Measuring Word of Mouth's Impact on Theatrical Movie Admissions

• Date: 01 December 2007
• DOI: http://doi.org/10.1111/j.1530-9134.2007.00160.x
• Published date: 01 December 2007

Measuring Word of Mouth’s Impact

on Theatrical Movie Admissions

CHARLES C. MOUL

Washington University Department of Economics

Campus Box 1208

St. Louis, MO 63130-4899

moul@artsci.wustl.edu

Information transmission among consumers (i.e., word of mouth) has received

little empirical examination. I offer a technique that can identify and measure

the impact of word of mouth, and apply it to data from U.S. theatrical movie

admissions. While variables and movie fixed effects comprise the bulk of observed

variation, the variance attributable to word of mouth is statistically significant.

Results indicate approximately 10% of the variation in consumer expectations

of movies can be directly or indirectly attributed to information transmission.

Information appears to affect consumer behavior quickly, with the length of a

movie’s run mattering more than the number of prior admissions.

1. Introduction

Despite its widespread theoretical implications in environments of

incomplete information, information transmission among consumers

(a.k.a. word of mouth) has received relatively little empirical support. In

this paper,I show that an existing method of detecting word of mouth is

overly broad, in that its empirical prediction of autocorrelated growth

can also be generated by a simple model of saturation in demand. I

instead offer a model of demand that can accommodate both saturation

and word of mouth, and then consider its implications within an error

components framework. Applying this technique to U.S. theatrical

admissions, my estimates suggest that word of mouth is statistically

and economically significant and that information travels quickly to the

average consumer in commonly observed situations. Simulations using

these estimates confirm that word-of-mouth can have large impacts on

how movies play out in theaters.

I thank the editor and two anonymous referees for comments on an earlier draft that

greatly improved the paper. Seminar participants at Washington University in St. Louis

and the DeSantis Center’s Business and Economics Scholars Workshop also provided

helpful feedback. The usual caveat applies.

C

2007, The Author(s)

Journal Compilation C

2007 Blackwell Publishing

Journal of Economics & Management Strategy, Volume16, Number 4, Winter 2007, 859–892

860 Journal of Economics & Management Strategy

In a world of incomplete information, consumers sharing informa-

tion about experience goods can play a critical role in moving economic

outcomes closer to the full information ideal. This intuitive insight has

been formalized by Ellison and Fudenberg (1993, 1995) who present

mechanisms for and implications of what they refer to as social learning.

Furthermore, either word of mouth or repeat purchases are essential in

the theoretical literature explaining how advertising can be used as a

signal of quality in a separating equilibrium (Nelson, 1974; Milgrom

and Roberts, 1986). The speed and manner of information transmission

among consumers, however,are inherently empirical questions, and it is

there that I will concentrate my efforts. Given the presumed importance

of word of mouth in the theatrical sector of the movie industry, these

results also potentially have bearing on the best responsesto information

transmission among consumers.

I discern information transmission by interpreting my model’s

residuals, and thus, while I will refer throughout to this transmission as

“word of mouth,”I am unable to distinguish between consumers shar-

ing information among themselves and information that is exogenously

revealed after a movie’s release (e.g., late movie reviews, published box

office announcements, etc.). With that caveat in mind, the general idea

of my approach is that word of mouth will be revealed in a specific

and well-defined manner.Products presumably have unique differences

between consumer expectations and realizations (the information that

is relayed by word of mouth). All products, however, will begin their

commercial lifespans in the absence of such information. If a product

stays available for a long enough time and enough consumers purchase

it and share the product’s true quality, then the original consumer

expectations will be supplanted by the conveyed realized quality.1This

systematic spread of information has implications for serial correlation

and heteroskedasticity, specifically that both will increase over a movie’s

theatrical run. My use of movie fixed effects in estimation alters this

prediction, in that heteroskedasticity and serial correlation based upon

these residuals will be nonmonotonic (specifically U-shaped) over each

movie’s run. The speed of information can then be inferred from the

nature and importance of asymmetry in these U-shaped relationships.

1. Herding therefore does not arise in this context, as signals (in the form of word of

mouth) as well as actions are observable. See Bikhichandi et al. (1992). This framework

also implicitly assumes that consumers are aware of all products’existence. Relaxing this

assumption and allowing word of mouth to affect the likelihood that a consumer knows

of a product’s existence is a potential extension, though a computationally burdensome

one. This burden arises as observed quantities are the weighted averages of all possible

combinations of consumer awareness. At the 50 weekly movies observed in my sample,

this would require the calculation of 250 −1 probabilities of each potential combination

of movies and 50 ∗249 informationally conditional quantities.

Theatrical Movie Admissions 861

This proposed error components approachdiffers substantially from the

prior literature which either exploits additional information regarding

the sign and extent of the information gap (e.g., Chevalier and Mayzlin,

2006) or takes advantage of detailed microlevel data (e.g., Foster and

Rosenzweig, 1995).

This paper is not the first to use movie admissions in an attempt to

document word of mouth. De Vanyand Walls (1996) show that informa-

tion transmission among consumers will cause autocorrelated growth

rates among movies. They then document rejections of a Pareto Law

relationship in favor of positively autocorrelated growth, and interpret

this as evidence of word of mouth. This paper makes two improvements

upon that work. First, I show that positively autocorrelated growth can

also be generated by a simple model in which consumers typically buy

a product at most once (as seems likely in the theatrical movie industry

and many others). Second, my approach can not only detect the presence

of word of mouth, but can also provide measures of the importance of

that information transmission.

The movie industry has been the subject of several other studies

in recent years. Moul (2001) finds that, from the late 1920s to early

1940s, studio revenues rose with experience using synchronous sound

recording technology and interprets this as evidence of qualitative

learning-by-doing. Einav (2007) decomposes the seasonal demand for

theatrical movies into an underlying component and the amplification

that arises from endogenous industry practices. Corts (1998) looks at

the issue of theatrical release timing in finding that a studio’s divisions

largely behave as an integrated whole. On the exhibition side, Davis

(2006) uses the location and prices of theaters to measure how cross-

price elasticities change with distance, and, on the production side,

Goettler and Leslie (2004) look at causes and consequences of two

studios cofinancing a movie. Finally, Ravid (1999) examines measures

of profitability over a movie’s entire commercial run rather than simply

revenues in a particular sector.

This paper makes some contributions to this literature beyond its

conclusion regarding the importance of word of mouth in theatrical

admissions. My initial estimates of demand are generally plausible and

conform well with implications of other studies. As expected, ignoring

and imperfectly addressing saturation in demand substantially distort

the predicted impacts of Oscar nominations, awards, and abbreviated

weeks. Estimates of the full model suggest that Oscar nominations have

a substantial impact on admissions in equilibrium, but indicate no such

comparable bump from winning the award. Prior commercial perfor-

mance of both starring cast and director have significant impacts on

consumer expectations in equilibrium. There is also strong evidence of