Академический Документы
Профессиональный Документы
Культура Документы
This paper examines the financial strategies employed in the process of film production
with particular emphasis on the performance of Warner Bros during the period from 1921
to 1940. The setting of film budgets is interpreted within a context of the financial risks
involved, and in particular, a portfolio theory approach is found to provide a useful
framework for describing and analyzing risk. The rapid growth of the film industry in the
interwar years, together with the economic volatility of the period, produced a variety of
risktaking strategies, and this paper attempts to assess the relative success of these
strategies. r 1998 Academic Press
* The authors thank seminar participants at the Universities of Cambridge, North London and East
London, and delegates to the Ninth International Congress on Cultural Economics who made a
number of very helpful observations on earlier drafts of this paper. In particular, the authors
acknowledge the specific contributions of Stuart Archbold, Mark Casson, James Foreman-Peck,
Bernard Hrusa Marlow, and two anonymous referees.
† To whom correspondence should be addressed: Mr Michael Pokorny, The Business School,
University of North London, 277-281 Holloway Road, London N7 8HN, England. Fax: 0171 753
5051; E-mail: m.pokorny@unl.ac.uk.
196
0014-4983/98 $25.00
Copyright r 1998 by Academic Press
All rights of reproduction in any form reserved.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
No. of Pages—25 First page no.—196 Last page no.—220
RISK ENVIRONMENT OF FILM MAKING 197
1 Annual Hollywood production declined steadily from the peak of 678 films in 1927 to 489 by
1932. Production lifted to around the 520 mark for the three years of 1934, 1935, and 1936, but fell
below 500 again toward the end of the 1930s. There were a number of forces at play in accounting for
this decline including the transformation to sound during the late 1920s and substantially enlarged
budgets per film toward the end of the 1930s. Undoubtedly, the severity of the depression in the early
1930s was an important factor in an account of the fall between 1930 and 1932 (Finler, 1988, Film
Daily Yearbooks).
2 Greenwald (1950).
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
198 SEDGWICK AND POKORNY
cinemas, while the relative success of A films depended on it. Herein lies the
source of strategic thinking for these combines. In needing to attract paying
audiences the cinema chains required hit productions. Yet from Sedgwick’s
(1997) work on the consumption of films and stars in Britain during the 1930s,
only a small number of films could expect to become the season’s hits; between
1932 and 1937 only 8 to 15 films per season, from a stock of releases which
increased from 648 to 803, earned more than five times the mean box-office take.
The quest for hit production was elusive, yet necessary if the combine was not to
become totally dependent on the products of rival studios, with the consequent
probability of being subjected to opportunistic behavior, and explains the dispro-
portionate budgets available for such gambles. As King (1986, p. 162) has written,
‘‘What the lucky producer (of a ‘hit’ production) has, therefore, is a monopoly
(copy) right to a film which will give his company access to his competitor’s
screen time for a price.’’
This paper sets out to chart Warners’ emergence during the 1920s, to become,
by the end of the decade, one of the major Hollywood players. Based upon the
recently discovered William Schaefer ledger of production costs and box-office
returns of feature films released by Warners between 1922 and 1951,3 it traces the
studio’s response to the rapid expansion in domestic and overseas demand during
that decade followed by, in marked contrast, the Great Depression and subsequent
slow and uncertain recovery up to 1941. From the ledger, it becomes clear that the
size and organization of Warner’s filmmaking budget reflected strongly a chang-
ing set of box-office expectations on the part of its senior executives. In particular,
by treating the annual film output as a portfolio of investments it can be shown
that these expectations affected the risk stance of the business, which in turn
manifested itself in the content and form of the films being made.
3 The authors are grateful to Mark Glancy for allowing us access to the complete Schaefer ledger in
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
RISK ENVIRONMENT OF FILM MAKING 199
TABLE 1
Selected U.S. Statistics, 1922–1941, 1929 Prices
Real
personal Real Annual
disp. recreational cinema Real Average real B–O/
Population Unemployment CPI income expenditure admissions annual admission recreational
Year (ms) (%) (1929 5 100) ($bn) ($m) (m) B–O ($m) price ($) expenditure
distribution, and exhibition assets. During this period Warners’ assets grew from
$5 million in 1925 to $230 million by 1930.4
As pointed out above, this strategy of vertical integration adopted by Warners
was common to the other four dominant firms in the industry. In order to finance
this strategy all five had become incorporated during the 1920s and to a greater or
lesser extent had incurred debt. The cost associated with reequipping the studios
and cinemas for sound had been particularly onerous coming as it did at the tail
end of the drive to acquire cinemas.5 The basis for this expansion rested ultimately
with the popularity of film at home and abroad and expectations of continued
growth in demand. Table 1 shows that during the 1920s cinema admissions in the
U.S. doubled from a weekly figure of 40 million in 1922 to 80 million by 1929,6
while real personal disposable income increased by over a third. The major
players all had extensive worldwide distribution operations by the end of the
decade reflecting the international phenomenon of film popularity.7 The dramatic
downturn in economic activity in the U.S. between 1929 and 1933 found these
firms in a particularly vulnerable state. Cinema attendances and box-office
receipts, in nominal terms, fell by a third during these years as unemployment
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
200 SEDGWICK AND POKORNY
soared, although given the fall in the Consumer Price Index from 100 in 1929 to
75.6 by 1933, the decline in real revenue (12%) was not as dramatic. During this
time personal disposable income almost halved although, of course, this was
mitigated by the fall in prices.
One important observation derived from the statistics in Table 1 concerns the
resistance of the industry to the decline in economic activity; between 1929 and
1931 box-office revenue actually increased in real terms. It was not until 1932 that
the Depression began to make a significant impact on the film industry. Certainly,
as a proportion of overall expenditure on recreational pursuits, cinema going
became relatively more popular, increasing its share from 17% in 1929 to 22% by
1931 before stabilizing at 21% from 1934 as recovery slowly began to get under
way. The importance of cinema going can be further emphasized by considering
expenditure on just spectator amusements (cinema going, theatre, and spectator
sports), and noting that cinema going accounted for 64% of such expenditures in
1923, increasing to 82% in 1930, a level which was maintained throughout the
1930s.8
Warners’ reported profits fell from $14,514,628 in 1929 to a loss of over $14
million in 1932.9 However, while Paramount, Fox, and RKO went into receiver-
ship and changed their management teams substantially, allowing financiers to
take a much greater role at both the strategic and operational level, Warners, along
with MGM (Loews), maintained the confidence of their shareholders and bankers
and retained their management structures.10 Warners’ response to the recession
was singular; overhead costs were driven down through selling cinemas while
operating costs were rigorously managed. Warners had developed a reputation for
tight control of filmmaking budgets during the 1920s.11 This served them in good
stead during the height of the Depression. As Roddick (1983, p. 10) argues
(b)udgets could not continue to spiral and since the actual physical cost of filming had
increased, a proportional reduction had to be made in non-technical costs—that is to say, in
sets, schedules, stars and story material. . . The new cost conscious budgeting and meticu-
lous planning that would characterise the studio system through the 1930s had come to stay.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
RISK ENVIRONMENT OF FILM MAKING 201
TABLE 2
Annual Revenues, Costs and Profits of Warner Bros. Films Produced during the Period From
1921/1922 to 1940/1941, 1929 Prices ($’000)
Coefficient of
Rate of Average variation of
Films Domestic Foreign Total Production Film return production production
Season produced revenue revenue revenue costs profits a (%) b costs costs
commercial impact of the onset of the Depression in the 1930/1931 season. While
the decline in Warners’ film output can be partly explained by the rationalization
of production which followed the acquisition of First National,12 the decline in
cinema attendances from 1930, charted in Table 1, led to a retrenchment across all
Hollywood studios. Warners’ box-office receipts halved for those films released
between the 1929/1930 and 1930/1931 seasons.13 Revenues continued to fall with
the 1931/1932 batch of films and thereafter fluctuated widely around an upward
trend. The 1937/1938 crisis in economic recovery is strongly reflected in the fall
in domestic revenues for both 1939/1940 and 1940/1941 releases.
Foreign sales contributed between a third and a half of receipts toward overall
12 Films continued to be presented under the First National trade name until 1941.
13 It is important to note that there is a misalignment between Tables 1 and 2 dates. The latter
represent the period of film release. Hence, given Warners’ use of a 14-month amortization schedule,
the releases of one year would be expected to earn at least part of their box-office revenue during the
following year. A further complicating factor in any time-series analysis is that the studio adopted a
financial year which ended during the last week of August.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
202 SEDGWICK AND POKORNY
Bros accounts for the period. They take no account of the company’s accumulated assets and
liabilities, and the accounting conventions applied to these. Also, they do not represent performance
within the strict limits of a trading year, since films released in one season generated revenue during
the following season. Rather they are to be interpreted as reflecting Warner Bros’ annual performance
with regard to its pure filmmaking activities.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
RISK ENVIRONMENT OF FILM MAKING 203
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
204 SEDGWICK AND POKORNY
(w)ere strongly influenced by the economic and organisational system under which they
were produced. The classic Hollywood style, with its insistence on the unproblematic,
seamless narrative whose advancement is controlled by the destiny of one or more
individual characters, was determined by the economics of the production system. (1983, p.
28)
Yet its strategy for survival, resulted in films of quality that audiences wanted to
see.
It was a factory geared to turning out product on a regular basis. But the product was in a
sense artistic: if a work of art is defined not by the conditions of its production but by the
circumstances of its consumption, then the product manufactured by Warners (and other
studios) was undoubtedly art. (1983, p. 25)
(1935/1936), the Adventures of Robin Hood (1937/1938), The Prince and the Pauper (1936/1937),
The Private Lives of Elizabeth and Essex (1939/1940), The Sea Hawk (1939/1940), The Life of Emile
Zola (1936/1937), and the Story of Louis Pasteur (1935/1936) are notable examples.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
RISK ENVIRONMENT OF FILM MAKING 205
FIG. 1. Box-office revenue against production cost, 1921/1922 to 1940/1941 (1929 prices).
22 An extreme example is The Singing Fool (1928/1929), the highest revenue generating film
during the period, which generated receipts of $5,916,000, from a production budget of just $388,000.
This compares with the average production cost during the period of $326,000.
23 In order to derive the net profit generated by a film, data are required on the distribution costs of
the film in addition to its production costs. Unfortunately, a serious weakness of the information
presented in the Schaefer ledger is the absence of data on distribution costs, and therefore such costs
have had to be estimated. Sedgwick (1994) examined data from another film studio, RKO, for the
period 1930–1941, from which distribution cost data could be derived directly. The sample of 155
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
206 SEDGWICK AND POKORNY
FIG. 2. Film profits against total costs, 1921/1922 to 1940/1941 (1929 prices).
distribution costs). Thus Fig. 2 also reflects, indirectly, the rates of return earned
by the films (the ratio of profits to total costs), and highlights more starkly the
conundrum which studio heads faced. Thus, while low to medium cost films were
more reliable in the sense that they were less likely to make losses, they tended to
generate only half of Warners’ seasonal profits during the best box-office years.
Even though some low cost films produced very high profits, these were relatively
small in number and their overall contribution to annual profits was necessarily
limited. Conversely, higher cost films were more likely to generate higher profits,
but at an increasing risk of incurring losses. In the absence of any clear indication
that profits were positively related to total costs, the decision to bet heavily on big
budget productions would appear to be more inherently risky than the pursuit of a
strategy based exclusively on low to medium cost production. Indeed, the
information contained in Fig. 2 suggests that the premium to risk fell as total costs
increased, not only does the variance of performance appear to increase with
costs, but also the average rate of return appears to fall.
A breakdown of the data into the period leading up to the Depression and the
decade or so that followed, however, reveals distinctive patterns of performance.
Fig. 3 shows profits against total costs for the 193 films produced up to
RKO films, adopted by Jewell (1994), produced a very high correlation of 0.97 between film revenues
and distribution costs, the more popular a film the greater the costs incurred in distributing it. On
average, distribution costs represented 37% of film revenues. By applying this ratio to Warners’
revenue data a profit figure for each film can be derived as the difference between the total box-office
revenue generated by the film and the sum of the production and estimated distribution costs. These
are the profit figures shown in Fig. 2.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
RISK ENVIRONMENT OF FILM MAKING 207
FIG. 3. Film profits against total costs, 1921/1922 to 1928/1929 (1929 prices).
1928/1929, while Fig. 4 shows profit performance for the 678 films produced
from 1929/1930. For the 1920s it is apparent from Fig. 3 that a pronounced
relationship existed between the variability of profit performance and the level of
total costs; in general, more variable profit performance was associated with
higher costs. In addition, there would appear to have been a positive association
between profits and costs; while there might have been higher risks associated
with high cost film production, the higher cost films tended to produce higher
profits, on average.
From Fig. 4 it can be seen that during the 1930s, while there was still a
tendency for the variability of profit performance to increase with costs, this was
not quite so pronounced as in the case of Fig. 3. However, and more significantly,
Fig. 4 no longer appears to reflect a strong positive relationship between profits
and costs; to the extent to which the production of higher cost films involved
greater risks, Fig. 4 would imply that there were limited returns to such risk taking
during the 1930s. Figure 4 also emphasizes the much higher incidence of
lossmaking films during the 1930s.
The relationship between film profits and total costs can be examined more
formally by regressing profits on costs and testing for the significance of the slope
FIG. 4. Film profits against total costs, 1929/1930 to 1940/1941 (1929 prices).
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
208 SEDGWICK AND POKORNY
coefficient. That is, we can test for the significance of b in the relationship
Pi 5 a 1 bCi 1 ei, (1)
where Pi is the profit generated by film i and Ci is the total cost of film i. In order to
derive a valid significance test, account must be taken of the apparent heterosce-
dasticity in this relationship. The simplest assumption which could be made, and
one that would appear consistent with Fig. 3 and 4, is that the variance of the eis
are related to total costs. The specific relationship which produced the most
satisfactory empirical results was
Var(ei) 5 bC 2i , (2)
where b is some constant. Thus dividing Eq. (1) throughout by Ci produces an
estimating equation with homoscedastic disturbances, and therefore the appropri-
ate regression equation for testing for the significance of b is
Pi 1
5a 1 b 1 ui. (3)
Ci Ci
Estimating Eq. (3) for the films produced during the period from 1921/1922 to
1928/1929 produces the estimated equation (t-statistics in brackets)24
Pi 1
5 235.638 1 0.53
Ci Ci
(4)
(3.19) (9.18)
R 2 5 0.051 n 5 192
Thus the estimate of b is positive and significant, implying that during the period
from 1921/1922 to 1928/1929 an increase of $1000 in the total cost of a film
generated, on average, an additional $534 in profit, albeit with increasing levels of
risk.
The corresponding estimated equation for the data in Fig. 4 is
(Pi/Ci) 5 43.008 (1/Ci) 1 0.063
(6.05) (3.30) (5)
R 2 5 0.051 n 5 678
Thus the estimate of b is still positive and significant, but it is significantly lower
than the value generated by Eq. (4) implying that during the 1930s, only $63
additional profit was generated per $1000 increase in costs; there were signifi-
cantly lower returns to risk in the 1930s. However, it is apparent from Fig. 4 that
higher profit levels still tended to be generated by the higher cost films, even if
24 There was one observation which proved problematical, and this was the lowest cost film
produced over the period, which generated extraordinarily high profits. This was Lights of New York
which was produced in 1927/1928 at a cost of just $23,000 but generated profits of $766,000. This
observation was dropped for the purposes of estimating Eq. (3).
EEH0691
@xyserv2/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 donn
RISK ENVIRONMENT OF FILM MAKING 209
this did not translate into higher rates of return. The difference between Figs. 3
and 4 is that the higher cost films in the 1930s also ran the risk of incurring
substantial losses. However, annual estimates of Eq. (3) throughout the 1930s
revealed considerable variation in the estimates of b, and years during which the
value of b was insignificantly different from zero, namely, the low rate of return
years of 1930/1931, 1931/1932, 1936/1937, 1937/1938, and 1939/1940.
It would appear that the quest for hit production was not only more risky, the
variance of profits increased with production budgets, but also that such produc-
tions generated lower average rates of return. A solution to this apparent paradox
can be found in the joint product qualities of the studio’s portfolio of films. As
argued above, the production and exhibition wing of Warners, along with the
other major combines, were strategically bound by the need to meet cinema
capacity utilization targets. Low budget B-movies were generally designed as
support features, while big budget super As were intended not only to play at the
organization’s principal cinemas as single features for extended runs, but also in
the cinemas of their rivals. The degree of programmed support they received and
the lengths of runs at these principal cinemas were a function of their popularity.
Lower budget A films fulfilled both purposes, again depending on individual film
popularity. For the greater part the combines were able to guarantee the exhibition
of their lower budget films in support or supported roles in their in-house cinemas.
(Columbia and Universal together with a host of ‘‘poverty row’’ producers, who
did not own their own cinemas, also specialized in the production of support
features.) With much lower budgets to amortize, such guaranteed billing was
sufficient to earn such films healthy and fairly predictable returns. Although super
A films would also play in the organization’s main cinemas, as principal
attractions they were in effect competing for audiences with rival attractions in
similarly priced cinemas within a locality. Audiences in general did not go to the
cinema to watch low budget films. They were attracted by stars and production
values. Accordingly, the studios, through their publicity departments, were very
conscious of keeping their contract stars before the general public, primarily via
fan magazines, newspapers, and radio outlets. However, given that capacity
utilization was paramount, in-house super A films would only take up screen time
on merit. It is for these reasons that the return to such productions was so variable,
particularly in the context of a Depression which saw audiences fail to surpass
their 1930 peak before the onset of America’s entry into the war. It is also for these
reasons that the correct approach to analyzing studio performance is not one based
simply on the relative performance of all films emanating from the studio, but
rather their interrelationship as part of a portfolio of films designed to enhance the
capacity utilization of the respective cinema chains, and of course through this the
market share and the profitability of the organization.
For each of the studios, accordingly, the quest for hit productions was
intrinsically linked to the risk stance which was taken. In turn this was related to
the willingness to bet heavily on potential film properties and stars as well as
supporting these with high quality, idiosyncratic human and physical capital. At
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
210 SEDGWICK AND POKORNY
the margin this decision was a function of the extent to which an organization
could increase its absolute profits from its portfolio of films by investing more
heavily in features which might be expected to become hits, the corollary being
dependence upon other super A producers, with all of the attendant problems
associated with strategic vulnerability.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
RISK ENVIRONMENT OF FILM MAKING 211
would have resulted in an unchanged annual budget and can be interpreted as the
minimum rate of return required on film production; a lower rate of return would
have resulted in reduced budgets, and hence a shift in resources away from film
production.25
The robustness of the statistical results in Eq. (6) can be further illustrated by
‘‘undifferencing’’ the data on the left hand side of the equation, regressing ln Ct on
Rt-1 and ln Ct21, and examining the stability of the resulting coefficients. The
resulting regression is shown in Eq. (7).
ln Ct 5 0.940 1 0.014 Rt21 1 0.881 ln Ct21
(1.55) (3.90) (14.58) (7)
R 5 0.939
2
n 5 19 d 5 2.54
Thus in moving from Eq. (6) to Eq. (7) we can note that the coefficients on Rt21
are stable in the sense that they differ insignificantly, and that the coefficient on ln
Ct21 in Eq. (7) is insignificantly different from 1, consistent with the first
difference transformation in Eq. (6). We can therefore conclude that Eq. (6)
provides an efficient statistical summary of the manner in which the annual film
production budget was set.
Once the annual film production budget was determined, the next decision
related to how the annual level of investment was then distributed across the set of
projects earmarked by the studio executives. The approach taken here is to
distinguish between three categories of film production, high, medium, and low
budget productions. However, to define each of these categories in absolute terms
would fail to recognize the changing nature of film production over time.
Innovations such as sound and color, together with a general increase in the
sophistication of film production meant that the nature of the film product was
very much an evolving one. Thus the cost of a high budget film in the mid-1920s,
say, would differ markedly from that of a high budget film in the late 1930s. In
crude terms this can be seen from the general increase in average production costs
in Table 2.
Consequently the extent of high budget (and hence high risk) film production in
any given year is measured here as the proportion of the total film budget which is
absorbed by the top 20% most costly films produced during that year. Similarly,
25 A perhaps simpler hypothesis is to assume that annual production budgets were determined by
cash flow. This could be examined by regressing D ln Ct on the proportionate change in cash flow in
t-1. Cash flow could be measured either in terms of film revenues generated, or more realistically, in
terms of film profits. However, both measures produced poor statistical results, and in the case of
profits, insignificant results. Such a result is perhaps not surprising as a close association between
production budgets and cash flow would imply that finance for film production was derived largely
from internal sources, in addition to implying that it is only cash flow which is of relevance, in contrast
to the rate of return which was generated by film production. Thus Warners certainly reallocated
investment funds internally among its various divisions and subsidiaries over the sample period, in
addition to raising finance from external sources. It is therefore more plausible to assume that the scale
of film production was determined by rate-of-return performance, rather than the simple availability of
internal finance.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
212 SEDGWICK AND POKORNY
TABLE 3
Annual Rates of Return by Budget Category
the extent of medium budget film production is measured by the proportion of the
film budget absorbed by the next 40% of films in rank order. The extent of low
budget film production is reflected in the proportion of the annual film budget
allocated to the bottom 40% of films. Such an approach also has the advantage of
standardizing for the number of films which were produced each year. While the
choice of these partitions, the top 20% of films, the middle 40% and the bottom
40%, is essentially arbitrary, they serve to provide an adequate reflection of the
budgetary decisions which were made with regard to the various categories of
film production.
Table 3 shows the annual rates of return which were generated by each of these
budget categories. The final column of the table also shows the annual rates of
return on the total film budget, from Table 2, for comparative purposes. The first
data period in the table aggregates during the period from 1921/1922 to 1924/
1925, in order to derive sufficient observations, given the very low level of film
production in the early 1920s. Table 4 shows the annual allocations of the
production budget and the proportionate contribution of each budget category to
annual profits.
The rates of return generated by the top 20% of films follows a clearly defined
cyclical path, already noted in the discussion of Table 2, with peaks in 1928/1929,
1932/1933, 1933/1934, and 1940/1941, and troughs in 1930/1931, 1931/1932,
and 1937/1938. Further, while the peaks are marginally greater than those of the
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
RISK ENVIRONMENT OF FILM MAKING 213
TABLE 4
Budget Allocations and Profit Contributions by Budget Category
other two budget categories, the troughs are considerably deeper. As argued above
in relation to the discussion of Fig. 2, the volatility of this cycle when coupled
with the comparative under performance of films in this category, characterize the
risk environment faced by Warners’ executives.
From Eq. (6) and Table 3 a set of expectations can be formed concerning the
share of the annual budget taken up by the top 20% of films, based upon the rates
of return of the previous period; where rates of return had been rising the
expectation would be that the proportion of the budget would increase in the next
period; where they had been in decline the opposite would occur. Table 4 shows
this to have been broadly the case. The cyclical pattern is once again repeated so
that in periods of downturn the proportion of funds going to big budget
productions becomes very much more equal across the categories. Indeed in the
1931/1932 season the most expensive 20% of productions only absorbed 27% of
the global budget, having fallen from 49% in 1928/1929.
To this point the nature of risk in film production had been discussed in
relatively broad terms, without having used a specific definition of risk or deriving
an explicit measure of risk. Using a portfolio theory interpretation of the process
of annual film production however suggests a natural measure. By interpreting the
annual film budget as a three asset portfolio, the return on the portfolio will be a
simple weighted average of the returns on each asset, where the weights are the
proportions of the investment budget allocated to each asset. Thus the aggregate
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
214 SEDGWICK AND POKORNY
TABLE 5
Estimates of Risk on Film Assets and Film Portfolio
rate of return on the annual film budgets in the last column of Table 3 is equivalent
to such a weighted average.26 An overall measure of risk then can be derived from
the risks associated with each asset. The commonly used measure of asset risk in
the portfolio theory literature is the variability in the rate of return on an asset, and
specifically, the standard deviation of the rate of return. In terms of the Warners’
data the risk associated with a given budget category, for any one year, can be
measured by the standard deviation of the rates of return on the films falling into
that category. These annual standard deviations are shown in columns 2, 3, and 4
of Table 5. Finally by combining these standard deviations it is possible to derive
an overall measure of portfolio risk.
The basic result of portfolio theory is that the risk on a portfolio is not just a
simple weighted average of the individual asset risks. Rather the expression for
overall portfolio risk will involve the covariances (or correlations) between the
rates of return of the assets making up the portfolio, reflecting the risk reduction
potential of a diversified portfolio. In particular, if rT denotes the rate of return on
26 There is a slight complication here as the rates of return in Tables 2 and 3 are net rates of return in
the sense that they incorporate imputed distribution costs in addition to production costs. So strictly
the weights which should be used are the proportions of total costs accounted for by each asset, rather
than the proportions of production costs only. However, the use of weights derived from total costs
result in only relatively small differences, and we prefer to focus on production costs as these are a
more direct reflection of decision making and risk taking.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
RISK ENVIRONMENT OF FILM MAKING 215
the top 20% of films and rM and rB denote the rates of return on the middle 40%
and the bottom 40% of films, respectively, then the overall rate of return on the
portfolio, rP, is given by
rP 5 xTrT 1 xMrM 1 xBrB, (8)
where xT, xM, and xB are the proportions of the total production budget allocated to
the top 20%, middle 40%, and bottom 40% of films, respectively. The risk
associated with the portfolio, as reflected in the standard deviation of the rate of
return on the portfolio, is derived from the variance of rP in Eq. (8), and specifically27
Var (rP) 5 x 2T Var (rT) 1 x 2M Var (rM) 1 x 2B Var (rB) 1 2xTxM Cov (rT, rM)
(9)
1 2xTxB Cov (rT, rB) 1 2xMxB Cov (rM, rB).
Alternatively, noting that Cov (rT, rM) 5 sTsMrTM, where sT is the standard
deviation of the rate return on the top 20% of films, sM is the standard deviation of
the rate of return on middle 40% of films, and rTM is the correlation coefficient
between the rates of return on the top 20% of films and the middle 40% of films,
we can write
s2P 5 x 2Ts2T 1 x 2Ms2M 1 x 2Bs2B 1 2xTxMsTsMrTM
(10)
1 2xTxBsTsBrTB 1 2xMxBsMsBrMB.
The risk on the portfolio, then, is measured by the square root of this expression
or sP.
Annual estimates of sP can be derived by substituting into Eq. (10) the standard
deviations of the rates of return in Table 5. The correlations between the rates of
return can be estimated by the time-series correlations between the rates of return
in Table 3.28 These estimates of sP are shown in the fifth column of Table 5.
While the rate of return and risk data in Tables 3 and 5 provide a useful insight
into the financial environment of film production during the 1920s and 1930s, the
data strictly reflect the outcomes of film production decisions, rather than the
explicit strategic decisions taken by Warner Bros during the period. Hence the risk
estimates in column 5 of Table 5 reflect risk which was experienced or incurred,
rather than the risks which were actually taken. That is, the portfolio risk measure
used here reflects the influence of environmental/external factors, in addition to
the explicit production decisions taken by Warners. However, it would be useful
to examine the way in which Warners adjusted their own risktaking strategy over
time in response to the return-to-risk performance of their film portfolios.
A measure of the risk stance taken by Warners in any given year can be derived
by considering the variability within their annual production budgets. Specifi-
cally, by interpreting the coefficient of variation of the production costs for a given
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
216 SEDGWICK AND POKORNY
asset as a measure of the risk taken on the asset, Eq. (10) can be used to derive an
estimate of the overall portfolio risk stance by substituting these coefficients of
variation for the various s terms (and then taking the square root). These
estimates are shown in the last column of Table 5.
It is instructive to graph the portfolio rates of return against estimates of the risk
taken in each year. This is shown in Fig. 5. In broad terms Fig. 5 provides
evidence that there are returns to risk, a general tendency for returns to increase
with the level of risk taken. In addition, higher levels of risk produced more
variable rates of return performance, as would be expected. Figure 5 also implies
a much stronger positive relationship between return and risk for the 1920s, than
is the case for the 1930s. This is consistent with the much more volatile
environment during the 1930s. The distinction between the risk environment of
the 1920s and the 1930s can be seen more clearly in Figs. 6 and 7, which present
separate graphs for each decade.
What these measures highlight is the cyclical nature of the risk stance adopted
by Warners in which increments in rate of return performance for any risk stance
taken in period t-1 led to an increase in the risk taken in period t, up to the point at
which the sequence is broken by an exogenous shock. The onset of the Great
Depression caused the studio to dramatically revise downward its risk stance, so
that it was a third of its pre-Depression level in the 1931/1932 season. Improving
box-office performance thereafter results in a more positive risk position, only to
be knocked back once again by the downturn in economic activity in 1938.
These observations can be formalized by considering a regression equation
which explains the change in Warners’ risk stance from year to year. The
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
RISK ENVIRONMENT OF FILM MAKING 217
dependent variable would be the proportionate change in risk taken. Denoting the
risk taken variable in Table 5 by RT, then the dependent variable is D ln RTt. Thus
it can be argued that the change in the risk stance would have been influenced by
the rate of return performance of the portfolio in the previous year relative to the
risk taken during that year. Such a variable would reflect the returns to the risk
which was actually taken, rather than the risk which in the event was experienced.
As such it might have been expected to influence the structure of the following
year’s film portfolio and in particular, the extent to which the risk taken on that
portfolio should change. The following relationship can be hypothesized, where R
denotes the rate of return on the film portfolio, as previously,
1RT2
R
D ln RTt 5 a 1 b 1 et. (11)
t21
In estimating Eq. (11) there were two observations which caused difficulties, and
these were the observations for 1929/1930 and 1932/1933. The 1929/1930 season
was the first season in which the First National production portfolio was included
with the Warners’ portfolio, following Warners’ acquisition of First National
during the previous year. This resulted in a 278% increase in the total production
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
218 SEDGWICK AND POKORNY
budget and a more than doubling of the number of films produced (see Table 2).
While this jump in the scale of film production was rational from Warners’
perspective given the rapidly growing demand at the time and the success they
were experiencing, they inherited a film portfolio which did not necessarily reflect
their own risk stance. Indeed, the First National portfolio consisted of films with a
lower average cost and lower variability than Warners’ portfolio in 1929/1930.
Consequently it might be argued that the risk stance taken in 1929/1930 was
somewhat more conservative than might have been the case had Warners’ had full
control over the construction of the 1929/1930 First National portfolio. In addition, it
might also be argued that such a large increase in the total production budget may have
induced a degree of conservatism in Warners’ own risk taking. For these reasons a
dummy variable for 1929/1930 was included in Eq. (11) (denoted by D30).
The problem with the 1932/1933 observation is the issue which was com-
mented upon in the discussion of the coefficient of variation data in Table 2 (see
Note 18). That is, the coefficient of variation of the low budget films in 1932/1933
was swamped by the six very low cost Westerns made in this year, producing an
overall coefficient of 50.9. By comparison the coefficient of variation of the low
budget films in the previous year was 17.9 and for the following year it was 24.3.
In turn this abnormally high coefficient dominates the RT measure for the
portfolio, even though the low budget films accounted for only 20% of the film
budget (the coefficient for the high budget and medium budget films for 1932/
1933 were 13.4 and 13.9, respectively).
Rather than drop these six observations from the sample the approach taken
was to interpret these six films as effectively two films. Thus they were divided
into two sets of three films and then costs, revenues, and profits were aggregated
across these two sets. The films each cost the same to produce and each generated
virtually identical profits. Such an approach can be justified by arguing that the six
films were of the same genre (Westerns), used the same stars (all starred John
Wayne), used a small core of technical staff, and effectively just repeated an
identical formula in their production. The result was that the coefficient of
variation for the low budget films in 1932/1933 was reduced to 25.79 (although
low budget films now accounted for 25% of the 1932/1933 production budget).
The RT measure produced was 15.0 (in contrast to 19.9 in Table 5).29
The resulting estimate of Eq. (11) is shown in Eq. (12).
D ln RTt 5 20.094 1 0.172 (R/RT)t21 2 0.836 D30t
(1.39) (2.91) (4.16) (12)
R 5 0.589
2
n 5 16 d 5 1.96
Therefore these regression results are consistent with the notion that portfolio risk
stance adjusted positively to return on risk.30 It is certainly not being suggested
29 Dropping the six films from the sample produced an RT value of 14.4.
30 Equation (7) can be further improved by adding a dummy variable for 1930/1931. Equation (7)
EEH0691
@xyserv2/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 donn
RISK ENVIRONMENT OF FILM MAKING 219
CONCLUSION
This paper makes a contribution to the growing field of research into film as a
business. A notable recent contribution is that of De Vany and Walls (1996), which
examined the nature of information dynamics and transmission mechanisms in
the generation of box-office revenues, and in turn how such processes result in
box-office ‘‘hits’’ and ‘‘flops.’’ However, their analysis derived from contempo-
rary data and an environment which was (and is) dominated by independent
production companies. By contrast, the analysis here focuses on the much more
tightly controlled studio system of the interwar years, and an industry which was
characterized by a high degree of vertical integration.
It is clear from the evidence presented that Warners responded to changing
economic and competitive circumstances in a strategic fashion: During the 1920s
expectations of an expanding market, principally at home but also overseas, led to
a series of heroic corporate decisions in the direction of vertical integration, while
the onset of the Depression caused it to produce fewer films and alter the cost,
genre, and star characteristics of its annual portfolio of films. Although it sold
some cinemas during these lean years, the vertical structure of its business, like
that of its main rivals, was left intact. In particular, three distinct stages can be
distinguished during the 1930s: The period of survival following the onset of the
Depression when average film costs were reduced dramatically; the slow return to
confidence as both annual portfolio and average film budgets crept up; and finally,
the move toward exclusive A film production schedules at the end of the period.
In drawing attention to the variability and structure of production budgets
through the application of portfolio theory it is clear that during Hollywood’s
classical period Warners took a strategic approach to risk, both in terms of the
global sum they were willing to invest for any single production season and the
manner in which it was spread across the set of production projects. The results in
predicts a relatively large increase in risk taking for 1930/1931, on the basis of satisfactory return to
risk performance of the 1929/1930 portfolio. However, this increased risk taking does not in fact take
place. The inclusion of the dummy variable for 1930/1931 might be justified by arguing that the
1930/1931 portfolio was constructed in the immediate aftermath of the 1929 Wall Street crash, and
thus in an environment of considerable uncertainty and volatility. Thus irrespective of the satisfactory
performance on the 1929/1930 portfolio, this major structural change inhibited risk taking severely.
Thereafter the approach to risk taking returns to that implied by Eq. (6). Including a dummy variable
for 1930/1931 increases R 2 to 0.813; the coefficient on the return to risk variable is virtually unchanged
at 0.174, but now with a t-statistic of 4.17; d remains insignificant.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh
220 SEDGWICK AND POKORNY
Table 5, together with the estimated models set out in Eq. (6) and (12) suggest that
Warners adopted a retrospective approach to risk, where the financial perfor-
mance of films in period t-1 proved to be a critical factor. This led to a cyclical
pattern in annual film production budgets and risks taken which closely mirrored
the behavior of the U.S. economy.
REFERENCES
Balio, T. (1995) (Ed.), Grand Design: Hollywood as a Modern Business, 1930–39. New York.
Barrios, R. (1995), A Song in the Dark: The Birth of the Musical Film. Oxford.
Bordwell, D., Staiger, J., and Thompson, K. (1985), The Classical Hollywood Cinema: Film Style and
Mode of Production to 1960. London.
De Grazie, V. (1989), ‘‘Mass Culture and Sovereignty: The American Challenge to European
Cinemas, 1920–1960.’’ Journal of Modern History 61, 53–87.
De Vany, A., and Walls, W. D. (1996), ‘‘Bose-Einstein Dynamics and Adaptive Contracting in the
Motion Picture Industry.’’ Economic Journal 106(439), 1493–1514.
Dimsdale, N. H. (1981), ‘‘British Monetary Policy and the Exchange Rate 1920–38.’’ Oxford
Economic Papers 33, 306–349.
Fine, B., and Leopold, E. (1993), The World of Consumption. London.
Finler, J. (1988), The Hollywood Story. London.
Glancy, H. M. (1992), ‘‘MGM Film Grosses, 1924–1948: The Eddie Mannix Ledger.’’ Historical
Journal of Film, Radio and Television 12, 127–144.
Glancy, H. M. (1995), ‘‘Warner Bros. Film Grosses, 1921–1951: The William Schaefer Ledger.’’
Historical Journal of Film, Radio and Television 15, 55–74.
Gomery, D. (1992), Shared Pleasures: A History of Movie Presentation in the United States. London.
Greenwald, W. I. (1950), The Motion Picture Industry: An Economic Study of the History and
Practices of a Business. Ph.D. dissertation, New York University.
Izod, J. (1988), Hollywood and the Box Offıce, 1995–1986. Basingstoke.
Jewell, R. B. (1994), ‘‘RKO Film Grosses, 1929–51: C. J. Trevlin Ledger.’’ Historical Journal of Film,
Radio and Television 14(1), 35–50.
King, B. (1986), ‘‘Stardom as Occupation.’’ In P. Kerr (Ed.), The Hollywood Film Industry. London.
Lumby, S. (1991), Investment Appraisal and Financing Decisions. (4th ed.). London.
Roddick, N. (1983), A New Deal In Entertainment: Warner Bros in the 1930s. London.
Sedgwick, J. (1994), ‘‘Richard B. Jewell’s RKO Film Grosses, 1929–51: The C. J. Trevlin Ledger: A
Comment.’’ Historical Journal of Film, Radio and Television 14(1), 51–59.
Sedgwick, J. (1995), ‘‘The Warners’ Ledger 1921–22 to 1941–42: A Comment.’’ Historical Journal of
Film, Radio and Television. 15(1), 75–82.
Sedgwick, J. (1998), ‘‘Cinema-Going Preferences in Britain in the 1930s.’’ In J. Richards (Ed.), The
Unknown 1930s: An Alternative History of the British Cinema, 1929–39. London.
Thompson, K. (1985), Exporting Entertainment: America in the World Film Market 1907–1934.
London.
U.S. Department of Commerce, Bureau of the Census (1975), Historical Statistics of the US; Colonial
Times to 1970.
EEH0691
@xyserv1/disk3/CLS_jrnl/GRP_eehj/JOB_eehj35-2/DIV_238a01 angh