The Influence of Variety on the Demand For Bundles of Musical Performances
ABSTRACT - The influence of variety OD subscription ticket sales was studied using box office data provided by a major U.S. opera company covering a five-year period. Performance timing, opera popularity, and production attributes (such as subtitles and premieres) were found to be significantly related to attendance. Measures inspired by Pessemier's concept of structural variety (1985) were found to be insignificant predictors of sales. The results indicate the need for additional validation of variety measures and research on the effect of variety in product assortments on aggregate demand.
Citation:
William J. Havlena and Susan L. Holak (1988) ,"The Influence of Variety on the Demand For Bundles of Musical Performances", in NA - Advances in Consumer Research Volume 15, eds. Micheal J. Houston, Provo, UT : Association for Consumer Research, Pages: 22-26.
The influence of variety OD subscription ticket sales was studied using box office data provided by a major U.S. opera company covering a five-year period. Performance timing, opera popularity, and production attributes (such as subtitles and premieres) were found to be significantly related to attendance. Measures inspired by Pessemier's concept of structural variety (1985) were found to be insignificant predictors of sales. The results indicate the need for additional validation of variety measures and research on the effect of variety in product assortments on aggregate demand. INTRODUCTION In addition to investigating the behavior of consumers with regard to utilitarian products, consumer researchers have examined the choice and consumption of leisure goods. Examples of such products and services include television and radio programs, magazines, sports events, musical performances, and theatrical productions (Hawes 1978; Holbrook, Chestnut, Oliva, and Greenleaf 1984; Holbrook and Lehmann 1981; Tinsley and Kass 1978). The primary benefits of such products are subjective and involve such concepts as fun, enjoyment, and intellectual stimulation (Holbrook and Hirschman 1982). Importance of Considering Bundles Of Items This research has tended to concentrate on the choice of individual items and the personal characteristics of consumers. However, in many cases consumers choose bundles of items, not single objects. For example, the current research examines consumer responses to subscription series of operatic performances. Many arts organizations offer similar subscription programs. Other situations might include book or record clubs, vacation packages, and clothing wardrobes. Even in cases where consumers choose single products, the producer is often faced with the problem of designing a package of goods. These items interact with one another to create a bundle from which the consumer may select individual objects. The producer's task is to create a bundle such that the entire assortment is attractive to the consumer. Thus, a television programmer designs a schedule to provide an assortment of shows which he hopes will appeal to a loyal audience. Similarly, a magazine or newspaper editor includes a bundle of features and articles to interest the reader. It is reasonable for consumers choosing such products to consider the degree of variety provided by the alternative product offerings. For example, in choosing several items of clothing, the purchaser may wish to maximize variety in order to provide suitable outfits for diverse occasions (e.g., the office, weekend trips, athletic activities, etc.). In choosing bundles of artistic products, the consumer may wish to avoid boredom by increasing the amount of variety within the bundle. This paper presents an econometric analysis of audience response to an array of subscription series offered by a major U.S. opera company. The aggregate analysis leads to conclusions regarding the preferences of audiences for this type of performance offering. Several operationalizations of the variety provided by alternative subscription packages are included in the analysis. In addition, some suggestions are offered concerning the extension of this investigation to the broader consideration of product bundles. LITERATURE REVIEW Econometric Analysis Of Arts Attendance A number of studies have been conducted to identify the determinants of aggregate demand for cultural events. The predictors fall into two classes: audience-related variables and-product-related variables. Skrzypczak (1970) used both types of variables to examine concert attendance for 23 major symphony orchestras. The cross-sectional regression indicated that metropolitan population, aggregate education measures, and musicians' salaries were the best predictors. However, due to the cross-sectional focus of the study no examination of repertory preferences was undertaken. On the other hand, Weinberg and Shachmut (1978) included only managerial and performance-related variables in their ARTS PLAN model: timing, type of performance (e.g., chamber music, dance, jazz), single vs. multiple performance, and performer popularity. Although the performing organization used in the study offered subscriptions in addition to single tickets, only overall attendance was analyzed. Cooper and Nakanishi (1978) used performance variables to explain ticket sales at a Los Angeles theater. Subscribers were found to prefer Saturday to weekday performances, matinees to evening performances, and higher-priced seats. Because subscribers purchased tickets for the entire season, individual performance attributes were not considered. The analysis of single-ticket sales was based on average sales for each play. Thus, timing variables were not included. Single-ticket buyers preferred comedies and musicals to dramas, classical to modern plays, and foreign plays. Variables representing cast attributes and newspaper reviews were also found to have a significant effect. Holak, Havlena, dc Kennedy (1986) incorporated both performance-related and audience-related variables in an analysis of attendance at Dallas Opera performances from 1975 to 1983. In addition, subscribers and single-ticket buyers were analyzed separately to ascertain differences in the relative effects of the independent variables on attendance for the two groups. Four variables proved to be significant predictors of overall (total) attendance: day of performance (weekends were more popular than weekdays), time of performance (matinees were more popular than evenings), type of opera (verismo operas were better attended than other types), and relative popularity. Subscribers were significantly more sensitive than nonsubscribers to the day and time of the performance, while nonsubscriber attendance was more heavily influenced by the popularity of individual operas. In addition, subscriber attendance was positively related to ticket price and symphony attendance. The relative unimportance of popularity for subscriber attendance may be explained by the lack of choice alternatives available to subscribers. Both Cooper and Nakanishi (1978) and Holak ct al. (1986) yield little insight into the effect of repertory decisions on subscriber behavior. The planning of the subscription bundle is, however, a crucial decision for the arts organization, given the emphasis on building the subscriber base. Variety-seeking and Product Assortments In one of the few empirical investigations of product assortments, Green and Devita (1974) examined the problem of preferences for item collections in the context of a single meal. Using a conjoint approach, the preference was described in terms of attributes of the component objects. McAlister and Pessemier (1982) provided a review of the variety-seeking literature as it pertains to consumer behavior. A number of studies model variety-seeking in terms of attribute inventories (Farquhar and Rao 1976, Jeuland 1978, McAlister 1982). McAlister (1982) developed a model of variety-seeking behavior in a dynamic framework based on the notion of attribute satiation. However, the emphasis in the model is on sequential choice behavior, rather than on the variety obtained by the selection of a single product. The idea of inventories is less appealing for nonphysical than for physical attributes, and the model does not directly incorporate variety (in the form of a desire for new or different stimuli). These studies tend to measure variety in terms of the number of different objects chosen, downplaying the degree of similarity between objects in the choice set. Pessemier (1985) presented a conceptualization of variety which posited that variety can be decomposed into two separate constructs: structural variety and temporal variety. He provided several functional forms for the measurement of structural variety in a set of objects. These measurements are based on Euclidean distances between objects in a perceptual space. Calculation of the variety in a set thus requires identification of the orthogonal attribute dimensions which characterize the space. Pessemier's operationalization of variety acknowledges the existence of a continuum ranging from no variety to extremely high levels of variety. The amount of structural variety inherent in a set can be described in terms of attributes uncovered by the scaling procedure. Although Pessemier and McAlister (1982) presented some empirical results concerning varied choice behavior, extensive validation of the variety formulations remains to be done. Unlike the individual-level focus of most research on variety, the current application involves aggregate data on sales, a surrogate measure for aggregate preference. Therefore, no individual data on attribute perceptions or object similarities are available to identify the relevant attributes as conceptualized by Pessemier (1985). The study instead seeks to compare the performance of a variety of possible attributes in explaining ticket sales. Several attribute-based variety measures are incorporated using the characteristics discussed by Pessemier (1985). The analysis then allows for the identification of the effects of these potential determinants of variety on preference at the aggregate level. DATA The data used in the current study were provided by a major U.S. opera company and include information on all 191 subscription packages offered between 1979 and 1983. The box office reports include information on total unit attendance and dollar sales for each performance during the period as well as subscription dollar sales for each performance. The reports also include an indication of the percentage of capacity represented by the total attendance for each performance. The average level of attendance was 79%, with only 7% of the performances filled within 1% of capacity. Total subscription sales for each season exhibited considerable variation, ranging from 81% to 133% of the average level over the five-year period. When sales are adjusted to reflect price changes, the total season sales still range from 85% to 113% of the five-season average. Thus, the sales include either new purchasers and/or multiple purchases and do not simply reflect switching behavior by a fixed set of subscribers. Model Specification Based on Holak et al. (1986), the model incorporates two general classes of variables: variables related to the timing and pricing of the subscription series and variables related to attributes of the opera performances themselves. In addition, a third group of variables hypothesized to be related to the level of variety contained in the subscription series was included. Each group of variables will be discussed separately. Timing & Pricing Variables Two dummy variables were used to represent three possible categories of performance times. WKENDMAT refers to matinee performances on Saturdays and Sundays, while WKENDEVE refers to evening performances on the same days. The reference category includes performances given on Tuesday through Friday evenings. The average ticket price for each subscription performance was incorporated into the model as AVGPRICE. Since unit subscription attendance was not available, the price measure is based on the overall average ticket price [(total dollar sales)/(total unit attendance)]. Performance Variables Five performance variables were included which are related to characteristics of the performances rather than the operas. The length of the subscription (typically 3 or 4 operas) was incorporated as LENGTH. Dummy variables were used to indicate the presence within a particular subscription offering of [1] a season opening night performance (OPENING), [2] a premiere of a new work or new production (PREMIERE), [3] a new production of an opera included in the series (NEWPRODN), and [4] an opera provided with simultaneous English translation projected above the stage (SUBTITLE). Opera types: The differing appeal of operatic offerings was incorporated through the use of two groups of measures. The first group includes ten dummy variables which indicate the presence or absence of ten categories of operas in each subscription. The variables, based on categorizations developed by Martorella (1982) and Holak et al. (1986), are: BAROQUE, CLASSIC, BELCANTO, GERMANY19, FRANCE19, ITALY19, VERISMO, EARLY20, CONTEMP, and OPERETTA. BAROQUE includes all operas composed prior to 1750, while CLASSIC contains works written during the latter half of the 18th and early years of the 19 h century (e.g., Mozart operas). BELCANTO includes Italian operas of the early to mid-19th century by composers such as Rossini and Donizetti. GERMANY19 comprises 19th-century operas by German composers (e.g., Weber, Wagner), while FRANCE19 denotes operas from the Romantic period by French composers (e.g., Bizet, Gounod). ITALY19 encompasses such mid- to late-19th century composers as Verdi, with VERISMO incorporating operas characterized by more direct emotional expression (such as the operas of Puccini). EARLY20 includes operas from the first half of the 20th century which were not more appropriately included in the verismo category. CONTEMP operas include all works composed during the past 30 years and contain a fair number of local and world premieres. OPERETTA encompasses a wide variety of works, from the operettas of Strauss, Gilbert and Sullivan, and Lehar to more popular Broadway-type shows. Serious/comedy: Within each category operas may be either comic or serious. This "comedy" attribute was included as a separate variable. COMEDY counts the number of comedies in the subscription. Languages: In addition to the categorization by musical style, the language in which the operas were presented was included as an attribute for each series. Four languages were represented in the operas: ENGLISH, ITALIAN, GERMAN, and FRENCH. The language-based variables are defined as the number of operas in the particular language contained in the series. Because the length of the series is a straight linear combination of the number of operas in each language (LENGTH=ENGLISH+ITALIAN+GERMAN+FRENCH), only three of the languages were included (ENGLISH, GERMAN, and FRENCH). Popularity measures: The second group of variables measures popularity of particular operas not accounted for by the popularity of the category. Four variables were considered in the attempt to capture the effect of subscription popularity on sales. A surrogate measure of overall popularity was obtained by counting the number of times each opera had been performed in the U.S. during the period of the study. Two different methods were used to standardize these raw frequency scores. First, an overall popularity measure for each opera was calculated by standardizing over all the operas in the set. Second, a type-based measure was created by standardizing the popularity of each opera within categories (i.e., using the category popularity mean and standard deviation). The popularity of component operas was posited to affect subscription popularity in several ways. Three measures are based on standardized popularity scores across all categories: AVGOVPOP measures the effect of a series' average popularity on sales, MAXOVPOP measures the effect of the most popular opera in each series, and NOHITS counts the number of operas which had popularity scores more than one standard deviation above the mean for all operas ("hits"). MAXPOPLR is based on the within-category popularity measure and is equal to the highest score across the operas in each series. Although the variables describe different ways in which subscribers may consider the popularity of a series, these four measures are highly correlated due to their common origin (the number of performances of each opera within the United Slates). Therefore, only AVGOVPOP and MAXOVPOP were included in the model; these measure the effect of a series' overall popularity and the additional effect of a particularly popular opera. Variety-related Measures Three types of variety measures were included in the regression model. These measures were inspired by Pessemier's (1985) concept of structural variety. In the absence of perceptual similarity data, objective attribute-based measures were developed. Pessemier's scheme measures variety based on orthogonal dimensions. However, the use of predetermined attributes in the current application rules out the use of independent dimensions. The variables in each group will be presented separately. Type-related: Four variables were developed to measure the amount of type-related variety: VI, RELVAR, and VMODIVHIGH. VI counts the number of different types found in the series, while RELVAR is based on the ratio of types to length of series. VMOD and VHIGH are dummy variables indicating moderate and high levels of variety (as represented by number of opera categories), respectively. The use of dummy variables might be necessary if the impact of variety is hypothesized to be nonlinear. Again, since the variables are highly ,correlated, only RELVAR was incorporated in the model presented here. Serious/comedy: The next variable indicates the presence (or absence) of comedy in a given subscription series (DCOMEDY). While COMEDY measures the level of the attribute, DCOMEDY serves as an additional dummy indicator of variety. That is, DCOMEDY is equal to zero if all the operas are the same type (either comic or serious) and is equal to one if the series includes both comedies and serious operas. Language-related: The last variety variable considers language and country of origin rather than period and musical style. The number of languages -included in a series (NOLANG) was included as a direct measure of language variety. MODEL The full regression model was based on the complete set of variables described earlier and appears in Table 1. A reduced form of the model was also tested. Specific opera attributes (such as categories and languages) were not included in this regression, but the variety indices based on the categories (RELVAR) and languages (NOLANG) were retained. Thus, the second model is nested within the complete model and the marginal explanatory power of the omitted variables can be tested using an F-test. RESULTS The pairwise correlations among the variables were examined for evidence of severe multicollinearity. Only 3 correlations exceeded .8: rVMOD,VHIGH. rVI,VHIGH. and rFRENCH,FRANCE19- Only the last of these involve a pair of variables included in the model. As expected, the popularity measures (AVGOVPOP, MAXOVPOP, MAXPOPLR, and NOHITS) were also highly correlated, with pairwise correlations ranging from .55 to .76. The correlation between the two comedy variables (rCOMEDY,DCOMEDY) was .67. The variables most highly correlated with attendance were SUBTITLE (r=.42), WKENDMAT (r=.35), and AVGPRICE (r=.28). REGRESSION RESULTS FOR FULL AND REDUCED MODELS As anticipated, the three variety measures were moderately correlated, indicating that the dimensions are not truly orthogonal. However, the correlations (rRELVAR,NOLANG=29, rRELVAR,DCOMEDY=.03, rNOLANG,DCOMEDY=25) were low enough to indicate that they are measuring distinct attributes. Nested Models The regression results for both models are presented in Table 1. The full model, which contains 27 independent variables, explained approximately 45% of the variance in attendance (adjusted R2=.36). The reduced model (13 variables) explained 41% of the variance (adjusted R2=.37). An F-test of the two models indicated no significant difference (F14,163=.68). Thus, the addition of the type, language, and comedy variables did not increase the explanatory power of the model. The findings concerning the independent variables are discussed in more detail below. Timing, pricing, & performance variables: Five variables were found to be significant predictors of subscription attendance (at a S% significance level) in both models: WKENDMAT, NEWPRODN, SUBTITLE, AVGOVPOP, and MAXOVPOP. Results for the first variable indicate that weekend matinees are significantly more popular than other performance times. The inclusion of a new production within a subscription (NEWPRODN) also has a positive impact on attendance, as does the use of subtitles (SUBTITLE). The last three variables are all indicators of overall (or historical) popularity. The average popularity of the series (AVGOVPOP) has a positive effect on attendance. However, the marginal effect of the most popular opera on attendance (MAXOVPOP) is negative, implying that one popular opera does not sell a generally unpopular series. Variety-related measures: None of the variety measures reached significance at the 5% level in either model. The number of languages (NOLANG) had a significant negative coefficient in the full model (at a 10% significance level), but the coefficient in the reduced model was very close to zero. Despite the good performance of the reduced model, the variety measures do not seem to improve the explanatory power. The six significant variables in both models are not directly related to the variety included in the subscription series. DISCUSSION Individual Performance Vs. Subscription Models The results of the current study appear to validate the findings of Holak et al. (1986), Cooper and Nakanishi (1978), and Weinberg and Shachmut (1978). Weekend matinees are popular with subscribers, as are new productions and subtitles. The latter two variables were not included in either of the previous studies. Unlike the Holak et al. study, the current research found no significant difference in attendance between weekday and weekend evenings. This may be due, in part, to regional differences between Dallas and the city used in this investigation. As expected, the popularity measures were more important in the present setting because each subscriber was offered a choice of subscription series. In that respect, the situation represents a hybrid of the two groups examined by Holak et al. and the subscriber attendance results closely reflect the combined attendance results found in the earlier study. Insignificance of Variety Measures None of the variety indices produced any significant effects. Two arguments can be developed to explain these findings. First, the measures themselves were developed based on available information according to concepts found in the literature (Pessemier 1985). However, the variables were developed without the use of perceptual data. Although face validity appears good (Martorella 1982), the convergent and predictive validity of these measures has not been tested. Thus, the variables incorporated in the model may not adequately represent perceived variety. Second, the variety incorporated in subscription series may not be related to total attendance. Although individuals may consider variety in their choice of subscription bundles, heterogeneity of preferences may mask the effect at the aggregate level. The nature and level of this heterogeneity may be influenced by audience/population characteristics, such as sophistication and prior experience. These alternative explanations imply different managerial responses. If the measures are not valid, then the impact of variety may be greater than the present study indicates. On the other hand, if the current measures are valid, then variety preferences are sufficiently heterogeneous that no global recommendations can be made concerning variety and optimal subscription design. These results suggest the need for additional validation of measures of the variety construct. The current study produces ambiguous findings concerning the predictive validity of the measure used here. More research is needed to develop and test the convergent and predictive validity of variety measures of the type proposed by Pessemier (1985). Future Research To address these issues, future research is being planned to investigate the concept of variety in opera subscriptions. Perceptual measures of structural variety will be tested at the individual level to establish their validity. These perceptual measures of variety will be compared to objective product attributes to provide a link to the current approach. In addition, structural and temporal variety in other types of product assortments will also be examined to develop an understanding of the nature and importance of variety in product assortments. REFERENCES Cooper, Lee G. and Masao Nakanishi (19783, "Extracting Choice Information from Box Office Records," Performing Arts Review, 8 (2), 193-203. Farquhar, Peter H. and Vithala R. Rao (1976), "A Balance Model for Evaluating Subsets of Multiattributed Items," Management Science, S (January), 528-539. Green, Paul E. and Michael T. Devita (1974), "A Complementarily Model of Consumer Utility for Collections of Items," Journal of Consumer Research, 1 (September), 56-67. Hawes, Douglass K. (1978), "Satisfactions Derived From Leisure-time Pursuits: An Exploratory Nationwide Survey," Journal of Leisure Research, 10 (4), 247-264. Holak, Susan L., William J. Havlena and Pamela K Kennedy (1986), "Analyzing Opera Attendance: The Relative Impact of Managerial Vs. Environmental Variables," Empirical Studies of the Arts, 4 (2), 175-188. Holbrook, Morris B. and Donald R. Lehmann (1981), "Allocating Discretionary Time: Complementarily Among Activities," Journal of Consumer Research, 7 (March), 395-406. Holbrook, Morris B. and Elizabeth C. Hirschman (1982), 'The Experiential Aspects of Consumption: Consumer Fantasies, Feelings, and Fun," Journal of Consumer Research, 9 (September), 132-140. Holbrook, Morris B., Robert W. Chestnut, Terence A. Oliva, and Eric A. Greenleaf (1984), "Play as a Consumption Experience: The Roles of Emotions, Performance, and Personality in the Enjoyment of Games," Journal of Consumer Research, 11 (September), 728-739. Jeuland, Abel P. (1978), "Brand Preference Over Time: A Partially Deterministic Operationalization of the Notion of Variety Seeking," in AMA Educators' Conference Proceedings, ed. Subhashi Jain, 43, 33-37. Martorella, Rosanne (1982), The Sociology of Opera, New York: Praeger. McAlister, Leigh (1982), "A Dynamic Attribute Satiation Model of Variety Seeking Behavior," Journal of Consumer Research, 9 (September), 141-150. McAlister, Leigh and Edgar A. Pessemier (1982), "Variety Seeking Behavior: An Interdisciplinary Review," Journal of Consumer Research, 9 (December), 311-322. Pessemier, Edgar A. (1985), "Varied Individual Behavior: Some Theories, Measurement Methods and Models," Multivariate Behavioral Research, 20, 69-94. Pessemier, Edgar A. and Leigh McAlister (1982), "Varied Consumer Choice Behavior A Theory, Some Empirical Results, and Their Practical Consequences," Report No. 82-111, Marketing Science Institute. Sknypczak, C. S. (1970), "Is There a Niche for a Major Symphony with Its Own Symphony Hall on Long Island?," in Community Support of the Performing Arts--Selected Problems of Local and National Interest, ed. A. Easton, Hofstra University Yearbook of Business, Series 7, 5, 163-202. Tinsley, Howard E. A. and Richard A. Kass (1978), "Leisure Activities and Need Satisfaction: Replication and Extension," Journal of Leisure Research, 10 (3), 191-202. Weinberg, C. B. and K. M. Shachmut (1978), "ARTS PLAN: A Model Based System for Use in Planning a Performing Arts Series," Management Science, 24, 654-664. ----------------------------------------
Authors
William J. Havlena, Southern Methodist University
Susan L. Holak, The University of Texas at Dallas
Volume
NA - Advances in Consumer Research Volume 15 | 1988
Share Proceeding
Featured papers
See MoreFeatured
R2. Brand-to-Brand Communications: How Consumers React to Flattery Between Brands
Lingrui Zhou, Duke University, USA
Katherine Crain, Duke University, USA
Keisha Cutright, Duke University, USA
Featured
Understanding Trust Formation in Peer-to-peer Social Commerce
Lena Cavusoglu, Portland State University
Deniz Atik, University of Texas Rio Grande Valley, USA
Featured
Cueing Backwards: Attention Processes in Multi-Attribute Choices
Antonia Krefeld-Schwalb, Geneva School of Economics and Management
Agnes Scholz, University of Zurich
Ursa Bernadic, Geneva School of Economics and Management
Benjamin Scheibehenne, Geneva School of Economics and Management