Using IMDB user ratings to study the relationship between character strengths and movie enjoyment, and the role of gender and gender representations in this relationship

(1)

1

Master thesis: “Using IMDB user ratings to study the relationship between

character strengths and movie enjoyment, and the role of gender and gender

representations in this relationship”

Name: Paulien Rouwendaal Student number: 5983126 Date: January 27th 2017

Degree: MSc. in Business Administration – Marketing Track Educational institution: University of Amsterdam

Supervisor: Frederik Situmeang

Statement of originality

This document is written by Paulien Rouwendaal who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(2)

2

Bridging media psychology and positive psychology literature, this is the first study to explore the relationship between character strengths portrayed in movies and movie enjoyment. Furthermore, it is the first study that uses Bechdel’s test for gender equality to study gender portrayals in movies and its role in the aforementioned relationship. Using a sample of 23,466 IMDB user ratings, regression analyses revealed that strengths Leadership and Zest of the leading movie character significantly increased movie enjoyment compared to other strengths. Furthermore, both males and females preferred movie characters with neutral strengths over movie characters with stereotypical masculine or feminine strengths. It was further revealed that, using Bechdel’s test, movies with more developed female characters are enjoyed just as much as male-centered movies. This study yields multiple managerial

applications. Screenplay writers could increase movie enjoyment by including certain strengths, potentially enhancing a movie’s box office performance. In the long run, not only movie studios, also streaming providers such as Netflix and other platforms that rely heavily on recommendation algorithms could benefit from these insights, as they could match character strengths of their users with those portrayed in movies in order to increase enjoyment. Furthermore, this study shows that stereotypical representations of males and females in movies are unnecessary as they do not increase movie enjoyment.

(4)

4

Introduction

The production of movies is a risky business, with revenue forecasts lacking precision and the probability distribution of box-office outcomes having an infinite variance (De Vany & Walls, 1999). This means that it simply cannot be predicted if a particular movie will be a hit or a flop, causing a studio either millions of euros in profits, or in losses (Terry, Butler and D’Armond, 2005). Previous studies showed that some variables are positively correlated to - though not guarantee – a movie’s box office performance, such as the production budget (Basuroy, Chatterjee & Ravid 2003), its “star power” (Elberse 2007), and the popularity of its genre (De Vany and Walls 1999). Nevertheless, uncertainty remains high, and to reduce this uncertainty for movie industry practitioners more research is necessary to understand what factors can increase box office performance.

Essential to generating box office revenue is movie enjoyment, as individuals who have enjoyed a particular media experience, such as a movie trailer or the first part of a trilogy, are likely to seek out that experience again in the future (Green, Brock & Kaufman, 2004). Furthermore, due to a positivity bias, people are more likely to post about products or services when they are satisfied with it than when they are dissatisfied with it (Gao et al., 2005), increasing the amount of online word-of-mouth (eWOM). Duan, Gu & Whinston (2008) found that the amount of eWOM in turn can positively affect BOP. Thus, when somebody enjoyed a movie, they are more likely to post something about it on the internet than when they did not like it. The amount of posts about a movie posted by professional critics, friends, family and colleagues in turn urges potential viewers to the box office themselves (Duan et al., 2008). Movie enjoyment thus has a central role in attracting more customers to the box office.

As enjoyment with movies is essential to attract more movie goers to the box office, it is relevant to study how movie enjoyment can be increased. For a movie to be enjoyable, it is necessary that viewers are able to identify themselves with the characters in the movie (e.g. Cohen, 2001; Cohen, 2006; Igartua, 2010). The general consensus in media psychology literature is that identification with movie characters increases when viewers perceive they are similar to the movie character, either demographically, psychologically, or both (see Cohen, 2006, for a full review). An example of identification with movie characters on the basis of psychological similarity is identification on the basis of character traits (e.g. Hoffner, 1996; Hoffner & Cantor, 1991; Hoffner & Buchanan, 2005). Hoffner & Buchanan (2005) found that viewers identified more with successful and admired movie characters, than with smart, funny

(5)

5

or violent characters. If some character traits lead to more identification and thus more movie enjoyment, it is interesting to study which character traits are the most effective.

Positive psychology literature made great contributions to the study of character traits in humans, identifying six virtues comprising 24 character strengths (Peterson & Seligman, 2004). In a cross-cultural examination of character traits, Lee, Foo, Adams, Morgan & Frewen (2015) found that some of these character strengths are universally prevalent in humans across cultures, confirming an earlier study by Park, Peterson & Seligman (2006). Movies in which the leading character has such strengths might be more enjoyable, and might be enjoyed by a bigger crowd than movies that do not show these character strengths. Media psychology literature has further shown that there are gender differences in the way viewers identify with male versus female movie characters (Hoffner & Buchanan, 2005). One

explanation for this is that males and females possess different character traits on the basis of which they identify with same and opposite sex movie characters. An alternative explanation is that there is a misrepresentation of genders in movies. To explore how these viewer and movie characteristics are related, the following research question will be tested: What is the

relationship between character strengths in movies and movie enjoyment, and what is the role of gender and gender representations in movies in this relationship?

This study will contribute to the current body of literature in multiple ways. Bridging positive psychology and media psychology literature, this study will be the first to test the relationship between character strengths in movies and movie enjoyment. Furthermore, this study will provide more insights in gender preferences for character strengths in movies, and if a more equal representation of genders in movies can enhance movie enjoyment. Using Bechdel’s test, a measure for gender representation in movies, Hickey (2014) and Linder et al. (2015) found that developed female characters were indeed greatly underrepresented in

roughly 50 % of all movies. Although their studies claim that audiences enjoy movies with more developed roles for females just as much as or even more than movies that are more male centered, no study to date has tested this assumption. This study aims at filling these gaps in the literature.

This research may yield some interesting managerial implications for the movie industry. A movie might become more enjoyable if it features a leading character with certain character traits, and this might or might not be dependent on the gender of the target audience. Using these insights could not only increase sales, but also reduce pre-release uncertainty about box office outcomes. Not only movie studios, also streaming providers such as Netflix and other platforms that rely heavily on recommendation algorithms might benefit from this

(6)

6

knowledge. Should they match the character traits and gender of viewers with character traits in movies and series in order to make better, more personalized recommendations, hereby increasing consumption? Furthermore, this research might further support earlier findings that movies, that feature women in more developed roles, equally perform at the box office as movies that don’t. This would suggest that the ‘celluloid ceiling’, that still dominates Hollywood, should not only be of concern to feminists or society. Taking into account that roughly 50 % of movie ticket buyers is female, equal representation of genders in movies is a tactic the movie industry could capitalize on.

As movie enjoyment can be operationalized by user ratings on the Internet Movie Data Base (IMDB), this study will use a dataset containing over 120.000 online movie ratings from IMDB.com. The dataset also contains a wide array of metadata on movies, as well as the users’ user names, from which the users’ gender can be derived. The sample size and variables in the dataset will allow to control for variables that are potential confounding factors in the conceptual model, such as a movie’s production budget and its star power. The dataset will be merged with data from Niemiec & Wedding (2014), who have classified over 1.400 movies on the basis of the character strengths of Peterson & Seligman (2004). For gender and gender representations in movies, additional data will be scraped from other websites and added to the dataset.

In the next sections, first an overview of relevant theory will be discussed. Once the conceptual framework is established, the method is discussed, followed by the results. In the discussion, the findings of this study will be reflected upon, drawing conclusions, limitations, and suggestions for future research.

Theoretical framework

Identification with movie characters

According to Cohen (2006, p.183), “It is not the mere exposure to entertainment that we enjoy, but the ability of entertainment content to distract us from ourselves and reveal to us novel and exciting experiences of others”. Put differently, for a movie to be a pleasurable experience, it is essential that viewers are able to identify themselves with the characters in the movie (Cohen, 2006; Igartua, 2010; Hoffner &Buchanan). Reviewing the large body of literature on identification with media characters, Cohen (2001) formulated the following definition: “Identification is a process that culminates in a cognitive and emotional state in which the audience member is aware not of him- or herself as an audience member, but rather

(7)

7

imagines being one of the characters in the text.” (Cohen, 2001, p. 252). Cohen’s definition implies that identification is not just an attitude or emotion, but a process in which one temporarily loses self-awareness, resulting in a heightened emotional and cognitive

connection with a character. It is different from identifying with social groups or leaders, in the sense that it is a response to carefully constructed textual features, a narrative that is intended to provoke identification. Identification furthermore is based on internalizing a view point, rather than a projection of one’s own identity onto someone or something else. People often identify with characters they find attractive or even wish to be like (Cohen, 2006; Von Feilitzen & Linne, 1975).

Identification and its relationship with movie enjoyment

Cohen (2001) conceptualized identification with movie characters along four dimensions:

affective empathy (empathy towards or sharing the feelings of the character), cognitive

empathy (sharing the perspective of the character, or the degree to which an audience member

understands the characters motivation for his/her behavior), motivational empathy (the degree to which the audience member internalizes and shares the goals of the character), and

absorption (the degree to which self-awareness is lost during exposure to the text). From this

conceptualization it is clear that empathy is a key process in the process of identification with media characters, which can explain why identification increases movie enjoyment. Feelings of empathy are enjoyable because they tap into individuals’ need for emotional stimulation (Eliashberg & Sawney, 1994). De Wied et al. (1994) found that empathic sadness was indeed positively correlated to the degree of enjoyment with a dramatic movie. Although viewers that highly empathized with movie characters were more distressed when something bad

happened to the character than low empathizers, they also reported greater enjoyment with the movie as a whole. Similarly, Buselle & Bilandzic (2009) and (Raney, 2002) found a positive relationship between empathy and movie enjoyment.

To clarify the concept of movie enjoyment a little more, according to Green et al. (2004), movie enjoyment refers to a pleasurable affective response to a stimulus (watching a movie). This does not mean that enjoyment with movies is synonymous to the experience of joy or happiness, as viewers can also enjoy watching sad genres (Oliver, 2008). From the perspective of the viewer, watching movies can be viewed as a trade-off: the viewer spends a considerable amount of time and money in exchange for an experience. The reward for this trade-off is the anticipated enjoyment from the experience. Enjoyment of movies has different determinants than enjoyment of non-narrative media such as science programs. Where

(8)

8

enjoyment from non-narrative media is likely to stem from the satisfaction of curiosity or information-seeking needs, enjoyment from narratives (such as movies and novels) stems from emotional engagement (Green et al., 2004). As will be discussed in more depth in the method section, user ratings on IMDB can be used as a measurement for movie enjoyment. That is why in this study, movie enjoyment and user ratings are used interchangeably.

Predictors for identification with movie characters

As identification with movie characters is essential for movie enjoyment, it is relevant to study what factors can increase identification. The strength and development of identification depends on multiple factors: the viewer characteristics, media character characteristics, and the text (directing, writing and acting) (Cohen, 2006). When viewer and media character characteristics are perceived to be similar in the eyes of the viewer, this will increase

identification (e.g. Cohen, 2006; Cohen, 2001; Lee, 1998; Maccoby & Wilson, 1957; Turner, 1993). Similarity between the viewer and the media character can occur on different levels. According to early identification research, similarity in age, sex, social class between the viewer and the character increased identification (Maccoby & Wilson, 1957). Viewers can also perceive psychological similarity between themselves and the movie character, on the basis of for example similar attitudes, beliefs, or character traits. For example, Turner (1993) found that similarity in attitudes was a better predictor of identification than similarity in looks. Children often identified with animals and kids that were older then themselves (Von Feilitzen & Linne, 1975). Psychological similarity seems to be a stronger predictor for identification than demographic similarity (Cohen, 2006).

The identification literature suggests that character traits of the media character are an important predictor for identification. For example, both Reeves & Greenberg (1977) and Reeves & Lometti (1979) found that children evaluated television characters based on typical human traits such as activity, strength and humor. Hoffner & Cantor (1991) found that certain character traits, especially humor, explained which characters were most liked. Hoffner & Buchanan (2005) identified six character traits (smart, successful, attractive, funny, violent, admired) and found that both men and women identified more strongly with characters that were successful and admired than with characters with other types of strengths, when judging characters of the opposite sex.

(9)

9

Character traits as a predictor for identification and movie enjoyment

So far, it was argued that identification with movie characters increases movie enjoyment, resulting in higher ratings for that movie. As character traits are an important predictor for identification, several character traits will now be identified in order to study their influence on movie enjoyment. Because people often identify with people they like or wish to be like, only character traits that are perceived as strengths are taken in to account.

A great resource for theory on character strengths is the positive psychology literature. Whereas mainstream psychology focuses on the treatment of psychopathology, positive psychology is focused on achieving a satisfactory life, so that pathology can be prevented (Seligman & Csikszentmihalyi, 2014). One of the aims of positive psychology is thus to build positive qualities within persons. In order to do so, Peterson & Seligman (2004) identified six virtues – namely wisdom, courage, justice, humanity, temperance and transcendence-

comprising 24 psychological traits of humans. A complete overview of these virtues and character strengths, and their definitions, can be found in Table 1. Niemiec & Wedding (2014) classified a list of over 1.400 movies on the basis of these character strengths, which can be found in their book Positive Psychology at the Movies. In order to decide whether a movie was a “positive psychology movie”, Niemec & Wedding used the following coding instruction (p. 393):

“1. […] Strong portrayal of a character strength that benefits self/ others; conflict/ obstacles that challenge the strength; strength is used to overcome adversity; an overall presentation that is uplifting or speaks deeply to the human condition.

2. Portrayal of a character’s signature strengths, which are strengths that are critical to who the character is. In addition, signature strengths are energizing and natural to use, are displayed across settings, and are recognized by others as highly characteristic of an individual (Peterson & Seligman, 2004).

3. The film holds great potential to generate cinematic elevation (i.e., it causes people to feel motivated to do good and increases their altruism) or cinematic admiration (i.e., it causes people to want to improve themselves and copy the model).”

Although Niemiec & Wedding’s (2014) list was originally meant for psychology practitioners to as use as inspirational content for their clients, the list is a useful source to study character strengths in movies. Their data will be used to study if movies that show

(10)

10

certain character strengths are more enjoyable than others, and outperform at the box office. For the sake of clarity, character strengths will be further referred to as ‘strengths’.

Table 1

Classification of Character Strengths

1. Wisdom and knowledge: cognitive strengths that entail the acquisition and use of knowledge. • creativity: thinking of novel and productive ways to do things.

• curiosity: taking an interest in all of ongoing experience.

• judgment: thinking things through and examining them from all sides. • love of learning: mastering new skills, topics, and bodies of knowledge. • perspective: being able to provide wise counsel to others.

2. Courage: emotional strengths that involve the exercise of will to accomplish goals in the face of opposition, external or internal.

• honesty: speaking the truth and presenting oneself in a genuine way. • bravery: not shrinking from threat, challenge, difficulty, or pain. • persistence: finishing what one starts.

• zest: approaching life with excitement and energy.

3. Humanity: interpersonal strengths that involve ‘‘tending and befriending’’ others. • kindness: doing favors and good deeds for others .

• love: valuing close relations with others .

• social intelligence: being aware of the motives and feelings of self and others. 4. Justice: civic strengths that underlie healthy community life.

• fairness: treating all people the same according to notions of fairness and justice. • leadership: organizing group activities and seeing that they happen.

• teamwork: working well as member of a group or team. 5. Temperance: strengths that protect against excess.

• forgiveness: forgiving those who have done wrong .

• modesty: letting one’s accomplishments speak for themselves.

• prudence: being careful about one’s choices; not saying or doing things that might later be regretted.

• self-regulation: regulating what one feels and does.

6. Transcendence: strengths that forge connections to the larger universe and provide meaning. • appreciation of beauty and excellence: noticing and appreciating beauty, excellence,

and/or skilled performance in all domains of life.

• gratitude: being aware of and thankful for the good things that happen. • hope: expecting the best and working to achieve it.

• humor: liking to laugh and joke; bringing smiles to other people.

• religiousness: having coherent beliefs about the higher purpose and meaning of life. Note. Reprinted from Character strengths in fifty-four nations and the fifty US states by Park, Peterson & Seligman, 2006, The Journal of Positive Psychology, 1(3), p. 119)

Testing the influence of strengths on movie enjoyment

24 strengths have been identified so far that might increase individuals’ identification with movie characters and thereby movie enjoyment. It is expected that some of these 24 traits

(11)

11

might increase identification more than others. As discussed earlier, Hoffner & Buchanan (2005) found that the some character traits led to wishful identification with movie characters of the opposite sex, whereas other character traits did not. Furthermore, while the prevalence of character traits in a population is highly cultural dependent and may vary greatly across individuals, there are some strengths according to Park et al. (2006), that are universally prevalent: the most prevalent in 54 countries and 50 US states were Kindness, Fairness, Honesty, Gratitude and Judgment. It is likely that movies, that contain one or more ‘universal character strength’, lead to more identification among more viewers through perceived psychological similarity, which will result in higher movie enjoyment. It is thus expected that movies that contain one or more universal strength result in higher online ratings:

H1a: Movies that depict one or more universal strength (Kindness, Fairness, Honesty, Gratitude or Judgment) receive a higher online rating than movies that don’t.

The role of gender in the relationship between strengths and movie enjoyment As discussed earlier, if a movie viewer perceives demographical and/or psychological similarity between themselves and the movie character, this will increase identification, leading to higher movie enjoyment. Whereas strengths are a basis for psychological

similarity, gender can be a basis for demographical similarity. For example, Hoffner (1996) found that male characters were liked by both boys and girls for their intelligence, and girls also liked male characters for their humor. However, female characters were judged by boys and girls solely based on their looks. Furthermore, Hoffner & Buchanan (2005) pointed out that whereas men and women both identified more strongly with successful and admired characters of the opposite sex, they found differences in the way male and female viewers judged characters of the same sex. Men identified with male characters whom they perceived as successful, intelligent, and violent. Women on the other hand identified more with female characters that they perceived as successful, intelligent, attractive and admired.

The differences that Hoffner (1996) and Hoffner & Buchanan (2005) found between males’ and females’ judgment of movie characters can be explained from two perspectives: 1) either females possess and admire different character traits than males, on the basis of which they identify with movie characters, or 2) gender differences in character traits do not exist, but the differences that were found in Hoffner (1996) and Hoffner & Buchanan (2005) are due to the portrayal of female and male characters in the movie. The current body of literature

(12)

12

seems to be inconclusive about the existence of psychological gender differences. Thus, in the remaining sections of this chapter, both perspectives will be theorized and hypothesized upon.

Perspective 1: Females and males are psychologically different

According to Huddy & Terkildsen (1993), there are pervasive and remarkably uniform differences in the personality traits ascribed to men and women. A typical woman is seen as warm, gentle, kind, and passive, whereas a typical man is viewed as tough, aggressive, and assertive (Huddy & Terkildsen, 1993). In the psychology literature, gender differences are widely acknowledged, and researchers have theorized on gender differences from either a social constructionist perspective, or an evolutionary perspective. According to the evolutionary perspective, stable gender differences in human species can be attributed to biological adaptations through evolution. A somewhat less radical view is the social constructionist perspective, which claims that gender differences across social contexts emerge from the meaning of male and female within particular social contexts (Wood & Eagly, 2002).

According to the evolutionary perspective, different psychological dispositions were built in females and males specific to their gender, due to evolutionary adaption to the conditions at hand in primeval times. Through these sex-specific evolved mechanisms, men and women differ psychologically and take on different social roles. Strengths such as

kindness and love would be more prevalent in females, because these are associated with their evolutionary role in nurturing the young (Eagly & Wood, 1999). Strengths associated with hunting and gathering, such as bravery, would be more prevalent in men (Eagly & Wood, 1999). In the social constructivist view, on the other hand, gender differences arise from the different expectations society has for men and women, which impose different restrictions and opportunities on each gender (Eagly & Wood, 1999). Gender differences are installed in men and women from birth as they are socialized differently. Whereas girls are socialized to be nurturing and prosocial, boys are socialized to suppress tears or feelings of pain, in order to be tough and invulnerable (Šikić-Mićanović, 1997). As a result, in adulthood, women are more likely to take on roles like homemaker and primary children care taker, while men take on roles in the paid economy and are the primary family provider (Eagly & Wood, 1999). Regardless the perspective, a large body of psychology research ascribes more nurturing traits to females, and more providing traits to males.

Some support for this view can be found in the literature on strengths. Brdar, Anić, & Rijavec (2011), argue that women typically score higher on so called strengths of heart while

(13)

13

men excel at more intellectual strengths. Indeed, strengths of kindness and love were

repeatedly found to be more prevalent in females than in males (Biswas-Diener, 2006; Brdar et al., 2011; Linley, Maltby, Wood, Joseph & Harrington, 2007; Miljković & Rijavec, 2008; Park, Peterson & Seligman, 2006; Shimai, Otake, Park, Peterson, & Seligman, 2006). Further differences were found by Shryack, Steger, Krueger & Kallie (2010), who found that love of learning, kindness, love, teamwork, appreciation of beauty and gratitude were more prevalent in females than in males. Linley et al. (2007) also found that women scored higher than men on interpersonal strengths, such as kindness and love, and social intelligence. The same pattern was also found for appreciation of beauty and gratitude, although the effect sizes were small.

Following this line of reasoning, it seems that more stereotypical female strengths, such as kindness, love, and appreciation of beauty, are more prevalent in females than in males. If this is true, it is expected that females enjoy movies, in which the leading character shows these strengths, more than males, due to perceived psychological similarity. When the leading character in such a movie is of the same sex, this effect will be expected to be even stronger, due to increased demographic similarity. To test/ this, hypothesis 2a will be tested.

H2a: Females give higher ratings to movies that depict kindness, love, and/or appreciation of beauty, than males. This relationship is stronger when the leading character of the movie is of the same sex/female, than when the leading character is of the opposite sex/male.

Similarly, a number of stereotypical male strengths appear to be more prevalent in males than in females, namely: self-regulation (Shryack et al, 2010), creativity (Linley et al., 2007), bravery (Shimai et al.) and curiosity (Brdar et al., 2011). It is expected that male viewers will appreciate movies more when the leading character shows such strengths, than their female counterparts. Furthermore, when the leading character is of the same sex, the perceived demographic similarity is increased, which will strengthen this relationship. To test this, hypothesis 2b will be tested:

H2b: Males give higher ratings to movies that depict self-regulation, creativity, bravery and/or curiosity, than females. This relationship is stronger when the leadingcharacter of the movie is of the same sex/male, than when the leading character is of the opposite sex/female.

(14)

14

Perspective 2: There are no differences between males and females, only how they are portrayed in movies

In the previous sections, it was theorized that the gender differences found by Hoffner (1996) and Hoffner & Buchanan (2005) can be explained by differences in character traits between females and males, due to either evolutionary adaptations, or institutionalization by their social context. But what if males and females are not so different after all? Although gender differences have been found in the studies discussed previously (e.g. Park, Peterson & Seligman, 2006; Linley et al., 2007; Shryack et al, 2010), often, the rankings of strengths in females and males were roughly the same. Also, there are inconsistencies among studies in this field. For example, Lee et al. (2015) found cross-cultural differences in individuals’ strengths, but they did not find significant differences in strengths between males and females despite a sample of over 110.000. Furthermore, the effect sizes in the study of Linley et al. (2007), who found that love and kindness were more prevalent in women and creativity more in men, were small. And although they found differences, they also found that four of the top five ‘‘signature strengths’’ of UK men and women overall were the same (judgment, fairness, curiosity, and love of learning). Lastly, the study of Brdar et al. (2011), who found that women scored higher than men on love and kindness, and men higher on curiosity, had relatively small sample sizes.

Contrary to studies that claim the existence of gender differences, the gender similarities hypothesis of Hyde (2005) states that males and females are similar on most psychological variables. This means that men and women, as well as boys and girls, are more alike than they are different. The meta-analysis of 46 articles performed by Hyde (2005), supports this notion. She found that 78 % of the articles that claimed to have found significant gender differences had small to close-to-zero effect sizes. Hyde (2005) criticizes the

“overinflated claim of gender differences” in academic literature, that only reifies harmful gender stereotypes of women as caring and nurturant and men as lacking in nurturance. Hyde (2005) points out that when gender differences are found in psychological studies, they are often a result of contextual circumstances of that given study. One such contextual

circumstance in the context of media psychology research is the portrayal of males vs. females in movies.

Thus, an alternative explanation might be that the differences in judgements of female and male strengths that were found in Hoffner (1996) and Hoffner & Buchanan (2005) can be explained by the strengths that are portrayed by male vs. female actors, instead of

(15)

gender-15

predisposed differences in character traits between males and females. Hoffner (1996) argued that his findings “may be due to the fact that male characters were more plentiful and had more exciting, interesting roles.” This notion is supported by the fact that literature has repeatedly shown that girls choose male characters as role models more often than boys choose female characters (e.g., Albert, 1957; Reeves & Miller, 1978).

A critical note on this notion is that it may be more socially acceptable for females than for males to behave in ways traditionally associated with the other sex (Huston, 1983), which resulted in them more often choosing male role models in earlier studies. Furthermore, women might have been portrayed in a more traditional, less inspiring way in the days of Albert (1957) and Reeves & Miller (1978). However, from the following section it will become clear that there still seems to be a misrepresentation of women in contemporary movies, which could explain gender differences in judgement of media characters. The next paragraphs will focus on that.

Bechdel’s test: a measure of gender representation

Gender inequality in Hollywood has been a hot topic lately in popular news media, with actresses such as Meryl Streep, Jennifer Lawrence, and the likes speaking out on lower pay rates and unfair standards compared to their male counterparts. But this gender bias is also reflected in the depiction of female movie characters, such as the lack of character

development and serious roles for women (Hickey, 2014). A popular test for gender bias in movies is Bechdel’s test (Hickey, 2014). There are three rules for a movie to pass the test: 1. It has to have at least two women in it, 2. Who talk to each other, 3. About something besides a man. The test, originally intended as "a little lesbian joke in an alternative feminist newspaper", stems from a cartoon from 1985, Dykes to Watch Out For, written by Allison Bechdel. Recently, the test has been gaining more research attention and has been used to detect male bias in movies (Agarwal, Zheng, Kamath, Balasubramanian & Dey, 2015; Hickey, 2014; Lindner, Lindquist, & Arnold, 2015; Scheiner-Fisher and Russell III, 2012). Although nowadays more movies pass the Bechdel test than in the 1970’s, Hickey (2014) noted that in the past two decades, this positive trend has plateaued, and the number of movies that pass the test has stayed below 50 % ever since.

There are some obvious limitations about the Bechdel test for measuring gender equality in movies, due to its generality. For example, a movie can still pass the test even when it shows highly stereotypical representations of women (for example, women only discussing topics such as babies or nail polish). Also, a movie such as Interstellar, which

(16)

16

starred Sandra Bullock as a developed character in a non-stereotypical role (an astronaut), fails to pass the test because she does not interact with another woman (Van Raalte, 2015). Although the test has these serious limitations, it’s the only quantitative measure to date for gender equality in movies that has been tested academically. Also, a lot of data is available about movies that pass or fail Bechdel’s test, for example on www.bechdeltest.com that contains coded data on almost 5.000 movies. As Lindner et al. (2015) pointed out, data on Bechdel’s test is often user generated and might be low in validity. But when they let researchers and graduates code a sample of 100 movies from www.bechdeltest.com to calculate the intercoder reliability with the user generated coding, they found a high

intercoder reliability (α = .83). Furthermore, Agarwal et al. (2015) used a Natural Language Processing (NLP) technique to link 964 movie scripts to data on the Bechdel test and IMDB data, revealing that women are indeed portrayed as less central and less important in movies that fail the test than movies that pass. These findings supported the long held belief that women are portrayed as less strong leaders and thinkers in popular media, and provided validation of the Bechdel test as a measure for gender representation in movies (Agarwal, 2015).

According to the first perspective, females would give higher ratings to movies that show stereotypical female strengths (H2a), and that males give higher ratings to movies that show stereotypical male strengths (H2b). These relationships were expected to be even stronger when the leading actor was of the same sex as the viewer, due to demographic similarity. But according to Hyde (2005), such stereotypical psychological differences

between men and women do not exist. Thus, in the alternative scenario, it is still expected that some strengths lead to more identification than others (H1a and H1b), but the gender

differences in preferences for character traits (H2a & H2b) are not expected. However, gender differences are expected in the sense that female and male leading characters might be

portrayed with stereotypical strengths.

Testing Bechdel’s test

As discussed earlier, Bechdel’s test has its limitations, and more research is needed into the validity/application of the test. As Agarwal et al. (2014) pointed out, movies that past the Bechdel test, generally portray women in a more central way than movies that failed. Would they also portray women and men in a less stereotypical way? It is assumable that movies in which women are portrayed more centrally and important, also portray them in a less

(17)

17

earlier discussion, kindness, love and appreciation of beauty were pinpointed as being stereotypical female. Thus, it is expected that movies that pass the Bechdel test and have a female in the leading role, show less of these strengths compared to movies that fail the test. To test this, H3a:

H3a: Compared to movies that fail Bechdel’s test, movies that pass the test and have a female in the leading role, show her with less stereotypical strengths.

In a similar fashion, it is expected that movies that pass the Bechdel test and have a male leading actor show less stereotypical male strengths than movies that fail the test.

Self-regulation, curiosity, bravery and creativity were identified as typical male strengths in earlier sections. Thus, it is expected that movies that pass Bechdel test show male characters in a less stereotypical way than movies that fail the test:

H3b: Compared to movies that fail Bechdel’s test, movies that pass the test and have a male in the leading role, show him with less stereotypical strengths.

Next, it is interesting to test what the influence is of a more equal gender representation on a movie enjoyment and BOP. The next section will focus on that.

Gender representation and the successfulness of a movie

Lindner et al. (2015) tested if movies that featured women living their lives independent from men underperformed at the box office compared to more male-centric movies. They found that male-centered movies greatly outperformed the more female-centered movies. Do

audiences then prefer male-centered movies? Rather, their results showed that movies that are more male-centric generally receive more production budget than female-centered movies, which enables those movies to become more successful. This ‘gendered distribution of resources’, or gender inequity, within movie studios is mainly due to institutionalized

perceptions and practices, keeping the celluloid ceiling in place (Lindner et al., 2015). This is comparable to Hickey’s (2014) results, who controlled for the production budget a movie received, when he found that movies that passed Bechdel’s test performed equally well as movies that failed.

Both Hickey (2014) and Lindner et al. (2015) use their findings to call for more gender equity in the movie industry, concluding that audiences probably enjoy Bechdel-movies just

(18)

18

as much or even more than non-Bechdel movies. They argue that equal representations of men and women in movies might even increase movie enjoyment, hereby attracting more consumers to the box office. However, no literature on Bechdel’s test to date has tested their assumption. That is why hypothesis 4a will be tested:

H4a: Movies that pass the Bechdel test receive a higher average online rating on IMDB than movies that don’t.

It is likely that the effect hypothesized in H4a is stronger for females than for males. Movies that pass the Bechdel test portray women in a more central and important way than movies that fail the test. When a movie shows more than one woman, talking about something other than men, it is likely that female viewers will both perceive more demographic similarity as well as psychological similarity. This will in turn increase identification among female viewers, and subsequent movie enjoyment. It is thus expected that females will give higher ratings to movies that pass the test, than to movies that don’t. For males, this effect may be absent due to the fact that, whether a movie passes the Bechdel test or not, male characters in movies are generally more prevalent and central in movies anyway. To test this assumption, hypothesis 3b will be tested:

H4b: Women give higher online ratings to movies that pass Bechdel’s test than to movies that fail Bechdel’s test. For men, this relationship is absent.

(19)

19

Method

In this chapter, the process of data-collection and cleaning, as well as the operationalization of variables, is discussed.

Data collection & Sample

For the purpose of this study, two datasets were created: a cross-sectional movie-level dataset in order to study gender representations in movies, and a user-level cross-sectional dataset, in order to study the relationship between character strengths, gender and gender representations and movie enjoyment. Below, the process of data collection for each dataset is discussed, as well as the sample characteristics.

Movie level Data

This study uses several sources of data on movies. The first source is the Internet Movie Database (IMDB.com), which is a subsidiary of Amazon.com and currently is the biggest online database for information on movies, television shows and games. IMDB has 70 million users, who are enabled to post reviews and ratings on all types of titles. Firstly, a secondary dataset with metadata on movies was obtained from Frederik Situmeang (this thesis’

supervisor). The dataset contained cross-sectional data on 5639 movie titles released between 1919 and 2014, such as its production budget, its genre and the movie studio by which it was produced. Additionally, data on the cast and crew of the movies was obtained from Frederik Situmeang.

The next source of data was Niemiec & Wedding’s book Positive Psychology at the

Movies (2014). Niemiec & Wedding studied the prevalence of 24 character strengths in

movies that were originally identified by Peterson & Seligman (XXX). Niemiec & Wedding’s book contains a list of over 1400 movies in the Appendix that are classified on the basis of the strengths of their leading character. According to their data, spirituality (22.20 % -131), love (21.52 % - 127), perseverance (12.03 % - 71) and creativity (10.51 % - 62) were the most common strengths portrayed in movies. Prudence (1.86% - 11), love of learning (2.71 %- 16), and perspective (2.7 % - 16) were the least common strengths portrayed in movies.

The last source for movie data was www.bechdeltest.com, which was scraped using Google Chromes Web Scraper tool. The website contains coded data on almost 6.634 movies that either passed or failed Bechdel’s test (a measure of character development of female movie characters), and has been used in previous research as well (for example by Hickey, 2014, to study its relation to box office performance). Of all movies on Bechdeltest.com, 57,8 % of the movies passed the test.

(20)

20

The datasets from IMDB.com, Niemiec & Wedding’s (2014) book and Bechdeltest.com were merged into a single file on the basis of movie titles, using the VLOOKUP function in Excel. However, the notation of movies differed among the files, hindering VLOOKUP to recognize some movie titles correctly. That is why the Excel columns containing the movie titles were “cleaned” first (examples can be found in Table X, Appendix X) before VLOOKUP was run.

User-level data

Besides movie-level data, user-level data on movies was scraped from IMDB as well,

resulting in a cross-sectional dataset containing 126.255 ratings and reviews by 41.800 unique user accounts. These were posted by IMDB users between December 22nd 2001 and February 21st 2014 about a total of 5.569 movie titles. This dataset contained users’ IMDB user name, the ID-number of the movie he or she rated and the score he or she gave to a movie (ranging 1 to 10). The user-level data and the movie-level dataset were merged on the basis of Movie ID-numbers using the Excel VLOOKUP function.

The last step in the data collection process was aimed at collecting data on first names by gender, in order to derive the gender of IMDB-userss. The program gender.c was used, which is a program written in the programming language C, and designed to determine a person’s gender based on a given first name. Gender.c was written by Michael (2008) on behalf of a German computer magazine and could be downloaded from

https://www.heise.de/ct/ftp/07/17/182/. Gender.c contains a database of 48.527 first names by gender from 33 countries. In Appendix 1, Table 2, an overview of gender.c’s classification rules is provided. Using the gender.c database and the VLOOKUP function in Excel, the gender of the IMDB users could be derived from their user name. A first inspection of the data revealed that users often used their real first and last name to create an account, partly due to the fact that IMDB allows users to create an account using their Facebook login. However, user names often contained all kinds of ‘noise’, such as numbers, punctuation marks, symbols and other notations that would hinder VLOOKUP in recognizing the first name. Thus, using Excel, first the user names were cleaned up before running VLOOKUP. Table 3 in Appendix 1 shows a few examples of names that were cleaned up and classified using this process.

(21)

21

Variables

Movie enjoyment. As a measure for movie enjoyment, IMDB user movie ratings were used

that were available in the IMDB user dataset. The rating system on IMDB.com allows the user to rate movies on a scale ranging from 1 (“awful”) to 10 (“excellent”). Single-item measurements of movie enjoyment are common within the media psychology literature (e.g., Greeson, 1991; Iguarta, 2010; Knobloch & Zillmann, 2002; Krcmar & Kean, 2004; Zillmann, 1989). For example, Igartua (2010) measured enjoyment with movies using a one-item measure, “to what extent did you like the film?”, which had an answering scale ranging from ‘0. I didn’t like it at all’, to ‘10. I liked it very much’. Validation of this approach can be found in the study of Gao, Greenwood, Agarwal & Jeffrey (2015), who showed that online rating systems can serve as a valid instrument for measuring actual population opinions. They found that 5-star online ratings were positively correlated with actual population opinions by comparing them with ratings on a 5-point attitudinal Likert scale. Thus, in line with previous studies, it is assumed that the IMDB rating scale (M = 7.30, SD = 3.088.) is a valid instrument for measuring movie enjoyment, and that a low score on the scale indicates low movie

enjoyment, and a high score a high movie enjoyment.

Strengths. Each of the 24 strengths as identified by Peterson & Seligman (2004) and

classified by Niemiec & Wedding (2014) was included in this study. 24 dummy variables were created for each character strength, coded as 0 (= movie does not contain the character strength), or 1 (= movie does contain the character strength). To check if none of the separate strengths actually reflected the same construct, it was decided to perform a Principal

Components Analysis (PCA). The dummy variables had a normal distribution: for each strength, the skewness or the kurtosis was lower than |1|. For example, the dummy variable love had a skewness of .920 and a kurtosis of .690.

Next, the 24 dummy variables were entered into a Principal Components Analysis (PCA) analysis using SPSS. Not all requirements for a PCA were met. Although Bartlett’s test was significant (χ² (276) = 750,65 p < .001), indicating that the correlation between the variables was high enough to perform PCA, Kaiser-Meyer-Olkins measure of sampling adequacy value was .047, which is well below the required .60. Through an initial analysis, the eigenvalues for each component in the data was obtained. Thirteen components had an eigenvalue higher than Kaiser’s criterion of 1, explaining 64.769 % of the variance. An inspection of the scree plot revealed the same number of components. These factors were retained and rotated using Oblimin with Kaiser normalization rotation. The results indicated that many items loaded strongly on more than one factor (high cross-loadings), which was

(22)

22

unsurprising regarding that the assumptions for performing a factor analysis were not met. Thus, it was decided not to construct scales with the character strength items, but to maintain all 24 items.

IMDB User Gender. In the user-level dataset, the gender of the user was classified by

the Gender.c program. In order to run the analyses, names that were masculine or likely masculine (M, ?M or 1M) were recoded as 0, and names that were feminine or likely feminine were coded 1. Karimi, Wagner, Lemmerich, Jadidi & Strohmaier (2016) calculated that the accuracy of gender.c was .74 which was considered acceptable. However, to increase the accuracy, a manual check was done on the user names that were classified as ‘?’, in order to identify more genders that gender.c did not recognize, due to the the notation of the user name. The manual check increased the amount of classified genders from 11183 to 15136, on a sample of 23465.

Leading actor gender. The gender of the leading actor was coded in a similar way as

the users’ gender. Again, males were recoded from M to 0, and females were recoded from F to 1. A manual check was performed afterwards on all the movies to ensure the coding was accurate.

Bechdel score. As a measure of gender equality, Bechdel’s test is used, assuming that

movies that pass the test show a more equal portrayal of men and women than movies that fail the test. Support for this notion can be found in Agarwal et al. (2015), who used a Natural Language Processing (NLP) technique to link 964 movie scripts to data on the Bechdel test and IMDB data. They discovered that women are indeed portrayed as less central and less important in movies that fail the test than movies that pass validating the Bechdel test as a measure for gender representation in movies. Data on Bechdel’s test is used from the website bechdeltest.com, were the following coding instruction is used to decide wether a movie passes or fails the test: 1. It has to have at least two women in it, 2. Who talk to each other, 3. About something besides a man. Movies that met all three requirements were coded 1 ( = movie passes the Bechdel test), the other movies were coded 0 (= movie fails the Bechdel test). Support for the use of data from www.bechdeltest.com can be found in Lindner et al. (2015). As Lindner et al. (2015) pointed out, data on www.bechdeltest.com is user generated and might be low in validity. But when they let researchers and graduates code a sample of 100 movies from www.bechdeltest.com to calculate the intercoder reliability with the user generated coding, they found a high intercoder reliability (α = .83).

(23)

23

Control variables

A number of variables were included in this study that was expected to potentially have some confounding effects in the models. Movies seem to perform better at the box office when they received a high production budget (Basuroy, Chatterjee & Ravid 2003); are produced by major Hollywood studios (Kim 2013); contain one or more major Hollywood actors or actresses, referred to as “star power” (Elberse 2007); entail a popular genre (De Vany and Walls 1999); or are a sequel (Elberse and Eliashberg 2003). As they might influence movie enjoyment as well, they are included in this study as control variables.

The variable adaptation of a book or play was included, as the expectation about a movie based on the original work might influence the actual movie experience. Previous studies showed that the enjoyment of a movie could indeed be influenced by expectations of its quality (Klaaren, Hodges, & Wilson, 1994; Wilson, Lisle, Kraft & Wetzel, 1989). For example, in the study of Klaaren, people with positive expectations about the movie experience enjoyed the experience more than people with neutral or negative expectations. For the same reasons, the variable sequel was included in the study. Similarly, when a movie is produced by a major Hollywood studio, this might signal quality to the movie goer, raising the expectations.

The variable star power was included because individuals often identify themselves with others who are like them or who they want to be like (Bandura, 1976; Fraser & Brown, 2002). The perceived similarity can be influenced by gender, ethnicity, age, or attractiveness from the actor or actress (Cohen, 2006). Actors and actresses are admired by many and can serve as references or role models for having special qualities such as outstanding

achievements or living a certain lifestyle (Bush, Martin, & Bush, 2000), which consumers also aspire to have (Choi & Rifon, 2007). During an actor’s or actress’s career, individuals have often developed strong feelings and attitudes towards him or her, in which identification plays an important role (Nijhuis, 2006).

As individuals might have distinct preferences for genres (Eliashberg & Sawhney, 1994), it was controlled for if a movie belonged to a popular genre (family, comedy, drama, action, adventure, horror, thriller/crime, romance, and science fiction/fantasy) or not. A movie’s production budget was included in this study as it might increase movie quality, hereby influencing movie enjoyment. A higher production budget might also mean that more money was spent on advertising, influencing movie-goers’ expectations. As this variable was highly skewed to the right (M = $ 27.025.482, SD = $ 41.127.908) production budget was

(24)

log-24

transformed to approximate a normal distribution. See Table 4 for a full overview of control variables included in this study.

(25)

25

Table 4.

Operationalization of Variables Included In This Study

Variable Label Operationalization Data Source Examplary Studies

User rating as a measure for movie enjoyment

MR Online consumer movie rating on a scale from 1 ( = awful 1 (=“awful”) to 10 (“excellent”)

IMDB.com Igartua (2010); Gao et al.(2015)

Character strengths depicted in a movie

CHAR One of the 24 character strengths as

identified by Peterson & Seligman (2004) that are shown in a movie

Book: Positive psychology at the movies by Niemiec & Wedding ()

Niemiec & Wedding ()

Gender of the user REVSEX Viewer is male (=0) or female (=1) heise.de/ct/ftp/07/17/182/ Karimi, Wagner, Lemmerich, Jadidi & Strohmaier (2016)

Gender of the leading character of a movie

ACTSEX Leading actor in the movie is male

(=0) or female (=1)

IMDB.com, Hoffner (1996), Hoffner & Buchanan ()

Bechdel score BECH A movie passes the test when it meets the following three requirements: 1) it has to have at least two [named] women in it, 2) who talk to each other, 3) about something besides a man (=1, 0 = otherwise).

BechdelTest.com Hickey (2014)

Production budget LN(BUDGET) Log-transformed production budget of a movie, in US $

IMDB.com Hennig-Thurau, T., Wiertz, C., & Feldhaus, F. (2015)

Genre GENRE _{Movie belongs to at least one out of}

nine major movie genres: family, comedy, drama, action,adventure, horror, thriller/crime, romance, and/or science fiction/fantasy (= 1 if appropriate or 0 otherwise)

IMDB.com Ho, Dhar, and Weinberg (2009)

Star Power STARS Movie contains one or more major stars (=1, 0 otherwise)

Quigley Publishing Swami, Eliashberg, and Weinberg (1999) Sequel SEQUEL Movie is a sequel (=1, 0 =

otherwise)

IMDB.com Hennig-Thurau, Houston, and Heitjans (2009)

Adaptation of book or play ADAPT Movie is an adaptation of a book or play (=1, 0 otherwise)

IMDB.com Hennig-Thurau, Wiertz Major Hollywood studio STUDIO Movie is produced by one of the six

major Hollywood studios (Warner, Fox,Universal, Sony, Paramount, Disney)

IMDB.com Elberse and Eliashberg (2003)

(26)

26

Results

Descriptives

The movie-level data file consisted of 590 movie titles, released between 1921 and 2011, with an average user rating a movie received was 27.02 million USD, (SD = 41.12 million USD). The movies received an average rating of M = 6.67, SD = 2.28. Of all movies in the sample, 54.8 % passed Bechdel’s test. The most common genres were drama (74.4 %), romance (29.8 %), comedy (28.5 %), thriller (15.5) and fantasy (12.1 %). The most common strengths featured in movies were Love (21%), Spirituality (8,7 %), Perseverance (7.8 %), and Honesty (7,3 %).

After merging the movie-level data with the data on IMDB user ratings, a dataset consisting of 23.465 ratings remained. These ratings were posted by 12.202 unique users about the 590 movies, posted between February 2nd 2006 and the 21st of January 2014. Of all users in the user-level dataset, 50.3 % were male, 14.9 % were female, 34.7 % were user names that were either androgynous or unknown. This dataset was used to test effects on user scores (hypotheses 1, 2a & b, 4a & b), whereas the movie-level dataset was used to test hypotheses 3a and b.

Hypotheses testing

The relationship between strengths and movie enjoyment

Hypothesis 1 predicted that movies that depict one or more universal character strength (Kindness, Fairness, Honesty, Gratitude or Judgment) receive a higher online rating than movies that don’t. A hierarchical multiple regression was used to assess this assumption, while controlling for a movie’s production budget, star power, whether or not the movie was of a popular genre, if the movie was a sequel or an adaption of a book or play, and if it was produced by a major studio. Preliminary analyses showed no violations of normality,

linearity, multicollinearity and homoscedasticity. The control variables were entered at Step 1, explaining 0.1 % of the variance in online ratings. At step 2, the 24 strengths were entered into the model, increasing the variance in online ratings explained by the model to 2.1 %, F (30, 20326) = 14.572, p < .001, hereby explaining an additional 2.0 % of variance, R squared change = .02, F change (24, 20326) = 17.04, p < .001.

` The strength Fairness significantly increased movie enjoyment (B = .33, p < .05), as well as Gratitude (B = .73, p < .001). However, Honesty, Kindness and Open Mindedness did not significantly increase movie enjoyment. An overview of the effects is presented in Table

(27)

27

5, ranked by the size of the standardized betas. The standardized beta’s revealed that the strongest predictors increasing movie enjoyment were not the expected strengths, but instead Leadership (beta = .08, B = 1.25, p < .001), Zest (beta = .08, B = 1.25, p < .001), and

Teamwork (beta = .0, B = .60, p < .001). Thus, although Gratitude and Fairness had a positive effect on movie enjoyment, they were not the best predictors for movie enjoyment. There were however strengths that even had a negative effect on movie enjoyment. For example, Self-regulation (beta = -.04, B = -.49, p < .001) and Love of learning (beta = .03, B = -.61, p < .001) had the strongest negative effect on movie enjoyment. Thus, hypothesis 1 was only partially supported.

Table 5

Summary of Multiple Regression Analysis for Strengths Predicting User Movie Ratings (N = 23466)

Variable B SE Beta Leadership 1.25 .12 .08*** Zest 1.30 .15 .07*** Self-regulation -.49 .11 -.04*** Teamwork .60 .11 .04*** Creativity .49 .10 .04*** Love of learning -.61 .18 -.03*** Gratitude .73 .16 .03*** Appreciation of beauty .47 .12 .03*** Humility .47 .14 .03** Bravery .03 .08 .03 Prudence -.51 .17 -.02** Curiosity -.39 .14 -.02** Spirituality -.19 .08 -.02** Perspective .50 .17 .02** Hope .42 .13 .02** Social intelligence .36 .12 .02** Perseverance .23 .08 .02** Love .20 .07 .02** Fairness .33 .14 .02* Judgment -.22 .12 -.01ms Forgiveness -.22 .12 -.01 Honesty -.16 .11 -.01 Kindness -.14 .11 -.01 Humor .05 .13 .00 R2 .021 F for change in R2 17.04

Note. Controlled for ln(budget), sequel, adaptation, star power, and

major studio.

*p < .05. **p < .01. ***p < .001. ns = not significant. ms = marginally significant.

(28)

28

Gender preferences and character traits

Hypothesis 2a predicted that females give higher ratings to movies that depict kindness, love, and/or appreciation of beauty, than males. This relationship is stronger when the leading character of the movie is of the same sex/female, than when the leading character is of the opposite sex/male. To test this hypothesis, separate hierarchical multiple regressions were run for each independent variable, beginning with the character strength appreciation of beauty. The control variables were entered at Step 1, explaining .1 % of the variance in user ratings. At step 2, the strength love, user gender, and actor gender were entered into the model to control for their direct effects on user ratings, explaining an additional 1.3 % of the variance in ratings, R² change = .01, F change (3, 13577) = 56.32, p < .001. The total variance of the model was 1.4 %, F = (9, 13586) = 20.86, p < .001. In this model, there was a positive relationship between user gender and user score (B = .75, p < .001), meaning that females in general gave higher ratings to movies than men. Also, there was a negative relationship between actor gender and user score (B = -.37, p < .001), meaning that movies with a female as a leading character received lower user ratings than movies with a male as leading

character. There was no significant effect of appreciation of beauty on user score. Then, at step 3, the direct effects as well as an interaction term of user gender * appreciation of beauty was entered into the model to test for the first part of hypothesis 2a. However, the model became insignificant. For the second part of hypothesis 2a, an interaction term of user gender*appreciation of beauty*actor gender was added to the model at Step 4. Still, the model stayed insignificant. The significant effects for appreciation of beauty are summarized in Table 6.

When this procedure was repeated for kindness and love, none of the models reached significance. Thus, no support was found for hypothesis 2a.

Table 6

Summary of Hierarchical Regression Analysis for Variables Predicting User Movie Ratings (N = 23466) Model 1 B SE Beta Appreciation of beauty ,31 ,13 ,02 User gender ,75 ,06 ,10*** Actor gender -,37 ,07 -,05*** R² ,01 F for change in R² 56,32

Note. Controlled for ln(budget), sequel, adaptation, star

(29)

29 *p < .05. **p < .01. ***p < .001. ns = not significant. ms = marginally significant.

Hypothesis 2b predicted that males give higher ratings to movies that depict self-regulation, creativity, bravery and/or curiosity, than females. This relationship is stronger when the leading character of the movie is of the same sex/male, than when the leading character is of the opposite sex/female. The procedure that was used for hypothesis 2a was repeated for hypothesis 2b, beginning with creativity. The variance of the model with the control variables was increased to 1.4 % (F = (9, 13.577) = 21.48, p < .00) after entering the direct effects of creativity gender and actor gender at Step 2, user, R² change = .01, F change (3, 13.577) = 58.17, p < .00. In this model, again there was a positive relationship between user gender and user score (B = -.75, p < .001), as well as a negative relationship between actor gender and user score (B = -.38, p < .001), which was in alignment with the findings for hypothesis 2a. No significant relationship between creativity and user score was found.

At Step 3, an interaction term of user gender * creativity was added to the model, to test the first part of hypothesis 2b. The variance explained by this model was 1.4 %, F = (1, 13.576) = 5.04, p < .05), not further increasing the variance of the previous model, R² change = .00, F change (13.576) = 5.04, p < .00. There was a significant negative interaction effect of user gender * creativity on user score (B = -.64, p < .05), meaning that male users indeed gave higher ratings to a movie when the leading actor showed creativity, than female users. This was in support of the first part of hypothesis 2b. At step 4, an interaction term of user gender * creativity * actor gender was added to the model, to test for the second part of hypothesis 2b. However, the model became insignificant. Thus, for creativity, hypothesis 2b was partially supported. An overview of the results of this analysis is provided in Table 7.

Table 7

Summary of Multiple Regression Analysis for Variables Predicting User Movie Ratings (N = 23466)

Model 1 Model 2 Model 3

Variable

B SE Beta B SE Beta B SE Beta

Creativity 0,27 0,11 ,02** 0,41 0,12 ,034** 0,41 0,12 ,03** User gender 0,75 0,06 ,10*** 0,79 0,07 ,107*** 0,79 0,07 ,11*** Actor gender -0,38 0,07 -,05*** -0,38 0,07 -,051*** -0,38 0,07 -,05*** User gender * creativity

0,41 0,12 ,034** 0,41 0,12 ,03** User gender * creativity

* Actor gender 0,79 0,07 ,107*** 0,79 0,07 ,11***

(30)

30

F for change in R² 21.48 5.04 ns

Note. Controlled for ln(budget), sequel, adaptation, star power, and major studio.

These steps were repeated for bravery, curiosity and self-regulation. Only the

hierarchical regression including bravery reached significance. Its outcomes are reported in Table 8. Support for the first part of hypothesis 2b was found, as there was a negative interaction effect of user gender and bravery (B = -.52, p < ,05). This means that male users indeed gave higher ratings to movies when the leading character showed creativity, than female users. However, when the interaction effect of user gender * bravery * actor gender was entered at Step 4, the model became insignificant, thus no support was found for the second part of hypothesis 2b.

Following from the testing op hypotheses 2a and 2b, two hierarchical regressions were performed for males and females separately, to further explore the relationship between strengths and user score among genders. For both genders, at Step 1, the control variables were entered, and at Step 2, all 24 strengths were entered into the model. The results of these regressions are presented in Table 9. Although this approach does not test if there are

significant differences between male and female

Table 8

Summary of Multiple Regression Analysis for Variables Predicting User Movie Ratings (N = 23466)

Model 1 Model 2 Model 3

Variable

B SE Beta B SE Beta B SE Beta

Bravery -0.03 0.09 0.00 0,06 0,10 0,01 0,06 0,10 0,01 User gender 0.75 0.06 0.10*** 0,79 0,07 0,11*** 0,79 0,07 0,11 Actor gender -0.38 0.07 -0.05*** -0,38 0,07 -0,05*** -0,39 0,07 -0,05 User gender * Bravery

-0,52 0,27 -0,02ms -0,59 0,29 -0,02

User gender * bravery * Actor gender

0,39 0,61 0,01

R² .011 .014 ns

F for change in R² 36.23 11.23 ns

Note. Controlled for ln(budget), sequel, adaptation of a book or play, star power, and major studio.

users, it does provide insights in how preferences for strengths in movies are ranked among men and women. Both males and females gave higher ratings to movies in which the leading