In 1992 arrests for prostitution in the United States ranked tenth in the number of arrests of female adults (1), with more than 55,000 women arrested for prostitution annually. Prostitutes frequently suffer from mental health problems such as depression, anxiety, and posttraumatic stress disorder (2,3,4,5), sexual victimization by customers (6), and potentially fatal sexually transmitted diseases (2,7,8,9,10,11,12). Women who are prostitutes have difficulty disengaging from the lifestyle of prostitution (2,4,13). The prevalence of prostitution and its profoundly negative mental health outcomes underscore the need to understand the factors that lead women to enter into this high-risk lifestyle.

Research on the precursors of prostitution has centered around three very different pathways: childhood sexual victimization, running away, and drug use. Each of these potential pathways suggests markedly different prevention and treatment strategies, and the need to understand their respective contributions is critical.

Researchers have hypothesized that childhood sexual victimization leads to deviant behavior, including prostitution (2,14). Some investigators have found higher rates of childhood sexual victimization among prostitutes than among nonprostitutes (13,15,16,17,18,19,20), although others have not (21,22). Studies have reported that 50 to 60 percent of juvenile and adult street prostitutes had been sexually abused before age 16 (4,5,17), often incestuously (16,17). In another study, nearly 70 percent of women prostitutes reported that childhood sexual victimization influenced their decision to engage in prostitution (5).

Others have studied rates of prostitution among known victims of sexual abuse. Widom (23) and Widom and Ames (24) found that sexually abused children had 27.7 times the odds of being arrested for prostitution as adults than did matched control subjects, although Benward and Densen-Gerber (25) found no differences in rates of prostitution between incest victims and non-incest victims in a sample of drug users.

Other researchers have argued that is it not sexual victimization per se that predicts entry into prostitution. Rather, it may indirectly lead to prostitution because sexually abused girls often run away from home. Prostitution may be one of the few viable economic alternatives for runaway girls living on the street. Family alienation, running away, and familiarity with street culture encourage a wide variety of deviant or criminal acts, including prostitution (22,25,26,27,28).

Indeed, studies have documented both high rates of sexual abuse among runaways (14,18,29,30,31) and high rates of running away among prostitutes (18). Compared with nonprostitute control subjects, girls and women who became prostitutes had run away much more often (27,32), left home permanently at significantly younger ages (15), and were living on their own earlier (16). However, other studies did not find that runaway behavior mediated the relationship between childhood sexual victimization and prostitution (19,23).

A third route of entry into prostitution besides childhood sexual victimization and running away may involve drug abuse. Drug abuse may leave girls both emotionally vulnerable and economically motivated to prostitute (33,34).

Although high rates of drug use have been documented among prostitutes, the relationship between drug use and prostitution is far from clear. Studies have demonstrated higher rates of drug abuse among women with a history of prostitution among female arrestees (34), adolescents (32), abused children (27), runaways (35), and clients of clinics treating sexually transmitted diseases, including HIV (20). However, Bour and colleagues (21) found the opposite in a small sample of detained delinquent children.

Many studies have found that drug use is as likely to have preceded entry into prostitution as to follow it (13,34,36,37,38,39), although Potterat and colleagues (20) found drug use usually preceded prostitution in their sample of prostitutes. In a more in-depth analysis, James (36) found that younger women tended first to start drugs and then enter prostitution, possibly as a support system, while women who entered prostitution later in life were less likely to become addicts.

In general, research on precursors to prostitution is limited. Most samples are small (15,16,17,21,22,25,27,34,36,37,40) or do not have ideal control groups (15,17,18,19,25,37). Only two studies have examined the impact of all three variables—childhood sexual victimization, running away, and drug use—on prostitution (18,19). This paper examines the effects of sexual victimization, running away, and drug use on entry into prostitution and the differential effects of these risk factors over the life course. Understanding distinct pathways into prostitution may highlight specific mental health services needed to prevent entry into prostitution and help women seeking a way out.

Subjects and sampling procedure

Female detainees entering the Cook County Department of Corrections in Chicago between 1991 and 1993 were interviewed. A total of 1,272 subjects were recruited and interviewed in the jail's intake area during routine processing.

We chose the Cook County Department of Corrections because its population is demographically similar to that of other large urban jails (41) and because its size ensured adequate availability of subjects. We stratified the sample to ensure adequate representation of rarer groups. A total of 40.4 percent of the sample were African Americans, 33.6 percent were non-Hispanic whites, 24.7 were Hispanics, and 1.3 percent were members of other ethnic groups. Subjects were between 17 and 67 years old (mean and median age, 28 years), with mean and median educational levels of 11 years.

All interviewers were female, had at least a master's degree, and were experienced in interviewing correctional and patient populations—characteristics likely to facilitate successful interviews with the study subjects (42,43). Subjects were interviewed confidentially and in private using a two- to three-hour structured interview protocol that included assessment of psychiatric diagnosis and high-risk behaviors.

The refusal rate was 4.2 percent. Details of the study procedure are available elsewhere (44). Because the instruments addressing prostitution and sexual victimization were finalized after the study began, the analyses reported here are based on data from 1,142 of the 1,272 subjects. Poststratification weights and inferential tests were adjusted accordingly.

Definition of variables

Prostitution.

We used two measures to assess the severity of a detainee's involvement with prostitution—age when the subject first solicited money for sex or engaged in prostitution (episodic prostitution), and age when the subject first started soliciting money for sex or engaging in prostitution more than once a week (routine prostitution). Routine prostitution does not imply duration, but rather frequency of prostituting behaviors. A positive response to having ever solicited money for sex may not indicate a pattern of sex-for-money exchanges. Engaging in sex-for-money exchanges more than once a week, however, is indicative of an established pattern of behavior.

Sexual victimization.

No widely accepted brief measure of childhood sexual abuse existed at the time of the study. We developed a childhood sexual abuse module based on work by Russell (14) and Browne and Finkelhor (31). We defined sexual victimization as an "unwanted sexual experience, like someone trying to touch you in ways you didn't want or trying to get you to touch them in ways you didn't want" before age 16.

Having run away.

Runaways were women who reported having run away from home and staying away overnight before age 15. Although some studies have examined the phenomenon of leaving home permanently (15,16), first runaway behavior during youth is commonly used in this type of research (19,27).

Drug abuse.

Drug abuse, assessed using the National Institute of Mental Health Diagnostic Interview Schedule, version III-R (45,46,47,48,49), was defined as a history of moderate or severe DSM-III-R cocaine or opiate abuse or dependence. A diagnosis of at least moderate substance use or abuse requires frequent intoxication or withdrawal symptoms that impede major role obligations. This definition distinguishes women with serious chemical dependencies from those whose drug use does not have a firm grip on their lifestyles and financial needs.

Cocaine and opiate abuse or dependence accounted for 84.6 percent of all moderate or severe substance abuse or dependence, including that of alcohol, in our sample. Including other types of substance abuse or dependence in the measure reduced the fit of the model, suggesting that the mechanism linking substance abuse to prostitution may be specific to cocaine and opiate abuse or dependence. Age of onset of drug abuse indicates the first time a symptom of cocaine or opiate abuse was present among those who developed serious disorders.

We examined univariate, bivariate, and finally multivariate relationships of our three independent variables—childhood sexual abuse, having run away, and drug abuse—with our dependent variable, prostitution.

Prevalence

Prostitution.

More than one-third of the sample (34.1 percent) reported ever having engaged in prostitution. Subjects who prostituted had their first prostitution experience at a mean age of 20.9 years (see t1). (Because the study used a complex sample design, standard deviations are not reported for univariate means.) Three-quarters of those who ever prostituted, or one-quarter of the women in jail (25.6 percent), reported having engaged in prostitution one or more times a week (routine prostitution), beginning at a mean age of 21 years. Bivariate analyses (t1) show that rates of prostitution differed by race, with white detainees the most likely to report a history of prostitution and Hispanic detainees least likely. African Americans who ever prostituted began, on average, at a younger age than whites or Hispanics, though age of entry into routine prostitution did not differ by race.

Detainees with less education were more likely to have ever prostituted, though routine prostitution did not vary by education. Although age of entry into prostitution and rates of routine prostitution were lower for detainees with the least education, these results are partly due to younger subjects' greater likelihood of still being in school.

Childhood sexual abuse.

More than one-third of the sample (35.3 percent) reported a history of childhood sexual abuse. Among those who were abused, the mean age of first abuse was 10.6 years. Rates of childhood sexual abuse did not differ by race or ethnicity. However, African Americans reported being older, on average, at first abuse (10.9 years), compared with whites (9.8 years) and Hispanics (9.9 years).

Having run away.

More than one-quarter of the sample (29.1 percent) reported a history of having run away, beginning at an average age of 12.8 years. African-American detainees were less likely to have run away (26.7 percent) than whites (39.1 percent) or Hispanics (35.6 percent), although no racial differences in age of first running away were found. Not surprisingly, detainees who had run away had the least education because running away interrupts normal completion of schooling.

Drug abuse.

Just more than half the sample (51.4 percent) met criteria for moderate or severe cocaine or opiate abuse or dependence. The average age at the first symptom was 22.8 years. Hispanics were significantly less likely to be diagnosed as having heroin or cocaine abuse or dependence (35.3 percent) than were African Americans (51.9 percent), whites (55.5 percent), and others (66.7 percent). African Americans reported experiencing the first symptoms of substance abuse or dependence at a slightly later age than others. Prevalence of drug abuse did not vary by level of education, although high school dropouts had a somewhat younger age of onset of drug abuse.

Bivariate analyses

Sexual abuse and prostitution.

Detainees who had experienced childhood sexual abuse had substantially higher rates of ever prostituting (44.2 percent, compared with 28.5 percent for detainees with no history of abuse) and of routine prostitution (34.6 percent, compared with 20.6 percent for nonabused detainees). On average, detainees with childhood sexual abuse also began prostituting at slightly younger ages than nonabused detainees. The average age of entry into routine prostitution did not differ by presence of childhood sexual abuse.

Having run away and prostitution.

Women who reported childhood sexual abuse first ran away at a slightly younger mean age than nonabused women (12.6 years, compared with 13.2 years for nonabused women). Women who reported ever having run away from home were significantly more likely than nonrunaways to have ever prostituted (44.7 percent, compared with 29.7 percent) and to have engaged in routine prostitution (35.6 percent, compared with 21.5 percent).

Detainees with a history of having run away entered prostitution at a younger age than those who did not run away. However, among those who entered routine prostitution, no age differences were found between those with a history of running away and those who did not run away.

Drug abuse and prostitution.

Substance abuse or dependence was strongly associated with prostitution. Women with moderate or severe cocaine or opiate abuse or dependence were more likely to have engaged in prostitution than detainees without these disorders (53.2 percent versus 13.9 percent) and more likely to have engaged in routine prostitution (40.9 percent versus 9.5 percent). Of women who reported prostituting one or more times a week and who had a diagnosis of moderate or severe cocaine or heroin abuse or dependence, 67.4 percent experienced the onset of substance abuse symptoms before beginning routine prostitution, 10.3 percent were uncertain which behavior started first or reported that both started in the same year, and 22.3 percent experienced the onset of substance abuse or dependence after the start of prostitution. Interestingly, substance abuse or dependence did not affect the age of onset of engaging in either episodic prostitution or routine prostitution.

In summary, the bivariate analyses suggested that all three variables—sexual abuse, having run away, and drug abuse—predicted entry into prostitution.

Multivariate analysis

To test for the relative influence and timing of each predicted effect, we performed an event-history analysis of timing of entry into routine prostitution. We modeled the timing of entry into prostitution using a refinement of discrete time logistic regression known as the piecewise constant nonproportional odds model (50). Discrete time logistic regression breaks the time line into discrete units. In our analysis, the odds ratios of association between entry into prostitution and discrete time periods and other independent variables were estimated. Because it is based on the odds ratio, discrete time logistic regression was amenable to our retrospective study design (51,52).

In addition, because we expected that the odds of entry into prostitution were not equal over the life course and that the effects of independent variables varied with time, we used the piecewise constant nonproportional odds model to analyze the timing of entry into prostitution. With this method, discontinuities in the baseline hazard, or instantaneous risk, of entering prostitution at various time periods are modeled by including dummy variables for each period, and effects that vary by time are modeled using interaction terms between each dummy variable and other predictors (50).

Other analytic survival techniques were not suited to our data. Quasiparametric methods such as Cox's proportional hazards model allow for an undefined baseline hazard, but the number of ties in our data rendered these models computationally troublesome and potentially biased (53,54). Fully parametric models of survival require the specification of an explicit survival function and were not appropriate for our data. Our study design required that we treat our data as retrospective and use an appropriate measure of association such as the odds ratio. In addition, our measure of time of entry into prostitution was not truly continuous, but was divided into one-year increments. Finally, as demonstrated below, the baseline hazard rate of entry into prostitution did not follow a continuous distribution, so we needed a method that could model nonproportionality directly (52,53).

We began by identifying discrete life periods for our analysis. (Details of this procedure are available from the authors.) The uncorrected hazard rate of entry into prostitution was quite low before age 15, was high from age 15 to 25, and then declined. We examined alternative breakpoints and settled on four life periods pertinent to entry into prostitution—the early adolescent years to age 15, the adolescent years 15 to 17, young adulthood from age 18 to 24, and adulthood over age 24. We checked our results against the baseline hazard function estimated with a Cox regression model when controlling for covariates and found a good fit between this hazard and that of our discrete time model. (Results of this comparison are available from the authors.)

Because our sample was stratified by race and type of criminal charge (misdemeanor or felony), we corrected inferential statistics by conditioning the analysis on these variables (55). Examination of these variables for time dependence determined that interactions with time variables were not needed. That is, the sample strata were treated as time invariant. Time-dependent covariates in the model—childhood sexual abuse, running away, drug abuse, and leaving school—were coded as present for each time period only if the measured onset was during or before the reported year of entry into prostitution (50).

t2 reports our complete and final models. The complete model includes childhood sexual abuse, having run away, leaving school, moderate or severe opiate or cocaine abuse or dependence, baseline function, and design variables. Only the design variables, the baseline function, childhood sexual abuse, and having run away had an impact on entry into routine prostitution. Neither the complete model nor the final model supported the hypothesis that drug abuse affects entry into prostitution.

We used backward removal based on likelihood ratio tests to select the final model. The final model reveals that childhood sexual victimization had a lifelong effect on entry into routine prostitution, doubling or nearly doubling the odds of entry into prostitution throughout the lifetime. Having run away, by contrast, affected entry into routine prostitution only in the early adolescent years, increasing the odds of entry into prostitution during that period by more than 40 times. In short, the analysis provided strong support for the hypothesis that sexual abuse and having run away influence entry into prostitution, although these risk factors were found to have distinctly different impacts over the life course. The hypothesis that drug abuse affects entry into prostitution was not supported by the complete or the final model.

Sample design variables had some effect on entry into prostitution. Non-Hispanic whites were most likely to report entry into routine prostitution, and Hispanics were the least likely. Detainees arrested for felonies were least likely to have prostituted routinely. These effects may have been due to the selection process leading to arrest and detention.

Our study design has several limitations that qualify our results. First, retrospective self-report data are potentially inaccurate. We used self-report data because they provide more complete information than official criminal records, which many researchers use, especially in studies of victimless criminal behaviors such as prostitution.

Second, measures of deeply personal material—particularly childhood sexual abuse—would have been more sensitive if subjects had been given more than one opportunity to disclose their experiences (14). Although we used experienced interviewers and were careful to ask questions about sexual behaviors and sexual abuse at the end of the interview, it is likely that our rates underrepresent the extent of abuse in the sample, as well as the extent of prostitution and drug use.

Third, sampling from a jail population yields a biased sample of prostitutes. Poor persons are more likely to be arrested and jailed before trial than are wealthier persons, and some types of prostitutes—for example, those who are addicts—are more likely than others to be arrested (36,56,57). In addition, some persons begin and end their careers as prostitutes without ever being jailed; they may either abandon their criminal careers, be imprisoned for long periods, or die. Thus our findings address the process of entry into prostitution for only detained prostitutes.

Despite these limitations, our findings indicate that a history of childhood sexual victimization is a powerful risk factor for entry into prostitution among women in jail. Our study extends and confirms findings about the relationship between childhood sexual abuse and prostitution in other populations (15,16,17,18,19,24).

Our results also confirm a relationship between running away and prostitution. Overall, female jail detainees who had run away were significantly more likely to prostitute than nonrunaways, and runaways also had an earlier onset of episodic prostitution. However, we found that running away was a risk factor for prostitution only in childhood and adolescence. These results may clarify conflicting findings in the literature about the association between running away and prostitution (19,23,27).

Women in jail with moderate or severe drug abuse or dependence were more likely to have run away. They were also more likely to have participated in prostitution either episodically or routinely. However, surprisingly, when running away and childhood sexual victimization were controlled for, drug abuse was not a risk factor for entry into prostitution. The onset of symptoms of drug abuse did not influence the timing of entry into routine prostitution. Our findings are consistent with those of other researchers who found that drug abuse is as likely to follow the onset of prostitution as to precede it (13,36,37).

Our findings—that entry into prostitution has distinct pathways—highlight areas on which mental health services could focus. First, early recovery of children who run away is necessary to reduce entry to prostitution. Services must address children's basic survival needs by providing a stable emotional and environmental support system. At the very least, youths who run away require a safe place to live, medical care, and the opportunity to develop work skills (18,27,58,59).

Second, the impact of childhood sexual abuse on entry into prostitution appears to persist over the life course. Victims of childhood sexual abuse need mental health services to help them come to terms with their victimization and restore a sense of mastery and control over their lives (2). The need for this help—and the potential impact on a victim's ability to make positive choices—does not diminish as women age.

Third, although preventing or treating drug abuse may not be the most powerful deterrent to entry into prostitution, intervention is still vital. Addict-prostitutes tend to be heavy users and report that narcotic use increases with further involvement in prostitution (33,36). Although entrance into prostitution before addiction may be due to effects of sexual abuse, entering prostitution after developing an addiction may be an economic necessity (36). Addiction complicates women's efforts to leave prostitution, and drug treatment needs to begin early in their prostitution careers (34). Prostitutes who use drugs are also at a significantly higher risk for HIV infection (60).

Most women who want to extricate themselves from prostitution have many service needs. Social services must provide alternative and crisis housing, employment and school counseling, and care of children; lack of child care is a significant barrier to women's accepting residential services (2,28,61,62,63). The most successful programs provide outreach to foster confidence that help is genuine and trustworthy (62,64).

Future research should examine dynamics that lead women into prostitution. For example, which aspects of abusive experiences place children at higher risk of entering into prostitution? What are potential protective factors in early childhood (24)? Understanding pathways to prostitution will aid in the development of prevention and treatment programs for sexually victimized girls and women.

This work was supported by grants RO1-MH-45583 and RO1-MH-47994 from the National Institute of Mental Health. The authors thank Cook County sheriff Michael F. Sheahan and former Cook County department of corrections executive director J. W. Fairman, Jr.

The authors are affiliated with Northwestern University Medical School in Chicago. Send correspondence to Dr. McClanahan, 675 North St. Clair Street, Suite 20-250, Chicago, Illinois 60611 (e-mail, mcclanahan@nmff.nwu.edu). This paper is part of a special section on mentally ill offenders.