Table of Contents
Blockbusters: Building Perceptions and Delivering at the Box Office
Joshua Glastein, Ohad Ludomirsky, Dean Lyettefi, Parag Vaish and Nitin J oglekar
Boston University School of Management, Boston MA 02215
Abstract
The Hollywood Stock Exchange (HSX) is an on-line market that tracks the perceived value of
movie talent and their product: the movies themselves, while they are in development or
production. We model the decision rules that drive this market place and estimate the underlying
decision parameters by calibrating the evolution of a selected sample of 23 movies released in
2001-2002. Our results show systematic differences in the decision rules followed by the market
for the eventual winners (a.k.a. the blockbusters) and the losers at the box office. Regression
analysis of combined decision parameters for winners and losers cannot explain the variance in
the box office performance. However, segmenting these data between winners and losers
provides selective insights about how the aggregate market perceptions evolve.
Key Words
Movie, Box Office, HSX, Actor, Director
Acknowledgements
We thank Professor Rogelio Oliva for making the summary statistics module available for this
paper. We thank Professor Jonathan Hibbard and a set of anonymous reviewers for commenting
on an earlier version of the paper. Additionally we thank the Hollywood Stock Exchange
(www.hsx.com)] for providing access to the time series data underlying their product.
Caveat
All findings are based on publicly available data from (fvww.hsx.com)] HSX has provided us
access more detailed data. However, our analysis does not account for some of the details. To
that extent, the findings in this paper are preliminary.
1. Introduction
Movies are a multi-billion dollar business. An average of 300 movies are developed and
released every year in the US. Developing movies is a very expensive endeavor. For instance,
the cost to produce Harry Potter and the Chamber of Secrets, one of last year’s blockbusters, was
$120.5 million. Its revenue from the box office was $255.5 million in the first 10 weeks after
release; this amount is expected to triple through post release market sales such as: DVD, rentals,
television rights and merchandise. Producing, releasing and launching movies amount to a
special case of the new product development (NPD) process. While the development process of
conventional products such as automobiles has been studied extensively (Clark and Fujimoto,
1991), the decisions of movie making have generally gone without much analysis within the
NPD literature.
The development side of movies is characterized as a risky business with nearly 95% of movie
projects getting canceled or failing to return the expected profit. This is primarily because
decision-makers, such as studio executives, distributors and merchandisers, are asked to commit
millions of dollars towards projects based on the speculated creative abilities of the talent, ie.
actors and a director. The problems are compounded because these decision processes are
flooded with scripts and themes that may not be in line with the studio’s preferences (Trip 1997).
Another uncertainty in the mix is how well a particular movie script/theme will be received.
Speculations about talents’ abilities are typically gauged based on the perceived public
perception. Studio executives attempt to match the right talent with the right script at the right
time to maximize the potential at the box office and the after market. The decision making
process is clouded by two layers of bias. The first bias originates from the customers’ perception
and the second originates from the executives’ perception of the market place.
In this paper, we argue that studying the evolution of public perceptions can shed light on factors
that drive these perceptions during development. We have crafted this argument by exploring a
data set from the Hollywood Stock Exchange (fwww.hsx.com)] an information clearing house
where visitors buy and sell virtual shares of talent and movies with a currency called the
Hollywood Dollar®. Hollywood Dollars are allocated to each new member at no cost because
Blockbusters: Building Perceptions and Delivering at the Box Office 2
they are a fictional currency. The company's virtual technology allows an unlimited number of
consumers to trade thousands of virtual entertainment securities in a fair and orderly, supply-and-
demand-based market. These data are an excellent proxy for the consumers’ perception of talent
and movies because of this free market environment. We explore the question: are market
assessments, and underlying decision rules, of a movie's stock on HSX a good predictor of the
movie's box office earnings and return on investment. We defined a decision rule as a numerical
composite of the weights that the market puts on different parameters that drive the stock price.
We conduct an aggregate analysis of a data set from a sample of 23 movies (and their associated
talent), released over a two-year period, using a simple single feedback loop model to estimate
decisions rule parameters. We then conduct regression analysis to statistically explore if the
underlying decision making process deals with a dynamic environment while accounting for
market feedback effects. The results suggest that, consistent with the findings of other dynamic
decision making situations (see Sterman 1989; Kampmann 1992), the HSX market as a whole is
a poor predictor of return on investment and box office success. However, when we segment the
data into populations of winner and losers, we illustrate that the estimated decision rules can
provide selective insights into the biases within the decision making process.
The rest of the paper is organized as follows. We begin by describing the decisions associated
with the movie development process and how these decisions reflect the reference modes derived
from the HSX market place. This is followed by a short discussion of the literature on estimating
decision rules in dynamic markets. We then describe our constructs, model formulation, data
collection, and analysis methodology. This is followed by a presentation of the statistical
findings. The paper concludes with a discussion of the limitation of our approach, managerial
implications of these findings, and possible extensions for this work.
2. Decisions Associated within the Movie Making Process
The movie making process consists of four very distinct stages, with equivalent “phases” and
“gates” in the conventional NPD terminology where projects are approved, recycled, or canceled
Blockbusters: Building Perceptions and Delivering at the Box Office 3
(Cooper, 1994; Ulrich and Eppinger 2000). These stages are - development, green light,
production, and distribution.
2.1 Development
A project in the ‘development’ stage has been typically sourced from books, characters, plays, or
simply an idea. It is not uncommon for movie projects to be simply titles at this stage with no
script attached. The development process consists of developing a story through numerous
iterations on a script. Content for a script will be found through story meetings, research,
interviews, and multiple writers rewriting drafts of scripts.
Projects can stay in the development stage for as long as a studio chooses. Studios will, from
time to time, sell projects to each other. Some projects are considered to be ‘fast tracked’ which
usually means that they are on track for being produced within a few months. Often times, titles,
which are fast tracked, have acting talent already associated with them, and script drafts are
customized to the particular talent.
Major studios (Disney, Wamer Bros., Universal, etc.) will spend millions of dollars developing
projects with many of them never reaching the next stage. Development expenses can range
from $100 thousand to $10 million per movie. Typically, the high development expense films
are those that require extensive research (documentaries) or extensive consulting for realism (e.g.
Pearl Harbor). Usually about 1 in 10 projects, which are in development, make it past the
development gate at the major studios.
2.2 Green Light
The green lighting process at most studios consists of negotiating talent assignment and
solidifying the production budget. It is typical for a producer to already be attached to the
project by this point either because the producer brought it to the studio, or because a producer
was attached during the development stage. The studio then faces a ‘chicken and the egg’
problem where a director will usually accept assignment to a project only if a certain
actor/actress is attached, and vice versa (unless of course the project was brought to the studio as
a package deal, which is not uncommon). Once the top acting talent, the director, the shoot
Blockbusters: Building Perceptions and Delivering at the Box Office 4
schedule, and the budget have been agreed to - the studio chairman will ‘green light’ the project
moving it to the production stage. Approximately 1 in 5 projects, that pass through the initial
development gate makes it through the green lighting process.
2.3 Production
During production the budget, plan, and schedule developed during the “green light” stage are
executed. The production budget consists of ‘above the line’ (ATL) and ‘below the line’ (BTL)
expenses. ATL expenses are talent related expenses, which can make up nearly 50% of the total
production budget of the film (e.g. The Sixth Sense). BTL expenses are the actual shooting
related expenses (i.e. set design, travel, craft services, etc.)
Studios are able to accurately predict the shoot schedule of most films, which in tum allows them
to accurately determine a release date. Along with the shoot schedule, a studio will take into
account their film pipeline (a.k.a. the film slate) and competition so as not to cannibalize their
revenues or mismatch the release against other films.
Once a reasonable amount of footage has been shot and a release date has been decided upon, the
studio’s marketing group can begin assembling the marketing collateral (i.e. trailers, one-sheets,
etc.) and the actual marketing investment. The marketing investment is typically not a function
of the production budget of the film, but rather the studio executive's expectation of the film’s
performance at the box office. Trailers are released to the public at this point creating an early
awareness of the film. These trailers often serve as the first point of contact with users of HSX
and typically cause the most noticeable increase in a movie's stock price. There is quite a bit of
time invested in positioning the movie appropriately in the marketplace since many viewers use
this form of marketing as their final decision on seeing the movie.
2.4 Distribution
Figure 1 shows the typical evolution of the cumulative box office revenue and the number of
theaters showing the movie. It is important to note that the box office returns evolve rather
quickly, in 10 weeks or so, compared with the development cycle that takes more than 100
weeks. In anticipation of the box office run, and concurrent with the marketing group’s
Blockbusters: Building Perceptions and Delivering at the Box Office 5
preparation, a studio’s distribution team decides the number of theatrical screens and the delivery
schedule. The number of screens is also based on the studio’s expectation of the film. Typical
blockbusters are released on about 3,500 screens (Spider-Man reached about 3,800 at its widest
point). Eliashberg et al. (2000) have modeled prerelease market evaluation of motion pictures,
however their work does not account for a web based perception-tracking systems.
Pearl Harbor
Weekly Box Office and # of Theaters
$80,000,000
$70,000,000
$60,000,000
$50,000,000
$40,000,000 —<—Box Office
$30,000,000 ——Theaters
$20,000,000
$10,000,000
$0
y § & ©
Se 6 6
wi Ra a
Sy Sy
ge” gh” wh” AP” WP
AM AP MO”
Oy We WO Le
Figure 1: Box Office Revenue and Theater C overage for the movie Pearl Harbor
The marketing group also works in conjunction with the film’s producer and director in
conducting test screenings. These test screenings are designed to give the marketing team an
idea of the demographics that the film appeals to (via analyzing data obtained in audience
screenings) and to indicate to the director portions of the film that may need to be changed. It is
not uncommon for a movie to make significant amendments to the film at this stage ranging from
the downplaying of a character to fundamentally changing portions of the movie. In some
isolated cases the release date is changed to match the public’s interests or to preempt
competition.
Blockbusters: Building Perceptions and Delivering at the Box Office 6
3.0 HSX and Decisions of Movie Making
The Hollywood Stock Exchange is an on-line marketplace. It treats Hollywood talent like
financial securities and allows for a market price to be determined through active trading. The
typical user of HSX is a person who has a fair amount of knowledge about the entertainment
industry, reads trade journals such as Variety, and has a socially curious interest in ‘the
business.’
Talent (producer, director, actor and writers) bonds are constantly traded without being de-listed
unless the talent is no longer active (i.e. retired, deceased, etc). Typically, a talent’s bond rises
and falls are directly attributable to their current productions (see Figure 2 for M. Night
Shyamalan's lifetime bond value. In his case there was an IPO on 11/24/99 right after the
release of The Sixth Sense. An IPO is an initial public offering, which for HSX means that it is
the first time that the stock or bond was available for purchase by investors.
Evolutions of a selected set of movie stocks, some winners and other losers, are the reference
modes that we will explore systematically in the second half of this paper.
1400
ry
4 1050 =
g 700 \
ZB 350
8 \_
0
6-Nov-99 20-Mar-01 2-Aug-02
Elapsed Time
Figure 2: Lifetime Bond Value for M. Night Shyamalan
When a movie is first announced, its Stock receives an IPO price and it is available for trading
until a few weeks past the release date of the film. During this time users can buy and sell the
stock at their own discretion. Figure 3 shows the stocks of “Harry Potter and the Chamber of
Secrets” (IPO on 5/01/2000 under the name HPOT2, released on 11/15/2002; Life = 135 weeks)
Blockbusters: Building Perceptions and Delivering at the Box Office 7
and “Ali” (IPO on 12/14/1999 and released on 12/25/2001; Life = 104 weeks). Trading is heavy:
HPO2 was traded 3,384,340 times over its life. Harry Potter tumed out to be a winner at the box
office where as Ali was a loser in terms of its box office collections.
As the movie gets closer to completion there is a greater amount of information known about the
movie, which in tum would affect the perceptions of HSX users. Preliminary movie trailers are
probably the most descriptive information that becomes available causing active buying and
selling shortly after trailer releases. Additionally, there are a number of entertainment
periodicals, which track movie productions very closely. The Hollywood Reporter has a weekly
issue that details changes to active productions throughout the industry. Presumably, with a
trained eye, one can see the impact of budget overruns and changes of release dates as being
positive or negative for the referenced stock.
St 200
oc
k
Va 100
lue
in
0 50 100
Elasped Time Since IPO in Weeks
— Harry Potter Il nnanann: Ali
Figure 3: Reference Modes for the Stocks of Harry Potter-II and Ali
(Data sets terminate upon theatrical release)
The evolution of stock and the estimated box office revenue is of interest in the equity analysis
community as well. Besides the obvious application of the model’s output to determine which
movie studio has the most powerful and potentially successful pipeline of products, it can also be
used to improve tax and financial reporting procedures. The improvement will come from having
better estimates of the lifetime revenues of the movies, which can then be more accurately
depreciated, based on incoming revenue (Lesley 1996).
Blockbusters: Building Perceptions and Delivering at the Box Office 8
We also view this data as a natural experiment for studying market dynamics. Data on market-
based assessments of products while they are under development are rarely available. In many
instances, scholars have studied the decision making process underlying such markets by running
controlled experiments. There is a rich tradition within the system dynamics literature for
estimating aggregate decision rules in controlled settings that simulate either idealized markets
(Kampman, 1992) or industrial situations such as Beer Distribution (Sterman 1989), Real Estate
(Bakken, 1993) and Service Supply Chains (Anderson and Morrice 2000). Much of this literature
illustrates that actors tend to ignore much of the dynamic stimuli, relying on their mental models,
and in general perform sub-optimally in such settings.
These data lend themselves to estimation of market decision rule parameters in a manner
suggested by Oliva (2003b). In his work, Oliva had used the calibration capability of the system
dynamics simulation tools to make optimal estimates of model parameters. Oliva has also argued
for the use of Theil Statistics to assess the goodness of fit between the observed and the
simulated data. In the following section we build on this methodology to inform our research
design and to develop cross-sectional estimates of decision rule parameters within our data set.
4. Research Design
Our research design seeks to:
(i) model and estimate the decision rules that guide the evolution of our reference mode,
namely the time history of a movie stock from it’s IPO to the release;
(ii) assess, in aggregate, if the estimated decision rule parameters influence the box office
and critical performance of these movies
4.1 Modeling the Decision Rule
We build on the system dynamics tradition (Sterman 2000) by modeling perception as a stock
that seeks to reach a target value. The target value is driven by the decision rules for the HSX
market place.
Movie_Stock(t+1) = Movie_Stock(t) (1)
+ {Target_Value(t) - Movie_Stock(t)} / Time to Adjust Perceptions
Blockbusters: Building Perceptions and Delivering at the Box Office 9
Target_Value (t)= Wi*Actor; (t)+ W2*Actor(t)+ W3*Actora(t) (2)
+ Wa *Director(t) + Wn* t +W.p* Completion Pressure(t)
Where:
* W1, W2, Ws, Wa, Wu, Wop are the unknown parameters within the decision rule. We assume
that these parameters are time invariant.
¢ Actor; (t) , Actor, (t), Actors(t) and Director(t) are known parameters for the talent pool’s
value in the HSX market place
¢ tis the elapsed (a.k.a. ramp up) time in weeks since the IPO of the movie on HSX
¢ Completion Pressure (t) = 1/ In (Release Date - t); the Release Date is measured as the
elapsed time since the IPO in weeks.
We justify the specification of the decision rule for the target value based both on literature and
on anecdotal evidence around practices within the movie industry. Recall from the discussion in
sections 1 and 2 that executives in the movie industry put a lot of emphasis on the choice of the
talent pool while funding the movies. Also, recall that these data are immediately available as
cues to the HSX users through a web-based interface. The choice of variables associated with
timing, i.e. elapsed time and completion pressure, has been identified as key drivers for
performance in the NPD literature (Ulrich and Eppinger 2000).
4.2 Estimation of Decision Rule Parameters
We follow the procedure suggested by Oliva (2003a) while calibrating the model outcome
against the time series for stock performance. The fitted model seeks to minimize the gap
between the estimated and the observed time series by selecting the decision parameters: W1,
W2, W3, Wa, Wr, Wp and Time to Adjust Perceptions. (see Appendix) The default settings
for the input parameters and optimization control parameters during the search are listed in the
appendix.
We save the time series for the fitted model and compute Theil statistics for goodness of fit using
a free-ware module made available by Oliva (2003b). While accepting the fit, we wish to
minimize the bias and maximize the co-variation between the fitted and observed data set.
Blockbusters: Building Perceptions and Delivering at the Box Office 10
Indicators of the goodness of fit statistics for the aggregate data are presented in section 5 along
with the results.
4.3 Decision Parameters and Ex-Post Performance of Movies
Estimated decision rule parameters are regressed against the ex-post performance of the movie at
the box office (Box Office) and the associated pay off (Pay Off) taken from www.baseline.com.
The regression models are specified as:
Box Office =o + Qi*Wit O2«W2+ O3«W3+ QaxWat O5*Writ O6«Wep +
7 »Time to Adjust Stock + &0 (3)
Pay Off = Bo + Bi+Wit By*W2+ Bs+Ws+ BasWat Bs+Wr+ BoxWep
B:« Time to Adjust Stock + €o (4)
Where:
* Box Office is the earnings (in millions of dollars) from the movie at the box office after 10
weeks of run.
« Pay Off =(4* Box Office - Cost)/Costs; The Box Office Earnings have been multiplied by a
factor of 4 to reflect the gains from the after market. This 300% mark up a commonly
assumed benchmark in the movie industry, it also reflects the idea that earnings from the box
office and the after market are correlated. Costs are the reported cost of development and
production.
© €o and €po are the noise terms.
We have ignored several fixed factors, e.g. timing of release (e.g. the 4th of July weekend effect)
and a Studio’s portfolio effect in these specifications.
4.4 The Data Set
We have collected data on a total of 23 movies released over a two-year window in 2001-2002.
The sample includes the top 12 movies that did well in the box office. The rest were selected
from a random sampling of the population. In some instances, complete data for lead actors were
not available on HSX because these actors were not traded when their movie IPO took place. In
such cases, we have used a truncated data set as long as complete data are available for at least
half the life cycle of the development; otherwise the movie was eliminated and the next movie
from the sample list is included.
Blockbusters: Building Perceptions and Delivering at the Box Office 11
We are also in the process of collecting weekly data from HSX’s database. For the purpose of
this analysis, our time series data capture the break points, with data for the intermediate weeks
generated via linear interpolation. Implications of this assumption on the interpretation of results
are discussed in section 6.
5. Results
We begin this section by presenting the calibration results, i.e. estimates of the decision rule
parameters and allied goodness of fit statistics. These are followed by the regression results.
5.1 Calibration of Decision Parameters
For the movie Harry Potter and the Chamber of Secrets (HPOT2) the values of estimated
parameters are W; = 0.6; W2 = 0.23; W3 = -.01; Wa= 0.49, Wm = -1.08, Wop = 543. The
corresponding Theil statistics are: R*2 = 0.9994; Bias = 0.00465; Variation = 0.04828;
Covariation = 0.947. Recall that HPOT2 has been a deemed as a blockbuster or a “winner,”
based on its box office retums.
On the other hand, “Ali” is a less successful movie and has been termed a “loser.” The values of
estimated parameters for Ali are: Wi = 0.0007; W2 =0.196; W3 =-0.012; Wa= -0.12, Wm =0.90,
Wop = 36. The corresponding Theil statistics are: R*2 = 0.959; Bias = 0.0019; Variation = 0.127;
Covariation = 0.871.
Recall from Figure 3 that both movies start out with relatively similar stock value and growth
pattern. However, Ali’s stock falters when it reaches close to the release date. This loss is
mirrored within the estimated parameters. We interpreted these parameters as follows; in the case
of Harry Potter, Wp drives the HSX stock price higher as the movie nears completion (i.e. the
completion pressure increases). On the other hand, for Ali, Wp does not drive the HSX stock
price up. Additionally, it is also interesting to note that the market does not place a premium on
the lead talent in Ali (i.e. Will Smith) but places a relatively higher emphasis on the talent pool in
Harry Potter. While it is difficult to generalize which parameter will have the largest impact on
the overall perception of movie stocks, we have shown that by analyzing the HSX data it is
Blockbusters: Building Perceptions and Delivering at the Box Office 12
possible to assess these parameters during movie development, based on the most up to date
information. The managerial implications for use of these parameter estimates will be discussed
further in Section 6.
We have compiled similar estimates for all the movies in the sample. We present the average
values (and the associated standard deviations) for the decision parameters in Table I, separated
into two sub-samples: box office winners and losers. Winners and losers are delineated based on
box office revenue for the purpose of discerning whether or not “Investors” use two distinct sets
of decision rules in valuing movie stocks. The corresponding Theil statistics, in Table II,
confirm the low bias and high co-variation in the estimated data set. Again in the manner
described above, the aggregate statistics suggest that the completion pressure parameter (W ,p) for
the winners and losers indicate different contributions. Whereas the elapsed time, i.e. the ramp
up, parameter (W,) shows comparable contributions.
Table I: Estimated Decision Rule Parameters
Time to Adjust wil W2 W3 wd Wep Wru
Box Office Losers[Average 5.395 -0.053 0.155 | -0.168 | -0.074 |-187.105| 0.824
(n =12) St. Dev. 2.457 0.100 0.403 | 0.475 0.425 | 974.808 | 0.686
Box Office Average 6.329 0.064 0.008 | -0.011 | 0.005 | 305.938] 0.823
Winners (n = 11) [St. Dev. 1.951 0.185 0.142 | 0.141 0.186 | 324.537 | 1.108
Table II: Theil Statistics for fit between the simulated and observed Stocks.
Sample Size] R72 Bias | Variation |Covariation |Mean Abs. % Error,
Box Office Losers [Average 82.417 0.958 | 0.002 0.023 0.976 0.083
(n=12) St. Dev. 29.516 0.078 | 0.004 0.037 0.037 0.063
Box Office Average 92.400 0.989 | 0.002 0.011 0.987 0.061
Winners (n=11) [St Dev. 52.846 0.009 | 0.002 0.014 0.015 0.070
5.2 Regression Analysis
We show the aggregate statistics for the dependent variables, i.e. the Box Office earings and the
Pay Off in Table III. We regressed the Box Office earnings against the estimated parameters
according to the specification in equation (3). The results are shown in Table IV.
Blockbusters: Building Perceptions and Delivering at the Box Office 13
Table III: Aggregate statistics for the dependent variables
Pay Off Box Office Earnings Cost
Aggregate Value Ratio* in Million $ in Million $
Average 5.886 161.130 92.457
Standard Deviation 2.756 81.684 25.646
n=23 * More is better
These results indicate that for the overall sample the variance in box office performance cannot
be explained by the estimated coefficients in the decision rules. However, the R? increases when
these data are divided into losers and winner sub-samples.
We also regress the pay off against the decision rule parameters following the specification in
equation (4). Recall that the pay off is defined as the ratio of profit and the cost. We illustrate in
table V that none of the models are statistically significant. Hence we conclude that payoff
expectations cannot be explored using the HSX data. On the other hand, the nearly 71% variance
in the “winner” sub-sample is explained by the model shown in equation (3), where as only 18%
variation box office performance of the “loser” sample can be explained by this model. The most
important conclusion is that the decision rules for the entire population should not be pooled
while analyzing the HSX data. This analysis also indicates that a and ag are statistically
significant coefficients. .
Table IV: Regression Results with Box Office Earnings as the Dependent Variable
Symbol |Corresponding |Winners Losers All
Parameter
Coefficient |p Coefficient |P Coefficient |p
% Intercept 206.84 0.03 {14.32 0.82 |97.30 0.18
en wil 489.70 0.11 |-37.09 0.95 |170.34 0.39
O W2 -9.05 0.97 |-361.68 0.27 |-41.69 0.88
3 W3 456.44 0.03 |-510.54 0.24 |-94.93 0.64
4 wd -306.53 0.13 |-110.51 0.53 |-117.01 0.30
Os Wop -0.13 0.14 |0.14 0.39 |0.08 0.22
OG Wu 25.51 0.19 |38.29 0.36 14.99 0.83
7 Time to Adj 2.32 0.73. |7.30 0.53 |7.82 0.43
N 11 12 23
Adj. R2 0.718 0.180 -0.067
Blockbusters: Building Perceptions and Delivering at the Box Office 14
Table V: Regression Results with Pay off as the Dependent Variable
Symbol |Corresponding |Winners Losers All
Parameter
Coefficient |p Coefficient |P Coefficient |p
Bo Intercept 3.80 0.73 (9.32 0.07 97.30 0.07
Bi Wi -3.30 0.81 (26.14 0.26 |-41.69 0.91
Bo W2 8.86 0.55 |-11.39 0.25 |-94.93 0.76
Bs W3 23.00 0.32 |-8.56 0.25 |-117.01 0.67
Ba Wd 4.16 0.79 1.48 0.68 —|0.08 0.59
Bs Wop (0.00 0.84 0.00 0.54 |4.99 0.41
Bs Wa 1.34 0.39 |-1.53 0.35 |7.82 0.81
B; Time to Adjust {0.43 0.77 |-0.68 0.16 |170.34 0.68
N 11 12 23
Adj. R2 -0.305 0.180 -0.235
6. Discussions
Our results show that it is possible to explore the perception of the market place about a movie,
while it is being developed, by estimating the decision rules followed by an interested set of
observers on HSX. We have also shown that one basic difference in winners and losers lies in
Completion Pressure. Using HSX as a sole source or measure of future box office potential can
be misleading. However, using the HSX data in conjunction with other cues can be useful. It
might be appropriate to reflect on why HSX data might be correlated with the eventual box
office performance and why some of the parameters, such as and 0g, are significant for the
winners. We speculate that the HSX data and underlying processes are indeed endogenous to the
perception of box office success and drive the mindset of the executives to make a movie
successful. This occurs because of the long development time (> 100 weeks on average) and the
relatively short box office run (~10 weeks). In the rest of this section, we discuss the implications
of our findings, their limitations and suggest some extensions that will improve this work.
6.1 Limitations
There are some confounding features within our results. It is difficult to explain why ay
contributes to the success, where as a and a do not? Our choice of assignment of the talent to
the slots Actor_1; Actor 2 and Actor_3 is arbitrary. The market may be putting different weights
Blockbusters: Building Perceptions and Delivering at the Box Office 15
on actors, but our measurement may confound these signals, and that may be a reason why the
intercept (ao) is significant. We also think that our data set is not detailed enough, and thus filters
out some of the high frequency (i.e. low time constant) event. This may be a reason why the time
to adjust the perception is not a significant contributor to the regression results.
Competition between studios is fierce given the high stakes of the industry. Our model does not
capture the following that could be factored in as fixed factors for exploring the decision rules:
Studio Portfolio Effects: A studio’s brand value and portfolio can have an impact on the
performance of films. For example, the Disney brand name gives parents a level of comfort
in knowing that the film will adhere to certain family value standards. A studio’s portfolio of
movies may speak to their ability to handle the scope of a slate of films during a given year.
For example, from a cash standpoint, few studios can afford numerous blockbuster (high
budget) films in the same year. Additionally, physical capacity constraints may become an
issue from a production standpoint. There may be a particular sound stage which is
appropriate for two separate films, but it can only be used by one film at a time, with up to
six months of exclusivity.
Film similarity: - On numerous occasions, the movie industry has turned out similar films
within a short time frame from each other. This will have an impact on the performance of
one or both films as moviegoers tend to have a low tolerance for perceived duplication.
Other factors that could be included in this analysis are the quality of script, awards, talent
synergies, actors’ extra curricular activities, and limited entertainment wallet.
6.2 Implications for Studio Executives
Typically, users of HSX are reacting to publicly available information such as daily reports from
the shoot location and trailers released to theaters, television and the Internet. HSX data does a
good job of predicting the box office successes while it does a poor job of predicting box office
failures. Having said this, a studio executive would be interested in knowing if their movie is not
on this track, because they can take actions, which can impact the publics’ perception. We have
described the decisions made by the executives in section 2.3 that take place beyond the green
light phase. For instance, the marketing investment is typically not a function of the production
budget of the film, but rather the studio executive's expectation of the film. Currently, some of
Blockbusters: Building Perceptions and Delivering at the Box Office 16
this positioning is done using market research. Arguably, HSX like measures would be useful to
measure and manage the trending effects, by adjusting the release of trailers and allied collateral.
Aside from the marketing angle, a studio may employ risk reduction strategies while the film is
being produced if preliminary perception is negative. These are typically done in two ways:
Pre-selling - A studio may find it economical to seek a third party to distribute the film
in a particular territory (usually because a 3 party may have a better distribution
infrastructure or because the studio has little faith in the film).
Co-production - The financial risk of a full-length feature film can be so great, that
studios would seek financial partnerships with other studios to reduce their down side
risk. If the movie under performs, then the lead studio loses less than if they carried the
entire cost of the film. Similarly, their upside is equally reduced.
We note that currently movie production costs are capitalized and amortized based on the
percentage of expected revenues that were actually earned in a period of time. It is common
practice for the estimate of the ultimate revenue to change on a period to period basis depending
on how well the movie is performing. Changing the denominator in the amortization equation
can have drastic effects on the financial health of the movie studio. In addition, changing
amortization rates sends mixed signals to investors, and raises “red flags” within the IRS.
Neither of these effects is beneficial. Applying our model in the pre-release time frame to a
studios portfolio could improve both tax and financial reporting methods. The application of a
quantified model based on the public’s perception of talent and timing will produce more
accurate results than an executive’s perception of the public’s perception. In essence, the model
can remove one layer of bias from the existing accounting methods.
6.3 Implications in the Post Release Market
Knowing the estimated box office performance of a movie well before it is released in theaters
can allow downstream distribution channels to plan accordingly. The life span of a movie is
significantly shorter than in previous generations. For example, the home entertainment release
date of a movie typically occurs just days after a movie is pulled from theaters, whereas in the
past, it was up to a year after the theatrical run. Theatrical box office is the best indicator of the
performance of the movie in all other distribution channels.
Blockbusters: Building Perceptions and Delivering at the Box Office 17
Having an understanding of a movie's performance before release could significantly affect the
way future movie deals are structured for Hollywood talent. If the talent of a movie has an
understanding of the performance of their current film, they can decide to accept or decline their
next project based on the anticipated success or failure of the movie.
6.4 Extensions
Data on the evolution of box office earnings of a typical movie, as shown in Figure 1, indicates
that the earnings follow a goal seeking behavior. Such a behavior is typically an outcome of a
classic market diffusion model. The decision rule specified in equation 2 can be made
endogenous to the market diffusion model. Such a model can then be used to study issues such as
the share of distribution channels, ancillary revenue analysis, and theater capacity allocation.
7. Conclusion
We have applied the system dynamics methodology to a novel class of development scenario,
i.e. the development of movies. We model the decision rules that drive this market place and
estimate the underlying decision parameters by calibrating the evolution of a selected sample of
23 movies released in 2001-2002. Our results show systematic differences in the decision rules
followed by the market for the eventual winners (a.k.a. the blockbusters) and the losers at the
box office. Regression analysis of combined decision parameters for eventual winners and losers
cannot explain the variance in the box office performance. However, segmenting these data
between winners and losers provides interesting insights about how the market perceptions
evolve. These simple results have managerial and accounting implications in the movie industry
and can be extended for the analysis of the NPD processes in other settings.
Appendix
Default settings for Optimization Process Using Vensim DSS 32 V 4.1
Type of Simulation: Calibration
Payoff Element: Movie Stock| Observed Stock/1
Optimizer: Powell-Random
Blockbusters: Building Perceptions and Delivering at the Box Office 18
Type: Linear
Max Iterations: 1000
Vector Points: 25
Absolute Tolerance: 1
Fractional Tolerance: 0.0003
Tolerance Multiplier: 21
Initialization of Calibration Parameters
Wi = 0.014 W2= 0.01 W3=0.03 Wa=0.02 Wn=0.01 Wop =0.01
Time to Adjust Perceptions: 1 Week
Constraint
0.5 <Time to Adjust Perceptions <7.0
References
Anderson, E.G., and DJ. Morrice (2000). “A Simulation Game for Service-Oriented Supply
Chain Management: Does Information Sharing Help Managers with Service Capacity
Decisions?” Production and Operations Management 9 (1), 44-55.
Bakken, B. (1993), “Learning and Transfer of Understanding in Dynamic Decision
Environments, ” MIT System Dynamics Group Publication D-4343.
Clark, K. and T. Fujimoto (1991), Product Development Performance. Boston: HBS Press.
Cooper, R. (1994), “Perspective: Third-generation New product processes,” Journal of Product
Innovation Management 11(1) 3-14.
Eliashberg, J., Jonker, J., Sawhney, M., Wierenga, B. (2000)."MOVIEMOD: An Implementable
Decision-Support System for Prerelease Market Evaluation of Motion Pictures,” Marketing
Science, 19(3), 0226-0243.
Kampman, C. (1992),“Feedback Complexity, and Market Adjustment: An Experimental
Approach,” MIT System Dynamics Group Publication D-4304.
Lesley, E. (1996), “Fatal Subtraction? Hollywood's Creative Accounting Gets a Rewrite,”
Businessweek March 11.
Oliva, R. (2003a), “Model Calibration as a Testing Strategy for System Dynamics Models,”
forthcoming European Journal of Operational Research.
Oliva, R. (2003b) “Vensim® Module to Calculate Summary Statistics for Historical Fit,”
available at hhttp://www.people.hbs.edu/roliva/research/sd/|
Sterman, J. (1989), “Modeling Managerial Behavior: Misperception of Feedback in Dynamic
Decision making,” Management Science 35(3), 321-339.
Sterman, J. (2000), Business Dynamics: Systems Thinking and Modeling for a Complex World.
New Y ork: Irwin-McGraw Hill.
Trip, G. (1997), “Turning Rough Takes Summer s Big Hits,” New Y ork Times, May 5, pD1.
Ulrich, K. and S. Eppinger (2000), Product Design and Development. New York: Irwin-
McGraw Hill.
ww.baseline.com] ex-post movie facts database.
Blockbusters: Building Perceptions and Delivering at the Box Office 19
Back to the Top
CC BY-NC-SA 4.0