The effectiveness of pre-release advertising for motion pictures: An empirical investigation using a simulated market

Abstract

We use data on a movie’s stock price as it trades on the Hollywood Stock Exchange, a popular online market simulation, to study the impact of movie advertising. We find that advertising has a positive and statistically significant effect on expected revenues, but that the effect varies strongly across movies of different “quality”. The point estimate implies that the returns to advertising for the average movie are negative.

Introduction

Companies often spend hefty sums on advertising for new products prior to their launch. That is particularly true for products in creative industries such as motion pictures, music, books, and video games (Caves, 2001), where the lion’s share of advertising spending typically occurs in the pre-launch period. Consider the case of motion pictures. Across the nearly 200 movies released by major studios in 2005, average advertising expenditures amounted to over $36 million, while average production costs totaled about $60 million (MPAA, 2005). On average, about 90% of advertising dollars were spent before the release date. In addition, fueled by an intense competition for audience attention, studios have significantly increased advertising expenditures: average advertising spending per movie jumped about 50% between 1999 and 2005. Of this, television advertising represented the largest cost—accounting for 36% of total advertising expenditures for new releases in 2005. As a result, film executives are under pressure to address the soaring costs of advertising, particularly television advertising. Universal Pictures Vice Chairman Marc Schmuger commented “It is a little startling to see spending skyrocket across the board. Clearly the industry cannot sustain a trend that continues in that direction” (Variety, 2004).

This view suggests that the escalation of advertising expenditures may reduce the returns to advertising, even drastically. Furthermore, the effectiveness of advertising is likely to differ across movies according to movie “quality”: if there is any information content in advertising, then advertising a movie of low quality might even drive away consumers rather than attract them.² How effective, then, is movie advertising? Are the returns to the marginal advertising dollar positive or negative? And, how does advertising effectiveness differ across movies?

Since advertising is a major instrument of competition in the movie industry, it follows that understanding the impact of movie advertising is central to an assessment of the current and future industrial organization of the movie industry. Unfortunately, disentangling the impact of movie advertising is quite difficult. The main reason is that studying the effect of advertising on box office receipts is confounded by the classic endogeneity problem: movies expected to be more popular also are likely to receive more advertising (Einav, 2007, Lehmann and Weinberg, 2000).³

In this paper, we attempt to shed light on the effectiveness of movie advertising by pursuing a different empirical strategy. Instead of looking at box office receipts, we look at the impact on a measure of sales expectations in the pre-release period. Our measure is the movie’s “stock price” as it trades on the Hollywood Stock Exchange, a popular online stock market simulation. This measure is sensible since a movie’s HSX stock price is one of the strongest predictors of actual box office receipts. The idea that market simulations can aggregate information that traders privately hold follows work by a growing number of researchers who use such simulations to gauge market-wide expectations or to identify “winning concepts” in the eyes of consumers.⁴ Beyond that, the HSX measure has two advantages over actual receipts measures. First, one can observe the entire dynamic path of a movie’s stock price (which is a measure of market-wide revenue expectations for the movie) prior to release, and therefore relate these to the dynamics in the advertising process as well. Second, one can sweep out any time-invariant unobserved factors that affect both advertising and expectations, by first-differencing both series. Since changes in the planned sequence of advertising expenditures within the twelve-week window prior to a movie release are difficult to execute for a variety of institutional reasons, one can argue that the first-differenced advertising series is plausibly exogenous over the sample period. We go beyond this by performing a series of robustness tests that examine how sensitive the resulting estimates of advertising effectiveness are to this identifying assumption.

Section 2 describes our data and variables used in estimation. We use data on weekly pre-release expectations, as measured by the HSX stock prices, for a sample of 280 movies that were widely released from 2001 to 2003. We obtain data on weekly pre-release television advertising expenditures for that same set of movies from Competitive Media Reporting (CMR), and measure quality using data from Metacritic.

Section 3 describes our empirical strategy to examine the relationship between movie-level advertising and market-wide expectations of the movie’s success. The model centers around two questions posed earlier. First, does pre-release advertising affect the updating of market-wide expectations? Second, how does this effect vary according to product quality?

The results, described in Section 4, indicate that the impact of advertising on pre-release market-wide expectations is positive and statistically significant. Furthermore, this effect is more pronounced for movies of higher “quality”. However, the model estimates imply that, on average, a one dollar increase in advertising increases expectations of box-office receipts by at most $0.65. We discuss the implications of these results for the “optimality” of current advertising expenditures in the industry.

Section 4 also presents a series of tests that examine the robustness of our results, in particular to the assumption that unobserved time-varying movie-specific effects do not bias the point estimates of the impact of advertising on expectations. In effect, we estimate the relationship between advertising and expectations for two samples separately: one where the sequence of advertising expenditures is plausibly exogenous, and another for which a studio’s ability or need to adjust advertising within the twelve-week window is arguably greater. We find that while the dynamics of the advertising process are indeed somewhat different in the two samples, the estimates of the effectiveness of advertising are not statistically different across the two samples. Section 5 concludes and discusses some implications for future research.