Asking About “Prostitution”, “Sex Work” and “Transactional Sex”: Question Wording and Attitudes Toward Trading Sexual Services

ABSTRACT

This study explored the impact of question wording on attitudes toward trading sexual services. There are no previous research agendas investigating how views on the trade in sex are susceptible to question word choice. We utilized an original survey to assess how three different concepts used to represent the exchange of sexual services impact on the reported level of acceptability among respondents. The concepts we selected were “prostitution,” “sex work”, and “transactional sex.” We also explored the contrasting effects of predictor variables based on question wording, such as general sexual attitudes and positive and negative associations with trading sex. We found that attitudes toward the trading of sex is partially a function of question wording. Respondents were more likely to translate negative associations with the exchange of sexual services into unacceptability when assigned a question using the term prostitution. The results indicated that prostitution is a stigmatized word, resonating more negatively with respondents as compared to sex work or transactional sex. When asked about transactional sex, respondents were less likely to translate an attitude into an assessment on acceptability, indicating that the term is vague. Sex work elicited more neutral responses on acceptability. The results suggest that sex work would be the most useful term to use in dialogue and debate regarding the trade in sexual services. Sex work does not appear to be imbued with the same sense of negativity as prostitution and the meaning of sex work seems clearer than transactional sex.

In this article, we explore how question wording impacts the public’s views on the acceptability of trading sexual services. In contemporary democracies, elections and referenda serve as the primary outlet for citizens to express their desires regarding the direction of laws and public policy. Beyond elections and referenda, and due to the relative infrequency of these events, politicians and political parties routinely rely on public opinion measures to guide governmental action. The fact that governments utilize public opinion as a barometer to direct policy action signifies an importance in confirming that the survey instruments utilized to gauge public opinion create valid measures. A longstanding survey research agenda demonstrates that the wording of questions can have quite large effects on support for a range of governmental policies (Bishop et al., 1978; Huber & Paris, 2013; Rasinski, 1989; Smith, 1987). In addition, studies find that question wording can impact a range of attitudes and opinions, such as belief in climate change (Schuldt et al., 2011), reported political partisanship (Sanders et al., 2002), and support for issues such as assisted suicide (Magelssen et al., 2016).

In this study, we extended this research agenda by exploring the impact of question wording on the acceptability of exchanging or trading sex for money – we use these terms interchangeably to refer to the actions of selling and paying for sexual services. People’s attitudes toward the trade in sex is a timely topic of inquiry. Over the past two decades, digital developments have facilitated the access of sex work-related content and services while also making the diversity of sex work apparent (Abel, 2021; Campbell et al., 2018; Jones, 2015; Nelson et al., 2020; Sanders et al., 2016, 2018). Online spaces are also used by sex workers’ organizations to get their political messages across. Meanwhile, concerns over issues such as women’s exploitation and human trafficking in the sex trade have informed policy debates and decision-making, resulting in a range of repressive measures targeting sex trade participants (Majic, 2020). How these conflicting trends resonate with the general public’s perceptions about the exchange of sexual services for money is, however, a question that requires further attention.

A great deal of academic and public debate exists discussing the advantages and disadvantages of using one concept over another when describing the trade of sexual services. Sex workers’ rights organizations commonly argue that “sex work” is the correct term to use when describing this activity due to the historical stigma and negative bias attached to the word “prostitution” (ICRSE, 2021; NSWP, 2017), which has long been associated with dishonorable and shameful sexual behavior (Benoit et al., 2019). Sex work is seen as a preferable alternative as it captures the labor involved in selling sex, as well the agency of the people concerned (Leigh, 2010). On the other hand, policing agencies and legal texts continue to refer to the practice as prostitution (Altemimei, 2013; Polismyndigheten, 2019). Another term is “transactional sex,” which has been used as an alternative in the context of social work and health care (McMillan et al., 2018). We implemented a study that investigated how question wording impacts views on the trade in sexual services. Further, we explored how predictor variables have contrasting effects based on question wording.

Our analysis utilized an original survey where respondents were randomly assigned a question wording variant asking about the acceptability of trading sexual services. In particular, the question asked “To what extent do you think that [_____________] is acceptable?”. The respondent was provided one of three different concepts that are commonly used to represent the trade in sexual services: 1.) prostitution, 2.) sex work, and 3.) transactional sex. Then, we compared the distribution of responses for the three different questions. Finally, we estimated three models that predicted the level of acceptability for each wording variant.

There were two research questions: 1.) how does question wording impact the distribution of responses on the acceptability of the exchange of sexual services? 2.) how does the relationship between independent variables and predicting acceptability differ based on question wording? We uncovered two overarching findings. First, there were important differences in views on the acceptability of trading sexual services based on question wording. Respondents were more likely to indicate “0 = not at all acceptable” when asked about prostitution as compared to those asked about sex work or transactional sex. Further, respondents selected “5 = neither acceptable nor unacceptable” at a statistically higher rate when asked about sex work in comparison to being asked about prostitution or transactional sex. The results provide support for the idea that prostitution is a stigmatized word, as well as the argument that sex work is not imbued with the same sense of negativity.

Second, respondents were more likely to translate other attitudes, and positive and negative associations with the causes and consequences of the sex trade, into an evaluation of the acceptability of prostitution when compared to being asked about sex work or transactional sex. For the acceptability of prostitution, liberal attitudes toward general sexual behavior, positive associations with trading sexual services, and negative associations were highly predictive. In comparison, negative associations with trading sexual services had no relationship with evaluations of the acceptability of sex work. Finally, neither positive nor negative associations had any relationship with predicting acceptability of transactional sex, and liberal attitudes toward general sexual behavior had only a small impact. Taken together, the results indicate that sex work and transactional sex are not as negatively biased as prostitution, and that use of the word prostitution might elicit negative evaluations based on views being highly influenced by other attitudes and associations.

Question Wording and Attitude Formation

The literature on survey research demonstrates that question wording can have a significant and substantive impact when measuring support for government policies (Bishop et al., 1978; Huber & Paris, 2013; Rasinski, 1989; Smith, 1987), partisanship and vote choice (Box-Steffensmeier et al., 2000; Sanders et al., 2002), generalized trust (Lundmark et al., 2016), belief in climate change (Schuldt et al., 2011), belief in conspiracy theories (Krosnick et al., 2014), and support for controversial topics (Magelssen et al., 2016). In addition, studies even find that the validity of eyewitness testimony (Loftus & Zanni, 1975), as well as self-reported victimization (Fisher, 2009), is impacted by question wording.

Some of the earliest research on question wording explored how questions that asked about governmental programs helping people at lower income levels could elicit contrasting results based on the words utilized to represent this group of people. For instance, Smith (1987) found that support for such initiatives was substantially higher when survey questions contain the phrase “poor people” in comparison to mentioning “people on welfare.” Rasinski (1989) uncovered a similar finding, but also demonstrated that additional question wording choices can impact a range of support for funding programs. In particular, they found differences in endorsing funding programs when comparing phrases such as “halting rising crime” vs. “law enforcement,” “dealing with drug addiction” vs. “drug rehabilitation,” “assistance to big cities” vs. “solving problems of big cities,” “assistance to blacks” vs. “improving conditions of blacks,” and “social security” vs. “protecting social security.” More recently, Huber and Paris (2013) confirmed these findings and demonstrated that the differences in support based on question wording were a function of respondents associating different programs with the phrases. Thus, when developing responses to questions, respondents rely on different heuristics based on the question wording in order to form their opinion. When a respondent is asked about “welfare” there is an immediate negative reaction to the word based on these heuristics, and it is more likely that a negative attitude is formed and expressed. In our study, we expected a similar reaction when respondents were asked about prostitution when compared to sex work and/or transactional sex. These expectations are formalized in the hypothesis section.

To further convey the importance of question wording, it is worth discussing studies that demonstrate that eyewitness testimony and even reports of victimization are impacted survey instruments. An earlier experiment by Loftus and Zanni (1975) had participants watch a video of a car accident and then immediately answer questions about the event that occurred in the film. One question the authors asked the participants was about seeing a broken headlight. The authors provided two variants of the question where they altered the article in the question wording. When using a definitive article (e.g., did you see the broken headlight?) there were fewer “don’t know” responses and a greater number of participants indicating witnessing events that had not taken place when compared to using an indefinite article (e.g., did you see a broken headlight?). The result is striking since the questionnaire was given immediately at the end of the video.

The impact of question wording is not limited to altering responses about events witnessed by the individual. Fisher (2009) found that question wording can impact respondent reports regarding their own experiences of being victims of crime. The author explored two nationally representative studies that asked college women about their experience with rape and assault: 1.) National College Women Sexual Victimization Study (NCWVS), and 2.) National Violence Against College Women Study (NVACW). The NCWVS gives behaviorally specific examples within their questions, while the NVACW did not give examples and the questions used legal technical wording. The percentage of respondents reporting personal victimization was larger for the NCWVS. The author argued that the specific scenarios allowed respondents to reflect on particular situations, which led to a higher victimization reporting rate. Given the serious nature of the victimization, the fact that question wording can have such a large effect on reporting indicates a need to regularly evaluate the impact of question wording on responses for survey questions.

Prostitution, Sex Work, and Transactional Sex

There exist only a few studies that set out to predict individual-level attitudes toward the trade in sexual services over the last thirty or so years (Basow & Campanile, 1990; Cosby et al., 1996; Hansen & Johansson, 2022; Jakosson & Kotsadam, 2011; May, 1999; Peracca et al., 1998; Räsaänen & Wilska, 2007; Valor-Segura et al., 2011; Vlase & Grasso, 2021; Yan et al., 2018). With the exception of Jakosson and Kotsadam (2011), who asked about buying and selling sex, these previous studies specifically used the word prostitution in their survey questions. In fact, Hansen and Johansson (2022) recognized that, “a survey question using the term ‘prostitution’ may elicit different responses than a question using the phrase ‘transactional sex’ or ‘sex work’” (p. 95). The idea behind this observation is that people’s attitudes are affected by the way the activity is labeled. Thus, there is a gap in the literature that needs to be filled exploring question wording.

Scholars in several fields have now abandoned the word prostitution due to its negative connotations (McMillan et al., 2018). Sanders et al. (2018) detailed an expansive history of the word showing that although the meaning of prostitution has changed over time it has generally been used in a negative sense. Similarly, Scoular (2015) highlighted how the term prostitution has historically been utilized in the context of defining morality and sin. In many eras, the “prostitute” has been seen as the antithesis of decency and the framework of prostitution has been deployed to condemn, control and marginalize certain groups of people, women especially (Agustín, 2007, 2008). Benoit et al. (2019) described how calls for actions about the so-called “prostitution problem” has risen “to the level of high politics during historical periods marked by global capitalism, international migration, and tensions related to entrenched gender, class, and race inequalities within and across nations” (p. 1905).

Bernstein (1999) argued that the word prostitution is generally selected with the intent to elicit thoughts of exploitation. Agustín (2007; 2008) and García (2012) identified the nineteenth century as an important definitional turning point in this regard. Around this time, influential social agents started considering prostitution as an immoral social evil from which women needed to be rescued and the role of the victim was assigned to those who sold sex. The campaign against the “traffic in women” from the early 1900s relied on this notion of victimhood, which continues to influence anti-trafficking measures to this day.

Prostitution continues to be the preferred term in the anti-trafficking context, where it is used in combination with concepts such as “sexual exploitation,” “sexual slavery” and “sex trafficking” in such a way that the lines between coercion and consent become unclear. In many instances, the consent of the people who sell sex is even regarded as irrelevant (Agustín, 2007, 2008; García, 2012). While the concept of “human trafficking” nowadays encompasses a variety of criminal acts that have nothing to do with the sex trade, the conflation between prostitution and human trafficking persists (Wagenaar & Altink, 2012; Wagenaar et al., 2017). The association between prostitution, exploitation and human trafficking is likely to resonate with the public, as stories about innocent girls and women who are forced into prostitution by unscrupulous traffickers have become a common feature in popular culture and news media (Johnston et al., 2014; Kinney, 2015; Snajdr, 2013; Szörényi & Eate, 2014).

Moreover, scholars who view paid sexual encounters as an intrinsic part of a patriarchal institution that allows men to exercise their ownership over women’s bodies often opt for the term prostitution (Pateman, 1999; Raymond, 1998). Some even criticize terms like sex work and commercial sex as enabling the exploitation of women (Niemi, 2010). The notion of prostitution as men’s violence against women also appears in contemporary policy debates regarding the sex trade, and policies that aim to abolish the sex trade have been justified under the auspices of this perspective (Johansson & Östergren, 2021; Östergren, 2018, 2020). In general, the term prostitution is deployed by groups and governments that lobby for repressive prostitution policies, like client-criminalization, the goal of which is to further chastise those who engage in paid sex (Weitzer, 2010; Östergren, 2018, 2020).

While the idea of the prostitute as a disreputable deviant has now largely been replaced by the idea that individuals who sell sex are exploited victims, the negative sense of prostitution remains. Beyond these negative connotations, the term prostitution is more commonly associated with a particular segment of the sex trade, mainly street-based sex workers who are not only among the most visible but also the most precarious and exposed (Armstrong, 2016; Monroe, 2005; Overall, 1992). Thus, there are two potential issues with use of the word prostitution in surveys. First, there is an inherent negative connotation with the word. Second, in the mind of survey respondents, when asked about prostitution they might be evaluating only a very narrow segment of the sex trade.

Overall, sex work and transactional sex are widely recognized by scholars as more encompassing and less stigmatizing terms (Bernstein, 1999; Overall, 1992; Sanders et al., 2018; Scoular, 2015). Sex work is used to describe a range of activities, from the in person “full service” of sexual intercourse, to striptease and online services like webcamming shows (Sanders et al., 2018). Bernstein (1999) pointed to the positive connotations of the term sex work, in that it implies economic empowerment and the expression of sexual freedom. McMillan et al. (2018) provided one of the most thorough discussions on word choice. The authors argued that sex work is the most appropriate and useful term because it positions paid sex “as a matter of labor and not culture or morality” (p. 1518). They pointed out that while prostitution has negative connotations of criminality and immorality, sex work signals that paid sex can be legitimate work. Sex worker activists have staunchly campaigned for using the term sex work when talking about adults who consensually exchange sexual services for money since the late 1970s-early 1980s (Leigh, 2010). However, the burden of the prostitution label and structural stigma have left sex workers with extremely limited leverage (Bruckert & Hannem, 2013), making it difficult to have any influence over framing. In general, sex worker advocacy groups have struggled to find support for their efforts to de-stigmatize sex work, and they have generally been excluded from the policy process (Bruckert & Hannem, 2013; Wagenaar & Altink, 2012; Wagenaar et al., 2017).

While transactional sex does not carry the same negative connotations as prostitution, nor the same charge as sex work, McMillan et al. (2018) believed that the term is too vague. It has also been used inconsistently, as both distinct from and synonymous with sex work and prostitution. The authors stated, “to a certain extent, all consensual sex is transactional in nature” (1520), calling into question the usefulness of transactional sex as opposed to sex work. Against this backdrop, we would expect questions using the word prostitution to elicit negative reactions, as well as a strong reliance on other attitudes in order to provide an assessment. In comparison, we would expect that sex work was more likely to result in assessments less impacted by negative associations. Finally, aligned with McMillan et al.’s (2018) argument, it was expected that predictor attitudes would be less likely to be translated into an assessment when asked about transactional sex due to the vagueness of the term.

Attitudes Toward the Exchange of Sexual Services

As stated, there exists only a handful of studies seeking to predict individual level attitudes toward the trade in sexual services, and the literature is more characterized by debates than trends. Since our study focused on differences in results based on question wording, we briefly summarize previous findings on predicting attitudes. In terms of sociodemographic trends, studies have found that age has a positive relationship with support for the exchange of sexual services (Cosby et al., 1996; Hansen & Johansson, 2022; May, 1999). Previous research has found both positive (Jakosson & Kotsadam, 2011; May, 1999; Vlase & Grasso, 2021) and negative (Hansen & Johansson, 2022) relationships between education and support for the trade in sexual services. One clear finding in the literature is that women are more likely than men to have negative views on the exchange of sexual services (Cosby et al., 1996; Cotton et al., 2002; Hansen & Johansson, 2022; May, 1999). While we included a range of sociodemographic variables in the regression models, we did not have any expectations regarding sociodemographic variables operating differently across the models.

Studies show a range of attitudes being correlated with positive views on trading sexual services. Several studies confirm that a liberal ideology and liberal attitudes translate into more positive evaluations of the sex trade (Cosby et al., 1996; Hansen & Johansson, 2022; May, 1999; Peracca et al., 1998; Valor-Segura et al., 2011; Vlase & Grasso, 2021). For instance, Valor-Segura et al. (2011) found that a greater commitment to gender equality was related to positive views on prostitution. In addition, Hansen and Johansson (2022) demonstrated that general attitudes toward sexual behavior strongly translated into views on the acceptability of prostitution. Finally, one implication of the existing research that has not yet been tested is how positive and negative associations with the trade in sexual services will impact an individual’s views. In the regression models, we included a range of attitudinal variables, as well as positive and negative associations. Our expectation was that since prostitution is a well-known and highly stigmatized word, attitudinal variables will be more strongly related to predicting acceptability of prostitution when compared to sex work and transactional sex. For instance, we expected negative association with the causes and consequences of trading sexual services to have a stronger relationship with predicting the acceptability of prostitution when compared to the other two question variants. We formalized these hypotheses as:

H₁:

Respondents will be less likely to select the neutral choice, and more likely to select “not at all acceptable,” when asked about the acceptability of prostitution when compared to being asked about sex work or transactional sex.

H₂:

Models predicting the acceptability of prostitution will perform better than models predicting the acceptability of sex work and transactional sex.

H₃:

Independent variables will have a larger effect on predicting respondents’ views on the acceptability of prostitution when compared to predicting the acceptability of sex work or transactional sex.

Data

To study the impact of question wording on predicting respondent assessment of the acceptability of trading sexual services, we conducted an original survey experiment through Amazon’s platform Mechanical Turk (MTurk) in conjunction with Qualtrics survey software. The survey was launched and concluded in a single day on 16 February 2022. Adults aged 18 and older residing in the United States were allowed to take the survey. Respondents were paid $1 for their participation, even if they did not answer all of the questions. The average amount of time it took for respondents to complete the survey was 3 mins and 35 seconds. Thus, if calculating an hourly wage rate, the respondents were compensated just over twice the U.S. national minimum wage. Overall, 614 respondents started the survey and 603 respondents completed all of the questions utilized in the multiple regression analysis.

Scholars have found that MTurk is an inexpensive tool for scholars to conduct survey research without jeopardizing the quality of the data. Buhrmester et al. (2011) found that MTurk samples were demographically more diverse than typical online samples and surveys of college students. Similarly, Thomas and Clifford (2017) found that MTurk online participants were as attentive as offline participants that are recruited for research. The authors demonstrated that online participants did not produce unreliable or invalid data, and that online participants were equally likely to be impacted by experimental designs in survey research. Likewise, Levay et al. (2016) showed that MTurk respondents did not differ from the popular national population-based surveys in unmeasurable ways. Thus, researchers are able to advance research using MTurk, as long as they acknowledge the ways their sample might differ from the population and how those differences could impact the results.

Our sample was quite similar to the population when exploring sociodemographic and attitudinal variables, with two exceptions. The sample had a slightly higher proportion of male respondents when compared to women. Additionally, after coding partisan leaners as partisans, the sample had a noticeably higher proportion of Democratic identifiers. Since we do not find gender or partisan gaps among respondents’ views on the acceptability of trading sexual services, there was no expectation that these trends in the sample are problematic or create biased inferences.

Dependent Variable and Method

The dependent variable measures the respondents’ view on the level of acceptability of trading sexual services. The base of the questions asks, “To what extent do you think that [_____________] is acceptable?”. The concept of trading sexual services was operationalized in three contrasting ways: “prostitution,” “sex work,” or “transactional sex.” Around one-third of the sample was randomly asked one variant of the question. As for the number of respondents contained in the full models, 200 respondents were asked about the acceptability of “prostitution,” 199 were asked about “sex work,” and 204 were asked about “transactional sex.” The respondents were asked to place the acceptability of the activity on a ten-point Likert-scale from 0 = not at all to 10 = to a great degree. The base language in the question, as well as the scaled response option, aligns well with large, reputable cross-national surveys that ask about the topic, such as the World Values Survey (WVS) and the European Values Study (EVS) (EVS/WVS, 2020). Since the dependent variable was a 10-point Likert-scale measure, we employed OLS linear multiple regression in order to estimate predictors of acceptability.

Independent Variables

We included a number of sociodemographic, attitudinal, and associative predictor variables in the models. First, there were a number of sociodemographic predictors included in the models, such as age, gender, race, education, and income. Since previous studies have found that sociodemographic variables are correlated with views on a range of sexual behaviors, we included all of the core demographics as predictors in the models. Here, however, we did not expect noticeable sociodemographic predictors. In Appendix D in the online supplementary material, we also provide models where we included prior experience purchasing sexual services as an independent variable. The models indicate that respondents who have purchased sex are more likely to view prostitution, sex work, and transactional sex as acceptable. That being said, there is the issue of endogeneity regarding the results (e.g., views on acceptability lead to purchasing sex).

Second, we included five attitudinal predictors based on prior research on the topic. A measure of religiosity was included in order to determine whether religious beliefs impact acceptability. Similarly, a measure of political ideology was utilized in order to determine whether a liberal ideology was associated with greater acceptability toward trading sexual services. Since partisan identification is increasingly growing in importance for predicting a range of views on society in the United States, a measure of partisanship is included. Due to the fact that trading sexual services is commonly viewed through a gendered or feminist lens, a measure of views on the importance that men and women have the same sexual freedom was utilized. Likewise, a variable was included that measured how acceptable the respondent viewed the act of non-committal casual sex. The coding scheme for the sociodemographic and attitudinal predictors, as well as descriptive statistics for these variables, are provided in Appendices A and B in the online supplementary material. Further, we provided the coding and descriptive statistics in these appendices for additional variables that we explored, but ultimately excluded because they produced no statistically significant relationships.

Third, a main area of exploration in the analysis was determining whether positive or negative associations with the causes and consequences of the sex trade have differing effects on a respondent’s view regarding the acceptability of trading sexual services based on question wording. For example, it is important to know whether negative associations with the exchange of sexual services is only related to acceptability when asked about “prostitution,” which has historically been a stigmatized word that elicits negative thoughts. Therefore, the respondents were asked their level of association with three commonly identified positive and three commonly identified negative causes and consequences related to trading sexual services. Following the experimental question, the respondents were asked, “how closely do you associate the exchange of sexual services with the following?”. Within the same slider question set, the respondent was first provided three positive associations: sexual freedom, companionship, and women’s empowerment. Then, the respondent was provided three negative associations: poverty, women’s exploitation, and human trafficking. It is worth noting that, at this point in the survey, the respondent was unable to go back and change their response to the experimental question. Using minimum residual factor analysis based on the groupings of three associations, two latent variables were estimated: 1. positive associations, and 2. negative associations. In Appendix C in the online supplementary material, correlation matrices, factor analysis statistics, and Cronbach’s Alpha scores are provided. These two latent traits are contained in the models as predictor variables.

Results – Descriptive Statistics

Figure 1 displays the density of responses for the three different survey question wordings. There are some noticeable and expected trends in the figure. First, the percentage of respondents indicating that prostitution is “0 – not at all acceptable” was statistically higher for prostitution (8.65%) than it was for sex work (3.4%) or transactional sex (6.28%). The result provides some initial support for the idea that the word prostitution elicits a more immediate negative evaluation. However, when collapsing the categories leaning toward unacceptable, all three questions wordings have a similar percentage of responses: prostitution = 28.45%, sex work = 27.67%, and transactional sex = 30.92%. The results yield no statistically significant differences in the overall negative evaluation percentages. Notably, only one-third of respondents provided a response that the trading of sexual services was unacceptable.

Figure 1. Acceptability of exchange of sexual services by question wording.

Respondents were statistically more likely to express a neutral position (5) on “sex work” when compared to prostitution or transactional sex.

On the other hand, there was a statistically significant difference in the percentage of respondents that selected the middle category (5), which expresses a neutral position on the activity. Overall, 11.54% of respondents selected the neutral category when asked about prostitution, 18.45% for sex work, and 12.08% for transactional sex. There was no statistically significant difference in the percentage of respondents that provided the neutral response for prostitution or transactional sex. However, respondents were statistically more likely to select the neutral response when asked about sex work when compared to the other two question wordings. The result provide partial support for $H_1$, which indicates that respondents would be less likely to express a clear opinion on the acceptability of sex work when compared to being asked about prostitution.

Over half of the respondents indicated some level of acceptability for the trading of sexual services. Around 59.62% of respondents indicated some level of acceptability when asked about prostitution, 53.88% for sex work, and 57% for transactional sex. The gap in acceptability between sex work and the other two questions wordings is a consequence of a greater percentage of individuals expressing a neutral position on acceptability. Thus, the term sex work does not appear to reduce the percentage of individuals thinking the trading of sexual services is unacceptable. Instead, the term appears to reduce the number of responses indicating acceptability in favor of a neutral response. The result indicates that the use of the term sex work could be slightly more unclear to individuals. While in survey research a researcher should utilize concepts and terms that are clear to citizens, use of the term sex work might allow policymakers and other actors advocating for policy change greater control over framing the debate regarding the trading of sexual services free from some preconceived negative bias.

Results – Predicting Evaluations of Prostitution, Sex Work, and Transactional Sex

Table 1 provides regression output from the three models predicting support for the trading of sexual services for each of the question wording variants the respondents encountered. An initial aspect of Table 1 to highlight are the measures of fit (adjusted $R^2$) for the three models. The adjusted $R^2$ is largest for the model predicting the acceptability of prostitution. In particular, the adjusted $R^2$ measure of fit indicates that around half (0.48) of the variance in views on the acceptability of prostitution is explained by the independent variables in the model. In comparison, the independent variables explain around 0.37 of the variability in views on the acceptability of transactional sex, and around one-third of the variance in predicting the acceptability of sex work. The result indicates that respondents are better able to translate their current attitudes on sexual behavior and associations with ideas related to the trade in sexual services into a view on the acceptability of prostitution when compared to sex work or transactional sex. Thus, respondents were more likely to provide a clearly defined preconceived position when a question asking about the trade of sexual services used the word prostitution. The results confirm $H_2$. Therefore, if policymakers or organizations want to engage citizens on the topic of trading sexual services from a more neutral position, the word prostitution should be avoided in favor of sex work or transactional sex.

Table 1. Predicting acceptability of exchange of sexual services by question wording.

	Prostitution	Sex Work	Transactional Sex
(Intercept)	1.81	3.15*	2.42*
	(1.07)	(1.06)	(1.16)
Age	0.01	−0.01	−0.02
	(0.01)	(0.01)	(0.01)
Woman	−0.10	0.20	0.10
	(0.32)	(0.34)	(0.31)
White	−0.62	−0.83*	0.15
	(0.41)	(0.41)	(0.40)
Education	−0.01	−0.08	−0.24
	(0.18)	(0.17)	(0.18)
Income	0.03	0.10	−0.02
	(0.06)	(0.06)	(0.06)
Religiosity	−0.07	−0.01	0.08
	(0.06)	(0.06)	(0.06)
Political Ideology	0.08	0.04	0.02
	(0.05)	(0.06)	(0.06)
Party ID – Independent	−0.84	−0.41	−0.02
	(0.48)	(0.47)	(0.42)
Party ID – Republican	−0.55	−0.37	−0.18
	(0.38)	(0.35)	(0.38)
Gender Equality	−0.02	−0.06	0.20*
	(0.08)	(0.08)	(0.07)
Non-Committed Casual Sex	0.68*	0.56*	0.45*
	(0.07)	(0.07)	(0.07)
Positive Associations	0.98*	0.50*	0.35
	(0.18)	(0.19)	(0.19)
Negative Associations	−0.57*	−0.32	0.02
	(0.17)	(0.18)	(0.18)
N	200	199	204
R2	0.52	0.38	0.41
adj.R2	0.48	0.34	0.37
Resid.sd	2.04	2.11	2.11

* indicates statistical significance at p < .05; Standard errors in parentheses.

Another aspect of Table 1 to explore are the differences between the models in terms of significant predictor variables and acceptability of prostitution, sex work, and transactional sex. In particular, here we focus on the attitudinal variables and associations since there were no statistically significant sociodemographic variables (with the exception of race when asked about sex work). First, there was only one predictor variable that was statistically significant across all three models. A respondent’s view on the acceptability of non-committed casual sex was positively related to views on the acceptability of prostitution, sex work, and transactional sex. However, the effect of attitudes on non-committed casual sex differed based on question wording. In Figure 2, we present predicted probabilities with 95% confidence bounds for the effect of the acceptability of non-committed casual sex on the three types of question wording. The effect of the variable was strongest when respondents were asked about prostitution. In particular, comparing an individual that responded that non-committal causal sex was “0 = not at all acceptable” to an individual that said that it was “10 = to a great degree acceptable” there was an average increase of 6.81 points in views on the acceptability of prostitution. In comparison, that increase is only an average of 5.65 points for sex work and 4.49 points for transactional sex. In all three models, the effect is quite large. However, the effect is noticeably larger when asked about prostitution – again conveying how attitudes on general sexual behavior have more of an impact on evaluations on prostitution when compared to the alternative terms for trading sexual services. The results confirm previous findings related to how attitudes on general sexual behavior impact views on the trade of sexual services.

Figure 2. Effect of attitudes on non-committed casual sex on acceptability of exchange of sexual services by question wording.

Predicted probabilities calculated with 95% confidence bounds; Other independent variables held at their median.

A second observation regarding predictor variables is that a belief that men and women should have the same sexual freedom was only statistically significant in the transactional sex model. That being said, the overall effect of the variable was incredibly small. In Figure 3, the predicted probabilities are plotted for the effect of attitudes on men and women having the same sexual freedom on the acceptability of the three questions. There was no effect for prostitution and sex work. When accounting for confidence bounds and comparing an individual that responded that it is “0 = not at all important” for men and women to have the same sexual freedom to an individual that responded “10 = to a great degree important” there was an increase of less than half of a point. It was worth noting that there was limited variance in responses to the question, which helps explain the lack of explanatory power of the variable in the models.

Figure 3. Effect of attitudes on gender equality sexual freedom on acceptability of exchange of sexual services by question wording.

Predicted probabilities calculated with 95% confidence bounds; Other independent variables held at their median.

Third, clear patterns emerge when exploring the impact of positive and negative associations with the causes and consequences of the sex trade. Positive associations had a statistically significant positive relationship with predicting the acceptability of prostitution and sex work. However, there was no statistical relationship with positive associations and views on the acceptability of transactional sex. In Figure 4, the effect of positive associations on the acceptability of the three different terms for trading sexual services are displayed. Positive associations had a large positive impact on the acceptability of prostitution. When comparing the lowest level of associations to the highest level of positive association there was an average increase of around 4.89 points in the acceptability of prostitution. However, the effect was half the size, at 2.49 points when predicting the acceptability of sex work. As stated earlier, there was no statistically significant impact of positive associations on the acceptability of transactional sex.

Figure 4. Effect of positive associations on acceptability of exchange of sexual services by question wording.

Predicted probabilities calculated with 95% confidence bounds; Other independent variables held at their median.

The final finding from the empirical analysis is perhaps the most important when it comes to the selection of terms to be used to represent the trade in sexual services. Model output conveys that negative associations had no statistically significant relationship with views on the acceptability of sex work or transactional sex. However, negative associations did have a negative and statistically significant impact on respondents’ views on the acceptability of prostitution. In Figure 5, predicted probabilities are displayed for the effect of negative associations on acceptability of the three types of questions. As the large confidence bounds demonstrate, there was no effect for sex work and transactional sex. However, there was a clear negative relationship in predicting the acceptability of prostitution. When accounting for the 95% confidence bounds, an individual who highly associated the negative concepts with the exchange of sexual services for payment evaluated the acceptability of prostitution a minimum of 1.1 points lower than an individual with no negative associations. Taken together, the results provide convincing support for $H_3$, which stated that predictor variables will have a much larger effect on views on the acceptability of prostitution when compared to sex work or transactional sex. Further, the results show that use of the word prostitution prompts clear translation of already pre-formed attitudes into an evaluation. In comparison, when asked about sex work and transactional sex, respondents were much less likely to convert pre-formed attitudes and ideas into a quick evaluation of the practice.

Figure 5. Effect of negative associations on acceptability of exchange of sexual services by question wording.

Predicted probabilities calculated with 95% confidence bounds; Other independent variables held at their median.

Conclusion

A longstanding research agenda on measuring public opinion for governmental policies demonstrates that survey question wording, and/or policy framing, can have a sizable impact on the proportion of respondents that supports a given policy (Bishop et al., 1978). For instance, asking survey respondents if they support policies that help “the poor,” rather than if they support “welfare” policies, leads to a dramatic increase in the percentage of respondents indicating support for such policies (Huber & Paris, 2013; Rasinski, 1989; Smith, 1987). Similarly, responses to questions about belief in climate change (Greenhill et al., 2014; Schuldt et al., 2011), levels of generalized trust (Lundmark et al., 2016), political partisanship (Sanders et al., 2018), support for assisted suicide (Magelssen et al., 2016), and even reported vote choice (Box-Steffensmeier et al., 2000) are highly susceptible to changes in question wording or policy framing.

In this study, we explored how wording impacts responses to survey questions asking about the acceptability of trading sexual services. Using an original survey, we assessed how three different concepts used to represent the trade in sexual services could impact views on the acceptability of the activity. The three different concepts were prostitution, sex work, and transactional sex. In addition, multiple regression models were estimated in order to test whether independent variables have contrasting effects on predicting the acceptability of trading sexual services based on the concept utilized in the survey question. To our knowledge, this is the first study to empirically explore how views on the trade in sexual services are susceptible to question wording.

The study generated two overarching findings. First, views on the level of acceptability of trading sexual services were partially a function of question wording. A higher proportion of respondents indicated “0 – not at all acceptable” when asked about prostitution when compared to those respondents that were asked about sex work and transactional sex. The result provides some support for the idea that prostitution is a stigmatized word that elicits immediate negative evaluations. In addition, a higher proportion of respondents indicated a neutral assessment (5 = neither acceptable nor unacceptable) when asked about sex work in comparison to prostitution or transactional sex. The result indicates that sex work may be a more fruitful starting point with which to engage in dialogue and debate regarding the trade in sexual services.

Second, respondents were more likely to translate other formed attitudes and associations into an attitude on prostitution when compared to being asked about sex work or transactional sex. There was only one instance where an attitude had a statistically significant relationship with predicting acceptability for all three question wording variants. Liberal attitudes toward general sexual behavior translate into views that the trade in sexual services was acceptable when asked about prostitution, sex work, and transactional sex. However, the effect of liberal attitudes toward general sexual behavior was largest when respondents were asked about prostitution and smaller when asked about sex work and transactional sex. When exploring the effect of positive associations with the exchange of sexual services, such as sexual freedom, women’s empowerment, and companionship, we found that positive associations increased the level of acceptability of prostitution and sex work. There was no relationship between positive associations and views on the acceptability of transactional sex. Finally, we found that negative associations with the exchange of sexual services, such as poverty, women’s exploitation, and human trafficking, only had an effect on respondents’ assessment of the acceptability of prostitution. Again, the result provides more evidence that prostitution is a stigmatized word.

What is the practical importance of this study as far as contemporary debates regarding policies toward the trade in sexual services are concerned? In their study on the usage of terms to describe the exchange of sexual services and their utility in HIV prevention research, McMillan et al. (2018) argued that sex work is the most appropriate term to use in research and dialogue on the topic. The authors argued that the term sex work frames the topic, “as a matter of labor, not culture or morality” (McMillan et al., 2018, p. 1518). In other words, sex work is not as stigmatizing as the word prostitution. In this study, we found support for this argument. McMillan et al. also asserted that transactional sex may be too vague of a concept to use among a mass public. Again, here we found that respondents were much less likely to translate commonly related attitudes into an opinion on the acceptability of transactional sex. Our results provide some support that the concept is less clear to respondents. In sum, we argue that policymakers, and more importantly sex workers’ rights organizations and other advocacy groups, should steer clear from the word prostitution – unless the goal is to elicit negative responses or further stigmatization. Individuals have pre-formed stigmatized attitudes toward the trade in sexual services when it is described as prostitution. In addition, when asked about prostitution individuals are more likely to translate negative associations with the exchange of sexual services into unacceptability. Thus, any policymaker or organization advocating for the rights and wellbeing of sex workers deploying the term prostitution would be working up hill and need to de-stigmatize the concept prior to engaging in dialogue on particular policies in order to achieve success.

Where is there a need for future research? This study represents only one sample of respondents. Therefore, scholars should continue to explore response differences based on question wording with additional samples. Future research should also explore several additional areas related to individual opinions on the trade in sexual services. First, this study did not uncover why some individuals associate negative outcomes with the exchange of sexual services, while other individuals are more prone to positive associations. Since these associations do impact views on the levels of acceptability of trading sexual services, it would be useful to take a step back to understand why there are differences among individuals in regards to their associations. Second, while the empirical analysis finds differences in the acceptability of the trade in sexual services based on the term that is being used in the question, more research is needed to determine which acts individuals recognize as encompassing the exchange of sexual services. As stated, it could be the case that each term is associated with a particular variant of the exchange of sexual services, such as prostitution with street-based activities. It would be important for future research to explore which activities respondents associate with each term in order to better understand the nuances of public opinion on the issue.

Ethical Review Process

University of Wisconsin – Parkside Institutional Review Board Compliance FY20-10

Acknowledgement

Authors listed alphabetically and contributed equally. The authors would like to thank the editors of The Journal of Sex Research, as well as the three incredibly helpful anonymous peer-reviewers. Additionally, a special thanks to Zaskia A. Hansen for her patience while we completed the study.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/00224499.2022.2130859.

Additional information

Funding

The authors reported there is no funding associated with the work featured in this article.