|DETERMINANTS OF UK BOX OFFICE SUCCESS: THE IMPACT OF QUALITY SIGNALS
Department of Economics
Lancaster University Management School
Lancaster LA1 4YX
Acknowledgement: Thanks are due to Edward Dutton and Xiaomei Zhou for their excellent research assistance. We thank David Peel, Sean Perkins of UK Film Council and seminar participants at University of Bath and Lancaster University for helpful and insightful comments.
This paper analyses the roles of various quality signals in the demand for cinema in the United Kingdom using a breakdown of advertising totals by category. Estimation of a three stage least squares model with data for 546 films released in the United Kingdom shows that the impacts of types of advertising on box office revenues vary both in channels and magnitudes of impact. For example, television advertising is shown to affect both the supply and demand sides of the cinema exhibition market. The impacts of television and press advertising are enhanced for films that are nominated for major awards. We also offer a more sophisticated treatment of critical reviews than hitherto by examining the spread (entropy) rather than just the mean rating. We find that the greater the range in ratings from UK newspaper film critics, the higher are box office revenues. UK box office revenues are also positively correlated with average US critic scores and with particular UK critics’ scores.
Films can be considered as experience goods in which the quality, as perceived by the consumer, is only fully revealed after the good is consumed (Nelson, 1970; Shapiro and Varian, 1999). Moreover, films usually yield greater utility to consumers on first viewing than repeat viewing and so should be regarded as inherently perishable. Recently, technological advances in home entertainment, such as wide-screen televisions and DVD players, have increased the quality of playback of films to rent or buy. Also, the film industry has become concerned about the growth of pirate recordings that are viewed at home. Despite the availability of new media for home viewing, cinema admissions have remained buoyant in both the US and UK. The experience of watching a film in a collective environment in front of a large screen, via paid admission, is still attractive to many. There has been a rising trend in UK cinema admissions over the period 1995-2004 with 170 million visits recorded for 2004 (UK Film Council Statistical Yearbook 2004/5). Hence, despite the growth of close substitutes, cinema-going remains a very popular activity, especially for people aged between 15 and 29.
Filmgoers may well have an expectation and an image of the film’s likely quality but this expectation can be either exceeded or unfulfilled. In choosing whether to go to a cinema to watch a film, an individual has available the title, genre, rating certificate, cast and director of the production. Sometimes the film will be a sequel, in which case the experience and record of the previous offering are factors that will determine willingness to view the follow-up product. Another attribute of a film is the size of budget. A large production budget could be taken as a signal of higher quality. Also, appearance of high profile A-list Hollywood stars could generate a positive quality signal as a large number of filmgoers may watch films mainly because of the appearance of a particular star. Indeed, many films are produced explicitly as ‘star vehicles’, to feature an A-list actor in a role that appeals to the public.
The majority of films released in the UK have been released previously in the US, although there is a growing tendency to shorten length of release gap between the two countries. There is also an increased incidence of simultaneous release, ostensibly in an effort to combat piracy. Nevertheless, most films released in the UK bring with them some prior history that could form useful information to the potential filmgoer. Excellent performance in terms of US box office returns could encourage UK filmgoers to watch a film. On the other hand, knowledge that a release has been substantially delayed could be a negative signal in itself.
These indicators may be supplemented by a number of other quality signals. The goal of this paper is to assess the relative importance of different quality signals for success of films released in the United Kingdom. These include the nomination and winning of awards, of which the most famous are the US Academy Awards or ‘Oscars’ (Nelson et al. 2001). We also consider critical reviews. As with many other entertainment products, such as books, music CDs and theatre productions, films are typically reviewed by critics in a number of newspapers and magazines and readers can use these reviews to help form a judgement on their expected utility of viewing the film. Increasingly, these reviews are summarised by easy-to-read numerical scores out of five or 10 to aid readers with high opportunity cost of time. A growing empirical literature is devoted to analysis of the impacts of third-party reviews in diverse settings such as Broadway shows (Reddy, Swaminathan and Motley, 1998) and books (Chevalier and Mayzlin, 2003; Sorensen and Rasmussen, 2004).
This critical review is important to the industry. Where favourable, extracts from critics’ reviews are used in advertising the film, especially in press and outdoor advertising. Some US research has highlighted the helpful role that positive reviews play in raising box office revenues (Eliashberg and Shugan, 1997; Reinstein and Snyder, 2005). As proposed by Eliashberg and Shugan, strong critical acclaim can mean either of two things. First, a positive review could be read by filmgoers and could influence their decision to view a film. Second, a positive review could simply be a proxy for film quality and so critical acclaim is a predictor of film success. Reinstein and Snyder (2005) distinguish empirically between these two functions of critics. They find a small but statistically significant influencing role for ratings offered by two high-profile film critics in the US which they identify as an influence effect, with positive reviews having a significantly larger effect on revenues than negative or mixed reviews.
We offer several innovations missing from the typical treatment of critical reviews in previous work. The first of these is to model the spread, or entropy, of reviews. Studies have typically focussed on average or total critic scores with less attention to dispersion of opinions (Chang and Ki, 2005; Elberse and Eliashberg, 2003; Litman and Kohl, 1989). A relevant question is whether a consensus of reviews, with a narrower range, influence box office returns positively or negatively.1
A further related question, which we consider below, is which critics, if any, command attention from the cinema-going public. Collins et al. (2002) use critical reviews based on a UK film magazine, Empire. This magazine is oriented towards cinema fans and its readership is more specialised and probably less representative of audiences than that of national newspapers. Moreover, the reviews published in Empire tend to be less critical and more enthusiastic than counterparts in several newspapers. Sample selection bias from choice of critics could contaminate estimation of impacts of reviews on box office revenues. Looking at the predictive effect of reviews, it could be the case that some critics are more in tune with the majority of the film-going audience than others. If so, this may mean that greater dispersion of critical opinions could be correlated positively with box office revenues, since some critics are better barometers of public opinion. We address this issue directly below.
The overwhelming majority of films in our sample are American productions that have been released in the US prior to UK opening. It is pertinent to ask whether there are spillover effects from American critical acclaim to British audience response. This can be tested using a measure of aggregate US critic score.
Nelson (1974) argued that advertising can be a signal of quality, as a firm will only be willing to invest in large advertising expenditures if confident about the quality of the product being promoted and the long run sales return to the advertising expenditure. These arguments were then formalised in the models of Kihlstrom and Riordan (1984) and Milgrom and Roberts (1986). However, there are numerous motives for advertising. Schmalensee (1978) developed a model in which advertising can signal product quality, but unlike later analyses he assumes a fixed industry price. Given the low variability in the price of cinema admissions tickets, his model may aid understanding of film advertising and its consequences. An important outcome in his 1978 model is that perverse equilibria may exist, in which low quality firms may have the largest advertising expenditure, sales and profits. While high quality firms may enjoy a greater long run sales effect of advertising, low quality firms can offset this if they have lower costs of production, giving rise to higher price cost margins.
In the film industry, advertising comprises four main types. First, there are advertisements placed in local and national press. Second, films selected for wide-scale national release are typically pre-announced by advertising posters on a large number of billboards and on buses. Third, films are advertised on commercial radio stations. Fourth, a selected number of films will be advertised on commercial television by means of short trailers. The purposes of these types of advertising are to make the public aware that the film is about to be released and to induce the public to visit a cinema to watch the film. The producers, distributors and exhibitors hope that diverse types of advertising will combine to maximise the audience for the film. Further, Nelson (1974) also argued that advertising will help consumers make better informed choices, such that purchases more closely match their preferences. Hence, the variety of advertising media and messages used may reflect attempts being made to ensure that advertising has a coordination role, better coordinating individuals’ film preferences and consumption.2
The effectiveness of alternative forms of advertising, as quality signals, is open to question. It may be the case that advertising is undertaken simply because stakeholders, including producers and actors, expect this to happen.3 Alternatively, Salop and Scheffman (1983) suggest that firms will engage in advertising to raise their rivals’ costs as competitors are forced to copy advertising expenditures to avoid risking market share. Advertising could be undertaken as strategic behaviour; if a rival production is heavily advertised then an appropriate strategic response may be to match the competitor’s effort. At any time, a newly released film must compete for screenings with other productions. As we shall show, it is possible that advertising could affect the supply-side of the cinema exhibition as well as, or instead of, the demand-side as is conventionally proposed in the literature. Although some research has found a positive influence of total advertising on box office revenues (Elberse and Eliashberg, 2003, Prag and Casavant, 1994) few studies have been able to disaggregate total advertising into its components.4 To date, Elberse and Eliashberg (2003) are the only authors to separate the impacts of total advertising on box office revenue through supply-side effects (number of screens allocated to particular films) and demand-side effects (customer response, controlling for number of screens). They did this using a structural three equation model, estimated not just for the US but four European countries as well, including the UK. Within the industry, much effort is made at persuading exhibitors to extend the length of run of a particular film. This effort is more likely to be successful if early box office returns are substantial. Total box office revenue over the life of a run, opening box office revenues and initial number of screens allocated to a film are all endogenous variables that film distributors wish to influence. In this paper, we follow the general approach of Elberse and Eliashberg (2003) by modelling initial screen allocation, opening box office revenues and total (gross) box office revenues as a system of recursive equations.
Various authors have tackled the problem of modelling determinants of box office success. Papers have focused on the roles of stars (Ravid, 1999, Elberse (2005)), R-rated films (Ravid and Basuroy, 2004) and the impact of winning Oscars (Nelson et al. 2001). So far, the literature on determination of box office revenues has given greater attention to the impacts of critical reviews than to the effectiveness of advertising (see Table 1). Moreover, very few studies consider critical reviews and advertisement as joint determinants of box office revenues. In this paper we consider both advertising and critical reviews. The essential research questions analysed here are, first, how and to what extent advertising expenditures can influence box office revenues when other quality signals are available and, second, the extent to which critical reviews are predictors of film success.
We also examine the potential for interaction between advertising expenditure and some of the other quality signals mentioned above. For example, it might be the case that heavy advertising could offset the adverse effects on box office returns from bad reviews. Alternatively, distributors may decide to cut their losses following bad reviews and actually reduce advertising expenditure. Advertising effort may be particularly strong for films that are nominated for awards as distributors see the opportunity to cash in on the greater prestige and recognition that awards bring about. Far-sighted distributors may spot award-winning potential in advance and may promote films more heavily in anticipation of a film gaining awards.
De Vany and Walls (1999, 2002) show that US cinema revenues follow a non-standard (i.e. non-normal) distribution in which the variance of log box office revenues is unbounded and high degrees of skewness are observed.5 The classical assumptions of well-defined mean and constant, finite variance are violated. Accordingly, De Vany and Walls model US box office revenues using various non-standard statistical distributions, such as the Pareto distribution, to handle the excess kurtosis prevalent in box office revenues (see also Walls, 2005, who applies the t-skew distribution). The principal cost of imposing non-standard, but more appropriate, distributions is that it is only possible to estimate a reduced form model. Computational complexity rules out estimation of a structural model. We wish to model the interaction between supply-side and demand-side influences on box office revenues using a system of equations. In response to the critique of De Vany and Walls, we subject the standard errors of our estimated effects to bootstrapping, so as to generate reliable t-statistics for inference (Davison and Hinkley, 1997; MacKinnon, 2002). The advantage of the bootstrap method is that it facilitates computation of robust standard errors in the presence of both non-normality of residuals and heteroskedasticity. Importantly in our context, the bootstrap method does not necessitate imposition of a particular distribution on our dependent variables.
Model Selection and Data
Our model structure is guided by the emphasis placed on certain variables by the industry and is in the spirit of Elberse and Eliashberg (2003). The UK film industry comprises production companies, distributors and exhibitors. We focus on the relationship between distributors and exhibitors. Distributors and producers share the expenses of promoting a film; most of the advertising budget is spent by distributors. The distribution sector is dominated by a small number of major studios.6 In the UK, the exhibition sector is also oligopolistic, with the six largest exhibitors owning 70 per cent of screens (UK Film Council, 2003).
The typical way in which films are released in the UK and US is as follows (see www.launchingfilms.com for a summary described by the UK Film Distributors’ Association). Distributors bring to exhibitors a ‘slate’ of films for release and the exhibitor will rent prints of the film for audience viewing. Opening dates are jointly negotiated between distributors and exhibitors. In the case of the UK, American films are normally released after a delay but some films are simultaneously released. The exhibitor decides how many screens to allocate to a particular film. Since most UK cinemas are now multiplex sites, the exhibitor has the option of using several screens in the same site for a potential blockbuster film. By UK law, the maximum initial period for which a distributor and exhibitor can contract a film’s opening release is two weeks. Consequently, the distributor will work hard to convince exhibitors that a promising film’s release is capable of being extended to mutual benefit of each party. The best signal that an extended run is worthwhile is the size of opening weekend box office revenues. Some films are ‘slow burners’, or ‘sleepers’ in industry parlance, examples in our data set being Billy Elliott and Shakespeare in Love 7. Exceptions apart, the decision to lengthen a release is largely affected by opening box office returns.8 Hence, within the industry, the first objective for a distributor is to secure a large number of opening screens. The second objective, quickly following the first, is to achieve sufficient opening week box office revenue so as to persuade exhibitors to extend a run. Of course, extension of a run for a film carries with it an opportunity cost since it is possible that screens could be allocated to other films with greater revenue potential.
In this framework, we have two demand-side variables: opening weekend box office revenues and total box office revenues. Of course, there are secondary markets for films, most notably purchase and rental of DVD and video copies. We suspend analysis of these increasingly important markets for later research and confine attention to UK cinema box office revenues. On the supply-side of the exhibition market we have number of screens. We treat this as determined prior to opening revenues and so we have a recursive system of variables in which the vectors X denote subsets of indicators of films determined by producers and distributors (prior US box office record, budget, advertising, presence of stars and high-profile directors, genre, certificate, identity of major studio) and Z denotes quality signals not determined by producers and distributors (critical ratings and award nominations).
Log Opening Screens = f(X1) (1)
Log Opening Revenue = g(X2, Z) (2)
Log Total Revenue = h(X3, Z) (3)
The treatment of total revenue as a distinct variable from opening revenue allows us to capture ‘word of mouth’ effects and the results of momentum established in the opening week. Although consciousness of a film’s quality may well be firmly established by quality signals early in release, many filmgoers may delay their attendance due to competing pressures on time or just to avoid long queues at the cinema.
Data on opening screens, US and UK opening revenues and total UK box office revenues were obtained from the box office section of the Internet Movie Database (www.imdb.com), henceforth the IMDB, which is a rich source of information about current and older films. All monetary values were converted to real sterling at 1996 prices. We found consistent data for 546 films released between 1999 and 2003.
3. Explanatory Variables