The Hotel Room Purchase Task: Effects of Gender and Partner Desirability on Demand for Hypothetical Sex in Individuals with Disordered Cocaine Use and Controls

Abstract

Hypothetical purchase tasks allow for rapid assessment of behavioral economic demand for numerous commodities and are useful in evaluating reinforcer pathologies, such as substance and behavioral addiction. Currently, there is not a task for evaluating demand for sex without requiring implicit engagement in sex work. The current study used a novel purchase task with hotel rooms for sex as the hypothetical commodity to assess demand for sex in individuals with disordered cocaine use, a population that frequently engages in risky sexual behavior. Adults meeting criteria for cocaine abuse or dependence (13 males, ten females) and non-cocaine-using controls (eight males, three females) chose hypothetical sexual partners from a series of photographs and endorsed two partners with whom they would most and least like to have sex. Participants then completed the hotel purchase task for both partners, wherein they reported how many nights at a hotel room, at prices from $10 to $1280 per night, they would purchase in a year. Demand intensity was significantly greater and demand elasticity was significantly lower for the most preferred relative to the less preferred partner. Males demonstrated significantly greater intensity and lesser elasticity for sex than females. Demand metrics did not differ between the cocaine and control group. This task may serve as a useful measure of demand for sex without requiring implicit hypothetical engagement in sex work. Future studies exploring the relation between task performance and other characteristics such as sexual dysfunction, in addition to acute substance administration effects, may further determine the task’s clinical utility.

INTRODUCTION

The field of behavioral economics combines principles of microeconomics and operant psychology to evaluate individual decision-making processes as they pertain to specific reinforcers or commodities (Bickel, Johnson, Koffarnus, MacKillop, & Murphy, 2014; Hursh, 1980, 1984; Lea, 1976). Within behavioral economics, the efficacy of a reinforcer is typically evaluated through a demand curve, which is generated by assessing consumption of a commodity across an array of monetary prices or response requirements. Consumption can then be modeled using nonlinear regression techniques (Hursh & Silberberg, 2008) to provide various demand metrics. Demand intensity, for instance, is a measure of consumption at the lowest price, whereas demand elasticity refers to the sensitivity of consumption to increases in unit price. Greater elasticity refers to proportionally greater decreases in consumption at higher prices compared to lower prices, and more inelasticity referring to lesser or no decreases in consumption at higher prices compared to lower prices. By evaluating the interaction between price and consumption, demand curve analysis allows for a multidimensional and robust assessment of reinforcement (Bickel & Madden, 1999; Johnson & Bickel, 2006).

Although assessments of demand using operant procedures provide the most rigorous experimental approach, they are time-consuming, often requiring many individual sessions. Furthermore, certain behaviors are not feasible for testing within the laboratory; in the case of drug dependence, for instance, provision of one’s drug of abuse may not be appropriate, especially for individuals in treatment (Jacobs & Bickel, 1999). In contrast, hypothetical purchase tasks provide a quick and efficient alternative to intensive laboratory assessments of operant demand. Through the use of an instructional set to provide context for the hypothetical purchase scenario, purchase tasks can assess behaviors across brief or long periods of time.

Hypothetical purchase tasks have been developed to assess demand for a number of commodities and are most commonly used to assess demand for substances of abuse, such as alcohol (MacKillop & Murphy, 2007; Murphy & MacKillop, 2006), tobacco (Jacobs & Bickel, 1999; MacKillop et al., 2008), e-cigarettes (Johnson, Johnson, Rass, & Pacek, 2017; Stein, Koffarnus, Stepanov, Hatsukami, & Bickel, 2018), marijuana (Aston, Metrik, Amlung, Kahler, & MacKillop, 2016; Collins, Vincent, Yu, Liu, & Epstein, 2014), anabolic steroids (Pope et al., 2010), heroin (Jacobs & Bickel, 1999), cathinone derivatives (“bath salts”; Johnson & Johnson, 2014), and cocaine (Bruner & Johnson, 2014). Hypothetical purchase tasks have also been successfully implemented in the assessment of nondrug reinforcers or behaviors such as food (Epstein et al., 2018; Sze, Stein, Bickel, Paluch, & Epstein, 2017), ultraviolet indoor tanning (Becirevic et al., 2017; Reed, Kaplan, Becirevic, Roma, & Hursh, 2016), Internet use (Acuff et al., 2018a), and erotica consumption (Mulhauser, Miller, Short, & Weinstock, 2018).

Demand metrics derived from these assays often strongly correlate with domain-specific outcomes, such as self-reported (Bruner & Johnson, 2014; Strickland, Lile, & Stoops, 2017) and actual (Amlung, Acker, Stojek, Murphy, & MacKillop, 2012) drug consumption and severity of dependence (Murphy & MacKillop, 2006; Skidmore, Murphy, & Martens, 2014; Strickland et al., 2017). Furthermore, Wilson, Franck, Koffarnus, and Bickel (2016) found robust correlations across prices between cigarettes purchased in a hypothetical purchase task and real cigarettes purchased with experimenter-provided money (although absolute levels of consumption were somewhat higher for hypothetical vs. real cigarettes), suggesting that hypothetical purchasing provides a reasonable proxy for modeling patterns of real-world engagement in a variety of reinforcing behaviors. Similar domain-specific trends have emerged in nondrug purchase tasks, such that demand for food correlates to body mass index (BMI) (Epstein et al., 2018), demand for indoor tanning access correlates with tanning frequency (Bericevic et al., 2017; Reed et al., 2016), demand for Internet access correlates with problematic Internet usage behaviors (Acuff et al., 2018a; Acuff, Soltis, Dennhardt, Berlin, & Murphy, 2018b), and demand for erotica is associated with degree of sexual compulsivity (Mulhauser et al., 2018).

The findings discussed above demonstrate the utility of a behavioral economic framework in evaluating reinforcement across commodities and behaviors; however, sexual behavior, which is a primary reinforcer and comprises a central component in human society and evolution, is conspicuously absent from the realm of demand analysis. This lack of assessment of sexual behaviors using behavioral economic principles is surprising, given that the analysis of “reinforcer pathologies”, which describes an overvaluation of engagement in a reinforcing behavior despite long-term negative consequences (Bickel, Jarmolowicz, Mueller, & Gatchalian, 2011), has been approached using a behavioral economic lens, yet primarily limited to assessments of drug dependence. Although compulsive sexual behavior, also referred to as “sex addiction” or “hypersexuality” (Walton, Cantor, Bhullar, & Lykins, 2017a), has not formally been designated as a psychiatric diagnosis, the overvaluation of sexual behaviors to the detriment of other aspects of life (i.e., marriage, job, etc.) fits squarely within the realm of reinforcer pathology theory. Previous evaluations of compulsive sexual behavior have demonstrated concurrent presence of depression and excessive life stress in individuals exhibiting compulsive sexual behavior (Schultz, Hook, Davis, Penberthy, & Reid, 2014; Storholm, Satre, Kapadia, & Halkitis, 2016; Walton, Cantor, & Lykins, 2017b), suggesting that the overvaluation of sexual behavior parallels similar findings in addiction in which overvaluation of drug use is comorbid with depressive symptoms (Davis, Uezato, Newell, & Frazier, 2008). Conversely, undervaluation of or aversion to sexual behavior is a central component of disorders such as acquired hypoactive sexual desire disorder or vaginismus (Cherner & Reissing, 2013; Goldstein et al., 2017; McCarthy & McDonald, 2009) and highly prevalent in psychiatric disorders such as major depressive disorder, anxiety disorders, and post-traumatic stress disorder (Barata, 2017; Phillips & Slaughter, 2000; Waldinger, 2015). Given the role of sexual behavior in general life and psychiatric disorders, a behavioral economic measure of sexual reinforcement would provide sex researchers and clinicians alike with a valuable tool for evaluating changes in the reinforcing aspects of sexual behavior after interventions in the laboratory or clinic.

In addition to potential emotional and social problems, overvaluation of sexual behavior in compulsive sexual behavior has also been associated with greater engagement in sexual risk behavior, such as sex with anonymous partners or decreased likelihood of utilizing protective measures in sexual encounters (Parsons, Rendina, Ventuneac, Moody, & Grov, 2016; Yeagley, Hickok, & Bauermeister, 2014). This is an especially salient issue, given the increasing incidence of sexually-transmitted infections like syphilis and gonorrhea, and the sustained incidence of HIV, especially among gay and drug-using populations (Center for Disease Control, 2018a, b; Scott-Sheldon & Chan, 2019). Analogous behaviors have been investigated using a behavioral-economic framework, specifically delay and probability discounting, that evaluated likelihood of condom use as a function to delay to condom access or probability of STI contraction (Johnson & Bruner, 2012; Johnson, Johnson, Herrmann, & Sweeney, 2015). These studies have also regularly reported gender differences in condom-use likelihood such that males were significantly less likely to use a condom relative to females (Johnson & Bruner, 2013; Johnson et al., 2015; Sweeney et al., 2019). These results are not entirely surprising, as differences in sexual behavior represent the largest behavioral gender difference, with men demonstrating more permissive attitudes toward casual sex, greater incidence of sexual intercourse, and greater frequency of sexual intercourse (Hyde, 2005; Oliver & Hyde, 1993); however, it may not be accurate to state that males find sexual behaviors more reinforcing than females, as gender differences in sexual satisfaction are minimal (Oliver & Hyde, 1993). Development of a hypothetical purchase task for partnered sexual outcomes (i.e. not erotica consumption) would provide a behavioral economic measure relevant for evaluating the reinforcing aspects of sexual behavior and may be applicable to risky sexual behaviors and evaluation of potential treatment and educational interventions for at-risk populations.

To our knowledge, only one study has directly assessed demand for sex acts by explicitly asking how many sex acts participants would purchase at various prices (Jarmolowicz, Lemley, Mateos, & Sofis, 2016); however, this method implicitly involves engagement in sex work, albeit hypothetical, which may influence responding in those with ethical concerns about this behavior, aversion to illicit behavior, or those who are otherwise unwilling to directly pay for sex. The current study attempted to address these potential limitations by developing and exploring the feasibility of a novel approach to hypothetical demand for sex by using nights at a hotel room explicitly used for casual sex with a consenting partner as the commodity for purchase rather than sex acts themselves. In addition to assessing the feasibility of the task, we aimed to assess potential differences in sexual reinforcement between individuals with disordered cocaine use, a population that is at high-risk for HIV contraction (Booth, Kwiatkowski, & Chitwood, 2000; Booth, Watters, & Chitwood, 1993; McCoy, Lai, Metsch, Messiag, & Zhao, 2004; Strathdee & Sherman, 2003), and demographically matched, non-cocaine-using controls, as well as potential differences between males and females, given the gender differences in sexual behavior mentioned above. We hypothesize that demand for sex will be greater in individuals with disordered cocaine use relative to the controls, in males relative to females, and for a more desirable partner relative to a less desired one.

METHOD

Participants

Participant demographics and recruitment were reported previously (Johnson et al., 2015). Participants were both individuals with disordered cocaine use (n = 23) and non-cocaine-using individuals (n = 11) recruited from the Baltimore area using flyers, Internet, newspaper, and radio advertisements. Informed consent was obtained from all individual participants included in the study. Inclusion criteria for both the cocaine abuse or dependence (cocaine) and non-cocaine-using (control) groups included being at least 18 years of age, having at least an eighth-grade reading level, and reporting having vaginal or anal intercourse with another person during their lifetime. Participants in the cocaine group met DSM-IV criteria for cocaine abuse or dependence (American Psychiatric Association), whereas participants in the control group reported no lifetime use of cocaine. Participants in both groups could meet criteria for abuse for drugs other than cocaine but could not meet dependence criteria for other drugs (excluding nicotine and caffeine). Exclusion criteria for both groups included self-reported serious head trauma, dementia, significant cognitive impairment, not currently sexually active (>5 years since last intercourse), or current diagnosis of major psychiatric disorder (such as schizophrenia, bipolar disorder, major depressive disorder) besides substance abuse/dependence.

Procedure

Following a telephone screening assessment, initially qualified participants came to the laboratory for an in-person screening. If qualified, participants remained in the laboratory for approximately 4 h to complete a variety of behavioral economic tasks described elsewhere (Johnson et al., 2015). In addition to the tasks, participants completed the HIV Risk Behavior Scale (HRBS) (Darke, Hall, Heather, Ward, & Wodak, 1991), which is an 11-item, validated scale assessing both drug-use and sexual-risk behaviors associated with HIV risk.

Before beginning the task, participants were instructed to pretend that they were not in a committed relationship. Similar to previous studies using the Sexual Delay Discounting Task (e.g., Johnson & Bruner, 2012; Johnson et al., 2015; Johnson, Herrmann, Sweeney, LeComte, & Johnson, 2017a), participants used a computer to view 60 individually-presented color photographs of diverse, clothed people (30 male, 30 female) and were asked to select photographs of individuals that they would consider having casual sex with based solely on physical appearance. The photo set was assembled from a variety of publicly available online repositories to provide a range of physical appearances. Next, participants identified from the subset of initially selected photographs the person they (1) most wanted to have sex with and (2) least wanted to have sex with. Participants who selected fewer than two photographs did not proceed further with the task.

Hotel Purchase Task

Participants completed two identical purchase tasks associated with the photograph they selected as the person with whom he/she most wanted to have sex, and the person with whom he/she least wanted to have sex. Participants read the following instructions and reviewed them with a research assistant to check for understanding before completing each task:

In the questions that follow we would like you to pretend that the person in the photograph wants to have sex with you on a regular basis. Neither of you are currently in a relationship nor is there any risk of pregnancy. However, even though you can have sex with this person whenever you want, you must always purchase a hotel room first. In the questions that follow, you will be asked to purchase individual nights in hotel room (i.e., one night at a time) for the upcoming year. Please answer the questions honestly and thoughtfully.

Pretend that this is the only opportunity to have sex that is available to you. You cannot have sex in any other location except for the hotel rooms you choose to purchase below. Also assume you have no other potential sexual partners. In other words, if you want to have sex at any time during the course of the upcoming year, you must do so in a hotel room that you purchase today. Prices for the rooms you may buy are listed below. You may buy as many as 365 nights for the room (1 years’ worth) or as few as 0 rooms.

Also, assume that the hotel room nights you are about to purchase are for your personal use only for having sex with the photographed individual. The hotel rooms cannot be used for any other reasons. You can’t sell the hotel rooms or give them to anyone else to use for any reason. All of the hotel room nights you buy are, therefore, for your own personal, sexual use with the photographed individual within the year.

Below is a list of various prices for the hotel room. In the space provided please indicate how many of these hotel rooms you would purchase at each of the prices listed in the column on the left. Please complete the entire table. If you wouldn’t purchase any hotel rooms at a particular price, please put “0”. Remember, only buy nights in the hotel room that you would personally use over the next year. If you have any questions, please ask us for help.

Although not formally included in the printed instructions, participants were verbally reminded by research staff to consider their current financial circumstances and not to spend beyond their available income and savings if purchasing patterns were judged by research staff to be unrealistically high. Participants then responded with how many nights they would purchase at a hotel over the next year at the following prices: $10, $20, $40, $80, $160, $320, $640, and $1280 per night.

Data Analysis

Orderliness of individual demand curves was determined according to three criteria previously used to eliminate nonsystematic data from purchase tasks (Bruner & Johnson, 2014; Stein, Koffarnus, Snider, Quisenberry, & Bickel, 2015). Purchasing functions were identified as nonsystematic if (1) units purchased at a given price were at least 20% greater than at the preceding price, (2) units purchased at the final price were not less than the first price by at least 10%, or (3) units purchased were greater than zero after zero consumption was reported at a lower price. Datasets violating these criteria were not included in the final analysis; however, datasets only violating criterion 2 were still included in analyses of demand intensity, given that they are still generally orderly, unlike those violating criteria 1 and 3. Datasets with 0 or unchanging consumption across prices still provide valuable information regarding consumption, but nonlinear curves cannot be fit to these data.

In GraphPad Prism (version 7.0c for Apple, Graphpad Software, La Jolla, California) group and individual demand data were log-transformed and modeled using the exponential demand equation (Hursh & Silberberg, 2008):

$$\log Q=\log Q_{\mathit{0}}+k(e^{-\alpha Q_{\mathit{0}}^C}-1)$$

In Eq. (1), Q refers to consumption at price C, k represents the range of consumption across prices, Q₀ represents demand intensity (consumption at the lowest price), and α is an elasticity-related rate constant describing the change in consumption as a function of price. Group data were described as the median number of rooms purchased at each price up to the price at which zero median consumption was endorsed. In order to plot the group data on logarithmic coordinates, zero consumption was changed to 0.1 for each group. For each individual dataset, at the first price at which 0 consumption was endorsed, the 0 was changed to 0.1 to allow for logarithmic evaluation, and subsequent 0 values were excluded from curve-fitting. The number of hotel rooms purchased at $10/room was used as Q₀ instead of the parameter derived from Eq. (1). The value of k was the range of all log-transformed possible consumption values, and was calculated as the difference between the log-transformed maximum observed consumption (log 365 nights = 2.56) and the transformed minimum consumption value (0 nights = log 0.1 nights = −1). For all applications of Eq. (1), k was set to 3.56 in order to directly compare individual α values across conditions. Values for Q₀ and α were highly-skewed and were square-root and natural-log transformed, respectively, for subsequent analyses. Transformed Q₀ and α values were analyzed using separate two-way analyses of variance with repeated measures to evaluate the between-group effects of either drug use or gender, and the within-subject effect of partner condition. Goodness of fit of individual demand curves was assessed using the standard deviation of residuals (Sy.x).

Scores from the five HRBS questions associated with sexual risk were summed together to calculate a composite sexual risk behavior subtotal (HRBS Sex Subtotal). Higher scores indicate greater HIV risk based on sexual behaviors. Sexual risk behavior scores were compared between the cocaine and control groups and males and females using a two-way ANOVA. Additionally, Pearson’s correlation coefficients were determined for the relations between HRBS Sex Subtotal scores and natural log-transformed α and square-root transformed Q₀ for both most- and least-preferred sexual partners.

RESULTS

A total of 56 potential participants were excluded from participation after screening (22 were below eighth-grade reading level; 16 were physically dependent on drugs other than cocaine; 8 used cocaine but did not meet abuse criteria; 3 had a current major psychiatric disorder; 3 demonstrated deceitfulness based on contradictory answers throughout screening; 2 were not currently sexually active; 1 provided BrAC > 0; and one did not choose at least a single picture in the partner selection component). Participant demographics for the control (n = 11) and cocaine (n = 23) groups are presented in Table 1. The two groups did not differ across demographic variables such as age, race, sex, marital status, income, education, or intelligence; however, the cocaine group consumed alcohol more frequently (M = 10.17 days/month, SD = 10.36) relative to the healthy controls (M = 2.43 days/month, SD = 4.70); t(32) = 2.35, p = .025, d = 0.96. There were no differences in tobacco or opiate use between groups. Of the 23 participants in the cocaine group, 20 participants (87%) met criteria for cocaine dependence, and 3 participants (13%) met criteria for cocaine abuse. Sex differences were determined for Quick Test intelligence scores such that scores were slightly lower for females (M = 40.77, SD = 3.56) than males (M = 44.29, SD = 3.41) (F(1, 33) = 7.148, p = .012), but no other sex differences or sex by group differences were determined for demographic variables (all ps > .05).

Table 1.

Participant sample demographics

Characteristic	Cocaine (n = 23)	Control (n = 11)
Age in years, mean (SD)	46.30 (10.93)	49.82 (11.69)
Sex, count (%)
Male	13 (57%)	8 (73%)
Female	10 (43%)	3 (27%)
Race, count (%)
African-American/Black	14 (61%)	7 (64%)
Caucasian/White	8 (35%)	4 (36%)
More than one race	1 (4%)	0 (0%)
Marital Status, count (%)
Non-married (single/separated/divorced/widowed)	20 (87%)	10 (91%)
Married	3 (13%)	1 (9%)
Education in years, mean (SD)	13.09 (1.68)	13.36 (1.57)
Monthly income, mean (SD)	$1,185.74 (825.77)	$1,149.82 (891.92)
Quick Test intelligence score, mean (SD)a	43.17 (3.45)	42.45 (11.69)
Substance Use Frequency in days/month, mean (SD)
Alcohol	10.17 (10.36)	2.43 (4.70)
Cannabis	5.24 (9.89)	1.59 (5.18)
Opioids	0.99 (3.15)	0.016 (0.03)
Cocaine	15.97 (9.11)	0 (0)

^aAmmons & Ammons (1962). Scores range from 2 – 50.

Bolded numbers indicate a significant difference between groups at α = .05

Demand data for all conditions followed the typical pattern of positively decelerating consumption as a function of price (Fig. 1) and were mostly orderly in that units purchased at one price were not greater than at the preceding price (Bruner & Johnson, 2014; Stein et al., 2015). Although all three criteria were used to evaluate orderliness of responses, there were no datasets violating criteria 1 and 3. One dataset from the “Most Sex” partner condition in the cocaine group was removed from elasticity analysis according to criterion 2 as the number of hotel rooms purchased at $1280 was the same as the number purchased at $10 (365 rooms purchased at every price); however, this participant’s data were still included for analyses of demand intensity. Similarly, in the “Least Sex” partner condition, one dataset from the control group and four datasets from the cocaine group were excluded from elasticity analysis according to criterion 2 for not purchasing any hotel rooms at any price as the slope was undefined with zero purchasing; however, these datasets were still used in analyses of demand intensity. Additionally, two datasets from the Cocaine group “Least Sex” condition were excluded from elasticity analysis for not purchasing any rooms at prices above $10, which rendered elasticity as undefined. Goodness-of-fit of the individual curves, as indicated by the standard deviation of the residuals (Sy.x), was acceptable for both the most-sex (M = 0.353, SEM = 0.051) and least-sex (M = 0.380, SEM = 0.070) partner conditions. Goodness-of-fit did not differ across partner, drug-use, or gender conditions.

Figure 1.

A: Demand curves for hotel rooms purchased for sex across prices as a function of drug use. Control group (Control- circles, dashed lines) and Cocaine group (Cocaine- squares, solid lines) are presented according to partner preference (Most Sex- filled symbols; Least Sex- open symbols). B: Demand curves for hotel rooms purchased for sex across prices as a function of gender. Males (circles, dashed lines) and females (squares, solid lines) are presented according to partner preference (Most Sex- filled symbols; Least Sex- open symbols).

The mean, transformed values for Q₀ and α are shown in Fig. 2 for both drug use groups and gender. The repeated-measures analyses of variance testing demand intensity by drug use group determined a significant main effect of partner condition (F(1, 32) = 34.02, p < .001, partial η² = 0.52, observed power = 1.00) with the “Most Sex” condition demonstrating greater demand intensity than the “Least Sex” condition, but no significant main effect of group (F(1, 32) = 0.72, p = .403, partial η² = 0.02, observed power = .131). A significant partner condition by drug use interaction was not detected for demand intensity (F(1, 32) = 0.13, p = .718, partial η² = 0.00, observed power = 0.13). Analyses of demand elasticity by drug use determined a significant main effect of partner condition (F(1, 24) = 61.80, p < .001, partial η² = 0.72, observed power = 1.0) with the “Most Sex” condition demonstrating lower elasticity than the “Least Sex” condition, but no significant main effect of group (F(1, 24) = .03, p = .854, partial η² = 0.05, observed power = 0.19). A significant partner condition by group interaction was not detected for elasticity (F(1, 24) = 1.29, p = .267, partial η² = 0.05, observed power = .05).

Figure 2.

Demand metrics by drug use group (top; A & B) and gender (bottom; C & D). Data are presented as mean (±SEM) of square-root-transformed Q₀ (left; A & C) and natural-log-transformed α (right; B & D). * p< .05 by partner condition. # p< .05 by gender.

Analyses of demand intensity by gender revealed significant main effects of partner condition (F(1, 32) = 39.35, p < .001, partial η² = 0.55, observed power = 1.0) and gender (F(1, 32) = 12.70, p = .001, partial η² = 0.284, observed power = 0.93), wherein intensity was significantly greater in the “Most Sex” relative to the “Least Sex” condition, and males demonstrating greater demand intensity than females. No significant gender by partner condition interaction was detected (F(1, 32) = 0.12, p = .731, partial η² = 0.00, observed power = 0.06). Analyses of demand elasticity by gender similarly determined significant main effects of partner condition (F(1, 31) = 58.00, p < .001, partial η² = 0.71, observed power = 1.0) and gender (F(1, 31) = 5.35, p = .030, partial η² = 0.18, observed power = 0.60), wherein elasticity was significantly less in the “Most Sex” relative to the “Least Sex” condition and males demonstrated less elasticity than females. No significant gender by partner condition interaction was detected (F(1, 31) = 0.50, p = .486, partial η² = 0.02, observed power = .10).

HRBS Sex Subtotal scores were significantly greater in the cocaine group (M = 9.48, SD = 5.00) than the control group (M = 5.27, SD = 4.96) (F(1,33) = 6.693, p = .015, partial η² = 0.182, observed power = .71), as previously reported (Johnson et al., 2015). HRBS Sex Subtotal scores did not differ between males (M = 7.24, SD = 4.75) and females (M = 9.54, SD = 6.01) (F(1,33) = 0.036, p = .852, partial η² = 0.001, observed power = .05), and no group by sex interaction was determined (F(1,33) = 2.431, p = .129, partial η² = 0.075, observed power = .326). HRBS Sex Subtotal scores were not significantly correlated with square-root transformed Q₀ for either the “Most Sex” (r(34) = .086, p = .630) or “Least Sex” conditions (r(34) = .065, p = .719), nor were they significantly correlated with natural-log transformed α for either the “Most Sex” (r(33) = .196, p = .275) or “Least Sex” conditions (r(27) = .022, p = .914).

DISCUSSION

These data indicate that the hotel room purchase task may serve as a useful measure of hypothetical demand for sex-related outcomes while circumventing scenarios analogous to sex work. Demand curve analysis indicated a monotonic decrease in the number of hotel rooms purchased as a function of price, as is commonly seen in hypothetical purchase tasks (e.g., Bruner & Johnson, 2014; Mackillop & Murphy, 2007; Petry & Bickel, 1998), and in operant demand studies (e.g., Johnson & Bickel, 2003; Johnson, Bickel, & Kirshenbaum, 2004). Moreover, demand for sex was well-described by an exponential demand function (Hursh & Silberberg, 2008). Furthermore, there were substantial differences in demand between partner conditions and gender, suggesting that the procedure is sensitive to certain within- and between-subject variables.

Beyond the effect of price itself, the largest and most-consistent differences in demand were observed across partner conditions. Specifically, demand was significantly higher for the most-preferred partner relative to the less-preferred partner, as we had initially hypothesized. These results replicate the findings of Jarmolowicz et al. (2016) which demonstrated orderly decreases in demand for hypothetical sex acts as a function of reduced partner preference.

These findings further highlight the role of partner preference in sexual decision-making processes. Studies evaluating the role of delay discounting on risky sexual behaviors have demonstrated that people more steeply discount delayed, condom-protected sex with a more-preferred over a less-preferred sexual partner (Berry et al., 2019; Collado, Johnson, Loya, Johnson, & Yi, 2017; Dariotis & Johnson, 2015; Hermann, Hand, Johnson, Badger, & Heil, 2014; Herrmann, Johnson, & Johnson, 2015; Johnson & Bruner, 2012, 2013; Johnson et al., 2017a, b; Koffarnus et al., 2016; Quisenberry, Eddy, Patterson, Franck, & Bickel, 2015), including the participants in the current sample (Johnson et al., 2015). Individuals who are more likely to choose an immediately-available, yet less-preferred, sexual partner over their most-preferred, but available after a delay, sexual partner are more likely to engage in risky sexual behaviors and have a higher number of sexual partners than those willing to wait for their preferred partner (Jarmolowicz et al., 2014; Lemley, Jarmolowicz, Parkhurst, & Celio, 2018). Similarly, a previous assessment of demand for sex acts demonstrated reduced demand for less desirable partners relative to more desirable ones (Jarmolowicz, Lemley, Mateos, & Sofis., 2016). Our data suggest a comparable sensitivity to partner desirability in a casual sex scenario without engagement in sex work through direct purchase of sex acts. Unlike the aforementioned discounting studies, however, we did not find any significant association between demand metrics for either partner and sexual risk behavior on the HRBS, suggesting the task may assess sexual decision-making processes related to reinforcement rather than risk behavior.

In general, males exhibited more inelastic demand for sex (i.e., lower sensitivity to price increases). In fact, of the 5 datasets excluded from elasticity analysis for not purchasing any hotel rooms, 4 were from female participants. Although both groups demonstrated reduced demand for the less preferred partner, females’ demand for their most-preferred partner was comparable in magnitude and shape to the males’ least preferred partner. Similar gender-related differences in partner-based decision making for sexual outcomes have been reported in the delay discounting literature (Jarmolowicz, Lemley, Asmussen, & Reed, 2015; Jarmolowicz et al., 2014; Johnson & Bruner, 2013; Wilson & Daly, 2004). These gender differences across studies and behavioral economic models for sexual outcomes may result in part from different mating strategies employed by males and females, with males showing relatively greater preference for short-term mating strategies and females showing relatively greater preference for long-term strategies (Buss & Schmitt, 1993). From an evolutionary perspective, a preference for numerous, short-term partners may be relatively advantageous for men regardless of partner quality, but females, conversely, may need to be more selective and tend to prefer fewer, higher-quality partners, as impregnation by a suboptimal male (in terms of both resource provision by the male and genetics) represents an opportunity cost considering females’ substantial parental investment and more limited maximal number of potential offspring (Buss & Schmitt, 1993).

We initially hypothesized that individuals with disordered cocaine use would have a substantially higher demand for sex relative to controls for a number of reasons: 1) cocaine use is associated with risky sexual behavior in humans (e.g., Booth et al., 1993, 2000); 2) cocaine administration increases sexual desire in humans (Johnson et al., 2017); 3) cocaine increases sexual behavior in nonhumans (Akins, Bolin, & Gill, 2017; Levens & Akins, 2004); and 4) individuals with disordered cocaine use discount delayed, condom-protected sex more than matched-controls (Johnson et al., 2015). Surprisingly, however, there were no differences in demand for sex between groups.

There are a number of potential reasons that we did not detect a difference in demand for sex between groups. First, the increased sexual risk behaviors associated with cocaine use may at least partially result from the acute effects of cocaine (Ober, Shoptaw, Wang, Gorbach, & Weiss, 2009), and the present study may not reflect such an effect because cocaine was not administered to participants. Discounting of condom-protected sex was significantly and dose-dependently increased (e.g., reduction in condom-use likelihood) as were ratings sexual desire in cocaine users under conditions of acute cocaine administration (Johnson et al., 2017). Second, our samples were determined to be underpowered to detect a between-subjects difference in demand between the cocaine and control groups. As such, a larger sample may be required to detect potentially subtle differences in sexual demand between individuals with disordered cocaine use relative to controls in nonintoxicated states, should they exist. Third, regarding previous discounting effects and the present findings, it is important to keep in mind that demand and discounting constitute unique processes and may be independent in this context, so group differences that emerge in one model may not be apparent in the other. Previous studies simultaneously evaluating delay discounting and hypothetical demand for marijuana or alcohol failed to demonstrate a significant relation between discounting and demand metrics, but they did demonstrate that the two tasks were uniquely associated with specific behaviors, such that demand was typically associated with frequency of use or total consumption, whereas discounting was associated with the severity of dependence (Acuff et al., 2018; Aston et al., 2016). Strickland et al. (2017), however, found significant correlations between discounting of cannabis, but not money, and hypothetical cannabis demand. Fourth, it is important to note that the Cocaine group consumed alcohol more frequently than the Control group, which may have limited our ability to detect cocaine-specific differences in behavior. However, given that alcohol increases sexual desire (e.g., Johnson et al., 2016) and is associated with increased sexual delay discounting (Jarmolowicz et al., 2013; Johnson et al., 2016), it seems reasonable that the additional drug use would likely only serve to increase demand for sex. Furthermore, concurrent alcohol use disorder among those with cocaine-use disorder is extremely common (e.g., Higgins, Budney, Bickel, Foerg, & Badger, 1994; Stinson et al., 2005), so matching groups on the basis of all substance use may diminish face validity and translatability. Future studies evaluating the relation between sex discounting and demand are necessary to understand how well the two processes are related and to explain group differences, or lack thereof, between drug users and nonusers.

We found no significant relations between demand metrics and HRBS Sex Subscale scores. Given that we found distinct differences in demand between the most- and least-preferred partners as well as gender-based differences, the current task may better model sexual reinforcement rather than sexual risk. Alternatively, the narrow range of risky sexual behaviors assessed and narrow time range (past 30 days) evaluated in the HRBS may not have provided a comprehensive enough assessment of sexual risk behavior to definitively relate the task to sexual risk behavior. The HRBS is primarily used for evaluating HIV-associated behaviors in drug users (Darke et al., 1991), so an alternative scale designed for broader populations, such as the Sexual Risk Scale (Turchik & Garske, 2009) or Safe Sex Behavior Questionnaire (DiIorio, Parsons, Lehr, Adame, & Carlone, 1992), may have provided greater resolution and a broader series of risky behaviors for validation purposes. Additional studies with the task will be needed to effectively rule out its ability to relate to sexual risk behaviors. Finally, given our limited sample size, we may have been underpowered to detect significant relations between the collected demand metrics and the sexual risk subscale on the HRBS.

The primary strengths of this study lie in the robust, parametric assessment of reinforcement provided by demand analysis and the laboratory-based investigation of reinforcing behaviors in a clinically relevant sample with demographically-matched controls. Despite these strengths, there are numerous limitations to consider when interpreting these findings. Perhaps most pertinent is the relatively small sample size assessed. Although our assessments of power determined we were sufficiently powered for the within-subjects analyses of partner preference and were adequately powered to detect gender differences, we were severely underpowered to confidently determine differences based on drug use or any interactive effects. Although we were able to detect main effects of gender, the low number of females in the control group severely hindered our ability to assess a group by gender interaction. The gender main effects provide some support for the ability of this task to detect individual differences in sexual reinforcement; however, we are unable to definitively claim how cocaine use and gender may interact to influence sexual behavior and replication in larger samples with balanced gender representation will be necessary to address this limitation.

Aside from the limited sample size, data regarding broad sexual behaviors were not collected for validation purposes and assessment of sexual risk was limited to a five-item subscale within a larger scale associated with HIV risk developed for use in drug-using samples (Darke et al., 1991). Additional studies including more generalizable assessments of sexual risk, compulsive sexual behavior, and sexual behavior broadly will be essential for determining the validity of the task for measuring clinically relevant sexual behaviors. Additionally, the task only evaluated sex as a general concept, rather than specifying condom-protected sex or additional risky components of sexual behavior (e.g., anal sex), which may have limited the ability to relate the demand metrics to our measure of sexual risk. Finally, although monetary-based demand assessments are valid for consumable commodities (e.g., drugs, food), sex outside of paid sex work is typically a free behavior and models based on monetary cost may not be the most-effective approach to assessing sexual reinforcement. Comparisons of the current task against novel variations using nonmonetary cost (e.g., effort, time) will be useful for determining the effectiveness of this task and optimizing measures for quantifying sexual reinforcement.

We believe this task will be a useful tool for sex researchers, behavioral economic researchers, and clinicians alike for its ability to evaluate sex demand in lieu of hypothetical engagement in sex work. In the realm of sex research, this task provides a quantified value of sexual reinforcement that can be used for exploring topics relevant to sexual behavior, such as the reinforcing properties of sex in normal vs. dysfunctional sexual behavior (e.g., compulsive sexual behavior, hypoactive sexual desire), the relation between sexual reinforcement and sexual violence, and how demand for sex may relate to attitudes toward and engagement in sex work for both sex workers and their clients. For behavioral economics research, this task can be implemented in combination with assessments of other reinforcing behaviors (e.g., demand for drugs, food, or other behaviors) to determine whether unique or interrelated mechanisms underlie the reinforcement across behaviors. In clinical practice, this task may provide a clinical assessment of response to treatment in individuals who overvalue (e.g., compulsive sexual behavior) or undervalue (e.g., hypoactive sexual desire disorder) sex and sexual behavior.

Future studies using this task could expand the literature on the behavioral economics of sexuality by relating sexual demand metrics to self-reported sexual behaviors, such as condom use or number of nonmonogamous partners, and standardized measures of sexual behavior. Similarly, determining the relation between discounting of sexual outcomes and sex demand would provide insight into the potential differences between these processes and their unique applicability to other sexual outcomes. Finally, studies evaluating factors that influence or change sex demand, such as sexual arousal or acute drug intoxication, will provide considerable insight into the nature of sexual behavior and how external factors influence decision-making processes for sexual outcomes.