- Email: [email protected]
HIV prevention and marriage: Peer group effects on condom use acceptability in rural Kenya
Social Science & Medicine 116 (2014) 169e177
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
HIV prevention and marriage: Peer group effects on condom use acceptability in rural Kenya Julia Cordero Coma n Trias Fargas 25-27, 08005 Barcelona, Department of Political and Social Sciences, Universitat Pompeu Fabra, Edificio Jaume I, Despacho 20.187, Ramo Spain
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 December 2013 Received in revised form 1 July 2014 Accepted 3 July 2014 Available online 4 July 2014
The twofold function of condom use e contraception and sexually transmitted disease protection e should be taken into account when understanding attitudes towards this practice. Emphasis on the interpretation of condom use as a protective practice conflicts with the norms of fidelity and trust, which regulate marriage. The alternative interpretation of condom use as a contraceptive method may be less problematic. This paper analyzes the extent to which the attitude of married men and women towards condom use with their spouses, and their actual use of condoms within marriage, are affected by their expectations about the dominant attitudes and behaviors in their peer group. I expect that a social consensus on understanding condom use as an HIV-preventive behavior will not make this practice more acceptable within marriage, while social acceptance of modern contraception and, more specifically, of the use of condoms for contraceptive purposes will. Two waves of a longitudinal survey from 1996 to 1999 in rural Kenya are analyzed using fixed-effects regression. Social support for each function of condom use is measured with indicators of the proportion of individuals in the peer group that use condoms for a particular purpose or have a positive attitude towards each of the uses, according to the respondent. The results support the hypothesis for men, but are inconclusive for women. © 2014 Elsevier Ltd. All rights reserved.
Keywords: Contraception Condom use HIV/AIDS Social networks Kenya Africa Marriage Fidelity
1. Introduction One key aspect of the current stage of the HIV epidemic in subSaharan Africa is that most new HIV infections in several countries take place in serodiscordant married or cohabiting couples. Dunkle et al. (2008) have found that 55e93% of the new infections via heterosexual intercourse in urban areas of Rwanda and Zambia occur in this type of couples. Heterosexual sex with a regular partner accounted for an estimated 44% of HIV infections in Kenya in 2006 (Gelmon et al., 2009). Similar or higher proportions are observed in other countries, such as Swaziland and Lesotho (Khobotlo et al., 2009; Mngadi et al., 2009). Contrary to common beliefs, getting married does not work as a preventive strategy against HIV infection (Glynn et al., 2001; Kelly et al., 2003). Therefore, preventive behavior within marriage has become crucial for the reduction of HIV incidence in the sub-Saharan region, and individuals seem to be more and more aware of the risk of infection involved in marital sex (Anglewicz and Kohler, 2009; Watkins, 2004). Nevertheless, spouses are often reluctant to introduce
E-mail address: [email protected]. http://dx.doi.org/10.1016/j.socscimed.2014.07.005 0277-9536/© 2014 Elsevier Ltd. All rights reserved.
condoms into their marital relationships (Chimbiri, 2007), and the reported use of this device within marriage is usually very low (see Demographic and Health Surveys). The salience of the Sexually Transmitted Disease (STD) protection function of condom use should be taken into account when understanding attitudes towards such a practice. HIV-prevention programs have imbued condoms with a meaning that conflicts with marital rules. Condom use has been promoted in sub-Saharan Africa as a particularly suitable preventive method in what has been considered risky sexual contexts, such as commercial sexual exchanges and sexual relations outside or before marriage (Foss et al., 2007; Maharaj, 2001). The contraceptive function of condom use has been somewhat overlooked. Family Planning (FP) programs have campaigned for highly effective contraceptive methods rather than less-effective barrier methods (Ali et al., 2004). Even less common programs designed to integrate STD prevention and FP (Askew and Maggwa, 2002) that have promoted dual protection (use of non-barrier methods and condoms simultaneously) may have reinforced the idea that condoms are only suitable for STD prevention. In fact, dual protection in marital sex has been found to be difficult to accept by both men and women (Maharaj, 2001; Maharaj et al., 2012). Sub-Saharan people tend to
170
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
agree that condoms should be used with partners who cannot be trusted ebecause they might be “promiscuous” (Chimbiri, 2007; Smith, 2007; Tavory and Swidler, 2009). The emphasis on the preventive function of condom use has made this practice incompatible with two of the main norms that regulate marital relations: trust and fidelity. As such, the suggestion of condom use signals that either oneself should not be trusted or that the partner is not trustworthy. Both interpretations are in conflict with the expected behavior of spouses. Married people, especially women, in subSaharan populations are expected to be sexually faithful (Akwara et al., 2003). In addition, public condemnation of sexual unfaithfulness by the numerous churches that are established in these countries (Agadjanian, 2005; Parsitau, 2009), together with the serious threat of the HIV epidemic to people's daily lives, is enforcing the normative disapproval of extramarital sex. The proposal of condom use then brings out the weakness of the relationship base and may lead to divorce or physical abuse (Achan et al., 2009; Muhwava, 2004). The conflict between condom use and marriage is reflected in the stark differences in the levels of condom use depending on the type of partner. The highest prevalence of condom use takes place in commercial sex, while the lowest levels are observed in regular or marital relationships (de Walque and Kline, 2011; Norman, 2003; Westercamp et al., 2010). Thus, condom use within marriage is still a rare phenomenon e in Kenya, the percentage of men who said they used a condom the last time they had sex with a spouse or cohabiting partner was 3.5% and 7% in 2003 and 2008, respectively, according to the Kenya Demographic and Health Surveys (KCBS and ORC Macro, 2004; KNBS and ICF Macro, 2010). These figures are lower when women's reports are considered. Emphasis on the interpretation of condom use as a contraceptive method however may be less problematic, since the acceptance of family planning has dramatically increased in sub-Saharan Africa, especially in the Southeastern countries (Cleland and Ali, 2006). It is clear that such an interpretation would not be a definitive solution, since condom use would only be resorted to by couples who want to stop having children or by couples that want to space births. Nonetheless, it could facilitate the negotiation of condom use in marital sex, and also in other sexual contexts, such as long-term extramarital relations. Moreover, despite the lower contraceptive efficacy of condoms in comparison with that of hormonal methods ethe annual failure rate being 2% for condoms and 0.3% for the pill or injectable contraceptives (Trussell et al., 2009) e a radical shift from contraceptive pills to condom use in the population would involve minor reproductive costs, in terms of abortion and unwanted births, according to a study of 16 developing countries conducted by Ali et al. (2004). Individual interpretation of condom use is thought to be strongly dependent on the prevailing meaning of condoms in the society in question. This paper explores the extent to which the attitudes of individuals towards condom use in marital relations in rural sub-Saharan settings are shaped by the perceived social acceptance of each of the two uses of condoms in their peer group. Married men and women in rural populations of Kenya are the units of analysis in this study. Kenya can be considered rather representative of the Southeastern African region in terms of several socioeconomic indicators and HIV prevalence rates (World Development Indicators 2008). 2. Social interactions and the attitude towards condom use In this paper, I propose an empirical strategy to examine whether a different socially-shared interpretation of a practice leads to different attitudes towards it. Health-behavior models that assume that behavior depends mainly on individual
knowledge, intentions or conscious control of one's life, such as the Social learning theory (Airhihenbuwa and Obregon, 2000), are unable to explain the puzzle of the disproportionate low levels of condom use among married individuals who are aware of, and worried about, the risk of HIV infection through unprotected marital sex (Anglewicz and Kohler, 2009; Watkins, 2004). An alternative approach needs to take into consideration the role that socially-shared understandings play on individuals' interpretation of actions and on the meaning that such actions give to their sexual relationships. As anthropologists have long pointed out, material objects are loaded with meaning. They may act as environmental stimuli that influence actions by making certain values, behavioral rules, or ideas salient (Bicchieri, 2006; Kay et al., 2004). Bringing out ideas such as infidelity, distrust or infection is likely to have important emotional consequences. As explained above, health campaigns have emphasized a particular aspect of condom use, and this may have determined the way people conceive of it. Nevertheless, some variation in attitudes can be observed among different social environments. Interpersonal communication is a key channel through which individuals become aware of dominant opinions and behaviors (Kohler et al., 2007; Watkins, 2004). With regard to protected sex, informal conversations allow individuals to update their beliefs about the reasons why the people around them use condoms, the characteristics of the contexts in which they do so, and their opinions about this practice. Interpersonal and group communication create the opportunity to re-negotiate the meaning of an action, and lead to a common understanding of the situations in which it takes place (Bicchieri, 2006). The main hypothesis of this paper is (a) that married individuals' view of the use of condoms in marital sex does not improve when they believe these devices to be commonly used to avoid HIV infection from extramarital partners, and (b) that married people become more likely to have a positive attitude when modern contraception is perceived to be widely accepted and, more specifically, when they believe that their peers use condoms for contraceptive purposes. The study focuses on personal attitude as it intends to provide evidence on the extent to which an individual is influenced by the peer group in which he/she is immersed. Given that condom use should not be conceived as a purely individual action, focusing on behavior would involve a different research design that enables us to consider the characteristics of both members of the couple, as well as other aspects that constrain the adoption of a practice. A proper examination of this question would require a couple-based analysis, which is not feasible in this paper due to lack of available data. Nevertheless, as an explorative analysis, the study checks whether analogous results are obtained when reported condom use within marriage is taken as the outcome to be explained. To my knowledge, the influence of the social group on the acceptance of condom use within marriage has not been examined from a quantitative perspective. The exception would be the study by Cordero-Coma and Breen (2012), who expected to observe a positive association between the perceived acceptance of extramarital relationships in the group e or weakness of the fidelity norm- and a married couple's likelihood of using condoms in rural Malawi. They did not find, however, a conclusive result in this regard. More recently, Anglewicz and Clark (2013) have explored the effect of the actual and perceived HIV statuses and the risk perceptions of both the respondent and his/her spouse on individual acceptance of condom use within marriage in rural Malawi. The role of the social group is, nevertheless, absent from their model.
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
3. Data and method I analyze longitudinal data that come from the Kenya Diffusion and Ideational Change Project (KDICP). The characteristics and aims of the survey have been described in Watkins et al. (2003), and the quality of the data has been thoroughly analyzed by Bignami-Van Assche et al. (2003). The data were collected in three waves: 1994, 1996, and 1999. Only the last two waves are analyzed here, because the first wave does not provide enough relevant information. Despite the fact that more recent data would be preferable, this dataset still offers a unique opportunity to examine the influence of social interactions on personal attitudes towards condom use in marriage in a sub-Saharan African context. The sampling frame in these surveys is ever-married women of childbearing age and their husbands (if currently married) in rural sites in the South Nyanza District. Given the aim of this research, the units of analysis are married individuals, both men and women. 3.1. Measures The dependent variable is a proxy measure of the respondent's attitude towards condom use in marital sex. Respondents were asked: “Would you feel comfortable suggesting to your spouse that you and he/she use condoms?”. The possible answers were yes or no. It could be argued that such an indicator measures not only the respondent's attitude towards this practice, but some other characteristics of the respondent and his/her marital relationship. It might be related to male and female skills with regard spousal communication on intimate matters, or to the balance of the bargaining power in the couple. Thus, the model also includes the following control variables: a two-category indicator that measures whether the respondent reported having ever talked to his/her spouse about the chances that his/her spouse or himself/herself might get infected with AIDS, and the dichotomous response to the question: Do you think it is acceptable for a woman to divorce an unfaithful husband?. Table 1 shows descriptive statistics for all the variables in the analysis. With regard to the explanatory variables, the KDICP survey asked with how many people the respondent had ever chatted about family planning (FP), on the one hand, and about AIDS, on the other. Then, the survey asked for information on up to four people with whom the respondent had chatted about each of the two topics. Those respondents who reported having ever had an informal conversation about FP with someone were specifically asked: Does [each of these people or network partners] use modern family planning with her/his spouse?, Which method is it? Condom use is one among various options. Those respondents who have talked about AIDS were asked: What does [each network partner] think is the best way to protect herself/himself from getting AIDS?, and “condom use with extramarital partners” was one of the answers formulated. Note that it is possible for us to observe two communication networks: one about FP and the other about AIDS. However, these networks may overlap, so that reported interlocutors in the discussion on one of the topics could be the same people mentioned as interlocutors in conversations about the other. The responses to the questions are used to construct two indicators: the proportion of people in the network that use condoms in marriage as a contraceptive method, and the proportion of network partners that consider the use of condoms with all extramarital partners as the best strategy to prevent HIV infection, both issues according to the respondent. The continuous measures were categorized to differentiate those respondents who reported a network in which such an opinion or behavior was dominant from the rest. Concerning the first indicator -the proportion of users of condoms for
171
Table 1 Summary statistics for the variables in the models, KDICP 1996 and 1999. Men 1996
Women 1999
N 526 564 26.4 27.8 Comfortable suggesting a CU to spouse b 42.7 (12.8) 43.6 (13.3) Age Level of education Never attended 9.9 8.0 Primary 59.1 57.1 Secondary or more 31.0 34.9 Monogamous 65.4 68.4 Religion Catholic 23.4 21.4 Protestant 64.4 69.7 Others 12.2 8.9 Suspects of infidelity 12.6 8.7 Wants no more children 27.6 31.6 Uses other FP methods 12.6 13.8 Acceptable for a woman to divorce 63.3 77.5 Ever talked to spouse about AIDS 73.4 84.6 Proportion of NP who support CU for HIV prevention Half or less 72.8 79.4 More than half 10.1 11.7 No AIDS network 17.1 8.9 Proportion of NP who use condoms as contraception Less than half 66.9 77.4 Half or more 9.1 5.0 No FP network 24.0 17.6 Proportion of NP who use modern contraception Half or less 45.1 59.1 More than half 30.3 23.4 No FP network 24.6 17.5 Proportion of NP who accept modern contraception Half or less 35.2 27.0 More than half 40.2 55.5 No FP network 24.6 17.5 a b
1996
1999
642 28.5
674 29.7
31.1 (7.9) 31.3 (8.5) 19.8 65.9 14.3 61.8
15.3 70.2 14.5 62.5
22.9 67.3 9.8 33.0 35.2 13.4 60.8 55.9
20.2 73.0 6.8 30.1 38.0 17.1 71.7 73.9
68.7 8.1 23.2
82.7 7.1 10.2
44.6 37.5 17.9
60.5 28.2 11.3
32.2 49.9 17.9
22.5 66.2 11.3
CU and NP stand for condom use and network partners, respectively. Means in italics, and standard deviations in parentheses.
contraception- I distinguish the respondents with networks in which half or more, and not more than half, of the network partners were believed to have this behavior from the rest, because otherwise that category would represent less than 5% of the male sample. A third category in each of the variables mentioned refers to individuals who have not chatted about AIDS or FP with anyone. Information regarding their beliefs about the dominant opinions in their social environment is missing. Given the modest size of the sample, I include them in the analysis as a separate category. Nonetheless, estimates for the subsample obtained after removing respondents with no reported network are also calculated. In the case of the female respondents, the perceived share of users of condoms for contraception in the peer group is so small e about 1% in both waves e that the statistical analysis for married women is not feasible. Nonetheless, two alternative indicators enable us to test whether the dominant attitude towards modern contraception overall in the social group has an impact on the dependent variable. Given that respondents who had chatted about FP were also asked whether they thought that each of the network partners approved of FP, two more measures have been constructed: the proportion of the peer group that uses modern FP and the proportion that approves of FP, both according to the respondent. The abovementioned explanatory variables are used as measures of the dominant attitudes and behaviors in the social network in relation to the two uses of condoms. In fact, even when the two communication networks (FP and AIDS) might not be composed of the same network partners, they are both considered proxies of the respondent's reference group. It is important, then, to be aware of
172
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
the characteristics of the interlocutors reported. Table 2 shows some features of the respondents concerning their networks. The uncensored size of the networks, which refers to the average total number of people with whom the respondent had chatted about each topic, is smaller for women in the two waves, and bigger in 1999 than in 1996 for both men and women. The individual accumulation of conversations over time explains, at least partially, the increase in interlocutors between waves, given that 75% and 72% of the married men and women in the sample in 1996, respectively, were reinterviewed in 1999. As mentioned above, respondents were asked to provide information about only four people at most with whom they had talked about FP, on the one hand, and AIDS, on the other. These limited samples of network partners compose what I call “censored networks”, which have an average size of more than three in all the cases. The KDICP survey does not allow us to know the respondents' criteria for selecting certain interlocutors to give information about, but Table 2 shows that censored networks are mainly composed of individuals of the same sex as the respondent, who are linked to him or her by strong ties (for example, in 1996, 70% of married men had chatted about FP only with men and 80% of married men had a FP communication network in which all the partners were reported to be confidants or friends, as opposed to acquaintances). Finally, about 35% of the respondents in 1999 had an AIDS communication network that overlapped with the FP network. In sum, it seems reasonable to consider that the indicators used in the paper about the proportion of network partners who have a particular behavior or opinion are proxy measures of dominant aspects in the immediate social network or peer group. The models include other factors that are expected to have a relevant influence on the attitude towards condom use within marriage. First of all, the respondent's interest in the outcomes of condom use is taken into account: the models consider whether the respondent (a) suspected that his/her spouse had been unfaithful, as an indicator of the perceived risk of getting infected through non-protected intercourse, and (b) reported his/her will to stop having children. A lack of available data prevents us from considering birth spacing intentions. The current use of other modern contraceptive methods that are more effective than condoms e the
pill, injectable contraceptives, IUD, or sterilization e is also taken into account, because it is expected to hinder the negotiation of condom use for pregnancy avoidance within marriage. Unfortunately, information about perceived affordability of condoms was not available. Respondents' age, level of education, type of marriage (monogamous or polygamous), and religion are also considered in the analysis. The positive influence of education on the acceptance of modern contraception and condoms has been observed in numerous empirical studies (de Walque, 2007; Zellner, 2003). Age is included as a continuous variable. Note that the analysis is restricted to women aged between 15 and 49. Women's capacity to discuss issues related to reproductive and sexual health with their husbands might be different in monogamous and polygamous marriages. Finally, the strong general opposition of the Catholic Church to artificial contraception is expected to induce more negative attitudes among its followers in comparison with other religions (Caldwell, 2000). However, previous studies in subSaharan countries have not provided conclusive evidence in this regard. This may be due to the vast variation in the adherence of local churches to the official position of their denomination on these issues (Freidman, 1995), or to differences in religiosity among individuals (Trinitapoli and Regnerus, 2006). An alternative dichotomous dependent variable that refers to the actual use of condoms in marriage is considered in the last section of the analysis. Respondents were asked the specific contraceptive method they currently used with their spouse. “Condoms” was one of the options, and multiple answers were not allowed. Only 7% and 5% of married men in 1996 and 1999, respectively, reported condoms as the contraceptive method used with their spouses. The corresponding percentages for women are so small (1% and 0.6%) that the statistical analysis for the female sample is not possible. 3.2. Method First, the attitude towards suggesting condom use to a spouse is analyzed cross-sectionally. The data from the two waves have been pooled in order to avoid working with too small sets of cases. This
Table 2 Description of the networks. Men
Family planning network N Uncensored size Proportion with at least one NP N (only those with one NP at least) Censored size Proportion with a censored network in which: All NPs are men All NPs are women All NPs are respondent's confidants or friends AIDS network N Uncensored size Proportion with at least one NP N (only those with one NP at least) Censored size Proportion with a censored network in which:a All NP are men All NP are women All NP are respondent's confidants or friends At least one NP is part of the FP network
Women
1996
1999
1996
1999
526 5.08 (5.91) 0.76 397 3.18 (1.01)
564 6.57 (7.03) 0.82 465 3.39 (0.90)
642 4.48 (4.73) 0.82 527 3.10 (1.01)
674 5.21 (5.24) 0.89 598 3.23 (1.00)
0.70
0.69
0.80
0.86
0.85 0.76
0.87 0.88
526 6.58 (7.89) 0.83 436 3.28 (0.97)
564 9.22 (10.18) 0.91 516 3.59 (0.77)
641 4.87 (5.92) 0.77 493 3.10 (1.05)
674 6.36 (7.05) 0.90 604 3.31 (0.99)
0.74
0.73
0.75 0.39
0.87 0.35
0.57 0.71 0.27
0.60 0.84 0.36
a In 1996, such proportions (in italics) refer to the network partners (NP) mentioned only as part of the AIDS censored network, because we are unable to identify the characteristics of the NP that were mentioned as people with whom the respondent also chatted about FP in this wave.
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
technique seems appropriate since there are no particular reasons to believe that the effects of the explanatory variables substantially vary from the first wave to the successive one. The relaxation of the assumption that errors are uncorrelated is possible by using clustered standard errors. The estimation of the influence of social interactions on individual behavior is challenging since it must deal with a potential self-selection problem. Social networks are rarely randomly distributed because individuals tend to select their interlocutors, usually preferring those with whom they share certain characteristics, attitudes and preferences. A positive correlation in a crosssectional analysis between the respondent's attitude and that of his/her network partners may simply reflect this systematic selection rather than provide evidence of the influence exerted by the social environment. The panel analysis with fixed effects is a suitable tool in this case, given that it allows us to control for timeinvariant observed and unobserved characteristics that may affect both the respondent's attitude and his/her likelihood of interacting with specific people (Behrman et al., 2001). A drawback of the fixed-effects model with a binary dependent variable is that the sample size might decrease notably, because only those individuals who experience a change between waves in the outcome variable are analyzed. This could lead to wrong conclusions, especially if the distribution of the explanatory variables in the subsample changes substantively. Linear probability models are then used as a robustness check, because the estimation of linear models with fixed effects does not imply a reduction in the number of analyzed cases. I adjust for the violations of the classical OLS model assumptions (normality and homoscedasticity of the disturbance term) by using robust standard errors (Behrman et al., 2001). 4. Results 4.1. Attitude towards condom use within marriage Table 3 displays the results of the fixed-effects logit regression of feeling comfortable suggesting condom use to their spouse, for married men and women. Note that only those respondents
173
interviewed in both waves, 1996 and 1999, are examined, and that the time-invariant variables e education and religion e drop out of the models. This is not the case in the cross-sectional analysis, in which all the variables are included (see the results in Table A.1 in the Appendix). Models 1, 2 and 3 (Table 3) differ from each other in the variable that they include to measure the acceptance in the social network of one of the uses of the condom, namely, contraception. Model 1 includes the most suitable measure, considering the aim of this research. According to Model 1, having a peer group in which half or more of the people are identified by the respondent as users of condoms for contraceptive purposes significantly increases the married men's likelihood of reporting a positive attitude towards condom use within marriage. Conversely, peer groups in which more than half of the partners are believed to consider that condom use outside marriage is the best HIV preventive strategy do not make any relevant difference to men's attitude. This result is coherent with the research hypothesis, according to which a perceived socially rooted interpretation of condoms as an HIVpreventive device does not facilitate the acceptance of its use in marriage, whereas a favorable social environment for condoms as a contraceptive method makes individuals more willing to introduce condoms in their marital sexual relations. The longitudinal analysis confirms that the influence of the peer group is not due to selection effects. Unfortunately, Model 1 cannot be estimated for women, as explained in the Measures subsection. We are only able to verify that a perceived widespread interpretation of condoms as suitable HIV preventive devices does not improve their acceptance in marital sex, even when controlling for the social approval of modern contraception. Table 3 shows that having a social network in which most people use modern contraception (Model 2) or approve of it (Model 3) clearly improves married men's attitude. The significance of the rest of the coefficients is similar to those in Model 1. Men are, then, unlikely to feel comfortable negotiating condom use when the use of modern contraception is not a prevailing behavior or is widely disapproved of in their peer group. In contrast, whether men perceive an extensive support of condom use for HIV preventive purposes has no influence on their attitude.
Table 3 Multivariate logit regression analysis with fixed effects of feeling comfortable suggesting CU to a spouse, married men and women in KDICP 1996e1999. Men Model 1
Women Model 2
Monogamous 0.061 (0.552) 0.495 (0.588) Suspects of infidelity 1.112* (0.489) 1.144* (0.525) Wants no more children 0.052 (0.396) 0.109 (0.423) Uses other FP methods 0.577 (0.483) 1.081* (0.512) Acceptable for a woman to divorce 0.563 (0.354) 0.353 (0.370) Ever talked to spouse about AIDS 0.940þ (0.562) 0.747 (0.546) Proportion of NP who support CU for HIV prevention (Ref: Half or less) More than half 0.212 (0.514) 0.05 (0.526) Has no AIDS network 0.548 (0.623) 0.419 (0.619) Proportion of NP who use condoms for contraception (Ref: Less than half) Half or more 1.317* (0.625) Has no FP network 0.937þ (0.541) Proportion of NP who use modern contraception (Ref: Half or less) More than half 1.535** (0.452) Has no FP network (0.710) (0.571) Proportion of NP who accept modern contraception (Ref: Half or less) More than half Has no FP network Year 0.066 (0.077) 0.071 (0.080) Constant N 232 232 þ
p < 0.1; *p < 0.05; **p < 0.01. Standard errors in parentheses.
Model 3
Model 2
Model 3
0.216 (0.574) 0.870þ (0.494) 0.083 (0.414) 0.699 (0.482) 0.437 (0.368) 0.736 (0.554)
0.135 (0.626) 1.320** (0.345) 0.286 (0.356) 0.548 (0.377) 0.055 (0.279) 0.462 (0.347)
0.214 (0.625) 1.360** (0.349) 0.254 (0.358) 0.597 (0.383) 0.074 (0.279) 0.503 (0.353)
0.155 (0.524) 0.295 (0.630)
0.506 (0.493) 0.060 (0.506)
0.440 (0.496) 0.022 (0.502)
0.029 (0.322) 0.020 (0.560) 1.129** (0.368) 0.500 (0.589) 0.146þ (0.081)
0.033 (0.071)
0.326 (0.353) 0.161 (0.591) 0.024 (0.069)
232
288
288
174
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
Table 4 Fixed-effects regression analysis e linear probability model e of actual CU, married men in KDICP 1996e1999. Model 4
Model 5
Model 6
LPM
LPM
LPM
0.041 (0.028) 0.086* (0.039) 0.042 (0.032) 0.050 (0.036) 0.008 (0.025) 0.071** (0.026)
0.049þ (0.027) 0.083* (0.039) 0.039 (0.032) 0.047 (0.036) 0.008 (0.025) 0.076** (0.027)
0.041 (0.042) 0.046 (0.034)
0.037 (0.043) 0.049 (0.034)
Monogamous 0.047þ (0.028) Suspects of infidelity 0.087* (0.040) Wants no more children 0.048 (0.032) Ever used other FP methods 0.028 (0.035) Acceptable for a woman to divorce 0.005 (0.025) Ever talked to spouse about AIDS 0.070** (0.026) Proportion of NP who support CU for HIV prevention (Ref: half or less) More than half 0.030 (0.044) Has no AIDS network 0.044 (0.032) Proportion of NP who use condoms for contraception (Ref: Less than half) Half or more 0.147* (0.060) Has no FP network 0.027 (0.030) Proportion of NP who use modern contraception (Ref: half or less) More than half Has no FP network Proportion of NP who accept modern contraception (Ref: half or less) More than half Has no FP network Year 0.006 (0.005) Constant 11.942 (10.020) N 794
0.088** (0.029) 0.015 (0.028)
0.006 (0.005) 11.953 (9.969) 794
0.049þ (0.027) 0.014 (0.029) 0.002 (0.005) 3.426 (10.766) 794
þ p < 0.1; *p < 0.05; **p < 0.01. Robust standard errors in parentheses.
However, none of the variables about the acceptance of modern contraception in the peer group are statistically significant in the longitudinal analysis of women (Table 3), in spite of the fact that the cross-sectional analysis shows similar results for both sexes (see Table A.1 in the Appendix). Concerning the control variables in the models, some findings from the panel analysis should be highlighted. The suspicion of infidelity has a substantive positive effect on the attitude towards condom use in marriage, particularly for women, whereas the desire to stop having births is not statistically significant. The desire for protection may thus be the main motivation for accepting the use of condoms within marriage, even when it is proposed to the spouse in terms of contraception intentions. This finding lines up well with the results in the study of rural Malawi by Anglewicz and Clark (2013). As a robustness test for all the logit models in Table 3, I also estimated linear probability regressions. The signs and significance of the coefficients barely change. In addition, I replicated the analyses presented in this and in the following subsection on a subsample in which individuals with no reported network were removed. The results are consistent with those shown in Tables 3 and 4. Details are available upon request.
spouse. This measure refers then to a regular, as opposed to an exceptional, use of condoms. It is, however, an imperfect indicator of use, because respondents could only mention one method, even if they used dual protection. Models 4, 5 and 6 in Table 4 are analogous to Models 1, 2 and 3, with a minor change in the specification: the models control for whether the respondent has ever used other contraceptive methods with his spouse, instead of whether he currently uses them. Only linear probability models are presented here, because the standard deviations of the estimated coefficients in the logit models are too high, due to the reduced sample size and the small proportion of positive values in the dependent variable. Table 4 displays the results, which are in line with what has been pointed out above. A married man's probability of using condoms in marital sex is positively influenced by whether he believes that a large share of his peers use condoms as a contraceptive method (Model 4), use modern contraception (Model 5), or accept its use (Model 6). On the contrary, the respondent's belief about the social acceptance of condoms for HIV prevention has no impact on his sexual behavior within marriage.
4.2. Use of condoms within marriage
The study has shown that the attitudes, and very likely behavior as well, of individuals are substantially influenced by their expectations about the dominant attitudes and behaviors in the peer group. Moreover, the empirical analysis has offered some evidence that the perceived social interpretation of condom use modifies the attitude of individuals towards the use of this device in marital sex. Perceiving that modern contraception and, particularly, the use of condoms for this purpose are widely accepted in the peer group makes individuals more likely to report feeling comfortable about suggesting condom use to their spouses. On the contrary, the personal attitude does not improve when perceiving condoms to be commonly understood as a suitable preventive practice against HIV in extramarital sex. Similar conclusions are extracted from an analogous analysis that takes the reported use of condoms within marriage as the dependent variable. Besides, the estimated influence of the peer group is unlikely to be driven by a selection effect.
A proper examination of condom use would be based on a couple-level analysis (Cordero-Coma and Breen, 2012) or, at least, it would take into consideration the characteristics of the two people involved (Anglewicz and Clark, 2013). Although the KDICP enables us to link married respondents to their current spouses, the final number of couples on which there is enough information is limited, which renders such an analysis unfeasible. In this section, I replicate the longitudinal analysis described above, but with a dependent variable that refers to actual condom use within marriage. This should be interpreted as a strategy to obtain additional evidence for the main argument in the paper, rather than as an attempt to provide a compelling explanation of condom use. The outcome variable distinguishes married men who reported condoms as the contraceptive method currently used with their
5. Discussion
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
Such findings support the argument defended here, that an alternative interpretation of condom use which emphasizes its contraceptive function rather than its suitability for HIV protection from untrustworthy sexual partners makes the acceptance of this practice within marriage possible. This was not a trivial question when the data examined here were gathered, and even less so nowadays, since marital sex is the main channel through which HIV is transmitted at the current stage of the epidemic in several sub-Saharan countries. Therefore, a relevant policy implication for HIV prevention campaigns can be derived from the study. The results suggest that the promotion of condom use should be guided by the aim of making such a practice compatible with the norms that regulate not only marriage, but also other steady relationships. In addition, the study provides some indirect empirical support for peer education programs and other activities aimed at groups, because they take into account the relevance of group communication for individuals to update their expectations about what the others know, do and think, which affects, in turn, their opinions and behavior. Replicating these analyses with more recent data would most likely provide similar results, given that norms of trust and fidelity still regulate marriage, and the widespread meaning of condoms is unlikely to have changed. It may be, however, that suggesting the use of condoms for contraceptive purposes is now less convincing as the association of condoms with ‘illicit sex’ becomes entrenched in society and people are more and more concerned about infection from spouses (Watkins, 2004). An important limitation of this study is that the analyses have only been partially done for married women, because very few female respondents reported using condoms with their spouses, on the one hand, and having network partners, most of whom are female friends, who use condoms as a contraceptive method within marriage, on the other. It might be that men tend to talk with their peers more about condoms than about other contraceptive methods, such as the pill, which are not directly under their control, and an analogous reasoning would work for women. In line with this, men might be more likely to report condoms as the contraceptive method employed when the couple uses dual protection and the survey forces respondents to identify just one method. Furthermore, studies of husband-wife discrepancies in the reporting of behavior have highlighted that men may be likely to overreport their use of contraceptive methods in survey interviews for reasons related to how they want to present themselves to such interlocutors (Cordero-Coma and Breen, 2012; Miller et al., 2001). This could explain the higher levels of condom use in marriage according to men, but it is unlikely to explain the respondents'
175
report of their peers' use. Moreover, the motives that lead individuals to misreport their behavior when having a conversation with their peers may not apply to the context of a survey interview. Besides being incomplete, the analysis of women has not offered conclusive evidence, given that the expectations about the peer group have not been found to have a substantive effect on their attitudes in the longitudinal analysis. Women's lesser control over their sexual life may hinder our understanding of their attitude towards introducing something new into their relationship if we do not take into account their husbands' characteristics. Anglewicz and Clark (2013) observed, however, that the reported acceptance of condom use in marriage by the spouse does not affect the attitude of the respondent in rural Malawi. Nonetheless, further research on the role of women's perception of the attitude of their partners, rather than of the reported attitude by their husbands, would contribute to unraveling this question. Our finding that the perceived acceptance of modern contraception in the peer group has no impact on women's attitude would also make sense if the contraceptive function of condom use were especially less salient for women. This lines up well with the fact that women used to be the main target of FP programs, particularly until the UN Conference in Cairo 1994 (McIntosh and Finkle, 1995), which favored the promotion of non-barrier contraceptive methods over that of condoms. Given that individuals tend to talk about reproductive health with people of their same sex, as observed here, the spread of an alternative interpretation of condom use could then be particularly difficult among women. Some other limitations of the study must be mentioned. The panel analysis with fixed effects is unable to deal with potential biases caused by unobserved time-varying characteristics. In addition, individuals that move from the survey village clusters have not been analyzed. The inclusion of these particularly interesting units of analysis would enrich the study, because they are more likely to experience changes in the characteristics of their social networks. The rooted connotations of condom use that link this practice to the idea of ‘illicit sex’ are, unfortunately, not the only obstacle for the use of this device. Individuals tend to complain of the reduction in sexual pleasure that condoms involve (Thomsen et al., 2004). Even fears of side effects and health risks believed to be caused by condom use are quite extended (Siegler et al., 2012). Nonetheless, this research suggests that the diffusion of an alternative interpretation of condom use that emphasizes its contraceptive function would help individuals to cope with the most relevant threats at this stage of the HIV epidemic.
Appendix Table A.1 Multivariate logit regression of feeling comfortable suggesting condom use (CU) to a spouse. Pooled data of married men and women, KDICP 1996 and 1999. Men A.1 Age 0.028** (0.009) Level of education (Ref: Never attended school) Primary 0.538 (0.372) Secondary 0.904* (0.390) Monogamous 0.123 (0.185) Religion (Ref: Catholic) Protestant 0.462* (0.198) Others 0.302 (0.351) Suspects of infidelity 0.967** (0.227) Wants no more children 0.392* (0.200) Uses other FP methods 0.044 (0.198)
Women A.2
A.3
A.2
A.3
0.030** (0.009)
0.027* (0.008)*
0.013 (0.010)
0.013 (0.010)
0.616 (0.361) 0.935* (0.381) 0.176 (0.188)
0.586 (0.377) 0.957* (0.399) 0.172 (0.188)
0.336þ (0.201) 0.540* (0.250) 0.149 (0.142)
0.311 (0.202) 0.508* (0.250) 0.134 (0.142)
0.417* (0.197) 0.194 (0.348) 0.914** (0.229) 0.284 (0.199) 0.244 (0.213)
0.388þ (0.203) 0.238 (0.340) 0.941** (0.222) 0.211 (0.199) 0.193 (0.208)
0.184 (0.172) 0.385 (0.267) 0.783** (0.132) 0.139 (0.153) 0.041 (0.181)
0.193 (0.172) 0.393 (0.271) 0.776** (0.133) 0.112 (0.154) 0.052 (0.179)
þ
(continued on next page)
176
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
Table A.1 (continued ) Men A.1
Women A.2
Acceptable for a woman to divorce 0.274 (0.170) 0.272 (0.170) Ever talked to spouse about AIDS 0.664** (0.229) 0.603** (0.227) Proportion of NP who support CU for HIV preventiona (Ref: Half or less) More than half 0.249 (0.217) 0.326 (0.213) Has no AIDS network 0.449 (0.337) 0.409 (0.328) Proportion of NP who use condoms for contraception (Ref: Less than half) Half or more 1.622** (0.256) Has no FP network 0.558* (0.251) Proportion of NP who use modern contraception (Ref: Half or less) More than half 0.906** (0.164) Has no FP network 0.460þ (0.254) Proportion of NP who accept modern contraception (Ref: Half or less) More than half Has no FP network Constant 1.731* (0.674) 1.686* (0.662) N 1090 1089
A.3
A.2
A.3
0.212 (0.166) 0.510* (0.225)
0.054 (0.142) 0.553** (0.150)
0.070 (0.142) 0.543** (0.149)
0.273 (0.220) 0.438 (0.324)
0.263 (0.225) 0.156 (0.204)
0.263 (0.226) 0.150 (0.205)
0.338* (0.140) 0.306 (0.226) 1.023** (0.187) 0.142 (0.281) 1.974** (0.680) 1089
1.806** (0.436) 1316
0.535** (0.154) 0.076 (0.244) 2.002** (0.442) 1316
þ p < 0.1; *p < 0.05; **p < 0.01. Clustered standard errors in parentheses. a The abbreviations CU and NP stand for ‘condom use’ and ‘network partners’, respectively.
References Achan, S.F., Akweongo, P., Debpuur, C., Cleland, J., 2009. Coping strategies of young mothers at risk of HIV/AIDS in the Kassena-Nankana District of Northern Ghana. Afr. J. Reprod. Health 13 (1), 61e78. Agadjanian, V., 2005. Gender, religious involvement, and HIV/AIDS prevention in Mozambique. Soc. Sci. Med. 61, 1529e1539. Airhihenbuwa, C.O., Obregon, R., 2000. A critical assessment of theories/models used in Health Communication for HIV/AIDS. J. Health Commun. 5 (Suppl.), 5e15. Akwara, P.A., Madise, N.J., Hinde, A., 2003. Perception of risk of HIV/AIDS and sexual behavior in Kenya. J. Biosoc. Sci. 35, 385e411. Ali, M.M., Cleland, J., Shah, I.H., 2004. Condom use within marriage: a neglected HIV intervention. Bull. World Health Organ. 82, 180e186. Anglewicz, P., Kohler, H.-P., 2009. Overestimating HIV infection: the construction and accuracy of subjective probabilities of HIV infection in rural Malawi. Demogr. Res. 20 (6), 65e96. Anglewicz, P., Clark, S., 2013. The effect of marriage and HIV risks on condom use acceptability in rural Malawi. Soc. Sci. Med. 97, 29e40. Askew, I., Maggwa, N.B., 2002. Integration of STI prevention and management with family planning and antenatal care in sub-Saharan Africa e what more do we need to know? Int. Fam. Plan. Perspect. 28 (2), 77e86. Behrman, J.R., Kohler, H.-P., Watkins, S., 2001. How Can We Measure the Causal Effects of Social Networks Using Observational Data? Evidence from the Diffusion of Family Planning and AIDS Worries in South Nyanza District, Kenya. MPIDR Working Paper 22. Bicchieri, C., 2006. The Grammar of Society. The Nature and Dynamics of Social Norms. Cambridge University Press, New York. Bignami-Van Assche, S., Reniers, G., Weinreib, A.A., 2003. An assessment of the KDICP and the MDICP data quality: interviewer effects, question reliability and sample attrition. Demogr. Res. Special Collection. 1, 31e76. Caldwell, J.C., 2000. Rethinking the African AIDS epidemic. Popul. Dev. Rev. 26 (1), 117e135. Chimbiri, A.M., 2007. The condom is an ‘intruder’ in marriage: evidence from rural Malawi”. Soc. Sci. Med. 64, 1102e1115. Cleland, J., Ali, M.M., 2006. Sexual abstinence, contraception, and condom use by young African women: a secondary analysis of survey data. Lancet 368, 1788e1793. Cordero-Coma, J., Breen, R., 2012. HIV prevention and social desirability: husbandwife discrepancies in reports of condom use. J. Marriage Fam. 74 (3), 601e613. de Walque, D., 2007. How does the impact of an HIV/AIDS information campaign vary with educational attainment? Evidence from rural Uganda. J. Dev. Econ. 84 (2), 686e714. de Walque, D., Kline, R., 2011. Variations in sub-Saharan African Countries. Stud. Fam. Plan. 42 (1), 1e10. Dunkle, K., Stephenson, R., Karita, E., Chomba, E., Kayitenkore, K., Vwalika, C., Greenberg, L., Allen, S., 2008. New heterosexually transmitted HIV in married or cohabiting couples in urban Zambia and Rwanda: an analysis of survey and clinical data. Lancet 371, 2183e2191. Foss, A.M., Hossain, M., Vickerman, P.T., Watss, C.H., 2007. A systematic review of published evidence on intervention impact on condom use in sub-Saharan Africa and Asia. Sex. Transm. Infect. 83, 510e516. Freidman, G.S., 1995. AIDS prevention and the Church. Kenya: mixed messages. AIDS Soc. 6 (2), 4. Gelmon, L., Kenya, P., Oguya, F., Cheluget, B., Haile, G., 2009. Kenya HIV Prevention Response and Modes of Transmission Analysis. National AIDS Control Council, Nairobi, Kenya.
Glynn, J.R., Carael, M., Auvert, B., Kahindo, M., Chege, M., Musonda, R.M., 2001. Why do young women have a much higher prevalence of HIV than young men? A study in Kisumu, Kenya and Ndola, Zambia. AIDS 15 (Suppl. 4), S51eS60. Kay, A.C., Wheeler, C.S., Bargh, J.A., Ross, L., 2004. Material priming: the influence of mundane physical objects on situational construal and competitive behavioral choice. Organ. Behav. Hum. Decis. Process. 95, 83e96. Kelly, R.J., Gray, R., Sewankambo, N., Serwadda, D., Wabwire-Mangen, F., Lutalo, T., Wawer, M.J., 2003. Age differences in sexual partners and risk of HIV-1 infection in rural Uganda. J. Acquir. Immune Defic. Syndr. 32, 446e451. Kenya Central Bureau of Statistics and ORC Macro, 2004. Kenya Demographic and Health Survey 2003. CBS and ORC Macro, Calverton, MD. Kenya National Bureau of Statistics and ICF Macro, 2010. Kenya Demographic and Health Survey 2008-2009. KNBS and ICF Macro, Calverton, MD. Khobotlo, M., Tshehlo, R., Nkonyana, J., Ramoseme, M., Khobotle, M., Chitoshia, A., Hildebrand, M., Fraser, N., 2009. Lesotho: HIV Prevention Response and Modes of Transmission Analysis. National AIDS Commission, Maseru, Lesotho. Kohler, H.P., Behrman, J.R., Watkins, S.C., 2007. Social networks and HIV/AIDS risk perceptions. Demography 44 (1), 1e33. Maharaj, P., 2001. Obstacles to negotiating dual protection: perspectives of men and women. Afr. J. Reprod. Health 5, 150e161. Maharaj, P., Neema, S., Cleland, J., Busza, J., Shah, I., 2012. Condom use within marriage: an assessment of changes in South Africa and Uganda. AIDS Care 24 (4), 444e450. McIntosh, C.A., Finkle, J.L., 1995. The Cairo conference on population and development: a new paradigm. Popul. Dev. Rev. 21 (2), 223e260. Miller, K., Zulu, E.M., Watkins, S.C., 2001. Husband-wife survey responses in Malawi. Stud. Fam. Plan. 32 (2), 161e174. Mngadi, S., Fraser, N., Mkhatshwa, H., Lapidos, T., Khumalo, T., Tsela, S., Nhlabatsi, N., Odido, H., 2009. Swaziland: HIV Prevention Response and Modes of Transmission Analysis. National Emergency Response Council on HIV/AIDS, Mbabane, Swaziland. Muhwava, W., 2004. Condom use within marriage and consensual unions in the era of HIV/AIDS in Zimbabwe. Afr. Popul. Stud. 19 (1), 119e141. Norman, L.R., 2003. Predictors of consistent condom use: a hierarchical analysis of adults from Kenya, Tanzania and Trinidad. Int. J. STD AIDS 14, 584e590. Parsitau, D.S., 2009. Keep holy distance and abstain till he comes: interrogating a Pentecostal Church's engagements with HIV/AIDS and the youth in Kenya. Afr. Today 56, 45e54. Siegler, A.J., Mbwambo, J.K., McCarty, F.A., DiClemente, R.J., 2012. Condoms “contain worms” and “cause HIV” in Tanzania: Negative Condom Beliefs Scale development and implications for HIV prevention. Soc. Sci. Med. 75 (9), 1685e1691. Smith, D.J., 2007. Modern marriage, men’s extramarital sex, and HIV-risk in Southeastern Nigeria. Am. J. Public Health 97, 997e1005. Tavory, I., Swidler, A., 2009. Condom semiotics: meaning and condom use in rural Malawi. Am. Sociol. Rev. 74 (2), 171e189. Thomsen, S.C., Stalker, M., Toroitich-Ruto, C., Ndugga, B.M., Mwarogo, P., 2004. Fifty ways to leave your rubber: how men in Mombasa rationalize unsafe sex. Sex. Transm. Infect. 80, 430e434. Trinitapoli, J., Regnerus, M.D., 2006. Religion and HIV risk behaviors among married men: initial results from a study in rural sub-Saharan Africa. J. Sci. Study Relig. 45 (4), 505e528. Trussell, J., Lalla, A.M., Doan, Q.V., Reyes, E., Pinto, L., Gricar, J., 2009. Cost effectiveness of contraceptives in the United States. Contraception 79, 5e14. Watkins, S.C., 2004. Navigating the AIDS epidemic in rural Malawi. Popul. Dev. Rev. 30 (4), 673e705.
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177 Watkins, S., Behrman, P., Kohler, H.-P., Zulu, E.M., 2003. Introduction to research on demographic aspects of HIV/AIDS in rural Africa. Demogr. Res. Special Collection. 1 (1), 1e30. Westercamp, N., Matsson, C.L., Madonia, M., Moses, S., Agot, K., Ndinya-Achola, J.O., Otieno, E., Ouma, N., Bailey, R.C., 2010. Determinants of consistent condom use
177
vary by partner type among young men in Kisumu, Kenya: a multi-level data analysis. AIDS Behav. 14 (4), 949e959. ^ te d’Ivoire. Zellner, S.L., 2003. Condom use and the accuracy of AIDS knowledge in Co Int. Fam. Plan. Perspect. 29 (1), 41e47.
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
HIV prevention and marriage: Peer group effects on condom use acceptability in rural Kenya Julia Cordero Coma n Trias Fargas 25-27, 08005 Barcelona, Department of Political and Social Sciences, Universitat Pompeu Fabra, Edificio Jaume I, Despacho 20.187, Ramo Spain
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 December 2013 Received in revised form 1 July 2014 Accepted 3 July 2014 Available online 4 July 2014
The twofold function of condom use e contraception and sexually transmitted disease protection e should be taken into account when understanding attitudes towards this practice. Emphasis on the interpretation of condom use as a protective practice conflicts with the norms of fidelity and trust, which regulate marriage. The alternative interpretation of condom use as a contraceptive method may be less problematic. This paper analyzes the extent to which the attitude of married men and women towards condom use with their spouses, and their actual use of condoms within marriage, are affected by their expectations about the dominant attitudes and behaviors in their peer group. I expect that a social consensus on understanding condom use as an HIV-preventive behavior will not make this practice more acceptable within marriage, while social acceptance of modern contraception and, more specifically, of the use of condoms for contraceptive purposes will. Two waves of a longitudinal survey from 1996 to 1999 in rural Kenya are analyzed using fixed-effects regression. Social support for each function of condom use is measured with indicators of the proportion of individuals in the peer group that use condoms for a particular purpose or have a positive attitude towards each of the uses, according to the respondent. The results support the hypothesis for men, but are inconclusive for women. © 2014 Elsevier Ltd. All rights reserved.
Keywords: Contraception Condom use HIV/AIDS Social networks Kenya Africa Marriage Fidelity
1. Introduction One key aspect of the current stage of the HIV epidemic in subSaharan Africa is that most new HIV infections in several countries take place in serodiscordant married or cohabiting couples. Dunkle et al. (2008) have found that 55e93% of the new infections via heterosexual intercourse in urban areas of Rwanda and Zambia occur in this type of couples. Heterosexual sex with a regular partner accounted for an estimated 44% of HIV infections in Kenya in 2006 (Gelmon et al., 2009). Similar or higher proportions are observed in other countries, such as Swaziland and Lesotho (Khobotlo et al., 2009; Mngadi et al., 2009). Contrary to common beliefs, getting married does not work as a preventive strategy against HIV infection (Glynn et al., 2001; Kelly et al., 2003). Therefore, preventive behavior within marriage has become crucial for the reduction of HIV incidence in the sub-Saharan region, and individuals seem to be more and more aware of the risk of infection involved in marital sex (Anglewicz and Kohler, 2009; Watkins, 2004). Nevertheless, spouses are often reluctant to introduce
E-mail address: [email protected]. http://dx.doi.org/10.1016/j.socscimed.2014.07.005 0277-9536/© 2014 Elsevier Ltd. All rights reserved.
condoms into their marital relationships (Chimbiri, 2007), and the reported use of this device within marriage is usually very low (see Demographic and Health Surveys). The salience of the Sexually Transmitted Disease (STD) protection function of condom use should be taken into account when understanding attitudes towards such a practice. HIV-prevention programs have imbued condoms with a meaning that conflicts with marital rules. Condom use has been promoted in sub-Saharan Africa as a particularly suitable preventive method in what has been considered risky sexual contexts, such as commercial sexual exchanges and sexual relations outside or before marriage (Foss et al., 2007; Maharaj, 2001). The contraceptive function of condom use has been somewhat overlooked. Family Planning (FP) programs have campaigned for highly effective contraceptive methods rather than less-effective barrier methods (Ali et al., 2004). Even less common programs designed to integrate STD prevention and FP (Askew and Maggwa, 2002) that have promoted dual protection (use of non-barrier methods and condoms simultaneously) may have reinforced the idea that condoms are only suitable for STD prevention. In fact, dual protection in marital sex has been found to be difficult to accept by both men and women (Maharaj, 2001; Maharaj et al., 2012). Sub-Saharan people tend to
170
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
agree that condoms should be used with partners who cannot be trusted ebecause they might be “promiscuous” (Chimbiri, 2007; Smith, 2007; Tavory and Swidler, 2009). The emphasis on the preventive function of condom use has made this practice incompatible with two of the main norms that regulate marital relations: trust and fidelity. As such, the suggestion of condom use signals that either oneself should not be trusted or that the partner is not trustworthy. Both interpretations are in conflict with the expected behavior of spouses. Married people, especially women, in subSaharan populations are expected to be sexually faithful (Akwara et al., 2003). In addition, public condemnation of sexual unfaithfulness by the numerous churches that are established in these countries (Agadjanian, 2005; Parsitau, 2009), together with the serious threat of the HIV epidemic to people's daily lives, is enforcing the normative disapproval of extramarital sex. The proposal of condom use then brings out the weakness of the relationship base and may lead to divorce or physical abuse (Achan et al., 2009; Muhwava, 2004). The conflict between condom use and marriage is reflected in the stark differences in the levels of condom use depending on the type of partner. The highest prevalence of condom use takes place in commercial sex, while the lowest levels are observed in regular or marital relationships (de Walque and Kline, 2011; Norman, 2003; Westercamp et al., 2010). Thus, condom use within marriage is still a rare phenomenon e in Kenya, the percentage of men who said they used a condom the last time they had sex with a spouse or cohabiting partner was 3.5% and 7% in 2003 and 2008, respectively, according to the Kenya Demographic and Health Surveys (KCBS and ORC Macro, 2004; KNBS and ICF Macro, 2010). These figures are lower when women's reports are considered. Emphasis on the interpretation of condom use as a contraceptive method however may be less problematic, since the acceptance of family planning has dramatically increased in sub-Saharan Africa, especially in the Southeastern countries (Cleland and Ali, 2006). It is clear that such an interpretation would not be a definitive solution, since condom use would only be resorted to by couples who want to stop having children or by couples that want to space births. Nonetheless, it could facilitate the negotiation of condom use in marital sex, and also in other sexual contexts, such as long-term extramarital relations. Moreover, despite the lower contraceptive efficacy of condoms in comparison with that of hormonal methods ethe annual failure rate being 2% for condoms and 0.3% for the pill or injectable contraceptives (Trussell et al., 2009) e a radical shift from contraceptive pills to condom use in the population would involve minor reproductive costs, in terms of abortion and unwanted births, according to a study of 16 developing countries conducted by Ali et al. (2004). Individual interpretation of condom use is thought to be strongly dependent on the prevailing meaning of condoms in the society in question. This paper explores the extent to which the attitudes of individuals towards condom use in marital relations in rural sub-Saharan settings are shaped by the perceived social acceptance of each of the two uses of condoms in their peer group. Married men and women in rural populations of Kenya are the units of analysis in this study. Kenya can be considered rather representative of the Southeastern African region in terms of several socioeconomic indicators and HIV prevalence rates (World Development Indicators 2008). 2. Social interactions and the attitude towards condom use In this paper, I propose an empirical strategy to examine whether a different socially-shared interpretation of a practice leads to different attitudes towards it. Health-behavior models that assume that behavior depends mainly on individual
knowledge, intentions or conscious control of one's life, such as the Social learning theory (Airhihenbuwa and Obregon, 2000), are unable to explain the puzzle of the disproportionate low levels of condom use among married individuals who are aware of, and worried about, the risk of HIV infection through unprotected marital sex (Anglewicz and Kohler, 2009; Watkins, 2004). An alternative approach needs to take into consideration the role that socially-shared understandings play on individuals' interpretation of actions and on the meaning that such actions give to their sexual relationships. As anthropologists have long pointed out, material objects are loaded with meaning. They may act as environmental stimuli that influence actions by making certain values, behavioral rules, or ideas salient (Bicchieri, 2006; Kay et al., 2004). Bringing out ideas such as infidelity, distrust or infection is likely to have important emotional consequences. As explained above, health campaigns have emphasized a particular aspect of condom use, and this may have determined the way people conceive of it. Nevertheless, some variation in attitudes can be observed among different social environments. Interpersonal communication is a key channel through which individuals become aware of dominant opinions and behaviors (Kohler et al., 2007; Watkins, 2004). With regard to protected sex, informal conversations allow individuals to update their beliefs about the reasons why the people around them use condoms, the characteristics of the contexts in which they do so, and their opinions about this practice. Interpersonal and group communication create the opportunity to re-negotiate the meaning of an action, and lead to a common understanding of the situations in which it takes place (Bicchieri, 2006). The main hypothesis of this paper is (a) that married individuals' view of the use of condoms in marital sex does not improve when they believe these devices to be commonly used to avoid HIV infection from extramarital partners, and (b) that married people become more likely to have a positive attitude when modern contraception is perceived to be widely accepted and, more specifically, when they believe that their peers use condoms for contraceptive purposes. The study focuses on personal attitude as it intends to provide evidence on the extent to which an individual is influenced by the peer group in which he/she is immersed. Given that condom use should not be conceived as a purely individual action, focusing on behavior would involve a different research design that enables us to consider the characteristics of both members of the couple, as well as other aspects that constrain the adoption of a practice. A proper examination of this question would require a couple-based analysis, which is not feasible in this paper due to lack of available data. Nevertheless, as an explorative analysis, the study checks whether analogous results are obtained when reported condom use within marriage is taken as the outcome to be explained. To my knowledge, the influence of the social group on the acceptance of condom use within marriage has not been examined from a quantitative perspective. The exception would be the study by Cordero-Coma and Breen (2012), who expected to observe a positive association between the perceived acceptance of extramarital relationships in the group e or weakness of the fidelity norm- and a married couple's likelihood of using condoms in rural Malawi. They did not find, however, a conclusive result in this regard. More recently, Anglewicz and Clark (2013) have explored the effect of the actual and perceived HIV statuses and the risk perceptions of both the respondent and his/her spouse on individual acceptance of condom use within marriage in rural Malawi. The role of the social group is, nevertheless, absent from their model.
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
3. Data and method I analyze longitudinal data that come from the Kenya Diffusion and Ideational Change Project (KDICP). The characteristics and aims of the survey have been described in Watkins et al. (2003), and the quality of the data has been thoroughly analyzed by Bignami-Van Assche et al. (2003). The data were collected in three waves: 1994, 1996, and 1999. Only the last two waves are analyzed here, because the first wave does not provide enough relevant information. Despite the fact that more recent data would be preferable, this dataset still offers a unique opportunity to examine the influence of social interactions on personal attitudes towards condom use in marriage in a sub-Saharan African context. The sampling frame in these surveys is ever-married women of childbearing age and their husbands (if currently married) in rural sites in the South Nyanza District. Given the aim of this research, the units of analysis are married individuals, both men and women. 3.1. Measures The dependent variable is a proxy measure of the respondent's attitude towards condom use in marital sex. Respondents were asked: “Would you feel comfortable suggesting to your spouse that you and he/she use condoms?”. The possible answers were yes or no. It could be argued that such an indicator measures not only the respondent's attitude towards this practice, but some other characteristics of the respondent and his/her marital relationship. It might be related to male and female skills with regard spousal communication on intimate matters, or to the balance of the bargaining power in the couple. Thus, the model also includes the following control variables: a two-category indicator that measures whether the respondent reported having ever talked to his/her spouse about the chances that his/her spouse or himself/herself might get infected with AIDS, and the dichotomous response to the question: Do you think it is acceptable for a woman to divorce an unfaithful husband?. Table 1 shows descriptive statistics for all the variables in the analysis. With regard to the explanatory variables, the KDICP survey asked with how many people the respondent had ever chatted about family planning (FP), on the one hand, and about AIDS, on the other. Then, the survey asked for information on up to four people with whom the respondent had chatted about each of the two topics. Those respondents who reported having ever had an informal conversation about FP with someone were specifically asked: Does [each of these people or network partners] use modern family planning with her/his spouse?, Which method is it? Condom use is one among various options. Those respondents who have talked about AIDS were asked: What does [each network partner] think is the best way to protect herself/himself from getting AIDS?, and “condom use with extramarital partners” was one of the answers formulated. Note that it is possible for us to observe two communication networks: one about FP and the other about AIDS. However, these networks may overlap, so that reported interlocutors in the discussion on one of the topics could be the same people mentioned as interlocutors in conversations about the other. The responses to the questions are used to construct two indicators: the proportion of people in the network that use condoms in marriage as a contraceptive method, and the proportion of network partners that consider the use of condoms with all extramarital partners as the best strategy to prevent HIV infection, both issues according to the respondent. The continuous measures were categorized to differentiate those respondents who reported a network in which such an opinion or behavior was dominant from the rest. Concerning the first indicator -the proportion of users of condoms for
171
Table 1 Summary statistics for the variables in the models, KDICP 1996 and 1999. Men 1996
Women 1999
N 526 564 26.4 27.8 Comfortable suggesting a CU to spouse b 42.7 (12.8) 43.6 (13.3) Age Level of education Never attended 9.9 8.0 Primary 59.1 57.1 Secondary or more 31.0 34.9 Monogamous 65.4 68.4 Religion Catholic 23.4 21.4 Protestant 64.4 69.7 Others 12.2 8.9 Suspects of infidelity 12.6 8.7 Wants no more children 27.6 31.6 Uses other FP methods 12.6 13.8 Acceptable for a woman to divorce 63.3 77.5 Ever talked to spouse about AIDS 73.4 84.6 Proportion of NP who support CU for HIV prevention Half or less 72.8 79.4 More than half 10.1 11.7 No AIDS network 17.1 8.9 Proportion of NP who use condoms as contraception Less than half 66.9 77.4 Half or more 9.1 5.0 No FP network 24.0 17.6 Proportion of NP who use modern contraception Half or less 45.1 59.1 More than half 30.3 23.4 No FP network 24.6 17.5 Proportion of NP who accept modern contraception Half or less 35.2 27.0 More than half 40.2 55.5 No FP network 24.6 17.5 a b
1996
1999
642 28.5
674 29.7
31.1 (7.9) 31.3 (8.5) 19.8 65.9 14.3 61.8
15.3 70.2 14.5 62.5
22.9 67.3 9.8 33.0 35.2 13.4 60.8 55.9
20.2 73.0 6.8 30.1 38.0 17.1 71.7 73.9
68.7 8.1 23.2
82.7 7.1 10.2
44.6 37.5 17.9
60.5 28.2 11.3
32.2 49.9 17.9
22.5 66.2 11.3
CU and NP stand for condom use and network partners, respectively. Means in italics, and standard deviations in parentheses.
contraception- I distinguish the respondents with networks in which half or more, and not more than half, of the network partners were believed to have this behavior from the rest, because otherwise that category would represent less than 5% of the male sample. A third category in each of the variables mentioned refers to individuals who have not chatted about AIDS or FP with anyone. Information regarding their beliefs about the dominant opinions in their social environment is missing. Given the modest size of the sample, I include them in the analysis as a separate category. Nonetheless, estimates for the subsample obtained after removing respondents with no reported network are also calculated. In the case of the female respondents, the perceived share of users of condoms for contraception in the peer group is so small e about 1% in both waves e that the statistical analysis for married women is not feasible. Nonetheless, two alternative indicators enable us to test whether the dominant attitude towards modern contraception overall in the social group has an impact on the dependent variable. Given that respondents who had chatted about FP were also asked whether they thought that each of the network partners approved of FP, two more measures have been constructed: the proportion of the peer group that uses modern FP and the proportion that approves of FP, both according to the respondent. The abovementioned explanatory variables are used as measures of the dominant attitudes and behaviors in the social network in relation to the two uses of condoms. In fact, even when the two communication networks (FP and AIDS) might not be composed of the same network partners, they are both considered proxies of the respondent's reference group. It is important, then, to be aware of
172
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
the characteristics of the interlocutors reported. Table 2 shows some features of the respondents concerning their networks. The uncensored size of the networks, which refers to the average total number of people with whom the respondent had chatted about each topic, is smaller for women in the two waves, and bigger in 1999 than in 1996 for both men and women. The individual accumulation of conversations over time explains, at least partially, the increase in interlocutors between waves, given that 75% and 72% of the married men and women in the sample in 1996, respectively, were reinterviewed in 1999. As mentioned above, respondents were asked to provide information about only four people at most with whom they had talked about FP, on the one hand, and AIDS, on the other. These limited samples of network partners compose what I call “censored networks”, which have an average size of more than three in all the cases. The KDICP survey does not allow us to know the respondents' criteria for selecting certain interlocutors to give information about, but Table 2 shows that censored networks are mainly composed of individuals of the same sex as the respondent, who are linked to him or her by strong ties (for example, in 1996, 70% of married men had chatted about FP only with men and 80% of married men had a FP communication network in which all the partners were reported to be confidants or friends, as opposed to acquaintances). Finally, about 35% of the respondents in 1999 had an AIDS communication network that overlapped with the FP network. In sum, it seems reasonable to consider that the indicators used in the paper about the proportion of network partners who have a particular behavior or opinion are proxy measures of dominant aspects in the immediate social network or peer group. The models include other factors that are expected to have a relevant influence on the attitude towards condom use within marriage. First of all, the respondent's interest in the outcomes of condom use is taken into account: the models consider whether the respondent (a) suspected that his/her spouse had been unfaithful, as an indicator of the perceived risk of getting infected through non-protected intercourse, and (b) reported his/her will to stop having children. A lack of available data prevents us from considering birth spacing intentions. The current use of other modern contraceptive methods that are more effective than condoms e the
pill, injectable contraceptives, IUD, or sterilization e is also taken into account, because it is expected to hinder the negotiation of condom use for pregnancy avoidance within marriage. Unfortunately, information about perceived affordability of condoms was not available. Respondents' age, level of education, type of marriage (monogamous or polygamous), and religion are also considered in the analysis. The positive influence of education on the acceptance of modern contraception and condoms has been observed in numerous empirical studies (de Walque, 2007; Zellner, 2003). Age is included as a continuous variable. Note that the analysis is restricted to women aged between 15 and 49. Women's capacity to discuss issues related to reproductive and sexual health with their husbands might be different in monogamous and polygamous marriages. Finally, the strong general opposition of the Catholic Church to artificial contraception is expected to induce more negative attitudes among its followers in comparison with other religions (Caldwell, 2000). However, previous studies in subSaharan countries have not provided conclusive evidence in this regard. This may be due to the vast variation in the adherence of local churches to the official position of their denomination on these issues (Freidman, 1995), or to differences in religiosity among individuals (Trinitapoli and Regnerus, 2006). An alternative dichotomous dependent variable that refers to the actual use of condoms in marriage is considered in the last section of the analysis. Respondents were asked the specific contraceptive method they currently used with their spouse. “Condoms” was one of the options, and multiple answers were not allowed. Only 7% and 5% of married men in 1996 and 1999, respectively, reported condoms as the contraceptive method used with their spouses. The corresponding percentages for women are so small (1% and 0.6%) that the statistical analysis for the female sample is not possible. 3.2. Method First, the attitude towards suggesting condom use to a spouse is analyzed cross-sectionally. The data from the two waves have been pooled in order to avoid working with too small sets of cases. This
Table 2 Description of the networks. Men
Family planning network N Uncensored size Proportion with at least one NP N (only those with one NP at least) Censored size Proportion with a censored network in which: All NPs are men All NPs are women All NPs are respondent's confidants or friends AIDS network N Uncensored size Proportion with at least one NP N (only those with one NP at least) Censored size Proportion with a censored network in which:a All NP are men All NP are women All NP are respondent's confidants or friends At least one NP is part of the FP network
Women
1996
1999
1996
1999
526 5.08 (5.91) 0.76 397 3.18 (1.01)
564 6.57 (7.03) 0.82 465 3.39 (0.90)
642 4.48 (4.73) 0.82 527 3.10 (1.01)
674 5.21 (5.24) 0.89 598 3.23 (1.00)
0.70
0.69
0.80
0.86
0.85 0.76
0.87 0.88
526 6.58 (7.89) 0.83 436 3.28 (0.97)
564 9.22 (10.18) 0.91 516 3.59 (0.77)
641 4.87 (5.92) 0.77 493 3.10 (1.05)
674 6.36 (7.05) 0.90 604 3.31 (0.99)
0.74
0.73
0.75 0.39
0.87 0.35
0.57 0.71 0.27
0.60 0.84 0.36
a In 1996, such proportions (in italics) refer to the network partners (NP) mentioned only as part of the AIDS censored network, because we are unable to identify the characteristics of the NP that were mentioned as people with whom the respondent also chatted about FP in this wave.
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
technique seems appropriate since there are no particular reasons to believe that the effects of the explanatory variables substantially vary from the first wave to the successive one. The relaxation of the assumption that errors are uncorrelated is possible by using clustered standard errors. The estimation of the influence of social interactions on individual behavior is challenging since it must deal with a potential self-selection problem. Social networks are rarely randomly distributed because individuals tend to select their interlocutors, usually preferring those with whom they share certain characteristics, attitudes and preferences. A positive correlation in a crosssectional analysis between the respondent's attitude and that of his/her network partners may simply reflect this systematic selection rather than provide evidence of the influence exerted by the social environment. The panel analysis with fixed effects is a suitable tool in this case, given that it allows us to control for timeinvariant observed and unobserved characteristics that may affect both the respondent's attitude and his/her likelihood of interacting with specific people (Behrman et al., 2001). A drawback of the fixed-effects model with a binary dependent variable is that the sample size might decrease notably, because only those individuals who experience a change between waves in the outcome variable are analyzed. This could lead to wrong conclusions, especially if the distribution of the explanatory variables in the subsample changes substantively. Linear probability models are then used as a robustness check, because the estimation of linear models with fixed effects does not imply a reduction in the number of analyzed cases. I adjust for the violations of the classical OLS model assumptions (normality and homoscedasticity of the disturbance term) by using robust standard errors (Behrman et al., 2001). 4. Results 4.1. Attitude towards condom use within marriage Table 3 displays the results of the fixed-effects logit regression of feeling comfortable suggesting condom use to their spouse, for married men and women. Note that only those respondents
173
interviewed in both waves, 1996 and 1999, are examined, and that the time-invariant variables e education and religion e drop out of the models. This is not the case in the cross-sectional analysis, in which all the variables are included (see the results in Table A.1 in the Appendix). Models 1, 2 and 3 (Table 3) differ from each other in the variable that they include to measure the acceptance in the social network of one of the uses of the condom, namely, contraception. Model 1 includes the most suitable measure, considering the aim of this research. According to Model 1, having a peer group in which half or more of the people are identified by the respondent as users of condoms for contraceptive purposes significantly increases the married men's likelihood of reporting a positive attitude towards condom use within marriage. Conversely, peer groups in which more than half of the partners are believed to consider that condom use outside marriage is the best HIV preventive strategy do not make any relevant difference to men's attitude. This result is coherent with the research hypothesis, according to which a perceived socially rooted interpretation of condoms as an HIVpreventive device does not facilitate the acceptance of its use in marriage, whereas a favorable social environment for condoms as a contraceptive method makes individuals more willing to introduce condoms in their marital sexual relations. The longitudinal analysis confirms that the influence of the peer group is not due to selection effects. Unfortunately, Model 1 cannot be estimated for women, as explained in the Measures subsection. We are only able to verify that a perceived widespread interpretation of condoms as suitable HIV preventive devices does not improve their acceptance in marital sex, even when controlling for the social approval of modern contraception. Table 3 shows that having a social network in which most people use modern contraception (Model 2) or approve of it (Model 3) clearly improves married men's attitude. The significance of the rest of the coefficients is similar to those in Model 1. Men are, then, unlikely to feel comfortable negotiating condom use when the use of modern contraception is not a prevailing behavior or is widely disapproved of in their peer group. In contrast, whether men perceive an extensive support of condom use for HIV preventive purposes has no influence on their attitude.
Table 3 Multivariate logit regression analysis with fixed effects of feeling comfortable suggesting CU to a spouse, married men and women in KDICP 1996e1999. Men Model 1
Women Model 2
Monogamous 0.061 (0.552) 0.495 (0.588) Suspects of infidelity 1.112* (0.489) 1.144* (0.525) Wants no more children 0.052 (0.396) 0.109 (0.423) Uses other FP methods 0.577 (0.483) 1.081* (0.512) Acceptable for a woman to divorce 0.563 (0.354) 0.353 (0.370) Ever talked to spouse about AIDS 0.940þ (0.562) 0.747 (0.546) Proportion of NP who support CU for HIV prevention (Ref: Half or less) More than half 0.212 (0.514) 0.05 (0.526) Has no AIDS network 0.548 (0.623) 0.419 (0.619) Proportion of NP who use condoms for contraception (Ref: Less than half) Half or more 1.317* (0.625) Has no FP network 0.937þ (0.541) Proportion of NP who use modern contraception (Ref: Half or less) More than half 1.535** (0.452) Has no FP network (0.710) (0.571) Proportion of NP who accept modern contraception (Ref: Half or less) More than half Has no FP network Year 0.066 (0.077) 0.071 (0.080) Constant N 232 232 þ
p < 0.1; *p < 0.05; **p < 0.01. Standard errors in parentheses.
Model 3
Model 2
Model 3
0.216 (0.574) 0.870þ (0.494) 0.083 (0.414) 0.699 (0.482) 0.437 (0.368) 0.736 (0.554)
0.135 (0.626) 1.320** (0.345) 0.286 (0.356) 0.548 (0.377) 0.055 (0.279) 0.462 (0.347)
0.214 (0.625) 1.360** (0.349) 0.254 (0.358) 0.597 (0.383) 0.074 (0.279) 0.503 (0.353)
0.155 (0.524) 0.295 (0.630)
0.506 (0.493) 0.060 (0.506)
0.440 (0.496) 0.022 (0.502)
0.029 (0.322) 0.020 (0.560) 1.129** (0.368) 0.500 (0.589) 0.146þ (0.081)
0.033 (0.071)
0.326 (0.353) 0.161 (0.591) 0.024 (0.069)
232
288
288
174
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
Table 4 Fixed-effects regression analysis e linear probability model e of actual CU, married men in KDICP 1996e1999. Model 4
Model 5
Model 6
LPM
LPM
LPM
0.041 (0.028) 0.086* (0.039) 0.042 (0.032) 0.050 (0.036) 0.008 (0.025) 0.071** (0.026)
0.049þ (0.027) 0.083* (0.039) 0.039 (0.032) 0.047 (0.036) 0.008 (0.025) 0.076** (0.027)
0.041 (0.042) 0.046 (0.034)
0.037 (0.043) 0.049 (0.034)
Monogamous 0.047þ (0.028) Suspects of infidelity 0.087* (0.040) Wants no more children 0.048 (0.032) Ever used other FP methods 0.028 (0.035) Acceptable for a woman to divorce 0.005 (0.025) Ever talked to spouse about AIDS 0.070** (0.026) Proportion of NP who support CU for HIV prevention (Ref: half or less) More than half 0.030 (0.044) Has no AIDS network 0.044 (0.032) Proportion of NP who use condoms for contraception (Ref: Less than half) Half or more 0.147* (0.060) Has no FP network 0.027 (0.030) Proportion of NP who use modern contraception (Ref: half or less) More than half Has no FP network Proportion of NP who accept modern contraception (Ref: half or less) More than half Has no FP network Year 0.006 (0.005) Constant 11.942 (10.020) N 794
0.088** (0.029) 0.015 (0.028)
0.006 (0.005) 11.953 (9.969) 794
0.049þ (0.027) 0.014 (0.029) 0.002 (0.005) 3.426 (10.766) 794
þ p < 0.1; *p < 0.05; **p < 0.01. Robust standard errors in parentheses.
However, none of the variables about the acceptance of modern contraception in the peer group are statistically significant in the longitudinal analysis of women (Table 3), in spite of the fact that the cross-sectional analysis shows similar results for both sexes (see Table A.1 in the Appendix). Concerning the control variables in the models, some findings from the panel analysis should be highlighted. The suspicion of infidelity has a substantive positive effect on the attitude towards condom use in marriage, particularly for women, whereas the desire to stop having births is not statistically significant. The desire for protection may thus be the main motivation for accepting the use of condoms within marriage, even when it is proposed to the spouse in terms of contraception intentions. This finding lines up well with the results in the study of rural Malawi by Anglewicz and Clark (2013). As a robustness test for all the logit models in Table 3, I also estimated linear probability regressions. The signs and significance of the coefficients barely change. In addition, I replicated the analyses presented in this and in the following subsection on a subsample in which individuals with no reported network were removed. The results are consistent with those shown in Tables 3 and 4. Details are available upon request.
spouse. This measure refers then to a regular, as opposed to an exceptional, use of condoms. It is, however, an imperfect indicator of use, because respondents could only mention one method, even if they used dual protection. Models 4, 5 and 6 in Table 4 are analogous to Models 1, 2 and 3, with a minor change in the specification: the models control for whether the respondent has ever used other contraceptive methods with his spouse, instead of whether he currently uses them. Only linear probability models are presented here, because the standard deviations of the estimated coefficients in the logit models are too high, due to the reduced sample size and the small proportion of positive values in the dependent variable. Table 4 displays the results, which are in line with what has been pointed out above. A married man's probability of using condoms in marital sex is positively influenced by whether he believes that a large share of his peers use condoms as a contraceptive method (Model 4), use modern contraception (Model 5), or accept its use (Model 6). On the contrary, the respondent's belief about the social acceptance of condoms for HIV prevention has no impact on his sexual behavior within marriage.
4.2. Use of condoms within marriage
The study has shown that the attitudes, and very likely behavior as well, of individuals are substantially influenced by their expectations about the dominant attitudes and behaviors in the peer group. Moreover, the empirical analysis has offered some evidence that the perceived social interpretation of condom use modifies the attitude of individuals towards the use of this device in marital sex. Perceiving that modern contraception and, particularly, the use of condoms for this purpose are widely accepted in the peer group makes individuals more likely to report feeling comfortable about suggesting condom use to their spouses. On the contrary, the personal attitude does not improve when perceiving condoms to be commonly understood as a suitable preventive practice against HIV in extramarital sex. Similar conclusions are extracted from an analogous analysis that takes the reported use of condoms within marriage as the dependent variable. Besides, the estimated influence of the peer group is unlikely to be driven by a selection effect.
A proper examination of condom use would be based on a couple-level analysis (Cordero-Coma and Breen, 2012) or, at least, it would take into consideration the characteristics of the two people involved (Anglewicz and Clark, 2013). Although the KDICP enables us to link married respondents to their current spouses, the final number of couples on which there is enough information is limited, which renders such an analysis unfeasible. In this section, I replicate the longitudinal analysis described above, but with a dependent variable that refers to actual condom use within marriage. This should be interpreted as a strategy to obtain additional evidence for the main argument in the paper, rather than as an attempt to provide a compelling explanation of condom use. The outcome variable distinguishes married men who reported condoms as the contraceptive method currently used with their
5. Discussion
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
Such findings support the argument defended here, that an alternative interpretation of condom use which emphasizes its contraceptive function rather than its suitability for HIV protection from untrustworthy sexual partners makes the acceptance of this practice within marriage possible. This was not a trivial question when the data examined here were gathered, and even less so nowadays, since marital sex is the main channel through which HIV is transmitted at the current stage of the epidemic in several sub-Saharan countries. Therefore, a relevant policy implication for HIV prevention campaigns can be derived from the study. The results suggest that the promotion of condom use should be guided by the aim of making such a practice compatible with the norms that regulate not only marriage, but also other steady relationships. In addition, the study provides some indirect empirical support for peer education programs and other activities aimed at groups, because they take into account the relevance of group communication for individuals to update their expectations about what the others know, do and think, which affects, in turn, their opinions and behavior. Replicating these analyses with more recent data would most likely provide similar results, given that norms of trust and fidelity still regulate marriage, and the widespread meaning of condoms is unlikely to have changed. It may be, however, that suggesting the use of condoms for contraceptive purposes is now less convincing as the association of condoms with ‘illicit sex’ becomes entrenched in society and people are more and more concerned about infection from spouses (Watkins, 2004). An important limitation of this study is that the analyses have only been partially done for married women, because very few female respondents reported using condoms with their spouses, on the one hand, and having network partners, most of whom are female friends, who use condoms as a contraceptive method within marriage, on the other. It might be that men tend to talk with their peers more about condoms than about other contraceptive methods, such as the pill, which are not directly under their control, and an analogous reasoning would work for women. In line with this, men might be more likely to report condoms as the contraceptive method employed when the couple uses dual protection and the survey forces respondents to identify just one method. Furthermore, studies of husband-wife discrepancies in the reporting of behavior have highlighted that men may be likely to overreport their use of contraceptive methods in survey interviews for reasons related to how they want to present themselves to such interlocutors (Cordero-Coma and Breen, 2012; Miller et al., 2001). This could explain the higher levels of condom use in marriage according to men, but it is unlikely to explain the respondents'
175
report of their peers' use. Moreover, the motives that lead individuals to misreport their behavior when having a conversation with their peers may not apply to the context of a survey interview. Besides being incomplete, the analysis of women has not offered conclusive evidence, given that the expectations about the peer group have not been found to have a substantive effect on their attitudes in the longitudinal analysis. Women's lesser control over their sexual life may hinder our understanding of their attitude towards introducing something new into their relationship if we do not take into account their husbands' characteristics. Anglewicz and Clark (2013) observed, however, that the reported acceptance of condom use in marriage by the spouse does not affect the attitude of the respondent in rural Malawi. Nonetheless, further research on the role of women's perception of the attitude of their partners, rather than of the reported attitude by their husbands, would contribute to unraveling this question. Our finding that the perceived acceptance of modern contraception in the peer group has no impact on women's attitude would also make sense if the contraceptive function of condom use were especially less salient for women. This lines up well with the fact that women used to be the main target of FP programs, particularly until the UN Conference in Cairo 1994 (McIntosh and Finkle, 1995), which favored the promotion of non-barrier contraceptive methods over that of condoms. Given that individuals tend to talk about reproductive health with people of their same sex, as observed here, the spread of an alternative interpretation of condom use could then be particularly difficult among women. Some other limitations of the study must be mentioned. The panel analysis with fixed effects is unable to deal with potential biases caused by unobserved time-varying characteristics. In addition, individuals that move from the survey village clusters have not been analyzed. The inclusion of these particularly interesting units of analysis would enrich the study, because they are more likely to experience changes in the characteristics of their social networks. The rooted connotations of condom use that link this practice to the idea of ‘illicit sex’ are, unfortunately, not the only obstacle for the use of this device. Individuals tend to complain of the reduction in sexual pleasure that condoms involve (Thomsen et al., 2004). Even fears of side effects and health risks believed to be caused by condom use are quite extended (Siegler et al., 2012). Nonetheless, this research suggests that the diffusion of an alternative interpretation of condom use that emphasizes its contraceptive function would help individuals to cope with the most relevant threats at this stage of the HIV epidemic.
Appendix Table A.1 Multivariate logit regression of feeling comfortable suggesting condom use (CU) to a spouse. Pooled data of married men and women, KDICP 1996 and 1999. Men A.1 Age 0.028** (0.009) Level of education (Ref: Never attended school) Primary 0.538 (0.372) Secondary 0.904* (0.390) Monogamous 0.123 (0.185) Religion (Ref: Catholic) Protestant 0.462* (0.198) Others 0.302 (0.351) Suspects of infidelity 0.967** (0.227) Wants no more children 0.392* (0.200) Uses other FP methods 0.044 (0.198)
Women A.2
A.3
A.2
A.3
0.030** (0.009)
0.027* (0.008)*
0.013 (0.010)
0.013 (0.010)
0.616 (0.361) 0.935* (0.381) 0.176 (0.188)
0.586 (0.377) 0.957* (0.399) 0.172 (0.188)
0.336þ (0.201) 0.540* (0.250) 0.149 (0.142)
0.311 (0.202) 0.508* (0.250) 0.134 (0.142)
0.417* (0.197) 0.194 (0.348) 0.914** (0.229) 0.284 (0.199) 0.244 (0.213)
0.388þ (0.203) 0.238 (0.340) 0.941** (0.222) 0.211 (0.199) 0.193 (0.208)
0.184 (0.172) 0.385 (0.267) 0.783** (0.132) 0.139 (0.153) 0.041 (0.181)
0.193 (0.172) 0.393 (0.271) 0.776** (0.133) 0.112 (0.154) 0.052 (0.179)
þ
(continued on next page)
176
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177
Table A.1 (continued ) Men A.1
Women A.2
Acceptable for a woman to divorce 0.274 (0.170) 0.272 (0.170) Ever talked to spouse about AIDS 0.664** (0.229) 0.603** (0.227) Proportion of NP who support CU for HIV preventiona (Ref: Half or less) More than half 0.249 (0.217) 0.326 (0.213) Has no AIDS network 0.449 (0.337) 0.409 (0.328) Proportion of NP who use condoms for contraception (Ref: Less than half) Half or more 1.622** (0.256) Has no FP network 0.558* (0.251) Proportion of NP who use modern contraception (Ref: Half or less) More than half 0.906** (0.164) Has no FP network 0.460þ (0.254) Proportion of NP who accept modern contraception (Ref: Half or less) More than half Has no FP network Constant 1.731* (0.674) 1.686* (0.662) N 1090 1089
A.3
A.2
A.3
0.212 (0.166) 0.510* (0.225)
0.054 (0.142) 0.553** (0.150)
0.070 (0.142) 0.543** (0.149)
0.273 (0.220) 0.438 (0.324)
0.263 (0.225) 0.156 (0.204)
0.263 (0.226) 0.150 (0.205)
0.338* (0.140) 0.306 (0.226) 1.023** (0.187) 0.142 (0.281) 1.974** (0.680) 1089
1.806** (0.436) 1316
0.535** (0.154) 0.076 (0.244) 2.002** (0.442) 1316
þ p < 0.1; *p < 0.05; **p < 0.01. Clustered standard errors in parentheses. a The abbreviations CU and NP stand for ‘condom use’ and ‘network partners’, respectively.
References Achan, S.F., Akweongo, P., Debpuur, C., Cleland, J., 2009. Coping strategies of young mothers at risk of HIV/AIDS in the Kassena-Nankana District of Northern Ghana. Afr. J. Reprod. Health 13 (1), 61e78. Agadjanian, V., 2005. Gender, religious involvement, and HIV/AIDS prevention in Mozambique. Soc. Sci. Med. 61, 1529e1539. Airhihenbuwa, C.O., Obregon, R., 2000. A critical assessment of theories/models used in Health Communication for HIV/AIDS. J. Health Commun. 5 (Suppl.), 5e15. Akwara, P.A., Madise, N.J., Hinde, A., 2003. Perception of risk of HIV/AIDS and sexual behavior in Kenya. J. Biosoc. Sci. 35, 385e411. Ali, M.M., Cleland, J., Shah, I.H., 2004. Condom use within marriage: a neglected HIV intervention. Bull. World Health Organ. 82, 180e186. Anglewicz, P., Kohler, H.-P., 2009. Overestimating HIV infection: the construction and accuracy of subjective probabilities of HIV infection in rural Malawi. Demogr. Res. 20 (6), 65e96. Anglewicz, P., Clark, S., 2013. The effect of marriage and HIV risks on condom use acceptability in rural Malawi. Soc. Sci. Med. 97, 29e40. Askew, I., Maggwa, N.B., 2002. Integration of STI prevention and management with family planning and antenatal care in sub-Saharan Africa e what more do we need to know? Int. Fam. Plan. Perspect. 28 (2), 77e86. Behrman, J.R., Kohler, H.-P., Watkins, S., 2001. How Can We Measure the Causal Effects of Social Networks Using Observational Data? Evidence from the Diffusion of Family Planning and AIDS Worries in South Nyanza District, Kenya. MPIDR Working Paper 22. Bicchieri, C., 2006. The Grammar of Society. The Nature and Dynamics of Social Norms. Cambridge University Press, New York. Bignami-Van Assche, S., Reniers, G., Weinreib, A.A., 2003. An assessment of the KDICP and the MDICP data quality: interviewer effects, question reliability and sample attrition. Demogr. Res. Special Collection. 1, 31e76. Caldwell, J.C., 2000. Rethinking the African AIDS epidemic. Popul. Dev. Rev. 26 (1), 117e135. Chimbiri, A.M., 2007. The condom is an ‘intruder’ in marriage: evidence from rural Malawi”. Soc. Sci. Med. 64, 1102e1115. Cleland, J., Ali, M.M., 2006. Sexual abstinence, contraception, and condom use by young African women: a secondary analysis of survey data. Lancet 368, 1788e1793. Cordero-Coma, J., Breen, R., 2012. HIV prevention and social desirability: husbandwife discrepancies in reports of condom use. J. Marriage Fam. 74 (3), 601e613. de Walque, D., 2007. How does the impact of an HIV/AIDS information campaign vary with educational attainment? Evidence from rural Uganda. J. Dev. Econ. 84 (2), 686e714. de Walque, D., Kline, R., 2011. Variations in sub-Saharan African Countries. Stud. Fam. Plan. 42 (1), 1e10. Dunkle, K., Stephenson, R., Karita, E., Chomba, E., Kayitenkore, K., Vwalika, C., Greenberg, L., Allen, S., 2008. New heterosexually transmitted HIV in married or cohabiting couples in urban Zambia and Rwanda: an analysis of survey and clinical data. Lancet 371, 2183e2191. Foss, A.M., Hossain, M., Vickerman, P.T., Watss, C.H., 2007. A systematic review of published evidence on intervention impact on condom use in sub-Saharan Africa and Asia. Sex. Transm. Infect. 83, 510e516. Freidman, G.S., 1995. AIDS prevention and the Church. Kenya: mixed messages. AIDS Soc. 6 (2), 4. Gelmon, L., Kenya, P., Oguya, F., Cheluget, B., Haile, G., 2009. Kenya HIV Prevention Response and Modes of Transmission Analysis. National AIDS Control Council, Nairobi, Kenya.
Glynn, J.R., Carael, M., Auvert, B., Kahindo, M., Chege, M., Musonda, R.M., 2001. Why do young women have a much higher prevalence of HIV than young men? A study in Kisumu, Kenya and Ndola, Zambia. AIDS 15 (Suppl. 4), S51eS60. Kay, A.C., Wheeler, C.S., Bargh, J.A., Ross, L., 2004. Material priming: the influence of mundane physical objects on situational construal and competitive behavioral choice. Organ. Behav. Hum. Decis. Process. 95, 83e96. Kelly, R.J., Gray, R., Sewankambo, N., Serwadda, D., Wabwire-Mangen, F., Lutalo, T., Wawer, M.J., 2003. Age differences in sexual partners and risk of HIV-1 infection in rural Uganda. J. Acquir. Immune Defic. Syndr. 32, 446e451. Kenya Central Bureau of Statistics and ORC Macro, 2004. Kenya Demographic and Health Survey 2003. CBS and ORC Macro, Calverton, MD. Kenya National Bureau of Statistics and ICF Macro, 2010. Kenya Demographic and Health Survey 2008-2009. KNBS and ICF Macro, Calverton, MD. Khobotlo, M., Tshehlo, R., Nkonyana, J., Ramoseme, M., Khobotle, M., Chitoshia, A., Hildebrand, M., Fraser, N., 2009. Lesotho: HIV Prevention Response and Modes of Transmission Analysis. National AIDS Commission, Maseru, Lesotho. Kohler, H.P., Behrman, J.R., Watkins, S.C., 2007. Social networks and HIV/AIDS risk perceptions. Demography 44 (1), 1e33. Maharaj, P., 2001. Obstacles to negotiating dual protection: perspectives of men and women. Afr. J. Reprod. Health 5, 150e161. Maharaj, P., Neema, S., Cleland, J., Busza, J., Shah, I., 2012. Condom use within marriage: an assessment of changes in South Africa and Uganda. AIDS Care 24 (4), 444e450. McIntosh, C.A., Finkle, J.L., 1995. The Cairo conference on population and development: a new paradigm. Popul. Dev. Rev. 21 (2), 223e260. Miller, K., Zulu, E.M., Watkins, S.C., 2001. Husband-wife survey responses in Malawi. Stud. Fam. Plan. 32 (2), 161e174. Mngadi, S., Fraser, N., Mkhatshwa, H., Lapidos, T., Khumalo, T., Tsela, S., Nhlabatsi, N., Odido, H., 2009. Swaziland: HIV Prevention Response and Modes of Transmission Analysis. National Emergency Response Council on HIV/AIDS, Mbabane, Swaziland. Muhwava, W., 2004. Condom use within marriage and consensual unions in the era of HIV/AIDS in Zimbabwe. Afr. Popul. Stud. 19 (1), 119e141. Norman, L.R., 2003. Predictors of consistent condom use: a hierarchical analysis of adults from Kenya, Tanzania and Trinidad. Int. J. STD AIDS 14, 584e590. Parsitau, D.S., 2009. Keep holy distance and abstain till he comes: interrogating a Pentecostal Church's engagements with HIV/AIDS and the youth in Kenya. Afr. Today 56, 45e54. Siegler, A.J., Mbwambo, J.K., McCarty, F.A., DiClemente, R.J., 2012. Condoms “contain worms” and “cause HIV” in Tanzania: Negative Condom Beliefs Scale development and implications for HIV prevention. Soc. Sci. Med. 75 (9), 1685e1691. Smith, D.J., 2007. Modern marriage, men’s extramarital sex, and HIV-risk in Southeastern Nigeria. Am. J. Public Health 97, 997e1005. Tavory, I., Swidler, A., 2009. Condom semiotics: meaning and condom use in rural Malawi. Am. Sociol. Rev. 74 (2), 171e189. Thomsen, S.C., Stalker, M., Toroitich-Ruto, C., Ndugga, B.M., Mwarogo, P., 2004. Fifty ways to leave your rubber: how men in Mombasa rationalize unsafe sex. Sex. Transm. Infect. 80, 430e434. Trinitapoli, J., Regnerus, M.D., 2006. Religion and HIV risk behaviors among married men: initial results from a study in rural sub-Saharan Africa. J. Sci. Study Relig. 45 (4), 505e528. Trussell, J., Lalla, A.M., Doan, Q.V., Reyes, E., Pinto, L., Gricar, J., 2009. Cost effectiveness of contraceptives in the United States. Contraception 79, 5e14. Watkins, S.C., 2004. Navigating the AIDS epidemic in rural Malawi. Popul. Dev. Rev. 30 (4), 673e705.
J. Cordero Coma / Social Science & Medicine 116 (2014) 169e177 Watkins, S., Behrman, P., Kohler, H.-P., Zulu, E.M., 2003. Introduction to research on demographic aspects of HIV/AIDS in rural Africa. Demogr. Res. Special Collection. 1 (1), 1e30. Westercamp, N., Matsson, C.L., Madonia, M., Moses, S., Agot, K., Ndinya-Achola, J.O., Otieno, E., Ouma, N., Bailey, R.C., 2010. Determinants of consistent condom use
177
vary by partner type among young men in Kisumu, Kenya: a multi-level data analysis. AIDS Behav. 14 (4), 949e959. ^ te d’Ivoire. Zellner, S.L., 2003. Condom use and the accuracy of AIDS knowledge in Co Int. Fam. Plan. Perspect. 29 (1), 41e47.