Advance Contracting, Word-of-Mouth, and New-Product Success in Creative Industries: A Quantification for Books

Abstract

The odds of success in creative industries (e.g., book, music, or movie) are often said to be particularly low. Furthermore, due to the nature of advance contracting between creator and producer, the standard approach to use sales as a success measure can be misleading from the producer's perspective. This article presents a novel approach to empirically identify producer success by incorporating the standard terms of contract between creator and producer into a parsimonious model of information diffusion (word-of-mouth) about new products. The model is applied to weekly sales data for a representative sample of novels. Estimation results indicate a success rate between 10% and 15% for this market.

Notes

¹“Ajoutez que, de compte fait, sur dix entreprises, il y en a une, et c'est beaucoup, qui réussit, quatre dont on recouvre ses frais à la longue, et cinq où l'on reste en perte” (Diderot, 1763/2003, p. 61) [“In addition, out of every ten ventures, there is perhaps one, at the most, which succeeds, four in which the costs are eventually recovered, and five where the trader does not recover his losses.” (trans. Lydia Mullholland in: Diderot's Letter on the book trade (1763), Primary Sources on Copyright (1450–1900), eds. L. Bentley & M. Kretschmer, www.copyrighthistory.org]. See Turnovsky (2003) for a review of the general reception of Diderot's article.

²Diderot's (1763/2003) rule is cited, for example, by Escarpit (1969, p. 123), Tietzel (1995, p. 38), and von Lucius (2005, p. 66).

³The common 80/20 rule is often asserted for classic industries. For example, the Food Marketing Institute (2002) reported failure rate estimates for new grocery products to be between 25% and 80%. To the best of my knowledge, there are no detailed estimates for other markets. The vast management literature on the subject, of course, focuses on the determinants of new-product success, and not on the aggregate success rate (for a meta-analysis, see Henard & Szymanski, 2001).

⁴Most European countries have laws that enforce resale price maintenance for books. See Benhamou, De Vrièse, and Guillon (2010) and my Humboldt University dissertation (Beck, 2008) for recent summaries of this debate.

⁵I thank an anonymous reviewer for highlighting the relevancy of these new media. Although the data analyzed in this article relate to a period in which social media, such as Facebook® or Twitter, were not yet (widely) available, their importance in social communication about creative products is clearly on the rise. In the Conclusion section, I discuss how the results of this article suggest managerial implications that apply equally or even particularly to these new social media.

⁶ Caves (2000) observed that the “distribution of consumers between 'buffs' and 'casuals' strongly influences the organization of an art realm” (p. 173).

⁷ Horvitz (1966), and subsequent literature, discussed in more detail why the seemingly more natural alternative of profit-sharing is rarely observed in academic (textbook) publishing. Dana and Spier (2001) showed that revenue-sharing is also valuable in manufacturer–retailer contracts when demand uncertainty is realized only after inventory decisions have been made.

⁸For example, the advance insures an author against publisher insolvency, and the publisher may save on transaction costs after release when a potentially large number of small royalty payments can be accounted against the advance.

⁹An auction is the optimal selling format from an author's viewpoint (Bulow & Klemperer, 1996), and has a long tradition in the book industry. For an analysis of Goethe's second-price auction, see Moldovanu and Tietzel (1998), and for further anecdotal evidence, see Hansmann and Kraakman (1992), De Vany and Walls (2004), and Meyer (2009).

¹⁰For authors, in contrast, total sales remain important ex post because they are associated with auxiliary revenues (e.g., from live performances or movie deals). Similarly, authors and publishers can have a conflict of interest with respect to copyright enforcement (Gayer & Shy, 2009).

¹¹I thank an anonymous reviewer for bringing this aspect to my attention.

¹²In contrast to Anglo-Saxon countries, in Germany there is a significant time lag of typically 12 to 24 months between hardcover and paperback editions of a title.

¹³See Schmidt-Stölting, Blömeke, and Clement (2011). The reasoning is similar for secondary markets other than paperback, such as audio books, dramatic performances, or films.

¹⁴If the negotiated advance payment is ρ, where 0 < ρ ≤ 1, publisher profits are negative ex post if there is no word-of-mouth and C > E[N _b ](1 – δ k).

¹⁵ Meyer (2009) claimed that “7 out of 10 titles do not earn back their advance,” but he did not provide further evidence, nor did he specify whether this statistic holds for the U.S. book market as a whole or for a specific segment (such as fiction or nonfiction; p. BR27).

¹⁶ Moul (2007), who quantified the average effect of word-of-mouth in motion picture revenues, also identified word-of-mouth through inter-temporal dynamics of weekly unit sales. However, the specific demand model underlying his analysis (nested logit) is very different from the new-product diffusion model presented here, variants of which are widely applied in the marketing literature.

¹⁷In most European countries, book prices are, by law, subject to resale price maintenance and, thus, invariant over time. Even in the unregulated U.S. market, inter-temporal price variation seems to be low (Clerides, 2002).

¹⁸The following specification was independently developed by Van den Bulte and Joshi (2007) and myself. In Beck (2007), I studied the estimation properties of the model with a Monte Carlo simulation, and presented detailed estimation results for four example titles. In Van den Bulte and Joshi, this model is a special case of a more general class of models, which I discuss in more detail here in the Comparison With Other Models section.

¹⁹It is assumed that casuals do not recommend the title to other people (for a discussion, see the Comparison With Other Models section).

²⁰See Van den Bulte and Joshi (2007) for a review. Variants and extensions of the Bass model (see Bass, 1969) have also been used in related literature on the diffusion of technological and organizational innovations across firms.

²¹The simulation results by Goldenberg, Libai, and Muller (2001) also indicate that, in the presence of word-of-mouth effects, advertising has a sales effect only at the very early stage of new-product diffusion.

²²See “Booksellers Say Author Tours, Oprah Most Effective for Marketing Books,” Book Publishing Report, September 27, 1999, Vol. 24, Issue 38; and “Suche nach Öffentlichkeit,” Handelsblatt, March 16, 2006, Issue 54.

²³Following the classification of the German book trade association, the segment “novels” (Segment Code 111) does not include genres such as crime, science fiction, or fantasy novels for which separate segments are defined.

²⁴In fact, each observation of sales of one, two, or three may arise from just one preordered copy because some points of sale from which the data were aggregated have a sample weight larger than one.

²⁵In a previous version of this article (Beck, 2009), I showed that the distribution for sales including all observed weeks is quite similar and only slightly broader than the one including only the first 26 weeks.

²⁶No other significant seasonal variation seems to be present. In a panel regression specification following Sorensen (2007)—S _iτ = (α _i + α_τ + β t _iτ)S _iτ–1 + ε _iτ—where τ denotes calendar weeks and t _iτ denotes Title i's weeks since release at Week τ, all off-December week fixed effects α_τ are insignificant.

²⁷More details on estimation and interpretation of a Christmas effect can be found in Beck (2009)

²⁸On April 9, 2011, Google™ Scholar counted 4,499 citations of Cleveland's (1979) article.

²⁹I classify converged estimates as degenerate if they fulfill at least one of the following conditions: (a) is smaller than .01 and not significantly different from zero, or (b) should be italicized theta. I lose the “cap” (or “hat”) when I do it/> is smaller than .01 or larger than .99 and not significantly different from zero or one (all with 95% confidence intervals).

³⁰Because _i does not include Christmas sales, it can be lower than Ni(T) for titles with a large Christmas share.

^a Number of sampled titles published by the respective publisher.