The Influence of Communication on Rock Concert Success
ABSTRACT - Regression and discriminant models were formulated to explain variability in several dependent indicators of the success of rock concerts. Predictor variables were promotional activities and subjective interest-in-performer variables. Interpretation of results and issues For further study are also suggested.
Richard Mizerski and Gary M. Mullet (1981) ,"The Influence of Communication on Rock Concert Success", in SV - Symbolic Consumer Behavior, eds. Elizabeth C. Hirschman and Morris B. Holbrook, New York, NY : Association for Consumer Research, Pages: 103-107.
Regression and discriminant models were formulated to explain variability in several dependent indicators of the success of rock concerts. Predictor variables were promotional activities and subjective interest-in-performer variables. Interpretation of results and issues For further study are also suggested. INTRODUCTION Promoting a rock concert is an activity that demands a perspective on marketing unlike almost all other consumer goods and most entertainment/esthetics discussed at this conference. Consider that the product being marketed changes from concert date to concert date, often vary dramatically. Unlike a symphony, which tends to be oriented to the same type of people from one week to the next, rock concert audiences often differ by age, socio-economic, and psychographic orientation. The audience may be note motivated to attend by which friends or types of people are going, the weather, security (e.g., smoking law enforcement), seating arrangements, and the atmosphere surrounding the-event" than they are by the music itself. The timing is extremely important, in that missing the peak of an act's popularity can be the matter of a month, with devastating impacts on the successful sale of that concert. ?revious to this work with a major rock promoter, the authors were involved in using conjoint analysis in an attempt to ascertain, from the consumer's perspective, how various elements were traded off in determining whether to see a particular act. It was found that three general features appeared to be the most salient-concert location, the general type of music, and certain aspects of the act itself. The-work appeared quite promising in that stress values of 0.8% were achieved, with .95 correlation between predicted ranks and actual ranks for sixteen different profiles. Armed with these heartening results, a rock promoter was approached, only to find that an analysis that works well in package goods marketing simply does not lend itself to the real world of the rock promoter. Unlike a product that can be amenable to subtle changes and reallocations, the promoter must work with many unmeasurables and fixed factors. For example, while a promoter may be quite interested in booking an act at a specific time, it is usually the
Regression and discriminant models were formulated to explain variability in several dependent indicators of the success of rock concerts. Predictor variables were promotional activities and subjective interest-in-performer variables. Interpretation of results and issues For further study are also suggested.
Promoting a rock concert is an activity that demands a perspective on marketing unlike almost all other consumer goods and most entertainment/esthetics discussed at this conference. Consider that the product being marketed changes from concert date to concert date, often vary dramatically. Unlike a symphony, which tends to be oriented to the same type of people from one week to the next, rock concert audiences often differ by age, socio-economic, and psychographic orientation. The audience may be note motivated to attend by which friends or types of people are going, the weather, security (e.g., smoking law enforcement), seating arrangements, and the atmosphere surrounding the-event" than they are by the music itself. The timing is extremely important, in that missing the peak of an act's popularity can be the matter of a month, with devastating impacts on the successful sale of that concert.
?revious to this work with a major rock promoter, the authors were involved in using conjoint analysis in an attempt to ascertain, from the consumer's perspective, how various elements were traded off in determining whether to see a particular act. It was found that three general features appeared to be the most salient-concert location, the general type of music, and certain aspects of the act itself. The-work appeared quite promising in that stress values of 0.8% were achieved, with .95 correlation between predicted ranks and actual ranks for sixteen different profiles.
Armed with these heartening results, a rock promoter was approached, only to find that an analysis that works well in package goods marketing simply does not lend itself to the real world of the rock promoter. Unlike a product that can be amenable to subtle changes and reallocations, the promoter must work with many unmeasurables and fixed factors.
For example, while a promoter may be quite interested in booking an act at a specific time, it is usually the"act@ that determines when s/he/they will be in the area. This is often tied to the act's promotion of a current hit, but the timing of when the act appears--before, during, or after the hit peaks--is not usually within the promoter's control. The facility to be used is also not a choice that offers great flexibility, in that some facilities are almost the sole franchise of a specific promoter. Add to this quandary the fact that rock acts are notorious for breaking dares (in fact, many promoters dare not book in excess of 3 months ahead for many acts) and one finds that the neat solutions offered by many .note cognitively-oriented techniques simply are not workable.
Our initial discussions revealed that four factors were generally considered to affect a concert's success; (1) the radio; (2) the "street talk" or general level or interest and discussion about the group; (3) the advertising/promotion effort behind the concert; and (4) the price of tickets as compared to alternate forms of entertainment (e.g., other concerts, events, etc.).
It was felt that a unique perspective on the differential effects of the various communication elements could be gained if data on these elements came from largely archival sources. The following report is an exploratory investigation into this area touching upon consumer esthetics.
Working with a major midwestern concert promoter, an extensive data bank was collected consisting of the promoter's advertising schedule among the various media used, daily ticket sales (usually begun 30 days before the date of the concert), gross receipts, paid attendance, the mix of general and reserved seating, and the capacity of the facility.
The promoter's advertising buys were limited to local radio time and regional print media. The promoter felt that concert exposure on the local [One station was a clear channel station situated between the concert city and a neighboring metro area.] rock-oriented stations was critical, yet had little insight as to the optimal scheduling necessary. His experience suggested that the best mix of stations was based on attempting to match the demographic/psychographic profile of the act's audience with that of the station. Although station formats differed somewhat--Top 40, album-oriented rock (AOR), and gradations between the two--the significant audience profile overlap made intuitive choices less than obvious.
Print ad purchases were split between two major metropolitan newspapers (morning and evening publications), several regional music magazines, and the occasional use of newspaper advertising in other major markets within 100 miles of the concert site. Newspaper purchases were generally limited to notices of the upcoming concert in the weekend edition. All results will note scheduling in terms of three general time periods preceding the concert date. The "short-term" will encompass a period From the day of the concert to 13 days before, and will be reported in terms of "x days before." Advertising from 14 days to one month, an "intermediate" period, will be reported as "x weeks before." Advertising that is significant in the modeling that reflects more than 4 weeks before the concert, a "long-term" period, is reported as "x + month before."
The facility for the concert was viewed by the promoter, and evident in our own preliminary research, as one of the most critical variables in a concert's success. This importance is due to factors such as the acoustics of the concert site, available parking and transportation convenience, and the quality of the seating. Although the promoter booked acts into more than one location, it was felt best to control as many factors as possible; therefore only those concerts that played in the largest and most prestigious site were analyzed. Also, location was not treated as a separate independent (predictor) since most of the concerts for which data were readily available were in this largest facility. Only one or two concerts played in some of the smaller sites and, hence, not enough information on relative location was available for a good dummy variable regression analysis.
This facility's seating capacity varied according to several factors. The set-up of the stage, the electronic amplification and speaker placement, the mix of the general and reserved seating, and security arrangements all affected the ultimate number of seats available. Therefore, seating capacity could vary by as much as twenty percent.
In addition to accounting for the "controllable" promotional activity of the promoter, the cooperation of the radio stations to which the promoter allocated his advertising expenditures (radio stations 1, 2, 3, 4 and misc., nominal identification only) was obtained. In a separate series of interviews, the programming directors and sales managers of each station provided us with data on the dates and number of times each concert was mentioned on the air and the daily airplay of records by the lead act and opening act (if any). This information was compiled from station logs.
In addition, each station developed its own weekly list of an album's and single's rank in the area, which was used as a separate variable. The airplay of an act was viewed as somewhat tied to that act's rank on the station's list, and generally deter-mined its 'rotation" on the playlist. The degree of association between the two depended upon the station's programming format with the top-40 type station revealing the highest association, an~ the A'OR type station having the least. Airplay was also affected by whether the station "sponsored" an act, or lent its name to the event, and usually became the prime beneficiary of the advertising allocation. Finally, airplay of an act was somewhat associated with the programmers' and disc jockeys' personal feelings toward the music and whether that music properly reflected the station's image. a group like Kiss, which has the majority of its appeal among the very young (e.g., preteen to 15 year olds), was rarely played by the AOR stations.
Most of the stations also offered special promotion tie-ins with the concert that may or may not have been related to their "sponsorship" of the concert. These promotions ranged from giving pairs of tickets, albums, and "T-Shirts," to use of a chauffeured limousine and a backstage meeting with the act. This promotion activity was also collected from archival sources.
Finally, in an effort to capture some measure of what the promoter believed was a critical predictor, each programming manager and the promoter were asked to rate the level of "street talk" surrounding each concert. This variable was measured on an eleven-point scale, running from 0 (no street talk) to 10 (the most street talk I've heard). Although this is a crude surrogate, it was felt that the raters were in a unique position to gauge this dimension in that they were in constant contact with the market, record marketing reps, and the record and broadcasting media. In essence, their jobs depended on the accuracy of their perceptions about this factor.
Given this extensive array of communication variables, an attempt was made to find which ones appeared to have the most influence on concert success. Twenty-eight rock concerts were evaluated. These spanned the music spectrum from "hard rock" (e.g., Bob Seger, Boston, Black Sabbath) to "mellow rock" (e.g., John Denver). One of the goals of this research is to produce several predictive models, and to identify lead and lag indicators to help the promoter and the stations better understand the dynamics of the rock concert business. The work is still in the exploratory stages and the present analyses and results are just the first steps in this process.
ANALYSES AND RESULTS
The controllable promotion (advertising) and "uncontrollable" communication variables were initially examined for dependency and interdependency relationships with correlation analysis using cross tabulations as a visual aid. Unfortunately, many of the variables had a substantial number of zero values because most of the promotion and sales variables were of a sporadic nature due to the closing of almost all ticket sales locations on weekends and holidays, and the normal "pulsing" of promotion activity. Therefore, the typical scattergram would look like the idealized figure shown and was accompanied by low correlation coefficients.
IDEALIZED SCATTERPLOT SHOWING ZERO VALUE PROBLEM
It was felt that the elimination of data points for which values on one variable were zero would provide misleading results for two reasons; (1) the zeros were "real," not missing data, because of the ticket office closings and pulsing of the promotion efforts: and (2) the resulting correlations would be unreliable because of the paucity of data points.
Step-wise regression analysis was chosen as the most suitable procedure for modeling the relative influence of the various communication factors (see Table 1) on rock concert success. Given the rather extensive number of variables for which data were collected and the strong possibility of multicollinearity, this procedure offers a heuristic method of determining an order of dominance for individual effects by having successive individual explanatory variables included using a sequential selection criterion of maximum partial correlation with the relevant dependent variable under study.
Two general types of models were developed using the step-wise procedure. The first step of analyses attempted to produce predictive models from among all the independent variables noted in Table 1. The second modeling task was to estimate individual models for the major rock-oriented radio stations, according to what each contributed to concert sales. Only the radio stations were analyzed in this manner because they were clearly the dominant medium in terms of advertising expenditure and promotional activity. This intentional underfitting can lead to substantial model bias- however, given the exploratory nature of this study, we felt the insights gained outweighted this theoretical deficiency.
INDEPENDENT VARIABLES USED IN ANALYSIS
Three variables were used to gauge concert success paid attendance, gross receipts, and per capita receipts --in order to provide some estimate of a consistent pattern of influence and to incorporate some crude control for differences In ticket prices across concerts. This latter factor is quite important in that there would usually be some price at which a concert will sell out, somewhat irrespective of any significant promotion activity, given the very high popularity of the acts booked into the approximately 18,000-seat facility. In addition, almost all of the concerts analyzed offered a combination of general and reserved seating which made it difficult to make a priori decision as to which criterion would be the most appropriate dimension for gauging the success of a concert.
In view of the exploratory nature of this investigation, an arbitrary value of p < .05 was selected as the appropriate significance level for each of the partial regression coefficients. For ease of comparison and interpretation, the resulting equations are presented as a group in Table 2. Note that the independent variables are not shown in order of their inclusion to the models under study and the contribution of each variable to R2 (explained variance) is omitted for proprietary reasons.
Negative signs should not be worrisome as long as one remembers that these are. partial regression coefficients, i.e., they show the effect of a one-unit change in that particular activity given the other activities are presented, on the average. Pragmatically they might be interpreted as over promotion effects, but that is still a tenuous interpretation (very tenuous).
REGRESSION MODELS OF PREDICTIVE COMMUNICATION FACTORS FOR ALL CONCERTS
Six variables are significant in the predictive model of paid attendance although the first two provide the majority of explained variance. The dominant factor was advertising in an intermediate period before the day of the concert on radio station 2, a very popular AOR (album-oriented-rock) station in the metropolitan area. The next most influential factor was placing notices of the concert, averaging a half-page, in newspapers outside of the trading area, and in specialty music magazines serving the geographic region. This element may be unique to the city studied in that the concert facility was located in a somewhat isolated region with similarly large cities (approximately 1,000,000 metro area population) about 60 miles away. Depending on the featured act, each concert could expect about 20% of the audience to come from outside the trading area.
The negative coefficients in this, and other, models would seem to indicate that working with other promotional activities, the particular activity has a deleterious effect on the dependent variable. As the activity, and dollars spent, increases, other things remaining constant, paid attendance shows a relative decrease. Tapping the potential concert-goers mind for a reaction to what might be called overpromotion should be a rewarding venture.
Radio station three's album give-aways was the third factor to enter the model (and this is the only promotion variable to enter this or any other model). Viewing the raw data, this factor appears associated with acts that reflected the station's programming format, and may be a surrogate for implicit support of the station, as possibly evidenced by the influence of its "mentions of the concert" in the per capita model. Still, neither its sponsorship nor its airplay, readily apparent surrogates, proved significantly influential.
The fourth most influential (partial) factor was ads placed in the leading morning newspaper in an intermediate time period before the event. This same factor was the second most important predictor for the per capita model, and when ads were placed even before this period, showed up as a significant predictor for the criteria of gross receipts as well as per capital receipts. The final two predictors accounted for relatively little explained variance, but may have been critical for several of the concerts. "Radio other" consists of the sporadic use of certain stations for several acts who drew special interest from minorities (e.g., the Commodores), country and western acts (e.g., Marshall Tucker), and acts whose audience demographics were skewed toward the very young (e.g., Kiss) or over 30 years old (e.g., Neil Diamond). Note that this type of medium was significant when used a relatively short time before the concert, and was somewhat more influential when modeling gross receipts.
Radio Four is a large clear channel station located between the concert city and a metropolitan area of similar size, and is relatively popular in both areas. Examination of ticket sales by geographic area reveals that an average of 20% of a concert's audience tended to come from that nearby market.
The total amount of explained variance is quite high (R2 = . 821) given six predictor variables, and fails to evidence significant autocorrelation using the Durbin-Watson statistic. This latter statistic was looked at since the data set was analyzed in Lime sequence from early in the calendar year to late in the calendar year; As will be noted later in the discussion, its significance plays a role in one of the other models. The data set, in essence, can be looked at as a (short) time series of observations on the various dependent variables studied.
We found three of the six media that dominated the paid attendance model were also significant predictors when modeling gross receipts. The additional two factors were simply other scheduling of two media vehicles. The first two predictors that dominate, Radio Two and Print Misc., enter in the same order as the previous model, and account for the majority of the explained variance. The third and fourth dominant factors were two short-term advertising schedules on specialty stations.
The final predictor of gross receipts, the long-term scheduling of a concert notice in newspaper one, accounted for very little additional explained variance. The total variance explained (R2 = .84) is again quite high given five predictor variables, and the model does not exhibit significant serial correlation of the residuals.
PER CAPITA RECEIPTS
A total of ten variables were significant predictors of per capita receipts, although only the components of ads and mentions of the concert for radio station three reflect anything but differences in scheduling emphasis from the paid attendance and gross receipts models. The very high level of variance explained (R2 = .92) suggests a rather good predictive model- however, the Durbin-Watson statistic did reveal a significant amount of serial correlation among the residuals. Because- the data input were in time-sequence order of concert, examination of the residuals indicated a tendency for concert goers to "underspend" early and late in the calendar year--possibly a holiday-related effect.
There is some consistency across the regression models representing different measures of concert success. Clearly advertising an intermediate period (somewhere two to four weeks) before the concert on radio station two is the most effective as it is the most dominant factor across all three models in terms of coefficient magnitude. Rather long-term advertising expenditures in specialty music magazines and regional newspapers outside the immediate trading area was the second most dominant predictor for both paid attendance and gross receipts. although it failed to show up as a significant source of variance explained for per capita receipts. The rest of the variables tended to reflect the importance of intermediate advertising expenditures on radio and the use of large print ads placed well ahead of the radio schedules.
OTHER REGRESSION MODELS
Least-squares equations for each radio stations' activities were also produced in an effort to explain the three criterion variables and to help break any multicollinearity between stations. For propriety reasons the specific models are not shown since enumeration of the components would make it relatively easy to identify the stations. However, Table 3 presents the explained variance related to the communications efforts of each station for the three measures of concert success.
INDIVIDUAL STRESS MODELS OF EACH STATIONS' CONTRIBUTION TO CONCERT SUCCESS
For every station except number three, the only factors showing a significant influence were involved in the advertising schedule. As in the overall models presented in Table 2, station two had a significant component of album give-away contests. The relative prominence of station two again shows up with the relatively higher R- values, although the generally low level of explained variance suggests that no single station is completely dominant in explaining/predicting concert success. The comments should be tempered by the possibility of significant modeling bias from consciously underfitting the regression equations.
Finally, again without showing the equations, the three criterion measures representing concert success were partititioned into quartiles. After giving each quartile a nominal level indication, discriminant analysis was applied to the data in an attempt to explain variations in attendance/receipts on the now qualitative scale.
Because of the small number of concerts to work with, validating the discriminant function through the use of a split sample approach was not utilized. A simultaneous discriminant analysis was performed, and the discriminating variables (again using the p < .05 level of significance, if significance has any meaning given the joint distribution function of the independent variables) misclassified, at most, 15% of the concerts. The results are displayed in Table 4.
The discriminant functions appear to be valid predictors although one cannot be overly optimistic about the level of the hit ratios because of the upward bias from classifying the same concerts as used in computing the function, and the results should be viewed with caution. In addition, it should be noted that a concert could fall into different quartiles depending upon the criterion used.
Given these cautionary notes it may still be of interest to discuss the variables, many of which appear for the first time in the data analysis. Radio station two again reveals its dominance first noted in the stepwise analyses although the discriminating ability involves album rank in per capita. Radio station three's album rank is also a discriminating variable for paid attendance, along with short-term advertising effort.
At first glance, the importance of ticket sales on the day of the concert is intuitively confusing. However, it is general knowledge among promoters that some concerts can have a significant number at walk-up sales on concert day. In the present analysis, 60% of the concerts were not sold out by concert day. Of those, an average of 11% of the total ticket sales were done on the day of the concert. Three concerts sold over 25% of their tickets on that day, with one selling 36% of the total!
Album rank appears as a significant discriminator for paid attendance and per capita discriminations. Interestingly, the rankings are from different stations. It is not clear at this point just how this factor manifests itself although it may be operating as a lead/lag factor. Additional analytic work is presently under way in this area.
SUMMARY AND CONCLUSIONS
What did not show up as significant predictors may be as interesting as the communication elements that did! For example, the variables of 'street-talk" and record "airplay" were not evident in any of the analyses. Several potential reasons beyond their relative importance may account for this. Measurement problems may be a factor for the "street talk" component in that the scale may not be appropriate since the "sample" surveyed. Because may not have accurately gauged that factor. "air-play' data came from station logs, measurement error is less likely for that variable, although deviations from logs and suggested playlists is a common occurrence.
Another reason that a variable's significance may be hidden is that the item in question may be correlated with an independent variable that did prove to be a significant source of explained variance (multicollinearitv). It is not intuitively obvious what "street talk" is correlated with, but airplay may be closely tied to album rank and perhaps advertising.
Another surprise was the lack of significant impact of the lead group (47% of the concerts had an opening act). Taking a cautionary note given this exploratory phase, it would not be prudent to totally discount this element's effect, yet one would have anticipated some influence. One reason behind this lack of measurable effect entails the process of how lead groups are determined.
Although these results are admittedly exploratory, the findings do suggest that marketer-controlled communication activity variables dominate the list of predictors for concert success. However, these results cannot be readily generalized even if further work continues to reflect these findings. Different markets may have significant differences in terms of their socioeconomic and psychographic profiles, their tastes in music and the general attitude of the potential market toward rock concerts as entertainment. In addition, the mix of media and the degree of cooperation provided by those media could significantly affect potential generalizability. Further work is continuing and it is hoped next to define significant lead and lag indicators of concert success, and to evaluate the impact of the concert"event" on the other elements of rock music such as the record sales of a featured act.
Richard Mizerski, Federal Trade Commission
Gary M. Mullet, University of Cincinnati
SV - Symbolic Consumer Behavior | 1981
Featured papersSee More
B3. The Effect of Temporal Distance on Online Reviews’ Recommendation Power: The Role of Spontaneous Retrieval and Perceived Trust
Kyu Ree Kim, Seoul National University
Wujin Chu, Seoul National University
E4. Doing Good for Nothing: Motive Inferences from the Probabilistic Profits of Prosociality
Ike Silver, University of Pennsylvania, USA
Jackie Silverman, University of Pennsylvania, USA
Evan Polman, University of Wisconsin - Madison, USA
Sam J. Maglio, University of Toronto Scarborough