He uses statistics as a drunken man uses lamp-posts—for support rather than illumination. – Andrew Lang
Does anyone remember number puzzles? In a way they were the predecessors of Rubik’s cube, and I used to love them as a child (though I haven’t played with one in decades). The puzzle is simply a small tray in which there are fifteen numbered tiles which can be slid vertically or horizontally; in the most basic form of the game one slides the tiles around until they’re in numerical order. There are, however, other ways to play; my favorite was to make the sequence go in a clockwise spiral rather than four rows. Another was to line up all the odd numbers in sequence, followed by the evens. Because the tiles couldn’t be removed from the tray without breaking the puzzle, one often had to do considerable back-and-forth shifting to solve it, much as one does with a Rubik’s cube (though obviously the cube is more complex).
Manipulating statistics to serve a particular agenda is something like a number puzzle, in that a skilled but unethical statistician has to do a lot of shifting numbers around to make the image within the frame look like he wants it to look; once that’s achieved it takes more shifting by a debunker to get the numbers back where they started. Of course, it’s an imperfect analogy; in the puzzle one sequence is as valid as another, whereas in real life that isn’t so. Also, the numbers of the puzzle move in only two dimensions while those in a study might move in several. And while cheating by removing the tiles from the tray and re-arranging them will leave obvious signs of tampering and may even break the toy completely, sometimes it’s difficult to tell exactly how a dishonest statistician rigged his numbers without careful analysis by another statistician of equal or greater skill.
Unfortunately, the average person is mathematically illiterate and so is oblivious to even the most egregious manipulation. Tyrannical laws, prohibitionist schemes and pure bigotry are routinely supported by the most outrageous numerical claims, and virtually nobody cries foul: huge fractions of the population are said to be victimized by one crime or another without anyone noticing; large numbers of similar examples are dismissed as unrepresentative anomalies; extremely rare phenomena are touted as indicative of crises or important correlations; physical and economic impossibilities are claimed to be routine occurrences; and statistical outliers are represented as averages or even majorities. And if anyone dares to point out any of it, the inevitable response is a torrent of illogic expressed in meaningless catchphrases such as “If it saves even ONE CHILD!!!!!” which would be false even if they were intellectually coherent.
There was an instructive example of this a few weeks ago. As regular readers know, the Swedish model of prostitution law is marketed by neofeminists and many politicians as a way to “have their cake and eat it, too” by criminalizing only the client side of the transaction and thus repressing women while pretending not to. But even ignoring the incredible injustice of the law, it simply doesn’t do anything it is marketed as doing; it does not decrease prostitution, but it does subject sex workers to police abuse and increases the incidence of STIs, “sex trafficking” and violence. Obviously, this doesn’t sit well with neofeminists, so they simply ignore or deny it; a few weeks ago Samantha Berg published an article claiming that the Swedish model actually reduces violence, and as you might expect it was touted all over the internet by prohibitionists. There’s only one problem with that: the figures she quotes “prove” no such thing, as pointed out by Wendy Lyon:
…there are two really massively important issues here…The 2007-2008 study asked sex workers if they had ever experienced violence, throughout their “entire career in prostitution (which could be anything from one day to 50 years)”. The newer study asked about violence in the past three years alone. These are two very different questions, which can’t possibly give rise to comparable answers…Both surveys recorded sex workers’ experience of violence…wherever it occurred. For the 2012 study, we have a breakdown: 70% of respondents said it only happened in Norway; 12% said Norway and elsewhere; 10% said only elsewhere; 8% didn’t answer…but that is all. We don’t know…which types of violence occurred in which country, or how many of the specific incidents occurred in which country. This makes it impossible to know how much of the reported violence even took place under the Nordic model. And we don’t have any of this data from the 2007-2008 study, so there’s really nothing for us to compare here at all…
Wendy also points out there’s some strange mumbo-jumbo going on in that “rape” and “threatened or forced into sex that was not agreed to” are listed as separate categories even though they’re synonymous, and I noticed that the researchers make the asinine statement that “many of the women would not characterize actual rape as rape” (as though their opinion of what happened to a woman is somehow more important than her own interpretation of it).
Wendy says she didn’t have to dig very deep to discover the problems, but regular reader Kevin Wilson (who is a professional research consultant) looked at it as well and had these things to say in addition to those Wendy listed:
The sample is almost certainly not representative of the sex work industry in Norway at all. Participants were recruited primarily through outreach programs and shelters, so it seems likely that the sample contains workers primarily from the more dangerous segments of the market. Only 17% of the sample is actually from Norway…These two samples are different groups of people, rather than the same group of people tracked over time. This isn’t wrong so much as not ideal for mapping out the intervening effects of criminalization. It’s possible that one sample or the other was more or less exposed to violence due to different sample composition (e.g., if one sample had more indoor workers). In my opinion, this study would have been much better had they either followed the same group of workers from Pre to Post, or asked their current sample about the violence they experienced Pre and Post…There are no margins of error presented, so it’s hard to know which sample-to-sample comparisons are significant/reliable. My speculation is that with such margins included, most of this paper would read as “no effect, no effect, no effect…” due to sample sizes; in many of cases we’re dealing with a lot of small number…the quality of evidence is so low that I’m not confident the findings are accurate or reliable. This isn’t to say that the study is bad in the same way as the Schapiro Group study, merely that the quality isn’t very good and so little can ultimately be concluded from this data.
Kevin also shared one other bit of data that Berg didn’t want her audience to see. If we really could make an apples-to-apples comparison as she pretends, we would see that some kinds of violence (including armed threat, kicking and strangulation) increased while others (including rape, robbery and being thrown out of a car) decreased. But take a look at this table, which shows the change in how much help sex workers have been able to obtain from various sources after experiencing that violence:
|No Support Required||-42%|
Keep in mind that the same caveats apply to this comparison as to the one on violence. But if we were able to make a valid comparison in violence levels, we would also see that there has been a catastrophic drop in whores’ ability to get help for any violence which does occur, and the fraction who had never felt they needed help was cut almost in half. Manipulating statistics is just a game for prohibitionists, a number puzzle intended to advance their agenda by presenting whatever picture they feel would do so most effectively; they are either so detached from reality that they don’t recognize the damage they cause, or so fanatical they just don’t care how many people have to suffer, be brutalized or even die for them to “win” the game.