Portrayals of men and women in Drum magazine (South Africa) advertisements

Researchers in the social sciences and mass communication fields have been studying gender portrayals in the media for a long time (e.g. Courtney & Lockeretz 1971; Lindner 2004; Morna & Ndlovu 2007) for three main reasons. Firstly, images in the media may provide an insight into the status of women in society (Hovland et al. 2005). Secondly, the media and, in particular, advertisements have an influence on audiences’ discourses, values and attitudes; hence the argument that advertisements reflect and shape the cultural landscape (Eisend 2010; Waters & Ellis 1996). Thirdly, longitudinal analysis in the study of portrayals may highlight developments in gender relations (Hovland et al. 2005). The portrayal of men and women in the media is an issue of concern in part because of gender stereotyping. Particularly, the woman’s image as a symbol of multiple culturally constituted meanings is known to attract attention and is widely used in the promotion of products (Tan, Ling & Theng 2002). Based on a meta-analysis of 64 studies of gender roles in television and radio advertisements, Eisend (2010:418) concludes that gender ‘stereotyping is prevalent in advertising’, especially with regard to occupational status (where gender equality is least likely to be manifested). Though gender stereotyping is not negative on its own, it may lead to oversimplified and inappropriate evaluations of people and therefore contribute to the restriction of access to opportunities to some. It is therefore ideal to pursue gender equality across gender stereotype components. However, the pursuit of gender equality as a social goal is only appropriate for occupational roles and some role behaviours. For physical characteristics and trait descriptors, the use of the actual distribution of the characteristics in the population is appropriate.


Introduction
Researchers in the social sciences and mass communication fields have been studying gender portrayals in the media for a long time (e.g.Courtney & Lockeretz 1971;Lindner 2004;Morna & Ndlovu 2007) for three main reasons.Firstly, images in the media may provide an insight into the status of women in society (Hovland et al. 2005).Secondly, the media and, in particular, advertisements have an influence on audiences' discourses, values and attitudes; hence the argument that advertisements reflect and shape the cultural landscape (Eisend 2010;Waters & Ellis 1996).Thirdly, longitudinal analysis in the study of portrayals may highlight developments in gender relations (Hovland et al. 2005).The portrayal of men and women in the media is an issue of concern in part because of gender stereotyping.Particularly, the woman's image as a symbol of multiple culturally constituted meanings is known to attract attention and is widely used in the promotion of products (Tan, Ling & Theng 2002).Based on a meta-analysis of 64 studies of gender roles in television and radio advertisements, Eisend (2010:418) concludes that gender 'stereotyping is prevalent in advertising', especially with regard to occupational status (where gender equality is least likely to be manifested).Though gender stereotyping is not negative on its own, it may lead to oversimplified and inappropriate evaluations of people and therefore contribute to the restriction of access to opportunities to some.It is therefore ideal to pursue gender equality across gender stereotype components.However, the pursuit of gender equality as a social goal is only appropriate for occupational roles and some role behaviours.For physical characteristics and trait descriptors, the use of the actual distribution of the characteristics in the population is appropriate.

Gender stereotypes
According Deaux and Lewis (1984), gender stereotypes may be based on trait descriptors (e.g.independence), physical characteristics (e.g.height), role behaviours (e.g.serving others) and occupational status (e.g.medical doctors).Eisend (2010) suggests that research on gender stereotyping may be divided into optimistic studies and pessimistic studies.Optimistic studies are those that conclude that women are increasingly less likely to be portrayed negatively and that there is some progress towards the realisation of gender equality.For example, to the extent that they observed declining trends in gender stereotyping, the studies by Sexton and Haberman (1974) and Venkatesan and Losco (1975) may be labelled as optimistic studies.They found that women were depicted in the traditional roles of housekeepers and as beautiful but dependent companions to men.Both studies also observed favourable changes such as the declining number of advertisements portraying women as housewives, mothers and other unfair or unrepresentative roles.Other optimistic studies recorded minor declines in gender stereotyping with the established magazines of the day continuing to represent women in traditional roles.The use of sex in advertisements was becoming more overt and visual with women being more likely than men to be portrayed this way (Lyonski 1983;Ruggiero & Weston 1985;Soley & Kurzbard 1986).
Pessimistic studies on the other hand are those that conclude that women are still portrayed in negative and stereotypical ways and in some cases increasingly so.Examples of pessimistic studies include Plous and Neptune's (1997) study, which found that females were four times more likely than males to be depicted with their bodies exposed.Another study found that women were portrayed in traditional roles, although a decline in this trend was observed (Klassen, Jasper & Schwartz 1993).Pessimistic studies have also shown that women were underrepresented in advertisements (Ahmed et al. 2004).When race was taken into account, the old stereotypes were evident along with new stereotypes based on race (as women were represented differently depending on their race) (Sengupta 2006).Even in cross-cultural studies involving highly industrialised western countries where women have made significant progress in the social and economic spheres, gender biases and stereotypes were evident.For example, Wiles, Wiles and Tjernlund (1995) found that men were more likely than women to be portrayed in working situations and also in high-level executive positions than women.A comparison of United States (highly industrialised country) and Arab world (emerging market countries) depictions of women found that the two regions differed significantly in that Arab world advertisements only featured women if their presence was related to the advertised product and if their bodies were fully covered (Al-Olayan & Karande 2000).

Gender portrayals in advertising
The differential treatment of men and women in advertising is prevalent in the literature.The nature and number of portrayals of women and men is indicative of the roles and the relative power assigned to them (Hovland et al. 2005).For example, activities like housekeeping may be perceived as being of lower social status and be employed as a means to construct and reproduce gender identities (West & Zimmerman 1987).Women and men also tend to be used to advertise different types of products (e.g.Bretl & Cantor 1988;Culley & Bennett 1976;Dominick & Rauch 1972;Marecek et al. 1978).To illustrate, men are more likely than women to be featured in advertisements for products that are typically employed in occupational or work settings (non-housekeeping products).On the other hand, women are more likely to be featured in advertisements for products that are typically used in the home (housekeeping products).When women are used in advertisements, they typically are not shown to make important decisions or do important things but perform domestic work and other at-home activities (Courtney & Lockeretz 1971;Maake 2006).Women are more likely to be used to drape products to make them more appealing (Chestnut, LaChance & Lubitz 1977;Reid & Soley 1983).The use of separate or unequal settings or places for women and men may be used to socially confine women to roles such as housekeepers, mothers, wives and sex objects (Welter 1966).Women are more likely to be portrayed in family scenes and as providers of service to others (e.g.husbands, family members and friends) (Maake 2006).This is consistent with the view that masculine or instrumental roles are all important, dominating and workplace oriented whilst feminine or expressive roles are nurturing, compliant and home oriented (Connell 1987).
Gender stereotypes are also constructed and perpetuated through the assignment of voice to message sources as narrators or spokespersons.The role assigned to the narrator and the target audience of the narration are important strategic decisions in advertising (Bell, Kravitz & Wilkes 2000).Some advertisements use sexually oriented appeals including varying the extent of dressing, facial exposure and body display.Sexual appeal refers to the extent of sexual explicitness in an advertisement (Putrevu 2008).The use decorative models is another form of sexual appeal (Reid & Soley 1983).The use of sexual appeals is more likely to be applied to female models (Soley & Kurzbard 1986) and may be found even in advertisements in general interest magazines like Drum (Putrevu 2008;Soley & Kurzbard 1986).
In the case of gender representations in South African television advertisements, Luyt (2011) found hierarchical social differences in gender portrayals between men and women with men being represented as dominant and women as subordinate.More specifically, men were more likely to be portrayed as primary visual actors and narrators than women whilst women were more likely to be presented in indoor settings and at home.In an investigation of race, class and gender undertones in the authorial voice in print advertisements, evidence of paternalism and male condensation was found (Maake 2006).Even though women were used as central characters, their authority tended to be marginalised by male voices.African women were particularly portrayed in stereotypical roles (e.g.carrying babies, performing domestic work and preoccupied with being an eligible bride).Depictions of women as carers and men as protectors were also identified.There was also a tendency to give more prominence to the advertised product than to the female models.This was not the case in advertisements featuring men.Women also tended to be foreshadowed by the work they performed or the services they provided to others.Further, women were rarely portrayed alone; they were either in family settings or performing household work.In family settings, gender hierarchy tended to be palpable with emphasis on women being the service providers.Goffman's (1979) view of gender representations in advertisements is that women and men are portrayed in social scenes in which power relations may be evident.Goffman proposed six categories for analysing portrayals in advertisements, namely relative size, feminine touch, function ranking, family scenes, ritualisation of subordination and licensed withdrawal, as explained in Table 1.These categories have been widely adopted and used by a number of researchers (e.g.Belknap & Leonard 1991;Hovland et al. 2005;Kang 1997;McLaughlin & Goulet 1999;Sirikaya & Sonmez 2000).Additional categories have also been proposed.Kang (1997), for example, added two additional categories to Goffman's framework, namely body display (are models dressed provocatively or appearing nude?) and independence or self-assurance (do models give the overall impression of independence?).Facial and body exposure communicates social and cultural messages.According to Archer et al. (1983), showing the face emphasises personality and character whilst not showing the face hides personality and character.Similarly, showing only the body or parts of the body dehumanises and objectifies the person (Cortese 1999;Kilbourne & Jhally 2000).

Socialisation effects of portrayals in advertising
Two related theories of socialisation, namely social learning theory (Moschis & Churchill 1978) and cultivation theory (Gerbner et al. 1986;Morgan & Signorielli 1990), are central to this study.Social learning theory postulates that the ideas and behaviours observed in socialisation agents such as advertisements shape people's attitudes and behaviours (Bandura 1977) and normalise what is observed (Nelson 2001).Repeated exposure to the ideas or observed behaviours increases the rate of retention and the impact (Clark, Martin & Bush 2001;Drèze & Hussher 2003), especially if these ideas and behaviours are perceived as superior.The use of social learning theory in the analysis of portrayals in advertising has been employed in other studies (e.g.Bailey 2006;Clark et al. 2001;McCullick et al. 2003;Skill & Wallace 1990).Cultivation theory proposes that perceptions of social reality are influenced by the kind of beliefs and attitudes people acquire as a result of long-term exposure to and interaction with socialisation agents such as advertisements (Gerbner et al. 1986;McQuail 2000;Morgan & Signorielli 1990).It also suggests that how people are portrayed may influence how they come to perceive themselves and relate to others in society (Shrum 1996).Based on the potential and actual impact of the media as explained by the two theories, it is important to understand gender portrayals in advertisements for practice and public policy reasons.

Research objective and hypotheses
The objective of this study is to determine how print advertisements in Drum magazine portray women in comparison to men.More specifically, the study seeks to assess the effect of gender on the profile (roles) assigned to women and men in advertisements, gender equity in the portrayals in advertisements and the use of sexual appeals in advertisements by testing the following null hypotheses: 1

Methodology Sampling procedure
The sample universe for the study was all the advertisements that appeared in Drum magazine in the period 1982-2006 that featured people whose gender could easily be determined.For this reason, advertisements featuring only children were excluded.Advertisements for clothing items where models were not involved in any activity other than posing to show off the clothing were excluded because they generally provided insufficient information regarding normal human activities.Advertisements also had to be large enough to be examinable.Advertisements therefore had to be at least one quarter of the A4 magazine page to be considered.To meet the criterion of representativeness, multistage sampling was employed.First, the years were selected and then months were sampled from the years.The following years were selected for inclusion : 1981, 1982, 1985, 1987, 1990, 1992, 1994, 1996, 1998, 2000, 2002, 2004 and 2006.The selection of the years and months from which magazines were sampled was largely influenced by availability and the condition of magazines.When a particular year could not be used, the following year was substituted for it.Once a year was selected, the next step was sampling the months from the years selected.The months were selected randomly using a random number table.For each month picked, a census of qualifying advertisements was included in the sample.Additional months were sampled until at least 25 advertisements were sampled.This sampling procedure resulted in the sample of 415 advertisements, excluding duplicate advertisements.In this study, duplicate advertisements (defined as those that are similar in all aspects excluding variations in size and colour) were coded only once in each year that they appeared.

Category development
In  Cortese's (1999), and Kilbourne and Jhally's (2000) categories were employed.Also included in the coding instrument was the use of models as narrators or spokespersons (Bell et al. 2000).

Coding procedure
The assistance of three university students of advertising was employed.To ensure that they worked independently from each other, the coders were unknown to each other and were trained separately.At the end of the training session each coder was provided with three advertisements and codebooks to content analyse on their own.Disagreements between the coders were resolved by discussion involving the researcher and the coders.At the end of this process, two coders were employed whilst the third was dropped for consistently providing unreliable coding.The remaining two coders were each asked to content analyse a convenience sample of 39 (7% of the total sample size) advertisements independently of each other.According to Lewis and Neville (1995), it is acceptable practice to check for validity and reliability by having approximately 5% of the advertisements recoded by independent coders.The percentage of agreement between the researcher and the coders is one way of checking for coder reliability (or internal reliability) (Krippendorf 1980).Using Holsti's (1969) definition (inter-coder reliability is the overall percentage of agreement between coders) and formula (Reliability = 2M/(N 1 + N 2 ), where M = total number of decisions on which the coders agree, N 1 = total number of decisions made by coder 1 and N 2 = total number of decisions made by coder 2), an average percentage of agreement of 92.6% was realised.This is well above the recommended minimum 80% (Kassarjian 1977).Based on this level of reliability, it was assumed that the clarity of categories and category dimension in the codebook or code sheet was acceptably high to proceed with the coding.

Analysis and results
The overall sample size used in this study was 415 advertisements.These advertisements fell into two groups: those featuring models of one gender only (294 advertisements accounting for 70.8% of the sample) and those featuring both women and men (121 advertisements accounting for 29.2% of the sample).For single gender, the gender breakdown of the advertisements was 169 for women only and 125 for men only.The data were analysed by partial least squares (PLS) analysis using the programme SmartPLS 2.0 (Ringle, Wende & Will 2005).The following section on measurement (external) model estimates presents statistics on which of the variables used to measure the latent variables were significant.The section on structural (internal) model estimates shows the extent to which gender predicts the latent variables profile, gender equity and sexual appeal.

Measurement (external) model estimates
To test the statistical significance of the indicator variables on the endogenous latent variables, t-values were generated using the bootstrap function of SmartPLS 2.0 (Ringle et al. 2005).For single gender ads (see Table 3), the t-values show that of the seven indicator variables for the latent variable profile, four were statistically significant: Setting (location) (p < 0.05), Family scene (p < 0.01), Product type (p < 0.01) and Typical users (p < 0.01).For the latent variable gender equity, none of the indicator variables was statistically significant (p < 0.05).Both the indicator variables used to measure sexual appeal, namely Dressing (p < 0.01) and Body display (p < 0.05), were statistically significant.
The path coefficients in Table 3 to Table 6 show the predictive power of the independent variable and may be interpreted in the same way as correlation or regression coefficients (McIntosh et al. 1994).A positive path coefficient indicates that a unit increase in the independent variable leads to a direct increase in the dependent variable in line with the size of the coefficient, in other words high scores on the independent variable predict high scores on the dependent variable & Jacobs 1986; McIntosh et al. 1994).Conversely, a negative coefficient indicates that a unit increase in the independent variable leads to a direct decrease in the dependent variable in line with the size of the coefficient, in other words low scores on the independent variable predict high scores on the dependent variable (Eccles & Jacobs 1986;McIntosh et al. 1994).Because gender was coded 1 for women and 2 for men, the negative path coefficients mean that women have higher scores than men on the predicted variables.According to Cohen's (1988) guidelines for interpretation, effect sizes may be defined as either small (r = 0.10), medium (r = 0.30) or large (r = 0.50).
For advertisements featuring both genders (see Table 4), eight of the nine indicator variables for profile were statistically significant according to the t-values.For gender equity, none of the nine indicator variables showed a statistically significant effect (p < 0.05).Finally, for sexual appeal, only one indicator was statistically significant (p < 0.05).

Structural (internal) model estimates
In single gender advertisements gender had a direct effect on all three endogenous latent variables, namely profile TABLE 2: Category dimensions for the physical characteristics of the message source.
For ads featuring both women and men, gender accounts for 12.8%, 17.9% and 5.4% respectively of the variance observed for profile, gender equity and sexual appeal (Table 6).However, only the total effects on profile and sex appeal were statistically significant (p < 0.01).Therefore, the null hypotheses H 1 and H 3 were rejected whilst the null hypothesis H 2 could not be rejected.
In both single gender advertisements and advertisements featuring women and men, the structural (internal) model estimates revealed that gender had statistically significant total and direct effects on profile and sexual appeal but not on gender equity.Therefore, null hypotheses H 1 and H 3 were rejected (i.e.gender influences both the profile and the use of sexual appeal).On the other hand, H 2 could not be rejected, suggesting that gender does not influence gender equity (not surprising given the non-significant indicator variables).

Discussion and conclusion
This study set out to determine the extent to which gender influences the profile assigned to models, gender equity in the way models are displayed and the employment of sexual appeal in print advertisements.Central to interpreting the findings of this study are the theoretical underpinnings in the social learning and cultivation theories which show that people learn from socialisation agents such as print advertisements.
The study found that for both single gender advertisements and advertisements featuring men and women together, the indicator variables used to measure gender equity were not statistically significant, indicating that the measurement items used were not significant predictors of gender equity.Consequently, null hypothesis H 2 could not be rejected.For single gender advertisements, gender correlated positively with profile and negatively with sexual appeal; hence, the null hypotheses H 1 and H 3 were rejected.The positive correlation with profile suggests that men are more likely to be absent from family scenes, home settings, housekeeping products and products targeted at women only.This is consistent with expectations based on prior research that show that men are more likely to be portrayed in masculine or instrumental roles whilst women on the other hand are more likely to be portrayed in expressive or feminine roles.The negative correlation between gender and sexual appeal suggests that women are more likely to be presented in suggestive or partially clad dressing and to have parts of their bodies focused on or amplified.This is consistent with the objectification of women.It is evident therefore that in single gender advertisements, there is some stereotyping even though the effect sizes range from small for sexual appeal to medium for profile.In advertisements featuring both genders, there is a significant negative correlation between gender and profile (medium effect size) and between gender and sexual appeal (small effect size).The negative correlation between gender and profile suggests that women are, amongst other things, more likely to be absent from family scenes, home settings, housekeeping products and products targeted at men only.This is difficult to explain in light of the finding in single gender advertisements.However, it appears that there is a difference in role portrayals depending on whether or not models of only one gender are employed in an advertisement.
To sum up, the roles assigned to models and the use of sexual appeal in advertisements is influenced by the models'   gender.Advertisers therefore treat models differently with regard to the roles assigned to models and the use of sexual appeal.Previous research evidence and theory indicate that the content of advertisements does influence the readers' self-perception and also how they perceive others.Within the constraints imposed on advertisers by the need for advertising creativity, for example, it is recommended that advertisers pay particular attention to the way they portray models and relationships in advertisements to avoid images that show women in a manner that subordinates and denigrates them.Whilst there can be no one common standard regarding what comprises negative portrayals in a multicultural context like South Africa, it is incumbent upon the advertisers to negotiate the subtleties involved to avoid externally imposed pressure on the sector for change.
This study has two main limitations.Firstly, although the decision to use Drum was a deliberate one, the decision limits the extent to which the results can be generalised to other South African magazines.Even though advertisements in Drum feature models of other ethnic groups, there is a possibility that the results could have been different if magazines targeted at different readers had been included in the study; hence, the recommendation is that future studies should include a wider range of South African magazines.Secondly, the use of dichotomous category dimensions for some of the categories included in the coding instrument could have reduced the explanatory potential of the data collected.For example, instead of coding the types of products advertised into housekeeping and non-housekeeping products, increasing the number category dimensions could have yielded more informative data.

TABLE 1 :
Explanation of coding categories based on Goffman'.
Ritualisation of subordinationPerson shown in positions and poses that imply inferiority or deference such as bowing or lowering oneself.Person shown on floor or bed or in other spatially lower positions.Person of other gender elevated above person.Person in bashful knee bend or leaning on someone.Person in canting posture where head or body is tipped lower.Person is smiling in response to others.Person is dressed like a child or posed like a child.Person is prey in mock assault or teased or being held possessively (around shoulder or hand).Licensed withdrawal Person shown as withdrawn from the scene or 'tuned out'.For example, person covers mouth or face with hands, turns gaze from others, with middle distance looks, dreamily talking on phone, emotional displays, snuggling or nuzzling others, being supported by others as in grief embrace.Person is 'tuned out' from the scene because they feel protected.Source: Hovland, R., MacMahan, C., Lee, G., Hwang J.-S. & Kim, J., 2005, 'Gender role portrayals in American and Korean advertisements', Sex Roles 53(11/12), 887-899

Table 2 .
Kang's (1997979) latent variables, namely gender, profile (i.e. the roles assigned to women and men in print advertisements), gender equity (i.e. as measured by the differential gender display of models in advertisements) and sexual appeal (i.e. the extent to which sexual appeal is exploited in advertisements), The latent variable gender equity in the use of women and men in print advertisements was assessed by nine categories outlined in Table2.These categories were mainly based onGoffman's (1979)framework for analysing gender portrayals.In addition,Kang's (1997),  Archer et al.'s (1983), Nature of narration, Ritualisation of subordination, Narrator's roles, Relative size, Target of narration, Feminine touch and Licensed Withdrawal.To measure the construct sexual appeal, three variables (Facial exposure, Body parts and Dressing) were used.Regarding the latent variable profile, eight manifest variables or categories were investigated as outlined in

variables Predictor variable Total effects R 2 t-value
**, Significant at p < 0.01