Many studies of prostitution can be faulted on methodological grounds. Some authors fail to describe how and where they contacted research subjects. Others fail to include comparison groups (nonprostitutes matched on demographic characteristics; e.g., age, social class), without which it is impossible to know if the findings reported for a prostitute sample differ significantly from those of nonprostitutes. Those few studies that do include appropriate control groups yield mixed results. Some find significant differences between prostitutes and controls on, for instance, history of childhood victimization, whereas others find no significant differences (Earls & David, 1989; Nadon, Koverola, Weitzer / FLAWED THEORY AND METHOD 937Copyright 2005. Permission granted by Sage Publications & Schludermann, 1998). When it comes to victimization in prostitution, studies are “often methodologically flawed and, moreover, contradictory” (Vanwesenbeeck, 2001, p. 259). Reliance on unrepresentative samples is widespread.
Although random sampling of sex workers and customers is impossible, too often the findings and conclusions drawn from convenience and snowball samples are not properly qualified as nongeneralizable. Victimization studies are a case in point. Street prostitutes appear to experience high rates of violence in the course of their work, but the samples used in most studies consist of people who contacted service agencies, were approached on the street, or were interviewed in jail (James & Meyerding, 1977; McKeganey & Barnard, 1996; Weisberg, 1985). The high victimization rates reported in such studies are thus vulnerable to selection bias: The most desperate segment of the population or those persons who are most frequently or seriously victimized may be especially likely to contact service providers or agree to interviews. Generalizing from prostitutes in custody to the population of prostitutes is also improper, just as with other types of incarcerated offenders. Yet the implications of this sampling bias typically are neglected in the published reports. Moreover, the victimization rates reported are often reproduced in the secondary literature and in newspaper reports without disclosing the sampling technique and its limitations.3
To cite just one example of this tendency: Silbert and Pines (1982, p. 127) studied 200 street prostitutes in San Francisco and reported that 45% had been robbed, 65% had been beaten, and 70% had been raped or had experienced a customer “similarly going beyond the work contract” (a bit vague). The authors hired interviewers who were former prostitutes, had been residents of a treatment facility in the city, and “had been victims of various assaults” when they worked as prostitutes (Silbert & Pines, 1982, p. 123). Despite the problematic orientation of the interviewers (given their past experiences) and the fact that the prostitutes interviewed were all drawn from the streets and from a single city, this study is one of the most frequently cited sources (by Farley and others) of evidence that violence is rampant in prostitution.