Using the Inequitable Gender Norms Scale and Associated HIV Risk Behaviors among Men at High Risk for HIV in Ghana and Tanzania

Abstract

Context:

Gender norms influence unintended pregnancy, maternal health, HIV/AIDS infection, and act as barriers to reproductive health services. The Gender Equitable Men (GEM) scale has been used widely in programs and research in African settings, but it has yet to be statistically validated.

Method:

We examined the internal and external validity of the Inequitable Gender Norms (IGN) subscale of the GEM scale in Tanzania and Ghana using a two-step, mixed-method process. Confirmatory factor analysis tested the internal validity of the subscale and regression tests identified associations between the IGN scale and several HIV risk-related variables.

Results:

The IGN scale was shown to be a useful measure of gender norms in both countries. Excluding two questions that measured attitudes toward homosexuality, the scale met the hypothesized single factor structure. Furthermore, the IGN scores were significantly associated with several HIV risk variables in both samples.

Conclusions:

The IGN scale is a robust measure of gender norms in these African countries. However, the role of attitudes toward homosexuality as a contributor to gender norms measurement needs further exploration. Our analyses provide a basis for using the IGN scale to provide a contextualized understanding of men’s perceptions of gender norms and to evaluate programs focused more equitable gender norms. We are aware of only one other measure of gender norms that has been statistically validated in the African Context.

Keywords

Africa gender culture health sexualities gender equality confirmatory factor analysis

Introduction

The World Health Organization (WHO) stressed the importance of gender in the HIV pandemic over a decade ago (Campbell 2004), and greater attention is being given to the importance of gender in HIV research (Akeroyd 2004; Campbell 1995, 2004; WHO 1995; Parker 2001; Varga 2003; Susser and Stein 2004). Likewise, US policy such as the Global Health Initiative’s focus on women and girls (US Government 2010) and Millennium Development Goal Number 3, the promotion of gender equality and empowerment of women (US Department of State 2005, United Nations 2011), are fueling an increasing interest in gender as a social construct.

Gender norms are widely believed to play a key role in HIV epidemiology in Africa. Polygamous relationships in Africa have been attributed to a norm of “free sexuality,” marriage and kinship structures, the high value placed on fertility (Caldwell, Caldwell, and Quiggin 1989), multiple sexual partners, and numerous children to ensure continuation of their lineage (MacDonald 1996; Caldwell, Caldwell, and Quiggin 1989; Meyer-Weitz et al. 1998). In many African traditional cultures, social norms often condone extramarital relations for men while, at the same time, discourage women from questioning such behavior (Meyer-Weitz et al. 1998; van de Wijgert et al. 1999). Men are often perceived within these cultures as having uncontrollable and insatiable sex drives, and are, therefore, expected to engage in multiple sexual partnerships, regardless of marital status (Campbell 2004; Varga 2003; Meyer-Weitz et al. 1998; McGrath et al. 1992). Not surprisingly in these same settings, male celibacy is often socially viewed as dangerous and perceived to lead men into homosexual activities or risky behavior of a nonsexual nature (Campbell 2004).

Additional studies in Africa indicate that power imbalances between men and women in Africa are responsible for much of the observed risk behavior among heterosexual men and subsequent spread of HIV (MacDonald 1996; Jewkes, Levin, and Penn-Kekana 2003; Dunkle et al. 2004; Martinez-Donate et al. 2004; Zierler and Krieger 1997; Maman et al. 2002; van der Staten et al. 1998). Men often are the recognized decision makers in sexual relationships, and decisions about safe-sex practices, in general, have been shown to be influenced by gender power imbalances (Varga 2003; Susser and Stein 2004). Other studies have found that gender norms influence unintended pregnancy, maternal health, HIV/AIDS infection, and are barriers to the utilization of reproductive health services (United States Agency for International Development [USAID] 2011a).

In response to this growing body of evidence, gender-based interventions have been implemented. For instance, a program called Stepping Stones, implemented in South Africa, aimed to improve sexual health through building stronger, more gender equitable relationships among couples. This intervention resulted in reduced levels of herpes simplex virus-2 incidence, reduced rates of intimate partner violence, and reduced rates of transactional sex and binge drinking within the program population (Jewkes et al. 2008).

Measuring Gender Norms

The growing knowledge of the association between gender norms and health behavior has fueled the development of a variety of questions and scales to measure the “gender norm” construct. In an effort to synchronize the various surveys, C-Change (2011) compiled a compendium of gender-related measurement instruments to facilitate dissemination and access to measures that have been previously used. The most frequently administered, modified, and cited scale is the Gender Equitable Men (GEM) scale, which is comprised of two subscales (Pulerwitz and Barker 2008). The GEM has been used to assess relationships between perceived gender norms and outcomes in projects including contraceptive uptake (Shattuck et al. 2011), HIV prevention (Exner et al. 2003), and gender-based violence (Pulerwitz et al. 2010).

Barker et al. (2011) used the GEM scale to assess gender equity in representative samples from six different countries: Brazil, Chile, Croatia, India, Mexico, and Rwanda. They found that men with more gender equitable attitudes were better educated, had more satisfying sexual relationships with their wives, were less likely to regularly abuse alcohol, and were less likely to have ever paid for sex.

GEM Scale Content and Structure

The GEM scale was developed as part of a project in Brazil designed to promote gender-equitable norms and behaviors among young men in order to prevent risky behaviors and thus curtail the spread of HIV/AIDS. The scale was based on a review of the literature on gender norms and formative research on the construction of gender norms in Brazil. Pulerwitz and Barker (2008) identified “social expectations for appropriate behavior of men” as the means by which to define gender norms. The GEM scale looks to measure men’s attitudes toward gender norms that directly influence health behaviors such as sexual and reproductive health and intimate partner violence. Based on their research, Pulerwitz and Barker conclude that a “gender-equitable” man was defined as one who:

Seeks relationships with women based on equality, respect, and intimacy rather than sexual conquest;

Seeks to be involved in household chores and child care, meaning that he supports taking both financial and caregiving responsibility for his children and household;

Assumes some responsibility for sexually transmitted infection prevention and reproductive health in his relationships;

Is opposed to violence against women under all circumstances, even those that are commonly used to justify violence (e.g., sexual infidelity); and

Is opposed to homophobia and violence against homosexuals.

The GEM scale contains twenty-four items spread across two factors: equitable gender norms (EGN, seven items) and inequitable gender norms (IGN, seventeen items). The overall combined reliability (internal consistency) of both scales was found estimated at α = .81 (EGN α = .77, IGN α = .85; Pulerwitz and Barker 2008).

One or both subscales of the GEM have been used in a variety of cultural settings, including Malawi (Shattuck et al. 2011), Tanzania, India, Brazil, China, Ethiopia, Kenya, and Uganda (C-Change 2011). Despite its repeated use as a predictor, the single latent construct of IGN has not been rigorously evaluated using hypothesis based confirmatory tests in an African setting. Designing and validating measurement tools is critical to assessing gender norms and gender-oriented interventions in a valid and reliable manner. The main purpose of this article is to validate a seventeen-item gender norm subscale—the IGN scale (Pulerwitz and Barker 2008)—among men in two African countries, Tanzania and Ghana. Using large samples, we tested the internal integrity of the scales and identified statistical associations between the IGN scores and HIV risk behaviors.

Method

The study was reviewed and approved in the United States by the Protection for Human Subjects Committee at Family Health International 360 and in Ghana by the Ghana Health Service Ethical Review Committee. In Tanzania, institutional review boards from three separate organizations oversaw the review and approval of the study—National Institute for Medical Research, Muhimbili University of Health and Allied Sciences, and The Ministry of Health and Social Welfare. Written informed consent was obtained from survey respondents, and each respondent received the local equivalent of US$5 for participating in the study in both countries.

The data we present in this article were collected among adult men in the cities of Tema, Ghana (June–November 2008) and Mbeya, Tanzania (September 2008–February 2009). We chose these sites due to their relatively high HIV prevalence. As of the most recent BSS (2003), Tema had one of the highest HIV prevalence rates in the country (6.4 percent), double that of the national rate (3.1 percent; Ghana AIDS Commission 2004) and Mbeya the highest HIV prevalence in Tanzania (13.5 percent) almost double the national average of 7 percent (Tanzania Commission for AIDS NBoSaOM 2005).

Local field staff administered a survey containing the IGN scale and questions about sexual behavior to 800 men in Tema and 807 men in Mbeya. Eligibility criteria for respondents included being eighteen to forty-nine years of age and reporting at least three different sex partners in the previous three months. The survey instrument and recruitment procedures were informed by participant observation and focus groups among the study populations. The survey was pretested among ten men from the study population at each site and revised before full implementation.

Time-space sampling was used to enroll a probability sample of men in both sites, as it is particularly useful for reaching populations that congregate in known and geographically bounded venues (MacKellar et al., 1996; Magnani et al. 2005; Muhib et al. 2001). Time-space sampling is composed of a three-step procedure in which mapped venues are the primary sampling units. A sample of venues is randomly selected from the universe of known venues in a study area. A specific time period associated with each venue is then randomly selected and participating venues are visited during the specified time period. For our study, venues within high risk transmission areas (HTA)—places where men at high risk for HIV congregate—were identified, listed, and confirmed in both cities through a combination of existing research data, local community outreach activities, observational research, and focus groups.

Within each HTA, we generated an exhaustive list of all venues. Similarly, a comprehensive list of all potential sampling dates was developed. From within the list of possible dates and venues, random selection of both was accomplished using a SAS program with uniform random number generators. A final set of eighteen venues in Tema and thirteen venues in Mbeya were selected for sampling.

Local field staff recruited men within the venues using systematic sampling methods (e.g., every nth man). Response rates were approximately 90 percent in both sites. Men who agreed to participate were then administered informed consent and interviewed in or near the venue.

In our study, the IGN scale was embedded within a larger behavioral survey. Through pretesting and staff examination at both sites, it was determined that the two questions on norms and perceptions of homosexuality could constitute a unique construct in these cultural settings. Therefore, those items were removed from this administration of the IGN, reducing it to fifteen items. Response options for the IGN were consistent with those suggested by the scale authors: disagree, agree, and strongly agree (Pulerwitz and Barker 2008).

Data Analysis

We hypothesized that the IGN scale is comprised of a single latent factor. To test this hypothesis in the two African contexts, we implemented a multistep analysis. First, confirmatory factor analysis (CFA) was used in each sample independently to determine whether the IGN scale measures perceived gender norms consistently with the single factor structure previously proposed by Pulerwitz and Barker (2008). Next, associations between the IGN scale and several HIV risk-related variables were tested, as suggested by Pulerwitz and Barker. We described the mean differences for the aggregate scale and scores for each item in the IGN across samples and tested differences (t-tests) across samples. We then tested the statistical validity of a single factor model for the IGN scale, as suggested by Pulerwitz and Barker, and used CFA methodology, as described by Thompson and Daniel (1996). We computed the CFA using MPlus (Muthén & Muthén, Los Angeles, California). The ratio of parameter estimates to corresponding standard errors was examined to determine which items loaded significantly onto this single factor. Initial analyses assumed independence of error terms (base model). Subsequently, modification indices were examined and those items with modification indexes greater than 4.0 were allowed to freely associate with one another (correlated model). The fit of the final model was tested and reported for both samples.

Several fit statistics were used to describe the results. The χ² statistic is the first measure of how well a model fits the data. However, it is greatly influenced by sample size and thus is rarely found to be nonsignificant in samples sufficiently large to legitimately perform CFA (Thompson and Daniel 1996). Therefore, an appropriate model fit was assessed using the Root Mean Square Error of Approximation (RMSEA), the Comparative Fit Index (CFI), and the Tucker–Lewis index (TLI), also known as the Nonnormed Fit Index (NNFI; Bentler and Bonett 1980). In summary, we assessed the fit of the model using the following criteria: χ² within three times the degrees of freedom (Segars and Grover 1993), RMSEA score below .06, CFI score above .95, and TLI or NNFI score above .95 (Hu and Bentler 1999).

Finally, associations between IGN scores and several associated variables were tested using multiple regression tests. Pulerwitz and Barker (2008) conducted similar tests of the original scale with their Brazilian sample. Specifically, they focused on three variables: education level, history of intimate partner violence, and condom use. Pulerwitz and Barker also elected to create tertiles with the participants’ mean scores using low, medium, and high support for equitable norms. We also tested education level and condom use, among others, but chose to use the factor scores from the latent factor IGN to test the relationship with the dependent variables.

The use of mean scores assumes that each item is weighted equally in the construct, while factor scores account for the item weights in the construction of the latent IGN factor. We tested relationships between the IGN factor scores and the following demographic and HIV risk variables: education level, number of sexual partners, frequency participants paid for sex, and condom use using linear and logistic regression, respectively. To assess the external validity of the IGN, we tested the scale’s relationships with several sexual risk variables and asked participants about the sexual behaviors of their peers (condom use, multiple partnerships, and paid sex).

Results

Participant Demographics and IGN Scale Scores

Table 1 provides demographic data of study participants for each site. The overall distribution of IGN scores for each sample was within standard range for skew (2.0) and the Mbeya sample has a significantly higher mean score (higher inequitable beliefs) than the Tema sample (p < .01), as shown in Table 2.

Table 1.

Demographic Characteristics of Sample Populations.

		Tema, Ghana	Mbeya, Tanzania
Sample size		800	807
Age	M (SD)	28.32 (5.96)	29.54 (6.25)
Monthly salary (US$)^a	M (SD)	169.88 (266.04)	195.17 (142.92)
STD (last twelve months)	Yes	132 (16.52%)	167 (20.80%)
STD (last twelve months)	No	667 (83.48%)	636 (79.20%)
Marital status	Married	129 (16.13%)	194 (24.04%)
	Living with someone	55 (6.88%)	105 (13.01%)
	Widowed	—	24 (2.97%)
	Single	616 (77.00%)	484 (59.98%)
Number of wives	1	119 (92.97%)	170 (87.63%)
Number of wives	>1	9 (7.03%)	24 (12.37%)
Number of children	M (SD)	0.72 (1.04)	0.92 (1.39)
	0	455 (58.41%)	440 (54.66%)
	1	167 (21.44%)	168 (20.87%)
	2	103 (13.22%)	168 (20.87%)
	3 or more	54 (6.93%)	89 (11.06%)
Ethnicity	Akan	424 (53.00%)	Nyakyusa	249 (30.86%)
	Gaadangbe	112 (14.00%)	Wasafa	92 (11.40%)
	Ewe	111 (13.88%)	Wakinga	105 (13.01%)
	Moledagbani	96 (12.00%)	Wachaga	70 (8.67%)
			Wapare	21 (2.60%)
			Other	270 (33.46%)
Religion	Christianity	641 (80.13%)	560 (69.39%)
	Islam	107 (13.38%)	246 (30.48%)
	Traditionalist	5 (0.63%)	1 (0.12%)
	None	46 (5.75%)	—

Note: M = mean; SD = standard deviation; STD = sexually transmitted diseases.

^aUS$ rate for Ghanaian Cedi calculated for September 1, 2008. 1 GHC = US$0.849. Tanzanian Shilling calculated for February 1, 2009. 1 TZS = US$0.00076. The following converting website was used: http://www.oanda.com/currency/converter/.

Table 2.

Inequitable Gender Norms Aggregated Scale Scores and Item Scores.

Aggregate Scores	M*	SD	Median	Mode	Skew	Kurtosis
Tema	1.97	0.40	1.97	1.93	0.17	−0.43
Mbeya	2.21	0.47	2.67	3.00	−0.02	0.57

Note: M = mean; SD = standard deviation. Higher score reflects more inequitable gender norms.

*p < .01.

Men in Mbeya scored significantly higher on ten of the fifteen scale items (Table 3). Five of the items where they differed focused on violence toward women and other people. The other items where they differed focused on the acceptance of multiple partners and women’s sole responsibility for contraceptive use and sexual desires.

Table 3.

IGN Item Sample Means and Standard Deviations.

Items	Tema	Mbeya
1. It is the man who decides what type of sex to have	2.49 (0.78)	2.55 (0.71)
2. A woman’s most important role is to take care of her home and cook for her family	2.44 (0.84)	2.51 (0.72)
3. Men need sex more than women do**	2.17 (0.92)	2.33 (0.82)
4. You do not talk about sex, you just do it*	1.68 (0.88)	1.59 (0.80)
5. Women who carry condoms on them are “cheap.”	1.96 (0.96)	2.03 (0.85)
6. A man must have sex with other women, even if things with his wife are fine**	1.84 (0.94)	2.25 (0.78)
7. There are times when a woman deserves to be beaten**	1.66 (0.89)	1.97 (0.84)
8. It is a woman’s responsibility to avoid getting pregnant**	1.98 (0.94)	2.42 (0.77)
9. A man should have the final word about decisions in his home**	2.37 (0.87)	2.66 (0.64)
10. Men are always ready to have sex	2.22 (0.92)	2.28 (0.82)
11. A woman should tolerate violence in order to keep her family together**	1.99 (0.95)	2.41 (0.78)
12. If a woman cheats on a man, it is okay for him to hit her**	1.67 (0.90)	2.09 (0.83)
13. If someone insults me, I will defend my reputation with force if I have to**	1.69 (0.88)	2.30 (0.83)
14. I would be outraged if my wife asked me to use a condom	2.07 (0.96)	2.08 (0.83)
15. It is okay for a man to hit his wife if she will not have sex with him**	1.24 (0.62)	1.67 (0.82)

Note: IGN = inequitable gender norms. Higher score reflects more inequitable gender norms. Item mean scores (standard deviations). Response options: (1) do not agree, (2) partially agree, and (3) agree.

Significant difference between mean scores: *p < .05. **p < .01.

CFA

We used CFA to test whether the data supported the single latent factor of IGN in both data sets independently as suggested by Pulerwitz and Barker (2008). The sequence of model decisions and subsequent results for both populations mirrored one another as described above. Examination of fit indices and the χ²difference tests revealed that the correlated model fit the data significantly better than the base model in both populations (p < .01). Regression weights revealed that all items loaded significantly onto the single IGN factor in both samples. The correlated models met all four of our criteria for a good fit (Table 4). There was no statistical justification for testing a reduced model for the IGN factor. Reviewing these fit statistics, we can say that the fifteen items of the IGN scale meet the hypothesized one factor structure for our samples from Tema and Mbeya.

Table 4.

IGN Subscale CFA Fit Parameters.

Analysis	n	χ²	df	p	χ²Δ	RMSEA	RMSEA 90% CI	CFI	TLI
Ghana
Base model (fifteen items)	800	267.47	73	*	—	0.058	[0.050, 0.065]	0.87	0.89
Correlated model (fifteen items)	800	133.80	70	*	133.67*	0.034	[0.025, 0.042]	0.96	0.96
Tanzania
Base model (fifteen items)	807	790.46	57	*	—	0.126	[0.119, 0.134]	0.80	0.91
Correlated model (fifteen items)	807	123.06	43	*	667.40*	0.048	[0.038, 0.058]	0.98	0.99

Note: CFA = confirmatory factor analysis; CFI = Comparative Fit Index; IGN = inequitable gender norms; RMSEA = Root Mean Square Error of Approximation; TLI = Tucker–Lewis index.

χ²Δ is the chi-square difference statistic for comparing the current model to the previous model in the table.

*p < .01.

In social science, researchers focus on the reliability of a scale (Cronbach’s α) to determine its effectiveness. We calculated the reliability scores for each sample. The Cronbach’s α for Tema was .72 and Mbeya .87, which are both above the generally accepted cutoff of 0.70 (Cortina 1993).

IGN and Sexual Risk

We found several significant relationships between the IGN factor scores and the dependent variables we included in our analysis (Table 5).

Table 5.

Associated Relationship between IGN Factors Scores, Demographic and Sexual Risk Behavior Variables.

		Ghana				Tanzania
Independent Variable	Dependent Variable	β/Parameter Estimates	F-Score/χ²	R ²	CI	β/Parameter Estimates	F-Score/χ²	R ²	CI
IGN factor scores	Number of partners
	No. sex partners last twelve months	1.70**	16.12	0.02	[0.87, 2.54]	0.42**	6.81	0.01	[0.10, 0.74]
	No. current sex partners	0.51**	14.45	0.02	[0.25, 0.78]	0.35**	33.30	0.04	[0.23, 0.47]
	Paid sex
	Frequency pay for sex	0.55**	35.34	0.21	[0.37, 0.73]	0.60**	67.42	0.078	[0.46, 0.74]
	Whether paid for sex (last twelve months, yes/no)^a	−0.45**	0.64		[0.53, 0.75]	−0.45**	32.04		[0.55, 0.75]
	Condom used during paid sex (last twelve months, Yes/No)^a	0.39*	1.46		[1.06, 2.01]	0.37**	7.0		[1.10, 1.91]
	Last paid sex, condom used? (yes/no)^a	0.28 (ns)	3.19		[0.97, 1.78]	0.46**	10.33		[1.20, 2.09]
	Condom use
	Frequency of condom use during vaginal sex	−0.19**	22.64	0.28	[−0.27, −0.11]	−0.20**	62.17	0.07	[−0.25, −0.15]
	STD in the last twelve months (yes/no)	−0.28*	0.76		[0.61, 0.94]	−0.38**	0.68		[0.57, 0.82]
	Condom use with last three partners (range 0–3)	−0.23	26.33	0.03	[−0.31, −0.14]	−0.23	43.43	0.05	[−0.29, −0.16]
	Peer norms
	How many friends use condoms with new sex partner?	−0.08**	12.82	0.02	[−0.12, −0.03]	0.02 (ns)	1.69	0.00	[−0.01, 0.05]
	How many friends had >2 sex partners	0.01 (ns)	0.53	0.00	[−0.02, 0.05]	−0.07**	14.27	0.02	[−0.10, −0.03]
	How many friends paid for sex last year?	0.10**	19.93	0.03	[0.06, 0.15]	−0.04*	6.90	0.01	[−0.07, −0.01]
	Education
	Level of school completed	−0.36**	73.03	0.08	[−0.44, −0.27]	−0.39**	72.91	0.08	[−0.48, −0.30]

Note: CI = confidence interval; IGN = inequitable gender norms; STD = sexually transmitted diseases.

^aRelationship tested using logistic regression. Other relationships tested using linear regression. ns = not significant.

Peer norm responses were valued: (1) none of them, (2) some of them, and (3) all of them. IGN factor was scored (1) do not agree, (2) partially agree, and (3) agree.

*p < .05. **p < .01.

Men with higher IGN (or more IGN) had lower levels of education in both Tema and Mbeya. In both samples, the IGN was associated with greater numbers of sex partners over the last twelve months (p < .01) and more concurrent partners (p < .01). Also within both samples, there was a positive linear relationship between the IGN and frequency of paid sex (p < .01), while an inverse relationship existed between the IGN and whether a participant paid for sex in the last twelve months (p < .01). Men in both samples with higher IGN scores were more likely to use condoms during paid sex in the last twelve months (p < .05) but less likely to use condoms during any vaginal sex (p < .01) and had lower rates of sexually transmitted diseases in the last twelve months (Tema = p < .05, Mbeya = p < .01).

Three variables had significant relationships with the IGN scores in one country and not the other. In Mbeya, higher scores on the IGN were associated with higher condom use during last paid sex (p < .01), while they were negatively associated with the number of friends with more than two sexual partners. In Tema, a significant negative relationship was found between the IGN and the number of friends who use condoms with new partners (p < .01). Participants also reported condom use during the last sexual encounter with their last three partners (range 0–3). The IGN factor score was found to have a negative association with this event-level condom use (p < .01) in both data sets.

Discussion

Based on our findings, the IGN scale is now a validated measure of gender norms in the two African contexts where we administered the study. The scale met the hypothesized single factor structure and was significantly associated with several HIV risk variables. These findings are important for several reasons.

First, our analysis has provided much needed statistical validation of a commonly used subscale. CFA provides several fit statistics that can be compared and used as a gauge for the structural integrity of the scale. The fit indexes of both data sets met the criteria determined at the outset of this article and the α for both samples was above the lower cutoff of .70 (Cortina 1993). Many researchers focus on the reliability of a scale (Cronbach’s α) to determine its effectiveness. This is not the purpose of the α score. Routinely, emphasis is placed on α despite its susceptibility to item number and number of response options (Cortina 1993; Crocker and Algina 1986; Jenkins and Taber 1977). The methods implemented in this study are more rigorous and reflect hypothesis testing. Our analyses provide a basis for using the IGN scale as an independent variable. Other measures of gender norms are available in the literature but few have much foundation in statistical validation.

As previously noted, we reduced the IGN based on pretesting and the recommendation of local researchers who reviewed the survey. Previously unpublished analyses (CFA) of two large data sets from Uganda and Kenya found that the model fit for the IGN improved to acceptable levels only after the removal of the two items referring to homosexuality in both data sets (Shattuck 2010). The initial data collection on the GEM scale (including the IGN) was done in Brazil, a country culturally different to those of East Africa. Our findings suggest that questions that include references to homosexuality should not be included within the same latent construct for these populations.

Previous gender scale development efforts by Luyt (2005) identified three latent factors among forty items from the earlier version of the Male Role Norms Inventory in South Africa (Levant et al. 1992): toughness, control, and sexuality. The eight items found to comprise the sexuality dimension used either the word “gay” or “faggot” and described implied homosexual acts (men sleeping in the same bed and kissing in public) and impotence (“cannot get an erection”). Luyt’s work supports our understanding that homosexuality is an important component of masculinity, but men’s perception of homosexuality is differentiated from other gender normative attitudes.

Quick examination of the legal and social context of homosexuality in Africa provides examples of the ongoing challenges faced by homosexual men and women on the continent and has merited its own attention in academic circles (Baldauf 2010; Day 2011). Some researchers assert that homosexuality is not part of traditional societies in Africa (Hrdy 1987), while others provide historical evidence that homosexuality has always existed in African communities, and that much of the current stigma associated with homosexuality is derived from the colonial redefinition of homosexuality as taboo and immoral (Murray and Roscoe 1998). Guadio (1998) suggests that we need to examine how Western constructions of sexuality and homophobia stigmatize homosexuals in African countries and result in oppression. Our conclusions about the measurement of gender norms and how homosexuality relates to it may also reflect the imposition of Western constructions of sexuality. In order to better measure change in gender norms, additional research that includes item development or modification at the level of the target community and confirmatory testing is needed.

Our findings also inform HIV prevention efforts. We identified several sexual behavior associations with the IGN scale that should be noted by programmatic and research staff, particularly when working with similar populations of men. Higher IGN scores were associated with increases in partners and frequency of paid sex, as well as decreases in condom use during sex. Nyanzi, Byanzi-Wakholi, and Kalina (2009) suggested that prevention efforts should embrace the “social and historical roots of phenomena—masculinity and manhood.” We concur with this suggestion and add that this social context is within a period of rapid change in African nations as more women enter the job market and male roles and responsibilities are in flux.

Programs focused on changing behaviors associated with risky sexual practices need to account for the underlying gender norms that influence those behaviors. Challenging gender norms includes the encouragement of critical awareness among men and women of gender roles and norms, the promotion of women’s position in communities and homes, challenges to the distribution of resources and allocation of duties between men and women, and addressing the power relationships between women and others in the community (Rottach, Schuler, and Hardee 2009; Pulerwitz et al. 2006; Barker et al. 2011).

Our data are relevant to gender norm theory as well. They support previous studies that demonstrate a link between HIV risk and gender roles and norms (Jewkes et al. 2008; USAID 2011a). They also suggest that we may need to reexamine our conceptualization of masculinity, at least in certain African contexts. The data presented illustrate that higher IGN were associated with increased number of sexual partners and paid sex. At the same time, our data show that higher IGN were associated with increased condom use during paid sex but decreased condom use during vaginal sex in general. These results need to be contextualized within what we know about masculinity and sexual risk.

Previous research suggests that boys internalize messages about their appropriate roles, which often promote male dominance and sexual promiscuity (Rivers and Aggleton 1999; Campbell 1995; Wingood and DiClemente 2000). These roles further perpetuate men as more knowledgeable about matters of sex and reproductive health than women (Marsiglio 1988). Research by Nyanzi, Byanzi-Wakholi, and Kalina (2009) has documented that promiscuity supports historical masculine stereotypes of strength and admiration. Among our study populations this might translate into the sexual risk patterns we observed. The social and physical motivation to have multiple partners, combined with the increased awareness that most men have about the risks of HIV, may have the desired effect of leading some men to use condoms with women they perceive as risky (i.e., a sex worker), but not the additional behavior of 100 percent condom use. Men are not using condoms with women who they perceive as more intimate and trusted, which is a behavior similar to women at high risk, although contextualized by very different power dynamics (Murray et al. 2007; Ngugi et al. 2007; Silberschmidt and Rasch 2001).

This raises an interesting question. If we interpret these findings from an inverse perspective, does it mean that men with more EGN lack a sort of “self-protection” trait related to unprotected paid sex? And if so, how is this trait related to the larger construct of masculinity? These questions deserve further investigation.

As with any research, our findings come with limitations. This study analyzed data from men with multiple partners. Our findings therefore may not reflect the normative gender beliefs among men in the general population of either country. A population-based study would be needed to answer this question. Nonetheless, from an epidemiologic perspective, the subpopulation from which we sampled plays a significant role in driving the HIV epidemic within Africa (Morris 1993). Our data suggest that it is socially acceptable for men in this subpopulation, in both sites, to have multiple sex partners, engage in transactional sex, and have unprotected sex in nonpaid sexual encounters. From a public health perspective, gender norms play an important role in shaping HIV risk behaviors that facilitate progression of the epidemic.

Footnotes

Appendix A

Table A1.

Dependent Variable List.

Items	Response Options
Peer norms
Q116. How many of your male friends always use a condom with new sex partners: all of them, some of them, or none of them?	1. All of them 2. Some of them 3. None of them 4. Don’t know
Q117. How many of your male friends had more than two different sexual partners in past three months: all of them, some of them, or none of them?	1. All of them 2. Some of them 3. None of them 4. Don’t know
Q118. How many of your male friends paid for sex in the past year: all of them, some of them, or none of them?	1. All of them 2. Some of them 3. None of them 4. Don’t know
Number of partners
Q401. In total, with how many different women have you had sex in the last twelve months?	[____\|____\|____]
Q406. In total, how many women are you currently having sex with, including: wives, girlfriends, sweethearts, and regular sex workers?	[____\|____]
Paid sex
Q402. In the last twelve months, did you pay a woman in exchange for sex?	Yes No
Q405. Did you use a condom every time you paid a woman to have sex in the last twelve months?	Yes No
Q403. In an average month, how often do you pay a woman for sex?	[____\|____]
Q404. The last time you paid a woman in exchange for sex, was a condom used?	Yes No
Condom use
Q407. How often do you use a condom during vaginal sex with a woman: always, usually, sometimes, or never?	1. Always 2. Usually 3. Sometimes 4. Never
Education
Q108. What is the highest level of school you completed? 1. None 2. Primary 3. Middle school/junior secondary school 4. Senior secondary/technical/vocational training 5. Postsecondary/advanced teacher training/polytechnic 6. University

Appendix B

Table B1.

Mean Scores: Sexual Risk Variables.

		Tema	Mbeya
Number of partners
Number of different women had sex with in last twelve months: average (standard deviation [SD])**		9.61 (10.29)	6.78 (4.37)
Number of current sex partners: average (SD)**		3.76 (3.27)	4.29 (1.67)
Paid sex
Frequency pay for sex: average (SD)		1.37 (2.25)	1.21 (2.05)
Whether paid for sex (last twelve months)	No	431 (53.94%)	409 (50.74%)
Whether paid for sex (last twelve months)	Yes	368 (46.06%)	397 (49.26%)
Every time paid for sex a condom was used (last twelve months)**	No	57 (18.87%)	119 (37.90%)
	Yes	245 (81.13%)	195 (62.10%)
Last time paid for sex a condom used?	No	63 (17.26%)	80 (20.41%)
Last time paid for sex a condom used?	Yes	302 (82.74%)	312 (79.59%)
Condom use
Frequency of condom use during vaginal sex (any woman)**	Never	59 (7.38%)	21 (2.61%)
	Sometimes	335 (41.88%)	225 (27.95%)
	Usually	186 (23.25%)	422 (52.42%)
	Always	220 (27.50%)	137 (17.02%)
Condom use last three partners, last sexual encounter	None	122 (15.25%)	77 (9.55%)
	One	119 (14.88%)	148 (18.36%)
	Two	274 (34.25%)	311 (38.59%)
	Three	285 (35.63%)	270 (33.50%)
Education
Level of school completed**	None	10 (1.25%)	8 (0.99%)
	Primary	36 (4.51%)	206 (25.56%)
	Middle school	213 (26.66%)	311 (38.59%)
	Secondary school	356 (44.56%)	138 (17.12%)
	Postsecondary School	105 (13.14%)	49 (6.08%)
	University	79 (9.89%)	94 (11.66%)

*p < 0.01.

Authors’ Note

Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This publication was made possible by Grant Number 1R01HD052429-01A2 from The National Institute of Child Health and Human Development (NICHD).

References

Akeroyd

2004. “Coercion, Constraints, and ‘Cultural Entrapments:’ A Further Look at Gendered and Occupational Factors Pertinent to the Transmission of HIV in Africa.” In HIV and AIDS in Africa: Beyond Epidemiology, edited by Kalipeni

Craddock

Oppong

Ghosh

, 89–103. Oxford, UK: Blackwell.

Baldauf

Scott

. 2010. “In Africa, Homosexuality Emerging as Hot-Button Issue.” Christian Science Monitor. Accessed May 27, 2011. http://www.csmonitor.com/World/Africa/2010/1208/In-Africa-homosexuality-emerging-as-hot-button-issue.

Barker

Contreras

J. M.

Heilman

Singh

A. K.

Verma

R. K.

Nascimento

. 2011. Evolving Men: Initial Results from the International Men and Gender Equality Survey (IMAGES). Washington, DC: International Center for Research on Women (ICRW) and Rio de Janeiro: Instituto Promundo.

Bentler

P. M.

Bonett

D. G.

. 1980. “Significance Tests and Goodness of Fit in the Analysis of Covariance Structures.” Psychological Bulletin 88:588–606.

Caldwell

J. C.

Caldwell

Quiggin

. 1989. “The Social Context of AIDS in Sub-Saharan Africa.” Population Development Review 15:185–234.

Campbell

C. A.

1995. “Male Gender Roles and Sexuality: Implications for Women’s AIDS Risk and Prevention.” Social Science & Medicine 41:197–210.

Campbell

C. A.

2004. “Migrancy, Masculine Identities, and AIDS: The Psychosocial Context of HIV Transmission in the South African Gold Mines.” In HIV and AIDS in Africa: Beyond Epidemiology, edited by Kalipeni

Craddock

Oppong

Ghosh

, 145–54. Oxford, UK: Blackwell.

C-Change. 2011. “Compendium of Gender Scales.” Accessed May 27, 2011. http://www.c-changeprogram.org/content/gender-scales-compendium/index.html.

Cortina

1993. “What Is Coefficient Alpha? An Examination of Theory and Applications.” Journal of Applied Psychology 78:98–104.

10.

Crocker

Algina

. 1986. Introduction to Classical and Modem Test Theory. New York: Holt, Rinehart and Winston.

11.

Day

Elizabeth

. “Why Was I Born Gay in Africa?” The Guardian, March 27. Accessed May 27, 2011. http://www.guardian.co.uk/world/2011/mar/27/uganda-gay-lesbian-immigration-asylum.

12.

Dunkle

Jewkes

Brown

Gray

McIntryre

. 2004. “Gender-based Violence, Relationship Power, and Risk to HIV Infection in Women Attending Antenatal Clinics in South Africa.” Lancet 363:1415–21.

13.

Exner

T. M.

Dorkin

S. L.

Hoffman

Ehrhardt

A. A.

. 2003. “Beyond the Male Condom: The Evolution of Gender-Specific HIV Interventions for Women.” Annual Review of Sex Research 14:114–36.

14.

Ghana AIDS Commission. (2004). 2004 HIV Sentinel Survey Report. Accra, Ghana: Ghana AIDS Commission.

15.

Guadio

Rudolf P.

1998. “Male Lesbians and Other Queer Notions in Hausa.” In Boy-Wives and Female-Husbands: Studies in African Homosexualities, edited by Murray

Stephen O.

Roscoe

Will

, 115–28. New York: St. Martin’s Press.

16.

Hrdy

D. B.

1987. “Cultural Practices Contributing to the Transmission of Human Immunodeficiency Virus in Africa.” Reviews of Infectious Diseases 9:1109–19.

17.

Bentler

P. M.

. 1999. “Cutoff Criteria for Fit Indexes in Covariance Structure Analysis: Conventional Criteria versus New Alternatives.” Structural Equation Modeling 6:1–55.

18.

Jenkins

G. D.

Taber

T. D.

. 1977. “A Monte Carlo Study of the Factors Affecting Three Indices of Composite Scale Reliability.” Journal of Applied Psychology 62:392–98.

19.

Jewkes

Rachel

Nduna

Levin

Jama

Dunkle

Puren

Duvvury

. 2008. “Impact of Stepping Stones on Incidence of HIV and HSV-2 and Sexual Behaviour in Rural South Africa: Cluster Randomised Controlled Trial.” British Medicine Journal 337:a506.

20.

Jewkes

R. K.

Levin

J. B.

Penn-Kekana

L. A.

. 2003. “Gender Inequalities, Intimate Partner Violence and HIV Preventive Practices: Findings of a South African Cross-sectional Study.” Social Science & Medicine 56:125–34.

21.

Levant

R. F.

Hirsch

Celentano

Cozza

Hill

MacEachern

Marty

Shendeker

. 1992. “The Male Role: An Investigation of Contemporary Norms.” Journal of Mental Health Counseling 14:325–37.

22.

Luke

2005. “Confronting the ‘Sugar Daddy’ Stereotype: Age and Economic Asymmetries and Risky Sexual Behavior in Urban Kenya.” International Family Planning Perspectives 31:6–14.

23.

Luyt

2005. “The Male Attitude Norms Inventory-II: A Measure of Masculinity Ideology in South Africa.” Men and Masculinities 8:208–29.

24.

MacDonald

1996. “Notes on the Socio-Economic and Cultural Factors Influencing the Transmission of HIV in Botswana.” Social Science Medicine 42:1325–1333.

25.

MacKellar

Valleroy

Karon

Lemp

Janssen

. 1996. “The Young Men’s Survey: Methods for Estimating HIV Seroprevalence and Risk Factors among Young Men Who Have Sex with Men.” Public Health Reports 111:138–44.

26.

Magnani

Sabin

Saidel

Heckathorn

. 2005. “Review of Sampling Hard-to-Reach and Hidden Populations for HIV Surveillance.” Acquired Immune Deficiency Syndrome 19:S67–72.

27.

Maman

Mbwambo

J. K.

Hogan

N. M.

Kilonzo

G. P.

Campbell

J. C.

Weiss

Sweat

M. D.

. 2002. “HIV-Positive Women Report More Lifetime Partner Violence: Findings from a Voluntary Counseling and Testing Clinic in Dar es Salaam, Tanzania.” American Journal of Public Health 92:1331–7.

28.

Marsiglio

1988. “Adolescent Male Sexuality and Heterosexual Masculinity: A Conceptual Model and Review.” Journal of Adolescent Research 3:285–303.

29.

Martinez-Donate

A. P.

Hovell

M. F.

Blumberg

E. J.

Zellner

J. A.

Sipan

C. L.

Shillington

A. M.

Carrizosa

. 2004. “Gender Differences in Condom-Related Behaviors and Attitudes among Mexican Adolescents Living on the U.S.-Mexico Border.” AIDS Education and Prevention 16:172–86.

30.

McGrath

Schumann

Pearson-Marks

Rwabukwali

Charles B.

Mukasa

Rebecca

Namande

Barbara

Nakayiwa

Sylvia

Nakyobe

Lucy

. 1992. “Cultural Determinants of Sexual Risk Behavior for AIDS among Baganda Women.” Medical Anthropology Quarterly 6:153–61.

31.

Meyer-Weitz

Reddy

Weijts

van den Borne

. 1998. “The Socio-Cultural Contexts of Sexually Transmitted Diseases in South Africa: Implications for Health Education Programmes.” AIDS Care 10:S39–55.

32.

Morris

1993. “Determining Male Fertility through Surveys: Young Adult Reproductive Health Surveys in Latin America.” Paper presented at the International Population Conference, Montreal, Canada, August 24–September 1.

33.

Muhib

Lin

Stueve

Miller

R. L.

Ford

W. L.

Johnson

W. D.

Smith

P. J.

and Community Intervention Trial for Youth Study Team. 2001. “A Venue-Based Method for Sampling Hard-to-Reach Populations.” Public Health Reports 116:216–22.

34.

Murray

Moreno

Rosario

Ellen

Sweat

Kerrigan

. 2007. “The Role of Relationship Intimacy in Consistent Condom Use among Female Sex Workers and Their Regular Paying Partners in the Dominican Republic.” AIDS and Behavior 11:463–70.

35.

Murray

Stephen O.

Roscoe

Will

. eds. 1998. Boy-Wives and Female-Husbands: Studies in African Homosexualities. New York: St. Martin’s Press.

36.

Ngugi

E. N.

Chakkalackal

Sharma

Bukusi

Njoroge

Kimani

MacDonald

K. S.

Bwayo

J. J.

Cohen

C. R.

Moses

Kaul

and Kibera HIV Study Group. 2007. “Kibera HIV Study Group. Sustained Changes in Sexual Behavior by Female Sex Workers after Completion of a Randomized HIV Prevention Trial.” Journal of Acquired Immune Deficiency Syndrome 45:588–94.

37.

Nyanzi

Byanzi-Wakholi

Kalina

. 2009. “The Negotiation of Masculinities by Motorbike Taxi-Riders in Masaka, Uganda.” Men and Masculinities 12:73–89.

38.

Parker

2001. “Sexuality, Culture, and Power in HIV/AIDS Research.” Annual Review of Anthropology 30:163–79.

39.

Pulerwitz

Julie

Barker

Gary

. 2008. “Measuring Attitudes toward Gender Norms among Young Men in Brazil: Development and Psychometric Evaluation of the GEM Scale.” Men and Masculinities 10:322–38.

40.

Pulerwitz

Julie

Barker

Gary

Segundo

Márcio

Nascimento

Marcos

. 2006. “Promoting More Gender-Equitable Norms and Behaviors among Young Men as an HIV/AIDS Prevention Strategy.” Horizons Final Report, Population Council, Washington, DC.

41.

Pulerwitz

Michaelis

Verma

Weiss

. 2010. “Addressing Gender Dynamics and Engaging Men in HIV Programs: Lessons Learned from Horizons Research.” Public Health Reports 125:282–92.

42.

Rivers

Aggleton

. 1999. Men and the HIV Epidemic, Gender and the HIV Epidemic. New York: HIV and Development Program, United Nations Development Programme.

43.

Rottach

Schuler

S. R.

Hardee

. 2009. Gender Perspectives Improve Reproductive Health Outcomes: New Evidence. Washington, DC: Population Reference Bureau.

44.

Segars

A. H.

Grover

. 1993. “Re-examining Perceived Ease of Use and Usefulness: A Confirmatory Factor Analysis.” Management Information Systems Quarterly 17:517–25.

45.

Shattuck

. 2010. “Assessing Gender Norms: Confirmatory Factor Analysis, the Inequitable Gender Norms Subscale of the Gender Equitable Men (GEM) Scale.” Unpublished Manuscript.

46.

Shattuck

Dominick

Kerner

Brad

Gilles

Katherine

Hartmann

Miriam

Ng’ombe

Thokozani

Guest

Greg

. 2011. “Encouraging Contraceptive Uptake by Motivating Men to Communicate about Family Planning: The Malawi Male Motivator Project.” American Journal of Public Health 101:1089–95.

47.

Silberschmidt

Rasch

. 2001. “Adolescent Girls, Illegal Abortions and Sugar-Daddies in Dar es Salaam: Vulnerable Victims and Active Social Agent.” Social Science & Medicine 52:1815–26.

48.

Susser

Stein

. 2004. “Culture, Sexuality, and Women’s Agency in the Prevention of HIV/AIDS in Southern Africa.” In HIV and AIDS in Africa: Beyond Epidemiology, edited by Kalipeni

Craddock

Oppong

Ghosh

, 133–43. Oxford, UK: Blackwell.

49.

Tanzania Commission for AIDS NBoSaOM. 2005. Tanzania HIV/AIDS Indicator Survey 2003-04. Calverton, MD: TACAIDS, NBS, and ORC Macro.

50.

Thompson

Daniel

L. G.

. 1996. “Factor Analytic Evidence for the Construct Validity of Scores: A Historical Overview and Some Guidelines.” Educational and Psychological Measurement 56:197–208.

51.

United Nations. 2011. “Millennium Development Goals.” Accessed June 13, 2011. http://www.beta.undp.org/undp/en/home/mdgoverview.html.

52.

USAID (United States Agency for International Development). 2011a. A Summary Report of New Evidence that Gender Perspectives Improve Reproductive Health Outcomes. Washington, DC: Population Reference Bureau.

53.

US Department of State. 2005. “Global Health Initiative.” Accessed June 13, 2011. http://www.ghi.gov/about/index.htm.

54.

US Government. September 2010. “Global Health Initiative.”

Washington, D.C.

55.

van de Wijgert

J. H.

Khumalo-Sakutukwa

Coggins

Dube

Sabada E.

Nyamapfeni

Prisca

Mwale

Magdalene

Padian

Nancy S.

. 1999. “Men’s Attitude toward Vaginal Microbicides and Microbicide Trials in Zimbabwe.” International Family Planning Perspectives 25:15–20.

56.

van der Staten

King

Grinstead

Vittinghoff

Serufilira

Allen

. 1998. “Sexual Coercion, Physical Violence, and HIV Infection among Women in Steady Relationships in Kigali, Rwanda.” AIDS and Behavior 2:61–73.

57.

Varga

2003. “How Gender Roles Influence Sexual and Reproductive Health among South African Adolescents.” Studies in Family Planning 34:160–72.

58.

WHO (World Health Organization). 1995. “Facing HIV/AIDS/STDs: A Gender-Based Response.” Amsterdam, The Netherlands: World Health Organization.

59.

Wingood

G. M.

DiClemente

R. J.

. 2000. “Application of the Theory of Gender and Power to Examine HIV-Related Exposures, Risk Factors, and Effective Interventions for Women.” Health Education & Behavior 27:539–65.

60.

Zierler

Krieger

. 1997. “Reframing Women’s Risk: Social Inequalities and HIV Infection.” Annual Review of Public Health 18:401–36.