1 Introduction

Titanic opened to the general public on December 19th 1997 to generally lukewarm reviews. Reviewers felt that the film’s screenplay and characterizations were the weakest elements of the film. Business that first day was steady but not spectacular. Throughout that first weekend, word of mouth increased patronage and, by Sunday, the film was experiencing huge audiences with sold out performances that would last for almost three months. The normal repeat viewing rate for a blockbuster theatrical film is about 5%. The repeat rate for Titanic was over 20%. People were not only telling their friends about the movie, they were returning to see it over and over again.

(Media Awareness Network 2007)

This study analyzes repeat consumption of identical cultural goods. The term ‘identical’ in this study is not meant to embrace the case of individuals who, for example, view a particular film at a movie theater and then see it some time later in another format, such as via a DVD player. Rather, it refers to cases where, for example, individuals (particularly, younger moviegoers) view the same film at a movie theater on more than one occasion i.e. experiencing the cultural good in precisely the same format. Hence the meaning of the term “repeat consumption” is somewhat different from the concept of repeat purchase in the marketing literature (where repeat purchases of a brand occur over time). In Sect. 2 the repeat consumption phenomenon is explored conceptually from both the (fairly obvious) economic viewpoint and the marketing approach. Section 3 outlines the nature of the data and its background. Section 4 sets out the model developed and the key results. Concluding remarks are offered in the Sect. 5.

2 The repeat viewing phenomenon

The time interval (or intervals) between consumption is considered to be an important basis for distinguishing among the various types of repeat consumption episodes identified. Long-time intervals may, for example, be associated with habit-based and ‘cult’ good consumption, the deriving of nostalgia-based utility and a desire to elicit associative memory triggers to unlock specific past life encounters and experiences. Linked episodes of cinema consumption might also be viewed in a rational addiction framework (Becker and Murphy 1988) such that there is a strong impact on current consumption from past consumption, ceteris paribus, as part of a consistent plan to maximize utility over time. However, Cameron (1999) finds that the evidence for this, using pooled cross-section time series data of UK cinema admissions (1965–1983) is not strongly supportive of the precise form of the rational addiction model as presented in Becker and Murphy (1988).

Repeat consumption potentially offers considerable commercial benefits. Boosting audience numbers is a key concern for the producers of new cultural goods as well as the presenting or exhibiting theater enterprises, including movie theaters. A somewhat under-researched means of enhancing first-year box office movie revenue is the encouragement of previous customers or audiences to repeat their consumption of the same cultural good, such as viewing a particular film again. This translates into a willingness to incur the direct costs (in terms of time and out-of-pocket expenditure) associated with each repeat viewing episode of a film in its first run in the cinema.

Repeat consumption can be linked to a number of explanations. First, where ex post utility exceeds expected utility; further consumption may yield higher levels of utility than alternatives. This phenomenon could be enhanced by an unanticipated utility premium (or consumption surprise). Arguably, such an effect is likely to be more prevalent among less experienced consumers. Second, repeat viewing may be motivated by a desire to refine the fuzzy preferences for the good being formed due its intrinsic complexity. For children in particular, repeated viewing of the same film may yield different experiences as different aspects of the film are noticed. For example, in Harry Potter the impact of the visual effects may have overshadowed the attending plot developments and therefore repeat viewing is desired to absorb the plot.

Third, fuzzy preferences may exist due to issues around the ambient consumption conditions during the first consumption episode. For example, the first viewing may have been interrupted by noisy and less considerate members of the audience. An individual’s prevailing mood and level of alertness may also influence the perceived quality of consumption. For example, if the first viewing was undertaken when tired, a repeated viewing when more alert would yield greater benefits. Fourth, repeat viewing can be a manifestation of being part of a momentary or transient fad, fashion, or cultural zeitgeist within which an individual wishes to participate in, or be associated with. Again, children and teenagers are more likely to be susceptible to such a desire. For example, the Harry Potter film clearly benefited from the success of the book series. However, such a momentary fad may not be related to the concept, but possibly the cast of the film. Such fads and fashions may thus be conceptualized in terms of market diffusion with the repeat viewers corresponding to the early adopters and the mainstream being the imitators.

Those who usually choose which film to see on the day or when they are at the cinema may also be more likely to repeat view, if only because their first choice may be sold out, they may then consider that repeat viewing would be more enjoyable than the alternatives on offer. In other words—an inventory ‘stock out’. Alternatively, if the other films available had already been seen, repeat viewing of the most preferred film may offer the greater utility.

Interdependence in consumer utility could be evident in repeat viewing. The experience of viewing a film is likely to be different when accompanied by others. Initial viewing may allow for a better assessment of the likely utility to other individuals. For example, such assessment may relate to determining the suitability of the film for a viewing by children, or to gauge its likely appeal to friends. Who the person is with when they view the film is also likely to influence the type of conversations that ensue about the film. These conversations could themselves lead to a desire for a repeat viewing. This study, however, is unable to discern whether repeat viewers do go alone (e.g., the possible case of lonely geek cultists) or with others.

It is helpful to move beyond standard economic considerations and look at some psychological insights. Repetition is a characteristic feature of play and has been associated with the process of learning in psychological enquiry. For example, Williams (1986) suggests repeatedly watching the same film, and/or extracts from films, turns a relatively passive passtime, with limited potential for learning into an experience which can facilitate both emotional and cognitive change and to which children and teenagers may be more susceptible. This change may, however, not necessarily be desirable. The theory of consumption values (Sheth et al. 1991) offers a potential explanation of repeat viewing. This theory suggests that consumer choice behavior is influenced by five values: functional, conditional, social, emotional, and epistemic. Epistemic value, the capacity to provide novelty and to arouse curiosity, would seem the most likely to influence film choice behavior. When a film creates a strong emotional reaction, seeing that film again may recreate those feelings, in which case the emotional value of a previously consumed film may outweigh the epistemic value of a new film. For the young this may be even more pertinent since for healthy development, both cognitive and emotional, a balance needs to be struck between being exposed to novel experiences, and achieving mastery and understanding of the familiar.

To our knowledge, there are no studies of cinema-going behavior based on consumption values. In the study of leisure behavior, a number of other frameworks have been employed. Elliott and Hamilton (1991) suggest that leisure choices are made based on simple choice tactics, rather than more complex decision processes. They found that ‘doing what friends were doing’, ‘doing something to match the mood the respondent was in’, and ‘doing something for a change’ were the main tactics used. Garlin and McGuiggan (2002) found that choosing a film to watch at the cinema, as opposed to watching a film on video or television entailed a higher level of involvement. Involvement in this context refers to the importance of the purchase to the consumer and the risks entailed in the decision. The greater degree of involvement in viewing films at the cinema perhaps reflects the additional time costs incurred over watching a film on television. They found that expressed preferences and choice were weakly related, perhaps, reflecting that as cinema-going is a social activity and choices are made according to group (rather than individual) preferences. Other studies have focused on predicting or modeling leisure participation based on demographics (see, for example, Prieto-Rodriguez and Fernandez-Blanco 2000; Borgonovi 2004; Collins and Hand 2005; Favaro and Frateschi 2007; Ateca-Amestoy 2008).

While different films will have different levels of appeal for men and women; overall there is no a priori expectation for a systematic gender bias in repeat viewing. For example, in the cinema and video industry audience research (CAVIAR) 2002 survey repeat viewers for Star Wars Episode 2: Attack of the Clones were predominantly male (90%) but repeat viewers for the Britney Spears debut movie Crossroads were predominantly female (72%). For Austin Powers: Goldmember, the gender split was virtually even.

The main research objectives of this article are to identify which sorts of films attract the most repeat view (defined as the films in the CAVIAR data the greatest number of respondents said they saw more than once) and to identify differences between types of film in terms of the incidence of repeat viewing. The second objective is to explore whether the propensity to repeat view can be predicted from statistical (specifically logit) analysis of moviegoer characteristics.

3 Data: background and description

We use data from the 2002 wave of the CAVIAR survey in the United Kingdom. The survey is conducted annually by a leading UK market research company—BMRB International for the Cinema Advertising Association. The primary purpose is to identify the characteristics of movie audiences to help in targeting on-screen advertising. The sample of 3,106 people is representative of the UK population with younger respondents being over-sampled to match the cinema-going population given that these age groups form the core of the moviegoing population. The survey sample is not restricted to cinema-goers and collects information on how often respondents go to the cinema, watches pre-recorded videos/DVDs, which other leisure pursuits they participate in, as well as television viewing, newspaper, and film magazine readership. The survey is conducted using face-to-face interviews. The survey covers frequency of moviegoing and the name(s) of the film(s) seen over a short period. The variables extracted from the data are set out in Table 1.

Table 1 Variable descriptions
Full size table

Most of the variables are self-explanatory, with the exception of the socio-economic grades which are specific to the UK. The grading uses six categories: A, B, C1, C2, D, and E. As very few respondents were in either the highest (A) or the lowest (E) categories; these were merged with categories B and D, respectively. Full definitions are provided by the Market Research Society (1991). We find that films aimed predominantly at children attract the highest number of repeat views, while those aimed at an older audience attract significantly fewer repeat views. We also examine the profile of repeat viewers of the 10 films which attracted the greatest number of repeat viewers in our sample. In general, repeat viewers tend to be younger but the gender balance differs according to film content as Table 2 indicates. For example, the proportion of males and females seeing Harry Potter and the Philosopher’s Stone more than once is roughly balanced (46% male, 54% female) whereas for Star Wars Episode 2: Attack of the Clones, repeat viewers were 90% male.

Table 2 Top 10 films with greatest number of repeat viewers
Full size table

Out of the films listed in Table 2, all except Ice Age and Crossroads ended 2002 among the 10 best performing films at the box office. While it is unlikely that repeat viewing drove their success alone, it must have been a factor.

Figure 1 shows that the majority of repeat viewers in our sample see only one film more than once, while 38% of repeat viewers had seen more than one film at least twice.

Fig. 1
figure 1

Frequency distribution of number of films seen more than once

Full size image

To assess the impact of the BBFC certificate (and hence the presence of sexual or violent scenes in the film) on repeat viewing we grouped the films in our sample according to certificate and calculated the average number of repeat viewers for each certificate (shown in Table 3). Owing to the small number of films in our dataset which attracted repeat views, the U and PG categories were merged, as were the 15 and 18 categories. As Table 3 shows, U/PG rated films attract a higher number of repeat viewers on average. Differences between the average levels of repeat viewing for different categories were highly statistically significant on a Kruskal–Wallis test; a post-hoc test shows that films in the U/PG category attract significantly more repeat viewers than films in the 15/18 category, but the differences between U/PG and 12 and between 12 and 15/18 are not significant (see Siegel 1988, for a description of the Kruskal–Wallis test and its accompanying post hoc test).

Table 3 BBFC certification and repeat viewing
Full size table

4 Model and analysis of results

The objective of this article is to identify the characteristics of repeat viewers. To determine whether demographic characteristics can be used to predict repeat viewing behavior, a logit model is used. If repeat viewing is determined solely by the film, we should not find any significant results; rather the effects should cancel out. If however, there is a subset of cinema-goers who are prone to repeat viewing, the logit model should identify their characteristics. The probability that person i repeat views can be expressed as:

$$ {\text{P}}(\hbox{Y} = 1) = 1/(1 + e^{{ - ^{{(\beta '{\mathbf{X}}_{i} )}} }} ),\,i\, = \, 1 ,\ldots ,\,{\text{N}} $$
(1)

where β is a vector of coefficients to be estimated by maximum likelihood and X i is a matrix of explanatory variables (see e.g. Long 1997 for a derivation of the logit model). We estimate two versions of the model, one including the “frequent” variable and one omitting it. It is arguably more likely that a frequent cinema-goer is more likely to repeat view. They may be more likely to have exhausted the films in their choice set before the films in the cinema change, or they may be more susceptible to the lure of the cinema over other leisure activities. However, including frequency on the right-hand side potentially introduces endogeneity. If the repeat viewing of a film generates an additional cinema visit, repeat viewing will increase viewing frequency. However, if repeat viewing displaces another film, viewing frequency will not change as a result of repeat viewing. Both are plausible, although the latter would require the number of cinema visits made to be fixed. Replacing the frequency with an instrumental variable would avoid the problem, but no appropriate instrument was available in our dataset. A second cause for concern is that the frequent and early viewer variables are potentially collinear; estimating the two models allows the extent of collinearity to be judged.

In the models estimated here there was generally a modest improvement in goodness of fit over the ‘naïve’ predictions obtained from using the existing proportions of repeat viewings in the data to predict repeat viewing. The first set of logit results are presented in Table 4.

Table 4 Repeat viewing model—logit results
Full size table

Table 4 shows the estimated coefficients (B) and the change in the predicted probability resulting from an increase in the value of the (dummy) independent variables from 0 to 1 (see Long 1997, p. 78 for a discussion of interpreting logit results).

One-tailed hypothesis tests show that membership of the two higher socio-economic groups, AB and C1 significantly increase the probability of repeat viewing. C2 however, is not significant, suggesting that those in group C2 are no more likely to repeat view than those in group DE (the base category). However, being in group AB or C1 does not raise the probability of repeat viewing by much (3 and 2.5% respectively). Instead, age seems to be the main influence on repeat viewing behavior; all of the age dummies, with the exception of 35–44 are statistically significant, again based on one-tailed tests with younger cinema-goers being the most prone repeat-viewing. Each of the age dummies, except 35–44 significantly increases the probability of repeat viewing. That the older age dummies are significant might just reflect that cinema-going is a group activity, with parents accompanying their children to the cinema.

As Tables 1 and 2 indicate, the films which gather the majority of repeat views are aimed at a family or teenage audience. Those who see a film soon after it is released (early viewers) are also significantly more likely to repeat view (early viewers are 6% more likely to repeat view). Although those who choose which film to see when they arrive at the cinema (last minute) are also significantly more likely to repeat view, the effect on the probability of repeat viewing is quite small (last minute viewers are 2% more likely to repeat view). The frequent variable does not appear to be collinear with any other variable; introducing it to the model does not influence the other variables. Frequent cinema-goers are also slightly more likely to repeat view, increasing the probability by 2.7%. However, even though this variable is shown to be significant by a one-tailed test, the direction of causation is not clear. Gender does not seem to have an effect as would be expected from the results presented in Table 1. Some films attract female repeat viewers; others attract more male repeat viewers. Location within the UK appears to have no effect.

We also estimated a separate model using only those respondents who are aged between 7 and 19 to see if there are other factors influencing repeat-viewing behavior in this age group. The results are shown in Table 5.

Table 5 Youth (7–19) model results
Full size table

The base category is 15–19 and the results indicate that those aged 10–14 are significantly more likely to repeat view (based on a one-tailed test), such that being 10–14 increases the predicted marginal probability of repeat viewing by 14%. Among Socio-economic groups only C1 is statistically significant, perhaps suggesting that the income constraint is not binding. The early viewer variable is still significant at the (one-tailed) 5% level but the ‘last minute’ variable is not.

4.1 Frequency of repeat viewing

We investigate whether those who repeat view more than once can be distinguished from those who repeat view only once and from those who have never repeated viewing. In effect, we examine the extent to which the frequency of repeat viewing can be predicted from demographic characteristics. Rather than using total frequency, we use three categories of frequency: never, once, and more than once. As there is an inherent order to these categories, we employ an ordered logit model.

In interpreting the results of an ordered logit, the significance of a coefficient is less informative than its impact on the predicted probabilities. As was found in the binary logit model (Table 4), all of the age dummies except for age 35–44 are significant, as are the social group AB and C1 dummies, and the early viewer and last minute variables (at the one-tailed 5% level). An ordered logit produces an estimate of how likely each respondent is to fall into each of the three response categories. In order to assess the effect of a change in a variable, we can calculate how these probabilities change when the variable changes. These changes in the probabilities appear in the three columns headed marginal effects in Table 6. The age dummies have the greatest marginal effect on the predicted probabilities. Taking the age 7–9 variable as an example, falling into this age category increases the likelihood of repeat viewing once by 0.121 and reduces the probability of repeat viewing more than once by 0.186. However, the ordered logit is very poor at identifying those who repeat view once and more than once. The significant variables seem to increase the likelihood of repeat viewing once and decrease the likelihood of repeat viewing more than once. We also estimated a logit model to try to classify repeat viewers into those who repeated view once and those who repeated view more often using the same explanatory variables. None were found to be significant. It seems that this particular aspect of repeat viewing behavior cannot be predicted from purely demographic characteristics. More information on the circumstances of each repeat viewing may improve the model’s performance, but is not available in our dataset.

Table 6 Repeat viewing: an ordered logit approach
Full size table

5 Concluding remarks

This article is, as far as we are aware, the first study of repeat consumption of identical cultural goods, and consequently is predominantly exploratory in nature. Our results suggest that there is a small section of the cinema audience who may see more than one film more than once. This study attempts to explain such audience behavior from an economic and psychological perspective. What distinguishes this group from other cinema-goers and the effect this group has on film performance has not been investigated. It is found that repeat viewers form a definite subset of cinema-goers who tend to be aged 10–14 and to see a film early in its run. On a practical level, if a film is seen soon after it is released, there is more time available to see it again. That those who choose which film to see on the day they see it or when they arrive at the cinema might suggest that films which are seen more than once are chosen in preference to films which are full, or could reflect a preference for something known over something new. The results also suggest that the extent to which the familiar is preferred over the new is related to age. Among older cinema-goers, the new would seem to be preferred to the familiar. There is, however, a limit to what models based on demographics can show. They show who tends to repeat view, but not why they do, nor why a particular film attracts more repeat views. While people with a larger share of economic resources tend to be the repeat viewers, children from higher socio-economic groups also tend to be more educated. As a consequence, it is difficult to determine whether children who repeatedly view the same film in the cinema do so simply because they can afford to, or because they have become used to repeat viewing in other contexts.

From a marketing perspective, repeat viewers may act as sources of word of mouth or as a form of quality signal. Their repeated choice of the film may be taken as an indication that it is worth seeing more than other films, though this aspect too could be usefully informed via the application of supplementary qualitative analyses. Repeat viewers also act as a valuable stream of revenue and seem to play a role in driving the success of some films at the box office.