What money buys: clients of street sex workers in the US
March 2006
Marina Della Giusta (University of Reading, UK)
Maria Laura Di Tommaso (University of Turin, Italy)
Isilda Shima (University of Turin, Italy)
Steinar Strøm (University of Oslo, Norway and University of Turin, Italy)
Abstract
Using data from a survey of clients of street sex workers (Monto, 1999), we present an
econometric model that explores the effect of personal characteristics and attitudes of clients
on their demand for prostitution, in particular to assess the relative importance of variables
related to power and control.
The results reveal the existence of two diametrically opposite profiles: one for clients
who declared never to have been with a sex worker or to have been only once, whom we label
“experimenters”, and one for the more experienced ones that we name “regulars”.
The ‘experimenters’ are typically white men, they like control and are gender-violent
and think that sex workers particularly like sex more than other women. Variety of
preferences is of no concern for them If married, they consume more paid sex. Thus the
“experimenters” are more of the macho type sex clients, and sex with prostitutes is a
complementary good to sex with their regular partners. “Regulars” on the other side are
typically non-white, single or with trouble in having stable relationships with women. The
regulars demand more sex with prostitutes the less gender-violent they are and the more
favourable they are to prostitution. They also favour variety of preferences in sex life. The
“regulars” thus seem to demand sex as a substitute for not having sex with a regular partner.
The “regulars” are typically the lonely losers.
Acknowledgments: Project, conferences, people at seminars…
1
1. Introduction
The social scientific literature on prostitution is vast (recent authoritative monographs
on the subject are O’Connell Davidson, 1998, and Lim, 1998) and representative of many
different views and concerns. A substantial part of the literature on prostitution consists of
studies of prostitution and its relationship with violence, health and drugs problems, and
international migration, and is often devoted to investigating the desirability of alternative
regulatory regimes and the definition of rights for sex workers (McKeganey and Barnard,
1996; O’Kane, 2002; Thorbek and Pattanaik, 2002; Kempadoo and Doezema, 1998; Tiggey et
al, 2002).
The industry is very fragmented and characterised by the presence of different types
of intermediaries, which play a particularly important role, ranging from more to less
exploitative (varying from slavery conditions to self-managed prostitutes and trade unions),
and appear to differ in the different sub-markets (Tiggey et al. 2000, Sharpe 1998, Garofalo,
2002). The activity is mostly entered for economic reasons, and appears to be a job done
largely but not exclusively by women (Aggleton, 1998; Sanchez Taylor, 2001). Supply and
working conditions depend partly on the availability of other livelihood opportunities (not just
other jobs and their pay, but also the working conditions) and partly on social stigma.
Criminalisation and social stigma make this a high-risk activity for prostitutes, both in terms
of health and of personal safety. Evidence suggests that specific skills are required in this
occupation, which include the ability to maintain emotional detachment and separate one’s
identity of prostitute from one’ other identities (Sanders, 2005; Chapkis, 1999; McKeganey
and Barnard, 1999), and the ability to defend oneself and cope with risks, given the
occurrence of violence against prostitutes is widespread and often little protection is offered
by institutions. As argued by Nussbaum (1999, p. 284), prostitution does require skills, but
different from the ones we are used to measuring. Motivations for entering prostitution are
often found to differ in other “higher” segments of the industry, such as those found by
Chapkis’ (1997) study of sex workers in parlours, escort agencies, and call girls, for whom
there appeared to have been more choice exercised both in entering the business, and in
controlling the conditions of their work and the way they felt about it. Evidence of the
different working conditions in different parts of the industry was also found in the second
episode of the Channel 4 programme that presented the stories of workers in different kinds of
prostitution (mainly parlour-based). Sex workers described their working conditions as safer
(in as far as the relationship with clients and pimps is concerned), and some of the workers
were lobbying for de-criminalisation or legalisation, following the experiences of countries
with a more open attitude towards prostitution which give legal protection to the workers
(Australia and the Netherlands).
2
The economic literature has traditionally approached prostitution either showing how
it is similar to other markets, or studying it as a form of crime and analysing the costs and
benefits of alternative regulatory regimes, generally agreeing that the main motivation behind
supply is an economic one (for a review, see Reynolds, 1986). More recent theoretical and
empirical contributions have focussed on modelling prices (Cameron et al, 1999; Moffat and
Peters, 2001; Edlund and Korn, 2002; Cameron, 2002), supply determinants (Cameron and
Collins, 2003), health risk and the effect of condom use on sex worker’s earnings (Rao et al,
2001; Gertler et al, 2003), and, more recently, the evolution of paid sex markets and the ways
in which urban spaces favour sexual transactions (Collins, 2004). The latter collection is
much broader in scope, with paid sex markets being studied as part of the wider sexual market
in which people seek partners for reasons that include deficiencies in amount or range of
sexual activities in which they participate, or diversification of sexual consumption (Collins,
2004, p.1634).
Whereas studies of sex workers are widespread, those who address the demand side
of the industry are harder to come by, and wanting to rigorously analyse demand
characteristics on the basis of empirical evidence can prove very difficult:
‘Presumably, the client has not been studied until very recently because his actions are not
perceived as morally reprehensible. A man who buys sex is viewed simply as a "man" doing
"what men do" and therefore there is nothing unique or interesting enough about his behavior
to justify research. There is no contradiction between legitimizing the client’s activities, and
preserving the smokescreen around the paid sex industry, since sex and sexuality are
considered “private” matters. And privacy is especially important in the case of purchased
sex, a potential source of embarrassment; a visit to a prostitute may be construed as failure to
obtain sex by consent or adultery. For this reason, paid sex is considered legitimate, even
“natural,” but part of a private realm that is best left un-discussed.
In the US 16% of men reported buying sex at least once in their lives, and 0.5 %
reported doing so at least once a year. In Finland, as in Russia, it was found that 10-13% of
men had purchased sex at least once. In Norway, the comparable figure is 11%, in Holland
14%, in Switzerland 19%, in London 7-10%, and in Spain 39%.29 Figures in the 70% range
have been recorded for Cambodia and Thailand, but these, too, appear to be imprecise
estimates. In the absence of precise data, the only formula most researchers agree on is that
the higher the degree of conservativeness, and the more rigid the social norms regarding the
place of women, the higher the demand for paid sex and the thicker the veil of secrecy
surrounding it’ (Ben-Israel and Levenkron, 2005: 13)
3
Pitts et al (2004) surveyed a sample of 1225 men in Australia and found that 23.4%
had paid for sex at least once, and reported paying for sex to satisfy sexual needs (43.8%),
because paying for sex is less trouble (36.4%), and because it is entertaining (35.5%), the
factors that the authors identified as accounting for 55% of the variance in motivations were
ease, engagement and arousal. Significantly, they found that there were not many significant
differences between men who had paid for sex and those who had not, other than being on
average older, less likely to have university education and to have had a regular partner.
The motivations of clients in the UK (who were all males and appeared to be
representatives of all sectors of society) were addressed in the third part of the Channel 4
series. These appeared to convey the impression that a connection existed between the effort
and costs associated with finding a sexual partner who would readily satisfy their sexual
preferences, and the straightforward and readily accessible option of prostitution.
Such characteristics and motivations of clients are analysed in the collection of
studies edited by Thorbek and Pattanaik (2002), which documents the increase in prostitution
across borders. A sort of “psychological” profile of male sex tourists is drawn on the basis of
their own descriptions of themselves and accounts of their experiences, which suggests that
many of them are finding relationships with others very difficult (either because they don’t
have the time or the skills required to meet people) and choose sex tourism as an “easier”
alternative, which does not imply any responsibility. As for the views they hold of sex
workers, it appears that both sexism and racism mix in determining a very marked distancing,
which allows clients to practically ignore and show no interest in the lives and working
motivations of the sex workers whose services they buy. Wider phenomena connected to
consumerism and globalisation are also clearly related to this industry, which reflects multiple
power structures: Marttila (2003) concludes from her study of Finnish clients ‘the sex
business is first and foremost about gendered, economic, social and cultural – global and local
– power structures. Structural inequalities, the new information and communication
technologies and increase in movement and moving of people have a considerable importance
in expansion of the sex industry as well. Hedonistic need for constant change and new
“products” grows demand for “exotic” prostitutes and thus sustains international sex trade.
Political changes and poverty (and especially feminization of poverty) offer “new bodies” to
the market and thus respond to the demand respectively. The demand thus has a major role in
sustaining the international sex trade and trafficking in women and girls. This again highlights
the importance of drawing attention to the client and to his position in the global sex trade.
Keeping in mind that most of the clients are men, it is also important to study how the
hegemony of men is structured and maintained in the context of globalising sex industry’
(Marttila, 2003, p.8)
4
Women clients are also engaging in sex tourism, as documented both in Thorbek and
Pattanaik, and in Sanchez Taylor (2001). The latter in particular offers a more in-depth
analysis of North American and Northern European women buying prostitution services of
young men in the Caribbean, in what they themselves describe as ‘romance holidays’.
Responses to her interviews suggest that, on the one hand, the women clients are mostly
reluctant to define what they engage in as prostitution, and on the other that their ideas about
the young men whose service they buy are deeply rooted in racist ideas about black men and
black men’s sexuality. The theme of inequality appears to be at the core of the relationship:
prejudices that allow the stigmatisation of another sex worker as fundamentally “different”
and inferior to oneself appear again and again in customers accounts. Thus both men (the vast
majority) and women demand prostitution services, and interviews with clients appear to
suggest that demand is underpinned by complex ideas of machism and racism which are at
play in the exchange of such services, suggesting that aspects of power and control are
essential to this transaction (McKeganey and Barnard, 1996; O’Kane, 2002; Thorbek and
Pattanaik, 2002, Kern, 2001). As often found in qualitative studies of inequality, these same
systems (patriarchy, racism, etc.) provide mechanisms for a partial subversion of the
stigmatisation, so that both prostitutes and clients tend to describe themselves as in control of
the relationship (Chapkis 1997; McKeganey and Barnard, 1996). Several studies also find that
clients want to feel mutual dependence and that it is not a pure market transaction. The effects
of personal characteristics (personal and family background, self-perception, perceptions of
women, sexual preferences), economic factors (education, income, work), as well as attitudes
towards risk (health hazard and risk of being caught where prostitution is illegal) are all likely
to affect demand. Usually, in the literature is found that the prostitution clients are attracted to
the prohibited nature of the encounter, lack interest in conventional relationships, desire
varieties of sex that regular partners do not provide, view sex as a commodity. Their decision
to approach to sex worker is influenced by the availability of sex workers, access to money,
perceived risk of getting caught or sexual disease.
Below we estimate demand for sex, and how this demand vary with the attitude of
clients towards paid sex and personal characteristics, on data from a survey of clients of street
sex workers (Monto, 1999). In the analysis we distinguish between those who are buying sex
for the first or second time, here named “experimenters” and the more experienced clients,
here name “regulars”. It turns out that the ‘experimenters’ like control and they are gender
violent. They are typically white and if married, they consume more paid sex. Thus sex with
prostitutes tends to be a complement rather than a substitute for having sex in a stable
relationship. “Regulars” are typically non-white, single or with problems in having stable
relationships with women and sex with prostitutes is a substitute for having in such stable
relationships.
5
The paper is organized as follows. First we present the data set. Next we specify
econometric models of demand for sex. Empirical results follow. In Section 4 we conclude..
2. The Monto dataset
The dataset contains background characteristics, attitudes, and reported behaviours of
arrested male clients of female street sex workers in five US cities over the period 1996-1999
(Monto, 2000). The data was collected in the context of two client intervention programmes
aiming to address the demand side of prostitution (Portland’s Sexual Exploitation Education
Project and San Francisco’s First Offender Prostitution Program, both aiming at prevention
efforts with clients, rather than with sex workers), asking clients questions about their sexual
behaviour (number and type of partners, frequency of sex, interest in pornography, age and
circumstances of first sexual encounter with a sex worker, sexual acts performed with sex
workers, condom use with sex workers), their attitudes toward premarital sex, homosexual
sex, extramarital sex and sex between adults and children, attitudes towards sex workers, the
legality of prostitution and violence against women. Background information about the clients
included race, educational level, sexual orientation, marital status, work status, socioeconomic status, age, parent’s marital status, and history of sexual or physical abuse, military
service, relationship history, and sexual preferences. Some of the men participating in the
programmes were required to do so as part of their sentence, others had reduced fines or the
arrest purged from their records in exchange for their attendance.
The sample includes 1342 respondents, and the questions addressed by the project
covered both the characteristics of men who solicit prostitution and their motivations, and
their views of sexuality, of women and of violence against women in particular1. (Monto,
2002, p.4). The analysis of the study uses a data set constructed on the basis of Monto Data
Set published by the U.S Department of Justice. The data are collected from self-administered
questionnaires. Arrested clients of street sex workers who accepted to participate to an
intervention program compiled a detailed anonymous questionnaire. Over 80 % of
participants completed questionnaires2.
1
Although the number of working sex workers in the US is difficult to estimate, the Department of Justice arrest statistics for
prostitution consistently exceed 100,000 per year (FBI, 1997; Barkan, 1997). These statistics tend to underestimate the number of
sex workers who are arrested each year. Prostitution- related activities may be processed under other statutes, such as nuisance
laws (San Francisco Task Force on Prostitution, 1996), and arrests of juvenile sex workers may be processed as status offences
(Alexander, 1987).
2
Though refusals constituted the largest single category of non-completions, language barriers and late arrivals also accounted
for a substantial proportion. Of these 1,342 respondents, 36 from San Francisco and 15 from Las Vegas completed a Spanishlanguage version of the questionnaire. Completing the English version of the questionnaire were 950 men from San Francisco,
254 from Las Vegas, 77 from Portland, and 10 from Santa Clara. The period is 1996-1999.
6
2.1 The characteristics of arrested clients
Table 1: Characteristics of arrested clients
Responses
Variable description
Race
56 %
White
Education
More than college after high school
Work Status
35 %
78%
5.6%
Full time
Part time
Average age
Age 25-36
Age 37-46
Age 37-56
Age 57-66
Age >66
38 years ( min=18 and max=84)
31,77%
26,23%
13,92%
3,92%
1,77%
Marital Status
Married
Never married
Married never with sex worker
Never Married - never with sex worker
Marriage description
Very happy
Not too happy
42%
34%
25,18%
16,16%
36%
20%
The majority of clients were white (56%), the rest were a mixture of other racial
backgrounds including Hispanic and Black African. Clients were relatively highly educated
(35% had some college after high school), which is an interesting fact since it contradicts the
general view that clients of sex workers have working class backgrounds. In addition, 78% of
the sample were in full time employment, 5.6% had a part-time job, and the rest were
unemployed, retired or students. The age of the clients approaching and arrested with sex
workers was on average 38, with a minimum age of 18 and the maximum age of 84. The
highest number of clients was found in the age group 25-45.
Regarding marital status, 42% declared they were married, while 34% were never
married; 36% said they were “ very happy” in their marriage, against 20% “not too happy”.
It is interesting to note that 20% of the sample claimed that they never have had
sexual relations with a sex worker.3 Among married men 141/560 claimed the same thing
3
Because men in the sample were almost all arrested while propositioning a decoy posing as a sex worker, it is possible that
some had never before sought out a sex worker or had not successfully completed the transaction. Men arrested for trying to hire
street sex workers appear to be less experienced prostitution clients, with more experienced clients better able to avoid arrest,
7
while among those never married only 16% denied having had sex with sex workers. An
explanation for this could be that clients are embarrassed about their behaviour, and fear
damaging their public reputation, and as results they deny the evidence. Referring to the
sexual behaviour and attitudes the arrested clients declared to have 1 sexual partner during
last 12 months (36 %), 2 partners (16 %), 3 partners (10 %) and more than 5 partners (12 %).
Regarding the frequency of sex during last year, the highest rate was 3 times per
month (20 %), about once a week (18 %), 2 or 3 times a week (16 %) and more than 3 times a
week (7 %). The first experience with a sex worker was found in the age group 18-25..
The most common circumstance of the first encounter with a sex worker was being
approached by a sex worker (25%), followed by “they approached the sex worker on their
own” (23%), and “a group of buddies set me up” (18%). The most frequent sexual act done
with the sex worker was “blow job/fellatio” (36%), followed by vaginal sex (10%), more than
2 acts (12 %) and “half and half (7 %). Only 56% of the sample declared that they always
used a condom, and 3% declared that they never use it, for more details see Table 2.
either due to knowledge of police procedures, familiarity with the sex workers themselves, or participation in off-street
prostitution. Monto (1999)
8
Table 2: Attitudes toward sexual behaviour
Variable description
Responses
Number of sex partners during
last 12 months
-
1 partner
2 partners
3 partners
more than 5 partners
Frequency of sex during last 12
36%
16%
10%
12%
months
-
3 times per month
Once a week
2-3 times per week
More than 3 time per week
The age when first with sex
worker
20%
18%
16%
7%
Min age = 10
Max age = 62
Circumstances when first with sex
worker
-
Were approached by sex workers
25%
They approached the sex workers on 23%
their own.
A group of buddies set them up
18%
Mostly done with a sex worker
36%
blow job / fellatio
10%
vaginal sex
12%
Checked more than 2 acts
7%
Half and half
Condom use with sex workers
56%
Always use it
3%
Never use it
Watch videos/ look at magazines
34%
Never
37%
Less than once a month
17%
1 to a few times a month
Sex with sex worker during last
year
-
Never
21%
Only one time
21%
More than 1 time but less than once 28%
per month
1 to 3 times per month
7.3%
Once or 2 times per week
1.3%
3-4 times per week
0.3%
5 or more times per week
0.37%
9
2.2 Motives for Seeking Prostitution
Arrested clients were asked to agree or disagree with 13 statements designed to reflect
popular and scholarly understandings of the reasons men seek out sex workers. Many
conventional understandings were supported by the results.
The percentage of agreement, strongly and somewhat, for the respective statements is as
follows:
1.
I have difficulty-meeting women who are not nude dancers or sex workers.( 20.9 %)
2.
I think most women find me unattractive physically. ( 22.13%)
3.
I want a different kind of sex than my regular partner. (37.60%)
4.
I am shy and awkward when 1 am trying to meet a woman. ( 38.4%)
5.
Would rather have sex with a sex worker than have a conventional relationship with
a woman. (16.4%)
6.
I am excited by the idea of approaching a sex worker. (40%)
7.
I don't have the time for a conventional relationship. (30%)
8.
I don't want the responsibilities of a conventional relationship. (26%)
9.
I like to have a variety of sexual partners. (38%)
10. I like to be in control when I'm having sex. (38.5%)
11. I like to be with a woman who likes to get nasty. (47.3%)
12. I need to have sex immediately when I am aroused. (28.8%)
13. I like rough hard sex. (17.6%)
From the responses it can be observed that a considerable number of clients appear to be
excited by the illicit, risky, or different quality of sex with a sex worker4. Findings also
suggest that some men pay for sex because they have difficulty becoming involved in
relationships. For some of these men prostitution is an attempt not only to have sex, but also
to establish intimate relationships with women. Kern (2001) got similar results. Among
clients, some of the men expressed that they had the time, energy, or interest also to engage in
a conventional relationship with a woman5.
4
Responses suggest that, for some clients, one of the appeals of prostitution is that it invites a self-focused, consumer oriented,
conception of sexuality in which one can conveniently meet sexual needs through purchase. Some of the arrested clients report
wanting a different kind of sex than their regular partner and liking rough sex, supporting the idea that some men seek out sex
workers because they can do things with them that other women might find unpleasant or unacceptable. Monto 1999
5
Overall, most of the items seem to reflect a sense of entitlement to sex among the respondents. Though their partners may not
be interested in a particular type of sex or though they don't have time to be involved in a relationship, they may feel that they
have a right to sexual access. Monto (1999)
10
2.3 The woman as the mirror of the man: commodification and attitude towards
violence
By definition the commodification of sexuality implies the exchange of something of value
for sexual access to someone's body. This study includes the evaluation of the degree to
which prostitution clients conceptualise sexuality as a commodity. The profile of the clients
indicates a strong relationship between frequency of prostitution encounters and this measure
of "commodification." 6. Prostitution clients differ in the degree to which they regard sexuality
as a commodity. However, the responses of the majority of these men arrested for trying to
hire a street sex worker do not indicate commodification. The study included also the analysis
of the relationship between prostitution and violence by exploring the “rape myth
acceptance”7 presented in Table 3.
This attitude implicitly demonstrates a tendency of
violence against women. The response rates presented in Table 3 indicates that the arrested
clients do not reveal attitudes that validate the “rape myth acceptance”. Actually, none of the
selected variables had a high response rate for the ‘agreeing strongly’ end of the spectrum, in
spite of the fact that most of them scored highly on the “disagree strongly” end of the
spectrum.
6
For four of the five attitude measures-not having time for a conventional relationship, not wanting the responsibilities
associated with conventional relationships, wanting to be in control during sex, and the need to have sex immediately when
aroused-fewer than half of the respondents agreed strongly or somewhat. For the fifth attitude item-preference for prostitution
over conventional relationships-only 19.3% agreed somewhat or strongly. Monto (1999)
7
Rape myths are attitudes that have been shown to support sexual violence against women. Rape Myth Acceptance measure
as a dependent variable. Rape myths are "prejudicial, stereotyped, or false beliefs about rape, rape victims, and rapists" (Burt,
1980, p. 217) that serve to justify or support sexual violence against women and diminish support for rape victims. They include
the idea that women who are raped are in some way responsible for the violence against them, the idea that women often lie
about being raped for selfish reasons, and the idea that only sexually promiscuous women are raped.
11
Table 3 : “Rape myth acceptance”
Variables
Stuck-up woman deserve a lesson.
Strongly agree
Agree somewhat
Disagree somewhat
Strongly disagree
Women hitchhiking get what they
deserve.
Strongly agree
Agree somewhat
Disagree somewhat
Strongly disagree
Provocative dress asks for trouble.
Strongly agree
Agree somewhat
Disagree somewhat
Strongly disagree
Rape victims have bad reputation.
Strongly agree
Agree somewhat
Disagree somewhat
Strongly disagree
Forced sex after necking’s woman fault
Strongly agree
Agree somewhat
Disagree somewhat
Strongly disagree
Going to home implies willing to
Response Rates
2%
3%
7%
75%
3%
5%
12%
69%
6.3%
21%
25%
38%
4%
10%
23%
50%
5%
10%
20%
55%
have sex
Strongly agree
Agree somewhat
Disagree somewhat
Strongly disagree
4%
16%
25%
45%
3. Empirical specifications
3.1 Factor Analysis
The Monto data set contains a large number of variables. To see whether it was
possible to reduce the number of variables we are dealing with, we performed a factor
analysis on the reasons of arrested clients to approach sex workers with the purpose of
uncovering a possible latent structure of these variables in the data set. The scope of using
factor loadings is to discern the factor structure of the data.
The choice of the number of factors is based on the rule of thumb. The rule of thumb
consists on the number of eigenvalues of pattern/correlation matrix, which is the covariance
12
matrix of the standardized variables8. Eigenvalues for a certain factor measures the variance
in all the variables, which are grouped into that factor. The ratio of eigenvalues is the ratio of
explanatory importance of the factors with respect to the variables. A low eigenvalue poorly
explains the variance of the variable. Thus the correlation between indicators and factors is
characterized by large loadings above 0.5, moderate loadings between 0.3 and 0.5 and small
loading below 0.3. In our case we have considered only the first two loadings.
Over 80 % of the participants completed the questioners constructing a sample of
1342 observations. In the factor analysis, from 100 variables, we have excluded those
variables, which has a percentage of missing values more than 22% and demographic
variables. The process provided 6 factors, as the number of eigenvalues exceeding 1 is 6. The
factor loadings for these 6 factors can be seen in Table 4.
Table 4. The result of the factor analysis
Factors
Eigenvalues Variables
Factor1
0.5305
c80 "FORCED SEX AFTER NECKING'S WOMAN'S
'gender violent'
0.5462
FAULT"
0.5814
c81 "WOMEN HITCHHIKING DESERVE RAPE"
0.6778
c82 "STUCK-UP WOMEN DESERVE A LESSON"
0.5036
c83 "SEX FUN IF WOMAN FIGHTS"
0.6396
c84 "SOME WOMEN LIKE BEING SMACKED"
c85 "WANT SEX MORE WHEN ANGRY"
Factor2
-0.6296
c74 "PROSTITUTION CREATES PROBLEMS"
'favourable to
-0.6586
c75 "COPS SHOULD CRACK DOWN ON
prostitution'
0.7296
PROSTITUTION"
0.6644
c86 "PROSTITUTION NOT WRONG"
0.5323
c88 "SHOULD LEGALIZE PROSTITUTION"
c91 "SHOULD DECRIMINALIZE PROSTITUTION"
Factor3
0.5301
c72 "SEX WORKERS LIKE SEX MORE"
'prostitution ok for 0.4821
c73 "SEX WORKERS LIKE SEX ROUGHER
sex workers'
0.5765
c96 "SEX WORKERS ENJOY WORK"
0.5483
c99 "SEX WORKERS LIKE MEN"
Factor4
'Relationship
burden'
0.4988
0.7108
0.6952
c61 "PREFER PROSTITUTION TO RELATIONSHIP"
c63 "NO TIME FOR RELATIONSHIP"
c64 "DON'T WANT RELATIONSHIP
RESPONSIBILITIES"
Factor5
0.4599
c62 "EXITED BY APPROACHING SEX WORKERS"
'Variety
0.5134
c65 "LIKE TO HAVE A VARIETY OF PARTNERS"
preference'
0.4755
c68 "LIKE WOMAN WHO GETS NASTY"
Factor6
0.4833
C52 "SERIOUS TROUBLE WITH PARTNER"
'relationship
0.7355
c53 "SEPARATED FROM PARTNER"
trouble'
0.6250
c54 "BROKE UP WITH PARTNER"
Variables c29 c44-c100. Observations=735. The c-s refer to the coding in the Monto dataset,
see Appendix 1.
8
Every standardized variable has a variance of 1 and if we would define this variable to be a factor, it
accounts for a common variance of at least 1. Therefore the argument is that a common factor is only
substantially relevant it if explains a common variance of more than 1. In PCA, each component
explains a variance equal to the corresponding eigenvalue of the correlation matrix and hence relevant
components correspond to eigenvalue larger than one. Wansbeek (2000)
13
The first factor that we have named “gender violent” is a predictor of violent
sexuality. It indicates that one of the motivations when clients approach the sex workers is the
attraction to violence, which can be satisfied by buying sex with sex workers. This factor is
somehow an explanation of the “myth of rape” but we could observe from the statistics most
of the arrested clients do not exhibit attitudes that would support violence against women.
Thus the violence toward woman through sex workers at the hands of clients is due to a
minority of clients.
The second factor named 'favourable to prostitution' is way of supporting prostitution
which indicates a commodified prospective toward the sexuality with the sex workers. There
is an intensive debate in the literature regarding the sexual commodification. Thus for some of
the clients we find that frequency of visiting sex workers is motivated by the treatment of sex
as commodity. When this factor is used in the demand analysis, we take into account the fact
that the eigenvalues for c74 and c75 are both negative. Thus c74 would then read
“Prostitution does not create problem” and c75 would read “Cops should not crack down on
prostitution”.
The third factor is 'prostitution ok for sex workers'. This factor loading is also an
indicator of justifying sex commodification and avoids the intrinsic feeling of treating of sex
as commodity.
The fourth factor refers to “Peter Pan”, men who prefer to interact with individuals
who can respond to their needs without demanding intimate relationships. Factor five is a
representative of the view that prostitution forms part of sex consumption, and can for
example serve to satisfy those sexual appetites that the regular partner is unwilling to satisfy9.
The sixth factor “relationship trouble” indicates that some of the clients pay for sex
not only to have sex but to establish an intimate relationship with woman as a quick
alternative.
3.2 Empirical estimation and results
The factor analysis results in a reduction in factors (to 6) that can affect demand for sex. We
will use these factors as explanatory variables in the demand equations for paid sex, together
with personal characteristics such as working status, educations, age, and occupation. The
9
The desire to "have a variety of sexual partners'' and "be in control during sex," and the need to "have sex immediately when I
am aroused" all point to this kind of self-focused sexuality that Blanchard (1994) calls "McSex" in his popular expose on "young
johns." According to one man he interviewed 'lit's like going to McDonalds; most people are looking for a good quick cheap
meal. It's satisfying, it's greasy, and then you get the hell out of there.'' Paying for sex because of the desire to have sex with
women with particular physical attributes, a motivation described by McKeganey (1994), also reflects a conception of sex as a
commodity. Monto (1999)
14
dataset does not contain information regarding the level of earnings, and hence we use some
of the personal characteristics as explanatory variables to proxy the income level.
We will use two specification of the demand for prostitution. The first specification is
an ordered logit model with four categories of having sex with a prostitute.
Let y*n be person n’s demand for having sex with a prostitute during a year. Here this
demand is considered as a latent variable. Let xn be a vector of explanatory variables that
affect demand.  is a vector of unknown coefficients. Moreover let n be a random variable.
We then have the following demand function for having sex with a prostitute:
(1) y*n  xnn ; n  1, 2,,, N
Let ynj be the observation of how many times the clients have had sex with a
prostitute during a year, j=1,2,3,4, where j=1 means that the client has not been with a
prostitute before he was observed and arrested, j=2 means that the client has been with a
prostitute once before, j=3 mean that he has had sex with a prostitute more than once, but less
than once per month, and j=4 if the client has had sex with a prostitute more than once per
month. Thus the ordered structure of demand is given by
(2) ynj 
1 if client n belongs to category j; j  1,2,3,4
0 otherwise
Let j denote the threshold in the ordering of the demand, we then have
yn1  1 if y*n  1
(3)
yn2  1 if 1  y*n   2
yn3  1 if  2  y*n  3
yn4  1 if 3  y*n
The thresholds j must satisfy 1<2<3. From (1) and (3) we get
(4) P(ynj  1)  P( j1  y*n   j )  P( j1  x n   n   j  x n)
We will assume that n is iid with c.d.f . P(n u)=F(u). The n-s are assumed to be
logistic distributed, with the first moment of the distribution equal to zero and the second
moment equal to 2/3. Thus
(5) F(u) 
1
1  e u
15
Now we can rewrite (4) to yield
(6) P(ynj  1)  F( j  x n)  F( j1  x n)
and where the distribution function F(.) is given in (5).
4
Note that
[P(y
j1
nj
 1)]  1so that P(yn 4  1)  1  F(3  x n)
The likelihood function of data is:
N
4
(7) L(, ) =  F( j - x n) - F( j-1 - x n) 
n=1 j=1

ynj

The coefficient vectors can then be estimated by maximizing this likelihood (or rather
the log likelihood). The estimates are given in Table 5, second column.
The first and the second category in the ordered demand structure is particular and
cover those who are the “first time approaching”, who we name “experimenters”. The third
and forth category cover the repeated frequenters, who we name “regulars”.
In Table 6 we give the estimates of the marginal effects of the ordered logit. From (6)
we get that
P(ynj  1)  F( j1  x n) F( j  x n) 


  ;for j  1,2,3,4
x n
x n
x n


(8)
From (5) and (8) we then have
P(y n1  1)
 F(1  x n)[1  F(1  x n)]
x n
(9)
P(y n 2  1)
 {F(1  x n)[1  F(1  x n)]  F( 2  x n)[1  F( 2  x n)]}
x n
P(y n3  1)
 {F( 2  x n)[1  F( 2  x n)]  F( 3  x n)[1  F( 3  x n)]}
x n
P(y n 4  1)
 {F( 3  x n)[1  F( 3  x n)]}
x n
Except for the first and last marginal effects, the terms in braces can have be positive or
negative.
In the second specification of demand we model the probability of being a “regular”
client. Let Unj be the utility for client n of being j-type of client. When j=1, the client is a
“regular” client and when j=0 he is an “experimenter”. We will assume that Unj is given by
16
(10) Unj  xn  j  nj ; j  0,1; n  1, 2,,, N
The vector xn is the same as in the ordered logit presented above, expect that it
includes 1 to allow for a constant, and  j is vector of alternative specific coefficients. By
assuming that nj is extreme value distributed (the double exponential distribution) with zero
expectation and a constant variance, and by assuming utility maximization we get the
following probability for being a “regular” customer:
K
(11) P(U n1  U n0 ) 
exp( 1k x nk )
k 0
K
K
k 0
k 0
K

exp(  k x nk )
k 0
K
exp(  0k x nk )  exp(  1k x nk ) 1  exp(  k x nk )
k 0
where
 k  1k   0k , and x n0  1.
Let yn1=1 if the individual has chosen to be a regular customer, and equal to zero
otherwise, and let n1(kkxnk ) be the choice probability in (9). Then the likelihood of the
data,
N
K
K
n 1
k 0
k 0
(12) L()  [n1 (   k x n )]yn1 [1  n1 (   k x n )]1 yn1
The coefficients k, k=0,1,,,,K are estimated by maximizing this likelihood (or rather the loglikelihood). The estimates are given in column 3 of Table 5.
Apart from the demand for prostitution we also estimate the demand for condom use
in order to analyze the peculiarity of clients behaviour with respect to this aspect. It would be
important to analyse how the determinants of the demand for sex with sex workers respond
with respect to the demand for condom use. Condom use is almost always negotiated directly
between the interested client and the street sex workers. Therefore, the client who requires the
use of condoms, signals that he has a more risk adverse attitude to approach the sex workers
for sex. The choice probability of using condom follows from a similar utility maximizing
procedure, with an additive random utility model, as the one that led to the likelihood in (10).
Estimates are given in the fourth column of Table 5. Details about the coding of the variables
are given in the Appendix.
.
17
Table 5: Estimation results
Variables
Demand for sex:
Ordered Logit
Probability of using
condom
0.160
(0.194)
0.655**
(0.281)
0.491***
(0.186)
-0.125
(0.170)
-0.312*
(0.173)
0.276***
(0.096)
0.017*
(0.009)
0.181*
(0.108)
Demand for sex:
Probability of being a
“regular” client
0.067
(0.243)
0.656*
(0.347)
0.201
(0.226)
-0.151
(0.209)
-0.118
(0.213)
0.220*
(0.118)
0.030***
(0.011)
0.274**
(0.136)
Education
Factor2
' favourable to
prostitution'
Factor3
'prostitution ok for
sex workers'
-0.159*
(0.094)
-0.199*
(0.112)
-0.400*
(0.222)
0.198**
(0.101)
0.200*
(0.124)
-0.102
(0.242)
Factor4
'relationship burden'
Factor5
' variety preference'
Factor6
' broken relationship '
Threshold 1
-0.536***
(0.112)
-0.968***
(0.121)
-0.026
(0.109)
0.788
(0.550)
2.233***
(0.559)
4.452***
(0.580)
-0.641***
(0.137)
-1.031***
(0.151)
0.006
(0.137)
-0.351
(0.266)
0.692***
(0.281)
0.482*
(0.293)
-2.501***
(0.692)
582
3.643***
(1.339)
570
Work status
Race
Job
Marriage
Control dislike
Age
Factor1
'gender violent'
Threshold 2
Threshold 3
Constant
# of observations
582
0.067
(0.474)
0.476
(0.564)
1.121**
(0.576)
-0.023
(0.415)
0.090
(0.412)
-0.062
(0.234)
-0.031
(0.020)
0.464*
(0.259)
Mcfaddens rho
0.14
0.18
0.71
Standard errors in parentheses. (Blank: Not significant. ***:Significant at 1%,
**: Significant at 5%, *:Significant 10%)
First we note that the threshold levels are increasing with the level of sex trips. The ordered
logit results imply that demand for having sex with prostitutes, in terms of frequency per year,
is the same across education levels, it is higher among full-time worker than individuals
18
working less hours, non-white individuals have more sex with prostitutes than white
individuals. Married individuals demand less sex with prostitutes than non-married, and the
more the individuals dislike having control when having sex, the more they have sex with
prostitutes. Sex with prostitutes is increasing with the age of the client.
When it comes to the interpretation of the factors we should keep in mind that the
more they disagree with the statement the higher is the score. Thus, the more the clients
dislike gender violence, the more they demand sex; the more favourable they are to
prostitution, the more they have sex with prostitutes, and the less they think that prostitution is
ok for the sex workers, the more they have sex with prostitutes! Furthermore the more they
have problems with having stable relationships, the more they demand sex and finally the
more they like variety in sex life, the more they demand sex with prostitutes.
The results are somewhat mixed compared to prior expectations, but as demonstrated
in Table 6, the overall results for the ordered logit in Table 5 shadow for differences in
behaviour across individuals with little experience with prostitutes (named “experimenters”)
and those with more experience (named “regulars). In Table 6 we distinguish between four
groups of clients. The first two are those who have never had sex with prostitutes before or
had sex once before. Clients in these groups we name “experimenters”. Clients in the two last
groups are named “regulars” because they have had sex with prostitutes at least more than one
time, but less than once a month (3rd group) or 1 to 3 times a month (4th group). Table 6 gives
the impact on demand for sex with prostitutes of marginal changes in the explanatory
variables, called marginal effects, within each group. The interesting result that emerges is
that the marginal effects imply that the “experimenters’ and “regulars” demand for sex with
prostitutes have opposing characteristics.
The “experimenters” demand more sex with prostitutes the less they work, more if they
are white opposed to non-white, more if they are married compared to not married, more the
younger they are and more the more they like to have control when having sex. The
“regulars” demand characteristics are quite the opposite, with a minor exception with regard
to the preference for having control (group 4). Even more interesting is the fact that the
impacts of the Factors 1-6 on demand for sex also are opposite. The experimenters demand
more sex with prostitutes the more gender violent they are, the less they are favourable to
prostitution, the more they think that the sex workers enjoy having sex, the less problem they
have in having a stable relationship with their regular partner, and the less like variety in their
sex life. For the “regulars” all of these effects are reversed.
Thus the ”experimenters” seem to be of the macho type, while the regulars are
typically loners seeking sex with prostitutes as a substitute for not having a relationship of a
more normal sort.
19
In Table 5 we also give the estimates of the probability of being a “regular” client
as opposed to being an “experimenter”. Comparing these results with the marginal effects for
the “regulars” derived from the ordered logit given in Table 6, we observe that the results are
quite similar, which is a further confirmation of the conclusions drawn above.
In Table 5 we also report the estimates from the use of condoms. Concentrating
on the significant parameters we note that the probability of using condoms is higher among
the non-white compared to the white individuals. The probability of using condoms is higher
among those who are opposed to gender violence relative to those who are not, and the
probability is higher the more you favour prostitution and the less you like variety in your sex
life. It is also interesting to note that among those with a broken relationship the probability of
using condom is higher than among those with a functioning relationship. Thus the users of
condoms seem more to fit the description of the ‘regulars” than the “experimenters” and
hence the non-use of condoms seem to be risk loving machos.
20
:
Table 6: Marginal effects in the ordered logit
Variables
Never with Once with
More than 1 time
prostitutes
but less then once
prostitutes
1 to 3 times per
month
per month
Education
Work status
Race
Job
Marriage
Control Dislike
Age
Factor1
'gender violent'
Factor2
'favourable to
prostitution'
Factor3
'prostitution ok
for prostitutes'
Factor4
'Realtionship
burden'
Factor5
'Variety
preference'
Factor6
'relationship
trouble'
-0.0269ˆ
(0.033)
-0.123**
(0.059)
-0.077***
(0.028)
0.02ˆ
(0.028)
0.051*
(0.0287)
-0.045***
(0.016)
-0.002**
(0.002)
-0.029*
(0.018)
-0.012ˆ
(0.014)
-0.033***
(0.008)
-0.044**
(0.018)
0.01ˆ
(0.014)
0.026*
(0.015)
-0.023***
(0.008)
-0.001*
(0.0008)
-0.015*
(0.0094)
0.027ˆ
(0.033)
0.113**
(0.048)
0.079***
(0.029)
-0.02ˆ
(0.028)
-0.052*
(0.029)
0.046***
(0.017)
0.002*
(0.0015)
0.030*
(0.018)
0.012ˆ
(0.014)
0.0429***
(0.015)
0.0425**
(0.017)
-0.010ˆ
(0.013)
-0.025*
(0.014)
-0.022***
(0.008)
0.001*
(0.0007)
0.014*
(0.0088)
0.026*
(0.015)
0.013*
(0.0083)
-0.026*
(0.015)
-0.012*
(0.0077)
-0.032**
(0.016)
-0.016*
(0.009 )
0.033**
(0.0172)
0.016*
(0.0083)
0.088***
(0.0186)
0.045***
(0.011)
-0.09***
(0.020)
-0.043***
(0.009)
0.159***
(0.02)
0.085***
(0.015)
-0.162***
(0.024)
-0.078***
(0.012)
0.004ˆ
(0.017)
0.002ˆ
(0.009)
-0.004ˆ
(0.018)
-0.002ˆ
(0.008)
Standard errors in brackets. ( ˆ non significant, * significant at (10%), **( 5%), ***( 1%)).
4. Implications and conclusions
…
21
References
Aggleton Peter Ed. (1998), Men who sell sex London: UCL Press.
Ben-Israel, Hanny and Numi Levenkron (2005)The Missing Factor: Clients of Trafficked
Women in Israel’s Sex Industry, The Hebrew University in Jerusalem: Top Print.
Boyle, F.M., Glennon, S., Najman, J.M., Turrell, G., Western, J.S. and Wood, C. (1997). The
sex industry: A survey of sex workers in Queensland, Australia Ashgate: Aldershot.
Cameron S The economics of Sin 2002
Cameron S., Collins A , Thew N, (1999), “Prostitution services: an exploratory empirical
analysis”, Applied Economics, 31, pg 1523-1529.
Chapkis, Wendy (1997) Live Sex Acts: women performing erotic labour Routledge: New
York.
Edlund L. and E. Korn (2002), “An Economic Theory of Prostitution”, Journal of Political
Economy, Vol 110, No 1.
Garofalo Giulia (2002) “Towards an economic theory of prostitution”, mimeo.
Gertler et al (2003) “Risky Business: the Market for unprotected commercial sex”, mimeo
Cornell University.
Lim, Lin Lean (1998) (ed.), The sex sector: the economic and social bases of prostitution in
Southeast Asia. Geneva: International Labour Office.
Marttila, A. (2003) ‘Consuming Sex: Finnish Male Clients and Russian and Baltic
Prostitution’ Presented at Gender and Power in the New Europe, the 5th European Feminist
Research Conference August 20-24, 2003 Lund University, Sweden.
May, Tiggey, Alex Harocopos, and Michael Hough (2000) For love or money: pimps and the
management of sex work London: Home Office, Policing and Reducing Crime Unit.
McKigney, Neil and Marina Barnard (1996) Sex work on the streets: sex workers and their
clients Buckingham: Open University Press.
Moffatt P.G., Peters S.A., “The pricing of Personal Services”, mimeo, Royal Economic
Society in Durham, Apr. 2001.
Monto, Martin A. (1999) Clients of street sex workers in Portland, Oregon, San Francisco and
Santa Clara, California, and Las Vegas, Nevada 1996-1999. University of Portland. Ann
Arbor: MI: Inter-university Consortium for Political and Social Research.
Nussbaum, Martha (1998), “Whether from reason or prejudice: taking money for bodily
services”, Journal of legal studies XXVII, n.2, part2.
Nelson J.A. and England P. (2002), “Feminist philosophies of Love and Work”, forthcoming
in Hypathia 17(2).
O’Connell Davidson J. (2002), “The rights and wrongs of prostitution” Hypathia 17(2)
Spring.
O’Connell Davidson J.(1998), Prostitution Power and Freedom, Polity Press, 1998.
Perkins R. (1991), Working Girls, Sex workers, their Life and Social Control Australian
Institute of Criminology.
Pitts, M.K., Smith, A.M.A., O’Brien, M. and S. Misson (2004) Who pays for sex and why?
An analysis of social and motivational factors associated with male clients of sex workers
Archives of Sexual Behaviour, 33 (4): 353-358.
Rao, V, Gupta I, Lokshin M., Jana J., “ Sex Workers and the Cost of Safe Sex: The
compensating Differential for Condom Use in Calcutta”, mimeo Jan 2001.
Reynolds, H. 1986, The Economics of Prostitution, Thomas, Illinois, USA.
Rodriguez L. Guven-Lisaniler F. and Ugural S., “Sex work and State Regulations in North
Cyprus” mimeo, presented at EAEPE Conference, Siena 2001.
Sanchez Taylor J (2001) Dollars are girl's best friend? Female tourists' sexual behaviour in the
Caribbean 749-760 Sociology: Identity politics in the workplace Vol 35, num.3.
Sanders, T. (2005) It’s just acting: sex workers’ strategies for capitalizing on sexuality
Gender, work and organization 12 (4): 319-324.
Sharpe, Karen (1998) Red light, blue light: sex workers, punters and the police Aldershot:
Ashgate.
22
Appendix
Dependent variables:
Version 1 – Dependent variable: frequency of demand for sex workers during last year by an
increasing order. The missing values and not applicable once are 21.5%.
The dependent variable (NPROS) takes values from 1 to 4.
1- never with the sex worker
2- once with the sex worker
3- more than 1 time but less then once per month
4- 1 to 3 times per month
Version 2- Refers to the estimation of the demand for sex workers if the dependent variable is
constructed as a dichotomous, taking value 1 if he demanded more than once sex worker
services last year and it takes value 0 if he replied none or once. This is useful to make a
distinction between those who are the first time users and the repeated once.
Version 3- Dependent variable is constructed as a dichotomous, taking value 1 if he uses
condoms and it takes value 0 if he never uses it.
Explanatory variables:
EDU1 = level of education. The explanatory variable of education, used in the estimation,
takes values 1 if the respondent has some college and more. Otherwise it takes value 0. The
missing values are at 1 %.
WORK = current work status is an independent variable that takes value 1 if working full
time and 0 otherwise. The %age of missing values is at 3%.
RACE = Race variable takes value 1 if black and other races, the white people takes value 0.
The missing values are at 2%.
JOB = type of the job held by the respondent. The explanatory variable (JOB) take value 1 if
the respondents held a job as executives/business managers and administration personnel, 0
otherwise. The missing values are at 17%.
MARR = marital status takes value 1 if the respondent is married and 0 otherwise. The
missing values are at 1%.
CONTROL = like control during sex, taking values from 1 to 4.
1- agree strongly
2- agree somewhat
3- disagree somewhat
4- disagree strongly
The missing values are at 8.2 %.
AGE= it refers to the age of the respondents, is continuous, with the minimu age 18 and the
maximum age 84. The missing values are at 7%.
F1 refers to factor 1 = gender violent , see the main variables are C80-C85:
c80 "FORCED SEX AFTER NECKING'S WOMAN'S FAULT"
23
c81 "WOMEN HITCHHIKING DESERVE RAPE"
c82 "STUCK-UP WOMEN DESERVE A LESSON"
c83 "SEX FUN IF WOMAN FIGHTS"
c84 "SOME WOMEN LIKE BEING SMACKED"
c85 "WANT SEX MORE WHEN ANGRY"
The variables c80-c85 rank from 1 to 4, respectively agree strongly up to diasgree strongly.
The missing values are at 10% in all variables c80-c85.
F2 refers to factor 2, favourable to prostitution, main variables are C74, C75, C86, C88,C91.
c74 "PROSTITUTION CREATES PROBLEMS" ( missing values at 8%)
c75 "COPS SHOULD CRACK DOWN ON PROSTITUTION" (missing values at 9%)
c86 "PROSTITUTION NOT WRONG" (missing values at 15%)
c88 "SHOULD LEGALIZE PROSTITUTION" (missing values at 15%)
c91 "SHOULD DECRIMINALIZE PROSTITUTION" (missing values at 15%)
The variables rank from 1 to 4, respectively agree strongly up to disagree strongly.
F3 refers to factor 3, prostitution ok for sex workers. Table 1 and the main variables are C72,
C73, C96, C99.
c72 "SEX WORKERS LIKE SEX MORE" ( missing values at 8%)
c73 "SEX WORKERS LIKE SEX ROUGHER ( missing values at 10%)
c96 "SEX WORKERS ENJOY WORK"( missing values at 17%)
c99 "SEX WORKERS LIKE MEN" ( missing values at 19%)
The variables rank from 1 to 4, respectively agree strongly up to disagree strongly.
F4 refers to factor 4, relationship burden, Table 1 and the main variables are C61, C63, C64.
c61 "PREFER PROSTITUTION TO RELATIONSHIP" ( missing values at 7.3 %)
c63 "NO TIME FOR RELATIONSHIP" ( missing values at 7.6%)
c64 "DON'T WANT RELATIONSHIP RESPONSIBILITIES" ( missing values at 8%)
The variables rank from 1 to 4, respectively agree strongly up to disagree strongly.
F5 refers to factor 5, variety preference Table 1 and the main variables are C62, C65, C68.
c65 "LIKE TO HAVE A VARIETY OF PARTNERS" ( missing values at 7.3 %)
c68 "LIKE WOMAN WHO GETS NASTY" ( missing values at 8.3 %)
The variables rank from 1 to 4, respectively agree strongly up to disagree strongly.
F6 refers to factor 6, Table 1 and the main variables are C52, c53, c54.
c53 "SEPARATED FROM PARTNER" ( missing values at 6 %)
c54 "BROKE UP WITH PARTNER" ( missing values at 7.6 %)
The variables takes value 1 if yes and 0 if no.
24