Questions concerning the effectiveness of advertisements that use or break traditional gender roles and the predictive value of gender attitudes in determining their effectiveness are not new (Eisend 2010; Grau and Zotos 2016; Wolin 2003). Yet most investigations to date have returned mixed findings (Zawisza and Cinnirella 2010), used often limited measures of gender attitudes and were conducted in Western countries (Zawisza et al. 2016). In an increasingly globalized market it is crucial to consider cross-national comparisons. These allow for testing the extent to which two key theories, the stereotype content model (SCM) and ambivalent sexism theory (AST), hold when applied to advertising in different countries. Two unanswered questions are the focus of the present paper: (a) Does the greater effectiveness of traditional paternalistic vs. nontraditional envious female portrayals in advertising hold across countries? and (b) Does the potential predictive value of gender attitudes differ as a function of sexism type (hostile and benevolent sexism), advertisement type, and country? In order to address these questions three samples from countries that vary in gender egalitarianism were examined: Poland (PL), South Africa (SA) and the United Kingdom (UK) (Zawisza et al. 2015). By recognizing the complex nature of sexist attitudes and by testing their predictive values cross-nationally, we address a key gap and inconsistencies in the literature on global gendered advertising. In doing so, we also contribute to our understanding of theory and its application to an advertising context.

Gendered Advertising Effectiveness across Cultures

Although the use of femvertising (i.e., advertising that shows women in higher status and empowered roles) is gradually increasing, content analyses show that women are still predominantly portrayed in traditional lower status and disempowered roles (Grau and Zotos 2016). There is some evidence of this pattern cross-nationally (Furnham and Paltzer 2010; Matthes et al. 2016). Is, however, such a traditional advertising approach effective? Research focusing on female gender roles in advertisements returns mixed results, ranging from greater preference for traditional portrayals for various products (Duker and Tucker 1977), equal effectiveness of (realistic) traditional vs. nontraditional portrayals in TV advertisements for food and cleaning products (as measured with attitude toward advertisements; Whipple and Courtney 1980), to greater effectiveness of nontraditional advertisement strategies for car repair services (as measured with attitudes toward advertisement, spokesperson, and purchase intent; Bellizzi and Milner 1991) and food (as measured with attitudes toward advertisement and purchase intent; Jaffe and Berger 1994). Recent research from various countries appears to provide more consistent results that report greater effectiveness of traditional (e.g., housewife) vs. nontraditional (e.g., businesswoman) advertisement types for mineral water (Zawisza and Cinnirella 2010), orange juice (as measured with affective and cognitive responses to the advertisement and purchase intent; Zawisza et al. 2016) or unisex perfume (as measured with purchase intent; Infanger et al. 2012). Infanger and Sczesny (2015) report similar results (as measured with purchase intent) albeit through the use of a different methodology that makes comparisons across advertisement types difficult.

Drawing firm conclusions, especially from early findings, is difficult due to methodological differences (e.g., in operationalizations of traditional and nontraditional portrayals of women, types of products, channels of communication, and ways of measuring advertising effectiveness). Yet the more recent literature seems to agree that greater preference for advertisements utilizing traditional female portrayals is due to their perceived greater warmth (Infanger et al. 2012; Infanger and Sczesny 2015; Zawisza and Cinnirella 2010). Warmth is one of two key dimensions of social perception proposed alongside competence by the SCM (Fiske et al. 2002). Warmth informs us of the intentions of others (helpful vs. harmful) whereas competence indicates their ability to enact these intentions. Four types of stereotypes result from the dimensions of warm and competence: paternalistic (people who are warm/ liked but incompetent/ disrespected), envious (those who are competent/ respected but disliked/ envied), contemptuous (disliked and incompetent), and admiration (liked, competent and, thus, admired). Traditional female subtypes, such as housewives or secretary, portray paternalistic stereotypes. Nontraditional female subtypes, such as career women, feminists or intellectuals, portray envious stereotypes (Eckes 2002). Our work focuses on such traditional/paternalistic (housewife) and nontraditional/envious (businesswoman) portrayals of women in advertising. The inherent association between warmth and liking may explain why warmth appears as a key determinant of advertising effectiveness (Du Plessis 2005), and hence, why characters perceived as warmer (e.g., housewife) result in higher advertising effectiveness than their colder counterparts (e.g., businesswoman; Infanger et al. 2012; Zawisza and Cinnirella 2010).

Yet it is uncertain whether the warmth advantage of the housewife advertisement strategy will hold cross-culturally. This is an important question given increasingly global markets. The studies we cited were predominantly conducted using Western samples and little is known of the effectiveness of such advertising strategies outside this broad cultural context. Because the SCM appears to apply across various social groups and nationalities and is now considered a universal model of social (Abele and Bruckmuller 2011; Cuddy et al. 2008; Cuddy et al. 2009) and brand (Kervyn et al. 2012) perception, we hypothesize that the traditional, paternalistic housewife advertisement strategy will be more effective than the envied businesswoman advertisement strategy, irrespective of country (Hypothesis 1). Indeed, such cross-cultural preference for the paternalistic over the envious advertisement strategy has been reported for male portrayals of househusband and businessman (Zawisza et al. 2016). However, no known research has tested this possibility for female portrayals. Moreover, Zawisza et al. (2016) largely focus on affective and cognitive responses to these advertisements as opposed to purchase intent as such. To test Hypothesis 1, three countries were chosen that differ in terms of national levels of sexism with the United Kingdom being relatively most gender egalitarian, Poland moderate, and South Africa least gender egalitarian (Zawisza et al. 2015).

Sexism and Advertising Effectiveness

Intuitively, gender attitudes should determine the effectiveness of gendered advertisements. Yet previous research returns mixed results. Gender role expectations (Putrevu 2004), gender identity (Morrison and Shaffer 2003), belonging to feminist organizations (Ford and Latour 1993), career vs. homemaker orientation (Barry et al. 1985), and gender role ideology (Baxter et al. 2016) have all been found to positively predict the effectiveness of such advertisements. On other occasions, such variables have not proven predictive (Bellizzi and Milner 1991; Duker and Tucker 1977; Whipple and Courtney 1980). More recent studies, which focus on sexism specifically, are no more conclusive. Zawisza and Cinnirella (2010) reported that attitudes toward women predicted only affective responses to gendered advertisements among British respondents but not their purchase intent. The authors, however, used old-fashioned gender attitude scales (Parry’s 1983, adaptation of Spence and Helmreich’s Attitudes Toward Women Scale for use in the UK) and dichotomized this continuous variable. Although they do not relate their findings directly to advertising effectiveness, Infanger et al. (2012) used the more up-to-date Ambivalent Sexism Inventory (Glick and Fiske 1996). They found benevolent sexism predictive of Swiss respondents’ positive reactions to traditional female advertisement characters whereas hostile sexism was predictive of more negative responses to nontraditional female advertisement characters. The latter was only the case when respondents were assessed under time pressure (i.e., when cognitive resources responsible for impression management were limited).

Our study points to two important observations. First, and consistent with AST, the two ambivalent forms of sexism maintain the status quo by rewarding compliance with traditional gender roles via benevolent attitudes and by punishing noncompliance with these roles via hostile attitudes. According to AST (Glick et al. 2000), traditional—or sexist—gender attitudes are not uni-dimensional. They are not simply negative but rather ambivalent: They consist of both evaluatively negative (hostile) and positive (benevolent) components. Hostile sexism (HS) indicates antipathy toward women who “are perceived as seeking to control men, whether through sexuality or feminist ideology” (Glick and Fiske 2001, p. 109). Benevolent sexism (BS), on the other hand, is characterised by positive but still patronizing beliefs about women (Glick and Fiske 2001). Together they maintain the status quo through a stick-and-carrot mechanism as seen, for example, in the reactions to advertisements in Infanger et al.’s (2012) Swiss sample. Similar findings were reported by others for both men (Sibley and Wilson 2004, New Zealand) and women (Becker 2010, Germany). Men reacted with increased BS and decreased HS to traditional and positive female subtypes in terms of their sexuality (i.e., Madonna type) but an increase in HS and decrease in BS was observed in their responses to nontraditional and negative female subtypes (e.g. whore type). Interestingly, Becker (2010) reported similar tendencies among German women who endorsed higher HS when thinking about nontraditional female subtypes (e.g., feminist or career women) but higher BS when thinking about traditional ones (e.g., housewife). We therefore predict that the stick-and-carrot functions of sexism will also affect the effectiveness of advertisements that utilize (non)traditional female portrayals. Specifically, HS will predict lower purchase intent for the nontraditional envied businesswoman portrayal (Hypothesis 2a) and BS will predict higher purchase intent for the traditional paternalistic housewife portrayal (Hypothesis 2b). We therefore extend our focus beyond Infanger et al.’s study to examine the effect of (non)traditional (envious vs. paternalistic) portrayals on purchase intent.

Second, the fact that in the context of advertising the findings pertaining to hostile sexism were only achieved under time-constraint conditions (Infanger et al. 2012) suggests that the measure is vulnerable to social desirability effects. These may be higher in cultures with higher egalitarian norms and thus have practical implications for globalized markets. Specifically, tools measuring sexisms may have limited predictive value in highly gender-egalitarian countries. Thus, a question remains whether the predictive role of sexism in advertising effectiveness will depend on culture.

Of relevance, AST has been shown to hold cross-nationally (Glick and Fiske 1996, 1999; Glick et al. 2004; Glick et al. 2000). HS and BS, being complementary ideological systems, have been found to correlate positively across 19 countries (Glick et al. 2000). The studies we cite supported the status quo-maintaining functions of BS and HS, even in the case of samples characterised by their relative gender egalitarianism, for example, Swiss students (Infanger et al. 2012), New Zealand male students (Sibley and Wilson 2004) and a general German female sample (Becker 2010). However, levels of hostile and benevolent sexism, as well as egalitarian norms, vary across countries (Zawisza et al. 2012, 2015). Sibley et al. (2007) argue that national levels of men’s BS depend upon the degree of social threat experienced in the society whereas their levels of HS depend on how competitive the society is (e.g., unequal and short in resources). They further argue that women too may embrace patriarchy under social threat, as evidenced in their sexism, “because the system provides at least some form of security (albeit only indirectly through men), especially when alternative sources of power and resources are not readily available” (p. 745). Glick and colleagues (Glick and Fiske 2001; Glick et al. 2000) similarly argue that, in highly sexist countries, women depend on men to a higher degree and the consequences of women challenging the status quo are more severe. Thus, women in such countries embrace BS to a greater extent than those in less sexist countries. Together, this research suggests that both men and women may embrace both forms of sexism more in less egalitarian countries. Therefore, we expect that the predictive power of HS and BS for advertising effectiveness will depend on country such that Hypotheses 2a and 2b should hold stronger for less gender egalitarian countries (i.e., PL and SA) than for more egalitarian ones (i.e., the UK; Hypothesis 3).

With this comparison in mind, our samples were chosen to come from the three countries (PL, SA and the UK) because they met key criteria for our comparisons. First, these countries differ demonstrably in their levels of sexism where the UK is relatively egalitarian, PL moderately so, and SA relatively non-egalitarian (Zawisza et al. 2015). The UK, as the most developed and stable democracy, served as a useful comparison against the two countries undergoing transition to democracy over a similar period of time (i.e., since the late 1980s). This criterion was important because democratization has been linked to liberalization (Inglehart and Norris 2003). Yet liberalization in the SA and PL followed from substantively different political and value systems. For example, opposition to apartheid in SA encouraged consideration of women’s rights as evidenced in its new constitution. Opposition to communism in PL on the other hand resulted in marginalised women’s rights due to their perceived association with communism’s forced emancipation (LaFont 2001). Thus a comparison of these three countries offers a unique cultural insight. To the authors’ knowledge ours is the first investigation of its kind and as such it presents a valuable and novel contribution to our understanding of gendered advertising globally.

Method

Participants

For the British sample, 158 participants were recruited from Winchester and Anglia Ruskin universities where 74 (47%) were female and 84 (53%) male. Their age averaged 22.15 years-old and ranged from 18 to 48 (SD = 7.29). They identified as White (136, 86%), Asian (11, 7%), Black (3, 2%), and other (8, 5%). For the Polish sample, 121 participants were recruited from University of Gdansk where 59 (49%) were female and 62 (51%) male. Their age averaged 21.96 years-old and ranged from 19 to 29 (SD = 2.88). All participants identified as White. For the South African sample, 171 participants were recruited from Cape Town University where 94 (55%) were female and 77 (45%) male. Their age averaged 19.89 years-old and ranged from 18 to 34 (SD = 2.13). They identified as White (94, 55%), mixed (29, 17%), Black African (25, 15%), Asian (21, 12%), and other (2, 1%). For each sample, participants were recruited by announcements distributed on-campus and via e-mail. Participants were remunerated a local equivalent of £5 for their participation.

Procedures, Design, and Measures

Participants were told that the study examined individual responses to different advertisements. Students participated either individually or in groups of a maximum of six. Each was provided with a questionnaire booklet including two printed advertisements: a traditional one and a nontraditional one (i.e., a within-subject manipulation). There were two versions of each advertisement type (i.e., Businesswoman: Bw1 and Bw2 for nontraditional and Housewife: Hw1 and Hw2 for the traditional). Participants were provided with one of eight possible advertisement combinations (e.g., Bw1 and Hw2), the order of which was counterbalanced. All participants were asked to evaluate the advertisements after which they completed measures of sexism and were fully debriefed. All the scales were back translated from English by independent translators for the Polish sample. Any ambiguities in the translations were resolved through discussions.

Advertisement Type

The women portrayed in the two sets of printed advertisements were carefully pre-tested. The envious nontraditional female portrayal set (housewives, or Hw) was perceived as nontraditional and masculine whereas the paternalistic traditional female role portrayal set (businesswomen, or Bw) was viewed as traditional and feminine (n = 18 students who did not participate in the main study and who rated these models on sematic differential scales anchored −3 to 3; all ps < .001). The characters were also matched in terms of attractiveness. With a similar scale anchored −3 (very unattractive) to 3 (very attractive); p = .686). (See Table 1s in the online supplement for a summary of descriptive and inferential statistics comparing the two ad sets. The online supplement also reports manipulation checks confirming similar perception of the Hw set as significantly more traditional and less liberal than the Bw set in all three countries.)

As noted previously, two versions of the advertisements depicted nontraditional portrayals whereas two depicted traditional portrayals. These were printed and prepared especially for the purpose of the experiment and are available on request. The nontraditional Bw1 advertisement portrayed a blond woman in her early 30s wearing a grey suit, carrying a brown briefcase, and walking against a non-descript outdoor setting (the background was blurry and the focus was on the woman). The Bw2 advertisement portrayed another smartly dressed brunette woman of similar age, holding an open brown briefcase from which she was reading while walking. Again, she wore a gray suit. The traditional Hw1 advertisement portrayed a casually dressed woman in her mid-30s in a kitchen preparing cookies. The traditional Hw2 advertisement portrayed the same woman, in the same setting, performing the same activities but also interacting with a child (a girl of about 10-years-old).

Other features of the advertisements were kept constant: in all cases the women were White and portrayed from their hips up looking away from the camera. The heading variably read: “Mothers/Professional women agree: until you try new X/Y orange juice you will never know what a real orange juice tastes like.” The product (i.e., a glass of orange juice surrounded by sliced oranges) was positioned in the middle right section of the advertisements. Orange juice was selected as the advertised product due to being neutral and unisex, and the brands were specifically developed for the current study to avoid issues with brand familiarity (i.e., brand “X” for the traditional advertisement and “Y” for the nontraditional one). The orange juice received a mean score of −.78 (SD = 1.11) and a modal score of 0 on the Product Gender Scale anchored which ranged from −3 (feminine) to +3 (masculine). It also received a mean score of 2.56 (SD = .98) and modal score of 2 on the Product Involvement Scale anchored from 1 (product requires little thought when purchasing) to 7 (a lot of thought).

Ambivalent Sexism

Glick and Fiske’s (1996) Ambivalent Sexism Inventory was used to measure sexist attitudes toward women. This is arguably one of the more sensitive explicit measures of sexism currently available (Glick and Fiske 1996; Glick et al. 2004; Glick et al. 2000). This tool uses a 6-point Likert-type response format from 0 (disagree strongly) to 5 (agree strongly), where half of the 22 items measure hostile (e.g., “Most women interpret innocent remarks or acts as being sexist”), and half measure benevolent (e.g., “Every man ought to have a woman that he adores”), sexism. After reverse coding as needed, the higher the average score, the higher benevolent or hostile sexism toward women. Evidence for the ASI’s discriminant and convergent validity has been provided by Glick and Fiske (1996), for its cross-national validity by Glick et al. (2000), and for its convergent and divergent validity in British samples by Masser and Abrams (1999). Evidence of invariance in PL, SA, and the UK is reported in Zawisza et al. (2015). Reliability coefficients for the HS scale range between .68 and .89 and for BS between .53 and .88 in cross-national samples (Glick and Fiske 1996; Glick et al. 2000; Masser and Abrams 1999; Viki et al. 2003; Zawisza et al. 2015). Cronbach’s alphas representing reliability of the measurement in the current studies are presented in Table 1.

Table 1 Descriptive statistics, correlations, and alphas for study variables within country
Full size table

Purchase Intent

A single-item declaring purchase intent likelihood was used. This item required participants to indicate the probability (0–10) that they would buy the advertised product.

Results

Analytic Strategy

To test Hypothesis 1, a 2 (Ad Type: Housewife or Businesswoman) × 2 (Gender: Men and Women) × 3 (Country: United Kingdom, Poland, South Africa) ANOVA was run with purchase intent as the dependent variable. We also used this test to determine if gender needed to be considered in further regression analyses with the continuous variable of sexism. Subsequently, and in order to test our hypotheses regarding sexism (Hypotheses 2a and 2b), a regression model with repeated measures for advertisement type was computed in line with Judd et al. (2001) recommendations. Participants’ gender, hostile sexism (HS), benevolent sexism (BS), and interactions between gender and both sexisms were entered as predictors. These were regressed on purchase intent (the criterion variable) in response to two experimental conditions (Housewife–Hw and Businesswoman–Bw advertisement types), as well as the difference between these advertisement conditions (Hw – Bw). A difference above zero indicates that purchase intent is stronger in the Hw (vs. Bw) advertisement condition, and vice versa (i.e., if the difference is below zero the purchase intent is stronger for Bw than for Hw). Thus the analysis allowed for testing all second and third order interactions between the variables of interest (advertisement type, sexism, and potentially gender).

Following Judd et al.’s (2001) and Jaccard and Turrisi’s (2003) suggestions, analyses were completed in several phases. First, means across both experimental conditions were compared in order to assess the overall main effect of the manipulation of advertisement type. Second, for the first criterion variable (e.g., purchase intent in Hw advertisement condition), stepwise regression was computed. In the first step, centered HS and BS were treated as input variables. Gender was inserted as a predictor in the second step. In the third step, both interaction products (HS x Gender; BS x Gender) were entered. The third phase involved repeating procedures from Phase 2 for the remaining criterion variables (e.g., purchase intent for Bw and the Hw – Bw purchase intent difference). If the interaction effect was significant (as indicated by a significant β value for the interaction term), further simple slope analyses were conducted following Cohen et al. (2003) recommendation. All steps described here were performed for each of the three country samples separately after which beta values were compared across the three countries to test Hypothesis 3.

Hypothesis Testing

Hypothesis 1 and Gender Effects

The three-way mixed design ANOVA revealed a significant main effect for advertisement type, F(1, 444) = 13.68, p < .001, ηp2 = .030, consistent with Hypothesis 1. The paternalistic Hw advertisement strategy resulted in higher purchase intent (M = 4.76, SD = 2.48) than the envious Bw advertisement strategy (M = 4.18, SD = 2.67). There also was a significant effect of participants’ gender, F(1, 444) = 11.52, p < .001, ηp2 = .025, such that women had higher purchase intent (M = 4.87, SD = 2.57) than did men (M = 4.24, SD = 2.46). A significant main effect of country emerged, F(2, 444) = 34.16, p < .001, ηp2 = .020, and post hoc tests revealed that the British respondents had lower purchase intent (M = 4.12, SD = 2.69) than did the Polish (M = 4.08, SD = 2.47, p = .003) and the South African ones (M = 4.54, SD = 2.50, p = .050) but there was no significant difference between the latter two groups (p = .232).

Lastly, there was a significant Gender x Country interaction, F(2, 444) = 7.73, p < .001, ηp2 = .034. When followed up, the independent, two-tailed, t-tests, indicated that British women had higher purchase intent to both advertisement types than did men, t(156) = 2.84, p = .005, d = .46, and the same difference held for Polish women and men, t(169) = 3.82, p < .001, d = .59 (see Table 1). There was no similar gender difference in the South African sample. Based on these findings, we decided that gender needed to be entered in our moderated regression analyses testing the role of sexism in the effectiveness of the advertisements. See Table 1 for descriptive statistics and correlations within each country.

Hypothesis 2: United Kingdom

For the purpose of testing Hypotheses 2a and 2b for each country, we followed statistical procedures recommended by Judd et al. (2001). Analyses were conducted using regression models. First a model for purchase intent in the Hw advertisement condition (used as a criterion variable) was computed where centered viewers’ HS and BS were treated as input variables. The model was significant (see Table 2). Viewers’ HS was the only significant predictor: Contrary to Hypothesis 2a and Hypothesis 2b, the higher the level of hostile sexism, the stronger the purchase intent for the Hw advertisement type (see Table 2). In the second step, gender was entered into the model. It was found that British women declared stronger purchase intent for Hw advertisements than did men. In the last step, two interaction products (HS x Gender; BS x Gender) were entered into the model. Although the full model was still significant, Adj. R2 = .12, F(2, 152) = 5.31, p = .006, they did not contribute significantly to the explained level of variance.

Table 2 Regressions predicting purchase intentions from sexism and gender within country and advertising type
Full size table

The same steps were performed for purchase intent in the Bw advertisement condition. Entering both sexism variables did not increase the explained variance significantly (F < 1), but adding gender to the model revealed a significant effect (see Table 2). British women declared stronger purchase intent for the Bw advertisement type than did British men. In the last step, the two interaction products (HS x Gender; BS x Gender) were entered into the model returning similar null results as for Hw purchase intent; however, this time the regression model lost its significance. Finally, the Hw – Bw purchase intent difference was treated as a criterion variable, but the regression models were not significant (see Table 2).

Hypothesis 2: Poland

Predictors were regressed on purchase intent for the Hw advertisement type. In the first step centered HS and BS were treated as predictors (see Table 2). We found that, in line with Hypothesis 2b, the higher the BS, the stronger the purchase intent for the Hw advertisement. In the second step, gender was inserted into the model. It appeared that Polish women declared stronger purchase intent for this advertisement type than did Polish men. In the third step, both interaction products (HS x Gender; BS x Gender) were added to the model, but a significant increase in explained level of purchase intent was not observed. However, the interaction between HS and Gender in the Hw advertisement condition was significant—full model: Adj. R2 = .18, F(5, 115) = 6.22, p < .001 (see Table 2).

Simple slopes analysis was performed in order to test this interaction. It was revealed that HS positively predicted purchase intent for the Hw advertisement type, β = .38, t(118) = 3.07, p = .003, but only among Polish women. Moreover, among highly hostile Polish individuals, women responded with stronger purchase intent to the Hw advertisement type than did men, β = .49, t(168) = 3.91, p < .001. The interaction is plotted in Fig. 1.

Fig. 1
figure 1

Interaction between Gender x Hostile Sexism (HS). Criterion variable was purchase intent measured in Husewiffe (Hw) ad condition in Poland

Full size image

Third, predictors were regressed on purchase intent for the Bw advertisement type. The first and the second steps of the analysis revealed significant effects. It appeared that, contrary to Hypothesis 2b, the stronger the BS, the higher the purchase intent for the Bw advertisement type. Additionally, Polish women declared stronger purchase intent for this advertisement type than did Polish men. A significant interaction effect was not detected. The full model was however significant (see Table 2 for details). Lastly, when the Hw – Bw difference in purchase intent between in the two experimental conditions was treated as a criterion variable, no significant effects were obtained (see Table 2).

Hypothesis 2: South Africa

Analyses were performed using a regression model. In the first step, centered HS and BS were added to the model; BS was the only significant predictor of purchase intent in the Hw advertisement condition. In line with Hypothesis 2b, the higher this type of sexism, the stronger the purchase intent in the Hw advertisement type condition. Adding gender in Step 2, and the two interactions products (HS x Gender; BS x Gender) in Step 3, returned no significant effects (Fs < 1). Moreover, the three-step model lost its significance, but BS was still a significant predictor of purchase intent for the Hw advertisement type (see Table 2 here for Steps 1 and 2; Table 2s in the online supplement for Step 3).

The same three steps were performed for purchase intent for the Bw advertisement type. After inserting centered HS and BS, it was observed that, in line with Hypothesis 2a, the higher the hostile sexism, the weaker the purchase intent for the Bw advertisement (see Table 2). In the second step, gender was added to the model, but this predictor did not reach significance (p = .08). In the third step, the two interaction products (HS x Gender; BS x Gender) were added to the model. A significant interaction between gender and HS was observed, ∆R2 = .05, F(5, 165) = 2.92, p = .015.

In order to test the nature of the interaction, we performed simple slopes analyses (Cohen et al. 2003). We found that among highly hostile sexist individuals, South African women had weaker purchase intent for the Bw advertisement type than did South African men, β = .28, t(168) = 2.24, p = .026. Additionally, it was revealed that for South African women, HS negatively predicted purchase intent in response to this advertisement, β = .49, t(168) = 2.85, p = .005. No other slopes were significant. The interaction is presented in Fig. 2.

Fig. 2
figure 2

Interaction between Gender x Hostile Sexism (HS). Criterion variable was purchase intent measured in Businesswomen (Bw) ad condition in South Africa

Full size image

Next, in the first step, the two predictors (HS and BS) were regressed on the Hw – Bw difference in purchase intent between the two experimental conditions. It was revealed that BS was a significant predictor of this difference. Specifically, the difference between the βs for relations between benevolent sexism and general purchase intent was significant (.20 vs. –.03, respectively), indicating that, in line with Hypothesis 2b, the higher BS, the stronger the purchase intent for the Hw over the Bw advertisement strategy (see Fig. 3). No other interactions reached significance (see Table 2).

Fig. 3
figure 3

Interaction between Gender x Hostile Sexism (HS). Criterion variable was the Hw – Bw difference between purchase intent in Housewife (Hw) and in Businesswoman (Bw) ad conditions in South Africa. Scores above zero (where Hw = Bw) indicate that purchase intent is stronger in the Hw (vs. Bw) advertisement condition

Full size image

Hypothesis 3: Comparisons of Countries

In order to test Hypothesis 3, regression coefficients (β) returned by the moderated regression analyses for different countries were compared. The comparison procedure was based on Clogg, Petkova, and Haritou’s (1995) recommendations and algorithms published by Faul et al. (2007). As before, the mixed design required calculating the difference scores between the two advertisement types to test the three-way interaction predicted by Hypothesis 3. G*Power 3 software was used to perform the calculations (e.g., Faul et al. 2009). Comparisons were conducted in pairs: (a) UK versus SA, (b) SA versus PL, (c) UK versus PL. Each comparison was performed in two steps (post hoc type of analysis): (a) obtaining details from regression analyses on differences between two βs, sample sizes, and errors (SDs) and (b) calculation of a statistical parameter δ to be contrasted with a standard critical value. The δ values that were higher than the critical value indicated that the two βs were statistically different. Additionally, Cumming’s (2009) analysis of contrasting bootstrapped confidence intervals of unstandardized regression coefficients (B) was also performed. Because it confirmed the findings obtained in the G*Power analysis, only the latter analysis is reported here.

Table 3 reports values for purchase intent as a criterion variable in each of the two advertisement type conditions. Because previous research revealed that British respondents exhibited a significantly lower level of sexism than participants in Poland or South Africa (Zawisza et al. 2015), Hypothesis 3 anticipated that sexism (especially the more overt type, HS) would predict purchase intent only in the latter two countries (Glick and Fiske 2001; Glick et al. 2000; Sibley et al. 2007). The findings partially support this prediction. Hostile sexism was generally less predictive of Hw advertisement effectiveness than benevolent sexism. Higher levels of benevolence to women went in hand with higher purchase intent for the Hw advertisement type in the less egalitarian countries (PL and SA), and it did not predict responses to the Bw advertisement type in any country. Hostile sexism on the other hand predicted greater purchase intent for the Hw advertisement strategy in the UK (i.e., the higher HS the higher the advertisement effectiveness) and in SA for the Bw advertisement strategy (i.e., the higher HS the lower purchase intent for the Bw advertisement).

Table 3 Comparisons of β coefficients returned by regression analyses performed on purchase intent as the criterion within the advertising type across the three countries
Full size table

Discussion

The key findings of our investigation are that, in line with Hypothesis 1, paternalistic, traditional-female, gender portrayals in advertising (i.e., housewife) are more effective than nontraditional ones (i.e., businesswoman) and this holds across three countries that differ in national levels of sexism. Moreover, in support of Hypotheses 2 and 3, viewers’ benevolent sexism positively predicted purchase intention for the paternalistic, traditional Housewife advertisement strategy and their hostile sexism negatively predicted their purchase intention for the envied, nontraditional Businesswoman advertisement strategy in relatively gender-conservative South Africa. However, surprisingly, in Poland viewers’ benevolent sexism positively predicted purchase intentions for both advertisement types, with the exception of highly hostile women, whereas in the United Kingdom, viewers’ hostile sexism positively predicted purchase intention for the Housewife advertisement.

These results support the generalizability of the SCM across the three countries. That is, the paternalistic housewife advertisement strategy triggered higher purchase intent than the envied businesswoman advertisement strategy in countries as diverse as PL, SA and the UK. This is in keeping with previous literature that reported the primacy of such paternalistic (or communal) over envious (or agentic) gender role portrayals in advertising. Yet it goes beyond Swiss (Infanger et al. 2012) or British (Zawisza and Cinnirella 2010) samples, male portrayals (Zawisza et al. 2016) or brand perception (Aaker et al. 2012; Kervyn et al. 2012) by focusing on two less egalitarian countries, female portrayals, and advertising context. It also confirms the universal nature of the dimensions of social perception (Abele and Bruckmuller 2011; Cuddy et al. 2008; Cuddy et al. 2009) and their cross-cultural influence by evidencing the warmth-over-competence primacy across three countries. Moreover, it goes beyond the previous research by showing that this applies directly to purchase intent in response to the advertisements and not only to the perception of the advertisement characters (Infanger et al. 2012) and ad and brand attitudes (Infanger and Sczesny 2015) researched previously.

Our findings also point to the importance of considering the type of sexism, advertising strategy, and country in determining the predictive value of sexism in explaining the effectiveness of gendered advertising that utilizes female portrayals. For example, in the UK, it was viewers’ hostile sexism that positively predicted higher effectiveness of the traditional female advertisement strategy. Viewers’ benevolent sexism was not predictive of the effectiveness of any of the advertisements. In PL, on the other hand, effectiveness of the same traditional housewife advertisement type was positively predicted by hostile sexism in women only and by benevolent sexism irrespective of participants’ gender. The effectiveness of the envied nontraditional businesswoman advertisement in this country was predicted positively by benevolent sexism. In SA the picture was somewhat different again. Here benevolent sexism positively predicted the effectiveness of the paternalistic housewife advertisement too (as in PL) and the preference of this advertisement over the envied businesswoman one. Moreover, the higher the hostility to women, the lower the effectiveness of the envied businesswoman advertisement in the SA, especially in women. Thus the pattern predicted by Hypothesis 2a and Hypothesis 2b held only in the SA, which is also in line with Hypothesis 3, which predicted such a pattern would hold especially in less gender-egalitarian countries.

The findings from SA are in keeping with theory and research suggesting that the two forms of sexism function together to maintain the status quo—rewarding desirable behaviors with benevolence and punishing undesirable behaviors with hostility (Becker 2010; Sibley and Wilson 2004). This is manifest in benevolently sexist responses being predictive of higher effectiveness of the traditional paternalistic advertising strategies and hostility being predictive of lower purchase intent for the nontraditional envied advertising strategy. Becker’s (2010) participants were a sample of German women from the general public, and Sibley and Wilson’s (2004) were male students from New Zealand. Thus, the comparability between these and our findings is limited. Of importance for theory development, Zawisza et al. (2015) suggest that there may be culture-specific ways in which the two sexist ideologies maintain the status quo. For example, in Poland, hostile sexism may be associated with feminism, which in turn has historically been frowned upon due to the links with forced emancipation under communism. This indeed fits the pattern obtained here: Women’s hostility to women manifests itself in higher support (i.e., purchase intent) for advertisements portraying traditional women whereas benevolence is directed to both. Thus, although benevolence does not differentiate perceptions of the two advertisement strategies, hostility does but in a different way than proposed by past research and theory (Becker 2010; Sibley and Wilson 2004).

However, in the United Kingdom, hostility seems to have had similar function as in Poland (e.g., it manifested itself in greater support for the traditional paternalistic advertisement strategy). The reasons for this may be different than is the case in PL. For example, some argue that the UK is experiencing the return to sexism after a period of high egalitarianism (Braun and Scott 2009; Crompton et al. 2005; Walter 2010). Because egalitarian norms that prohibit overt sexism are strong, hostile sexism in this country may manifest itself more subtly (e.g., in greater support/social reward for the traditional female gender roles and not in overt punishment through lack of acceptance for the nontraditional roles).

Although further research is needed to shed light onto the exact mechanisms underlying these responses, our study is unique in highlighting the theoretical possibility that there may be multiple, country-specific routes to maintaining the gender status quo. It also sheds some light on inconsistent past findings concerning the predictive role of gender attitudes in determining advertising effectiveness (Zawisza and Lobban 2015). These reactions are dependent on the type of sexism (hostile vs. benevolent), advertisement type (traditional/paternalistic vs. nontraditional/envied), country, and, to an extent, on the respondents’ gender.

Our direct comparison between countries showed that hostile sexism, compared to benevolent sexism, was generally less predictive of the effectiveness of the paternalistic traditional advertisement strategy and more predictive of the effectiveness of the nontraditional envied advertisement strategy. This pattern corroborates the general theoretical notion proposed by AST that the two ideologies condition responses consistent with the status quo (Becker 2010; Sibley and Wilson 2004). The findings pertaining to benevolent sexism supported Hypothesis 3: Benevolent sexism was more predictive in the two less egalitarian countries than in the UK but only (and as expected) with regard to the traditional paternalistic advertising strategy. However, findings regarding hostile sexism supported Hypothesis 3 only partially: It was more predictive of the effectiveness of the nontraditional envious advertisement strategy in the least gender-egalitarian country (SA) than in the egalitarian UK and moderate PL. Surprisingly, however, hostile sexism was also most predictive of the effectiveness of the traditional paternalistic advertisement strategy in the egalitarian UK. As we discussed, the last finding may reflect country-specific ways in which (hostile) sexism operates to maintain the status quo.

Our findings also show that women generally reported higher purchase intent than men (i.e., especially in PL and in the UK), which could be attributed to women’s (vs. men’s) easier identification with the female characters in the advertisements (Dimofte et al. 2015). Hostilely sexist women in particular responded differently in PL and SA. They preferred the traditional paternalistic advertisement strategy more than men did in PL, but in SA, they preferred the nontraditional envied advertisement strategy less than men did. It may be that women in these relatively gender non-egalitarian countries, by virtue of their lower status, are particularly sensitive to the normative function(s) of hostile sexism because they are the ones who have the most to lose if they transgress these social norms (Glick and Fiske 2001; Glick et al. 2000; Sibley et al. 2007). The fact that both advertisement strategies were received better in PL and in SA than in the UK may be explained in terms of higher expectations for the quality of advertising in a market as saturated and established as the British one.

Limitations and Directions for Future Research

Our study focuses on attitudes to women and advertisements that portray female gender roles. This work goes beyond previous findings regarding male gender roles (Zawisza et al. 2016). A similar investigation into the effectiveness of other paternalistic and envious portrayals (e.g., the poor and the rich, the elderly and the young, or brands originating from different countries; Glick et al. 2002) would be beneficial to fully examine the generalisability of the SCM across different stereotypes in advertising context. The inclusion of more gender non-egalitarian and moderate countries in a study of this kind would help to determine further the level of gender egalitarianism required to affect the purchase intent for the products advertised. We also note that future studies should control for mode of participation (groups vs. individual) in an attempt to better control potential influence of group presence on social desirability levels. Whereas this was not controlled here, participation mode was random in all three countries. Those participating in groups operated on individual work stations, thus minimizing interactions and potential for any group influences.

Other sample-specific limitations include participants’ age and ethnicity. Although the distributions of age and ethnicity in our samples were too limited for meaningful comparisons, some studies have shown that sexism increases with age (Gaunt 2012; Glick et al. 2002a) and others have shown no relation between these two variables (Mikołajczak and Pietrzak 2014). The fact that our advertisement characters were White may have affected the responses of our Participants of Color (Perkins et al. 2000; Whittler 1991). Whereas race was not a focus here, our additional analysis controlling for participants’ ethnicity returned findings similar to those reported here. Still, future research is needed in order to establish clear links between gender representations in advertising, participants’ age, their ethnicity, and advertising effectiveness.

Moreover, our investigation focused on one product that was unisex and low-involving. It is possible that the traditional paternalistic housewife advertising strategy simply works for such household-related products or that the tested product (i.e., orange juice) was more popular in one of our three countries. However, similar paternalistic (male) characters were reported to boost the effectiveness of advertisements for various products in American samples. These included washing-up liquid (Debevec and Iyer 1986), a cup of coffee, and a work-related personal computer (Garst and Bodenhausen 1997). Further studies could fruitfully test other products and control for their popularity across cultures. It is feasible, for example, that the pattern of findings would be different for high-involving product (e.g., cars or durables). Indeed, Zawisza and Pittard (Zawisza and Pittard 2015; Zawisza 2016) report that a paternalistic (male) advertisement strategy was less effective than the envious one for unisex but high-involving products such as smartphones. This is due to the proposed greater relevance of the competence (vs. warmth) dimension to high-involving products. The authors suggest an extension of the SCM model for use in advertising contexts by incorporating this relevance principle. Further research could focus on testing the generalizability of such a “relevance-SCM” model across cultures.

Practice Implications

In practical terms our findings suggests that the use of the traditional paternalistic housewife portrayal is globally more effective than the nontraditional envious businesswoman one—at least for low-involving, unisex products such as orange juice. However, before the advertisers and marketers decide to abandon the use of nontraditional female gender portrayals, they should consider the relevance of the warmth concept to their product and target group (Zawisza and Pittard 2015). As the recent example of the backlash directed at the “Beach body ready’” campaign illustrates (https://www.youtube.com/watch?v=g5p1TETIuIo), consumers’ changing values and attitudes to gender roles need be monitored.

Our findings also suggest monitoring which attitudes would be most informative in predicting the effectiveness of specific advertising strategies in specific countries and among women and men. Marketers in South Africa are advised to monitor benevolent sexism when considering traditional paternalistic female portrayals in their advertisements, but if their target audience is women, then hostile sexism should be monitored. British marketers should measure hostile sexism in preparation for a traditional paternalistic female advertisement strategy. Polish marketers preparing such a campaign should measure both types of sexism, especially if their target audience includes men. However, if they are interested in launching a nontraditional envious businesswoman advertising strategy, pre-tests involving benevolent sexism would be more informative. All advertising and marketing practitioners should, however, also consider broader socio-economic consequences of reinforcing traditional gender stereotypes because these are known to have a number of negative effects on, especially female, audiences (Davies et al. 2005) as well as on purchase intent (Lee et al. 2011. In fact, new legislation regarding gendered content in advertising is being currently considered by Advertising Standards Authority in the UK (ASA Report 2017).

Conclusions

Our findings have two main theoretical and related practical implications. They provide further evidence for cross-national generalizability of the SCM (Fiske et al. 2002), supporting the universality of the two dimensions of social perception (Fiske et al. 2007) and, uniquely, their cross-cultural applicability to global gender advertising. They also point out that advertisers need to take a more nuanced approach to the task of predicting the effectiveness of gendered advertisements utilizing female gender roles. Given the continuously changing expressions of sexism, its increasing social rejection, and negative social effects, measuring such psychographics is crucial for advertising effectives in global markets.