Distribution conventionality in the movie sector: an econometric analysis of cinema supply

MANAGERIAL AND DECISION ECONOMICS

Manage. Decis. Econ. 30: 517–527 (2009)

Published online 20 April 2009 in Wiley InterScience

(www.interscience.wiley.com) DOI: 10.1002/mde.1469

Distribution Conventionality in the MovieSector: An Econometric Analysis

of Cinema Supply

Alan Collinsa,�, Antonello E. Scorcub and Roberto Zanolac

aDepartment of Economics, University of Portsmouth, Portsmouth, UKbDepartment of Economics, University of Bologna, Bologna, Italy

cDepartment of Public Policy and Public Choice, University of Eastern Piedmont, Turin, Italy

This paper empirically analyzes the impact of several factors on a ‘conventionality index (CI)’

in the specific context of the cinema exhibition sector. To our knowledge, it is the first timethat a standard CI has been constructed for this purpose. Econometric analysis of the

determinants of variation in this index provides decision-makers with an empirical focus for

analyzing distributional aspects of the movie exhibition market, with particular emphasis onproduct differentiation. Specifically, (i) do cinemas based in a city area have a different or

‘specialized’ focus in contrast to cinemas in small towns? or (ii) do multiplexes have a different

or more specialized focus in comparison with cinemas? To this end, cross-sectional

econometric models are estimated to help analyze these effects in three Italian regions for asample of cinemas covering the 2006 season. Copyright r 2009 John Wiley & Sons, Ltd.

1. INTRODUCTION

Many studies have explored the extent of productdifferentiation within arts organizations. Such workhas tended to focus on high cultural pursuits suchas opera, dramatic theater or symphony orchestraconcerts (Pierce, 2000; Dowd et al., 2002; O’Haganand Neligan, 2005; Neligan, 2006). It is typicalfor these activities to require substantial publicfinancial support and/or substantial corporatesponsorship arrangements. Accordingly, manyinvestigations (typically involving the constructionof a ‘conventionality index’ (CI)) have sought tofocus on the extent to which subsidized/sponsoredperformances disproportionately feature ‘safe’ or‘established’ repertoire and eschew ‘new’ or

‘experimental’ work. The hypothesized rationale isthat by such a strategy it is believed more likely thatcontinuity of public funding (and/or majorcorporate sponsorship arrangements) can besecured or renewed. The alternative hypothesis,however, can also be explored. This relates towhether a measure of ‘insulation’ from commercialpressures afforded by significant public fundingand/or corporate sponsorship can actually fostergreater artistic experimentation and radicalism toembrace the ‘shock of the new’.

In cinema exhibition, private sector financingoverwhelmingly dominates, though there are manypublicly supported arts cinemas and in Italy at least,there is also a significant state film distributorworking among the private sector film distributiongiants. Nevertheless, despite the market orientation,expressions of consumer dissatisfaction regardingthe limited extent of offered product differentiationin various towns’ and cities’ cinemas andmultiplexes have arisen (see, for example,

*Correspondence to: Department of Economics, PortsmouthBusiness School, University of Portsmouth, RichmondBuilding, Portland Street, Portsmouth, Hampshire PO13DE, UK. E-mail: [email protected]

Copyright r 2009 John Wiley & Sons, Ltd.

Hubbard, 2000, 2002). It is thus of policy interest toexplore the extent to which such expressions areempirically valid. Decision-making regarding thedisbursement of public or sponsorship funding offilm products and specialist or art house movietheaters could also be informed by such an evidencebase. While there may be much anecdotal chatterand perceptions that movie theaters are showing toomany similar films, the degree to which this is trulythe case in a given geographical area is an openempirical question.

This is the first paper, to our knowledge, toconstruct an orthodox CI in the specific context ofcinema (movie theater) exhibition. That said, inthe specific context of the Boston metropolitanarea, Chisholm et al. (2006) have developed arelated index—a ‘similarity index’—in the contextof film scheduling. This is used to investigate thedeterminants of strategic product differentiation inthe industry as conditioned by contractualconstraints. In general, they find that productdifferentiation is inversely related to market sizeand that more similar film programming isfound within movie theaters under commonownership.

This study uses data from three representativeItalian regions. Cross-sectional econometric modelsare estimated to help analyze variation in this CI. Byso doing a particular mode of empirical enquiry isenabled to help frame analysis of some distributionalaspects of the movie exhibition market. Specifically,(i) do cinemas based in a city area have a different ormore ‘specialized’ focus in contrast to cinemas insmall towns? or (ii) do multiplexes have a differentor more specialized focus in comparison withcinemas? While differences in consumers’ movie-going behavior have been analyzed with respect tomultiplex and non-multiplex movie theaters (Collinset al., 2005), supply-side differences in terms ofmovie distribution conventionality have not beensimilarly explored.

This paper is organized in the following manner.Section 2 explores some of the economics ofproduct differentiation that can inform work onmovie exhibition conventionality and presents aguiding theoretical framework to this study.Background on the movie exhibition sector inItaly is then presented to contextualize the studyapplication. Section 4 outlines the modeling strategyadopted with respect to both the construction of theCI and the CI model structures. The data aredescribed in the following section and the results of

model estimation are presented and discussed inSections 6 and 7. The final section, Section 7, offerssome concluding remarks.

2. PRODUCT DIFFERENTIATION AND

MOVIE CONVENTIONALITY: A BRIEF

RETROSPECT

In a given time period (1 year in our case) a CI forthe jth single-screen cinema (or for the jth screen)may be defined as

CIj ¼Xn

i¼1pI

� �=nj ð1Þ

where pi is the number of movie runs in the sampleunder scrutiny and nj is the total number of moviesperformed on the screen under scrutiny, in thesame time period. For example, if in 2006 the jthsingle-screen cinema has always been open 365 outof 365 days,1 in order to compute the CI, thesedays of opening have to be divided by the totalnumber of movies offered by the cinema. To thisend, we approximate one movie run to one day, asif the number of daily runs were constant each dayand for every movie. Clearly, it may be the casethat the number of daily runs increases duringweekend. Moreover, some movies (Once upon atime in America, for example) are quite long withrespect to the typical film length, and in this case areduction in the number of daily runs could occur.However, such adjustments have been neglected inwhat follows.

More generally, the CI corresponds to theaverage number of runs of a given cultural goodis shown or performed by a given cinema, ortheater, opera house, orchestra and balletcompany in a given time period. It is a metricthat has been widely used to analyze formally therepertoire diversity or degree of productdifferentiation by theaters and orchestras (Pierce,2000; Dowd et al., 2002; O’Hagan and Neligan,2005; Neligan, 2006).

The second stage in such analyses is to identifyand estimate (usually in a single-equationframework) the explanatory variables of a seriesof CIs. The typical set of independent variables fortheaters CIs pertain to, inter alia, the proportion ofsubsidy and subscription income of a given theatervenue or performing company, a political index ofthe government subsidy awarders, some simplemeasure of competition (e.g. number of public/

A. COLLINS ET AL.518

Copyright r 2009 John Wiley & Sons, Ltd. Manage. Decis. Econ. 30: 517–527 (2009)DOI: 10.1002/mde

private theaters in the relevant municipality), arange of venue-specific characteristics, populationsize and a range of socio-economic variables forthe municipality in which the venue lies.

Hitherto this approach has not been used toanalyze product differentiation in commercialtheater (e.g. New York Broadway, London WestEnd) or cinemas/movie theaters of any kind. Withrespect to movie theaters, key differences with thedramatic theater studies concern (i) the degree ofreliance on commercial revenue streams and (ii)that while theater (or symphony orchestra) seniormanagement actually oversee the selection,production and performance of repertoire; thismay not be the case for a given movie theater. Thislatter point leaves room for some differentarguments in the case of a movie theater’s CIfunction and the need to be informed by a corpusof economic theory not largely contingent on thetheme of expected and revealed public sectorpreferences.

While many do exist, most individual movietheaters are not independent firms or entities.Rather, they are outlets in a multiple chain ofmovie theaters owned by an exhibitor firm. Localmanagement at each outlet typically has verylimited discretion over the ‘slate’ of newly releasedfilms that are shown. Decision-making by exhibitorfirms may also be constrained by means of highequity cross-ownership or full forward verticalintegration between distributor and exhibitorfirms. Even where this is not the case, exhibitorfirms still have to buy bundles of films (block-booking) negotiated with very complex contractualarrangements from movie distributor companies. Inaccordance with profit maximizing economictheory, each bundle will typically contain somefilms expected to be high earners and some lowearners (Adams and Yellen, 1976; Dansby andConrad, 1984; Salinger, 1995; Stigler, 1968).

The contracts feature various box office andmovie theater concessions (food, drink, etc.)revenue apportionment formula and possibly‘holdover’ clauses mandating extensions to filmscreening days in the event of particular box officetargets being achieved. Ornstein (1994) andOrbach (2004) provide a detailed analysis of thecontractual frameworks governing movie theateroperations. The latter work also provides usefulinsights to help account for the persistent lack ofprice competition in this sector, particularly giventhat history shows box office pricing was not

always predominantly uniform. He does notattribute this to the efficient market hypothesis,but rather to ‘pressure’ from distributors seekingto minimize a range of demand and supply-siderisks. Shaked and Sutton (1982) providetheoretical support for this view that firms willuse vertical product differentiation to relax pricecompetition.

Regardless of how the power of distributorsand exhibitors ensured non-price competitiondominates market conduct was achieved, itrenders the degree of price discrimination and itsdomains or levels (i.e. film-based, environment-based movie theater) to be a more significantpractical focus of economic enquiry. Lancaster(1990) provides a panoramic survey of the generalliterature and from the individual firm’s perspectiveidentifies three approaches to analyzing productvariety by multi-product firms. The first relates tothe production side and focuses on cost advantagesfrom joint production and economies of scope. Thesecond relates to demand-side concerns inbalancing the increased revenue from greaterproduct variety against the loss of scale economiesin the production of each variant. The third relatesto the deployment of product variety as a form ofstrategic preemption in the product space to achieveentry deterrence objectives.

The development, diffusion and rate of growthof the multiplex phenomenon could potentially beviewed in the light of all three of these reasons, butso could the vertical or quasi-vertical integrationof distributors and exhibitors, increasing marketconcentration and possibly the heavy reliance onparticular ‘proven’ plot formulas, ‘date movie’appeal and star actors/actresses.

Hotelling (1929) first set out the idea of firmscompeting on more than one level—in his caseprice and location. The upshot of the basicproposition was that the equilibrium competitivelocation for two firms operating in a linear marketwith distinct terminal boundaries and uniformlydistributed customers was to be as close as possibletogether without actually colliding. As Lancaster(1990) noted, ‘ybeyond pure geographicallocation, the proposition is concerned with thedegree of differentiation rather than the number ofgoods, answering the question: if there are twofirms in the market, will their best strategy be tochoose products which are very similar or verydifferent?’ (p. 197). So emerged the so-calledPrinciple of Minimum Differentiation.

AN ECONOMETRIC ANALYSIS OF CINEMA SUPPLY 519


Yet in subsequent years his equilibrium resultswere heavily challenged and found not to hold forcases of more than three firms and when variousother assumptions were relaxed or modified. Therevision by D’Aspremont et al. (1979) suggestingquadratic rather than linear transport costssuggested a maximum rather than minimumdifferentiation equilibrium outcome. Bester (1998),however, suggested that imperfect informationabout the vertical (quality) characteristics ofgoods (films) reduces distributors/exhibitorsincentives for horizontal product differentiation(other films/genres) and thus leading to minimumproduct differentiation. In spatial terms, theimplication is that firms (movie theaters) will tendto choose head-on competition by agglomeration ata given location. This means that consumers canincidentally benefit from imperfect informationconcerning product (film) quality.

Typically these studies tend to assume, asHotelling did in his original work, that consumersare uniformly distributed. When this particularassumption was dropped, De Palma et al. (1985)identified renewed theoretical support for theMinimum Differentiation Principle. Hence, non-uniformly distributed consumers in geographicaland/or product characteristics space (i.e. consumerheterogeneity in space and tastes) providesome theoretical basis to justify expectations of ahigh CI score in the context of movie theaterexhibition.

Chisholm and Norman (2004) investigatedthe degree to which the Minimum Differentia-tion Principle extends to location choices bymulti-product firms of different sizes supplyingdifferentiated goods to consumers with hetero-geneous tastes. They found that multi-productfirms disperse their products if consumer hetero-geneity is low or distance between markets is high.

3. THE MOVIE EXHIBITION SECTOR IN

ITALY

An important and motivating focus for this studyis the current supply structure of cinemas andscreens in Italy. Up to a few years ago, the openingof a new cinema was subject to stringent publicregulations. In fact, the regulation resulted in asevere rationing of openings and in the progressiveaging of supply: the vast majority of cinemas weresingle screen, often with historical locations in

downtown areas. In recent years the sector hasbeen liberalized. The changes resulted in theopening of many multi-screen cinemas (from 5 to7 screens) or multiplexes (eight or more screens) atthe expenses of smaller single-screen structures.2 Infact, in a few years, several of the aged downtowncinemas underwent major refurbishment and havebeen transformed into multi-screen cinemas. Inother cases newly built megaplexes have risen inthe urban suburbs. The increasing proportion ofnew-build cinemas can be shown to be connectedwith increased quality in terms of the viewingexperience (particularly in terms of comfort ofthe seats, sight lines, etc.) as well the introductionof up-to-date marketing and organizationaltechniques. For example, the number of cinemasclosed in July and August (a common practice forsmall single family-owned and run single-screencinemas) has sharply reduced. These developmentsare summarized in Table 1, which shows amoderate decline in the total number of cinemas(�6.3%) together with a significant growth of thetotal number of screens (110.2%).

On the basis of such data it is possible toassert that the average size of an Italian cinemahas slightly increased from 2.2 to 2.65 screensper cinema during the period 2004–2007. The

increase, however, comes in only in part from theincreased size of individual cinema complexes (assummarized in Table 1 the average number ofscreens has risen by 10%, from 4.35 to 4.80). Ofprincipal importance has been the increase in themarket share of the larger types of cinemas (an8.8% increase, from 465 to 506). The overall effecthas been a 20% increase in the number of screens

Table 1. The Structure of the Italian MovieDistribution Industry, 2004–2007

Year Singlescreen

Multiplex Total

No. of cinemas 2004 778 465 12432005 779 496 12752006 713 497 12102007 659 506 1165

No of screens 2004 778 2024 28022005 779 2237 30162006 713 2349 30622007 659 2428 3087

Average screen per cinema 2004 1 4.35 2.252005 1 4.51 2.372006 1 4.73 2.522007 1 4.80 2.65

Source: Cinetel.



in the larger cinemas, which overshadows the15.3% drop in the number of single-screencinemas.

Some further evidence is provided in Table 2,showing that admissions to single-screen cinemasin 2004 was slightly above 20% but that 3 yearslater the figure dropped to 13.6%.

As a consequence of a moderate increase in thenumber of admissions (from 97.2 to 103.6 millionsof tickets sold, Table 2, 15.8%) and of the strongincrease in the number of screens (from 2802 to3087, Table 1, 110.2%), the average attendanceper screen dropped by just under 4% over the 4years considered. Further interesting trends can bediscerned if the data are disaggregated by cinemasize. The average admission for single-screencinemas has collapsed dramatically from 26 591tickets sold in 2004, to 21 387 in 2007, with a19.6% decrease (�19.6%), whereas the drop formulti-screen cinemas, albeit significant, remainslimited to �3.3%. Therefore, the redistribution ofcinemagoers seems to be in the favor ofmultiplexes.

Perhaps unsurprisingly, the multi-screencinemas are located in provincial capitals morefrequently (59% of the cases) than single-screencinemas (45% of the cases); for multi-screencinemas the size of the market (measured interms of the average number of the inhabitantsof the municipalities and of the neighborhoods) is

larger than for single-screen cinemas (Table 3).The average GDP per capita in municipalities withmulti-screen cinemas, however, is only slightlyhigher than in municipalities with single-screencinemas. Overall, these differences in the marketstructure correspond to an average concentrationindex that is higher for multi-screen cinemas thanfor single-screen cinemas.

These developments have evolved at a differentpace within the various regions of Italy. The datashow that in the more urbanized Northern area,the extent of the decline of single-screen cinemashas been so thorough that they have been almostrendered extinct.3 While in the South, this processof transformation is lagging well behind that of theNorth. In the Central regions (with the notableexception of Rome) an ‘intermediate’ position4 canbe discerned.

As for the type or genre of movies, Italianpreferences seem similar to those of the rest of theEurope. As summarized in Table 4, US moviesdominate the box office and most of thedistributors are subsidiaries of the American-based majors. The market share for national filmsfluctuates around 22–25% of the total numberof admissions (with some very temporary increasesin the case of a few particularly successful Italianhits), compared with a 55–60% share for the USmovies.

Regardless of movie origins, cinemaconsumption in Italy has been and still remainshigh. It still remains, by and large, the mostpopular activity among the range of performingarts. Figure 1 presents the pattern of participationover recent years. It is immediately apparent thatwithin a broadly stable picture for the otherperforming arts, there is some evidence ofincreasing cinema going as compared with othercultural activities.5

From the perspective of this study’s focus ofenquiry, the identification of different businessstrategies and outcomes associated with largeand small cinema establishments is of centralimportance. This enables some assessment of the

Table 2. Movie Attendance in Italy, 2004–2007

Singlescreen (%)

Multiplex(%)

Total

Attendance 2004 21.13 78.87 97 927 1922005 17.57 82.57 90 641 6032006 15.20 84.80 92 275 4642007 13.60 86.40 103 632 460

Average attendanceper screen

2004 26 596 38 160 34 949

2005 20 444 33 457 30 0542006 19 672 33 312 30 1362007 21 387 36 877 33 571

Source: Cinetel.

Table 3. Single-Screen and Multi-Screen Markets and Concentration

Provincialcapital (%)

GDP 2004(avg. Euro per capita)

Populationmunicipality(av. number)

Populationneighborhood(av. number)

Full CI

Multiplex 158 0.59 22755.87 922221.9 468306.9 10.48Single screen 124 0.45 21584.65 372281.5 212821.8 9.15



extent to which cinemas of different sizes exhibitdifferent conventionality indices. This can bequantified by computing CIs per screen and percinema (thereby aggregating the several screens ofa multiplex into a single ‘average’ screen).

4. THE MODEL

The purpose of the paper is to empiricallyinvestigate the impact of several factors on theCI (Pierce, 2000; O’Hagan and Neligan, 2005;Neligan, 2006) as described in Section 2. In ouranalysis we distinguish between CIs related toscreens (single CI sample) and to cinemas (full CIsample). In the first case, the multiplex’s screensare regarded as single cinemas, while in the secondcase each of the multiplex’s screens are consideredas if they were a single cinema.

In the case of the kth multiplex the corres-ponding ‘average’ CI can be computed as follows:

CIk ¼XNj¼1

Xni¼1

pij

!, XNj¼1

nj

!ð2Þ

where pij indicates the number of runs of the movieI on the screen j of the kth cinema, nj indicates the

total number of movies in the period of time underscrutiny for screen j. Hence, the numerator of theCI indicates the total number of movie runs for allthe N screens of the cinema k, whereas in thedenominator there is the total number of moviesthat were presented by the cinema. Obviously, amultiplex or a megaplex could easily developsome product differentiation strategies: popular,expected hits, that are expected to have longerruns, are set on larger capacity screens.Alternatively, ‘highbrow’ movies, that attract fewmoviegoers, are set in smaller capacity screens (butif one such movie turns out to be an unexpectedsuccess, it is indeed likely to be moved on a largerscreen). In this respect, the determinants of the‘average’ screen’s strategy might differ from thatof the single screen.

In explaining variations in a CI, demographicand economic variables are used to identifydemand-side indicators that are expected to reflectlocal characteristics for specific productions (Pierce,2000). The size of the market is measured bypopulation, pop. This is expected to influenceconventionality positively to the extent that longerfilm runs are necessary to satisfy higher and morestable levels of demand. To this end, it is useful toadopt a spatial weight matrix approach (Werck etal., 2008). This involves multiplying the number ofinhabitants in municipality j in Province k with aspatial weight matrix relevant to cinema i, whoseelements are equal to one if municipality is part ofthe relevant market for cinema i and 0 otherwise.Up to three different specifications were considered.In the first it is assumed that the relevant market islimited to the municipality in which the cinema islocated, pop1; in the second specification therelevant market is extended to enclose contiguousmunicipalities, pop2; finally, in the thirdspecification it is assumed that the relevant marketcoincides with the whole Province, pop3. Thesecond specification is ultimately determined to bethe most appropriate and is the one considered inthe empirical specification. It might be reasonablyexpected that a cinema based on a metropolitanarea would have scope to host more specializedrepertoire than a cinema of equal size in a smalltown, due to audience segmentation and marketniches, over which they enjoy some market power(Neligan, 2006). Further, the influence of word ofmouth mechanisms might require longer timeperiods among larger populations, and thereforecontribute to this result.

Table 4. Revenue Shares on the Italian MovieMarket, Italian and the US Movies, 2002–2007

Italya USA

2002 22.2 60.12003 21.8 64.52004 20.3 61.92005 24.7 53.82006 24.8 61.92007 31.7 55.4

aComprises Italian co-productions.

0

10

20

30

40

50

60

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2005

2006

2007

theatre cinema

museums and exhibitions classical music concerts

other music concerts sport events

Figure 1. Participation in performing arts events in Italy(1993–2007). Source: Istat. Note: data for 2004 are not

available.



Additionally, the age structure represents apotentially important feature of movie turnover.The percentage of the population aged up to 14,age14 is a market segment whose characteristics areexpected to differ from the more mature moviegoers(aged 15–64). Younger people (and those with less‘established’ movie-going behavior) may be stronglyinfluenced by fads and be subject to different timeconstraints, etc. As a consequence, the overall effecton movie turnover is complex and the expected signcannot readily be predicted in advance.

Following the empirical literature on conven-tionality (Pierce, 2000; O’Hagan and Neligan, 2005;Neligan, 2006), per capita income, gdp, is introducedas a potential explanatory variable. Since cinemaconsumption is expected to increase with income(Sisto and Zanola, 2008), it is likely to register alower movie turnover.

Finally, we include a dummy for multiplexes,multiplex, to catch structural aspects of themarket. Since multiplexes screen blockbusters,whose demand may be more durable than lesscommercial movies, they are expected to show alower level of movie turnover.

In summary the model being tested in this paperis as follows:

CIi ¼ fðpopi; age14i; gdpi;multiplexiÞ ð3Þ

5. DATA

The analysis covers a cross-section of three Italianregions in 2006, which have been considered byItalian cinema industry analysts as representativeof the Italian market.7 These are, namely,Piemonte, Lazio and Campania. CIs areconstructed using data provided by Cinetel, acompany specializing in the collection andcompilation of motion picture industry data. Percapita income data are drawn from the Minister ofEconomy’s Dichiarazione dei redditi presentatedalle persone fisiche nel 2005. Data on per capitapopulation income derives from the Istat’s (2006)Annual Statistics. Structural data have beenderived from Cinetel. Table 5 lists the basicdescriptive statistics for the dependent variableCI and the key independent variables.

6. RESULTS

As indicated in the previous section, the dependentvariable is the CI, calculated as the ratio between

programming days and the total number of moviesscreened. However, two distinctive definitions areconsidered: (i) single CIs, calculated for eachscreen (and in this case multiplexes areconsidered as several different screens) or (ii) fullCIs, in which case multiplexes are considered as aunique weighted screen.

Before proceeding, we test for heterosked-asticity of residuals. Both the White’s test andthe Breusch–Pagan test have been performed. Thenull of homoskedasticity is rejected at the 1% levelso that heteroskedasticity seems to be a problem inour sample. To this aim, the standard errorsare adjusted to provide heteroskedastic consistentstandard errors using a heteroskedasticity consis-tent regression matrix.

Normality of residuals is required for the t andF statistics to be valid. Kernel density estimationindicated problems with the normality of residuals.Additionally, the Shapiro–Wilks test for normalityrejects the null hypothesis of normality at the 1%level. Having concluded that conventionality is notnormally distributed, we select an appropriatetransformation of the variable to convert it into anormally distributed one. Results indicate that thelog transformation is the most appropriate.

Regression results using the single CI sampleare summarized in Table 6. Standard errorsand variance–covariance matrices of the log-specification of Equation (3) have been computedby using the Huber/White/sandwich heterosked-asticity-robust procedure.

In testing for the correct model specification,a link test was undertaken. This test computes anew regression model using as predictor variablesthe risk score from the original model plus itssquare (quadratic term). If the additionalquadratic term is significant, there is evidenceof model misspecification. In our case, theestimated model does not display a significantquadratic term, providing no evidence of incorrectmodel specification when per capita income isexcluded.

Table 5. Summary Statistics (Single CI Sample)

Variable Obs. Mean SD Min Max

Single CI 870 10.408 4.273 1.583 40.083Pop 870 1 296 396 1 268 905 11 447 3 113 130p14 870 0.144 0.020 0.109 0.183gdp 870 22 480.63 3102.12 15 603 27 463Multiplex 870 0.856 0.351 0 1



Contiguous population displays a positive signat the 1% significance level. Consistent with priorexpectations, larger contiguous areas (pop) featurehigher conformity. Regarding the populationstructure (p14), the coefficient turns out todisplay a negative sign significant at the 5%level. A potential explanation may be linkedto the peculiar characteristics of this market,where teenagers are perhaps more prone tofaddy consumption and, consequently, a highermovie turnover. Per capita income is positiveand statistically significant at the 1% level. A 1%increase in income level will increase conven-tionality by 0.34%, suggesting the existence of amoderate-income effect on the size of the market.

Focussing on the structural variables, themultiplex coefficient (multiplex) is statisticallysignificant at the 1% level and shows a positiverelationship with conventionality. This result shedssome light on the scheduling characteristics ofmultiplexes, which tend to display a lower turnoverof movies due to higher audience numbers.8

Given that data on single screens are used in thisstudy, it is not possible to directly capture thedistinctive strategies of multiplexes in aggregate.Single CIs treat multiplexes as several differentscreens, where the programming strategy for a givenscreen is totally independent to those of the otherscreens (although part of the same multiplex).

With this shortcoming in mind, the previousanalysis has been extended to the full CI sample,whose basic descriptives are summarized inTable 7.

Table 8 lists the regression results of theHuber/White/sandwich heteroskedasticity-robustprocedure for the full CI sample. Due to multi-collinearity with population, the dummy variablefor provincial capital was dropped. Again, it wasfound that a log transformation helped to make

conventionality more normally distributed andno evidence of misspecification emerged from thelink test.

In the case of the full CI sample, theHuber/White/sandwich heteroskedasticity-robustregression seems to suggest the same picture thatemerged from the single CIs sample: coefficientsdisplay the same signs as in Table 6, and arestatistically significant, with only the exception ofthe multiplex coefficient. In this case, conven-tionality appears to be similar within the samplefor both single screens and multiplexes. On thecontrary, conditioning in terms of being(or not) a multi-screen is important in explainingthe differences in single CIs. This result is hardlysurprising: as full CIs average out our measure ofconcentration, the differences between single-screen and multi-screen cinemas are reduced andbeing (or not) a multi-screen cinema becomes lessimportant in terms of concentration strategy.

7. CONCLUDING REMARKS

This paper has sought to analyze the developmentof movie programming in Italy. The cinema sectorin Italy is undergoing a major transformation.Small single-screen cinemas are of diminishingimportance but remain, for the time being, asignificant element in the cinema exhibition sector.

Table 6. Regression Results (Single CI Sample)

Single CI Coefficient Robust SE t Beta

pop 0.079 0.010 7.97 0.343p14 �1.679 0.705 �2.38 �0.090gdp 0.350 0.112 3.13 0.129Multiplex 0.134 0.045 3.02 0.125cons �2.149 1.080 �1.99

Number obs. 870F(4, 865) 58.60R2 0.240Link test �0.6063868Mean VIF 1.70

Table 7. Summary Statistics (Full CI Sample)

Variable Obs. Mean SD Min Max

Full CI 282 9.896 5.362 1.583 40.083Pop 282 1 036 370 1 214 768 11 447 3 113 130p14 282 0.144 0.022 0.109 0.183Gdp 282 22 240.87 2945.753 15 603 27 463Multiplex 282 0.560 0.497 0 1

Table 8. Regression Results (Full CI Sample)

Full CI Coefficient Robust SE t Beta

Pop 0.094 0.028 3.43 0.328p14 �2.472 1.583 �1.56 �0.116Gdp 0.717 0.291 2.46 0.202Multiplex 0.061 0.052 1.16 0.064Cons �5.878 2.790 �2.11

Number obs. 282F(4, 277) 27.28R2 0.270Link test 0.4368184Mean VIF 1.89



Moreover, cinemas tend to be located in a numberof major cities, or close to them. On the basis ofsuch cinema characteristics, more homogeneousprogramming might be expected to be anobservable feature of the market. Thus, thecentral research question relevant to this studyand guiding the empirical work has been toidentify whether developments in the Italiancinema sector have led to an increasinglyhomogeneous or heterogeneous supply of movies?

In order to answer to this question, a CIapproach has been adopted. The CI is a metricthat has been widely used to formally analyze therepertoire diversity or degree of productdifferentiation by theaters and orchestras. Yet tothe best of our knowledge, it is the first time thatthis approach has been applied to a principallyprivate sector market, such as the case of cinema.

The analysis was conducted on a sample ofthree Italian regions in 2006 (Piemonte, Lazio andCampania), considered by industry experts asrepresentative of the full Italian market. The

main conclusion that can be drawn from thisstudy is that conventionality dropped slightlywhen the index was computed for multiplexesconsidered as a unique-weighted screen relative tothe single-screen case.

However, when the main determinants of theCIs were analyzed, the effect of the multiplexeswas actually to increase the conventionality ofthe supply, taking into account the effect of otherexplanatory variables. Additionally, conven-tionality is also increased by both income andmarket dimensions: the greater the income and themore extensive are the spatial dimensions of themarket, the higher is the level of conventionality.

APPENDIX A

The structure of the movie distribution industry inthe Northern, Central and Southern Italianregions is given in Table A1.

Table A1. The Structure of the Movie Distribution Industry in the Northern, Central and Southern ItalianRegions, 2005–2006

Cinemas

Northern regions Central regions Southern regions2005 2006 2005 2006 2005 2006

Total 455 437 480 454 319 318Of which single screen 59.34 56.29 56.88 55.51 66.77 64.78Of which 2–4 screens cinema 25.71 26.77 28.96 29.30 26.02 26.73Of which 5–7 screens cinema 5.71 6.64 6.25 6.83 2.51 2.83Of which megaplexes 9.23 10.30 7.92 8.37 4.70 5.66

Screens


Total 1157 1198 1213 1195 613 657Of which single screen 23.34 20.53 22.51 21.09 34.75 31.35Of which 2–4 screens cinema 25.41 25.38 30.67 30.29 33.28 32.27Of which 5–7 screens cinema 13.14 14.11 14.34 15.56 7.18 8.07Of which megaplexes 38.12 39.98 32.48 33.05 24.80 28.31

Admissions (� 1000)


Total 37 401 37 956 36 615 37 225 16 654 17 070Of which single screen 14.66 12.34 14.37 13.18 30.80 25.42Of which 2–4 screens cinema 21.71 20.78 26.34 24.94 22.72 22.72Of which 5–7 screens cinema 14.62 15.38 16.05 16.84 7.57 9.91Of which megaplexes 49.00 51.51 43.24 45.04 38.92 41.95



NOTES

1. In Italy many single-screen cinemas are still family-owned and run. In these cases the cinema is usuallyclosed in certain periods (typically in summer—thelow season), the numerator of the CI is reducedaccordingly.

2. This trend is portrayed in the movie Cinema Paradisoby G. Tornatore that won the 1989 Academy Award(Oscar) for Best Film in a Foreign Language.

3. Many single-screen cinemas have been transformedinto multiplexes.

4. Tables are not shown for the sake of brevity, but areavailable upon request from the authors.

5. The latter figure refers to the share of inhabitants (aged6 or more) that participate in the stated cultural activityand it does not distinguish between strong and weakconsumers. The data on admissions refers to all ticketssold and weights households on the basis of theirintensity of consumption, regardless of the age.Therefore, these two figures offer different perspectivesfor the analysis of the movie market.

6. Therefore, in the case of a cinema with 2 screens,both open for 350 days per year, where on one screenwas shown 27 movies and 35 shown on the other, theCI will be (350�2)/(27+35).

7. As suggested before, the adjustments in the differentregions of Italy evolved at a different pace. Oursample is representative of all the situations:Piemonte is in the North of Italy, Lazio in theCenter and Campania in the South.

8. We also check for the existence of some regionalidiosyncratic effect by introducing regional dummyvariables. However, they are not statistically significantconfirming the substantial homogeneity of the sample.

REFERENCES

Adams WJ, Yellen JL. 1976. Commodity bundling andthe burden of monopoly. Quarterly Journal ofEconomics 90: 475–498.

Bester H. 1998. Quality uncertainty mitigates productdifferentiation. RAND Journal of Economics 29:828–844.

Chisholm DC, McMillan MS, Norman G. 2006.Product differentiation and film programmingchoice: do first run movie theaters show the samefilms? NBER WP 12646.

Chisholm DC, Norman G. 2004. Heterogeneouspreferences and location choice with multi-productfirms. Regional Science and Urban Economics 34:321–339.

Collins A, Hand C, Ryder A. 2005. The lure of themultiplex? The interplay of time, distance and cinemaattendance. Environment and Planning A 37: 483–501.

Dansby RE, Conrad C. 1984. Commodity bundling.American Economic Review (Paper and Proceedings)74: 377–381.

D’Aspremont C, Gabszewicz JJ, Thisse J-F. 1979. OnHotelling’s stability in competition. Econometrica 17:1145–1151.

De Palma A, Ginsburgh V, Papageorgiou Y, ThisseJ.-F. 1985. The principle of minimum differentiationholds under sufficient heterogeneity. Econometrica 53:767–781.

Dowd TJ, Liddle K, Lupo K, Borden A. 2002.Organizing the musical canon: the repertoires ofmajor U.S. symphony orchestras, 1842 to 1969.Poetics 30: 35–61.

Hotelling H. 1929. Stability in competition. EconomicJournal 39: 41–57.

Table A1. Continued.Admissions per cinema

Northern regions Central regions Southern regions

2005 2006 2005 2006 2005 2006

Total 82.20 86.86 76.28 81.99 52.21 53.68Of which single screen 20.31 19.03 19.27 19.46 24.08 21.06Of which 2–4 screens cinema 69.41 67.41 69.40 69.80 45.58 45.64Of which 5–7 screens cinema 210.31 201.24 195.83 202.23 157.63 187.89Of which megaplexes 436.38 434.47 416.66 441.24 432.07 397.83

Admissions per screen

Northern regions Central regions Southern regions

2005 2006 2005 2006 2005 2006

Total 32.33 31.68 30.19 31.15 27.17 25.98Of which single screen 20.31 19.03 19.27 19.46 24.08 21.06Of which 2–4 screens cinema 27.62 25.94 25.93 25.65 18.54 18.30Of which 5–7 screens cinema 35.97 34.53 33.76 33.70 28.66 31.91

Of which megaplexes 41.56 40.82 40.19 42.45 42.64 38.50



Hubbard P. 2000. Save our screens? Town and CountryPlanning 69: 299–301.

Hubbard P. 2002. Screen-shifting: consumption, ‘Risklessrisks’ and the changing geographies of cinema.Environment and Planning A 34: 1239–1258.

Lancaster K. 1990. The economics of productdifferentiation: a survey. Marketing Science 9: 189–206.

Neligan A. 2006. Public funding and repertoireconventionality in the German public theatre sector:an econometric analysis. Applied Economics 38:1111–1121.

O’Hagan J, Neligan A. 2005. State subsidies andrepertoire conventionality in the non-profit Englishtheatre sector: an econometric analysis. Journal ofCultural Economics 29: 35–57.

Orbach BY. 2004. Antitrust and pricing in the motionpicture industry. American Law and EconomicsAssociation Annual Meetings, Working Paper No. 40;1–45. http://law.bepress.com/alea/14th/art40.

Ornstein SI. 1994. Motion picture distribution, filmsplitting and antitrust policy. Hastings Communicationsand Entertainment Law Journal 17: 415–444.

Pierce JL. 2000. Programmatic risk-taking by Americanopera companies. Journal of Cultural Economics 24: 45–63.

Salinger MA. 1995. A graphical analysis of bundling.Journal of Business 68: 85–98.

Shaked A, Sutton J. 1982. Relaxing price competitionthrough product differentiation. Review of EconomicStudies 49: 3–13.

Sisto A, Zanola R. 2009. Cinema Attendance in Europe,Applied Economics Letters (forthcoming).

Stigler GJ. 1968. A note on block booking. In: TheOrganization of Industry, Stigler GJ. (ed.). R. D.Irwin: Homewood, IL, 165–170.

Werck K, Stultjes M, Grinwis P, Heyndels B. 2008.Budgetary constraints and programmatic choices byFlemish subsidized theatres. Applied Economics40(18): 2369–2739.



Distribution conventionality in the movie sector: an econometric analysis of cinema supply

Documents

Transcript of Distribution conventionality in the movie sector: an econometric analysis of cinema supply