11
Cinematic Success, Aesthetics, and Economics: An Exploratory Recursive Model Dean Keith Simonton University of California, Davis Although the reputation of creative artists is based largely on the merit of their work, the latter can sometimes be assessed in several different ways that may not necessarily agree. This lack of evaluative consensus is perhaps most apparent in cinematic success; this can be judged by film critics (initial and final), movie awards (picture, dramatic, visual, technical, and music), and box office performance (including both first weekend and later gross). Previous research not only shows that these success criteria may not always agree, but also that the criteria may have distinct aesthetic and economic antecedents. However, because the success criteria emerge at distinct points across time, a recursive model can be developed that describes the relationships among the criteria as well as their differential dependence on the predictive factors most frequently identified in the literature. The model was constructed using a sample of 1006 English-language, live-action, feature-length narrative films released between 2000 and 2006. The resulting equations indicate the complexity of cinematic success. None- theless, overriding this complexity is the fundamental contrast between film as art and film as enter- tainment. Keywords: cinema, creativity, critics, awards, economics “By their works ye shall know them,” or so goes a common paraphrase of Matthew 7:20. This certainly holds true of the creative arts. Just imagine how much our knowledge of Michelan- gelo Buonarroti would change without his frescoes in the Sistine Chapel. Or contemplate William Shakespeare without Hamlet, Ludwig van Beethoven without the Fifth Symphony, Leo Tolstoy without War and Peace, Martha Graham without Appalachian Spring, Akira Kurosawa without The Seven Samurai, or Frank Gehry without the Bilbao Guggenheim Museum? Furthermore, empirical research indicates that an artistic creator’s reputation is mostly dependent on the aesthetic merit of his or her greatest works (e.g., Simonton, 1977, 1991b; see also Simonton, 1991a). Indeed, in the case of the “one-hit wonders” an artist’s place in the eyes of posterity may be mostly predicated on a single work (Kozbelt, 2008). Remove the Canon in D from Johann Pachelbel’s corpus and what do you have left on which to hang his fame? The dependence of a creator’s reputation on his or her work then raises the issue of how to judge artistic products. This question is not easily answered because there may be multiple ways of as- sessing the value of a given work. This variety of assessment approaches may not be problematic if they agreed with each other, as appears to be the case in opera (Simonton, 1998). But rival measures sometimes fail to display a strong consensus, and at times may outright disagree. Making matters all the worse, alter- native forms of aesthetic judgment may bear complex relationships with each other, so that they are not fully independent. We may know artists by their works, but that knowledge has a complexity that is almost as hard to fathom as the works themselves. Perhaps nowhere are these complexities more conspicuous than in cinema. Clearly not all films are equally successful, but how exactly are we going to gauge the degree of success? Consider the following three possibilities: First, if we define success according to artistic merit, then we may decide to rely on the opinions of film critics whose job is to evaluate lots of movies (Boor, 1990, 1992; Simonton, 2004a; Zickar & Slaughter, 1999). Yet which critics will we use? For example, some critics pass judgments early in a film’s theatrical run, whereas others may wait much later after the film has come out in video or DVD (Simonton, 2005a). Whose assessment best reflects the film’s intrinsic value? And does it matter that the final critics might be strongly influenced by the earlier critics rather than represent totally independent assessments? Second, instead of trusting film critics we might place our faith in the various and numerous film awards and nominations that are presented each year by prestigious professional organizations (Ginsburgh, 2003; Simonton, 2004b). The most prominent in- stances are the annual Oscars bestowed by the Motion Picture Academy of Motion Picture Arts and Sciences. Still, Oscars are granted in many categories of cinematic achievement, not all of which have the same relevance for gauging aesthetic quality. How should an Oscar for best picture or best screenplay be weighed relative to an Oscar for best sound mixing or best makeup? Also, the Oscars and other awards may be affected by extraneous factors, such as halo and sympathy effects, that could bias the outcomes (Collins & Hand, 2006; Pardoe & Simonton, 2008). As an added complication, movie awards of various kinds may be influenced by initial critical judgments, and the awards also may themselves influence the judgments of the final critics. So the alternative measures may be causally interconnected. Does interdependence then have repercussions for an eventual appraisal? Correspondence concerning this article should be addressed to Dean Keith Simonton, Department of Psychology, One Shields Avenue, University of California, Davis, CA 95616-8686. E-mail: [email protected] Psychology of Aesthetics, Creativity, and the Arts © 2009 American Psychological Association 2009, Vol. 3, No. 3, 128 –138 1931-3896/09/$12.00 DOI: 10.1037/a0014521 128

Cinematic success, aesthetics, and economics: An exploratory recursive model

Embed Size (px)

Citation preview

Cinematic Success, Aesthetics, and Economics: An ExploratoryRecursive Model

Dean Keith SimontonUniversity of California, Davis

Although the reputation of creative artists is based largely on the merit of their work, the latter cansometimes be assessed in several different ways that may not necessarily agree. This lack of evaluativeconsensus is perhaps most apparent in cinematic success; this can be judged by film critics (initial andfinal), movie awards (picture, dramatic, visual, technical, and music), and box office performance(including both first weekend and later gross). Previous research not only shows that these successcriteria may not always agree, but also that the criteria may have distinct aesthetic and economicantecedents. However, because the success criteria emerge at distinct points across time, a recursivemodel can be developed that describes the relationships among the criteria as well as their differentialdependence on the predictive factors most frequently identified in the literature. The model wasconstructed using a sample of 1006 English-language, live-action, feature-length narrative films releasedbetween 2000 and 2006. The resulting equations indicate the complexity of cinematic success. None-theless, overriding this complexity is the fundamental contrast between film as art and film as enter-tainment.

Keywords: cinema, creativity, critics, awards, economics

“By their works ye shall know them,” or so goes a commonparaphrase of Matthew 7:20. This certainly holds true of thecreative arts. Just imagine how much our knowledge of Michelan-gelo Buonarroti would change without his frescoes in the SistineChapel. Or contemplate William Shakespeare without Hamlet,Ludwig van Beethoven without the Fifth Symphony, Leo Tolstoywithout War and Peace, Martha Graham without AppalachianSpring, Akira Kurosawa without The Seven Samurai, or FrankGehry without the Bilbao Guggenheim Museum? Furthermore,empirical research indicates that an artistic creator’s reputation ismostly dependent on the aesthetic merit of his or her greatestworks (e.g., Simonton, 1977, 1991b; see also Simonton, 1991a).Indeed, in the case of the “one-hit wonders” an artist’s place in theeyes of posterity may be mostly predicated on a single work(Kozbelt, 2008). Remove the Canon in D from Johann Pachelbel’scorpus and what do you have left on which to hang his fame?

The dependence of a creator’s reputation on his or her work thenraises the issue of how to judge artistic products. This question isnot easily answered because there may be multiple ways of as-sessing the value of a given work. This variety of assessmentapproaches may not be problematic if they agreed with each other,as appears to be the case in opera (Simonton, 1998). But rivalmeasures sometimes fail to display a strong consensus, and attimes may outright disagree. Making matters all the worse, alter-native forms of aesthetic judgment may bear complex relationshipswith each other, so that they are not fully independent. We mayknow artists by their works, but that knowledge has a complexitythat is almost as hard to fathom as the works themselves.

Perhaps nowhere are these complexities more conspicuous thanin cinema. Clearly not all films are equally successful, but howexactly are we going to gauge the degree of success? Consider thefollowing three possibilities:

First, if we define success according to artistic merit, then wemay decide to rely on the opinions of film critics whose job is toevaluate lots of movies (Boor, 1990, 1992; Simonton, 2004a;Zickar & Slaughter, 1999). Yet which critics will we use? Forexample, some critics pass judgments early in a film’s theatricalrun, whereas others may wait much later after the film has comeout in video or DVD (Simonton, 2005a). Whose assessment bestreflects the film’s intrinsic value? And does it matter that the finalcritics might be strongly influenced by the earlier critics ratherthan represent totally independent assessments?

Second, instead of trusting film critics we might place our faithin the various and numerous film awards and nominations that arepresented each year by prestigious professional organizations(Ginsburgh, 2003; Simonton, 2004b). The most prominent in-stances are the annual Oscars bestowed by the Motion PictureAcademy of Motion Picture Arts and Sciences. Still, Oscars aregranted in many categories of cinematic achievement, not all ofwhich have the same relevance for gauging aesthetic quality. Howshould an Oscar for best picture or best screenplay be weighedrelative to an Oscar for best sound mixing or best makeup? Also,the Oscars and other awards may be affected by extraneous factors,such as halo and sympathy effects, that could bias the outcomes(Collins & Hand, 2006; Pardoe & Simonton, 2008). As an addedcomplication, movie awards of various kinds may be influenced byinitial critical judgments, and the awards also may themselvesinfluence the judgments of the final critics. So the alternativemeasures may be causally interconnected. Does interdependencethen have repercussions for an eventual appraisal?

Correspondence concerning this article should be addressed to Dean KeithSimonton, Department of Psychology, One Shields Avenue, University ofCalifornia, Davis, CA 95616-8686. E-mail: [email protected]

Psychology of Aesthetics, Creativity, and the Arts © 2009 American Psychological Association2009, Vol. 3, No. 3, 128–138 1931-3896/09/$12.00 DOI: 10.1037/a0014521

128

Third, it is possible that the previous two choices betray asnobbish elitism (cf. Holbrook, 1999; Wanderer, 1970). In lieu ofasking film critics or the voting members of professional societies,it might be better to seek the answer in the moviegoers who buythe tickets to see the film. Rather than inquire whether a film is amasterpiece or a turkey, perhaps we should find out whether amovie was a blockbuster or a flop. In short, the focus should be onfilm as entertainment business rather than film as “art for art’ssake” (Simonton, 2005b). Is popular art any less art than elitist art?Yet even this alternative criterion raises issues. Should we look atbox office gross at the end of the theatrical run? Or will the returnson the first weekend of release provide a sufficient indicator? If thelatter suffices, then how much is that determined by the number ofscreens on which the film opens? Are wide-release films generallysuperior or inferior to those that do the art house circuit? Inaddition, how are box office indicators really related to criticalevaluations? Do initial critics influence financial performance ordoes that performance influence final critics? And how do movieawards fit in with all this? Are the Academy voters at all swayedby box office returns, or do they favor financial flops? And can anaward convert a financial loser into a winner?

Underlying all three sets of cinematic criteria is a copioussupply of various antecedents, some economic and others artistic.The most obvious antecedent is the film’s production cost orbudget (Basuroy, Chatterjee, & Ravid, 2003; Prag & Casavant,1994). Is it a big-budget or small-budget product? Not only doesthis factor have critical implications for many of the criteriaalready mentioned (Simonton, 2005a), but also is contingent on ahost of influences that may or may not have independent effects onthe alternative criteria of cinematic impact (Simonton, 2005b).These influences include the film’s running time, its genre, itsrelation with previous films as a sequel or remake, whether or notthe director was involved in writing the script, whether or not thescript was adapted from another source, such as a novel or play,whether or not the original author participated in that adaptation,and so forth.

Fortunately, a very large empirical literature has emerged on thedeterminants of a film’s success has judged by different criteria(for review, see Simonton, in press). In principle, this researchshould enable us to specify a model that describes the mostprobable direct and indirect effects of diverse variables on thecritical acclaim, movie awards, and box office as well as theeffects connecting these same three sets of criteria. In practice,however, the research published to date does not allow us toconstruct such a model. The studies vary so much with respect tosample and variable definitions that the findings are not fullycomparable (Simonton, in press). Indeed, sometimes the method-ological differences are so great that results fail to replicate fromone study to the next (Chang & Ki, 2005). Consequently, it isnecessary to conduct a single investigation that imposes uniformsampling and variable definitions throughout. That is the mainpurpose of the current investigation.

I begin by defining a set of core variables that will be placed inthe following temporal order: final critics (from four movie guideratings), later gross (domestic box office after the first weekend),movie awards (dramatic, visual, technical, music, and picturehonors from multiple sources), initial critics (from Metacritic.com), early gross (first weekend domestic box office), number of

screens (for first weekend showing), and production budget (orcost). By order I mean that final critics can be a function of one ormore variables later in the list, later gross a function of one or morevariables later in the list (after excluding final critics), and so on.That is, the ordering is recursive (Loehlin, 2004). I then define asecond set of variables that can serve as potential antecedents ofone or more of the preceding variables, including budget. Theseantecedents include season of release, timing, Motion PictureAssociation of America (MPAA) ratings, genre, and other at-tributes of the screenplay. The inventory of possible antecedents isbased on the variables that have most consistently emerged asimportant predictors in past research (Simonton, in press). Lastly,multiple regression analysis can be applied to discern the mostlikely relations among these variables (Cohen, Cohen, West, &Aiken,, 2003, chap. 12). What should emerge is a tentative systemof equations representing the first approximation to a recursivepath model (Sclove, 2007).

Method

Except when stated otherwise, all data came from the InternetMovie Database (http://www.imdb.com/) and Metacritic (http://www.metacritic.com/). The former was especially valuable giventhe comprehensive nature of the available information.

Sample

The sample consisted of 1006 live-action, English-languagefeature-length, narrative films that had a theatrical run in theUnited States between 2000 and 2006. Thus, documentaries, ani-mations, and shorts were omitted along with films in any languagebesides English (cf. Holbrook, 1999; Simonton, 2007b). Theseomissions rendered the films maximally comparable with regard tothe variables to be defined below. The beginning and endingrelease dates were also selected to have maximal access to certainessential variables. Furthermore, each film in the sample met atleast one of the following three criteria:

1. The film earned a nomination or award in a main cate-gory of cinematic accomplishment from one or more ofthe following sources: the Academy of Motion PictureArts and Sciences (“Oscars”), the Hollywood ForeignPress Association (“Golden Globes”), the British Acad-emy of Film and TV Arts (“BAFTAs”), the Los AngelesFilm Critics Association, the Chicago Film Critics Asso-ciation, the Broadcast Film Critics Association, or theOnline Film Critics Society. A “main category” signifiesrecognition for picture, direction, screenplay, lead maleand female actors, supporting male and female actors,film editing, cinematography, art direction, costume de-sign, makeup, visual effects, sound effects editing, soundmixing, score, and song. This criterion guarantees thatthe sample includes the most acclaimed films releasedeach year. Annually about 50 films are bestowed withsome special honor.

2. The film received a minimum of 20 reviews according tothe criticism compilation available at Metacritic.com.This criterion ensures that the film was seen by a sizable

129CINEMATIC SUCCESS

number of film critics, and, presumably, by a large num-ber of moviegoers. Although films covered by Metacrit-ic.com average range between 1 and 71 reviews, rela-tively few obtain more than three dozen. Additionally,almost all films that are honored with awards or nomi-nations receive 20 or more critical reviews. Hence, filmsthat meet this criterion but not the first criterion can beconsidered “also-rans” for a given year of releases.

3. The film was “dishonored” with a Golden Raspberryor “Razzie” for worst picture, director, male or femalelead actor, male or female supporting actor, screen-play, or song (Golden Raspberry Awards Foundation,http://www.razzies.com/). This criterion allows the in-clusion of films that occupy the bottom of the distributionin cinematic achievement. Previous research has shownthat films that get Razzie’s are pretty much antithetical tofilms that earn Oscars in the same categories (Simonton,2007c). That is, bad films are mirror images of goodfilms (cf. Baumeister, Bratslavsky, Finkenauer, & Vohs,2001).

Although the preceding three criteria did not use box-officereturns as a criterion, the sample still included all top-grossingfilms in a particular year (according to Craddock, 2008). Thatinclusion results mostly from the second criterion. Even if a filmmay earn not one award nomination and still bring in megabucksin the theaters, it is impossible for a blockbuster to attract reviewsfrom fewer than 20 film critics. Big money makers cannot be soignored by any print or electronic publication that features filmreviews.

Although 1469 films met at least one criterion, 463 films wereomitted because they were missing basic financial data, particu-larly budget information. A film’s production cost is often seen asproprietary information that is only publicized at the producer’sdiscretion, and it is often not released at all, leaving industryexperts to provide estimates (cf. Basuroy, Chatterjee, & Ravid,2003; Simonton, 2005a). Because such data are essential to as-sessing the process behind critical acclaim, films lacking these datawere dropped. Even so, the sample remains highly representativeof the films appearing in the theaters each year. First, the samplestill exhibits extreme variation on all of the key variables tobe examined in this investigation. Second, the sample remainsquite large, constituting a respectable proportion of the films thatmost consumers and critics would likely see. Indeed, a moviegoerwould have to watch almost three films per week to claim first-hand experience with the cases defining this sample.1

Measures

For convenience the variables can be grouped into two catego-ries: endogenous and exogenous. With one exception, endogenousvariables can serve as a both dependent and independent variablesin one or more equations, whereas exogenous variables can onlyserve as independent variables in any given equation. Expressedmore loosely, exogenous variables can only be causes, whereasendogenous variables can also be effects.

Endogenous variables. Here I will therefore define the vari-ables according to their most likely recursive ordering from themost endogenous to the most exogenous:

1. Final critics. Four movie-guides were used to assess theopinions of critics after a film was released in video or DVDformat (viz., Craddock, 2008; Maltin, 2007; Martin & Porter,2006; Walker, 2007). Where necessary the ratings were convertedto a 5-point scale where 1 � turkey or bomb and 5 � masterpiece.The average of the four scores was then computed, yielding acomposite score that was very close to being normally distributed.Not only was the mean near to the middle of the distribution, butthe scores ranged from 1 star to a bit less than 5 stars (M � 2.74,SD � 0.69, range � 1.00–4.84). The measure was also highlyreliable, particularly for a 4-item composite (� � .81). This con-stitutes the ultimate endogenous variable, the only one that intheory could be a function of all other variables defined in thisexploratory study.

2. Later gross. This variable includes all domestic grossearned after the first weekend expressed in millions of U.S. dollars(i.e., total gross minus first-weekend gross; M � 34.43, SD �45.13, range � 0–339). Because this variable tends to be ex-tremely skewed, the scores were subjected to a logarithmic trans-formation.2 The result more closely approximates a normal distri-bution (M � 2.68, SD � 1.56, range � �0.69–5.83).

3. Movie awards. The nominations and awards each filmacquired were used to define the following five measures ofspecial recognition: (a) picture honors (M � 0.03, SD � 0.14,range � 0–1.46; � � .87); (b) dramatic honors (direction, screen-play, lead male and female actors, supporting male and femaleactors, film editing; M � 0.03, SD � 0.11, range � 0–0.92; � �.82); (c) visual honors (cinematography, art direction, costumedesign, makeup; M � 0.03, SD � 0.12, range � 1.11; � � .80);(d) technical honors (visual effects, sound effects editing, soundmixing; M � 0.05, SD � 0.18, range � 0–1.50; � � .78); and (e)music honors (score, and song; M � 0.03, SD � 0.11, range �0–1.41; � � .48).3 The honors had been bestowed by the sevenorganizations used as sample criteria along with the more special-ized awards granted by the following: Producers Guild of America(picture), Directors Guild of America (direction), the WritersGuild of America (screenplay), the Screen Actors Guild (all fouracting categories), the American Cinema Editors (film editing), theAmerican Society of Cinematographers (cinematography), the So-ciety of Motion Picture and TV Art Directors (art direction), the

1 Preliminary analyses revealed that films missing budget data are notrandomly selected from the larger sample. On the contrary, they tend toreceive fewer and lower critical evaluations, earn few if any movie awardsor nominations, and garner little if anything in the box office. In short, byany success criterion the omitted cases represented the most marginal filmsin the available pool. So the only restriction on the substantive results isthat they apply to the films that film critics, industry professionals, andmoviegoers are most likely to have seen and appreciated – or deprecated asin the case of the Razzie winners.

2 Because of the extreme skew of box office criteria, alternative estima-tion procedures will sometimes yield slightly different parameter estimates(Walls, 2005a). However, to maintain consistency in interpretation I haveused the same estimators for all equations, namely ordinary least squares.

3 This reliability is lower than the others not only because the compositemeasure contains the fewest items (just two), but also because outstandingscores are often found in very different movies than outstanding songs(e.g., dramas versus musicals; see Simonton, 2007a, 2007b).

130 SIMONTON

Costume Designers Guild (costume), the Motion Picture SoundEditors (sound effects editing), and the Grammy Awards (scoreand song). Prior studies have shown that awards received fromthese more specialized guild organizations are correlated withthose bestowed by the more general professional and criticalsocieties on which the sample was based (Simonton, 2004b,2004c). For all five measures a nomination was assigned 1 pointand an award 2 points (see also Basuroy et al., 2003; Ginsburgh,2003; Simonton, 2002; Sochay, 1994). The dramatic, visual, tech-nical, and music clusters were based on an earlier factor analyticstudy (Simonton, 2004c) and further validated in follow-up studies(e.g., Simonton, 2005a, 2005b; cf. Simonton, 2007b).

4. Initial critics. The ratings posted at Metacritic.com, whichscore films on a 1–100 point scale; the mean of this variable wasvery close to the middle of the distribution and at the same time thescores cover nearly the complete range (i.e., M � 51.47, SD �17.40, range � 9–94). Its distribution also closely approximatesthe normal. These reviews normally appear very early in a film’stheatrical run, and almost always before the movie award seasonbegins.

5. Early gross. The first-weekend domestic gross expressed inmillions of U.S. dollars (M � 13.05, SD � 17.53, range � 0–136).Like later gross, this was log-transformed (see Footnote 1), yield-ing a measure closer to a normal distribution (M � 1.66, SD �1.61, range � �0.69–4.92).

6. Screens. The number of screens on which the film wasprojected during the opening weekend (M � 1791.39, SD �1287.13, range � 1–4152). This was also log-transformed toreduce skew, albeit in this instance the resulting distribution wasbimodal (M � 6.18, SD � 2.60, range � 0.00–8.33). Filmsreceive either wide release or an art-house distribution. It should beemphasized that the initial screen count indicates the success offilm marketing: A film only gets wide release on first weekend ifits promoters manage to convince distributors that they have theopportunity to capitalize on a potential blockbuster. Unlike anyother criterion, this factor is a judgment of potential rather thanactual success.

7. Budget. The estimated production (or “negative”) costsassessed in millions of U.S. dollars (M � 38.07, SD � 36.23,range � 0–270). Because this measure is also highly skewed, itwas subjected to a logarithmic transformation that renders itsdistribution more approximately normal (M � 3.13, SD � 1.19,range � �0.69–5.60). This variable is the most exogenous of theendogenous variables because it can have a direct or indirect on allof the preceding variables, but cannot be a function of thosevariables. Still, it can be a function of the predictor variablesdefined in the next section.

Exogenous variables. This set of predictors is very heteroge-neous (see Simonton, 2005b, for details) but can be grouped intothe following nine categories:

1. Release season. A series of zero-one dummy variables thatrecorded the season within the year in which the film was firstreleased: (a) Easter (March and April; M � 0.18, SD � 0.38,range � 0–1), (b) summer (May through August; M � 0.31, SD �0.46, range � 0–1), and Christmas (i.e., November and December;M � 0.19, SD � 0.39, range � 0–1). These are traditionally majorholidays or vacation periods in which more moviegoers tend to

flock to the theaters, and the Christmas season is the periodfavored for the release of films that have good prospects for movieawards (see also Chang & Ki, 2005; Sochay, 1994).

2. Timing. the Film’s run time in minutes (M � 108.46, SD �18.03, range � 71–231).

3. MPAA ratings. Zero-one dummy variables to record theratings of G (M � 0.01, SD � 0.10, range � 0–1), PG (M � 0.09,SD � 0.29, range � 0–1), PG-13 (M � 0.42, SD � 0.50, range �0–1), and R (M � 0.46, SD � 0.50, range � 0–1). NC-17 and NRwere too infrequent to justify the inclusion of correspondingdummy variables.

4. Genre. Zero-one dummy variables registered the genre cat-egories of drama (M � 0.58, SD � 0.49, range � 0–1), comedy(M � 0.44, SD � 0.50, range � 0–1), romance (M � 0.23, SD �0.42, range � 0–1), and musical (M � 0.02, SD � 0.18, range �0–1). These were selected because they are by far the mostcommon and have also been shown to influence cinematic out-comes (Simonton, in press). It should be noted that these are notmutually exclusive categories (e.g., a romantic comedy will haveunit values for both of the corresponding dummy variables). As aresult, they do not represent a pure nominal scale.

5. Real events. Zero-one dummy variables recorded whetherthe screenplay was based on a true story (M � 0.08, SD � 0.38,range � 0–1), including a biopic about a real person (M � 0.03,SD � 0.32, range � 0–1), where the latter is a subset of theformer. In particular, nearly half of true stories are biopics.

6. Successor film. Two zero-one dummy variables registeredwhether the film was a remake (M � 0.10, SD � 0.30, range �0–1) or sequel (M � 0.08, SD � 0.27, range � 0–1).

7. Writer-director. A rank-category variable noted whether thedirector was directly involved in writing the screenplay. In partic-ular, writer-director � 0 if the director received no credit for thescreenplay, � 1 if he or she received partial credit, and � 2 if heor she received full credit (M � 0.61, SD � 0.83, range � 1–2).

8. Adaptation. A final set of zero-one dummy variables re-corded whether the script was adapted from a play (M � 0.04,SD � 0.19, range � 0–1), novel (M � 0.20, SD � 0.40, range �0–1), short story (M � 0.03, SD � 0.16, range � 0–1), andnonfiction (M � 0.04, SD � 0.19, range � 0–1), or some othersource (M � 0.05, SD � 0.22, range � 0–1) as well as whether theoriginal source was a literary classic (M � 0.04, SD � 0.19,range � 0–1) or best seller (M � 0.03, SD � 0.16, range � 0–1),a Broadway production (M � 0.004, SD � 0.063, range � 0–1),a prize winner (M � 0.01, SD � 0.11, range � 0–1) or a financialhit (M � 0.001, SD � 0.032, range � 0–1). Additionally, arank-category variable registered whether the original author wasdirectly involved in the adaptation, where author-adaptor � 0 if theauthor received no credit for the screenplay, � 1 if he or shereceived partial credit, and � 2 if he or she received full credit(M � 0.05, SD � 0.30, range � 1–2).

9. Release year. A series of zero-one dummy variables thatrecorded the year in which the film had its first US release. In moredetail, a dummy variable was generated for the years 2001 (M �0.14, SD � 0.35, range � 0–1), 2002 (M � 0.17, SD � 0.37,range � 0–1), 2003 (M � 0.14, SD � 0.34, range � 0–1), 2004(M � 0.13, SD � 0.34, range � 0–1), 2005 (M � 0.15, SD � 0.35,range � 0–1), and 2006 (M � 0.13, SD � 0.34, range � 0–1),with the year 2000 left as the baseline defining the intercept. As the

131CINEMATIC SUCCESS

means (proportions) suggest, the films are distributed fairly evenlyacross the consecutive years. The counts range from 131 (year2006) to 169 (year 2002). These dummy variables have no sub-stantive interest but rather are included to control for year-to-yearfluctuations in the central endogenous and exogenous variables.For example, these variables effectively adjust for inflation insofaras the latter affect the financial data.

Results

The overall strategy is to construct a recursive model by con-ducting a series of ordinary least-squares multiple regression anal-yses (Sclove, 2007; see, e.g., Simonton, 1977).4 I begin with theultimate endogenous variable, namely, the final critic evaluations.This criterion can be regressed on the other endogenous variablesand all exogenous variables. The nonsignificant predictors will beprogressively deleted using a backward stepwise procedure. Animportant note: because of the large sample size, “significance” isboth substantive and statistical. Almost all effects that are signif-icant at the .05 level or better will also feature standardized partialregression coefficients that are around .05 in absolute value, mean-ing that increasing the predictor by one standard deviation willincrease the criterion by 1/20th of a standard deviation. The latteris comparable to what is called a “medium effect” in meta-analyses(cf. Hunter & Schmidt, 1990). Needless to say, some of the effectswill be much larger than this.

Once the substantively and statistically significant predictors areidentified, I will then attempt to determine the antecedents of theendogenous variables that emerged as predictors for a given en-dogenous variable. The end result will be a series of regressionequations with the following as dependent variables: final criticevaluations (movie guide ratings), later (post first weekend) grossdomestic earnings, movie awards (picture, dramatic, visual, tech-nical, and music), early (first-weekend) gross domestic earnings,initial critic evaluations (Metacritic), number of screens on the filmopened its first weekend, and the production budget. It is possibleto end up with more than seven equations because movie awardsconsist of five distinct measures, albeit not all may emerge aspredictors of either final critics or later gross. In fact, as will beseen shortly, only two out of five awards measures entered therecursive model, yielding a total of eight equations.5

Final Critics: Movie Guide Ratings

When the film ratings of the final critics are regressed on allpotentially antecedent variables, and the nonsignificant predictorsare deleted, one obtains the results shown in the first two numericalcolumns of Table 1. The first of these columns gives the unstand-ardized partial regression coefficients (b) that express the relationin the original variable units while the second gives the standard-ized partial regression coefficients (�) specifying the relation withregard to z-scores. The latter provide the path coefficients for arecursive model.

Clearly the single most potent predictor is the initial criticalevaluations, indicating that critical judgments are highly stableacross time. It is largely for this reason that the equation accountsfor 70% of the variance in the criterion. Still, the final criticalassessments are also a positive function of dramatic and technical

honors, timing, a PG rating, the drama genre, and bestseller adap-tation, but a negative function of screens (and release in 2006). Inother words, movie-guide ratings are highest for films that (a)received awards and nominations in the categories of direction,writing, acting, film editing, visual effects, sound effects editing,and sound mixing, (b) were highly praised by critics during thefilm’s theatrical run, (c) were longer than average, (d) were ac-cessible to a large audience and yet deal with serious drama, (e)were adapted from a bestselling novel or work of nonfiction, and(f) opened on a small number of screens (e.g., did the “art-housecircuit”).

Of significance, these postrun appraisals are not influenced bythe film’s box office, whether early or late, nor even the productionbudget. The absence of later gross as an antecedent is especiallytelling. It signifies that final critics and later gross represent twolargely independent but final judgments of cinematic success, oneaesthetic and the other commercial (see also Simonton, 2005a). Infact, the two criteria have a small zero-order correlation (r � .09,p � .01) that vanishes to zero once their respective causes arepartialed out.

4 In the case of fully recursive models, equation-by-equation parameterestimation yields results virtually indistinguishable from estimation strat-egies in which all parameters are estimated at once using generalized leastsquares or maximum likelihood estimators. At the same time this approachhas the advantage that it provides statistics that are accessible to anyonewith training in multiple regression.

5 Given that the total number of defined endogenous and exogenousvariables is very large – nearly a thousand coefficients – it is not possibleto publish the complete correlation matrix. However, I would be glad tosupply such a matrix to any interested researcher.

Table 1Multiple Regression Analysis: Unstandardized (b) andStandardized (�) Partial Coefficients

Final critics Later gross

Predictor b � b �

Dramatic honors 0.617 .09��� 3.279 .22���

Technical honors 0.194 .05�

Initial critics 0.027 .67��� 0.018 .20���

Early gross 0.786 .81���

Screens �0.024 �.09��� �0.081 �.13���

Christmas release 0.332 .08���

Budget 0.238 .18���

Timing 0.004 .10���

MPAA rating: PG 0.138 .06�� 0.241 .04��

MPAA rating: R �0.205 �.07��

Genre: Drama 0.070 .05�

Genre: Comedy 0.216 .07���

Adaptation: Bestseller 0.205 .05��

2003 release �0.152 �.03�

2005 release �0.193 �.04��

2006 release �0.103 �.05�� �0.267 �.06���

Note. For the first dependent variable, intercept � 1.018 and R2 � .70( p � .001; adjusted-R2 � .69); for the second, intercept � 0.117 and R2 �.77 ( p � .001; adjusted-R2 � .77).� p � .05. �� p � .01. ��� p � .001.

132 SIMONTON

Later Gross: Total Domestic Box Office SinceFirst Weekend

The last two columns of Table 1 provide the regression resultsfor the gross domestic earnings since the first weekend. Thiscriterion shares only a few predictors in common with the previousone. Films that make a lot of money later in their theatrical run aremore prone to receive dramatic honors and earn high praise fromthe initial critics, to open on a small number of screens, and to havea PG rating (but again not be released in 2006). But dramatichonors have even more predictive power for this criterion, whereasinitial critical reviews have much less predictive power. Even morefascinating is the conspicuous impact of early gross earnings. Justas initial critic assessments is the best predictor of final criticassessments, so is early gross (first weekend only) the best pre-dictor of later gross (remainder of domestic run). Like predictslike; given that the standardized partial regression coefficient is.81, this temporal consistency effect largely explains why thisequation accounts for an impressive 77% of the variance in thecriterion.

In any case, although later gross loses some of the variables inthe equation for final critic assessments—namely, technical hon-ors, timing, the drama genre, and bestseller adaptation—this suc-cess criterion gains more predictors than it loses. Films that reg-ister big in the box office after the first weekend are less likely tobe R rated but are more likely to be comedies, to be released duringthe Christmas season, and to have big budgets (and to be releasedin the years 2003, 2005, and 2006). Such films provide yourstandard holiday faire.

Perhaps the only oddity in this equation is the negative effect ofwide release. Yet this negative association becomes less surprisingwhen we remember that the effect is estimated after controlling forearly or first-weekend gross. Although the zero-order correlationbetween later gross and screen count is .62 ( p � .001), the numberof screens and first-weekend earnings are correlated to an evenhigher degree, namely, .90 ( p � .001).6 So once the sharedvariance between early gross and screens is removed, the screencount ends up with a negative relation with later gross. Thisimplies that the number of screens can be too high for the potentialaudience, allowing moviegoers to see the film who might other-wise have to wait until the film shows up in their local theater (seealso Radas & Shugan, 1998). Given that the equation for finalcritic judgments does not include early gross, the negative effectfor the two criteria does not presume the same underlying process.

Movie Awards: Dramatic and Technical Honors

Judging from Table 1, most of the movie award measures had nopredictive relevance. The only exceptions are the dramatic andtechnical honors. This more restricted outcome permits us toconfine the analysis to just these two measures. Even best picturenominations and awards can be ignored.

Table 2 shows the two multiple regression equations that resultafter eliminating the nonsignificant predictors. A few of the effectsreplicate across both sets of movie awards: Honors in either caseare positively associated with initial critics’ ratings, with runningtime, and with Broadway adaptations, but negatively with adapta-tions of former hits of any kind. Other effects have the signsreversed for the two award criteria: (a) early gross has a negative

relation with dramatic honors but positive with technical honors;and (b) adaptations from the classics have a negative relation withthe former but positive with the latter. Lastly, there are severalinstances where a predictor only applies to one of the two awardcriteria. On the one hand, dramatic honors are a positive functionof budget and Christmas release but a negative function of authoradaptations and writer-directors, and summer release. On the otherhand, technical honors are a negative function of an R MPAArating, the drama genre, and adaptation from a prizewinning work.On the whole, then, films that receive nominations and awards inthe categories of direction, writing, acting, and editing are ratherdistinct from those that earn honors for special visual effects,sound effects editing, and sound mixing (see also Simonton,2005b). Curiously, notwithstanding the differences in the equa-tions, the two sets of movie awards can be predicted to the samedegree. In each case 30% of the variance is explained.

Because initial critics and early gross entered both predictionequations in Table 2 (even if not in an identical fashion), it nowbecome imperative to turn to these two variables.

6 This phenomenon is an example of statistical suppression (Maassen &Bakker, 2001). Because of correlations among the independent variables,sometimes the standardized partial regression coefficient either (a) is largerin absolute value in comparison to the zero-order correlation coefficient or(b) has a sign opposite to that of the zero-order correlation. Suppressioneffects are often a repercussion of a high degree of collinearity, such as thatseen here. Nonetheless, because of the large sample size, the high corre-lation between early gross and screens has minimal impact on the results(i.e., the variance inflation factor declines as a function of N). Moreover,the correlation notwithstanding, these two variables are substantively dis-tinct. It would make no sense to combine them into a single measure, andthey do not have identical antecedents (compare Tables 3 and 4).

Table 2Multiple Regression Analysis: Unstandardized (b) andStandardized (�) Partial Coefficients

Dramatic honors Technical honors

Predictor b � b �

Initial critics 0.002 .39��� 0.002 .22���

Early gross �0.012 �.18��� 0.014 .13���

Summer release �0.016 �.07�

Christmas release 0.017 .06�

Timing 0.001 .16��� 0.004 .38���

MPAA rating: R �0.027 �.08��

Genre: Drama �0.026 �.07�

Budget 0.008 .09�

Adaptation: Classic �0.033 �.06� 0.103 .11���

Adaptation: Hit �0.305 �.09�� �0.472 �.09��

Adaptation: Broadway 0.220 .13��� 0.423 .15���

Adaptation: Prizewinner �0.108 �.07�

Author adaptation �0.025 �.07�

Writer-director �0.012 �.10��

Note. For the first dependent variable, intercept � �0.186 and R2 � .30( p � .001; adjusted-R2 � .29); for the second, intercept � �0.459 andR2 � .30 ( p � .001; adjusted-R2 � .30).� p � .05. �� p � .01. ��� p � .001.

133CINEMATIC SUCCESS

Initial Critics: Metacritic Evaluations

Adopting the same procedure as in the preceding sections, wearrive at the results displayed in the first two numerical columns ofTable 3. Film critics who evaluate films during their theatrical runare positively swayed by early gross, Christmas release, run time,the drama genre, adaptations of prizewinners, novels and nonfic-tion works, and writer-directors, but are negatively influenced bysuperwide release, romances and musicals, big budgets, and se-quels. Apparently, the early reviewers prefer film as art rather thanfilm as entertainment (cf. Holbrook, 1999; Wanderer, 1970).Around 38% of the variance in their judgments might be attributedto this preference. Plainly, these critics favor more serious, small-budget films that still manage to make money despite opening onrelatively few screens.

Now we need to compare these statistics with those in the lasttwo columns of Table 3.

Early Gross: First-Weekend Box Office

Although early gross returns has a strong relation with initialcritic verdicts, its predictors are far from identical. To be sure, thetwo criteria share an appreciation for films with long running timesand a dislike for romances. But the two criteria have oppositerelations with the number of screens, big budgets, sequels, andnovel adaptations. Moreover, box office performance in the firstweekend is associated with predictors that bear no connection withinitial critical judgments. Such initial returns are a positive func-tion of summer release (and 2003 release), but a negative functionof an R MPAA rating and scripts based on true stories. Good firstweekend earnings are allotted to good cinematic entertainment.

The most striking contrast in Table 3 may be the differentialrepercussion of the number of screens on which a film opens:highly negative for initial critics but highly positive for earlygross—yet for both criteria these exhibit the largest effect sizes.Nevertheless, only for early gross is the number of screens sub-stantially more impressive than the other effects. As a result, forthis criterion we can say that the opening screen count constitutesthe chief reason why 84% of the variance is explained!

Number of Screens on First Weekend Release

The observation that concludes the last section compels us toinspect the first two numerical columns of Table 4. It shouldbecome immediately apparent that a single predictor dominatesthis equation as well: production budget. The more money aproducer spends on the film the more he or she wants to guaranteea substantial return. This need is especially urgent given that themajority of films lose money in the box office (De Vany, 2004).What better way to endure handsome returns that to have themovie open in superwide release. Even if the film gets bad re-views, it still might turn a profit if enough of the curious show upon the first weekend. In light of this necessity, the remainingpredictors may provide clues as to what attracts numerous mov-iegoers to the first weekend debut. It turns out that only onepredictor besides budget has a positive implication, namely, theromance genre. All of the remaining predictors have negativeimplications: summer or Christmas release, the film’s runtime, anR rating, the drama or musical genre, almost any adaptation(whether from a classic, a novel, nonfiction work, or a miscella-neous source such as a comic), an author adaptation, and a writer-director—all negative! Put together, these positive and negativepredictors account for 58% of the variance in the number ofscreens at the film’s opening.

It may seem strange that release during the summer months orthe winter holidays is negatively associated with screen count.After all, these are the most popular times for releasing films. Butthat very popularity has an adverse consequence. Because thenumber of available screens stays fairly constant during the year,the only way to accommodate a surplus of films is to have themopen on fewer screens, on the average. It is noteworthy that thenegative effect is over twice as strong for Christmas release. Thisperiod is prime time for showing films that have a good shot at themovie awards. That would enhance this competitive effect all themore.

Budget: Estimated Production Cost

By now we are more than ready to discern the predictors ofbudget—a variable that predicts post first weekend gross, dramaticawards and nominations, initial critical opinion, first weekendgross, and the number of screens on opening weekend. The out-come of the same methodology is exhibited in the last two columnsof Table 4. Because half of the variance is accounted for, this is notby any means an elusive variable. The largest effect is that ofruntime: Longer films cost more to make. Remakes and sequelsalso cost more, as do adaptations from novels and miscellaneoussources, as well as films released during the Christmas season.These conditions put more pressure on the producer to create a

Table 3Multiple Regression Analysis: Unstandardized (b) andStandardized (�) Partial Coefficients

Initial critics Early gross

Predictor b � b �

Early gross 6.148 .57���

Screens �5.028 �.75��� 0.478 .77���

Summer release 0.238 .07���

Christmas release 2.413 .05�

Timing 0.245 .25��� 0.006 .06���

MPAA rating: R �0.131 �.04��

Genre: Drama 4.119 .12���

Genre: Romance �2.693 �.06� �0.178 �.05���

Genre: Musical �6.630 �.07��

Budget �1.944 �.13�� 0.173 .13���

Sequel �3.504 �.05� 0.552 .09���

True story �0.178 �.04��

Adaptation: Prizewinner 10.076 .06�

Adaptation: Novel 4.371 .10��� �0.123 �.03�

Adaptation: Nonfiction 6.865 .07��

Writer-director 2.237 .11���

2003 release 0.136 .03�

Note. For the first dependent variable, intercept � 47.385 and R2 � .38( p � .001; adjusted-R2 � .37); for the second, intercept � �2.448 andR2 � .84 ( p � .001; adjusted-R2 � .84).� p � .05. �� p � .01. ��� p � .001.

134 SIMONTON

truly superior product, one better than previous films and evenbetter than the book!

The biggest negative effect is that of writer-directors. Othernegative associations are found for R-rated films, dramas, come-dies, or musicals, adaptations from the classics, and author adap-tations. Ironically, it seems that more serious cinema costs less tomake than lighter products. It is cheaper to make art than toproduce entertainment.

Finally, despite the strong correspondence between budget andscreens, their prediction equations are noticeably different. Evenwhen they include the same predictor, the effect is often of acontrary sign. To be specific, opposing effects are seen for timing,novel adaptations, miscellaneous adaptations, and Christmas re-lease.

Discussion

Current Contributions

This empirical investigation constitutes a first attempt at arecursive model that connects the major criteria of cinematicsuccess with their principal aesthetic and economic antecedents(Simonton, in press). The model’s equations were all estimatedusing the same large sample (viz., N � 1006), a sample thatrepresents the full range of values on the central success criteria(e.g., from masterpieces to turkeys, from blockbusters to flops,from art-circuit to superwide release, from low-budget to big-budget films). Moreover, the amount of variance accounted for ineach endogenous variable ranges from .30 (for movie awards in thedramatic and technical categories) to .84 (for early domesticgross). Lastly, the equations contained all of the most important

variables to emerge from past empirical research on cinematicsuccess (Simonton, in press). Taken together, these assets addsome credence to the resulting equations, however exploratory ortentative. As a consequence, it is possible to draw two generalconclusions about this creative and aesthetic phenomenon:

First, cinematic products are decidedly segregated into works ofart and works of entertainment (Holbrook, 1999; Simonton,2005b). The two most endogenous success criteria—final filmcritics (movie guide ratings) and later domestic gross (after firstweekend) – are not only independent of each other, but also havedivergent antecedents. Most conspicuously, the single most potentpredictor of final critic evaluations is the initial critic evaluations(during the theatrical run) while the single most powerful predictorof later gross is the early gross (during the first weekend). Al-though the initial critic assessments are positively influenced byearly gross, the two preliminary indicators of cinematic successalso have conflicting predictors. Thus, whereas the initial criticsrespond negatively to big-budget films that open on a large numberof screens—particularly if the films are sequels—the first weekendbox office receipts respond positively to the exactly same factors.

Second, and in line with the last statement, no predictor enjoysconsistently positive effects on all criteria variables. The predictorthat comes closest such broad utility is the film’s running time.Long-running films are positively favored by final film critics,dramatic and technical awards and nominations, initial film critics,and early domestic gross. Even so, timing bears no connectionwith later gross and a negative relation with the number of screenson opening weekend.

Other predictors come and go with even greater latitude, andwith equally capricious changes in sign. This predictor unpredict-ability is perhaps most prominent for the adapted script variables:(a) final critics more highly rate bestseller adaptations and nothingelse; (b) later gross ignores adaptation characteristics of any kind;(c) dramatic honors (direction, screenplay, acting, and editing) goto adaptations of Broadway shows but not to adaptations of clas-sics, hits, and adaptations by the original author; (d) technicalhonors (visual effects, sound effects editing, and sound mixing) goto adaptations of classics and Broadway shows but not to adapta-tions of hits and prizewinners; (e) initial critics favor adaptationsof prizewinners, novels, and nonfiction works; (f) early gross isnegatively associated with adaptations of prizewinners; (g) numberof screens is negatively predicted by adaptations of classics, nov-els, nonfiction, and miscellaneous sources, as well as by authoradaptations; and (h) production budget is positively associatedwith novel and miscellaneous adaptations but negatively associ-ated with classic adaptations and adaptations by the originalauthor.

Third, because of these intricacies, cinematic success is noteasily predicted. Suppose, for example, that one wanted calculatethe impact of writer-directors on final critic evaluations. Althoughthis variable is missing from Table 1, it does show up in Tables 2,3, and 4. The results reported in the latter tables indicate thatwriter-directors have predictive relevance for some of the predic-tors of the two final critic assessments. Some of these indirecteffects are positive and others negative. The resulting mediatedinfluences can end up either positive or negative. Thus, on the onehand, final critics are positively influenced by dramatic honorswhich are negatively influenced by writer-directors, yielding a

Table 4Multiple Regression Analysis: Unstandardized (b) andStandardized (�) Partial Coefficients

Screens Budget

Predictor b � b �

Summer release �0.369 �.07��

Christmas release �1.126 �.17��� 0.173 .06�

Timing �0.014 �.10��� 0.029 .44���

MPAA rating: R �0.484 �.09��� �0.567 �.24���

Genre: Drama �0.872 �.17��� �0.634 �.26���

Genre: Comedy �0.138 �.05�

Genre: Romance 0.396 .06��

Genre: Musical �0.907 �.06�� �0.594 �.09���

Budget 1.431 .65���

Remake 0.300 .08��

Sequel 0.409 .09���

Adaptation: Classic �1.011 �.07��� �0.315 �.05�

Adaptation: Novel �0.415 �.06�� 0.175 .06�

Adaptation: Nonfiction �1.066 �.08���

Adaptation: Miscellaneous �0.721 �.06�� 0.494 .09���

Author adaptation �0.425 �.05� �0.212 �.05�

Writer-director �0.233 �.07��� �0.356 �.25���

Note. For the first dependent variable, intercept � 4.583 and R2 � .58( p � .001; adjusted-R2 � .58); for the second, intercept � 0.756 and R2 �.50 ( p � .001; adjusted-R2 � .49).� p � .05. �� p � .01. ��� p � .001.

135CINEMATIC SUCCESS

negative indirect effect. On the other hand, final critic evaluationsare a negative function of the number of screens which is a positivefunction of budget which is a negative function of writer-directors,producing a positive indirect effect. Even though the overall effecthappens to be positive, it is not obvious that this would be the casejust looking at the equations involving writer-director as an inde-pendent variable.

Future Directions

I would be the first to admit that the current study can only beconsidered exploratory—hence the subtitle. Although it has builtupon an extensive body of published research (Simonton, in press),it also has taken that research into novel directions. Accordingly,it is my hope that this preliminary investigation will inspireadditional researchers to investigate cinematic success withmore thoroughness than has been the case in the past. Besidesadopting the measures used here— especially a wide-rangingsample and broad variable inventory—future research can in-troduce several necessary improvements. The following twoenhancements stand out:

First, it is clear that many additional variables should be tested forinclusion in the appropriate equations. Among the missing variables,probably the most noticeable is the absence of “star power” (Canter-bery & Marvasti, 2001; Desai & Basuroy, 2005; Wallace, Seigerman,& Holbrook, 1993). Does it make a difference whether the marqueefeatures A-list headliners? Although this factor might be partly incor-porated in the dramatic honors variable, it is possible to have starpresence in the absence of corresponding acting awards (e.g., Harri-son Ford, who has never won a major honor). Other variables thatmight be usefully added to the system of equations include: (a) honorsreceived at film festivals (Ginsburgh & Weyers, 1999); (b) actualconsumer film evaluations based on moviegoer surveys and Niesenratings of TV broadcasts (Holbrook, 1999; Taylor, 1974), (c) content-analytical attributes of the screenplay (Beckwith, 2007; Eliashberg,Hui, & Zhang, 2006); (d) prior experience, previous collaborations,and team quality of the cast and the core crew (Cattani & Ferriani,2008; Ferriani, Corrado, & Boschetti, 2005); (e) expenditureson marketing and promotion (Prag & Casavant, 1994); (f)distribution by a major (Warner, Universal, Columbia, Para-mount, Twentieth-Century Fox, etc.; Chang & Ki, 2005; Litman& Kohl, 1989); and (g) market competition at time of release(Litman & Ahn, 1998; Sochay, 1994). As seen already in thisinvestigation, it is highly likely that these added predictors willhave inconsistent effects on the several criteria of cinematicsuccess. These inconsistencies will partially reflect the persis-tent contrast of film as art versus film as business.7

Second, the inventory of variables eventually must be subjected toa more sophisticated analysis than was implemented here. The sim-plicity of the present model is apparent in the presumed temporalordering of the endogenous variables: final critics, later gross, movieawards, initial critics, early gross, screens, and budget. Although thisrepresents the most plausible ordering for the vast majority of films,there are occasional exceptions. Films distributed in superwide releasewill often preview the film for major critics to provide advancepromotion (unless the film is thought to be a dud, in which casethe critics will be denied the opportunity to forewarn the public). Inthese cases many critic evaluations will come before or during the first

weekend rather than after, as hypothesized in the model. Anotherproblematic case is when a film completes its theatrical run early inthe year so that its final gross cannot receive a boost during the movieaward season. Although such films are most often markedly inferiorto those released later in the year, and therefore are less likely to evenreceive an award nomination, exceptions are possible (cf. Nelson,Donihue, Waldman, & Wheaton, 2001). However rare these excep-tions may be, future research should try to accommodate the fullcomplexity of the phenomenon. Possible solutions include (a) intro-ducing multiplicative terms to handle interaction effects or moderatedrelationships (e.g., Basuroy et al., 2003; Simonton, 2004c) and (b)breaking the theatrical run into more finely grained weekly data (e.g.,Deuchert, Adjamah, & Pauly, 2005; Krider, Li, Liu, & Weinberg,2005). The full implementation of these intricacies would probablyalso require the application of more complex analyses, such as mul-tilevel modeling (see Silvia, 2007).8

Regarding cinematic success, the Oscar-winning screenwriter Wil-liam Goldman once remarked “Nobody knows anything” (Walls,2005b, p. 177). With more research along the lines suggested, thisobservation should more increasingly become “Somebody knowssomething!”

7 An anonymous reviewer raised the interesting question of whether thecontrast between the two types of film reflects the divergent types ofmeasurement. On the one hand, film as art is assessed via critic evaluationsand movie awards, both subjective judgments. On the other hand, film asbusiness is gauged by box office performance, as defined by a set ofobjective measures. To address this issue properly would require theexistence of subjective consumer assessments taken sometime during afilm’s theatrical release. Presumably, these measures would predict boxoffice indicators far better than critic evaluations, despite being subjectiverather than objective. Unfortunately, consumer surveys are invariablyconducted long after the theatrical run, most often when broadcast ontelevision or released on DVD (Holbrook, 1999; Taylor, 1974). As a result,these evaluations would already be contaminated by input from bothsubjective and objective assessments (e.g., movie awards and box office).

8 A different anonymous reviewer inquired why structural equationmodeling (SEM) was not used in the current study. The answer is twofold(cf. Loehlin, 2004). First, SEM is most appropriate for confirmatoryanalyses in which a hypothesized causal model has strong justification inprior theoretical and/or empirical research. This inquiry was clearly ex-ploratory. Second, SEM only yields highly discrepant results from multipleregression when (a) the structural model is nonrecursive and (b) a mea-surement model must be incorporated to accommodate measurement errors(by means of latent variables; e.g., Simonton, 1991c). But the recursivemodel has a reasonable justification in the temporal ordering of the keyvariables, and only three variables in the equations are composite indicatorsthat can be specified as latent variables (viz. final critics, dramatic honors,and technical honors). Besides, the main consequence of using multipleregression analyses instead of SEM is that the amount of variance ex-plained (the R2s) will be underestimated. In that sense, the results areconservative.

References

Basuroy, S., Chatterjee, S., & Ravid, S. A. (2003). How critical are criticalreviews? The box office effects of film critics, star power, and budgets.Journal of Marketing, 67, 103–117.

Baumeister, R. R., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001).Bad is stronger than good. Review of General Psychology, 5, 323–370.

Beckwith, D. C. (2007). Personal values of protagonists in best pictures

136 SIMONTON

and blockbusters: 1996–2005. Unpublished doctoral dissertation, Say-brook Graduate School and Research Center.

Boor, M. (1990). Reliability of ratings of movies by professional moviecritics. Psychological Reports, 67, 243–257.

Boor, M. (1992). Relationships among ratings of motion pictures byviewers and six professional movie critics. Psychological Reports, 70,1011–1021.

Canterbery, E. R., & Marvasti, A. (2001). The U.S. motion picturesindustry: An empirical approach. Review of Industrial Organization, 19,81–98.

Cattani, G., & Ferriani, S. (2008). A core/periphery perspective on indi-vidual creative performance: Social networks and cinematic achieve-ments in the Hollywood film industry. Organization Science, 19, 824–844.

Chang, B.-H., & Ki, E.-J. (2005). Devising a practical model for predictingtheatrical movie success: Focusing on the experience good property.Journal of Media Economics, 18, 247–269.

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied MultipleRegression/Correlation Analysis for the Behavioral Sciences (3nd ed.).Mahwah, NJ: Erlbaum.

Collins, A., & Hand, C. (2006). Vote clustering in tournaments: What canOscar tell us? Creativity Research Journal, 18, 427–434.

Craddock, J. (Ed.). (2008). VideoHound’s golden movie retriever 2008.New York: Gale.

Desai, K. K., & Basuroy, S. (2005). Interactive influence of genre famil-iarity, star power, and critics’ reviews in the cultural goods industry: Thecase of motion pictures. Psychology and Marketing, 22, 203–223.

Deuchert, E., Adjamah, K., & Pauly, F. (2005). For Oscar glory or Oscarmoney? Academy Awards and movie success. Journal of CulturalEconomics, 29, 159–176.

De Vany, A. (2004). Hollywood economics. London and New York:Routledge.

Eliashberg, J., Hui, S. K., & Zhang, Z. J. (2006). From storyline to boxoffice: A new approach for green-lighting movie scripts. Working paper,The Wharton School, University of Pennsylvania.

Ferriani, S., Corrado, R., & Boschetti, C. (2005). Organizational learningunder organizational impermanence: Collaborative ties in film projectfirms. Journal of Management and Governance, 9, 257–285.

Ginsburgh, V. (2003). Awards, success and aesthetic quality in the arts.Journal of Economic Perspectives, 17, 99–111.

Ginsburgh, V., & Weyers, S. A. (1999). On the perceived quality ofmovies. Journal of Cultural Economics, 23, 269–283.

Golden Raspberry Awards Foundation. Accessed from http://www.metacritic.com/

Holbrook, M. B. (1999). Popular appeal versus expert judgments of motionpictures. Journal of Consumer Research, 26, 144–155.

Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Cor-recting error and bias in research findings. Newbury Park, CA: Sage.

Internet Movie Database (n.d.). Accessed from http://www.imdb.com/Kozbelt, A. (2008). One-hit wonders in classical music: Evidence and

(partial) explanations for an early career peak. Creativity ResearchJournal, 20, 179–195.

Krider, R. E., Li, T., Liu, Y., & Weinberg, C. B. (2005). The lead-lagpuzzle of demand and distribution: A graphical method applied tomovies. Marketing Science, 24, 635–645.

Litman, B. R., & Ahn, H. (1998). Predicting financial success of motionpictures: The early ‘90s experience. In B. R. Litman (Ed.), The MotionPicture Mega-Industry (pp. 172–197). Boston: Allyn & Bacon.

Litman, B. R., & Kohl, L. S. (1989). Predicting financial success of motionpictures: The ‘80s experience. Journal of Media Economics, 2, 35–50.

Loehlin, J. C. (2004). Latent variable models: An introduction to factor,path, and structural analysis (4th ed.). Hillsdale, NJ: Erlbaum.

Maassen, G. H., & Bakker, A. B. (2001). Suppressor variables in pathmodels. Sociological Methods & Research, 30, 241–270.

Maltin, L. (Ed.). (2007). Leonard Maltin’s 2008 movie & video guide. NewYork: Signet.

Martin, M., & Porter, M. (2006). Video movie guide 2007. New York:Ballantine Books.

Metacritic (n.d.). Accessed from http://www.metacritic.com/Nelson, R. A., Donihue, M. R., Waldman, D. M., & Wheaton, C. (2001).

What’s an Oscar worth? Economic Inquiry, 39, 1–16.Pardoe, I., & Simonton, D. K. (2008). Applying discrete choice models to

predict Academy Award winners. Journal of the Royal Statistical Soci-ety: Series A (Statistics in Society), 171, 375–394.

Prag, J., & Casavant, J. (1994). An empirical study of the determinants ofrevenues and marketing expenditures in the motion picture industry.Journal of Cultural Economics, 18, 217–235.

Radas, S., & Shugan, S. M. (1998). Seasonal marketing and timing newproduct introductions. Journal of Marketing Research, 35, 296–315.

Sclove, S. L. (2007). Path analysis. In N. J. Salkind (Ed.), Encyclopedia ofmeasurement and statistics (Vol. 2, pp. 745–748). Thousand Oaks, CA:Sage.

Silvia, P. J. (2007). An introduction to multilevel modeling for research onthe psychology of art and creativity. Empirical Studies of the Arts, 25,1–20.

Simonton, D. K. (1977). Eminence, creativity, and geographic marginality:A recursive structural equation model. Journal of Personality and SocialPsychology, 35, 805–816.

Simonton, D. K. (1991a). Career landmarks in science: Individual differ-ences and interdisciplinary contrasts. Developmental Psychology, 27,119–130.

Simonton, D. K. (1991b). Emergence and realization of genius: The livesand works of 120 classical composers. Journal of Personality and SocialPsychology, 61, 829–840.

Simonton, D. K. (1991c). Latent-variable models of posthumous reputa-tion: A quest for Galton’s. G. Journal of Personality and Social Psy-chology, 60, 607–619.

Simonton, D. K. (1998). Fickle fashion versus immortal fame: Transhis-torical assessments of creative products in the opera house. Journal ofPersonality and Social Psychology, 75, 198–210.

Simonton, D. K. (2002). Collaborative aesthetics in the feature film:Cinematic components predicting the differential impact of 2,323 Oscar-nominated movies. Empirical Studies of the Arts, 20, 115–125.

Simonton, D. K. (2004a). The “Best Actress” paradox: Outstanding featurefilms versus exceptional performances by women. Sex Roles, 50, 781–795.

Simonton, D. K. (2004b). Film awards as indicators of cinematic creativityand achievement: A quantitative comparison of the Oscars and sixalternatives. Creativity Research Journal, 16, 163–172.

Simonton, D. K. (2004c). Group artistic creativity: Creative clusters andcinematic success in 1,327 feature films. Journal of Applied SocialPsychology, 34, 1494–1520.

Simonton, D. K. (2005a). Cinematic creativity and production budgets:Does money make the movie? Journal of Creative Behavior, 39, 1–15.

Simonton, D. K. (2005b). Film as art versus film as business: Differentialcorrelates of screenplay characteristics. Empirical Studies of the Arts, 23,93–117.

Simonton, D. K. (2007a). Cinema composers: Career trajectories for cre-ative productivity in film music. Psychology of Aesthetics, Creativity,and the Arts, 1, 160–169.

Simonton, D. K. (2007b). Film music: Are award-winning scores andsongs heard in successful motion pictures? Psychology of Aesthetics,Creativity, and the Arts, 1, 53–60.

Simonton, D. K. (2007c). Is bad art the opposite of good art? Positiveversus negative cinematic assessments of 877 feature films. EmpiricalStudies of the Arts, 25, 143–161.

Simonton, D. K. (in press). Cinematic success criteria and their predictors:The art and business of the film industry. Psychology & Marketing.

137CINEMATIC SUCCESS

Sochay, S. (1994). Predicting the performance of motion pictures. Journalof Media Economics, 7, 1–20.

Taylor, R. A. (1974). Television movie audiences and movie awards: Astatistical study. Journal of Broadcasting, 18, 181–186.

Walker, J. (Ed.). (2007). Halliwell’s film & video guide 2008. New York:Harper Collins.

Wallace, W. T., Seigerman, A., & Holbrook, M. B. (1993). The role ofactors and actresses in the success of films: How much is a movie starworth? Journal of Cultural Economics, 17, 1–27.

Walls, W. D. (2005a). Modelling heavy tails and skewness in film returns.Applied Financial Economics, 15, 1181–1188.

Walls, W. D. (2005b). Modeling movie success when “Nobody knows

anything”: Conditional stable-distribution analysis of film returns. Jour-nal of Cultural Economics, 29, 177–190.

Wanderer, J. J. (1970). In defense of popular taste: Film ratings amongprofessionals and lay audiences. American Journal of Sociology, 76,262–272.

Zickar, M. J., & Slaughter, J. E. (1999). Examining creative performanceover time using hierarchical linear modeling: An illustration using filmdirectors. Human Performance, 12, 211–230.

Received June 30, 2008Revision received October 17, 2008

Accepted October 22, 2008 �

138 SIMONTON