At the 2011 World Championships, anti-doping researchers were out in full force, subjecting large populations of athletes to blood tests, and two surveys: UQM and SSC. This produced no less than 4 prevalence estimates, each with its own strengths and limitations.
The IAAF collected blood samples from 1808 athletes, 569 of which were endurance athletes. From many parameters in the blood samples, they computed an Abnormal Blood Profile Score (ABPS), and compared the distribution of ABPS values with reference populations, to estimate blood doping prevalence. Among the endurance athletes (569), they estimated 18% blood doping, 15% for males and 22% for females. Note this is limited to estimated blood doping for endurance athletes, and does not include all doping or non-endurance athletes (1239). The method of accounting for altitude was also listed as a limitation, as data on prior exposure to altitude was not collected, and instead was arbitrarily assigned.
In parallel, a group of surveyors administered a questionnaire to 1203 athletes, that included two anonymous surveys: an anonymous survey using the Unrelated Question Model (UQM) and a second using Single Sample Count (SSC). The results of the UQM method has been discussed frequently here.
The UQM asks volunteers to answer either a birthday question, or a doping question, based on a random selection criteria. This approach ensures individual anonymity, while allowing to compute an overall doping prevalence estimate, using probability to factor out the noise of the birthday question. This assumes that all participants respond to the survey as instructed. The paper describing the UQM result produced two prevalence estimates: an initial estimate of 43.6% before any discussion, and a lower-bound estimate of 31.4%, based on empirical observation that the fastest responses were highly-biased towards doping. One weakness is that the UQM method cannot identify the degree nor the form of any non-compliance by its design, and the paper presenting the results identified at least 8 different forms of non-compliance, and various mathematical equations to adjust the results, given the factor of non-compliance -- something unknown. Put another way, the result of 43.6% is based on at least 8 unknown variables of non-compliance.
The UQM paper also produced a table of the results of the SSC responses, promising to discuss it in detail in an upcoming paper. After another lengthy delay, the results of the SSC data have finally been published. The SSC is similar to UQM, composed of birthday questions and a doping question, but with a different design. The volunteers are all asked 4 birthday questions, and one doping question, and to respond with a number between 0 and 5, how many answers are "yes". This design gives enough data to provide a doping prevalence estimate, as well as a survey non-compliance estimate. The results of the SSC survey were an estimated doping of 21.2%, and estimated non-compliance of 31.9%.
What is the "real" prevalence then? With blood samples looking only at blood values, and two survey methods similarly suggesting ~30% survey non-compliance, it appears to be somewhere in the ballpark of 18% - 44%, with some varying degrees of confidence.
SSC:
