The Determinants of Motion Picture Box Office Performance: Evidence from Movies Produced

in Italy
Author(s): M. BAGELLA and L. BECCHETTI
Source: Journal of Cultural Economics, Vol. 23, No. 4 (November 1999), pp. 237-256
Published by: Springer
Stable URL: http://www.jstor.org/stable/41810700 .
Accessed: 24/06/2014 23:29

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Journal of Cultural Economics.

http://www.jstor.org

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
Wá Journal 23: 237-256,1999.
Economics
ofCultural 237
© 1999Kluwer Publishers.
Academic intheNetherlands.
Printed

The Determinants of Motion Picture

Box Office Performance: Evidence

from Movies Produced in Italy*

M. BAGELLA and L. BECCHETTI

Università , Roma
TorVergata , FacoltàdiEconomia diEconomia
, Dipartimento e Istituzioni,
snc,00133Roma
ViadiTorVergata , Italy

Abstract Thepaper anempirical

provides ofboxoffice
analysis formovies
performance produced
inItaly
between 1985and1996.Descriptive documents
evidence inthetotal
a decrease number of
films anda
produced sharp in
reduction daily and
revenues perscreen
dailyadmissions
during the
sampleperiod.Intheeconometricanalysisvarious
alternative ontheimpact
hypotheses oftheex
antepopularity andcastofactors
ofdirectors onboxofficeperformances infavour
arerejected of
a quadratic with
specification between
externalities
positive thetwofactors.
Theeconometric cross-
sectional alsodocuments
evidence that thenetimpactofsubsidies is irrelevant
ontotaladmissions
andthatthesignificantly
lowerperformance of films
subsidised dependson the
lowerex antepop-
oftheir
ularity castanddirectors.
Results ontheimpactofspecialisation andproduction
genres
housesshowthatvery fewofthem(i.e.,thecomic genre
specialisation andtheFilmauroproduction
house)havea significant impact
marginal onboxoffice after
performances director
controlling and
actors
effects.

Keywords:
boxoffice movie
effect,
superstar
performance, economics

1. Introduction

An empiricalanalysison thedeterminants of filmattendancemayprovideuseful

information on severalinterestingissues in the literatureof movie economics.
First,it may shed lighton specialisationgenres(i.e. western,comedy,cartoon)
ofdomestically producedmoviesinvestigating whichone bestmatchesconsumer
tastes.Second,itmayshowhowmuchtheex antepopularity1 of castand director
matters in determining thesuccessof a moviethrough an evaluationof theirmar-
ginalcontribution to filmrevenueswhichin turnmayalso be usefulto producers
whenestablishing actorsanddirector rewards.Third,itmaymeasuretheneteffect
of production companieson commercialperformance.
and distribution Fourth,it
may evaluate whether or not subsidised films(in the specificcase of Italy) are
profit-makers providing new elements in thedebate between advocates and oppon-
entsof movie subsidies.This last issue is of particularrelevancegiventhe long
controversial debateon thepublic/private featuresofmoviesand,morein general,
ofall performing arts(Blaug-King,1976;Baumol-Bowen,1976; Scitovsky, 1976).

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
238 M.BAGELLA
ANDL.BECCHETTI

Whensummingup themostrelevantarguments in thisdebate,it is possibleto

the
justify government's patronage performing withat leastfiverationales.
of arts
First,performing artsbroadentherangeofculturaloptionsavailableto current and
future generations. These optionsimprove individuals' quality of lifewith positive
effectson theirutilityfunctions and,indirectly, also on theirproductivity. The ar-
gument of future is
generations particularly important as, even ifcurrent taxpayers
may decide notto contribute to theirown culturaland spiritualedification, their
intergenerational altruism should induce them to transmit a richer culturalheritage
as a bequest to the newcomers(Baumol-Bowen,1966). A second paternalistic
rationaleis thatfastindustrialgrowthmay have alteredthe preferences of the
averageconsumerwhichis "acquiringleisurefasterthanthepreparation forusing
it" (Scitovsky,1972). Statesupportto performing artsmaythenbe interpreted in
thisperspective as theactionof a benevolentplannerwhichhelpsagentsto enjoy
and enrichtheirlives.
The third,moredisputableargument statesthatgovernment subsidiesmustbe
providedto developand supportculturalidentity through thecreationof master-
pieces whichincreasetheinfluenceandprestigeofdomesticculture.To acceptthis
argument we mustimplicitly assumethatnationalprestige- a somewhatabstract
good - indirectly entersdomesticconsumers'andproducers'preferences or,more
likely, that it positively affects of
preferences foreign consumers inducingthem
to consumegreaterquantitiesof domesticproducts.A fourth rationaleis thatthe
development of artisticheritagegeneratesimmediatepositiveexternalities forall
thoseeconomicactivitiesdevelopingclose to areas whereartis performed. The
lastargument, theso called 4'Baumöl's disease",statesthattheindustrial revolution
induceda dramaticincreasein artproduction costsrelativeto thoseofstandardised
goods which are the of
output Tayloristic industrial processes.Government pat-
ronageis thenurgedin thiscase to shareincreasingrelativeinputcostsin theart
industry.
Can thesefivearguments be equally maintainedto thepurposeof justifying
subsidiesto theparticular and narrower setof artproductsrepresented bymovies?
The answeris yes,providedthatwe referto thosemoviesthatcan be considered
a formof art.Witha strongsimplification theinternational movieindustry may
be seen as offering heterogeneous qualityproductsfallingin a two dimensional
productrange,in whichmovies are positionedaccordingto theirartisticcontent
and to theintensity of special effects.Two oppositeextremesofthisrangemaybe
consideredtheso-called"filmd'autore"(witha low capitalintensity and a high
artisticmerit)and the "special-effects" film(witha low artisticmeritand a high
capitalintensity). The "filmd'autore"is in facta relatively morelabourintensive
product in which the idea of the directoris the crucial productive factor. The special
effectfilmis a capital-intensive product where a large contribution to productivity
is givenbyphysicalinfrastructures and technicalmachinery, and wheretheroleof
theplot(and thenofthedirector'sskills)tendsto be scarcelyrelevant. Manyfilms
are obviouslynot clusteredaroundthesetwo extremesas thereare filmswhich

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
THEDETERMINANTS
OFMOTION BOXOFFICE
PICTURE PERFORMANCE 239

arenotspecialeffectintensiveand have a low artisticcontent,and filmsin which

directorscombinehighcapitalintensity withrelevantartisticcontent.
In thelightof thisscenarioit seemsclearthatthefivearguments whichjustify
artsubsidisation in generalmaybe appliedonlyto thoseproductswithhighartistic
merits.In particularit is evidentthatthepositivecorrelationbetweentechnical
progressand marginalproductivity of labourinputsexistsmuchmorein filmsap-
the
proaching "special effects" rather thanthe"filmd'autore"extreme.This leads
totherejectionoftheBaumol's diseaseargument for"special-effects"filmswitha
low artisticcontent.
The presentpapertriesto shedsomelightoversome oftheissuesraisedabove
(theeffectof specialisationgenres,productionand distribution companies,cast
and directorex antepopularityand statesubsidieson movie theatricalsuccess)
witha descriptive and econometric analysison box officeperformance of movies
producedin Italybetween1985 and 1996.2
Thepaperalso extendstheexistingliterature bypresenting empiricalresultsre-
gardinga yetunexplored Europeanmarket3 andbyadoptinga newmethodological
approachwhichaddressessome of theproblemsunsolvedin previousempirical
paperssuch as the nonnormality of theresiduals.The focus of thepaperis the
relationship between the popularity of humaninputsand box officeperformance,
whosepresumednonlinearity and superadditivity has been underscrutiny in re-
centtheoretical and empiricalliterature on the"superstar'seffect"(Rosen, 1981;
Chung-Cox,1994; Hamlen,1991, 1994).
The descriptive analysispresentedin thepaperdocumentsa reductionin the
numberofmoviesfinancedin therelevantsampleperiod,and a sharpreduction in
per screendaily admissions and a
revenues,paralleledby positive trend in absolute
admissions.
The econometric analysispresentsnon-nestedtestson different a priorispe-
cificationsevaluatingwhethertheex antepopularity of directorand cast of actors
has linear,nonlinearand interaction effects
on box officeperformance. Diagnostic
resultsseemto showthatthequadraticspecification withinteraction effectsis the
mostcorrectspecification to represent theimpactofqualitylevelsofhumaninputs
(directorand cast of actors)into outputmeasuressuch as the averagenumber
of admissions(Bagella, 1997). As remembered by Hamlen(1991, 1994) who is
confronted witha similarproblem,thisresultin supportof thepresenceof super-
stardom effects mustbe consideredwithcare giventhatitis extremely sensitiveto
thescale adoptedforourqualityfilter.
Econometricresultsalso documentthatveryfew additionalinputs(i.e. the
comicspecialisationgenreand theFilmauroproduction house) have a significant
positiveimpact on low, medium and highpercentiles thedistribution
of of total
admissions.A resultof particularinterestalso shows thatsubsidisedmoviesdo
notexhibitsignificantly lowerperformances in termsof totaladmissions,netof
otherfactors,whiletheyclearlyunderperform on averagevis-à-visnonsubsidised
moviesbecauseofthesignificantly lowerex antepopularity oftheirhiredcast.

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
240 M.BAGELLA
ANDL.BECCHETTI

2. Database and Empirical Results

The descriptive and empiricalanalysiscontainedin thispaperis based on theau-

thors'databasewhichgathersinformation on all moviesproducedin Italyfrom
1985 to 1996. Foreach moviewe considerproducingor co-producing companies,
totaladmissions,distribution companies,box officerevenuesand programming
days,ex antepopularity of actorsand directorsand availability of statesubsidies.
Thisinformation is retrievedfromthestatistical Appendixpublishedin theweekly
issues of a journal called Gazzettadel Cinema. The criterionfollowedby the
journal (whichis the officialsourceof ANICA, theNationalAssociationof the
ItalianMovie Industry, and producesdetailedstatisticalinformation on domest-
in
ically producedmovies) determining the cast of actorsis that of includinga
different numberof themforeach movie by includingall main charactersand
excludingactorsplayingcameo roles.
Withinourdatasetwe considerthetotalnumberof admissionsas a dependent
variableto measurebox officeperformance.4
In orderto measuretheseparateeffectof actorsand directors on filmviewing
we construct an indexof actors' cast and directorpopularity before thefilmis on
of
thescreen.The indexis the average(in a sort "Delphi" or "Panel of experts"
method)of an independent judgementof threeinfluential moviecritics5writing
forItaliannewspapers(//Tempo , Il Messagero,Paese Sera).6
A detaileddescription of thedatabase,inclusiveoftheaveragescoreattributed
to cast and directorforeach movie,is providedin an Appendixavailablefromthe
authorsuponrequest.
The descriptiveempiricalanalysisgives a firstmeasureof whatcan be con-
sideredthemaintrendsin moviebox officeperformance afterthecrisisofthelate
seventies.The crisisgeneratesa reductionin thetotalnumberofdomestically pro-
ducedmoviesperyear(froma peak of 107 in 1988 to 63 in 1996) and a concurrent
increasein the averagestay of movies on the screen(almostfourtimeshigher
in 1996 thanin 1985) (Table I).7 Per screendaily admissionsfordomestically
producedmoviesdropby almosta halffrom1985 to 1996 (234 to 154).
The mix of genres("genericinematografici ' such as
comedy,western,etc.),
revealsthatItaliancinematraditionally specialisesin comedies.The roomleftfor
othergenres,whicharetraditionally strongin theU.S. movieindustry (adventure,
in
western,horrorand thrillerfilms), Italy is less thantenper cent (Figure 1).
This relativespecialisationof domesticallyproducedmoviesis paralleledby a
generally betterperformance ofcomediesoverothergenres.In fact,eventhoughall
specialisationgenressuffer a substantial dropduringthesampleperiod,comedies
maintaina higherlevel of per screendaily admissionswithrespectto dramatic
filmsand othergenres8(Table II).
In theeconometric analysiswe examinewhichfactorsdetermine thebox office
performance ofmotionpicturesproducedin Italyfocusingon threecrucialissues.
The firstis relativeto howpopularity ofhumaninputs(director and castof actors)

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
 OFMOTION
THEDETERMINANTS PICTURE PERFORMANCE
BOXOFFICE 241

TableI. Boxoffice
performance formovies
indicators
producedinItaly
(1985-1996)
Year Numberof Averageper Average
movies screen
daily release
on
in admissions lengtha
produced
Italy thebigscreen
1985 82 234.3 282.4
1986 80 264.1 300.5
1987 84 233.3 280.3
1988 107 164.0 498.0
1989 92 136.8 430.0
1990 92 154.1 402.8
1991 82 155.3 441.8
1992 81 142.7 563.7
1993 79 140.5 468.4
1994 62 130.1 562.2
1995 63 125.6 694.3
1996 63 154.2 817.7
s
a Releaselength
= where s is thetotalnumber
i=1
onwhich
ofscreens themovieis releasedand8¡ is the
numberoftotal daysonthei-thscreen.
release

1. Specialisation
Figure genre formovies
split inItaly
produced (1985-1996).

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
242 M.BAGELLA
ANDL.BECCHETTI
perscreen
TableIL Average dailyadmission
forcomedies
andthriller-horror
films in
produced
(1985-1996)
Italy
Year Comedies Dramatic
movies Others
per Number
Average of per Number
Average of per Number
Average of
screen
daily movies screen
daily movies screen
daily movies
admissionsproducedin admissionsproducedin admissionsproducedin
Italy Italy Italy
1985 258.1 52 207.9 17 161.0 7
1986 295.7 52 197.3 17 232.4 8
1987 247.1 52 225.1 20 169.8 9
1988 174.6 66 146.8 28 149.1 12
1989 166.7 43 114.3 23 92.6 18
1990 171.6 54 123.4 29 128.2 5
1991 171.3 51 136.9 25 91.3 4
1992 164.5 46 117.0 30 62.4 2
1993 183.7 38 99.2 36 109.7 5
1994 161.3 31 104.4 26 70.8 4
1995 137.5 37 114.8 20 53.0 2
1996 164.9 36 138.0 22 208.2 2
Eachoneoftheaggregated specialisation
genres(forwhich statistical
valuesarepresented
in
theabovetable)includes
several
disaggregated
specialisation
genres tothefollowing
according
taxonomy:Comedies (Comic,
Comedy, Sentimental,
Musical-Opera,Cartoon);Others
(Horror,
Detective
Stories,
War-Adventure,
Fantastic,
Erotic,
Documentary); Dramatic
films(Dramatic,
Historical, Melodrama
Generational, Socio-Political,
Biography, Source:
Esistential). CEIS-Tor
database.
Vergata

affectsmotionpictureperformance. The secondfocuseson therelativeimpactof

statesubsidisation. The thirdregardstherelativeimpactof additionalfactorssuch
as theorganisational and marketing capacityof production housesand theItalian
viewerstastespecificity affectingtherelative
success of specialisationgenres.
Withregardto thefirstissue,a well establisheda prioritheoretical hypothesis
on theimpactof humanfactoron sports,artsand letters,knownas theRosen's
(1981) superstar phenomenon, claimsthatmarketsize and revenuesin thesefields
of activitytendto be skewedtowardmoretalentedpeople fortwo mainreasons:
the limitedsubstitutability of hightalentswithrelativelylesser talentsand the
economiesofscale in theconsumption ofthesetypesofproducts.9 Ourtheoretical
a prioriis that,also formotionpictures,a relativelylowerex antepopularity is a
poorsubstitute forhigherpopularity so thathumaninputsshouldhavea nonlinear
impacton totaladmissions.In addition,consideringthattheinteraction between
talentedpeople oftengeneratespositiveexternalities, we also postulatethatthe
interactionbetweendirector andcastofactors'popularity mustgenerateadditional
positiveexternalities on motionpictureperformance.

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
THEDETERMINANTS
OFMOTION BOXOFFICE
PICTURE PERFORMANCE 243

As a consequencearguments fromtheoretical
literature
induceus to consider
thefollowingspecification
fortheeconometric
analysis:

Tadm = a0 + «1 (DEAP)2 + a2(AEAP)2 + a3(AEAP * DEAP)2 +

+ a^Subs + a5dgdp-fa^dpr + a7Forb+ Nocopr+
19 17 (1)
+ Y fl(QQiM)/Dprod; + Y ^(QQiM)¿Dtypes¿-
+ S' ,
i=1 k='

wherethedependentvariable(Tadm)is represented by thetotalnumberof admis-

sions.The regressors includequadraticdirector'sand actors'ex antepopularities
(respectivelyDEAP2 and AEAP2), a quadratichumaninputinteraction factor
*
(DEAP AEAP)2, a set of dummies for producers(Dprod)10 and specialisa-
tion genre(Dtypes) effects.The 20 main producersof domesticallyproduced
filmsbetween 1985 and 1996 are ACADEMY, ARTASS, BIM, CDI, CEN-
TURYF, CECCHIG, CIDIF, COLUMBIA, CRC, DARC, DLF, FILMAURO, IIF,
INDREG, LUCE, MEDUSA, MIKADO, PENTA, TITANUS, UIP. The 18 Dtypes
are the following:WAR-ADVENTURE, BIOGRAPHY, CARTOON, COMIC,
COMEDY, DOCUMENTARY, DRAMATIC, HISTORICAL, EROTIC, EXIST-
ENTIAL, FANTASTIC, GENERATIONAL, HORROR, MELODRAMA, MUSI-
CAL-OPERA, DETECTIVE STORY, SOCIO-POLITICAL, SENTIMENTAL.11
The otherthreedummiesin theEquation(1) pickup theeffectsof admittance
restriction
forviewersyoungerthan18 (Forb),of filmsubsidisation12 (Subs) and
ofabsenceofparticipation offoreigncoproducers(Nocopr).A finalvariablewhich
mustbe takeninto accountis the one of time-related effectson domestically
produced movies (i.e. changes in real ticketprices,seat capacity,tastesaffecting
overallfilmdemandand real income)whichobviouslyaffecttheperformance of
twomoviesproducedin twodifferent years. We then model
explicitly year-related
effectsbyintroducing yearlychangesin income(dgdp) and yearlychangesin real
ticketprices(dpr).13
(1) is testedagainstthefollowingalternatives
Specification whichcombinein
different
wayslinear, nonlinearand interaction effectsof human inputson motion
pictureperformance:

Tadm = ß$ + ß' (DEAP) + ft(AEAP) -f-ftSubs + dgdp+ ftdpr+

19
+ ftForb+ jSyNocopr + Y] Ö(LM)/Dprodi+
tí (2)
17
+ ^2 *(LM)*Dtypest + e2
k=]

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
244 M.BAGELLA
ANDL.BECCHETTI

Tadm = yo+ y' (DEAP) + y2(AEAP) + (AEAP * DEAP) +

+ y4Subs+ y5dgdp+ y^dpr+ y7Forb+ y8Nocopr+
19 17 (3)
+ £3
+ ^2 ^(LLiM)jfcDtypeSfc
+ ^2/ ^(LLiM)/Dprodř-
î=i k=1

Tadm = 5o H-5i (DEAP)2 + $2(AEAP)2 + ^Subs + 54dgdp+

19
+ 55dpr+ S6Forb+ á7Nocopr+ ^ 0(QM);Dprod¿ +
«'=1 (4)
17
+ k+ £4•
^*(QM)fcDtypes
k=1
Let us define(1) as the"quadraticwithquadraticinteraction" model(QQIM),
andalternatives (2), (3) and (4) respectivelyas the "linear"model (LM), the"linear
withlinearinteraction" model(LLIM) and the"quadratic"model(QM). In carry-
ingoutmodelestimateswe mustsolve someimportant problemswhichare likely
to bias our results.The distribution of the dependent variable(and also of other
performance variables such as totalrevenues and per screendailyadmissions)is
highlyskewedwithhigh(low) densityforlow (high) revenue picturesas clearly
shownby Table III.14 Almost40 per centof domesticallyproducedmoviessell
less than10,000tickets, earnless than100 millionliras(around40,000 pounds),15
whilethetoptenpercentof successfulmoviessell morethan350,000ticketsand
earnmorethan2.5 billionliras(around1 millionpounds).
Residuals fromOLS regressionscarriedon thefourmodels describedabove
do notpass normality tests.16 This violatesone of thefundamental assumptions of
OLS regression andprevents traditionalinference on parameters significance from
beingcorrect.
In orderto solve thisproblemwe adopttheapproachof McKinlay-Richardson
(1991) and Clare et al. (1997) and use a GMM-HAC (GeneralisedMethodofMo-
mentsHeteroskedasticity and Autocorrelation consistentapproach)17whichsets
thecalculatedcorrelations of theinstruments and disturbances as close to zero as
possible,according to a criterion a
providedby weighting matrix. GMM relies
on weakerassumptionsthanOLS and does notrequirenormality and constancy
of varianceof regressionresidualsforthe correctestimateof the parameters'
confidenceintervals.As withClare et al. (1997) thesame regressorsare used as
instruments.
In testingforthecorrectmodelspecification, a preliminary evidenceofthelin-
ear/nonlinearimpactofhumanfactorsonbox officeperformance is providedbythe
correlationmatrixwhichillustrates thelinksbetweenthedependentvariableand
varioustypesof director'sand cast of actors'popularity impact.Table IV shows
thatthequadraticandlinearinteraction factors(DEAP* AEAP2 andDEAP* AEAP)
have the strongest correlation withtotaladmissions,followedby quadraticindi-

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
THEDETERMINANTS
OFMOTION
PICTURE
BOXOFFICE
PERFORMANCE 245
TableIII. Cumulative
distribution
ofmainquantitative
indicators
oftheatrical for
performance
movies inItaly
produced
Percentile TotaladmissionsPerscreen
daily Daysofrelease Revenues
(1,000
admissions liras)
0 17 8.5 1 99
5 552 43.07 8 4,136
10 1041.2 60.25 13 7,962
15 1833 69.6 21 12,818
20 2613.8 79.23 32 19,831
25 3803 86.83 43 28,136
30 5541.8 99.04 57 41,174
35 7887.4 108.93 72 53,144
40 10600.8 120.45 95 79,709
45 15357.8 128.35 133 115,468
50 21427 137.86 170 156,638
55 31129.6 147.18 220 221,810
60 45205.4 160.31 289 316,169
65 63020 175.77 355 458,946
70 90857.6 196.77 437 651,218
75 120949 220.01 552 825,562
80 162692.2 247.95 714 1,197,432
85 242130 285.45 952 1,692,751
90 354505.6 344.3 1282 2,588,637
95 646842 416.48 2024 5,070,734
100 3514740 914.36 6300 35,013,675

Mean 130098.1 171.455 469.148 2,312,193

Median 21427 137.862 171 156,638
N.obs. 977 977 977 977

Normalitytest 72993.76
(0.00) 733.32(0.00) 8500.53(0.00) 142263.3
(0.00)
J.B.(p-value)

vidualeffects betweenthesefactors
(DEAP2 andAEAP2). The strongcollinearity
the of
anticipates difficulty highlightingthe impactof themon total
differential
admissions.
GMM estimatesshowthattheQQIM specification yieldsthebestfit(Table V).
of the QQIM model againstall otherspecifications
The (slight)superiority is
successfullytestedthroughone-by-oneDavidson-MacKinnon(1993) J-testfor
non-nestedmodels.18FitteddependentvariablesfromLM, LLIM and QM spe-

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
246 M.BAGELLA
ANDL.BECCHETTI
TableIV.Correlation
matrix
of totaladmissions
andvarious of human
specifications factors
boxoffice
affecting performance
AEAP AEAP2 DEAP DEAP2 AEAP* AEAP2* TADM
DEAP DEAP2
AEAP 1.000
AEAP2 0.954 1.000
DEAP 0.548 0.561 1.000

DEAP2 0.474 0.507 0.947 1.000

AEAP*DEAP 0.735 0.792 0.874 0.874 1.000
AEAP2*DEAP2 0.590 0.685 0.749 0.835 0.933 1.000
TADM 0.467 0.542 0.492 0.500 0.591 0.591 1.000
Variable DEAP:director
legend: exantepopularity; exantepopularity;
AEAP:actor TADM:total
admissions.

cifications(separatelyevaluated)do nothave anysignificant impactas additional

regressors in the QQIM model (if we consider as a thresholdthe 95 per cent
significancelevel), while the estimatedQQIM dependentvariablehas signific-
ant f-values when consideredas additionalregressorin LM, LLIM and QM
specifications (Table VI).
GMM resultsfromtheQQIM estimateprovidesomepreliminary answerstoour
threequestions.First,theex antepopularity ofhumaninputsaffects in a nonlinear
waybox officeperformance andtheinteraction betweenthetwofactors'popularity
has a positiveimpacton totaladmissions.The limitedrelevanceof separatequad-
raticfactors(DEAP2 andAEAP2) in theQQIM specification is probablydue tothe
highcollinearity betweenthesetworegressors and thequadraticinteraction effect
evidencedbythecorrelation matrix(Table IV).
Withregardto our second question,a quite surprising resultshowsthatsub-
sidised filmsdo not have a significantly lower performance in the econometric
analysis in terms of totaladmissions,daily revenues and per screendailyadmis-
sions (Table V). Simple subgroupmean comparisonsshow thatsubsidisedfilms
have on average83 per screendaily admissionsagainst884 fornon subsidised
films,and thattheaverageactors'and director'sex antepopularity of subsidised
films(respectively 0.52 and 0.07) is muchlowerthanthatof non subsidisedfilms
(respectively 1.43 and 0.64).
The jointinterpretation of theseresultssuggeststhattheneteffectof subsidies
on themeanof thedependentvariableis irrelevant, and thattheabove mentioned
considerabledifference in averagedailyand totaladmissionsis explainedby the
farlowerex antepopularity of cast and directorsof subsidisedfilmsas compared
to non subsidisedfilms.

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
THEDETERMINANTS
OFMOTION
PICTURE PERFORMANCE
BOXOFFICE 247
ofboxoffice
TableV.Thedeterminants variable:
(dependent
performance totaladmissions
divided
by10,000)
with Quadratic
Quadratic model Linear model with
Linear linear
inter-
quadratic model
interaction
action
model
Constant Coefficient f-Stat.Coefficientf-Stat.Coefficient t-
Stat.
f-Stat.Coefficient
-979.12-0.96 -766.36-0.72 -428.92-0.39 -908.18-0.86
Human AEAP2 1.51 1.75 3.47 6.50
factors DEAP2 0.50 0.34 4.45 4.72
(AEAP*DEAP)2 0.90 2.39
AEAP 6.92 5.62 -1.35 -0.65
DEAP 10.96 5.83 -4.80 -1.28
(AEAP*DEAP) 9.73 3.87
Production ACADEMY -2.57 -0.78 -3.93 -1.19 -5.33 -1.62 -3.82 -1.18
houses ARTASS -1.60 -0.56 -^.63 -1.59 -6.82 -2.17 -1.89 -0.63
BIM* -3.36 -0.80 -6.34 -1.44 -8.51 -1.80 -4.09 -1.01
CDI -1.13 -0.27 -3.93 -0.99 -4.43 -1.11 -1.27 -0.30
CENTURY 3.65 1.46 3.49 1.37 1.69 0.56 3.52 1.31
CIDIF -0.30 -0.09 -2.84 -0.84 -5.41 -1.59 -0.31 -0.10
COLUMBIA 7.77 1.56 4.90 0.90 3.96 0.68 7.82 1.47
CRC -0.03 -0.01 -1.09 -0.41 -3.65 -1.10 -1.36 -0.51
DARC 0.69 0.11 -3.37 -0.60 -5.61 -1.03 0.75 0.12
DLF -0.94 -0.31 -3.74 -1.22 -5.80 -1.82 -1.71 -0.56
FILMAURO 30.94 3.21 27.58 2.79 28.52 2.70 30.46 3.06
IIF -2.92 -0.93 -4.47 -1.30 -4.92 -1.43 -4.15 -1.31
INDREG -1.97 -1.17 -1.45 -0.82 -1.19 -0.64 -1.96 -1.18
LUCE* -1.39 -0.60 -3.73 -1.60 -5.44 -2.20 -1.98 -0.86
MEDUSA* -3.43 -1.22 -6.64 -2.13 -8.09 -2.44 -4.12 -1.39
MIKADO* 0.09 0.04 -0.69 -0.32 -1.57 -0.67 0.61 0.28
PENTA* 4.40 1.20 2.44 0.63 2.91 0.70 4.76 1.25
TITANUS -0.55 -0.16 -2.90 -0.84 -4.13 -1.21 -0.73 -0.22
UIP* -0.12 -0.03 -4.28 -0.95 -7.10 -1.53 -1.95 -0.42
WAR-
Specialisation 1.51 0.94 2.91 1.58 2.53 1.31 0.77 0.47
genres ADVENTURE
BIOGRAPHY 0.59 0.19 -3.49 -0.98 -4.86 -1.41 0.71 0.22
CARTOON -0.37 -0.12 -5.04 -1.11 -7.84 -1.51 3.96 0.69
COMIC 15.76 3.32 13.21 3.00 13.86 3.02 15.56 3.53
DOCUMENTARY -0.25 -0.13 1.22 0.58 -2.37 -0.72 0.77 0.29
DRAMATIC -0.51 -0.36 -0.68 -0.49 -1.70 -1.16 -1.47 -1.01
HISTORICAL 3.59 0.44 -0.68 -0.08 -1.17 -0.13 3.07 0.36
EROTIC -1.04 -0.32 -2.56 -0.85 -2.58 -0.87 0.02 0.01
EXISTENTIAL -0.10 -0.04 -0.48 -0.19 -0.08 -0.03 0.68 0.27
FANTASTIC 1.33 0.86 4.34 2.07 4.21 1.82 0.51 0.34
GENERATIONAL 1.39 0.45 0.92 0.30 0.05 0.01 2.02 0.68
HORROR 10.01 1.96 9.44 2.01 9.00 1.88 9.12 1.62
MELODRAMA4.79 1.19 0.07 0.02 -1.29 -0.43 4.61 1.18
MUSICAL-OPERA -8.83 -1.45 -10.53-1.29 -11.50 -1.32 -11.34-1.63
DETECTIVE 3.52 1.72 6.33 2.48 6.42 3.32 2.99 1.32
STORIES
SOCIO-POLITICAL 6.11 2.08 3.34 1.31 2.50 0.93 5.98 2.12
SENTIMENTAL -0.46 -0.21 -1.00 -0.39 -1.78 -0.55 0.10 0.04
Other DGDP 0.24 0.96 0.38 0.72 0.42 0.38 0.39 0.74
variablesDPR -0.49 -0.96 -0.38 -0.72 -0.21 -0.38 -0.46 -0.87
SUBS -0.24 -0.22 1.81 1.21 1.27 0.85 -0.38 -0.33
NOCOPR 0.89 0.26 2.52 0.74 1.29 0.36 -0.56 -0.16
FORB 4.11 1.12 3.85 1.20 3.13 1.06 4.02 1.19
BIENNALE 0.46 0.17 -0.64 -0.23 -0.10 -0.04 0.01 0.00
/{-squared
Adjusted 0.48 0.44 0.38 0.44
N.ofobs. 977 977 977 977
The
Legend: table and f-statistics
coefficients
presents forthe with
"quadraticquadratic model
interaction" the
(QQIM), model
"linear" (LM),
the with
"linear linear
interaction"
model(LLIM) andthe model
"quadratic" (QM)(see inthe
(l)-(4)
Equations The
text). variable
dependent
istotal
admissions
(divided The
by10,000). include
regressors the ofthe
indexes director's
andactors'
exante DEAP
(respectively
popularity
andAEAP),aset
ofdummies
for (Dprod
producers whichare ACADEMY,
specifically ARTASS,BIM,CDI, CENTURYF, COLUMBIA,
CIDIF,
CRC,DARC, DLF,FILMAURO,
IIF,
INDREG, LUCE,MEDUSA, MIKADO,PENTA,TITANUS,UIP. variables
Starred indicate who
producers
arealso and
distributors)
specialisation
genre effects
(Dtypeswhich
are WAR-ADVENTURE,
specifically BIOGRAPHY,CARTOON,COMIC,
COMEDY, DOCUMENTARY, DRAMATIC, HISTORICAL, EXISTENTIAL,
EROTIC, FANTASTIC,GENERATIONAL, MELO-
HORROR,
DRAMA, MUSICAL-OPERA,DETECTIVE STORIES,SOCIO-POLITICAL, Other
SENTIMENTAL). three
dummies upthe
pick effects
of
admittance for
restriction
viewers than
younger 18(Forb),offilm and
subsidisation
(Subs) ofabsence offoreign
ofparticipation coproducers
(Nocopr). are
Equations with
estimatedaGMM Method
(Generalised ofMoments) with
approach and
Heteroskedasticity Con-
Autocorrelation
sistent Matrix.
Covariance The functional
Bartlett's formofthekernel
isused the
toweight incalculating
covariances the matrix.
weighting The
NeweyandWest(1994) bandwidth
automatic isadopted
procedure todetermine inside
weights for
kernels The
same
autocovariances. regressors
areused
asinstruments.

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
248 M.BAGELLA
ANDL.BECCHETTI
TableVI.Davidson
McKinnon
y-test
onnon-tested
hypotheses
QQIM QM LM LLIM
QQIM - - 0.15 0.13
QM 0.01 - 0.Û01 0.07
LM 0.0001 0.000 - 0.0001
LLIM 0.032 0.046
Legend: Thetablepresents values
probability fortherejection
of
thenullhypothesisthat
thedependent
variable
estimatedinthe
vertical
headermodelhasnoadditional
explanatory
powerwhen
as regressor
introduced inthehorizontal
header
model. Thetest
is notperformed
when thevertical
headermodel inthe
is nested
horizontal
header
model.

As forthethirdquestion,whenlookingat theresultson thenetimpactofdiffer-

entItalianproduction houseson thedependentvariable,we findthatonlyFilmauro
has a significantpositiveeffecton totaladmissionsin all ofthefourmodels.19
Resultson therelativeimpactofspecialisationgenreson thedependent variable
give us some hints on the tastesof Italian movie consumers. The positiveand
significant effectof the comic genreon totaladmissionsshows thatthe choice
of producingthesetypesof filmshas an independent positiveeffecton box office
revenuesnetofex antecastanddirector popularity.Evidenceon thepositiveeffect
of socio-political,horrorand detectivestoriesis muchweakerand notconfirmed
acrossall modelspecifications.
The above describedGMM resultsméasuretheimpactof selectedregressors
on the mean of the dependentvariable.Thé highlyskeweddistribution of this
variablemakes theseresultsless generaland inducesus to integrate themwith
an analysisoftheregressor's impacton themedian,or on different quantilesofthe
distribution.To obtainconsistent standarderrorestimates(in spiteoftheresiduals
nonnormality) we hereprefertheGould (1992) approach,whichobtainsstandard
errorestimatesusing bootstrapresampling,ratherthanthe methodof Koenker
and Basset (1982) and Rogers(1993) whichunderstates standarderrorsin case
ofhomoskedastic errors.20
In Table VII we reporttheestimateresultson theimpactofhumanfactors, pro-
ductionhouses,specialisationgenresand otherregressors on themedian,as well
as the25thpercentile andthe75thpercentile ofthedistribution
oftotaladmissions.
Theseestimatesconfirm thattheGMM estimateon thevariablesaffecting themean
oftotaladmissionsprovidesa good representation ofthedeterminants shapingthe
distribution of the dependentvariable.Again, quadratichumanfactorsplus the
interaction coefficient,theFilmauroproduction house and thecomicgenrehavea
significantimpacton theselectedpercentiles ofthedependentvariable.21 The only
relevantdifference is thatthe director'spopularityseems to have no significant

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
THEDETERMINANTS
OFMOTION
PICTURE
BOXOFFICE
PERFORMANCE 249
TableVILThedeterminants
of theatrical in Italy.Quantile
of moviesproduced
performance
estimates
regression
Var:
Dep, medianoftotal Dep,
Var25th oftotal
quantileadmissions Dep,Var:
0.75 oftotal
quantile
admissions
(thousands)(thousands) admissions
(thousands)
QQIMmodel LLIMmodel QQIMmodel
Constant f-Statistic
Coefficient Coefficient
f-Statistic
Coefficient
ř-Statistic
-7.76 -1.03 -2.28 -0.52 11.60 0.90
Human factors
AEAP2 7.93 3.12 AEAP 2.70 3.23 16.04 2.67
DEAP2 10.90 2.75 DEAP 0.96 0.27 20.54 2.76
(AEAP *DEAP)2 6.78 3.89 (AEAP *DEAP)3.74 5.98 9.58 4.57
Production
houses
ACADEMY -3.65 -0.50 -2.32 -0.30 -9.95 -0.34
ARTASS -2.11 -0.45 -0.04 -0.01 -8.26 -0.53
BIM* -0.34 -0.02 -4.84 -0.40 -10.07 -0.38
CDI 16.61 0.66 1.83 0.11 10.57 0.31
CENTURY 7.18 1.10 5.41 2.54 10.41 0.19
CIDIF -4.79 -0.45 -1.74 -0.27 -14.57 -0.79
COLUMBIA 12.83 1.30 18.25 3.21 19.82 0.38
CRC 0.22 0.02 -3.40 -0.65 1.93 0.15
DARC 7.26 0.15 -1.49 -0.04 -3.10 -0.03
DLF 3.58 0.95 4.72 2.15 3.16 0.39
FILMAURO 138.90 1.84 94.70 3.12 417.05 2.55
IIF 5.11 0.49 -1.25 -0.25 -5.63 -0.17
INDREG -3.34 -1.12 -1.86 -1.73 -11.64 -1.86
LUCE* -2.59 -1.02 -0.24 -0.11 -10.83 -1.71
MEDUSA* 2.81 0.37 0.42 0.11 -5.57 -0.30
MIKADO* -1.91 -0.22 -1.46 -0.62 -9.55 -0.92
PENTA* 19.94 1.81 3.59 0.92 26.55 0.90
TITANUS 10.37 0.72 4.53 0.89 43.00 1.88
UIP* -10.34 -0.63 5.54 0.69 12.09 0.12
Spécialisation
genres
WAR- 1.26 0.55 0.19 0.06 -1.01 -0.36
ADVENTURE
BIOGRAPHY 32.73 1.48 7.57 0.78 0.99 0.03
CARTOON -6.19 -0.49 -2.31 -0.31 -15.33 -0.50
COMIC 65.47 2.75 27.90 3.25 144.16 4.27
DOCUMENTARY -1.99 -0.19 1.30 0.18 -5.30 -0.34
DRAMATIC -0.08 -0.07 -0.75 -0.35 -1.01 -0.48
HISTORICAL-11.87 -1.00 -5.22 -0.59 -10.67 -0.69
EROTIC -8.52 -1.12 0.14 0.05 -4.02 -0.29
EXISTENTIAL3.35 0.98 0.86 0.21 5.50 0.28
FANTASTIC 1.23 0.15 -0.03 -0.01 -0.93 -0.06
GENERATIONAL 5.36 0.34 0.87 0.16 1.66 0.11
HORROR 73.52 1.82 14.96 0.72 134.14 1.29
MELODRAMA 51.74 1.65 62.89 2.19 2.12 0.06
MUSIC.-OPERA -65.75 -1.98 2.77 0.07 -7.91 -0.07
DETECTIVE ST. 1.78 0.68 2.24 0.30 -3.69 -0.28
SOCIO-POL. 39.84 1.70 10.51 2.09 95.58 1.92
SENTIMENTAL -0.58 -0.11 0.33 0.10 -4.94 -0.84
Othervariables
DGDP 2.11 0.36 3.18 0.55 3.17 0.22
DPR -5.63 -1.13 -4.25 -0.29 3.30 -0.49
SUBS 0.74 0.65 0.32 0.26 -0.20 -0.11
NOCOPR 10.34 1.53 3.60 0.87 1.45 0.12
FORB 14.31 1.90 4.44 1.58 21.89 1.52
BIENNALE -7.65 -1.95 -6.96 -2.67 -11.88 -2.05
PseudoR2 0.27 0.13 0.42
N.ofobs 977 977 977
The
Legend: table coefficients
presents and f-statistics
forthemedianand
75th
quantile ofthe
regression with
"quadratic interaction"
quadratic
and
(QQIM),
specification for
the25th
quantile ofthe
regression with
"linearlinear
interaction" (LLIM)
specification (see (1)and
Equations
(3)inthe The
text). variable
dependent istotal
admissions.
The include
regressors the
indexes
ofthedirector's
andactors'
exante
popularity
DEAP
(respectively and aset
AEAP), ofdummiesfor
producers which
(Dprod are ACADEMY,
specifically ARTASS,BIM, CENTTJRYF,
CDI,
CIDIF,
COLUMBIA, CRC,DARC, DLF,FILMAURO,IIF,
INDREG, LUCE,
MEDUSA,MIKADO, PENTA,TITANUS,UIP.
Starred
variables
indicate who
producers arealso and
distributors)
specialisation
genreeffects which
(Dtypes are WAR-
specificallyADVENTURE,BIOGRAPHY,
CARTOON, COMIC, COMEDY, DOCUMENTARY, DRAMATIC, HISTORICAL,
EROTIC,
EXISTENTIAL,FANTASTIC,
GENERATIONAL,
HORROR, MELODRAMA, MUSICAL-OPERA, DETECTIVESTORIES,SOCIO-POLITICAL,
SENTIMENTAL). three
Other dummies
pickup
theeffects
ofadmittance
restriction
forviewers than
younger offilm
18(Forb), subsidisationand
(Subs) ofparticipation
ofabsence offoreign
coproducers Standard
(Nocopr). errors
areestimated
with
the
Gould (1992)
bootstrapping 50trials.
using
technique Standard
errors with
convergence
ahigher
numberoftrials
ischecked
by theauthors.

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
250 M.BAGELLA
ANDL.BECCHETTI
TableVIIL Davidson
McKinnon J-test
on non-nested
hypo-
theses
onquantile estimates
regression
QQIM QM LM LLIM
Median
regression
QQIM - - 0.804 0.458
QM 0.002 - 0.004 0.014
LM 0.000 0.000 - 0.000
LLIM 0.306 0.332
0.25Quantile
regression
QQIM - - 0.003 0.623
QM 0.028 - 0.018 0.056
LM 0.000 0.077 - 0.003
LLIM 0.958 0.969
0.75Quantile
regression
QQIM - - 0.933 0.772
QM 0.000 - 0.000 0.000
LM 0.000 0.000 - 0.996
LLIM 0.172 0.553
Legend:Thetablepresents values
probability fortherejection
of
thenullhypothesis
that
thedependent
variable
estimated
inthe
vertical
headermodelhasnoadditional
explanatory
powerwhen
introduced
asregressor
inthehorizontal
headermodel.Thetest
isnotperformed
when thevertical
header
model isnested
inthe
horizontal
headermodel.

effectson the25thpercentileof totaladmissionsor,in otherterms,seemsnotto

improvetotaladmissionsforlow successmovies.
significantly
A somewhatexpectedresultthoughis that,whentestingwhethertheQQIM
is thecorrectspecification
forestimatingthedeterminants of quartilesof thede-
pendentvariabledistributionwiththeDavidson-MacKinnon(1993) procedure, we
cannotdiscriminate anymorebetweentheQQIM and theLLIM. Furthermore the
LLIM seemsto be preferred in the25thpercentileestimate(Table VIII). For this
reasonwe do presentLLIM estimatesinthe25thpercentile regression (TableVII).

3. Conclusions
Severaltheoretical
papersin theliteratureofarteconomicsemphasisedsimilarities
betweenthetwo activitiesof investing in high-techindustrialsectorsand produ-
cing movies.The highlyskeweddistribution of measuresof movieperformance
withtheextremely highprobabilityof low success filmsconfirms thatproducing
a movie is an extremelyriskybet also in Italy.Giventhispreliminary evidence

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
THEDETERMINANTS
OFMOTION BOXOFFICE
PICTURE PERFORMANCE 251

a relevantquestionis thento understand whichdeterminants makethisbet safer.

To investigatetheissue thispaperpostulatesthatmotionpicturesuccess maybe
consideredas theoutputof a three-factor production function theinputsof which
are representedby two heterogeneous types of labour (the cast of actorsand the
and
director) by a residualfactor affected by several variables which includecon-
sumertastesand production housesability.In thespecial case of moviesthemain
determinant ofthequalityofthesetwofactors(whichis equivalentto theintensity
oftheirutilisation in physicalproduction, and generallydetermining higherlevels
of outputas faras thelaw ofpositivemarginalproductivity holds) identified
is by
theirex antepopularity. The existingliterature on the effectof talentin sectors
suchas arts,sportsand literature suggeststhatthehumaninputqualityhas positive
nonlineareffectson motionpicturesuccess (Rosen, 1981,De Vany- Walls, 1997).
The empiricalresultsof thispaperdescribetheperformance of moviesproduced
in Italyin the last twelveyearsand are broadlyconsistentwiththisconceptual
framework.
Descriptiveevidenceshowsat first howtheItalianmovieindustry appearsto be
clearlylabourintensive, givenitsdespecialisation in genresimplyinghugeinvest-
mentin capitalinputssuchas studio'sinfrastructure and special-effects.22During
thetwelveyearsanalysedin thepaperwe finda reductionin thetotalnumberof
domestically producedmoviesand a sharpdropin dailyrevenuesand admissions.
Econometric estimatesoftheimpactof variousdeterminants of motionpicture
successare performed withtwomethodologies (Generalised Method of Moments
and quantileregressionswithbootstrapping) whiletakingintoaccounttheprob-
lems generatedby nonlinearity in thedependentvariableand in OLS residuals.
The GMM resultsconfirm thatthetheatrical performance of motionpicturespro-
ducedin Italyseemsto be cruciallyinfluenced by the impactof cast of actorsand
director'sex antepopularity plus an interaction term whichexpressesadditional
positiveexternalities on performance generatedby thetwo humanfactors.Evid-
enceinfavourofa superstardom effect is subjectto thepotentialbias in thequality
filteradopted.23
Empiricalresultsfromquantileregressions confirm thatalso othercriticalval-
ues of the distribution of the dependentvariableare positivelyaffectedby the
abovementioned humanfactors, thoughtheirimpacton the25thpercentile oftotal
admissionsseemsto be onlylinear.
Anotherinteresting resultis thatthenetimpactof subsidieson filmrevenues
is notnegative,eventhoughsubsidisedfilmshaveon averagea significantly lower
numberof totaladmissionsvis-à-visnon subsidisedfilms.This outcomeseems
justifiedby the significantly lowerex antepopularityof actorsand directorsof
subsidisedfilms.
Finally,resultson theimpactof specialisationgenreson totaladmissionsshow
thatcomic filmsmeetItalianconsumertastesand pay an extrabonusin termsof
box officerevenuesnet of actorand directorex ante popularity. Resultson the
net impactof productionhouses on theatricalperformance try evaluatehow
to

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
252 ANDL.BECCHETTI
M.BAGELLA

much the producers'managerialskillsmatterin determining filmsuccess. The

resultsshowthatonlyone producer(Filmauro)has a marginalpositiveimpacton
movieattendance, whileall othercompanieshaveirrelevant
netofotherfactors, or
negativeeffects.

Notes
* attheConference ofCentro StudiCinematografici andCEIS-Tor on
Paperpresented Vergata
"Determinants ofItalian movie performance". Theauthors particularlythank twoanonymous
referees andL. Hinna, C. Kosulich, M. Lo Cicero, A. Pasquale, M. Repetto, C. Tagliabue,
A. Zagari, foruseful comments andsuggestions andA. Simoncini fortheprecious research
assistance. Though thepaperis theresult ofjointefforts, Section 1 maybe attributed toM.
Bagella, Section 2 toL. Becchetti whileconclusions arejointly written.Theusualdisclaimer
applies.
1. Wedefine ourquality index as "exantepopularity" because thepopularity ofdirector andcast
ofactors playing ina movie isevaluated before their taking partinitandtherefore doesnottake
intoaccount theimpact ofthat movie onthepopularity itself.
2. Given theinformation available inItalytheaimofthepaper must necessarily becircumscribed
totheanalysis ofboxoffice performance. Infact, whilepriceandattendance dataarepublic
andavailable, dataonhomevideoandTV revenues forindividual films arenotreleased by
producers anddistributors andmaynotbe inferred from aggregate information contained in
balance sheets. Though theauthors arewellaware that boxoffice receiptsaccount fora declining
shareofmovie revenues after thedevelopment ofnewsources ofreceipts suchas homevideos
andnetwork TVrevenues (approximately 1/3inthefirst halfofthenineties according tofigures
from Prag-Casavant, 1994),they nonetheless believe that aninquiry onboxoffice performance,
apartfrom itsrelevance, mayprovide important information inmeasuring filmappreciation and
total
revenues, given thathome videoandnetwork TVreceipts arerelated tofilm success onthe
screen.A similar argument maybefound inDe Vany- Walls(1997).
3. Forempirical analyses onthedeterminants ofmotion pictureperformance ondifferent datasets
seeCameron (1986),Smith andSmith (1986),Wallace etal.(1993),Prag-Casavant, (1994).
4. Being a nonmonetary measure, thisvariabledoesnotrequire deflating corrections(which maybe
problematic formultiyear releases) eventhough itisobviously affected bytime effects relative
tochanges inprices, income andtastes ofviewers. Thisvariable is then preferred tomonetary
proxiesoftotal gross admissions suchas"rentals" (thepercentage rateoftotal gross admissions
tobepaidtothedistributor) usedbySmith andSmith (1986)andPrag-Casavant (1994),andis
preferred toother nonmonetary measures suchasperscreen dailyadmissions andtotal number
ofrelease dayswhich maybebiased bydifferences incapacity oftheatres inwhich movies are
released.
5. Whenchosen fortheproject thesecritics provided their evaluation ofthetrack record ofcast
members andmovie directors under therecommendation ofignoring theexpostsuccess ofthe
inobject.
artists
6. Smith andSmith (1986)useAcademy Awards as human input quality indexes, whilePrag-
Casavant (1994)follow a methodology which is more similartotheoneadopted inthispaper
byevaluating whether thefilm castincludes nostar, risingorfalling starsorestablished stars.
The"Delphi" approach which wefollow is adopted intheliterature when theevaluation made
byfieldexperts is themostreliable (ortheonlyavailable) method formeasuring a certain
variable.Relevant applications maybe found inmonetary economics withtheevaluation of
regulated/unregulated prices indifferentproduct markets (Dexter-Levi-Nault, 1993)inregional
economics withtheanalysis oftheeconomic impact ofregional investment policies (Fearne,

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
THEDETERMINANTS
OFMOTION
PICTURE PERFORMANCE
BOXOFFICE 253

1989,Miller, 1993)andinfinance with theexanteevaluation ofequity riskpremium (Prat,

1996).Fora methodological survey oftheDelphiapproach see Hwang-Ming (1987)and
Wright-Ayton (1987).
7. Thislastphenomenon must bealsopartially attributed totheemergence, alsoinItaly, oflarge
multiplex theatricalchains,which improved exhibitors' opportunities forriskdiversification and
determined longer releaselengths together with lower perscreen dailyadmissions.
8. Ifweexclude the1996dataforother genres which isclearly anoutlier.
9. Theempirical relevance oftheRosen's superstar phenomenon inthemotion picture market is
confirmed byDe Vany- Walls(1997)whofind boxoffice revenues being highly convex inrank
ona largesample ofU.S.movies. Incontrast, theempirical results ofHamlen forthemarket of
popular music (1991,1994)donotfind anyevidence ofsuperstardom andarecommented by
theauthors aspartiallydepending ontheimperfect quality filter
adopted. Inthesameway, itis
then obvious that results
eventually findingtheevidence ofsuperstardom inthemovie sector are
subject tothesamecaveats onthequality filter,though theconvexity inrankproperty offilm
revenues applies alsotooursample.
10.Intheeconometric estimates thecoefficientsforproduction houses measure theindependent im-
pactof these companies on box office revenues, netof ex antecast anddirector popularity. They
therefore areaimedatassessing themarginal effectofproducers' managerial skillsincluding
marketing strategies,castanddirector wagepolicies andskills intheoverall organisation offilm
production.
11.Onespecialisation genre (COMEDY)andoneproduction house(CECCHIG)aredropped in
order toinclude anintercept intheestimate. Thecorrespondence between aggregated specialisa-
tiongenres (presented inFigure 1andTableII)anddisaggregated genres usedintheeconometric
estimates is described inTableII legend. Disaggregated genres aretaken from themostin-
fluentialstudies onmovieconsumers inthedomestic literature(CSC, 1996;Campari, 1983;
DellaFornace, 1981;Grazzini, 1980).Thespecialisation genre classificationis notexclusive
andmovies with mixed characteristics maybeclassified inmore than onegenre.
12.Moviesubsidies inItaly(regulated byL. 1213/1965 andimplemented byL. 153/1994) provide
softloans(loanswhose interestrates are40%ofmarket rates) covering upto70percentoftotal
production costsforfilm produced inItaly,andupto90percentoftotal production costs(with
interestrates at30%ofmarket rates)forfilms of"recognised nationalcultural interest" that are
produced inItaly.
13.Wallace etal.(1993)choose another approach bysubtracting meanvalues from their dependent
variable ofmovie rental
income. Wealsotried thealternative ofintroducing time dummies for
eachyearwithout findinganysignificant difference asregards theempirical resultscommented
intherestofthepaper.
14.Using thelogoftotal admissions doesnotsolvetheproblem asthenullhypothesis ofnormality
isrejected alsoforthisvariable at99.99percent (Jarque Bera20.36).
15.When weconsider anaverage exchange rate ofabout 2,500lirasperpound forthesample period.
16.TheJarque-Bera normality testscarried onQQIM,LLIM,LMandQMspecifications estimated
with OLS indicate thatthenullhypothesis ofnormality forregression residuals is rejected at
99.999%.
17.Fora description oftheGMMapproach, seeHansen (1982)andHansen-Singleton, (1982).
18.According totheDavidson-McKinnon (1989)approach, giventwonon-nested specifications
Ml andM2,specification Ml is thecorrect oneiffitted valuesofthedependent variable from
M2donothaveexplanatory power when estimating Ml, whilefitted valuesofthedependent
variable from Ml haveexplanatory power when estimating M2.Thetestmaynotbeconclusive
if,forinstance, fittedvalues from Ml andM2arenotsignificant when addedasregressors inthe
alternative specification.

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
254 M.BAGELLA
ANDL.BECCHETTI

19. Thisresult maybeexplained bythefactthat Filmauro isalsoa distributor

thatiswellknown to
thesector's expertsbecause ofitsaggressive promotion policytargeted toa restricted
menu of
produced movies. Thedomestic literature(Sesti,
1996)emphasizes thatFilmauro's
strategyisto
maximise boxoffice revenues inorder toincreaseitspower when bargaining TVsalesreceipts
with TVcompanies.
20. Bootstrapping consists
ofdoingrepeated random drawings,withreplacements, from a given
population andestimating given statisticsateachreplication.Itis a particularly
usefulmethod
inpresence ofa nonnormal population inwhich thesampling distributionistoohardtocalculate
andstandard errorscannot be computed byusingtraditional analyticalformulas.With boot-
strapping thepopulationsampling isobtained bysimulationontheassumption that
theobserved
distributionis a goodproxy fortheunderlying population distribution.50-200drawings are
usually consideredsufficienttoobtain reliablestandarderrorsestimates (Mooney-Duval, 1993).
21. Significantparameters estimated in thesequantile regressions proveto be stablewhenwe
increase thenumber ofresample trials.
22. A plausible explanation is thatItalian producers,facingthecompetition oftheforeign (U.S.)
movieindustry (heavilybasedoncapital intensiveproduction) followed intheselastyearsa
nichestrategy relyingon thecreative ofItalian
qualities young directorsandonthecomic
skillsofmostfamous domestic actors, inthiswaythewell-known
revitalising "commedia
all'italiana".
23. A potential problem is thatthereduced interval
ofthescalenecessarily limits
thevariability
ofjudgements andmayreduce distances betweenlowandhigh popularitywithrespecttowhat
might haveemerged forinstance ina 0-10scale.

References

Bagella,M.,1997,"Economia e Finanzadell'Industria Mimeo.

Cinematografica".
Bagella, M. andBecchetti, L. (1995)"TheBuy-Out PropertyRight ShareChoiceinFilmFinan-
cing:Financial Rationing,Adverse Selection andtheBayesian Dilemma". Journal
ofCultural
Economics 19:279-304.
Baumol, W.J. andBowen, W.G.(1976)"Arguments forPublicSupportofthePerforming Arts",in
M.Blaug(ed.),TheEconomics oftheArts,Martin RobertsonandCompany.
Blaug,M.andKing,K. (1976)"DoestheArtCouncil KnowWhat itisDoing?",inM.Blaug(ed.),
TheEconomics oftheArts, MartinRobertson andCompany.
Cameron, S. (1986)"TheSupply andDemand forCinema Tickets:SomeU.K.Evidence". Journal
ofCultural Economics 10.
Campari, R.(1983)Il Racconto delFilm:Generi, Personaggi,
Immagini. Bari.
Laterza,
Clare,Smith, andThomas(1997)"U.K. StockReturns andRobust Testsof Mean- Variance
Efficiency". JournalofBanking andFinance , 641-660.
Chung, K.H.andCox,R.A.K.(1994)"AStochastic ModelofSuperstardom: AnApplication ofthe
YuleDistribution". Review ofEconomics andStatistics76(4):771-775.
CSC(Centro StudiCinematografici)(1981)ComesiLeggeunFilm:Introduzione Critica
ai Generi
e ai Fenomeni CSCed.,Roma.
Cinematografici.
Davidson, RussellandMacKinnon, James G. (1993)Estimation andInference inEconometrics.
Oxford UniversityPress.
DellaFornace, L. (1981)Comesi LeggeunFilm:Introduzione Criticaai Generie ai Filoni
Cinematografici. Bulzoni.
De Vany, A.S.andWalls, A.(1997)"TheMarket forMotion Pictures:
Rank,Revenue andSurvival".
Economic Inquiry35(4):783-797.

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
THEDETERMINANTS
OFMOTION
PICTURE
BOXOFFICE
PERFORMANCE 255

Dexter,A.S.,Levi,M.D.,andNault, B.R.(1993)"Freely Determined versusRegulated Prices:Im-

plications fortheMeasured Linkbetween Money andInflation".
Journal-of-Money, -Credit
, and
-Banking 25(2):222-230.
Efron,B. andTbishirani, R.(1986)"Bootstrap Measures forStandardErrors,Confidence Intervals
andOther Measures ofStatistical
Accuracy". Statistical
Science1:54-77.
Fearne,A. (1989)"TheCAPin1995- A Qualitative Approach toPolicy Forecasting".European-
Review ofAgricultural-Economics 16(1):113-127.
Finn,A.,Hoskins, C., andMcFadyen, S. (1996)"Telefilm CanadaInvestment inFeature Films:
Empirical Foundations forPublic Policy". Canadian PublicPolicy22(2):151-161.
Gould, W.W.(1992)"Quantile Regressions withBootstrapped Standard Errors".StataTechnical
Bulletin 9: 19-21.
Grazzini,G.(1980)LeMilleParole delCinema. Bari.
Laterza,
Hamlen, W.A.(1991)"Superstardom inPopular Music:Empirical Evidence".Review ofEconomics
andStatistics 73(4):729-733.
Hamlen, W.A.(1994)"Variety andSuperstardom inPopular Music".Economic Inquiry 32(3):395-
406.
Hansen, L. (1982)"LargeSampleProperties ofGeneralised Methods ofMoments Estimators".
Econometrica 50:1029-1054.
Hansen, L. andSingleton, K. (1982)"Generalised InstrumentalVariablesEstimationinNon-Linear
Rational Expectations Models".Econométrica 50:1269-1286.
Heilbrun,J.andGray, C.M.(1993)TheEconomics ofArtandCulture: AnAmerican Perspective.
Cambridge University Press.
Hwang, C.L.andMing, J.(1987)Group Decision Making under MultipleCriteria:Methods and
Applications Lecture NotesinEconomics andMathematical SystemsSeries, Vol.281,p. 400,
Springer, NewYork, Berlin,London, andTokyo.
Koenker, R. andBasset,G. (1982)"Robust TestsforHeteroskedasticityBasedon Regression
Quantiles". Econometrica 50:43-61.
McKinlay, C.andRichardson (1991)"Using GeneralisedMethod ofMoments toTestMean-Variance
Efficiency". JournalofFinance 46(2):511-527.
Miller,M. (1993)"Enhancing Regional Analysis withtheDelphiMethod". Review ofRegional
Studies 23(2):191-212.
Mooney, C.Z. andDuval,R. (1993)Bootstrapping: A Nonparametric Approach to Statistical
Inference. SagePublications, Newbury Park, CA.
Newey, Whitney andWest,Kenneth (1994)"Automatic Lag Selection in Covariance Matrix
Estimation". Review ofEconomic Studies 61:631-653.
Prag,J.J.andCasavant, J.(1994)"AnEmpirical Study ofDeterminant ofRevenues andMarketing
Expenditures intheMotion Picture
Industry". JournalofCulturalEconomics 18(3): 217-235.
Prat,G. (1996)"Experts' Expected StockReturn andtheValuation Modelin U.S.".Revue-
Economique 47(1): 85-110.
Rosen,S. (1981)"TheEconomics ofSuperstars". American Economic Review 71(5): 845-858.
Rogers,W.H.(1993)"Calculation ofQuantile Regression StandardErrors".
StataTechnical Bulletin
13:18-19.
M.(1996)La ScuolaCinematografica
Sesti, Italiana.
MarsilioEditore.
Scitovsky,T. (1976)"What's Wrong with theArtsis Whatis Wrong withSociety", inM. Blaug
(ed.),TheEconomics oftheArts, Martin RobertsonandCompany.
Smith,S.P.andSmith, V.(1986)"Successful Movies: A PreliminaryEmpirical Analysis". Applied
Economics 18(5):501-507.
Vogel,H.L.(1994)Entertainment Industry Economics.A Guide forFinancial Analysis. Cambridge
University Press.

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions
256 M.BAGELLA
ANDL.BECCHETTI

Wallace, M. (1993)"TheRoleofActors
A.,andHolbrook,
W.T.,Seigerman, andActresses
inthe
HowMuchIs a MovieStarWorth?".
SuccessofFilms: Journal
ofCultural-Economics
17(1):
1-27.
G. andAyton,
Wright, P.(1987)Judgmental
Forecasting.
Wiley, NewYork,
Chichester, Brisbane,
andToronto.

This content downloaded from 185.2.32.134 on Tue, 24 Jun 2014 23:29:00 PM

All use subject to JSTOR Terms and Conditions