Sex Work Population Project

Who are We?

We are a loosely knit group of researchers and allies with an interest in the realities of those involved in the adult industry in Canada. We focus on using publicly available archival data in combination with relevant qualitative research to provide a more realistic picture of who does sex work in Canada and how they engage with the industry and society at large.

The papers listed below, representing over 15 years of history, all have associated datasets. Given that no research on the sex industry is definitive, your input is important: Readers are encouraged to download, review, and re-analyze the data associated with each paper as well as the stand alone datasets described below.

Published Peer-Reviewed Research


Research that is available to the public but may have not been scientifically reviewed. All research is peer reviewed by our advisors in the industry.

Other Writing


All published papers and preprints have an associated project. For researchers who would like to explore the ad data further, the following anonymized datasets are available.

  • Kennedy, L. (2024). Ad Views. (cached) public
    Ad pages from a classifieds site that provides statistics on views were captured daily to see how often ads were viewed. Data was collected between July 9 and August 10, 2023. During this period, ads were viewed median 128 times per day per ad (IQR 64-248, mean 195, SD 234, N=62582 ads). Client interest based on demographic data does not seem to conform to literature on price (image; see also Nelson et al. 2019 below). Summing the maximum daily visits for each region and ad category suggests that there were up to 433,962 daily visitors to the site.
  • Kennedy, L. (2024). Anonymized Canadian Classified Ads 2007-2023. request access
    All ad texts used in our studies in tab delimited format with most identifying information removed. Metadata from the 2021-2022 collection which provided consistent metadata for ethnicity is included. A new dataset used in the “What was the effect of the PCEPA on Canadian sex work advertising?” paper is added. This contains the full database, anonymized, that was used for the population comparison. Note that data from the source sites is no longer accessible.
  • Downloaded files: These are the ad pages that were originally downloaded. As this data is very sensitive, containing potentially identifying information, please contact us if you would like to review it.

Open Research Questions

Our interests are evolving. These points represent some unresolved questions at the moment:

  • What is the best way to integrate qualitative and archival research in this area?
  • How can we create the most accurate and complete data possible?
  • How can researchers effectively share archival data without revealing personally identifying information?
  • How can we improve models to better understand the data?
  • Is there a role for computer simulation in this type of research?

Ethics Statement

We understand that privacy and security are critically important for sex workers and allies. We also understand that reproducible results and collaboration are essential for research to progress.

  • We respect the privacy of our collaborators. Contributors will never be asked to reveal their identity unless they so choose.
  • All archival data was publicly available at the time of collection and was collected in accordance with the policies of the archival sources at the time of collection.
  • All data collected is stored in a secure manner.
  • Any datasets shared with researchers are anonymized.


Some relevant background research from external sources.


Canadian academic research:

Canadian NGO reports:


Via e-mail:

With the migration of sex workers to online advertising in Canada, a substantial body of research has emerged on how they communicate with prospective clients. However, given the enormous quantity of archival material available, finding representative ways to identify what sex workers say is a difficult task. Numerical analysis of commonly used phrases allows for the analysis of large numbers of documents potentially identifying themes that may be missed using other techniques. This study considers how Canadian sex workers communicate by examining how the word “no” was used by online advertisers over a 15-year period. Source materials consisted of three collections of online classified advertising containing over 4.2 million ads collected between 2007 and 2022 representing 214456 advertisers. Advertisers and demographic variables were extracted from ad metadata. Common terms surrounding the word “no” were used to identify themes. The word “no” was used by 115127 advertisers. Five major themes were identified: client reassurance (54084 advertisers), communication (47130 advertisers), client race (32612 advertisers), client behavior (23863 advertisers), and service restrictions (8545 advertisers). The probability of there being an association between an advertiser and a major theme was found to vary in response to several variables, including: time period, region, advertiser gender, and advertiser ethnicity. Results are compared with previous work on race and risk messaging in sex work advertising and factors influencing client race restrictions are considered. Over time, the restriction related themes of client behavior, service restrictions, and client race became more prominent. Collectives, multi-regional, cis-female, and Black or Mixed ancestry advertisers were more likely to use restrictions.

How long indoor sex workers stay employed in collectives is a poorly understood aspect of sex worker agency in industrialized democracies. This study provides estimates of turnover, the rate at which workers leave employment, using a subsample of 76 collectives representing 3545 workers over a one-year period. All the collectives provided data on individual workers via external websites. The collectives were identified in a larger random sample of 783 advertisers from a popular Canadian classifieds site used by sex workers, all of whom provided URLs as part of their ad contact information. Monthly between October 2022 and October 2023, individual workers associated with the subsample of advertisers were identified from web pages maintained by these advertisers and scheduling data was collected where available. Worker turnover was estimated based on whether workers were visible one month to the next. Over the year, estimated turnover ranged from 12.0% to 16.0% (mean 14.2% SD 1.1%). Turnover was not affected by month or number of workers in the collectives. Mean 41.1% workers (SD 23.5%, N=51 advertisers) were scheduled on any given day. Workers were visible for a mean 5.5 months (SD 4.5) with those visible for one month being the largest single group. Most sex workers in collectives are likely not permanent full time employees, and the extremely brief work histories of many suggest that failure in the industry may be common for this subpopulation.

Background: Most sex worker population studies measure population at discrete points in time and very few studies have been done in industrialized democracies. The purpose of this study is to consider how time affects the population dynamics of contact sex workers in Canada using publicly available internet advertising data collected over multiple years.

Methods: 3.6 million web pages were collected from advertising sites used by contact sex workers between November, 2014 and December, 2016 inclusive. Contacts were extracted from ads and used to identify advertisers. First names were used to estimate the number of workers represented by an advertiser. Counts of advertisers and names were adjusted for missing data and overcounting. Two approaches for correcting overcounts are compared. Population estimates were generated weekly, monthly and for the two year period. The length of time advertisers were active was also estimated. Estimates are also compared with related research.

Results: Canadian sex workers typically advertised individually or in small collectives (median name count 1, IQR 1-2, average 1.8, SD 4.4). Advertisers were active for a mean of 73.3 days (SD 151.8, median 14, IQR 1-58). Advertisers were at least 83.5% female. Respectively the scaled weekly, monthly, and biannual estimates for female sex workers represented 0.2%, 0.3% and 2% of the 2016 Canadian female 20-49 population. White advertisers were the most predominant ethnic group (53%).

Conclusions: Sex work in Canada is a more pervasive phenomenon than indicated by spot estimates and the length of the data collection period is an important variable. Non-random samples used in qualitative research in Canada likely do not reflect the larger sex worker population represented in advertising. The overall brevity of advertising activity suggests that workers typically exercise agency, reflecting the findings of other Canadian research.