This is a follow up thread to the previous one on my experience doing archival research on online advertising My unfortunate involvement in sex work research.
Based on the numbers I was able to generate, I would guess that if you knew 125 women who were between 20-49 in 2016 you probably knew a sex worker. I think facebook estimated that on average people have 10 close friends so if 5 of those were women 3 degrees of separation would be enough to connect any person in Canada with an SP.
However, this isn't the most intriguing result IMO. In order to avoid ambiguity of workers having multiple contacts being counted too many times I implemented a way to identify groups or clusters of contacts (phone or email) that tended to appear in ads together. To be included in a cluster the phone or email only had to appear in one ad together. The method I used, technical name is DBSCAN, is very good at finding comprehensive clusters using fairly simple criteria. However, this could create clusters that could contain multiple workers some of whom might not know each other. As the time period being examined increases this problem is more likely to occur.
Firstly, is it actually a problem? One way to mitigate this is to try and find out how many people a cluster represents. This was done by identifying and counting names.
Secondly, I don't think anyone has seriously studied these ad hoc networks of sex workers - most likely because they didn't know they exist.
On the whole, even when building these clusters of co-occurring contacts over a two year period the vast majority were quite small - an email and a phone being connected was the most common. However, as usual with this data set, there were really extreme outliers. The most extreme was a group of 700 (estimated) workers many of whom had been part of a collective of workers from BC. This is a breakdown for the period from Nov 1 2014 to Dec 31 2016:
"cluster_size" is the number of contacts per cluster - which roughly correlates with the number of people (anything 4 or less probably represents individual advertisers). The 1024 cluster is the one I was describing above. However it appears that there were others that might be similar - not too many but some.
I was wondering if anyone here has had contact with one of these groups and how does this contrast with working entirely independently or for an established business?
More information: https://populationproject.ca/
Based on the numbers I was able to generate, I would guess that if you knew 125 women who were between 20-49 in 2016 you probably knew a sex worker. I think facebook estimated that on average people have 10 close friends so if 5 of those were women 3 degrees of separation would be enough to connect any person in Canada with an SP.
However, this isn't the most intriguing result IMO. In order to avoid ambiguity of workers having multiple contacts being counted too many times I implemented a way to identify groups or clusters of contacts (phone or email) that tended to appear in ads together. To be included in a cluster the phone or email only had to appear in one ad together. The method I used, technical name is DBSCAN, is very good at finding comprehensive clusters using fairly simple criteria. However, this could create clusters that could contain multiple workers some of whom might not know each other. As the time period being examined increases this problem is more likely to occur.
Firstly, is it actually a problem? One way to mitigate this is to try and find out how many people a cluster represents. This was done by identifying and counting names.
Secondly, I don't think anyone has seriously studied these ad hoc networks of sex workers - most likely because they didn't know they exist.
On the whole, even when building these clusters of co-occurring contacts over a two year period the vast majority were quite small - an email and a phone being connected was the most common. However, as usual with this data set, there were really extreme outliers. The most extreme was a group of 700 (estimated) workers many of whom had been part of a collective of workers from BC. This is a breakdown for the period from Nov 1 2014 to Dec 31 2016:
| cluster_size | number_of_clusters |
|---|---|
| 2 | 8531 |
| 4 | 2606 |
| 8 | 763 |
| 16 | 219 |
| 32 | 65 |
| 64 | 19 |
| 128 | 5 |
| 256 | 1 |
| 1024 | 1 |
I was wondering if anyone here has had contact with one of these groups and how does this contrast with working entirely independently or for an established business?
More information: https://populationproject.ca/
Last edited:
