1 Introduction

Concert ticketing has grown increasingly sophisticated in recent years with many popular musicians now regularly utilizing various types of second-degree ticket price discrimination (Courty & Pagliero, 2014). Most musicians typically offer a range of tickets at different prices for the same concert, a strategy akin to versioning, where prices vary based on observable quality differences. Furthermore, musicians are increasingly experimenting with their ticket offerings by bundling a range of complementary goods and services with typical concert tickets to create what are known as “VIP Packages”.Footnote 1

Utilizing second-degree price discrimination strategies should enable concert producers to manage capacity and enhance revenues by permitting consumers to choose their ideal ticket, based on their preferred combination of ticket characteristics. However, currently, there is limited empirical evidence regarding how consumers value the various attributes of a typical concert ticket, primarily due to challenges in acquiring reliable data.

Additionally, while consumers are likely to agree on the factors that drive quality differences in concert tickets, such as the distance from the stage and the quality of the view, there is likely significant heterogeneity in consumer preferences regarding these characteristics. Furthermore, for characteristics like ticket type—whether it is a general admission versus a reserved seating ticket—opinions likely diverge on what constitutes a more desirable feature. Understanding how consumers value different ticket attributes, and how these values vary among consumers, can help musicians craft better ticket pricing strategies.

This study seeks to bridge this gap by adopting a novel approach to empirically analyze consumer preferences for concert ticket characteristics. Specifically, experimental data on consumer ticket choices were obtained from a carefully designed stated preference discrete choice experiment, coupled with data on respondent characteristics believed to affect concert demand. The choice experiment was crafted by replicating the layout of a typical concert venue with distinct areas that facilitate the use of ticket segmentation. The gathered data were analyzed using choice modeling techniques, providing estimates of how various ticket characteristics influence consumer choice, and highlighting the degree of heterogeneity in these estimates. Finally, these estimates were employed to explore how a typical consumer might react to alterations in ticketing attributes, and to assess the impact of innovations in ticketing practices on consumer welfare within the concert industry.

The analysis of the experimentally generated data provides several key insights. First, estimates of the impact of different venue areas on consumer utility suggest that consumers strongly prefer certain areas, with this preference intensity largely determined by proximity to the stage. However, there is limited evidence of preference differentials between venue areas further from the stage. Additionally, there is limited evidence of within-area preference heterogeneity. This suggests that the extent of both inter- and intra-venue area price discrimination that artists can employ within a given venue is limited.

Second, this paper uncovers new evidence of deterministic preference heterogeneity for reserved seating tickets. The estimates indicate that the average consumer with children, along with older consumers, tend to value reserved seating tickets more than other consumers, who are generally indifferent between reserved seating and general admission tickets. This insight could be useful for artists and promoters in deciding the optimal distribution and pricing of reserved seating compared to general admission tickets for a concert. However, the study does not find any other evidence of systematic variations in preferences or price sensitivities based on observable consumer characteristics. When combined with earlier findings, this implies considerable constraints exist on the use of conventional price discrimination strategies by musicians in setting ticket prices.

Finally, this study offers new insights into how consumers value and benefit from recent ticketing innovations, specifically VIP packages. The estimates indicate that consumers are prepared to pay considerable additional sums for tickets that include a range of VIP services and that these services substantially enhance consumer welfare at current prices. This discovery presents an alternative avenue for musicians to boost concert revenues, which is especially important given the constraints they may face in their ability to utilize standard forms of price discrimination.

The remainder of this paper is structured as follows. Section 2 provides an overview of the relevant literature and delineates this paper’s contribution. Section 3 outlines the methodology employed to generate and model data on consumer ticket choices. Section 4 presents the results of the choice modeling and consumer welfare estimates, and Sect. 5 discusses these results and explores avenues for future research.

2 Related literature

A modest body of literature exists that investigates the determinants of consumer choice in the concert industry. Predominantly, these studies focus on factors affecting concert attendance and aggregate demand. A significant portion, including works by Montoro-Pons and Cuadrado-García (2011), Mortimer et al. (2012) and Christensen (2022), examines the impact of technological advancements in the recorded music industry on overall concert demand.

The studies by Favaro and Frateschi (2007) and Montoro-Pons and Cuadrado-García (2016) are currently the only ones exploring the individual-level factors influencing concert attendance. Favaro and Frateschi (2007) analyze Italian consumers’ preferences and consumption patterns across different music formats, identifying a correlation between consumer characteristics and preference diversity. Conversely, Montoro-Pons and Cuadrado-García (2016) investigate the factors driving concert attendance among Spanish individuals, aiming to identify distinct consumer groups and suggest policies to enhance concert attendance. Although these studies address key aspects of individual-level concert consumption determinants, they do not explore the impact of various ticket offerings on consumer attendance or the variation in consumer preferences across different ticket characteristics.

The relationship between preference heterogeneity and ticket pricing within the concert industry has garnered notable attention. In their analysis of ticket price discrimination, Rosen and Rosenfield (1997), propose that the return to price discrimination is contingent on the extent of preference heterogeneity among various consumer groups. While this hypothesis finds empirical backing at the aggregate level in studies like Courty and Pagliero (2012), it remains unexplored at more granular levels, despite the increasing utilization of price discrimination by popular musicians in setting ticket prices (Courty & Pagliero, 2014; Thompson, 2023).

Additionally, empirical evidence is scarce on the value live music consumers attribute to the novel ticket offerings tied to VIP packages. To the best of the authors’ knowledge, the sole existing evidence on consumer valuation of VIP packages in the concert industry is provided by Halcoussis and Mathews (2007), who report that consumers are prepared to pay significant premiums for concert tickets bundled with opportunities to meet the headline band.

Numerous studies have focused on analyzing consumer preferences for ticket attributes within the broader performing arts context, predominantly in the theater industry. Initially, early studies used discrete choice experiments to investigate both aggregate preferences and the diversity in preferences for theatrical productions at a single theater (Grisolía & Willis, 2011, 2012). Subsequent studies, however, utilized data from actual ticket purchases to model consumer choice behavior and delve into ticket preferences (Baldin & Bille, 2018; Ozhegova & Ozhegov, 2020; Grisolía & Willis, 2016). In certain instances, researchers combined ticket purchase data with stated preference data obtained from discrete choice experiments, enhancing modeling accuracy. Additionally, these research methodologies have been applied to various cultural sectors, analyzing consumer preferences for different film and TV consumption formats (McKenzie et al., 2019), book formats (Crosby, 2019), heritage conservation strategies (Throsby et al., 2021), and new copyright compensation schemes in the recorded music industry (Handke et al., 2016).

This study makes several significant contributions to the existing literature. First, it contributes to the literature on concert demand by investigating the demand for various common ticket characteristics, as well as analyzing the degree of heterogeneity in these estimates within the sample. Moreover, this study delves into the demand for VIP packages, a relatively unexplored area in current research. The growing prominence of these ticketing practices suggests there likely exists significant demand for these services, and that their provision may make live music consumers better off. A deeper understanding of consumer preferences for ticket characteristics, including VIP tickets, can provide valuable insights that musicians and promoters can leverage to optimally price tickets for concerts and improve their concert income. Finally, this study contributes to the literature investigating consumer preferences for ticket attributes in the performing arts more broadly, employing a novel technique to model consumer preferences for general ticket attributes in the absence of actual purchase data.

3 Methodology

3.1 Choice modeling theory

In the literature concerning consumer preferences and choice, researchers commonly utilize a variety of discrete choice models based on random utility theory (RUT). Developed by Lancaster (1966), this consumer choice theory posits that individual utility from consuming a good comprises two components: a deterministic component, associated with the good’s attributes, and a random error component. It is typically assumed that consumers make choices aimed at maximizing their utility. A consumer’s utility can be expressed as follows:

$$\begin{aligned} U_{ij} = V_{ij} + \varepsilon _{ij} \end{aligned}$$
(1)

where \(U_{ij}\) is the utility of consumer i from alternative j in the choice set J, \(V_{ij}\) is the deterministic component of utility, associated with observable attributes of alternative j, and \(\varepsilon _{ij}\) is an unobservable error component.

Assumptions regarding the distribution of the unobservable error component are crucial in determining which choice model to employ. A common assumption is that error terms are independently and identically distributed following an extreme value type 1 distribution, leading to the adoption of the multinomial logit model (MNL). Additionally, the MNL model explicitly models the deterministic component of consumer utility, as outlined in RUT. The MNL model can be expressed as follows:

$$\begin{aligned} U_{ij} = \beta _{j} X_{ij} + \varepsilon _{ij} \end{aligned}$$
(2)

where \(X_{ij}\) represents the vector of attributes of alternative j observed by consumer i, and \(\beta _{j}\) denotes the vector of parameters that measure the average effect of alternative j on consumer utility. Consequently, \(\beta _{j} X_{ij}\) represents the deterministic component of utility as expressed in RUT, whereas \(\varepsilon _{ij}\) constitutes the random error component. In the context of consumer ticket choice, \(X_{ij}\) could include ticket characteristics like price and the venue area to which the ticket grants access.

The MNL is the most basic form of choice model. It estimates the probability that consumer i selects ticket alternative j from the available set J as follows (McFadden, 1973):

$$\begin{aligned} P_{ij}=\frac{e^{\beta _jX_{ij}}}{\sum _{k=1}^{J}e^{\beta _kX_{ik}}}\ \quad \forall \quad j\ne k \end{aligned}$$
(3)

where \(P_{ij}\) denotes the probability that consumer i chooses alternative j in choice set J. The coefficient vector, \(\beta _j\), is estimated by maximizing the corresponding log-likelihood function, which reflects how closely the model replicates the observed choices in the data.

While the MNL model is practical and beneficial for preliminary inferences, its assumption about the distribution of error terms leads to the potentially limiting and unrealistic independence of irrelevant alternatives (IIA) property. The main consequences of this property is that the estimated parameters are assumed to remain constant across the population, and substitution patterns among alternatives are uniform. In essence, this assumption equates to a lack of heterogeneity in individual preferences. Consequently, the MNL model’s applicability is limited in various contexts, given the likelihood of individuals having unique preferences for ticketing options. Should individuals exhibit heterogeneous preferences, the MNL model tends to yield inconsistent parameter estimates and demonstrate poor predictive accuracy.

Advanced choice models mitigate the IIA assumption, enabling a more comprehensive analysis of heterogeneity among consumers of a specific good. Among these, the mixed multinomial logit (MMNL) model is particularly noteworthy. The MMNL model relaxes the IIA assumption by incorporating individual-specific taste parameters, thus allowing for a more flexible distribution of error terms. However, this enhanced flexibility comes with the trade-off of significantly increased computational complexity. In the MMNL model, the utility function for consumer i consuming the jth alternative is expressed as follows:

$$\begin{aligned} U_{ij} = \beta _{ij} X_{ij} + \varepsilon _{ij} \end{aligned}$$
(4)

where \(\beta _{ij} = \beta _{j} + \eta z{_i}\) and \(z{_i}\) is a random draw from some underlying (multivariate) distribution. In this context, the \(\beta _{ij}\) vector of random parameters is allowed to vary across both individuals and alternatives, and the distribution of the error terms are correlated with the individual taste parameters. Consequently, this formulation enables analysts to account for unobserved variations in individual choice behavior, leading to improved model fit, more realistic substitution patterns, and consistent parameter estimates.

In the MMNL model, calculating choice probabilities involve integrating over the distribution of the random parameters. With multiple observations per individual, it is commonly assumed that sensitivities vary among individuals but remain constant within an individual, indicating interpersonal taste heterogeneity. The probability of observing a specific sequence of choices is determined as the product of individual logit formulas:

$$\begin{aligned} P_{ij}=\int _{\beta _{ij}}{\frac{e^{\beta _{ij}X_{ij}}}{\sum _{k=1}^{J}e^{\beta _{ik}X_{ik}}}f(\beta _{ij})}d\beta _{ij} \end{aligned}$$
(5)

where \(f(\beta _{ij})\) represents the density function of the random parameter distribution. Given that the same sensitivities apply across all choices made by a consumer, the integration over the density of \(f(\beta _{ij})\) applies to all of an individual consumer’s choices combined, rather than to each one separately. The coefficients of interest are still be estimated by maximizing the corresponding log-likelihood function. However, as the log-likelihood function lacks a closed-form solution, it must be approximated using simulation methods. This simulated log-likelihood is derived by averaging the simulated probabilities of choosing \(P_{ij}\) over R draws, before summing up over N consumers.

This paper will initially analyze the data using a MNL model to evaluate the validity of the underlying choice data and to gain some initial insights into the deterministic drivers of preference heterogeneity. Following this, the methodology will extend to the MMNL model. The MMNL model estimates will facilitate an analysis of preference heterogeneity within and across various ticket classes, to give insights that may be useful in guiding the development of a structured ticket pricing schedule. Additionally, the parameters derived from the MMNL estimation will be employed to analyze consumer behavior and welfare in response to changes in the availability and characteristics of different ticket attributes, with a key focus on the availability of VIP packages.

3.2 Discrete choice experiment

Analyses of consumer preferences and demand frequently utilize data derived from actual consumer behavior in relevant markets. In the current context, this would include data on concert ticket sales at the level of the individual consumer, or aggregated across ticket classes. However, obtaining relevant revealed preference data are challenging, primarily due to the proprietary nature of concert ticket sales data.

Various stakeholders in the music industry, including trade publications like Pollstar and Billboard, collect and disseminate information on ticket sales and concert revenues. However, these sources of data are inadequate for two primary reasons. First, they lack specific data on ticket prices and sales across different ticket classes for individual events, providing only concert-level average prices and total ticket sales. The absence of this detailed information precludes making inferences about individual preferences for concert tickets in scenarios where tickets of different characteristics are sold for the same event. Second, these sources do not provide details on the demographics of ticket purchasers, information crucial for analyzing how demand fluctuates based on observable consumer traits.

Employing a stated preference discrete choice experiment provides a solution to circumvent these data limitations. Fundamentally, these techniques are designed to replicate real market behavior by presenting consumers with realistic choice scenarios from which they are to choose their most preferred option, thus enabling analysts to gather information on individual preferences. In the context of concert ticketing, this involves designing an experiment that closely approximates the environment of the concert ticket purchasing decision.

Typically, consumers encounter a range of ticket alternatives when selecting tickets for a concert. Each ticket alternative represents a bundle of various ticket characteristics, collectively impacting consumer choice. By carefully selecting pertinent characteristics of concert tickets, it is possible to design a discrete choice experiment that yields valid data on consumer ticket choices for analysis. This data can be used to make inferences about consumer behavior, using choice modeling techniques to systematically analyze the impact of changes in ticket characteristics on consumer choice.

3.2.1 Overview of experimental design

Ensuring that the data yield valid estimates of consumer preferences necessitates careful planning and consideration in the design of a stated preference discrete choice experiment. The choice experiment must be crafted such that it mirrors the real-life decision-making environment as accurately as possible. Therefore, it is crucial to capture the essential characteristics or attributes that significantly influence consumer ticket choice.

The framing of alternatives and the choice architecture presented to respondents are well-known to influence response quality in discrete choice experiments (Schwirplies et al., 2019; Ahi et al., 2023). Additionally, studies have shown that choice experiments focusing on decisions in online environments yield more accurate results with online sampling (Collins et al., 2011). Consequently, considerable effort was invested in designing a choice experiment that closely mirrors the actual online ticket purchasing environment. This objective was achieved by selecting attributes for the ticket alternatives that reflect the characteristics commonly encountered in online ticket purchases. Additionally, the choice experiment itself was conducted online to maintain consistency with the typical ticket purchase environment.

When buying concert tickets from major ticketing platforms, consumers frequently encounter a venue map that delineates areas of access for each ticket, in the case where the venue is segmented. To mimic this experience, a hypothetical venue map was created for the choice experiment. The map was modeled on a medium-large indoor venue situated in Sydney, Australia. This venue features a typical theater-style layout, incorporating elements common to many concert venues. However, an important caveat is that this study’s external validity is contingent on the extent to which other concert venues resemble the layout of this particular venue.

In this study, the area of a venue that a ticket grants access to is an important ticket attribute, facilitating the analysis of preferences for various venue areas and the examination of substitution behavior across these areas. Consequently, the hypothetical venue map developed for the choice experiment also served to delineate the levels of the Venue Area attribute. The number of distinct areas, or levels, for the Venue Area attribute was selected to strike a balance between experimental efficiency and external validity. Following a pilot study, six distinct areas were identified as the optimal number of levels for the Venue Area attribute. Each area differs in terms of distance from the stage, elevation, and viewing angle. Variations in distance and angle were communicated to participants through the hypothetical venue map, as displayed in Table 1. Additionally, participants were informed that Areas 1 and 2 were located on the ground floor, Areas 3 and 4 on an elevated level, and Areas 5 and 6 on the upper level.

Another attribute in this study is Area Type, distinguishing whether a venue area is designated as general admission or reserved seating. These two categories represent the standard ticket types typically offered to consumers, with each usually defining an entire venue area as determined by the event organizer. The key difference is that reserved seating tickets provide access to a specific seat, akin to an airplane ticket, while general admission tickets usually do not guarantee a particular seat. While general admission tickets can be used for seating areas, they are more commonly associated with standing-only areas. These distinctions were explicitly communicated to participants before they commenced the choice experiment.

The experiment also included an attribute indicating whether a given ticket included a VIP package. Incorporating this attribute was essential to model consumer preferences for these emerging ticket offerings. The levels of the VIP Package attribute were carefully chosen to encompass various services that are generally the most appealing aspects of VIP packages offered by popular musicians. Six levels were chosen for the VIP package attribute, each comprising different combinations of three VIP services: meeting the headlining artist, taking a backstage tour, and accessing the venue early to watch the soundcheck.Footnote 2

The last but arguably most crucial ticket attribute included in the experiment was the ticket price. This significance arises from the fact that the price attribute enables the estimation of price sensitivities, and consequently, the determination of elasticity, willingness-to-pay (WTP), and consumer surplus estimates in response to changes in attribute levels. To ensure the range of prices observed in the choice experiment was realistic, a dataset was constructed consisting of concert pricing schedules set by a variety of popular, currently touring musicians. Using this data, the typical prices of the least and most preferable concert tickets were determined. Typically, prices for the least preferable tickets ranged from $50 to $100, while those for the most preferable exceeded $1000. To strike a balance between realism and comprehensive coverage of price ranges pertinent to VIP packages, twelve levels were selected for the Price attribute. The price levels ranged from $100 to $1200, in increments of $100.

Table 1 provides an overview of the attributes and levels used in the choice experiment, along with an example of the seating map presented to respondents.

Table 1 Overview of attributes, levels and hypothetical seating map
Full size table

After finalizing the attributes and their levels, various additional design aspects must be considered. These include determining whether to use labeled or unlabeled alternatives, the number of alternatives to present in each choice set, and the total number of choice sets to be presented to respondents. Aiming to maximize data acquisition while minimizing cognitive burden from additional choice tasks, twelve choice sets with four unlabelled alternatives each were deemed optimal (Caussade et al., 2005). Each choice set presented respondents with four generically labeled (A, B, C, and D) ticketing alternatives, from which they were asked to select their preferred option. Additionally, respondents had an opt-out option available if none of the provided choices were satisfactory.

Finally, an experimental design is required to determine the sets of alternatives that are to be presented to the respondents. Experimental design algorithms optimize the configuration of choice experiments, considering factors such as the quantity of attributes, their levels, the alternatives per choice set, and the overall number of choice sets. For this study, the experimental design was generated using the Ngene software package. An S-efficient design was employed, aimed at minimizing the necessary sample size for statistically significant parameter estimates at a 95% confidence level (Rose & Bliemer, 2013). Drawing on estimates from the pilot study and previous design choices, the algorithm estimated a minimum sample size of 52 respondents. Specific restrictions were applied to attribute combinations to exclude unrealistic and clearly dominated attribute pairings from the design. The use of such constraints means that it is not possible to achieve the desirable feature of attribute level balance. However, it was deemed more important to achieve a realistic design in this context. With these parameters in place, NGene generated an experimental design featuring 12 choice sets, each comprising four alternatives, and showcasing experimental variation in attribute levels both within and between these sets.

3.3 Survey and participant information

The choice experiment was conducted via an online survey presented to participants. This survey comprised various questions aimed at collecting additional data on respondents’ characteristics and providing context to the experiment. Adequately contextualizing the choice environment is crucial in this context to ensure the generation of valid data for analysis. Given the wide variety in musical styles and genres characterising popular touring musicians, alongside substantial preference diversity among consumers, a generic framing of the choice experiment might lead to biased parameter estimates, particularly if consumers are prompted to choose tickets for an event they would not realistically attend.

To address this concern, several introductory questions were presented to the respondents. The first question required respondents to select from a list of 35 popular and currently touring musicians those they would prefer to see in concert, assuming no attendance constraints. These musicians were chosen such as to offer choices from five broad genres of contemporary popular music: Pop, Rock/Alternative, HipHop/RnB, Dance/Electronic and Classical.Footnote 3 A follow-up question then asked respondents to choose their most preferred musician to see live from their previously indicated selections. The chosen musician’s name was incorporated into each choice task, prompting respondents to make ticket selections as if considering attending this musician’s concert. If respondents did not prefer any of the listed artists, they could specify an artist of their choice.

This served two purposes. First, it allowed the prompting question to be appropriately contextualized. Respondents were instructed to assume their selected artist was performing at a nearby venue, that they themselves were available on the concert night, and that no constraints existed that would hinder their attendance. Thus, respondents’ only decision involved selecting a ticket after reviewing available options on the event’s ticketing website. Figure 1 illustrates an example of one choice set encountered by respondents in the experiment, featuring Taylor Swift as the selected artist.

Second, this procedure allowed for the creation of a variable to control for differences in concert ticket preferences associated with musical genre, which was achieved by grouping artists by the five broad genres of popular music as outlined above. Accounting for this is important, for example, if individuals who prefer pop music exhibit different sensitivities to ticket attribute variations compared to fans of other genres. Given that the aim of this study is to assess the preferences of the typical concert goers, rather than how these preferences differ across genres, it is preferable to simply control for these potential differences.

Fig. 1
figure 1

Example choice set

Full size image

Additionally, the survey included questions to gather standard sociodemographic information from respondents, including their sex, age, marital status, income, residential area, educational level, and employment status. A question about whether respondents had children was also included, given its known influence on concert attendance (Montoro-Pons & Cuadrado-García, 2016). Respondents were further queried about their typical consumption of recorded and live music, including their frequency of concert attendance, how often they attended concerts with friends, their daily hours spent listening to recorded music, and their primary medium for consuming recorded music. The aim of these questions was to investigate the existence of systematic relationships between respondents’ sociodemographic traits and music consumption habits and their ticket preferences.

Participants for the study were sourced from a panel organized by a prominent multinational market research firm. Recruitment focused on United States residents who met specific criteria set by a pre-screening question. This question inquired if respondents had attended a concert in a stadium or similar large venue within the past 5 years, serving both to prime the survey, as well as target consumers with a preference for contemporary, popular musicians. This recruitment process resulted in a total of 230 U.S. respondents participating in the choice experiment. Out of these 230 respondents, 15 were excluded from the data set due to response quality issues. The remaining respondents, who each made selections across 12 choice sets, provided a total of 2,580 choice data points for analysis.

Table 2 presents summary statistics detailing the sociodemographic characteristics of the survey respondents. Assessing the representativeness of this sample in terms of sociodemographic characteristics, relative to the U.S. concert attendee population, is challenging given the scarcity of other reliable survey data. Nevertheless, this sample broadly aligns with samples obtained in previous studies conducted in other countries. A summary of music consumption statistics is provided in Table 7 located in Appendix A.

Table 2 Descriptive statistics of survey respondents
Full size table

3.4 Model specification

Data from the discrete choice experiment were analyzed and modeled using the Apollo choice modeling software (Hess & Palma, 2019). Apollo enables users to estimate a broad spectrum of choice models and conduct pre- and post-estimation analyses in the R programming language.

The estimation of choice models begins with specifying the utility function that is assumed to be faced by respondents. For this study, the overall utility function comprised five distinct utility functions: one for each unlabeled alternative and another for the opt-out option. Each utility function was identically specified for each unlabeled alternative. Variations in overall utility are thus determined by the differences in attribute levels among the alternatives. The utility function for the opt-out option consisted solely of an alternative-specific constant (ASC) and its associated parameter, which represents the average utility of non-attendance relative to choosing a ticket.

Controls for the concert’s genre were integrated into each alternative’s utility function. This approach aimed to capture variations in ticket attribute preferences potentially linked to concert genre preferences. This is important, for instance, if individuals who prefer pop music exhibit different sensitivities to ticket attribute variations compared to fans of other genres. Separate dummy variables were assigned to each of the five broad genres of popular music, reflecting the choices of musicians made by respondents during the priming stage. For example, if a respondent selected Taylor Swift as their preferred artist, the Pop dummy variable in their utility specification would be set to 1.

In the utility function for each alternative, a singular parameter was assigned to both the Price and Area Type attributes. The Price parameter conveys the estimated change in marginal utility associated with a $1 increase in the average concert ticket price, assuming all other factors remain constant. On the other hand, the Area Type parameter reflects the estimated change in marginal utility associated with a transition from a general admission to a reserved seating ticket.

Conversely, the Venue Area and VIP Pack variables were dummy-coded in each utility function. The dummy-coding approach assigns a distinct parameter to each level of a categorical variable. This allows for the estimation of nonlinear marginal effects of changes in attribute levels on respondent utility. To ensure the convergence of the estimation algorithm, one parameter of each dummy-coded variable must remain constant. Consequently, the parameters for all other levels of a dummy-coded variable represent utility deviations relative to this baseline level. For Venue Area, the sixth venue area was selected as the baseline level, while for VIP Pack, the baseline level represented the absence of a VIP package.

The MMNL model was estimated with seven additional random parameters to represent the distribution of consumer preferences for the ticket price, each venue area, and the outside option. The incorporation of these random parameters allows for a more comprehensive examination of consumer preferences by letting the estimated coefficients vary across respondents. In other words, this accommodates the existence of differences in ticket attribute preferences across respondents. The random parameters, representing the variance or the spread of the assumed distribution of preferences, must also be estimated. For the Price parameter, a negative log-normal distribution was assumed, limiting the probability density function to negative values and thereby ensuring realistic estimates of individual price sensitivities, which should not be positive. In contrast, the parameters for each Venue Area and the Outside Option were assumed to follow a normal distribution. Parameter distribution estimates were derived from 1000 MLHS (Modified Latin Hypercube Sampling) inter-individual random draws.

Finally, to explore whether differences in ticket attribute preferences are driven by observable consumer characteristics, interactions between respondent sociodemographic characteristics and the Price and Area Type attributes were specified in the MMNL model. Selection of the relevant characteristics was based on simple likelihood-ratio tests, evaluating the relative increase in log-likelihood resulting from the inclusion of additional attributes. Only those interactions that resulted in statistically significant changes in model log-likelihood remained in the final model.

4 Results

4.1 Choice model estimates

Table 3 presents the results derived from the estimation of the MNL and MMNL choice models. The columns in the table are organized according to the type of model estimated. Each set of columns report the coefficient estimates and standard errors for each attribute from each estimated model. Robust standard errors, calculated using the Huber sandwich estimator (Louviere & Woodworth, 1983), were used to correct for correlations in choices among respondents in panel data contexts.

The model statistics section outlines evidence in favor of the inclusion of the seven random parameters in the MMNL model. The MMNL model exhibits a significantly higher log-likelihood than the MNL model, indicating a substantially better model fit associated with the inclusion of these random parameters. Furthermore, the MMNL model’s AIC and BIC values, which penalize the inclusion of additional parameters, are considerably lower than those of the MNL model. This indicates that the increase in complexity from transitioning from the MNL to the MMNL model is warranted by the improvement in model fit.

Table 3 Parameter estimates of MNL and MMNL choice models
Full size table

Across all estimated models, the coefficients associated with the Price attribute consistently demonstrate a statistically significant negative impact on utility. Additionally, the inclusion of a randomly distributed price parameter estimation is necessary, evidenced by the statistically significant estimated dispersion parameter. This signifies that notable heterogeneity in respondents’ price sensitivity exists. The average price sensitivity is slightly higher in the MMNL model relative to the MNL model.

However, interactions between covariates and the Price parameter yielded statistically and economically insignificant coefficients, with likelihood-ratio test results suggesting these variables did not notably enhance the MMNL model’s predictive power. This suggests that, after accounting for standard ticket characteristics, no specific consumer traits lead to varied price sensitivities, implying that all heterogeneity in this sensitivity is random and unobservable to the concert organizer.

The coefficients estimated for each level of the Venue Area attribute reveal a consistent and logical pattern in consumer preferences across different venue areas. Consistently across all models, the coefficients for Venue Area 1 and Venue Area 2 indicate a positive and statistically significant association with utility, suggesting an average consumer preference for proximity to the stage. Conversely, the coefficients for Venue Area 4 and Venue Area 5 lack statistically significant deviation from zero in all models, implying that respondents view these areas as having comparable utility to Venue Area 6. Finally, the estimates for Venue Area 3 vary between the MNL and MMNL models. The MNL estimates indicate a statistically significant effect on the utility of the respondent when moving from Venue Area 6 to 3, but this estimate is no longer statistically significant in either of the MMNL model outputs. Additionally, variance parameter estimates for each venue area’s random parameter suggest varying degrees of within-area preference heterogeneity. These estimates indicate that there exists limits to the ability of musicians to implement standard forms of price discrimination with respect to different areas of a venue.

Estimates for the Area Type coefficient exhibit a consistent pattern between the MNL and baselines MMNL models. In both models, estimates indicate that the average consumer gains utility when switching from a general admission to a reserved seating ticket. However, incorporating specific covariates, namely Children and Age, significantly modifies the relationship between ticket type and consumer utility. These estimates imply that respondents with children, as well as older respondents, exhibit a more pronounced preference for reserved seating tickets compared to the average consumer. The inclusion of these covariates in the MMNL model renders the estimated coefficient for Area Type statistically insignificant from zero. This suggests that the estimated preference for reserved seating tickets is primarily driven by consumers with the aforementioned characteristics and that the average consumer without children, as well as the average younger consumer, are indifferent between reserved seating and general admission ticket types.

There is likely a strong correlation between consumer age and whether they have children. Indeed, the data reveal a relatively strong positive correlation between respondents having children and their age. Despite this, the results suggest that these variables individually drive preferences for reserved seating tickets. This implies that respondents with children and older respondents are likely more willing to pay a premium for reserved seating tickets compared to the average consumer.

The coefficients of the VIP Pack dummy variables quantify the marginal utility of each VIP package relative to the absence of a VIP package. Within the MNL model, VIP Pack 1 and VIP Pack 5 coefficients demonstrate the most significant estimated utility impacts. However, notable differences emerge between the estimates from the MNL and MMNL models. Estimates from the MMNL models consistently reveal greater marginal utilities for all VIP packages compared to not having a VIP package, and exhibit less variation in parameter estimates than the MNL model. Estimates linked to various VIP packages indicate a consumer preference for meeting and greeting the headlining artist, with less interest in backstage tours or early access to event soundchecks within VIP packages.

Overall, the estimates presented in Table 3 are consistent with previous expectations about how changes in the level of each attribute influences consumer utility. Additionally, incorporating random parameters to address random heterogeneity seems essential, as indicated by the significant coefficients for most random parameter distribution estimates and evidence from various model selection criteria. However, evidence linking consumer characteristics to preference heterogeneity is lacking. No covariates have been identified as causing significant variations in price sensitivity among consumers, with only a couple found to influence ticket type preferences.

Additionally, latent class modeling was attempted to systematically investigate the relationship between consumer characteristics and preference heterogeneity, by attempting to group consumers into separate classes based on similar preferences and demographic characteristics. Various latent class model specifications were estimated, with no models consistently converging when the number of classes was set above one. This failure to identify distinct classes of consumers reinforces the fact that there is little evidence of substantial preference heterogeneity associated with consumer characteristics. This is turn implies that limits exists with regards to musicians’ ability to practice price discrimination by targeting specific ticket types at particular consumer groups. Furthermore, the evidence on variation in venue area preferences implies that there are limits to the returns musicians can generate by employing between—and within—venue area price discrimination.

Conversely, the availability of VIP packages seems to significantly impact consumer utility and could offer musicians an alternative means to boost their concert revenues. Gauging how much more consumers are willing to pay for different VIP services or packages can consequently assist musicians in developing effective pricing strategies for these novel offerings.

4.1.1 Between-area price elasticities

While understanding how distance from stage influences consumer ticket preferences is important, a deeper investigation into how consumers react to price changes in each venue area is more important for developing pricing strategies. Such analysis facilitates an understanding of both the limitations and opportunities for musicians to refine their pricing, by examining the extent of consumer substitution between venue areas in response to price changes. This section provides estimates of both own- and cross-price elasticities for each of the six venue areas included in the choice experiment.

Given that the MNL model produces unrealistic substitution patterns due to the IIA assumption, own- and cross-price elasticities for each venue area were estimated using the MMNL model only. Formally, own-price elasticities are computed as:

$$\begin{aligned} E^{P_{in}}_{X_{ikn}} = \dfrac{\partial P_{iq}}{\partial X_{ikq}} \times \dfrac{X_{ikq}}{P_{iq}} = \dfrac{\int _{\beta _{i}}\dfrac{dU_i}{dx_{ik}}P_i(1-P_{i})f(\beta _i)d\beta _i}{\int _{\beta _i}P_jf(\beta _i)\partial \beta _i} \end{aligned}$$
(6)

and cross-price elasticities are computed as:

$$\begin{aligned} E^{P_{in}}_{X_{i'kn}} = \dfrac{\partial P_{iq}}{\partial X_{i'kq}} \times \dfrac{X_{ikq}}{P_{iq}} = \dfrac{\int _{\beta _{i}}\dfrac{dU_i'}{dx_{i'k}}P_{i}P_{i'}f(\beta _i)d\beta _i}{\int _{\beta _i}P_if(\beta _i)\partial \beta _i} \end{aligned}$$
(7)

where \(P_i\) and \(P_{i'}\) are the choice probabilities of alternatives i and \(i'\), respectively. The probability integrals are approximated by demand simulations to a 1% increase in the price for each venue area. Due to the nonlinear nature of the MMNL model, aggregating individual-specific elasticities could result in biased estimates. To mitigate this issue, sample average elasticities were computed using the probability-weighted sample enumeration method (Louviere et al., 2000), defined as follows:

$$\begin{aligned} E^{\overline{P_{i}}}_{X_{i'k}} = \Bigg (\sum _{n=1}^N \hat{P}_{jn}E^{P_{in}}_{X_{i'kn}}\Bigg )/\sum _{n=1}^N\hat{P}_{jn} \end{aligned}$$
(8)

where \(\hat{P}_{in}\) is the estimated probability and \(\overline{P_{i}}\) is the aggregate probability of i, and \(E^{P_{in}}_{X_{i'kn}}\) is the individual elasticity of individual n for alternative i in response to a change in the level of attribute k.

Table 4 presents the estimated own- and cross-price elasticities derived from the MMNL model. Overall, the estimated elasticities demonstrate logical substitution patterns, characterized by negative own-price elasticities and positive cross-price elasticities, indicating that each venue area acts as a substitute for the others. The own-price elasticity estimates reveal inelastic demand for areas 1 and 2, near-unit elasticity for area 3, and elastic demand for areas 4 through 6. This underscores the correlation between proximity to the stage and price sensitivity, where consumers show less sensitivity to price changes in areas closer to the stage, but this sensitivity heightens in areas farther away.

The pattern of estimated cross-price elasticities is suggestive of asymmetric substitution behavior (Blattberg & Wisniewski, 1989), as respondents show a stronger inclination to substitute toward the front rather than the back of the venue in response to price increases in a particular area. This finding is logical, considering that venue areas closer to the stage were estimated to contribute significantly more to utility compared to those toward the back. Moreover, there appears to be stronger demand for adjacent seating areas.

The estimated own-price elasticities for venue areas, apart from those near the stage, are all of magnitudes that suggest that relatively low prices are optimal for most areas of the venue. Increasing the prices of venue areas further from the stage leads to a large decrease in demand that is only partly filled by increased demand for other venue areas, implying that some consumers will choose to not attend the event. Given the desire to sell out concerts, musicians should refrain from setting prices too high for seats and venue areas furthest from stage. This, again, implies a limit on the amount of revenue musicians can derive from ticketing for any given concert.

Certain cross-price elasticities for area 4 are unexpected; estimates indicate consumers that show a higher propensity to substitute toward area 4 following price increases in area 1 or area 2 rather than to other, nearer venue areas. This suggests that area 4 is perceived as a closer substitute to area 1 and area 2 than to other areas. It is not clear why this is the case. One potential explanation relates to the design of the choice experiment. The parameters employed by the experimental design algorithm led to an unbalanced attribute design, where seating areas 4 and 6 featured least frequently in the overall experiment. This design aspect may have contributed to the discrepancies in estimated cross-elasticities observed in the simulations. Nonetheless, the overall picture painted by the elasticity estimates correspond with the evidence displayed in Table 3 and provide greater insight into the consumer preference patterns for specific venue areas.

Table 4 Own and cross-price elasticity estimates for different venue areas
Full size table

4.2 VIP package valuation

4.2.1 Willingness-to-pay for VIP services and packages

Considering the significant contribution of each VIP package to consumer utility, assessing how much consumer are willing to pay for each package would be important. This assessment is crucial in determining musicians’ potential to increase their ticketing revenue through the provision of VIP packages when optimally pricing other ticket attributes.

Table 5 displays the WTP estimates for both individual ticket attributes as derived from the MNL and MMNL models. For the MNL model, WTP estimates for individual attributes are directly computed as ratios of the coefficient of attribute k to the estimated price coefficient, as reported in Table 3. The standard errors for these estimates are determined using the Delta method (Daly et al., 2012). For reliable WTP values from the MMNL model, it is essential to consider the randomly distributed nature of the price parameter. Accordingly, the MMNL model was re-estimated in WTP space using the unconditional parameter estimates (Hensher et al., 2015) to derive WTP values.

Table 5 further presents WTP estimates for combinations of ticket attributes that represent various ticket bundles. VIP packages usually comprise a bundle of services, including opportunities to meet and greet artists, backstage tours, and access to particular venue areas and ticket types. Employing the Delta Method, WTP estimates for various combinations of VIP services were determined. In the MNL context, this involved calculating the ratios of the sum of the relevant attribute coefficients to the price coefficient, and in the MMNL context, this involved aggregating the WTP coefficients.

In each instance, Table 5 displays the marginal WTP for either an individual attribute, or a combination of attributes, relative to their respective baseline levels. For example, the WTP estimates for “Average Ticket” and “Front Section Ticket” reflect the average consumer’s valuation of a typical venue area compared to area 6, and of venue area 1 compared to area 6, respectively. Additionally, the estimates for a ticket bundle that combines the opportunity for an “Artist Meet and Greet” and entry to the “Front Venue Area” signifies the value consumers assign to meeting the headlining artist and experiencing the concert from the front venue area, compared to not meeting the artist and viewing from area 6.

Overall, the WTP estimates appear plausible and indicate that the average concert attendee places significant value on VIP packages. On average, consumers are willing to pay substantial additional amounts for the complete set of VIP services. Furthermore, consumers are willing to pay, on average, an additional $500 for a comprehensive VIP package that includes a reserved seat in the venue area nearest the stage. This suggests that by providing consumers the option to purchase VIP packages for specific concerts, musicians can notably boost their concert earnings without altering admission prices for different venue areas.

The WTP estimates for standard VIP packages are modest when compared to their typical market prices. However, two important facts must be taken into consideration. First, typical VIP packages contain many more goods and services than the hypothetical packages included in this choice experiment. The size of the packages in this experiment was constrained by sample size and financial limitations, which restricted the number of attributes, or services, that could be modeled. This limitation naturally affects the value consumers assign to these packages. Second, the sample of live music consumers who answered the survey likely does not represent the typical purchasers of VIP packages. Respondents were drawn from a population that participates in surveys for small financial rewards and likely have lower income and wealth levels compared to typical VIP package consumers. Therefore, these estimates reflect the average consumer’s valuation of VIP packages, which is likely much lower than the valuation by consumers at the higher end of the income distribution.

Table 5 Willingness-to-pay estimates for various ticket bundles
Full size table

4.2.2 Consumer surplus from VIP packages

Given the substantial WTP estimates for VIP packages presented in the previous section, it is evident that consumers derive a considerable amount of utility from these offerings. This section will formally assess the impact of VIP package offerings on consumer utility by presenting estimates of the consumer surplus derived from consuming VIP packages. This can be approximated by estimating the change in expected consumer surplus from a scenario where VIP packages are available to one where they are not.

Expected consumer surplus represents the indirect utility, quantified in monetary terms, that an individual derives from a specific choice situation, as described by (Train, 2009):

$$\begin{aligned} E(CS_{i}) = \frac{1}{\beta _{price}}\ln \Biggl (\sum ^{J}_{j=1}e^{V_{ij}}\Biggr ) + C \end{aligned}$$
(9)

where \(V_{ij}\) is the contribution to consumer i’s indirect utility of alternative j and C denotes an unknown constant that accounts for a consumer’s absolute level of utility, which remains unobservable. The term \(\sum ^{J}_{j=1}e^{V_{ij}}\) corresponds to the denominator of the logit choice probability formula, with the distinction that the summation extends over all alternatives in the choice set. Consequently, consumers experience higher utility in scenarios where the choice set consists of alternatives that encompass a greater number of desirable attributes.

The change in expected consumer surplus signifies the monetary compensation an individual would need to offset utility changes caused by a modification in the level of an attribute, thereby maintaining their welfare at a constant level. For instance, should the availability of a VIP package for a particular ticket, or set of tickets, be discontinued, the change in expected consumer surplus would measure the required ticket price reduction for an individual to remain indifferent between the original and modified ticket options. Therefore, the change in expected consumer surplus due to attribute level modifications is calculated as the difference in expected consumer surplus before and after the attribute change:

$$\begin{aligned} \triangle E(CS_{i}) = \frac{1}{\beta _{price}} \Biggl [\ln \Biggl ( \sum ^{J}_{j=1}e^{V_{ij}^1} \Biggr )-\ln \Biggl (\sum ^{J}_{j=1}e^{V_{ij}^0}\Biggr ) \Biggr ] \end{aligned}$$
(10)

where \(V_{ij}^0\) refers to indirect utility for consumer n from alternative j in choice task t, in the initial state, and \(V_{ij}^1\) refers to the corresponding indirect utility in the altered state.

Table 6 displays estimates of average changes in consumer surplus associated with changes in the provision of VIP packages by musicians from both the MNL and the MMNL choice models. Indirect utility for each consumer was determined by substituting the estimated parameter values from the choice models and the corresponding attribute levels for each choice situation into Eq. 10. For the MNL model, this approach produces uniform welfare estimates for each respondent across all choice tasks due to the IIA property.

Within the MMNL framework, where random parameter estimates vary among respondents, the estimation of individual consumer surplus becomes more complex. Owing to the random price parameter, each respondent exhibits varying levels of consumer surplus associated with specific attribute levels. To accommodate this heterogeneity, the unconditional distribution of each respondent’s random parameter estimates was averaged, yielding individual-specific taste parameters. Subsequently, the expected consumer surplus for each individual, in each choice situation, was estimated by incorporating the respective attribute levels and individual-specific parameters to compute the individual-by-choice set indirect utilities, both before and after VIP package removal. This information was then employed to determine the average change in expected consumer surplus, first by averaging the individual changes and then, aggregating them across choice situations to ascertain an average measure for the choice experiment.

The results associated with each model’s estimates reveal that consumers gain significant benefits from the availability of VIP packages at existing ticket prices. Should the offering of VIP packages be discontinued in the choice experiment, the estimates suggest that the average consumer would need a significant reduction in the typical ticket price to remain indifferent between alternatives in the original and new choice scenarios. The average estimate from the MNL model exceeds that of the MMNL model, likely due to the greater estimated average price sensitivity within the MMNL models. Nonetheless, the MMNL model’s results highlight significant gains in consumer welfare attributable to the availability of VIP packages. Furthermore, the distribution estimates derived from the MMNL model enable the calculation of a confidence interval for the expected change in consumer surplus estimate. Despite variations in the expected consumer surplus change estimates, the upper and lower bounds of this confidence interval demonstrate that the vast majority of respondents benefit from VIP packages being included in ticketing options. Assuming that the choice experiment’s respondents accurately represent concert goers, these findings indicate that the average consumer benefits significantly from the availability of VIP packages.

Table 6 Estimates of consumer from the provision of VIP packages
Full size table

5 Summary and conclusion

This paper represents a novel attempt to model consumer preferences for concert tickets. In doing so, it assesses the value consumers attribute to the typical characteristics of standard concert tickets and demonstrates the significant value they place on novel ticketing innovations. In turn, this study makes several important contributions.

First, Sect. 4’s estimates reveal that consumers prefer venue areas closer to the stage, but are generally indifferent between areas further from the stage. Although the external validity of these findings is naturally limited, they represent the first empirical evidence of a significant pattern in consumer preferences for concert tickets related to the distance a seat or venue area is from the stage.

Second, this paper offers new evidence highlighting deterministic preference heterogeneity in concert ticket choices. Specifically, it reveals that consumers with children, as well as older consumers, have a notable preference for reserved seating tickets, whereas other consumers display indifference between reserved seating and general admission tickets. This is a unique insight and one which may be of use to musicians looking to make improvements to their ticket pricing strategies. For example, musicians who have a significant portion of their fan base as parents might consider dedicating more venue areas to reserved seating and setting the ticket prices accordingly, particularly given the barriers to attendance experienced by consumers with children Montoro-Pons and Cuadrado-García (2016).

Finally, this paper provides new insights into the valuation of VIP packages and the welfare benefits they offer to the average live music consumer. The findings indicate that, on average, consumers are prepared to pay significantly more for a standard VIP package, and the presence of these packages contributes to considerable consumer surplus at prevailing ticket prices. Consequently, as musicians increasingly offer these packages, they are likely addressing a previously unmet demand from consumers, while also enhancing their concert revenues. While the WTP estimates are somewhat modest, it is important to recognize that these estimates are from a sample of live music consumers that are unlikely to be representative of those consumers that predominantly purchase VIP packages. Therefore, these may be considered as conservative estimates of the true WTP for, and surplus derived from, standard VIP packages.

The results of this study have several important implications for contemporary musicians and concert organizers. First, it appears that there are limitations to the extent of conventional price discrimination that musicians can effectively implement. The scant evidence of preference heterogeneity both within and between venue areas suggests that setting varied prices for different seats within a single area, and across different venue areas, may yield limited returns. Additionally, no standard or observable sociodemographic characteristics were identified as influencing price sensitivity, with only limited evidence pointing to deterministic preference heterogeneity regarding ticket types. Consequently, musicians and concert organizers may face challenges in boosting ticket revenues through the adoption of more sophisticated pricing strategies with respect to standard ticket attributes.

Conversely, my findings suggest that musicians have an alternative avenue for increasing concert revenue through the provision of desirable goods and services bundled in VIP packages. Crucially, the consumer value attributed to VIP packages seems closely linked to the chance of meeting the headlining artist, suggesting that musicians ought to include this service to maximise their earnings from VIP ticket sales. However, the drawback of this approach is that such a service requires significant involvement from the headlining artist, making it personally more demanding.

While this study’s results have the potential to aid musicians in refining their ticket pricing strategies, several important considerations must be acknowledged. First, it is essential to recognize the supply-side implications linked to offering these packages. Given that supplying VIP packages may involve high marginal costs, musicians might be hesitant to offer such services unless consumer valuations significantly outweigh these supply costs.

Second, musicians might hesitate to offer VIP packages due to other non-financial costs, such as concerns about fans’ fairness perceptions. The literature suggests that musicians refrain from fully leveraging their market power in setting ticket prices due to a desire to be seen as behaving fairly (Sonnabend, 2016, 2019). Research investigating the impact of VIP package provision on consumer satisfaction has revealed that the social visibility of such services diminishes satisfaction among non-VIP customers and that this effect is mediated by perceived violations of fairness norms (Fombelle et al., 2015; Cao, 2021). This becomes more salient when the perceived value of VIP services is amplified by VIP consumers’ desire to utilize the experience to facilitate positive social comparisons (Fombelle et al., 2015). Therefore, if the supply of VIP packages is perceived by fans as unfair, musicians might avoid offering VIP packages. However, these concerns could be mitigated if musicians opt to cross-subsidize non-VIP tickets as part of their pricing strategies for VIP tickets. Future studies could concentrate on examining the influence of VIP package provision on fans’ perceptions of fairness and exploring potential remedies for these issues.

Additionally, the focus of this study was on the optimal setting of posted ticket prices. This contrasts with the use of dynamic pricing, an alternative pricing strategy that is increasingly utilized by popular musicians, as well as producers in other entertainment industries (Courty & Davey, 2020; Sweeting, 2012). Such a pricing strategy allows ticket prices to vary based on observed demand, and theoretically allows a musician to maximise revenue while utilizing a venue’s full capacity. While an important innovation in concert pricing, it is infeasible to integrate this pricing strategy in the context of the current study.

However, it remains important to understand how musicians can improve their posted pricing strategies, given the uncertainty associated with the future use of dynamic pricing. Indeed, the use of dynamic pricing may be constrained by consumer concerns associated with perceived fairness, and the disdain consumers typically display for non-traditional allocation methods (Sonnabend, 2019; Roth, 2007).Footnote 4 Indeed, important parallels exist between the contemporary experience with dynamic pricing and that of ticket auctions, the use of which has declined over time despite evidence that it enabled the market to work more efficiently (Budish & Bhave, 2023). If consumers continue to respond with repugnance to non-traditional pricing strategies in the music industry, understanding how musicians can engage in optimal posted ticket pricing when organizing concerts will remain important.

Finally, the applicability of these results is limited by the specific type of musician under consideration. All respondents in the choice experiment made their selections under the assumption that they were attending a concert by a highly popular superstar musician. It is likely that consumer preferences for concert tickets depend on the type of musician in question, and therefore, these results may not be broadly representative of all consumers’ ticket preferences. Exploring how concert ticket preferences differ for early-career and non-superstar musicians would be important for understanding how the musicians can improve their ticket pricing and maximize their success.