Scale
Content
Economies of Scale in the Media Industry
Marc BOURREAU* Michel GENSOLLEN** Jérôme PERANI***
July, 2002
Abstract The economic literature considers that in the media industry production is characterized by high fixed costs and economies of scale. However, empirical evidence seems to belie that idea, for very often the programs and films with large audiences also have high production costs. In this article we propose a program production cost function that reconciles the theory of high fixed costs in content production with empirical evidence. To that end, we develop an empirical estimation of the program cost function in pay-television, and put forward some ideas to explain why production cost is an increasing function of the targeted audience.
JEL Codes: L82; L15. Keywords: Media economics; Program production cost; Economies of scale; Stars.
*
ENST, Department of Economics, 46 rue Barrault, 75634 Paris Cedex 13, France. Tel: 33 1 45 81 72 46. Fax: 33 1 45 65 95 15. Email: [email protected].
**
ENST, Department of Economics and France Télécom. E-mail: [email protected]. Canal Plus. E-mail: [email protected].
***
1
1. Introduction
Since Samuelson (1958), the economic literature1 has considered that production in the media industry (films, TV programs, etc.) is characterized by high fixed costs and economies of scale. For example, costs of creating a TV program schedule will be high when first copy costs were high.2 Once the schedule has been published the incremental cost of physical distribution to an additional consumer is very low or even nil. Broadcast television should therefore exhibit large economies of scale. The same naturally applies to the production of content.
Yet empirical studies have shown that, very often, the more successful a TV program or film, the more it costs. In particular, this correlation has been highlighted by Litman (1983) and Ravid (1999) for cinema. The existence of a relation between production costs and audiences is also recognized in the economic literature. Spence and Owen (1977) think that it may be necessary to increase production budgets to draw larger audiences. Owen and Wildman (1992) consider that a program's production costs can influence its potential audience: “The cost of producing a television program is independent of the number of people who will eventually see it. (The production cost, however, may very well influence how many people will want to see it).”
How can the theoretical idea that programs are goods with high fixed costs – which implies the existence of economies of scale in program production – be reconciled with empirical
1
For a review of the economic literature on television, cf. Owen and Wildman (1992); for the cinema and entertainment industry in general, cf. Vogel (2001).
2
Cf. Waterman (1990: 291) and Owen and Wildman (1992: 151).
2
evidence that program costs are positively correlated with their audiences? The aim of this article is to explain this paradox by proposing, for films and television programs, a program production cost function that reconciles the theory of high fixed costs with empirical evidence suggesting that marginal costs are not negligible.
To this end, we first present an empirical analysis of the cost function of a television channel. More precisely, based on a statistical analysis of program costs and audiences of French payTV channel, Canal Plus, and French cable channels, we show that the production costs of a television channel's program schedule are not unrelated to its maximum audience. This analysis highlights the theoretical problem mentioned above, i.e. that no economies of scale are observed for what is commonly called the "production" of television programs.
To illuminate the issue of possible economies of scale in content production, we propose a program production cost function for cultural goods such as cinema or television. This enables us to account for empirical results (the absence of economies of scale) and logical evidence (production costs cannot depend on ex post audiences). We consider that costs of a program or film increase in relation to the audience targeted by the producer, and that there is a correlation between this expected audience and the actual audience.
In the second part of the article, we consider the nature of the relationship between production costs and targeted audiences, based on the example of cinema and television. In the case of the media, production costs depend on expected audiences through specific, so-called "talent" or "star" production factors which inform and orient demand and impact on social processes needed to develop the consumption of experience goods. These "stars" are able to negotiate remuneration based on fame, i.e. on the mean audiences they draw. Sometimes they are even
3
remunerated on the basis of ex post audiences. Thus, the content production function seems to be more or less comparable to a publishing function.
2. Dependence between production costs of information and audiences
In this section we first assess the cost function of a program schedule for a pay television channel (French pay-TV channel, Canal Plus) and a cable channel. We consider only the content production phase, not the distribution and diffusion phases.3 The question of economies of scale in the media industry thus relates only to the initial production phase. We show that in both examples the program schedule cost is largely dependent on the maximum audience of the channel. Then, we propose a program production cost function that reconciles the theory of high fixed costs with empirical evidence.
2.1. Estimation of production costs for a pay-TV channel
Canal Plus was created in 1984 and it was at the time the first pay-TV channel in France. In 2001, it had 4.9 million subscribers. Canal Plus has production costs relating to commercialization (promotion and advertising costs, cost of decoders, costs of acquisition and management of subscribers). We assess only total programming costs here, that is, the cost of the program schedule. In order to do so, we compare the annual mean number of subscribers4
3
In the case of cinema, we distinguish the production of the negative from the diffusion phase, consisting in the sorting of positive copies and their distribution to cinemas. In the case of television, we distinguish production of the program schedule from its diffusion.
4
The number of annual subscribers of year N is equal to the mean number of subscribers on 31 December of years N-1 and N.
4
and the programming costs of Canal Plus between 1986 and 1996.5 The data are drawn from Canal Plus annual reports between 1991 and 1996. The number of subscribers provides a proxy for the maximum audience, i.e. the audience that the channel could draw if all households able to receive it watched it for the entire duration of the schedule.
We thus obtain 11 values for the program schedule cost, depending on the number of subscribers. The small size of the sample obtained does not allow a precise adjustment of the cost function. We have limited ourselves here to showing the relation between program costs and audiences.
We perform a logarithmic regression of the cost of the program schedule (expressed in millions of 1996 francs) related to the number of subscribers (in millions of subscribers), ln(PROG )=a*ln(ABO )+b . We find that a =1,2297 , b=−10,676 and that the regression is statistically significant (R²=0,986). Figure 1 below presents the cost of the Canal Plus program schedule in relation to the channel's number of subscribers.
[FIGURE 1 SOMEWHERE HERE]
The analysis of Canal Plus program schedule costs suggests a linear increase of total schedule costs with the channel's maximum audience. This finding must be qualified, for the number of points used for this estimation is small. We can nevertheless note that the schedule cost is a function of the number of subscribers.
5
We can consider that program purchasing conditions hardly changed for Canal Plus between 1986 and 1996. During that period Canal Plus was the only pay-television channel in France. The satellite multichannel package TPS was introduced in 1997 only, when it started to compete with Canal Plus for rights to cinema films and football matches.
5
2.2. Estimation of production costs for a cable channel
In the case of cable channels, the same result obtained for Canal Plus is found for their total production costs. We carried out an identical analysis of French cable channels in 1995.6 More precisely, we tried to determine whether there was a relation between the total budget of a cable channel7 and its number of subscribers. The total budget of a cable channel is a good estimate of program costs, in so far as broadcasting costs vary little from one channel to the next (in 1995 these were equal to the annual rate for renting an Astra or Eutelsat analogue satellite capacity).
We thus obtained ten values for ten cable channels (cf. Table 1). We estimated a logarithmic regression of the channels' budgets (in millions of francs) in relation to the number of subscribers of each channel, ln(CT )=c*ln(ABO )+ d . We found that c=0,8999 , d =−7,8118 and that the results obtained were statistically significant (R²=0,8841).
As observed for the cost of the Canal Plus program schedule, we found that the budget of a cable channel in 1995 increased with the number of subscribers to the channel. Once again, this finding has to be qualified in so far as the number of points is limited (ten points).
[TABLE 1 SOMEWHERE HERE]
[FIGURE 2 SOMEWHERE HERE]
6
That year enabled us to take into account the satellite subscribers of each channel since digital multichannel packages were launched the following year, in 1996. We thus consider that the 260,000 mean subscribers of channels present on analogue Canal Satellite, for the year 1995, have no impact on channels' budgets.
7
The total budget is equal to the difference between the channel's income and its profits.
6
2.3. Production cost of information and expected audience
We have presented two estimations which show that the programming cost of a pay-television channel is not a fixed cost in relation to the channel's maximum audience.
Yet in the case of television channels a distinction must be made between the content cost and the price for purchasing that content from copyright owners. This price can be defined by complex contracts and may depend more or less directly on ex post audience scores. In the case of television, for example, broadcasting rights on content are negotiated at prices which depend on an audience estimated ex ante, and sometimes also on the audience measured ex post. The relation that we have shown between programs costs and maximum audiences could thus be related, at least partly, to contractual clauses indexing the purchase prices of these programs over actual audiences or maximum audiences. In other words, we have shown that the cost of a program schedule increases in line with the maximum audience, but does this hold true for the production cost of a program?
Studies undertaken in the film industry provide a positive answer to this question. Ravid (1999) shows a strong correlation between the gross takings of a cinema film – hence, the number of entries – and its production cost. In his study, the production cost explicitly excludes the costs of marketing, distribution and remuneration of participants, which are indexed on actual revenue. Litman (1983) and Litman & Kohl (1989) confirm that result.
Is it possible to reconcile this result with the idea that program production for television or cinema has high fixed costs? This is what we now wish to do, by proposing a formalization for the program production cost function.
7
When they make a program, producers (TV channel, studio, independent producer, etc.) anticipate an audience, the expected audience, a* . We suggest that this audience is not only expected but is actively targeted during the production, and that the program cost, C , is a function of the expected audience, a* . For example, if C is an increasing function of a* ( ∂C ∂a* > 0 ), the program cost will increase in line with the size of the targeted audience. In the following section we shall revert to the relation between content production cost and targeted audience a* .
Once it has been produced, the program will be broadcast and will draw an actual audience, a. If it costs a program producer too much to target a bigger expected audience, then ∂C ∂a* > 0 . On the other hand, for a set value of a* , the production cost is obviously independent of the actual audience measured ex post, a. We then have ∂C ∂a = 0 .
For example, imagine that the producer can perfectly predict the actual audience. In a market equilibrium, since there are rational expectations, the producer's predictions prove to be correct, which means that a * = a .8 The function of production cost with rational expectations is then C ( a ) . Figure 3 illustrates the construction of a function of production cost with rational expectations. In this figure we assumed that the average cost of programming based on actual audiences could be CM ( a* , a ) = C ( a* ) a = a * a . The average production cost with rational expectations is then simply CM ( a, a ) = 1 . Naturally, this example is extreme, for
8
For example, one could imagine the following sequence. Initially the producer, aware of the relation between investment and expected audiences, chooses to invest in a program. Then viewers watch the program or not. In perfect information the producer can predict the viewers' behaviour in the second stage. We shall see below that this expectation is difficult for a film or particular program (variance of a is very great or infinite). Producers and broadcasters must use more complex strategies discussed in the second part of this article.
8
producers are not able to perfectly anticipate the success of the programs they produce, as shown in the following section.
The formalization that we propose enables us to reconcile the idea of program production with high fixed costs and the empirical observation which shows that the function of cost, observed ex post, does not have large economies of scale depending on the actual audience, a. While it is true that the average production cost decreases in line with the actual audience, for a set targeted audience (since ∂C ∂a = 0 ), the average production cost observed ex post is not necessarily strictly decreasing (for ∂C ∂a* > 0 ).
Note that the cost function suggested here for film and TV program production can be likened to a demand function for goods with externalities. Economides (1996) shows that the willingness to pay for an n-th good, when the targeted quantity of goods sold is n * , is p n, n * . This demand function decreases with n and increases with n * , due to the externality effect. The demand function with rational expectations is then p (n, n ) , assuming that at the equilibrium n = n * .
(
)
[FIGURE 3 SOMEWHERE HERE]
3. Relation between programming cost and expected audience
In the previous section we introduced the hypothesis of a relation between the cost of a program, C , and the expected audience of that program, a * . In this part we specify the nature of that relation.
9
If we exclude the case in which certain production factors are remunerated in relation to actual audiences,9 the production costs of program content cannot be linked to the ex post audience unless production factors depend on the expected audience (it is necessary to spend more to target a vaster audience, at least on average), and unless there is a correlation between the expected audience at the time of production and the actual audience in the end.
We are going to argue that particular inputs, introduced into the economic literature under the name of "talent", are probably the main source of the relation between production costs and targeted audience. These "talents", for example film stars, are identifiable by demand, and can serve to inform and orient supply. In so far as they contribute directly to the production of demand they can, after achieving a degree of autonomy, impose modes of remuneration which reflect a sharing of created value rather than payment for a job done. This trend, now distinct in the television industry, has been clearly observed in the case of the film industry.10
In this section we first analyse the nature of the relation between demand targeted ex ante, a * , and demand achieved ex post, a. We then define the notion of "talent" and the economic characteristics of this type of production factor. Lastly, we explain the absence of economies of scale of the production function in empirical studies, by the mode of remuneration of talents.
9
This remuneration must be counted not in production costs but as a share of profits.
10
Especially since 1948, the year of the US Supreme Court ruling in a case against Paramount, which banned vertical integration between production and exploitation. This ruling marked the transition between the Star System of studios and the emergence of independent stars. The greater a star's fame (i.e. their mean capacity to induce a demand: a*), the more advantageous their contracts will be, because of they directly induce audiences. Sometimes these contracts are even indexed on gross takings or on profits (i.e. dependent on the ex post audience: a). The relation between this industrial trend and the mode of remuneration of talents is described by Chisholm (1993) who shows that by losing control of distribution, studios also lost the possibility to invest in the construction of a specific asset, the "brand" of a star when s/he was bound to the studio by a long-term exclusivity contract.
10
3.1. Nature of the relation between a and a*
Due to the extreme uncertainty concerning a, scriptwriter William Goldman commented that in the media "No one knows anything". This uncertainty characterizing the media largely explains the nature of relations between the demand targeted at the time of production ( a* ) and demand finally expressed and satisfied (a).
If these two quantities were closely correlated, information goods would be normal goods;11 costs would depend directly on demand, not by temporal adjustment but by the capacity to accurately predict the public's tastes, and to adjust production ex ante, with precision. If, on the other hand, a and a* were unrelated, the very notion of expected demand would be meaningless (producers would not target a specific demand when making their products). In this case there would be no empirical correlation between production costs and audiences (an extreme case which does exist for certain types of "craft" production, e.g. there is no correlation between the time a writer takes to write a book and the number of readers of the book).
Empirical research in the film industry suggests that the media industry is characterized by extreme variability of ex post audiences, a. In a study of 300 films released between May 1985 and January 1986, De Vany and Walls (1996) show that distribution of audiences,
11
In the case of ordinary goods and services, production costs depend on the demand finally satisfied, in so far as the process of producing products or delivering services can be adapted: supply can be adjusted to demand as it is revealed. Only irreversible initial investments and product-specific R&D induce initial economies of scale. Economies of learning can subsequently develop. For information goods, it is as if all production costs had to be agreed ex ante. Once the content (book, film, program, etc.) is made, satisfaction of demand induces no more production costs, only distribution-exploitation costs. In so far as ordinary goods incorporate more and more R&D and have an increasingly short lifespan, the media economy is becoming a general reference used to explain the functioning of a "dematerialized" economy.
11
takings and profits are of the Pareto-Lévy type, with infinite variance (sometimes even with an infinite mean).12 In this case, even if at the production level an audience a * is targeted and even if, ex post, a correlation is found between a and a * , the result in both cases is unpredictable (no one knows anything). Almost all profits derive from a few rare cases, from the precarious balance between a few disasters13 and a small number of extraordinary successes.
Such variability of demand is related to the very nature of the media. Because information is an experience good (the usefulness of which is not known ex ante) and a network good (more useful when consumed by others as well), the process of acquisition of information by potential consumers, from those who have already consumed, plays a crucial part in the formation of demand. This may be a pure phenomenon of imitation14 or a more complex process of information exchange.15 In any case, we observe cascades (or avalanches!) of decisions leading to extremely uncertain results.
12
These are distributions such as Prob(X>x) is of the order of x• for x big enough with • between 0 and 2. In the case of cinema, we generally find values of • of the order of 1.5 (see De Vany and Lee, 1996).
13
This concentration of income is increasing with time. In the late 1940s, 1% of films (the best) brought in 2% of all income; in the early 1960s they earned 6%, and in 1993, 14% (in the latter case only two films were concerned). See Weinstein (1998).
14 15
Such as consumption based on imitation or fashion; see Bikhchandani et al. (1992).
In the case of ordinary goods and services, production costs depend on the demand finally satisfied, in so far as the process of producing products or delivering services can be adapted: supply can be adjusted to demand as it is revealed. Only irreversible initial investments and product-specific R&D induce initial economies of scale. Economies of learning can subsequently develop. For information goods, it is as if all production costs had to be agreed ex ante. Once the content (book, film, program, etc.) is made, satisfaction of demand induces no more production costs, only distribution-exploitation costs. In so far as ordinary goods incorporate more and more R&D and have an increasingly short lifespan, the media economy is becoming a general reference used to explain the functioning of a "dematerialized" economy.
12
The strategies of media content producers thus naturally concern control of meta-information diffusion processes needed for the formation of audiences, rather than production itself. There are two such types of strategy: • Incorporation of production factors making it possible to attract attention to the content (film, program, book, etc.) and, simply through their presence, to attract a specific audience. This may consist of certain original characteristics of the film, e.g. the novelty of special effects, the fact that it is a sequel to a well-known film or based on a success in another medium (book, TV series). Certain actors or film producers (the stars) thus act as audience prescribers. • Efforts to achieve direct control over demand through vertical integration between production and the various phases of distribution-exploitation. This was the strategy developed by studios until the 1948 Supreme Court ruling (U.S. v. Paramount). When the rarity of means of distribution or a monopoly position enables producers to impose a certain type of content, it becomes possible for them to build up a reputation and then to take advantage of the rent thus created (extracted by means of long-term contracts with audience prescribers).
In this context the notion of "talent" relates less to the ability to create quality content than to the capacity to directly induce a demand through a brand effect. Brand effects make it possible to reduce consumers' uncertainty as to the nature of a new product.
13
3.2. The role of talent in the formation of demand
To describe information goods such as films, economists have introduced the notion of input of the work of "talent" (producers, actors, scriptwriters, etc.) along with input of production work (lighting engineers, cameramen, etc.) and capital input (décor, special effects, etc.).16 The inputs of "talent" are presented as non-substitutable and "rare". They are responsible for a large part of the quality of content, for this quality is considered to be an objective variable judged in the same way by everyone (vertical differentiation).
By introducing the notion of "talent", standard models try to account for the originality of the media industry. Yet stars are not distinguished from other inputs (capital, work) only by their rarity and low level of substitutability. It is also because they can still be recognized by audiences after the production phase that they are able to act on processes of information exchange prior to the consumption of experience goods. Such exchanges of information are all the more necessary in so far as consumers' tastes, which guide their choices, vary widely in time and from one individual to the next.
Thus, the notion of talent is not only related to the "artistic" character of inputs. There are also artistic inputs which, because they are not directly known to consumers, play a part that is no different from that of ordinary work inputs. For example, designers in the luxury industry, because they are unknown to the general public, are remunerated independently of the success of the products they define. Likewise, most actors are not stars, in so far as their participation in a film has no statistically identifiable impact on audiences. For instance, De Vany and
16
See, for example, Crandall (1972) and Owen, Beebe & Manning (1974). Crandall (1972) distinguished only between inputs of "talent" and other input.
14
Walls (1999) identified only 19 stars in all North American actors and film producers during the period 1984-1996.17
The very small number of stars is often explained by a "winner takes all" effect because mediocre talent is seen simply as an imperfect substitute for eminent talent, whereas demand is only for the best.18 An alternative explanation could be proposed: it is the generalized costs of gathering, storing and processing data needed to estimate, ex ante, the usefulness obtained ex post from information goods which ultimately limits the number of signals that can be taken into account. In the case of films, for example, consumers are prepared to devote limited time to processing the meta-information needed to guide their choices. The number of stars identified by De Vany and Walls (1999) could therefore be interpreted in the following way: an informal database on about 20 actors represents a maximum load as far as a consumer's attention is concerned.
The rarity of talent should not be analysed as rarity of an ordinary production factor. Stars in themselves are not rare; there are few of them because consumers' attention is rare and because the time they have to seek information for their consumption of experience goods is limited. Consequently, factors predicting the usefulness of information goods (films, programs, books, etc.) also acquire a value. These factors are used to inform consumers and to
17
Most of whom are actors (Warren Beatty, Sandra Bullock, Jim Carrey, Kevin Costner, Tom Cruise, etc.). The only star producers whose names create a specific demand are Spielberg and Coppola.
18
See Rosen (1981). The author writes, for example: "Lesser talent often is a poor substitute for greater talent. The worse it is, the larger the sustainable rent accruing to higher quality sellers because demand for the better sellers increases more than proportionately: hearing a succession of mediocre singers does not add up to a single outstanding performance. If a surgeon is 10 percent more successful in saving lives than his fellows, most people would be willing to pay more than a 10 percent premium for his services."
15
orient demand. We can thus reasonably talk of an "attention economy"19 in which the essential resource lies in the ability to draw the public's attention to a particular experience good and to reduce uncertainty on the usefulness that can be derived from it.
From this point of view, the rarity of talent is not inherent to processes of formation of demand; it is relative to them. There are only a few actors who please the public, not because the fact of acting well or being likeable is "rare", but because actors orient demand and the current process of ex ante evaluation of ex post quality is costly in terms of attention. If this process changes, for example with the spread of the Internet, the nature of this rarity will also change.
The fact remains that currently stars play an essential part in the success of films, not by substantially reducing the risk taken by the producer (variance of distribution of takings is, in any case, infinite) but20 • by increasing takings: median takings are US$21m for films without stars, compared to US$38m for films with stars; • by substantially increasing the probability of the film being very successful,21 although this probability remains extremely slight.
19
A notion introduced by Michael H. Goldhaber in 1997 during the conference "Economics of Digital Information" (Cambridge, MA, January 23-26, 1997): The Attention Economy: The Natural Economy of the Net. The author defends the idea that the only really important rarity today is consumers' attention. While this argument may seem excessive, the fact remains that, at least for experience goods (information goods of radically new products, for example), the processing of information needed for consumption represents most of the cost borne by the consumer. For this type of good, the time budget is more constraining than the financial budget.
20 21
See De Vany and Walls (1999): it concerns films in the period 1984-1996. For De Vany and Walls (1999), a "hit" is a film with gross takings of over US$50m.
16
3.3. Remuneration of talent
As defined previously, "talent" inputs are characterized by their ability to act on information processes that generate demand for experience goods. These inputs (film stars, TV hosts, etc.), because they are at least partially responsible for the formation of demand, are in a good position to demand remuneration based more or less directly on audiences.
In the case of cinema, it seems that stars have progressively managed to capture most of the surplus they generate. In a study on a random sample of 180 films released between 1991 and 1993, Ravid (1999) found that while stars increase the box-office takings of films, they capture the income thus created and finally play no part in its financial success.
Contracts binding actors to producers vary widely.22 Classically, the following are distinguished: • a fixed payment for each service: actors who are not well-known are remunerated in this way; this was typically the case of silent film actors who, initially, were not known by name in the public at large;23 • payment by salary, with long-term contracts:24 in so far as actors inform audiences on the nature of films and on their potential usefulness, studios directly organize their fame, "packaged" in the form of typical characters;25 this is the "star system";
22 23
See Weinstein (1998).
The first actress to be known by her name was Florence Lawrence. She was employed by Independent Motion Pictures Company (the ancestor of Studios Universal). Before 1910 she was simply called "The IMP Girl".
24 25
Mary Pickford was the first actress to have a long-term contract, from 1913.
There was the "dangerous man", the "folk hero", the "boudoir dandy", the "sinful woman", etc. (see Chisholm, 1993). Actors were related to one of these types, sometimes even to a particular character, as in the case of Johnny Weismuller as Tarzan.
17
•
remuneration dependent on gross income: mostly a percentage of income, in excess of a set level; in these conditions the star shares with the producer a single type of risk – the film's success;
•
remuneration dependent on net profits: in this case, the star shares two types of risk: the risk related to audience size, but also that of not meeting production costs (a risk over which the actor has no control); this type of contract often leads to legal dispute26 and the courts readily consider it as one-sided.
While it may seem natural for stars to try to capture the rents they generate through their presence, the fact of them doing so by means of risk-sharing contracts (dependence on a) rather than fixed remuneration (dependence on a* ), may seem strange. It is generally thought that private individuals are more sensitive to risk than firms. Various attempts have been made to explain this situation.27
First, participation in takings or profits can be an incentive for the actor to work well. In this context of moral hazard it is difficult to see why the sharing exists primarily for stars and not equally generously for all actors or all those who participate in making a film. Moreover, the actor can benefit from private information on the probability of success and the producer buys that knowledge. However, such asymmetry of information is highly improbable and there is no reason for an actor to have better knowledge of risks than the studio (besides, "no one knows anything"!). Finally, the risks are so great in the entertainment industry that risksharing is necessary. It does not stand to reason that stars are more sensitive to risk than those
26
Weinstein (1998) cites recent cases, with the arguments of the Court of Justice, in particular: Buchwald v. Paramount Pictures Corp. (1990); Batfilm Productions v. Warner Bros. Inc. (1994); and Estate of Jim Garrison v. Warner Brothers et al. (1996).
27
See Weinstein (1998) for details of this discussion.
18
who make decisions in the studios. The fact that stars (who are often very rich) frequently give up all up-front payment if that is the only way a film can be produced,28 attests to this.
In any case, talent inputs are remunerated by producers according to their fame (that is, according to the mean ex ante audience, a* ) or according to the actual audience they generate (a). The relation between production costs and actual audience, a, is thus twofold. First, this relation exists through the link between production costs and expected audiences, a* , and through the correlation between a and a* . Second, since the remuneration of talent inputs is indexed on the actual audience, the marginal cost of an additional consumer is not nil. It is an apparent cost since the considered production factor (e.g. a film actor) is not used more because there is an additional consumer. The marginal profit increases and is distributed between the producer and the talent input.
3.4. The role of talent inputs in the case of television
The preceding analysis of relations between producers and talent inputs was based primarily on examples from the "pure" case of cinema. It can also be illustrated by examples from the more complex case of television.
In the case of US broadcast television, Woodbury, Besen and Fournier (1983) analysed a sample of 99 series programd in the US in 1977 and 1978 on the three national broadcast networks, ABC, CBS and NBC. These authors showed that the producers of these series were
28
The most frequently cited examples are: James Stewart for Winchester '73, Georges Lucas for Star Wars, Tom Hanks for Forrest Gump, Dustan Hoffman for Wag the Dog, James Cameron for Titanic, Anthony Hopkins for Amistad, Robin Williams for Good Will Hunting, etc.
19
remunerated in accordance with the "popularity" of the programs they had made (i.e. the ex post actual audience, a). For example, if a program draw a larger audience than expected, the network distributed part of the surplus to the producer.29 More recently, the case of the series ER broadcast on NBC illustrates the relation between audience and program value (i.e. cost for the channel). At the end of the year 1997 the producer of the series, Warner Bros., planned to negotiate on the basis of 10 million dollars per episode.30 In January 1998 the two parties reached an agreement to renew the broadcasting contract on the basis of 13 million dollars per episode.31
In the case of French broadcast television, the market power of talent input is illustrated by the case of host-producers. There are very few of these production companies in which the main shareholder bundles the animation and production of the program, and they are subject to little pressure from potential entrants. The guarantee of audiences, combined with the host's fame, enables them to maintain their position in program schedules far more easily than other types of producer.32
The first half of the 1990s was characterized by bigger audiences of programs produced by these hosts, which de facto enhanced their bargaining power in the market. Figure 4 presents mean costs, for TV channels, of the different short-lived programs broadcast in 1993, in
29
For Woodbury, Besen & Fournier (1983), the network tried to maintain good relations with these producers in order to obtain new programs, and relinquished part of the surplus in order to ensure good performance from them.
30 31 32
Cf. C. Littleton, "Seinfeld exit jolts NBC", Variety, 29 December 1997 See: "The Thursday-Night Massacre", The New York times Magazine, 20 September 1998.
Benzoni and Perani (1996) describe a “virtuous circle” for these host-producers between audience, fame and market power. The bigger the audiences of a host, the greater her/his fame among televiewers and distributors, for the latter can maximize their advertising income by using her/his services. A host's bargaining power with distributors increases along with her/his fame: that is the virtuous circle of success. A star-host can then bargain with distributors on the basis of her/his qualities as a host and the services of her/his production company.
20
relation to the audiences of those programs.33 As shown, programs with high mean costs are presented by famous hosts, whether their audiences are small (J.-M. Cavada with "La Marche du Siècle"; B. Pivot with "Bouillon de Culture") or large (J.-P. Foucault with "Sacrée Soirée" and M. Drucker with "Stars 90").
This analysis shows that in the case of television as well, talent inputs capture part of the surplus they create, in terms of subscription revenue or advertising revenue generated by the program.
[FIGURE 4 SOMEWHERE HERE]
4. Conclusion
The economic literature generally considers that the media industry is characterized by high fixed costs and economies of scale. More precisely, the content production phase is considered to have fixed costs. In this article we have shown that this analysis does not correspond to the findings of empirical research. In the case of both television and cinema, there seems to be a correlation between the audience of a program or film and its production cost. This property has been demonstrated by estimating the cost function of pay-television in France, which suggests that the production cost of a program is an increasing function of audience.
To illuminate this paradox we introduced a cost function of a program or film, in which production costs depend on the targeted audience at the time the content is produced. This
33
In certain cases the "costs" of these programs for channels can be purchase prices and not production costs.
21
cost function is logically independent of the actual audience. Once a program has been produced its cost cannot depend on the number of viewers, but it can increase with the targeted audience and thus statistically with the actual audience when these two variables are correlated.
The paradox described at the beginning of this article and the proposed explanation provide insight into the economic logic of the media and its originality.
First, we note that certain production factors index their remuneration on actual demand. In this case it is hardly surprising that production costs depend on ex post demand. As this is a form of profit-sharing, it is more accurate to take into account in production costs only remuneration that is independent of ex post audiences.
Apart from the case in which profits are shared with certain production factors, there is an indirect relation between production costs and ex post demand. This relation stems from the fact that the producer targets a certain audience at the time of production, and consequently chooses certain production factors.
To account for the relation between production costs and targeted demand, the explanatory model cannot be limited to the classic schema linking production factors to an objective quality of the product, a best quality generating a superior demand. Producers make use of talent inputs capable of drawing the public's attention and of guiding its choices. These inputs have a price that increases in direct relation to their effectiveness in influencing social processes of generation of demand. However, due to the very nature of these processes, ex post audiences vary widely. It is never possible to substantially reduce the risk, irrespective of
22
the costs incurred to target an audience. But in the very rare cases of hits, talent inputs have a considerable effect on takings.
In short, the originality of the media stems from the nature of the information good. Demand does not rely primarily on the goods themselves and their characteristics (e.g. their "formal quality") because consumers are in any case incapable of accurately estimating ex ante their potential ex post usefulness (experience goods). Demand depends primarily on the processes through which consumers can be informed and production is essentially customized through action on these processes. Talent inputs act directly on demand; they belong less to the production function of content (film, program, book, etc.) than to the production function of social devices allowing that demand to develop.
This analysis could open interesting perspectives for the study of the media industry (and, more generally, experience goods) from an empirical and theoretical point of view. Several questions are raised. How do talents emerge? Can possible forms of the production of this type of input be specified? For example, is it possible to create stars (and at what price?) when demand is not controlled (when the Star System was created studios did have some control over demand)? Can producers create temporary stars at a low cost and thus avoid, to a certain extent, sharing profits with them? The case of a program like Big Brother will be interesting in this respect, as will the development of films without actors (Final Fantasy). Finally, is it really necessary to produce content? Is it not becoming worthwhile to invest most costs in processes of generating demand (the case of the film Blairwitch Project would be interesting to analyse from this point of view)?
23
Acknowledgements The views expressed in this paper are not necessarily those of France Télécom or Canal Plus.
24
References
Benzoni, L. and J. Perani, 1996, Concurrence, dominance et contestabilité : Les animateursvedettes de télévision et le fonctionnement du marché des programs de divertissement, Communications & Strategies 23, 143-172.
Bikhchandani, S., Hirshleifer, D. and I. Welch, 1992, A Theory of Fads, Fashion, Custom, and Cultural Change as Informational Cascades, Journal of Political Economy 100, 992-1026.
Chisholm, D.C., 1993, Asset Specificity and Long-Term Contracts: the Case of the Motion Picture Industry, Eastern Economic Journal 19, 143-155.
Crandall, R., 1972, FCC regulation, monopsony and network television program costs, Bell Journal of Economics 3, 483-508.
De Vany, A. and W.D. Walls, 1996, Bose-Einstein Dynamics and Adaptive Contracting in the Motion Picture Industry, Economic Journal 106, 1493-1514.
De Vany, A. and W.D. Walls, 1999, Uncertainty in the Movie Business: Does Star Power reduce the Terror of the Box Office, Unpublished Working Paper, University of California, Irvine [http://marshallinside.usc.edu/mweinstein/].
Economides, N., 1996, The economics of networks, International Journal of Industrial Organization 14, 673-699.
25
Litman B.R., 1983, Predicting Success of Theatrical Movies: An Empirical Study, Journal of Popular Culture 16, 159-175.
Litman, B.R. and L. Kohl, 1989, Predicting financial success of motion pictures: The 80’s experience, Journal of Media Economics 2, 35-49.
Owen, B.M., Beebe, J.H. and W.G. Manning, 1974, Television Economics (Lexington Books, New York).
Owen, B.M. and S.S. Wildman, 1992, Video Economics (Harvard University Press, Cambridge).
Ravid, S.A., 1999, Information, Blockbusters and Stars: A Study of the Film Industry, Journal of Business 72, 463-492.
Rosen, S., 1981, The economics of superstars, American Economic Review 71, 845-858.
Samuelson, P.A., 1958, Aspect of public expenditure theories, Review of Economics and Statistics 40, 332-338.
Spence, M. and B.M. Owen, 1977, Television programming, monopolistic competition and welfare, Quarterly Journal of Economics 91, 103-126.
Vogel, H.L., 2001, Entertainment Industry Economics: A Guide for Financial Analysis, 5th edition (Cambridge University Press, Cambridge).
26
Weinstein, M., 1998, Profit Sharing Contracts in Hollywood: Evolution and Analysis, Journal of Legal Studies 27, 67-112.
Woodbury, J.R., Besen, S.M. and G.M. Fournier, 1983, The determinants of network television program prices : Implicit contracts, regulation and bargaining power, Bell Journal of Economics 14, 351-365.
27
TABLES AND FIGURES
8,5
ln(annual cost of schedule for Canal Plus)
8
7,5
7
6,5
6 13,8
14
14,2
14,4
14,6
14,8
15
15,2
15,4
ln(number of subscribers)
Figure 1 : Cost of the Canal Plus program schedule in relation to number of subscribers
5,5
5
ln(budget in 1995)
4,5
4
3,5
3 12,4
12,6
12,8
13
13,2
13,4
13,6
13,8
14
14,2
ln(average number of subscribers in 1995)
Figure 2 : Quadratic regression curve budget=f (subscribers) in 1995
28
5
4 CM( 10 , a ) CM( 20 , a ) CM( 30 , a ) CM( 40 , a ) CM( 50 , a ) 2 CMar( a ) 3
CM PROG a * = 50, a
(
)
CM PROG (a, a )
1
0 10 20 30 a 40 50
Figure 3 : Construction of a production cost function with rational expectations
29
35,00
Stars 90
30,00
La nuit des Césars
25,00
Sacrée Soirée Les marches de la gloire
Cost per minute
20,00
Dimanche Martin
15,00
La marche du siècle
10,00
Perdu de vue Bas les masques
Que le meilleur Le juste prix Bouillon de culture gagne 5,00 La chance aux chansons Questions pour un Français, si vous champion Des chiffres et parliez des lettres 0,00
0
5
10
15 Audience
20
25
30
Figure 4 : Average cost per minute for channels showing short-lived programs in 1993 (cost per minute as a function of audience)34
34
Source for audience: Médiamétrie. Source for program costs: TVSD, July 1993; Télérama, n°2261, may 1993.
30
Average number of cable subscribers Euronews Eurosport Canal J Planète MCM Paris Première Canal Jimmy Série Club La Chaîne Météo Nouvelle chaîne 1,203,587 1,112,716 1,036,716 928,716 914,216 843,216 776,216 468,452 325,000 300,000
Budget (million francs) 131.5 157.1 95.4 71.0 85.2 93.3 74.2 50.0 40.0 35.0
Table 1 : Subscribers and budget of cable channels in 199535
35
Sources for the mean number of subscribers: AVICA, Canal Plus, Ecran Total. Sources for the budget of channels: commercial court records, CSA (broadcasting regulatory authority), the press. The mean budget of a new channel is calculated on the basis of projected budgets in 1995-1996 of Odysée (FF35m), France Courses (FF15m), Free One (FF50m), Canal Soleil (FF50m), Festival (FF42m), Voyage (FF35m), two planned channels of MK2 (FF20m) and Téva (FF50m). The mean number of subscribers of the new channel is equal to the expected number of subscribers if the channel obtains a contract with one of the following cable operators: Lyonnaise Communications, Compagnie Générale de Vidéocommunication, France Télécom Câble, EDF Videopole, Réseaux Câbles de France.
31
Marc BOURREAU* Michel GENSOLLEN** Jérôme PERANI***
July, 2002
Abstract The economic literature considers that in the media industry production is characterized by high fixed costs and economies of scale. However, empirical evidence seems to belie that idea, for very often the programs and films with large audiences also have high production costs. In this article we propose a program production cost function that reconciles the theory of high fixed costs in content production with empirical evidence. To that end, we develop an empirical estimation of the program cost function in pay-television, and put forward some ideas to explain why production cost is an increasing function of the targeted audience.
JEL Codes: L82; L15. Keywords: Media economics; Program production cost; Economies of scale; Stars.
*
ENST, Department of Economics, 46 rue Barrault, 75634 Paris Cedex 13, France. Tel: 33 1 45 81 72 46. Fax: 33 1 45 65 95 15. Email: [email protected].
**
ENST, Department of Economics and France Télécom. E-mail: [email protected]. Canal Plus. E-mail: [email protected].
***
1
1. Introduction
Since Samuelson (1958), the economic literature1 has considered that production in the media industry (films, TV programs, etc.) is characterized by high fixed costs and economies of scale. For example, costs of creating a TV program schedule will be high when first copy costs were high.2 Once the schedule has been published the incremental cost of physical distribution to an additional consumer is very low or even nil. Broadcast television should therefore exhibit large economies of scale. The same naturally applies to the production of content.
Yet empirical studies have shown that, very often, the more successful a TV program or film, the more it costs. In particular, this correlation has been highlighted by Litman (1983) and Ravid (1999) for cinema. The existence of a relation between production costs and audiences is also recognized in the economic literature. Spence and Owen (1977) think that it may be necessary to increase production budgets to draw larger audiences. Owen and Wildman (1992) consider that a program's production costs can influence its potential audience: “The cost of producing a television program is independent of the number of people who will eventually see it. (The production cost, however, may very well influence how many people will want to see it).”
How can the theoretical idea that programs are goods with high fixed costs – which implies the existence of economies of scale in program production – be reconciled with empirical
1
For a review of the economic literature on television, cf. Owen and Wildman (1992); for the cinema and entertainment industry in general, cf. Vogel (2001).
2
Cf. Waterman (1990: 291) and Owen and Wildman (1992: 151).
2
evidence that program costs are positively correlated with their audiences? The aim of this article is to explain this paradox by proposing, for films and television programs, a program production cost function that reconciles the theory of high fixed costs with empirical evidence suggesting that marginal costs are not negligible.
To this end, we first present an empirical analysis of the cost function of a television channel. More precisely, based on a statistical analysis of program costs and audiences of French payTV channel, Canal Plus, and French cable channels, we show that the production costs of a television channel's program schedule are not unrelated to its maximum audience. This analysis highlights the theoretical problem mentioned above, i.e. that no economies of scale are observed for what is commonly called the "production" of television programs.
To illuminate the issue of possible economies of scale in content production, we propose a program production cost function for cultural goods such as cinema or television. This enables us to account for empirical results (the absence of economies of scale) and logical evidence (production costs cannot depend on ex post audiences). We consider that costs of a program or film increase in relation to the audience targeted by the producer, and that there is a correlation between this expected audience and the actual audience.
In the second part of the article, we consider the nature of the relationship between production costs and targeted audiences, based on the example of cinema and television. In the case of the media, production costs depend on expected audiences through specific, so-called "talent" or "star" production factors which inform and orient demand and impact on social processes needed to develop the consumption of experience goods. These "stars" are able to negotiate remuneration based on fame, i.e. on the mean audiences they draw. Sometimes they are even
3
remunerated on the basis of ex post audiences. Thus, the content production function seems to be more or less comparable to a publishing function.
2. Dependence between production costs of information and audiences
In this section we first assess the cost function of a program schedule for a pay television channel (French pay-TV channel, Canal Plus) and a cable channel. We consider only the content production phase, not the distribution and diffusion phases.3 The question of economies of scale in the media industry thus relates only to the initial production phase. We show that in both examples the program schedule cost is largely dependent on the maximum audience of the channel. Then, we propose a program production cost function that reconciles the theory of high fixed costs with empirical evidence.
2.1. Estimation of production costs for a pay-TV channel
Canal Plus was created in 1984 and it was at the time the first pay-TV channel in France. In 2001, it had 4.9 million subscribers. Canal Plus has production costs relating to commercialization (promotion and advertising costs, cost of decoders, costs of acquisition and management of subscribers). We assess only total programming costs here, that is, the cost of the program schedule. In order to do so, we compare the annual mean number of subscribers4
3
In the case of cinema, we distinguish the production of the negative from the diffusion phase, consisting in the sorting of positive copies and their distribution to cinemas. In the case of television, we distinguish production of the program schedule from its diffusion.
4
The number of annual subscribers of year N is equal to the mean number of subscribers on 31 December of years N-1 and N.
4
and the programming costs of Canal Plus between 1986 and 1996.5 The data are drawn from Canal Plus annual reports between 1991 and 1996. The number of subscribers provides a proxy for the maximum audience, i.e. the audience that the channel could draw if all households able to receive it watched it for the entire duration of the schedule.
We thus obtain 11 values for the program schedule cost, depending on the number of subscribers. The small size of the sample obtained does not allow a precise adjustment of the cost function. We have limited ourselves here to showing the relation between program costs and audiences.
We perform a logarithmic regression of the cost of the program schedule (expressed in millions of 1996 francs) related to the number of subscribers (in millions of subscribers), ln(PROG )=a*ln(ABO )+b . We find that a =1,2297 , b=−10,676 and that the regression is statistically significant (R²=0,986). Figure 1 below presents the cost of the Canal Plus program schedule in relation to the channel's number of subscribers.
[FIGURE 1 SOMEWHERE HERE]
The analysis of Canal Plus program schedule costs suggests a linear increase of total schedule costs with the channel's maximum audience. This finding must be qualified, for the number of points used for this estimation is small. We can nevertheless note that the schedule cost is a function of the number of subscribers.
5
We can consider that program purchasing conditions hardly changed for Canal Plus between 1986 and 1996. During that period Canal Plus was the only pay-television channel in France. The satellite multichannel package TPS was introduced in 1997 only, when it started to compete with Canal Plus for rights to cinema films and football matches.
5
2.2. Estimation of production costs for a cable channel
In the case of cable channels, the same result obtained for Canal Plus is found for their total production costs. We carried out an identical analysis of French cable channels in 1995.6 More precisely, we tried to determine whether there was a relation between the total budget of a cable channel7 and its number of subscribers. The total budget of a cable channel is a good estimate of program costs, in so far as broadcasting costs vary little from one channel to the next (in 1995 these were equal to the annual rate for renting an Astra or Eutelsat analogue satellite capacity).
We thus obtained ten values for ten cable channels (cf. Table 1). We estimated a logarithmic regression of the channels' budgets (in millions of francs) in relation to the number of subscribers of each channel, ln(CT )=c*ln(ABO )+ d . We found that c=0,8999 , d =−7,8118 and that the results obtained were statistically significant (R²=0,8841).
As observed for the cost of the Canal Plus program schedule, we found that the budget of a cable channel in 1995 increased with the number of subscribers to the channel. Once again, this finding has to be qualified in so far as the number of points is limited (ten points).
[TABLE 1 SOMEWHERE HERE]
[FIGURE 2 SOMEWHERE HERE]
6
That year enabled us to take into account the satellite subscribers of each channel since digital multichannel packages were launched the following year, in 1996. We thus consider that the 260,000 mean subscribers of channels present on analogue Canal Satellite, for the year 1995, have no impact on channels' budgets.
7
The total budget is equal to the difference between the channel's income and its profits.
6
2.3. Production cost of information and expected audience
We have presented two estimations which show that the programming cost of a pay-television channel is not a fixed cost in relation to the channel's maximum audience.
Yet in the case of television channels a distinction must be made between the content cost and the price for purchasing that content from copyright owners. This price can be defined by complex contracts and may depend more or less directly on ex post audience scores. In the case of television, for example, broadcasting rights on content are negotiated at prices which depend on an audience estimated ex ante, and sometimes also on the audience measured ex post. The relation that we have shown between programs costs and maximum audiences could thus be related, at least partly, to contractual clauses indexing the purchase prices of these programs over actual audiences or maximum audiences. In other words, we have shown that the cost of a program schedule increases in line with the maximum audience, but does this hold true for the production cost of a program?
Studies undertaken in the film industry provide a positive answer to this question. Ravid (1999) shows a strong correlation between the gross takings of a cinema film – hence, the number of entries – and its production cost. In his study, the production cost explicitly excludes the costs of marketing, distribution and remuneration of participants, which are indexed on actual revenue. Litman (1983) and Litman & Kohl (1989) confirm that result.
Is it possible to reconcile this result with the idea that program production for television or cinema has high fixed costs? This is what we now wish to do, by proposing a formalization for the program production cost function.
7
When they make a program, producers (TV channel, studio, independent producer, etc.) anticipate an audience, the expected audience, a* . We suggest that this audience is not only expected but is actively targeted during the production, and that the program cost, C , is a function of the expected audience, a* . For example, if C is an increasing function of a* ( ∂C ∂a* > 0 ), the program cost will increase in line with the size of the targeted audience. In the following section we shall revert to the relation between content production cost and targeted audience a* .
Once it has been produced, the program will be broadcast and will draw an actual audience, a. If it costs a program producer too much to target a bigger expected audience, then ∂C ∂a* > 0 . On the other hand, for a set value of a* , the production cost is obviously independent of the actual audience measured ex post, a. We then have ∂C ∂a = 0 .
For example, imagine that the producer can perfectly predict the actual audience. In a market equilibrium, since there are rational expectations, the producer's predictions prove to be correct, which means that a * = a .8 The function of production cost with rational expectations is then C ( a ) . Figure 3 illustrates the construction of a function of production cost with rational expectations. In this figure we assumed that the average cost of programming based on actual audiences could be CM ( a* , a ) = C ( a* ) a = a * a . The average production cost with rational expectations is then simply CM ( a, a ) = 1 . Naturally, this example is extreme, for
8
For example, one could imagine the following sequence. Initially the producer, aware of the relation between investment and expected audiences, chooses to invest in a program. Then viewers watch the program or not. In perfect information the producer can predict the viewers' behaviour in the second stage. We shall see below that this expectation is difficult for a film or particular program (variance of a is very great or infinite). Producers and broadcasters must use more complex strategies discussed in the second part of this article.
8
producers are not able to perfectly anticipate the success of the programs they produce, as shown in the following section.
The formalization that we propose enables us to reconcile the idea of program production with high fixed costs and the empirical observation which shows that the function of cost, observed ex post, does not have large economies of scale depending on the actual audience, a. While it is true that the average production cost decreases in line with the actual audience, for a set targeted audience (since ∂C ∂a = 0 ), the average production cost observed ex post is not necessarily strictly decreasing (for ∂C ∂a* > 0 ).
Note that the cost function suggested here for film and TV program production can be likened to a demand function for goods with externalities. Economides (1996) shows that the willingness to pay for an n-th good, when the targeted quantity of goods sold is n * , is p n, n * . This demand function decreases with n and increases with n * , due to the externality effect. The demand function with rational expectations is then p (n, n ) , assuming that at the equilibrium n = n * .
(
)
[FIGURE 3 SOMEWHERE HERE]
3. Relation between programming cost and expected audience
In the previous section we introduced the hypothesis of a relation between the cost of a program, C , and the expected audience of that program, a * . In this part we specify the nature of that relation.
9
If we exclude the case in which certain production factors are remunerated in relation to actual audiences,9 the production costs of program content cannot be linked to the ex post audience unless production factors depend on the expected audience (it is necessary to spend more to target a vaster audience, at least on average), and unless there is a correlation between the expected audience at the time of production and the actual audience in the end.
We are going to argue that particular inputs, introduced into the economic literature under the name of "talent", are probably the main source of the relation between production costs and targeted audience. These "talents", for example film stars, are identifiable by demand, and can serve to inform and orient supply. In so far as they contribute directly to the production of demand they can, after achieving a degree of autonomy, impose modes of remuneration which reflect a sharing of created value rather than payment for a job done. This trend, now distinct in the television industry, has been clearly observed in the case of the film industry.10
In this section we first analyse the nature of the relation between demand targeted ex ante, a * , and demand achieved ex post, a. We then define the notion of "talent" and the economic characteristics of this type of production factor. Lastly, we explain the absence of economies of scale of the production function in empirical studies, by the mode of remuneration of talents.
9
This remuneration must be counted not in production costs but as a share of profits.
10
Especially since 1948, the year of the US Supreme Court ruling in a case against Paramount, which banned vertical integration between production and exploitation. This ruling marked the transition between the Star System of studios and the emergence of independent stars. The greater a star's fame (i.e. their mean capacity to induce a demand: a*), the more advantageous their contracts will be, because of they directly induce audiences. Sometimes these contracts are even indexed on gross takings or on profits (i.e. dependent on the ex post audience: a). The relation between this industrial trend and the mode of remuneration of talents is described by Chisholm (1993) who shows that by losing control of distribution, studios also lost the possibility to invest in the construction of a specific asset, the "brand" of a star when s/he was bound to the studio by a long-term exclusivity contract.
10
3.1. Nature of the relation between a and a*
Due to the extreme uncertainty concerning a, scriptwriter William Goldman commented that in the media "No one knows anything". This uncertainty characterizing the media largely explains the nature of relations between the demand targeted at the time of production ( a* ) and demand finally expressed and satisfied (a).
If these two quantities were closely correlated, information goods would be normal goods;11 costs would depend directly on demand, not by temporal adjustment but by the capacity to accurately predict the public's tastes, and to adjust production ex ante, with precision. If, on the other hand, a and a* were unrelated, the very notion of expected demand would be meaningless (producers would not target a specific demand when making their products). In this case there would be no empirical correlation between production costs and audiences (an extreme case which does exist for certain types of "craft" production, e.g. there is no correlation between the time a writer takes to write a book and the number of readers of the book).
Empirical research in the film industry suggests that the media industry is characterized by extreme variability of ex post audiences, a. In a study of 300 films released between May 1985 and January 1986, De Vany and Walls (1996) show that distribution of audiences,
11
In the case of ordinary goods and services, production costs depend on the demand finally satisfied, in so far as the process of producing products or delivering services can be adapted: supply can be adjusted to demand as it is revealed. Only irreversible initial investments and product-specific R&D induce initial economies of scale. Economies of learning can subsequently develop. For information goods, it is as if all production costs had to be agreed ex ante. Once the content (book, film, program, etc.) is made, satisfaction of demand induces no more production costs, only distribution-exploitation costs. In so far as ordinary goods incorporate more and more R&D and have an increasingly short lifespan, the media economy is becoming a general reference used to explain the functioning of a "dematerialized" economy.
11
takings and profits are of the Pareto-Lévy type, with infinite variance (sometimes even with an infinite mean).12 In this case, even if at the production level an audience a * is targeted and even if, ex post, a correlation is found between a and a * , the result in both cases is unpredictable (no one knows anything). Almost all profits derive from a few rare cases, from the precarious balance between a few disasters13 and a small number of extraordinary successes.
Such variability of demand is related to the very nature of the media. Because information is an experience good (the usefulness of which is not known ex ante) and a network good (more useful when consumed by others as well), the process of acquisition of information by potential consumers, from those who have already consumed, plays a crucial part in the formation of demand. This may be a pure phenomenon of imitation14 or a more complex process of information exchange.15 In any case, we observe cascades (or avalanches!) of decisions leading to extremely uncertain results.
12
These are distributions such as Prob(X>x) is of the order of x• for x big enough with • between 0 and 2. In the case of cinema, we generally find values of • of the order of 1.5 (see De Vany and Lee, 1996).
13
This concentration of income is increasing with time. In the late 1940s, 1% of films (the best) brought in 2% of all income; in the early 1960s they earned 6%, and in 1993, 14% (in the latter case only two films were concerned). See Weinstein (1998).
14 15
Such as consumption based on imitation or fashion; see Bikhchandani et al. (1992).
In the case of ordinary goods and services, production costs depend on the demand finally satisfied, in so far as the process of producing products or delivering services can be adapted: supply can be adjusted to demand as it is revealed. Only irreversible initial investments and product-specific R&D induce initial economies of scale. Economies of learning can subsequently develop. For information goods, it is as if all production costs had to be agreed ex ante. Once the content (book, film, program, etc.) is made, satisfaction of demand induces no more production costs, only distribution-exploitation costs. In so far as ordinary goods incorporate more and more R&D and have an increasingly short lifespan, the media economy is becoming a general reference used to explain the functioning of a "dematerialized" economy.
12
The strategies of media content producers thus naturally concern control of meta-information diffusion processes needed for the formation of audiences, rather than production itself. There are two such types of strategy: • Incorporation of production factors making it possible to attract attention to the content (film, program, book, etc.) and, simply through their presence, to attract a specific audience. This may consist of certain original characteristics of the film, e.g. the novelty of special effects, the fact that it is a sequel to a well-known film or based on a success in another medium (book, TV series). Certain actors or film producers (the stars) thus act as audience prescribers. • Efforts to achieve direct control over demand through vertical integration between production and the various phases of distribution-exploitation. This was the strategy developed by studios until the 1948 Supreme Court ruling (U.S. v. Paramount). When the rarity of means of distribution or a monopoly position enables producers to impose a certain type of content, it becomes possible for them to build up a reputation and then to take advantage of the rent thus created (extracted by means of long-term contracts with audience prescribers).
In this context the notion of "talent" relates less to the ability to create quality content than to the capacity to directly induce a demand through a brand effect. Brand effects make it possible to reduce consumers' uncertainty as to the nature of a new product.
13
3.2. The role of talent in the formation of demand
To describe information goods such as films, economists have introduced the notion of input of the work of "talent" (producers, actors, scriptwriters, etc.) along with input of production work (lighting engineers, cameramen, etc.) and capital input (décor, special effects, etc.).16 The inputs of "talent" are presented as non-substitutable and "rare". They are responsible for a large part of the quality of content, for this quality is considered to be an objective variable judged in the same way by everyone (vertical differentiation).
By introducing the notion of "talent", standard models try to account for the originality of the media industry. Yet stars are not distinguished from other inputs (capital, work) only by their rarity and low level of substitutability. It is also because they can still be recognized by audiences after the production phase that they are able to act on processes of information exchange prior to the consumption of experience goods. Such exchanges of information are all the more necessary in so far as consumers' tastes, which guide their choices, vary widely in time and from one individual to the next.
Thus, the notion of talent is not only related to the "artistic" character of inputs. There are also artistic inputs which, because they are not directly known to consumers, play a part that is no different from that of ordinary work inputs. For example, designers in the luxury industry, because they are unknown to the general public, are remunerated independently of the success of the products they define. Likewise, most actors are not stars, in so far as their participation in a film has no statistically identifiable impact on audiences. For instance, De Vany and
16
See, for example, Crandall (1972) and Owen, Beebe & Manning (1974). Crandall (1972) distinguished only between inputs of "talent" and other input.
14
Walls (1999) identified only 19 stars in all North American actors and film producers during the period 1984-1996.17
The very small number of stars is often explained by a "winner takes all" effect because mediocre talent is seen simply as an imperfect substitute for eminent talent, whereas demand is only for the best.18 An alternative explanation could be proposed: it is the generalized costs of gathering, storing and processing data needed to estimate, ex ante, the usefulness obtained ex post from information goods which ultimately limits the number of signals that can be taken into account. In the case of films, for example, consumers are prepared to devote limited time to processing the meta-information needed to guide their choices. The number of stars identified by De Vany and Walls (1999) could therefore be interpreted in the following way: an informal database on about 20 actors represents a maximum load as far as a consumer's attention is concerned.
The rarity of talent should not be analysed as rarity of an ordinary production factor. Stars in themselves are not rare; there are few of them because consumers' attention is rare and because the time they have to seek information for their consumption of experience goods is limited. Consequently, factors predicting the usefulness of information goods (films, programs, books, etc.) also acquire a value. These factors are used to inform consumers and to
17
Most of whom are actors (Warren Beatty, Sandra Bullock, Jim Carrey, Kevin Costner, Tom Cruise, etc.). The only star producers whose names create a specific demand are Spielberg and Coppola.
18
See Rosen (1981). The author writes, for example: "Lesser talent often is a poor substitute for greater talent. The worse it is, the larger the sustainable rent accruing to higher quality sellers because demand for the better sellers increases more than proportionately: hearing a succession of mediocre singers does not add up to a single outstanding performance. If a surgeon is 10 percent more successful in saving lives than his fellows, most people would be willing to pay more than a 10 percent premium for his services."
15
orient demand. We can thus reasonably talk of an "attention economy"19 in which the essential resource lies in the ability to draw the public's attention to a particular experience good and to reduce uncertainty on the usefulness that can be derived from it.
From this point of view, the rarity of talent is not inherent to processes of formation of demand; it is relative to them. There are only a few actors who please the public, not because the fact of acting well or being likeable is "rare", but because actors orient demand and the current process of ex ante evaluation of ex post quality is costly in terms of attention. If this process changes, for example with the spread of the Internet, the nature of this rarity will also change.
The fact remains that currently stars play an essential part in the success of films, not by substantially reducing the risk taken by the producer (variance of distribution of takings is, in any case, infinite) but20 • by increasing takings: median takings are US$21m for films without stars, compared to US$38m for films with stars; • by substantially increasing the probability of the film being very successful,21 although this probability remains extremely slight.
19
A notion introduced by Michael H. Goldhaber in 1997 during the conference "Economics of Digital Information" (Cambridge, MA, January 23-26, 1997): The Attention Economy: The Natural Economy of the Net. The author defends the idea that the only really important rarity today is consumers' attention. While this argument may seem excessive, the fact remains that, at least for experience goods (information goods of radically new products, for example), the processing of information needed for consumption represents most of the cost borne by the consumer. For this type of good, the time budget is more constraining than the financial budget.
20 21
See De Vany and Walls (1999): it concerns films in the period 1984-1996. For De Vany and Walls (1999), a "hit" is a film with gross takings of over US$50m.
16
3.3. Remuneration of talent
As defined previously, "talent" inputs are characterized by their ability to act on information processes that generate demand for experience goods. These inputs (film stars, TV hosts, etc.), because they are at least partially responsible for the formation of demand, are in a good position to demand remuneration based more or less directly on audiences.
In the case of cinema, it seems that stars have progressively managed to capture most of the surplus they generate. In a study on a random sample of 180 films released between 1991 and 1993, Ravid (1999) found that while stars increase the box-office takings of films, they capture the income thus created and finally play no part in its financial success.
Contracts binding actors to producers vary widely.22 Classically, the following are distinguished: • a fixed payment for each service: actors who are not well-known are remunerated in this way; this was typically the case of silent film actors who, initially, were not known by name in the public at large;23 • payment by salary, with long-term contracts:24 in so far as actors inform audiences on the nature of films and on their potential usefulness, studios directly organize their fame, "packaged" in the form of typical characters;25 this is the "star system";
22 23
See Weinstein (1998).
The first actress to be known by her name was Florence Lawrence. She was employed by Independent Motion Pictures Company (the ancestor of Studios Universal). Before 1910 she was simply called "The IMP Girl".
24 25
Mary Pickford was the first actress to have a long-term contract, from 1913.
There was the "dangerous man", the "folk hero", the "boudoir dandy", the "sinful woman", etc. (see Chisholm, 1993). Actors were related to one of these types, sometimes even to a particular character, as in the case of Johnny Weismuller as Tarzan.
17
•
remuneration dependent on gross income: mostly a percentage of income, in excess of a set level; in these conditions the star shares with the producer a single type of risk – the film's success;
•
remuneration dependent on net profits: in this case, the star shares two types of risk: the risk related to audience size, but also that of not meeting production costs (a risk over which the actor has no control); this type of contract often leads to legal dispute26 and the courts readily consider it as one-sided.
While it may seem natural for stars to try to capture the rents they generate through their presence, the fact of them doing so by means of risk-sharing contracts (dependence on a) rather than fixed remuneration (dependence on a* ), may seem strange. It is generally thought that private individuals are more sensitive to risk than firms. Various attempts have been made to explain this situation.27
First, participation in takings or profits can be an incentive for the actor to work well. In this context of moral hazard it is difficult to see why the sharing exists primarily for stars and not equally generously for all actors or all those who participate in making a film. Moreover, the actor can benefit from private information on the probability of success and the producer buys that knowledge. However, such asymmetry of information is highly improbable and there is no reason for an actor to have better knowledge of risks than the studio (besides, "no one knows anything"!). Finally, the risks are so great in the entertainment industry that risksharing is necessary. It does not stand to reason that stars are more sensitive to risk than those
26
Weinstein (1998) cites recent cases, with the arguments of the Court of Justice, in particular: Buchwald v. Paramount Pictures Corp. (1990); Batfilm Productions v. Warner Bros. Inc. (1994); and Estate of Jim Garrison v. Warner Brothers et al. (1996).
27
See Weinstein (1998) for details of this discussion.
18
who make decisions in the studios. The fact that stars (who are often very rich) frequently give up all up-front payment if that is the only way a film can be produced,28 attests to this.
In any case, talent inputs are remunerated by producers according to their fame (that is, according to the mean ex ante audience, a* ) or according to the actual audience they generate (a). The relation between production costs and actual audience, a, is thus twofold. First, this relation exists through the link between production costs and expected audiences, a* , and through the correlation between a and a* . Second, since the remuneration of talent inputs is indexed on the actual audience, the marginal cost of an additional consumer is not nil. It is an apparent cost since the considered production factor (e.g. a film actor) is not used more because there is an additional consumer. The marginal profit increases and is distributed between the producer and the talent input.
3.4. The role of talent inputs in the case of television
The preceding analysis of relations between producers and talent inputs was based primarily on examples from the "pure" case of cinema. It can also be illustrated by examples from the more complex case of television.
In the case of US broadcast television, Woodbury, Besen and Fournier (1983) analysed a sample of 99 series programd in the US in 1977 and 1978 on the three national broadcast networks, ABC, CBS and NBC. These authors showed that the producers of these series were
28
The most frequently cited examples are: James Stewart for Winchester '73, Georges Lucas for Star Wars, Tom Hanks for Forrest Gump, Dustan Hoffman for Wag the Dog, James Cameron for Titanic, Anthony Hopkins for Amistad, Robin Williams for Good Will Hunting, etc.
19
remunerated in accordance with the "popularity" of the programs they had made (i.e. the ex post actual audience, a). For example, if a program draw a larger audience than expected, the network distributed part of the surplus to the producer.29 More recently, the case of the series ER broadcast on NBC illustrates the relation between audience and program value (i.e. cost for the channel). At the end of the year 1997 the producer of the series, Warner Bros., planned to negotiate on the basis of 10 million dollars per episode.30 In January 1998 the two parties reached an agreement to renew the broadcasting contract on the basis of 13 million dollars per episode.31
In the case of French broadcast television, the market power of talent input is illustrated by the case of host-producers. There are very few of these production companies in which the main shareholder bundles the animation and production of the program, and they are subject to little pressure from potential entrants. The guarantee of audiences, combined with the host's fame, enables them to maintain their position in program schedules far more easily than other types of producer.32
The first half of the 1990s was characterized by bigger audiences of programs produced by these hosts, which de facto enhanced their bargaining power in the market. Figure 4 presents mean costs, for TV channels, of the different short-lived programs broadcast in 1993, in
29
For Woodbury, Besen & Fournier (1983), the network tried to maintain good relations with these producers in order to obtain new programs, and relinquished part of the surplus in order to ensure good performance from them.
30 31 32
Cf. C. Littleton, "Seinfeld exit jolts NBC", Variety, 29 December 1997 See: "The Thursday-Night Massacre", The New York times Magazine, 20 September 1998.
Benzoni and Perani (1996) describe a “virtuous circle” for these host-producers between audience, fame and market power. The bigger the audiences of a host, the greater her/his fame among televiewers and distributors, for the latter can maximize their advertising income by using her/his services. A host's bargaining power with distributors increases along with her/his fame: that is the virtuous circle of success. A star-host can then bargain with distributors on the basis of her/his qualities as a host and the services of her/his production company.
20
relation to the audiences of those programs.33 As shown, programs with high mean costs are presented by famous hosts, whether their audiences are small (J.-M. Cavada with "La Marche du Siècle"; B. Pivot with "Bouillon de Culture") or large (J.-P. Foucault with "Sacrée Soirée" and M. Drucker with "Stars 90").
This analysis shows that in the case of television as well, talent inputs capture part of the surplus they create, in terms of subscription revenue or advertising revenue generated by the program.
[FIGURE 4 SOMEWHERE HERE]
4. Conclusion
The economic literature generally considers that the media industry is characterized by high fixed costs and economies of scale. More precisely, the content production phase is considered to have fixed costs. In this article we have shown that this analysis does not correspond to the findings of empirical research. In the case of both television and cinema, there seems to be a correlation between the audience of a program or film and its production cost. This property has been demonstrated by estimating the cost function of pay-television in France, which suggests that the production cost of a program is an increasing function of audience.
To illuminate this paradox we introduced a cost function of a program or film, in which production costs depend on the targeted audience at the time the content is produced. This
33
In certain cases the "costs" of these programs for channels can be purchase prices and not production costs.
21
cost function is logically independent of the actual audience. Once a program has been produced its cost cannot depend on the number of viewers, but it can increase with the targeted audience and thus statistically with the actual audience when these two variables are correlated.
The paradox described at the beginning of this article and the proposed explanation provide insight into the economic logic of the media and its originality.
First, we note that certain production factors index their remuneration on actual demand. In this case it is hardly surprising that production costs depend on ex post demand. As this is a form of profit-sharing, it is more accurate to take into account in production costs only remuneration that is independent of ex post audiences.
Apart from the case in which profits are shared with certain production factors, there is an indirect relation between production costs and ex post demand. This relation stems from the fact that the producer targets a certain audience at the time of production, and consequently chooses certain production factors.
To account for the relation between production costs and targeted demand, the explanatory model cannot be limited to the classic schema linking production factors to an objective quality of the product, a best quality generating a superior demand. Producers make use of talent inputs capable of drawing the public's attention and of guiding its choices. These inputs have a price that increases in direct relation to their effectiveness in influencing social processes of generation of demand. However, due to the very nature of these processes, ex post audiences vary widely. It is never possible to substantially reduce the risk, irrespective of
22
the costs incurred to target an audience. But in the very rare cases of hits, talent inputs have a considerable effect on takings.
In short, the originality of the media stems from the nature of the information good. Demand does not rely primarily on the goods themselves and their characteristics (e.g. their "formal quality") because consumers are in any case incapable of accurately estimating ex ante their potential ex post usefulness (experience goods). Demand depends primarily on the processes through which consumers can be informed and production is essentially customized through action on these processes. Talent inputs act directly on demand; they belong less to the production function of content (film, program, book, etc.) than to the production function of social devices allowing that demand to develop.
This analysis could open interesting perspectives for the study of the media industry (and, more generally, experience goods) from an empirical and theoretical point of view. Several questions are raised. How do talents emerge? Can possible forms of the production of this type of input be specified? For example, is it possible to create stars (and at what price?) when demand is not controlled (when the Star System was created studios did have some control over demand)? Can producers create temporary stars at a low cost and thus avoid, to a certain extent, sharing profits with them? The case of a program like Big Brother will be interesting in this respect, as will the development of films without actors (Final Fantasy). Finally, is it really necessary to produce content? Is it not becoming worthwhile to invest most costs in processes of generating demand (the case of the film Blairwitch Project would be interesting to analyse from this point of view)?
23
Acknowledgements The views expressed in this paper are not necessarily those of France Télécom or Canal Plus.
24
References
Benzoni, L. and J. Perani, 1996, Concurrence, dominance et contestabilité : Les animateursvedettes de télévision et le fonctionnement du marché des programs de divertissement, Communications & Strategies 23, 143-172.
Bikhchandani, S., Hirshleifer, D. and I. Welch, 1992, A Theory of Fads, Fashion, Custom, and Cultural Change as Informational Cascades, Journal of Political Economy 100, 992-1026.
Chisholm, D.C., 1993, Asset Specificity and Long-Term Contracts: the Case of the Motion Picture Industry, Eastern Economic Journal 19, 143-155.
Crandall, R., 1972, FCC regulation, monopsony and network television program costs, Bell Journal of Economics 3, 483-508.
De Vany, A. and W.D. Walls, 1996, Bose-Einstein Dynamics and Adaptive Contracting in the Motion Picture Industry, Economic Journal 106, 1493-1514.
De Vany, A. and W.D. Walls, 1999, Uncertainty in the Movie Business: Does Star Power reduce the Terror of the Box Office, Unpublished Working Paper, University of California, Irvine [http://marshallinside.usc.edu/mweinstein/].
Economides, N., 1996, The economics of networks, International Journal of Industrial Organization 14, 673-699.
25
Litman B.R., 1983, Predicting Success of Theatrical Movies: An Empirical Study, Journal of Popular Culture 16, 159-175.
Litman, B.R. and L. Kohl, 1989, Predicting financial success of motion pictures: The 80’s experience, Journal of Media Economics 2, 35-49.
Owen, B.M., Beebe, J.H. and W.G. Manning, 1974, Television Economics (Lexington Books, New York).
Owen, B.M. and S.S. Wildman, 1992, Video Economics (Harvard University Press, Cambridge).
Ravid, S.A., 1999, Information, Blockbusters and Stars: A Study of the Film Industry, Journal of Business 72, 463-492.
Rosen, S., 1981, The economics of superstars, American Economic Review 71, 845-858.
Samuelson, P.A., 1958, Aspect of public expenditure theories, Review of Economics and Statistics 40, 332-338.
Spence, M. and B.M. Owen, 1977, Television programming, monopolistic competition and welfare, Quarterly Journal of Economics 91, 103-126.
Vogel, H.L., 2001, Entertainment Industry Economics: A Guide for Financial Analysis, 5th edition (Cambridge University Press, Cambridge).
26
Weinstein, M., 1998, Profit Sharing Contracts in Hollywood: Evolution and Analysis, Journal of Legal Studies 27, 67-112.
Woodbury, J.R., Besen, S.M. and G.M. Fournier, 1983, The determinants of network television program prices : Implicit contracts, regulation and bargaining power, Bell Journal of Economics 14, 351-365.
27
TABLES AND FIGURES
8,5
ln(annual cost of schedule for Canal Plus)
8
7,5
7
6,5
6 13,8
14
14,2
14,4
14,6
14,8
15
15,2
15,4
ln(number of subscribers)
Figure 1 : Cost of the Canal Plus program schedule in relation to number of subscribers
5,5
5
ln(budget in 1995)
4,5
4
3,5
3 12,4
12,6
12,8
13
13,2
13,4
13,6
13,8
14
14,2
ln(average number of subscribers in 1995)
Figure 2 : Quadratic regression curve budget=f (subscribers) in 1995
28
5
4 CM( 10 , a ) CM( 20 , a ) CM( 30 , a ) CM( 40 , a ) CM( 50 , a ) 2 CMar( a ) 3
CM PROG a * = 50, a
(
)
CM PROG (a, a )
1
0 10 20 30 a 40 50
Figure 3 : Construction of a production cost function with rational expectations
29
35,00
Stars 90
30,00
La nuit des Césars
25,00
Sacrée Soirée Les marches de la gloire
Cost per minute
20,00
Dimanche Martin
15,00
La marche du siècle
10,00
Perdu de vue Bas les masques
Que le meilleur Le juste prix Bouillon de culture gagne 5,00 La chance aux chansons Questions pour un Français, si vous champion Des chiffres et parliez des lettres 0,00
0
5
10
15 Audience
20
25
30
Figure 4 : Average cost per minute for channels showing short-lived programs in 1993 (cost per minute as a function of audience)34
34
Source for audience: Médiamétrie. Source for program costs: TVSD, July 1993; Télérama, n°2261, may 1993.
30
Average number of cable subscribers Euronews Eurosport Canal J Planète MCM Paris Première Canal Jimmy Série Club La Chaîne Météo Nouvelle chaîne 1,203,587 1,112,716 1,036,716 928,716 914,216 843,216 776,216 468,452 325,000 300,000
Budget (million francs) 131.5 157.1 95.4 71.0 85.2 93.3 74.2 50.0 40.0 35.0
Table 1 : Subscribers and budget of cable channels in 199535
35
Sources for the mean number of subscribers: AVICA, Canal Plus, Ecran Total. Sources for the budget of channels: commercial court records, CSA (broadcasting regulatory authority), the press. The mean budget of a new channel is calculated on the basis of projected budgets in 1995-1996 of Odysée (FF35m), France Courses (FF15m), Free One (FF50m), Canal Soleil (FF50m), Festival (FF42m), Voyage (FF35m), two planned channels of MK2 (FF20m) and Téva (FF50m). The mean number of subscribers of the new channel is equal to the expected number of subscribers if the channel obtains a contract with one of the following cable operators: Lyonnaise Communications, Compagnie Générale de Vidéocommunication, France Télécom Câble, EDF Videopole, Réseaux Câbles de France.
31
