated for a 1-year period, the data suggest that nearly 5 percent (4.9 percent) of college women are victimized in any given calendar year. Over the course of a college career—which now lasts an average of 5 years—the percentage of completed or attempted rape victimization among women in higher educational institutions might climb to between one-fifth and one-quarter.
So let’s look at what this paragraph says and does not say.
-
Note that even Fisher says there are problems with projecting her figures over the whole year in part because of that summer month issue that I mentioned above. Fisher gives a warning that is a dire as the one that said that the pre-1993 and post-1993 editions of the NCS shouldn’t be compared. Nenena heeded the NCS warning that strengthened her argument but – surprise!!! -- ignores this one that weakens it.
-
As I mentioned above, the math for figuring how many people will be raped over a year cannot be done in the simplistic manner that Fisher uses here because simply multiplying the numbers based on the percentage of the year the data is from ignores the fact that as the year progresses, more of the rapes will be committed against women who have already been raped. 4.9 is too high a figure. I checked the math and the equation she is using is (12 – 6.91)/6.91 * 2.8 + 2.8 = 4.9 But that equation is entirely too simple to get an accurate figure when we know about the occurrence of rapes committed against people who were already raped. I don’t mind her using her formula in this paragraph because the paragraph is only intended to bring attention to the problem and not to specifically find the number of rape victims for a whole year, but if Nenena and I are going to be looking to see the actual number of rape victims, we can’t rely on that math.
-
Fisher is using the average for how long women go to school in the year 2000, but her data is from 1996 and Koss’s is from 1982. The average time women spend in school has been on the rise for all that time. For instance, the NCES estimates that between 1980 and 2005, the percentage of women getting Bachelor’s degrees rose 8.4%, but the percentage of women getting Master’s degrees rose 9.9% (up 18% over Bachelor) and for Doctoral degrees it rose 19.1% (127% over Bachelor). As more women get advanced degrees, the average amount of time they spend in college lengthens. We don’t know what length of time was used to calculate the 1 in 4 Koss figure, but it almost certainly had to be shorter than Fisher’s from the year 2000. If we compare the Koss and Fisher ratios, we might be seeing something more related to women going to school longer than the frequency of rapes not decreasing. This is why I think using their actual figures works better than using these ratios.
-
“Might climb” is not the way one phrases a conclusion. This section of the report is being used to explain why the report is important, not to make hard conclusions. If Fisher was trying to make people ignore her study rather than heed it as she is here, she might have phrased that last sentence as “Over the course of a college career—which now lasts an average of 5 years—the percentage of completed or attempted rape victimization among women in higher educational institutions might sink to between one-eighth and one-ninth.” Fisher’s expression of “might climb to between one-fifth and one-quarter” is not a conclusion; it’s a worst case scenario based on assumptions that she tells us in this very paragraph are dubious.
But let’s compare them anyway because Nenena wants us to.
OK, Koss said that 3.8% of female college students were raped over a 6 month period and Fisher, using math that she herself finds dubious, says that we might find that 4.9% of college women were raped over a year. If we cut Fisher’s dubious figure in half to get a 6 month figure to compare to Koss, we get 2.45% over 6 months.
Nenena stated repeatedly that the decline was “AT MOST by 20%.” And guessing from the capital letters there, I’d say she thinks the decline was actually significantly less. At other points, she said the decrease was between 0 and 20%. However, when we look at the actual numbers in Koss and Fisher, the difference between Koss’s 3.8% and Fisher’s dubious 2.45% is 35.5%, which is a lot more than the 0 to 20% that Nenena said it was. I will remind you that Fisher’s number is artificially inflated relative to Koss’s because of the summer month exclusion and the mathematical problem with calculating the total from a fraction of the whole. Therefore, the rate actually dropped AT LEAST by 35.5%. In fact, I feel comfortable saying SIGNIFICANTLY OVER 35.5%.
The BJS found a drop of about 52.3% between 1982 and 1996. Comparing the 52.3% drop and the 35.5% drop, we can see that there was less of a drop in the amount of rape victims in the Koss and Fisher comparison than there was in the NCS studies. While Nenena might say that that is evidence that the BJS is wrong in its findings of rapes decline since 1973 or since 1993, I would suggest that her conclusion is overreaching because:
-
It violates Nenena’s own rule of not comparing studies that use different methods. We would have to first ignore that we are comparing Koss and Fisher, which used different methods, and then ignore that we were comparing those dubious results to another survey that used different methods from those two, most notably that the NCS is a survey of rape among all people, not just college students. Despite Nenena’s claim that limiting the survey to just college students makes the survey MUCH more accurate when looking at rapes as a whole, the opposite of that assertion is true.
-
Koss and Fisher measure only 2 points in time and therefore, cannot reliably tell us about the time in-between or after. It is essentially impossible to discover a trend based on a comparison of only 2 data sets. If we could do that, then if you flipped a coin and it came up heads twice, you could reliably figure that the coin would come up heads about 100% of the time. The NCS gives us new data every year. With the multiple studies, we can see trends that are invisible when you only have 2 points on your graph. The NCS had 12 flips of the coin between 1993 and 2005. If it came up heads 12 times, I’d be more inclined to say there was a trend.
-
The 35.5% drop between Koss and Fisher ignores at least 2 major differences in the studies, the summer month exclusion from Fisher’s study and the mathematical problem with using 6.91 months to calculate the rate for the whole year. The difference must be higher than 35.5% based on the conclusions regarding the influence of college attendance on rape and the instances of people being raped more than once that both studies supported overwhelmingly. Just for fun (and yes, I know that playing with the math of statistical analysis should be no one’s idea of fun), I decided to see what would happen if I lowered Fisher’s number by 25% because it seemed well within the margin of error given the 2 factors that were inflating Fisher’s numbers relative to Koss’s. Turns out Koss and Fisher is 52.6%, which is awfully close to the 52.3% of the BJS. That 25% deflation I used is just an estimate and should in no way be considered factual, but it illustrates how close the actual difference in college rapes between 1982 and 1996 might be to the BJS figures for the nation as a whole.
-
No matter what you want to call it, 35.5% or significantly over 35.5% or 52.6%, the difference is significantly greater than the 0 to 20% that Nenena has been proffering as evidence.
And so we can now wonder what Nenena will say next. How will Koss and Fisher go from being trustworthy studies we can depend on to flawed studies that we must dismiss? What new flip-flopping hypocrisy will be trotted out to prove that men will always be monsters and that feminism has been a failure? What study will -- this time for sure! -- prove that the fear mongering of the media is justified and that women are helpless in making the world a better place? How will she next undermine the funding of anti-rape programs by pointing at their supposed failure via studies she hasn’t read?
I can’t wait to see how her arguments transform like Skrulls from trusted friends to hated enemies, her secret invasion of studies that aren’t what they seem. These two things aren’t alike and can’t be compared ... or can they? We don’t know which stance is real and which will later be revealed to be false! Who do we trust: Rennison, Koss, Fisher??? Rennison and Fisher worked in the government, so they must be Skrulls, influencing things from within! But wait, Koss studied rapes in the 70s, and we all know how Bendis loves to play with those 70s characters! Sweet Christmas! She’s the Skrull and no doubt dressed in a gaudy yellow blouse and steel tiara! Move over Bendis! Nenena’s arguments have more intrigue, more tricky twists and reversals than the Marvel Universe, and they are just as much a fiction.
Read and post comments |
Send to a friend
]]>
