Thursday, May 25, 2017

About The WomanMeansSomething Violence Database

So, a man by the name of Paul Dirks has created a "Sexual Violence Database" which he claims shows a strong correlation between jurisdictions where trans rights have been provided in law and an increase in sexual predation in female designated spaces.

I downloaded the raw data and took a look through it.  First, the database is fairly small.  At this point, it contains some 255 entries.  It encompasses incidents dating back as far as 1984, and as recently as 2017, in the USA, Canada and United Kingdom.  At a glance, it appears to be a compilation based on whatever he was able to find in the news.  To be fair, most of the data dates from 2004 to present, so I will ignore pre-2004 data since there is only a handful of such incidents.

2004 - 2 incidents
2005 - 2 incidents
2006 - 2 incidents
2007 - 1 incident
2008 - 1 incident
2009 - 3 incidents
2010 - 6 incidents
2011 - 11 incidents
2012 - 20 incidents
2013 - 22 incidents
2014 - 31 incidents
2015 - 52 incidents
2016 - 68 incidents
2017 - 32 incidents

If you look at this, you would think that there is a sudden increase in the number of "male predators invading female spaces".  After all, from 2009 to 2016 represents a 2200% increase in reported incidents.

However, there are a few problems to be considered.  First, Mr. Dirks has gathered his data by trawling through internet archives.  This leaves us with a few problems.  First of all, as good as internet archives are, we have to remember that they are not guaranteed to be complete.  Second, we have to recognize that news media in general seldom reports all such incidents.  For the most part, only the most salacious stories are going to make the headlines.

The apparent increase could simply be a function of perceived public interest in such stories, a little like the way that for a while after a spectacular storm, even minor weather events become part of the local news.  To get a perspective on this, I went to the Statistics Canada Uniform Crime Reporting Survey 2015 summary, which contains the following:
In 2015, there were almost 21,500 police-reported sexual assaults, the majority (98%) of which were classified as level 1 sexual assault. Between 2014 and 2015, the rate of sexual assault level 1 increased 3% to 58 per 100,000 population. The rates of sexual assault level 2 also increased (+13%) with a total of 377 incidents reported in 2015, or a rate of 1 per 100,000 population (about the same level as reported in 2013). In contrast, the rate of the most serious sexual assaults (level 3) declined 11% in 2015 with 104 incidents (12 fewer than in 2014) (Table 5).
So, let's put this in perspective.  Dirks' data shows 52 incidents in 2015.  That is 0.24% of the total reported sex crimes in Canada alone (and Dirks' data encompasses Canada, the USA and the UK).  When we isolate Dirks' statistics to Canada only, there is a total of 7 incidents, or 0.032% of the total sexual assaults reported in the country.

Unfortunately, Statistics Canada doesn't provide records at the level of detail that would allow for easy identification of how many sexual assaults occur in the scenario that Dirks and his allies are so worried about.  In terms of frequency, relative to the number of sexual assaults reported each year in Canada, 0.032% seems low - very low - hardly evidence of a shocking increase of any kind.

But, there's a more unsettling aspect to Dirks' data.  If his claim bore any truth value, we would expect to see a marked increase in the number of "men" presenting as "female" entering women's washrooms/change facilities etc. with malicious intent.  Across all countries, and years, there is 20 records in Dirks' database which match that criterion.  That's 7.8% of the records, with between 2 and 5 entries recorded per year between the years 2008 and 2015.  Of these, Canada is only represented in 2 entries, the UK in 2, and the remaining 16 taking place in the USA.  Canada is a country of some 35,000,000 people.  We are talking about 2 incidents - in different years - in the entire country that meet this somewhat bizarre criteria.

So, does this bear even the slightest resemblance to an epidemic of "predators taking advantage"?  Not in any meaningful sense.  It's not even a significant percentage of the raw data in Dirks' sample, much less when held up against the scale of sexual assaults reported in Canada.

There's another dimension to this that I think needs to be discussed further.  Dirks' data set presupposes that either implicitly, or explicitly, the perpetrator of any incident is using the idea that granting Transgender people equality rights under the law as a framework to support their malfeasant behaviour.  Since few of the cases in the database even involve someone who made the claim that they were transgender prior to committing the offence, we really cannot justify that assumption.  It's a bit like noting that there are a lot of cars speeding on a road going past an area where there have been a lot of break-and-enter crimes, and concluding that speeding drivers must have something to do with the break-and-enter crimes.  Statistically, one might even find a significant correlation using a simple correlation test, but that doesn't mean that the two measures are in fact causally connected to each other.  In fact, I'd wager that if you stopped and asked all of the speeders on the road, most would tell you they were in a hurry to get home/to work/whatever - nothing at all to do with the break-and-enter crimes.

Further, we need to question the criteria that Dirks is using for selecting candidate articles.  He seems to have taken the rather broad approach of simply assuming that any crime where a male person has entered a female space is implicitly associated with the granting of equality rights to transgender people in a particular jurisdiction.  No attempt is made to verify that there is any actual connection, he's just hoping that there will be enough incidents to make it look as if there is.  Gathering the kind of data that Dirks wants is actually very difficult work.  Simply trawling through news archives is going to give a very incomplete picture, one which is biased by the coverage themes of the time, as well as the availability of old archives.  Additionally, not all of the incidents that Dirks includes even constitute criminal offences.  There are several which were obvious "baiting" incidents (such as a man attempting to sue a women's-only gym for discrimination), and other incidents which are of similarly dubious motivations.

For Dirks to have any kind of reasonable conclusions, he needs a much more comprehensive dataset, and he needs it to be based on a more meaningful basis than a convenience sample of what he could dredge out of online news archives.  At this point, any inferences he draws from this dataset should be looked upon with considerable skepticism.  Lastly, I would point to the fact that in the entire dataset that Dirks has compiled, there is exactly 2 cases where someone who claimed to be transgender actually assaulted someone.  That person is now serving an indefinite sentence as a dangerous offender.

Remember, Bill C-16 doesn't legalize being a sexual predator.  That is still a crime.