Invisible Offenders: A Study Estimating Online Sex CustomersResearch ReportThis study and executive summary were developed from funding from Thorn: Digital Defendersof Children, ASU Office of Sex Trafficking Intervention Research and supported by the PhoenixPolice Department.Research Team from the Office of Sex Trafficking Intervention Research (STIR)Dominique Roe-Sepowitz, MSW, Ph.D., STIR DirectorKristine Hickle, MSW, Doctoral Candidate, STIR Associate Director of Research DevelopmentJames Gallagher, M.Admin, STIR Associate Director of Research Innovation and Lieutenant, ViceEnforcement Unit, Phoenix Police DepartmentJessica Smith, MA, STIR Project CoordinatorEric Hedberg, Ph.D., Arizona State University, Faculty Associate ASU STIR and Thorn: Digital Defenders of Children 20131

Background InformationProstitution, the exchange of sex for money, drugs or other things has received increasedattention in recent years from the media, advocacy groups, law enforcement and legislators witha specific focus on the victimization of minors in sex trafficking situations. Sex trafficking, sextrading and sex exchange are all used to describe the exchange of sex between a buyer and aprovider or prostituted person. If the person is under the age of 18 years old or is prostituted byanother person using force, fraud or coercion, it is defined as sex trafficking by the Victim ofTrafficking and Violence Protection Act (2000).Sex buyers of prostituted and sex trafficked persons are poorly understood as there arenumerous challenges to detecting and studying them. The foci of this study, online sex adcustomers, are hidden offenders who are rarely exposed to the public except by episodic targetedenforcement by police (Sanders, 2008). Online sex ad customers experience a lower risk ofbeing caught by police than street-level prostitution customers due to the insulation provided bythe relative anonymity inherent in internet-based solicitation. Other risk-mitigating factorsinclude the fact that online sex customers remain out of (physical) sight of law enforcementwhile soliciting for sex, the arrangements are made by phone or email and the sex exchange isdone in private in a hotel, brothel or private home, while street-level prostitution customers makesex exchange deals and many times engage in sex acts in public spaces where they are morelikely to receive law enforcement attention. The overall number of sex buyers arrested in theUnited States is unknown as current national crime data collection does not separate personscharged with solicitation, prostitution or pandering.An active anti-trafficking movement has grown in the United States in the past decadeand has increased the public’s awareness of the role of demand for prostitution and of the buyerof sex in sex trafficking. Demand reduction groups have begun to gain traction drawingattention to the behavior of customers and suggesting and developing prevention techniques.Some states and cities have begun to implement increasingly stringent penalties for customers,with a specific focus on those buying sex from minors. Other actions geared towards identifyingsex buyers and changing their behaviors have burgeoned, including techniques using the publicshaming of offenders through posting their personal information on websites and billboards, butlittle is known about the deterrent effect of these interventions. Sex buyers, except those caughtin targeted stings, are elusive, complex to research and, in most cases, hidden in plain sight inour communities. Because of these challenges, they continue to be enigmatic. For this study wefocus only on customers of online sex ads.About this StudyThe authors were provided with funding from Thorn: Digital Defenders of Children andthe ASU Office of Sex Trafficking Intervention Research to develop a comprehensive researchagenda related to exploring the demand for prostitution in the United States. This study is thefirst product of this effort and is an attempt to better understand the scope of demand for sexfrom online sex advertisements. This study was done in consultation and partnership with lawenforcement personnel, who assisted in the development of the methodology and goals of thisstudy to specifically address the gaps in operational knowledge as to the scope of demand, theincreasingly technological nature of solicitation, and the impact of demand on the overall sextrafficking landscape.2

Our definition of demand for this study focused exclusively on phone-based (calls,voicemails, and texts) responses to decoy online sex advertisements of a female offering sexualservices. We did not make any contact with respondents and the study was approved by theArizona State University Institution Review Board to use deception to collect phone numbersfrom the subjects.The issue of sex buying is multifaceted and complex, including different types of buyersvarying from the hobbyist, men who regularly buy sex from different sellers (Milrod & Monto,2012), occasional users, and first time users. This study is focused on the response by potentialcustomers to sex ads placed online in 15 cities. For the purpose of developing a probabilityestimate, we used an innovative methodology that has been successfully used in ecology,population biology and demographic research to determine the estimated or probable size of apopulation that is complex to assess called capture/recapture.Previous Research Used to Build This StudyResearch on prostitution demand, conducted through john schools and online discussionboards, has well documented why men buy sex from girls and women (Milrod et al, 2012;Monto, 2004; 2010; Monto and Hotaling, 2001; Shively, Kliorys, Wheeler, & Hunt, 2012) but isnot sufficient to estimate the population or the extent of demand. Previous attempts at estimatingthe population of sex customers have been made through social surveys including the GeneralSocial Survey (as cited in Monto, 2010; Smith, Marsden, & Hunt, 1972-2010) with an estimatethat 14% of men surveyed had previously bought sex and the National Health and Social LifeSurvey (as cited in Monto, 2010; Michael, Gagnon, Laumann & Kolata, 1994), which found that16% of men had visited a prostitute in their lifetime. Unsupported media reports have estimatedthat between 16 and 80 percent of men pay for sex (Bennetts, 2011).A recently released study by the Chicago Alliance Against Sexual Exploitation (CAASE)analyzed the attitudes and beliefs of online sex customers towards the sex sellers they frequentand the efforts of law enforcement to thwart them (Janson, Durchslag, Mann, Marro, & Matvey,2013). The CAASE study also found that online buyers used websites to locate and contact sexsellers throughout Illinois, and travel to different areas to purchase it, suggesting that anyresponse to demand must be coordinated to be effective.Two research groups have previously attempted to estimate the population of sex buyersin a specific geographical area. The Shapiro Group attempted to estimate how many men werebuying sex from ‘young’ girls from online venues in Georgia. Their study, although valiant intheir goal, was rife with methodological issues and assumptions and did not result in a robust andreplicable model (Pinto, 2011). Brewer, Roberts, Muth and Potterat (2008) were the first toattempt the complex task of creating an estimate of the size of the population of streetprostitution customers in the United States based on data from six cities using a version ofcapture/recapture sampling called list matching. Their study explored arrest records andmatched them with county health clinics, jails and outreach interviews. Brewer et. al (2008)created a model estimating the size of the population of male street prostitution customers byusing the population base of each city and found their estimates to range from .3 percent to 1.4percent during 1988-2002. This study builds upon the methods used by Brewer et. al (2008) butuses data collected from just one source, which more closely adheres to the assumptions of thecapture/recapture model.3

The goal of this study was to answer the following research question:What is the probable population estimate for males seeking to purchase sex from afemale using a prominent online sex ad venue?Research MethodsThis study used a capture/recapture sampling technique, which has been used in ecologyand population biology, as well as demography research. Previous capture/recapture studieshave been used to estimate the density of a population of animals such as jaguars in Brazil(Soisalo & Cavalcanti, 2006) and tigers in India (Karanth et. al, 2002), as well as estimates ofpopulations such as drug users in London (Hickman et al, 2002), heroin users in Australia(Larson, Stevens, & Wardlaw, 1994), and type 2 diabetes in the United Kingdom (Ismail,Beeching, Gill & Bellis, 1999). This sampling technique creates a model to estimate the size of apopulation by matching individuals from two random samples. Using this method, this studydeveloped an estimate of the total size of the online sex ad customer population from the exposedpopulation, all males in the metro area over the age of 18 (American Fact Finder, 2011) thatappear in two samples taken from the same population (people who responded by phone to anonline sex ad placed on two different days).An example of capture/recapture sampling is the counting of how many deer are in awooded area at a given time. A spotter is sent to the woods to photograph as many deer as theycan during a 10-hour period. The photographs would be examined to identify each deer by someunique feature and a list would be created. A short time later, a spotter would return to thewooded area and photograph as many deer as they could in a 10-hour period. Those photographsare compared with the photos from the first data collection and the overlap is counted as capturerecapture cases, which create the formula to determine the size of the deer population. In boththis study of online sex ad customers and the example of estimating the deer population in awooded area, the overall population is not closed and to minimize issues of attrition and newrecruits, we attempted to keep the time between collecting samples as short as possible.Recapture for this study is a caller making contact (call, voicemail or text) for both of the adsplaced in that same city.The team placed decoy ads online in 15 cities at two different times (seven days apart) tocollect information about calls and texts from potential online sex buyers. The cities, all ofwhose police departments were made aware of the study included: Atlantic City, Boston,Baltimore, Chicago, Houston, Kansas City, Las Vegas, Miami, Minneapolis, New York City,Phoenix, Portland, San Diego, San Francisco and Salt Lake City.Ads were placed on both (casual encounters) and (adultentertainment, escorts). The identical ad was placed at 2pm (local time) on Fridays twice, oneweek apart and calls/texts were recorded, including phone numbers and messages.4

Study assumptions necessary to consider when interpreting the findings from this study are: All men over the age of 18 that live in each city are potential customers for online sexads. We realize that this is a significant assumption and excludes tourists and shortterm residents but using the information from American Fact Finders was necessary todevelop the percentage estimate for each city. The ad placed on was normative to all of the other ads posted (adultentertainment, escort section) on regarding ad language andphotograph, specifically for individuals advertising sexual services by females andtargeted at male customers. The ad was designed with reference to three ads that hadbeen posted previously by law enforcement for customer stings. Contextualdevelopment assistance was provided by the Lieutenant of a Vice Enforcement Unit inthe 5th largest city in the United States (Phoenix, Arizona) and by a researcher withextensive experience analyzing online sex ads. The callers (customers) probably called other sex ads posted on duringthe 24 hours after our ads were posted. The ads were placed on two average Fridays in late spring 2013.FindingsTable 1: Population and Percentage Estimates of Online Sex Ad Customers.City% of Males in City Who CallSex Ads (ConfidenceIntervals)Estimated Sex AdCustomer PopulationAverage # of Ads Posted in a 24 hourPeriod (Friday 2pm)Atlantic City1.4%(0.5% - 3.2%)10,275206Baltimore1.8%(1.% -2.1%)17,766211.5Boston7.6%(4.8% -10.3%)130,416247Chicago2.4%(1.4% -3.1%)83,478518.5Houston21.4%(13.8% -29%)169,920472Kansas City14.5%(9.1% -17.9%)106,62498Las Vegas13.5%(9.1% -19.9%)99,910515Miami6.6%(4.2% 167New York City3.9%(0%-7.6%)21,514341.55

Phoenix4.9%(3.4%- 6.4%)78,412307.5Portland3.7%(2.6% -4.8%)31,282145.5San Diego3.1%(0% -7%)36,890310San Francisco.6%(.1% -1.3%)9,50496Salt Lake City2.6%(.6%-4.7%)10,67587.5The ads in all cities were taken down almost immediately after beingposted. In response to the two ads placed on, we received a total of 677 contacts(texts and calls) from 415 unique phone numbers of online sex ad customers in the 15 citiesduring the seven days after the ads were placed. The majority (69.6%) of contacts were madeduring the first 24 hours after the ad was posted ranging from 48% in San Francisco to 90% inNew York City. The contacts were from 105 area codes and ranged from one to nine calls/texts(M 1.5) per caller. Recaptured phone numbers (i.e. online sex ad customers who called bothweeks in response to ads placed online in the same city) were found in six cities. Both Baltimoreand Chicago had one recaptured number, two were found in Salt Lake City and Atlantic City,and three were found in Portland and Phoenix.The estimated population size of male online sex ad customers was calculated for eachcity and then anchored to the metropolitan area population to determine the percentage estimate.The estimated population size ranged from 9,504 in San Francisco to 169, 920 in Houston. Thenumber of ads placed during the same 24 hour periods as our ads were placed ranged by cityfrom 87.5 other ads in Salt Lake City to 518.5 other ads in Chicago.When applying the estimated population to create the percentage estimate we found onaverage, for all 15 cities, one out over every 20 males over the age of 18 was estimated to besoliciting online sex ads. In acknowledgement that some of the sex ad customers may not befrom the city where the ad was placed, we conducted analytics of the area codes and differencesbetween texts and voice calls. Area codes of the callers being from the metro area of theidentified city was found to range between 54.7% (Portland) to 88.6% (Kansas City) except forAtlantic City which was an outlier at 17.9%. We acknowledge that some of these calls may befrom hotel phones or phone applications that create false numbers.AnalyticsThe exposed populations for this study are all males over the age of 18 years old in themetro-area of each city. These were determined using census estimates from American FactFinder, specifically the one year estimate from the American Community Survey from the mostrecent year available, 2011. This is the basis by which we calculated the percentage estimate ofthe male population in each city soliciting online sex ads. This exposed population is thedenominator and calculated with the number of unique calls/texts for each ad (captured for 24hours twice divided by number of recaptured phone numbers during the second 24 hours) and the6

average number of ads posted on (escort section) in each city (counted for 24hours twice). This model reports the number of active online sex ad customers on the first datacollection date (mid-June 2013) and considers issues of attrition (customers no longer buying sexfrom online ad source) and new customers (who are entering the market to buy sex online in thatcity for the first time) (Chapman, 1951).Technique we used to find the total population estimate of online sex ad responders (N):N (n1 1) (n2 1) /Average # of adsm 1n1 number of phone numbers from the first ad responsesn2 number of phone numbers from the second ad responsesAverage # of ads average number of ads posted on escorts section during the 24 hours after the adwas placed X2m number of phone numbers recapturedDiscussionMany assumptions were made to reach these results but we believe the estimates arereasonable given previous estimates and our adherence to the statistical model ofcapture/recapture. We attempted in every decision possible to make the most conservativechoice. This study is the first to estimate online sex buyers using this methodology. The modelwe used has strengths and weaknesses and the probability estimates were higher than theresearchers expected in some cities. Regarding Houston, the city with the highest percentage, norecapture contacts were made for both the first and second ad, thus we did not have a sense of theboundaries of the population and the number is the direct calculation of other ads placed duringthat same time period and the number of contacts. Even if we used the most conservativeprobability estimate of 13.8%, the estimate of 169,920 would account for one of every sevenmales over the age of 18 in the Houston metropolitan area.A strength of this study is that the method of data collection and analyses can easily bereplicated and changes in the estimated population over time or pre/post an intervention can becalculated. Limitations of this study include that we were only able to gather useable data fromone website ( in 15 metro areas and we did not make any contact with any of thecustomers so verification of their intent to solicit sex from the posted ad was not conducted.Another limitation includes our base population. We elected to use the census numbers fromAmerican Fact Finder, which estimates the adult male population in a geographic area but doesnot consider fluctuation in population like tourism. We included all contacts in our study butwere unable to be sure who was a local resident or tourist except from our attempt to use areacodes to link to local numbers. A final consideration is that perhaps all men over the age of 18are not motivated to solicit sex through online sex ads, but to determine the real number ofpotential customers created a methodological challenge we were not able to estimate with anyaccuracy, thus we used the American Fact Finder census as the base population to develop thepopulation percentage estimate.7

No major weather activities or events occurred during the two dates the ads were posted butin Miami, a NBA Finals game was held the day before the second ad was placed. A number oftrends in the data were noted including: Atlantic City contacts were the least likely of all of the 15 cities to be from a local areacode. Kansas City contacts were made via call or voicemail (i.e. no texts) and had the highestrate of local area codes (88.6%); they were also the most persistent, with 50% beingrepeat callers. Most of the total contacts were via voicemail except in Miami and Salt Lake City wherecontacts were primarily by text (61.5% and 76.5%).Future ResearchThis study is a first step towards gaining a better understanding of the size of thepopulation of online sex buyers. Other information that would be helpful to develop this area ofresearch, as well as assist in the development of legislative and law enforcement actions, wouldinclude having better participation by online sex ad providers like about thebehavior of viewers and the number of clicks each advertisement receives.The new information provided in this study about the size of the online sex buyerpopulation is simply a contribution to the field of demand research and ads a new piece to thepuzzle. The intention of this study is to inform law enforcement, advocacy group and policymakers about the scope of the illegal behavior of buying sex from ads posted on websites and tobegin a dialogue about how to best integrate this new information into the demand reductionactivities currently being implemented.Contact Information:Dr. Dominique Roe-Sepowitz: [email protected], 602-496-0093James Gallagher, M. Admin. at [email protected]; 602-426-1081Eric Hedberg, Ph.D. at [email protected] Schmidt at [email protected]

ReferencesBennetts, L. (2011, July 25). The john next door. Newsweek, 158(4), 60-63.Brewer, D., Roberts, J.M., Muth, S., & Potterat, J. (2008). Prevalence of male clients of street prostitute women inthe United States. Human Organization, 67, 346-356.Chapman, D.G. (1951). Some properties of the hypergeometric distribution with applications to zoological samplecensus. University of California Publications in Statistics, 1, 131-160.Hickman, M., Cox, S., Harvey, J., Howes, S., Farrell, Frischer, M., Stinson, G., Taylor, C., and Tilling, K. (2002).Estimating the prevalence of problem drug user in inner London: A discussion of three capture-recapturestudies. Addiction, 94, 1653-1662.Ismail, A., Beeching, N., Gill, G., & Bellis, M. (1999). Capture-recapture-adjusted prevalence rates of type 2Diabetes are related to social deprivation. QJM: International Journal of Medicine, 92, 707-710.Karanth, K., Nichols, J, Kumar, S. & Hines, J. (2006). Assessing tiger population dynamics using photographiccapture-recapture sampling. Ecology, 87, 2925-2937.Larson, A., Stevens, A., & Wardlaw, G. (1994). Indirect estimates of ‘hidden’populations: Capture-recapturemethods to estimate the number of heroin users in the Australian capitol territory. Social Science andMedicine, 39, 823-831.Michael, R. T., Gagnon, J. H., Laumann, E. O., & Kolata, G. (1994). Sex in America: A definitive survey. Boston,MA: Little, Brown.Monto, M. A. (2004). Female prostitution, customers, and violence. Violence Against Women,10, 160-188.Monto, M. A. (2010). Prostitutes’ customers: Motives and misconceptions. In R. Weitzer (Ed.), Sex for sale:Prostitution, pornography and the sex industry (2nd ed., pp. 233-254). New York, NY: RoutledgeMonto, M., & Milrod, C. (2013). Ordinary or peculiar men? Comparing the customers of prostitutes with anationally representative sample of men. Journal of Offender Therapy and Comparative Criminology,online. doi: 11.1177/0306624X1340487Monto, M., & Hotaling, N. (2001). ‘Predictors of Rape Myth Acceptance among Male Customers of StreetProstitutes. Violence Against Women, 7, 275–293.Pinto, N. (2011, March 23). Women’s funding Network Sex Trafficking Study is Junk Science. Village Voice,Retrieved from ence/Sanders, T. (2008). Paying for pleasure: Men who buy sex. Portland, OR:Willan.Shively, M., Kliorys, K., Wheeler, K., & Hunt, D. (2012). National overview of prostitution and sex traffickingdemand reduction efforts (Final Report 2008-IJ-CX-0010). Retrieved 96.pdfSmith, T. W., Marsden, P. V., & Hout, M. General Social Survey, 1972-2010 [Cumulative File] (ICPSR31521-v1).Storrs: Roper Center for Public Opinion Research, University of Connecticut/Ann Arbor, MI: Interuniversity Consortium for Political and Social Research [Distributors], 2011-0805.doi:10.3886/ICPSR31521.v1Soisalo, M., & Cavalcanti, S. (2006). Estimating the density of a jaguar population in the Brazilian Pantanal usingcamera-traps and capture-recapture sampling in combination with GPS radio-telemetry. BiologicalConservation, 129, 487-496.Victims of Trafficking and Violence Protection Act (2000). Retrieved on August 25, 2013 from

of Children, ASU Office of Sex Trafficking Intervention Research and supported by the Phoenix Police Department. Research Team from the Office of Sex Trafficking Intervention Research (STIR) Dominique Roe-Sepowitz, MSW, Ph.D., STIR Director Kristine Hickle, MSW, Doctoral C