The prevalence of HIV infection varies with age and marital status with infection rates being influenced by the age- and marital-status-related frequency of intercourse and, most obviously, the infectivity of an individual’s sexual partner, or partners. Thus HIV/AIDS prevalence is generally observed to be higher among the divorced or separated than among either the single or the currently married and higher yet among widows and widowers. For example, in urban Kenya and Zambia, among both women and men aged 15-19 and 20-24 HIV was more prevalent among the ever-married than the never-married. More specifically, in rural eastern Zimbabwe, HIV sero-prevalence
(adjusting for such factors as age and gender) was 2.7 times higher among married than single people, 5.5 times higher among the divorced and separated, and 7.9 times higher among the widowed. That prevalence was highest among those whose previous spouses had died may reflect nothing more than that the spouse died of AIDS, but the interpretation of the high prevalence of HIV among the separated and divorced is unclear. Is it that their spouse divorced them because of their extramarital affairs, one of which happened to give them HIV? Is it that they divorced their spouse because of their spouse’s extramarital affairs, the source of the HIV infection? Or did they contract their HIV after their previous marriage ended?
In an earlier application of the microsimulation model employed here we simulated the proportions of newly-weds in three regions of rural Malawi who were already HIV-positive. We then made a first attempt at illustrating the implications for ultimate HIV infection of whether couples subsequently maintained a sexually exclusive relationship, or not: in the latter case, both women and men had a propensity to engage in extramarital affairs, and some men had a propensity to have sex with bar girls. We found a large effect of such extramarital sex on the ultimate acquisition of HIV. In the simulations of the southern region, for example, when couples were faithful to one another after marriage, around one-twentieth of women were infected with HIV by the age of 50 (the infection having been introduced into the marriage by sexual behaviour preceding it); in contrast, when both husbands and wives had affairs and half the husbands had a tendency to obtain sex from bar girls, the proportion of women infected by the age of 50 was nearly one-quarter.
We refer to that illustration as a “first attempt” because it incorporates two unrealistic assumptions: that marriages end only through death, and that the surviving spouse does not remarry. In actual fact, however, divorce rates are high in Malawi, as we shall go on to demonstrate, and remarriage is common. Moreover, there are indications that rural Malawians are coming to see divorce as a means to protect themselves against being infected with HIV by an unfaithful spouse.
Our goal in this paper is to assess the impact of time spent in and out of marriage on the lifetime risk of infection with HIV. There are various countervailing forces. Simply put, since intercourse is more frequent within marriage than outside it, an infected married person poses a greater risk to his or her spouse than to a casual extramarital sexual partner. On the other hand, extramarital partnerships and partnerships between marriages expose an unmarried individual to more heterogeneity in the infection status of their partners, thus increasing risk. HIV prevalence may rise in the presence of divorce and remarriage if divorce increases individuals’ average number of sexual partners because they have affairs after divorce and before they remarry, or simply because they do remarry. Conversely, HIV prevalence may fall if divorce removes a significant number of HIV-negative individuals from a marriage with an HIV-positive spouse: in the absence of divorce such individuals would be parted from a spouse only by death, thus increasing the risk that they would become infected themselves.
We investigate the following specific questions:
1. What are the implications for the lifetime probability of infection with HIV of high rates of marital dissolution over the life course?
2. What are the implications for the lifetime probability of infection of periods between marriages?
3. What are the implications for the lifetime probability of infection of trade-offs between fidelity in marriage and divorcing a partner one knows, or suspects, is HIV positive?
We address these questions by means of a microsimulation model. Microsimulation is a stochastic computer-dependent technique for simulating a set of data, one record at a time, according to predetermined probabilistic rules. The individual simulated records — life histories in the present case — are amalgamated and the resulting data set analyzed as though it had been derived in a conventional manner through prospective observation, interview, and continuous monitoring of STD and HIV status. Three characteristics of the method are that the problems to which it is applied depend in some important way on probability; that experimentation is impracticable; and that the creation of a set of realistic formulae is impossible. Each of these characteristics makes the method particularly useful in a variety of demographic applications, including the present one.
The model incorporates behavioural data pertaining to marriage, divorce and remarriage, and to premarital and extramarital affairs and sexual encounters, the data sources and the data themselves being described in the following section. The model also incorporates biological data pertaining to the aetiology of HIV and other STDs and in particular their transmission probabilities and durations of infectiousness.
For the present paper we produce models only for the southern region of Malawi. We plan subsequently to construct separate models for each of the three regions surveyed because their distinctive lineage and kinship systems influence marriage. Rumphi District, in the north, follows a patrilineal system of kinship and lineage: residence is ideally patrilocal, inheritance is traced through sons, and bridewealth is required. These conditions imply later marriage, which is indeed demonstrable in our data. Balaka District, in the south, follows the matrilineal system, ideally with matrilocal residence and with token gifts at marriage, although both patrilocal and matrilocal residence occurs; this is the region where marriage is earliest. Mchinji District, located in the central region, follows a less rigid matrilineal system whereby residence may be matrilocal or patrilocal, and, again, only token gifts at marriage. In addition, the three districts differ according to the predominant ethnicity, religion and language; the language difference, in particular, suggests that the attitudes and behaviours in the three communities may also be distinct.
We first trawl through quantitative and qualitative data collected under the aegis of the Malawi Diffusion and Ideational Change Project to gain insights into the local meaning of marriage and divorce, the causes of divorce, and the levels and patterns of divorce and remarriage. We use these data also in an attempt to assess the extent to which women and men are sexually active with people to whom they are not married. After briefly describing the microsimulation model we exploit the experimental character of the microsimulation method by applying it to examine the effects on women’s lifetime acquisition of HIV of observed patterns of marriage and divorce, and of various strategies to avoid being infected by a spouse.
