African Population Studies Union for African Population Studies ISSN: 0850-5780 Vol. 19, Num. 1, 2004, pp. 143-164

African Population Studies/Etude de la Population Africaine, Vol. 19, No. 1, April 2004, pp. 143-164

Ethnicity and Marriage Patterns in Mozambique

Carlos Arnaldo¹

Centro de Estudos de População, Universidade Eduardo Mondlane, Caixa Postal 257, Maputo -MOZAMBIQUE

¹ The author would like to thank the Mozambican National Institute of Statistics for providing the data used in this paper, and to Professor Jack Caldwell, Dr David Lucas Dr Heather Booth, and Ms Zitha Mokomane, all of the Australian National University, for their useful comments on earlier versions of the paper.

Code Number: ep04008

ABSTRACT

Marriage in sub-Saharan Africa has been commonly described as early and universal. Yet, its patterns vary across countries and even within countries among different cultural groups. Mozambique is culturally heterogeneous with both patrilineal and matrilineal systems of social organisation represented. Using data from the 1997 Census and 1997 MDHS, this article examines differentials in marriage patterns among five ethnic groups in Mozambique: Tsonga, Sena/Ndau, Lomwe/Chuwabo, Macua and Others, a residual group. Multivariate analyses are used to explore whether the reported ethnic differentials in age at first marriage, polygyny and marital dissolution can be attributed to ethnicity or to other characteristics that distinguish the ethnic groups. The findings are consistent with culture-associated differentials. After controlling for the socio-economic and demographic characteristics of women, the differences remain, with the matrilineal ethnic groups (Macua and Lomwe/Chuwabo) having an earlier age at marriage, lower prevalence of polygyny and higher marital dissolution than the patrilineal ethnic groups (Tsonga and Sena/Ndau).

Marriage in sub-Saharan Africa has been commonly described as early and universal (van de Walle, 1968; Lesthaeghe, 1989) and this situation has partly been blamed for the persistence of high fertility in the region (e.g. Gould and Brown, 1996). However, the region is far from homogenous. Marriage patterns vary across and within countries among different ethnic groups (Lesthaeghe and Eelens, 1989; Lesthaeghe, Kaufmann and Meekers, 1989). Such variations could be due to both cultural and socio-economic factors. This article aims to assess the effect of ‘culture’ (ethnicity) on marriage patterns (age at marriage, polygyny and marital dissolution) in Mozambique. The study seeks to test the hypothesis that ethnic differences in marriage patterns are a reflection of socio-economic differences among the individual members of different ethnic groups rather than ethnicity per se.

While the relative effect of ethnicity on demographic behaviour has been assessed in several countries, including some in sub-Saharan Africa (e.g. Kollehlon, 1989; Benefo, Tsui and Johnson, 1994; Addai, 1999a, b; Addai and Trovato, 1999; Zulu, 2001), in Mozambique such studies are still lacking, although the country has an ethnically heterogeneous population.

THEORETICAL BACKGROUND

Theoretically, group differences in demographic behaviour (especially fertility) have been interpreted in terms of two hypotheses: the ‘character-istics hypothesis’ and the ‘norms/cultural hypothesis’ (Goldscheider, 1971; Addai, 1999a; Addai and Trovato, 1999). Although these hypotheses have mainly been used to explain ethnic differences in fertility (Kollehlon, 1989; Addai and Trovato, 1999), contraceptive use (Addai, 1999a) and the duration of postpartum sexual abstinence (Zulu, 2001), they can also be useful in explaining ethnic differences in marriage patterns.

The characteristics hypothesis assigns ethnic differences in demographic behaviour (marriage patterns in this case) to differing socio-economic characteristics of individual members of different ethnic groups. Inherent in this hypothesis is the assumption that people of different ethnic background become assimilated into the socio-economic structure of the larger society (within the nation) as their socio-economic status improves through increased education, urban residence and modern occupations (Addai and Trovato, 1999). Ethnic affiliation can, therefore, be treated as an indicator of socio-economic status, degree of urbanisation and social mobility of the groups in a particular society (Goldscheider, 1971:273). Marriage pattern differences among ethnic groups that may exist at one point in time are considered temporary phenomena, representing a social or cultural lag, which is often indicative of social and economic backwardness or deprivation (Goldscheider, 1971). Thus, irrespective of their ethnic background, women who possess the same socio-economic characteristics are expected to have similar marriage patterns. Therefore, once differences in socio-economic characteristics are controlled for, ethnic variations in age at marriage, and the propensity to enter polygynous marriages and to experience a marital dissolution should disappear.

An alternative explanation of ethnic differences in marriage patterns is based on the ‘norms/cultural hypothesis’. This hypothesis assigns similar importance to the role of ethnicity/culture in explaining ethnic differences in marriage patterns (Addai, 1999a; Addai and Trovato, 1999). Group norms, ideals and beliefs are assumed to be important factors in determining age at marriage, propensity to enter polygyny and experience of marital dissolution.

One of the important ethnic differences is that between matrilineal and patrilineal systems of descent. Since late marriage is associated with greater independence for women, which is more likely to be found in a matrilineal system than in a patrilineal one (Kaufman and Meekers, 1998), age at marriage is expected to be later among matrilineal ethnic groups than among the patrilineal ones (McDonald, 1985; Lesthaeghe et al., 1989; Kaufman and Meekers, 1998). Matrilineal societies are also expected to have lower prevalence of polygyny than the patrilineal societies because the matrilocal or uxorilocal residence of spouses that characterises the former societies makes it difficult for a man to bring a new wife home to his first wife’s kin unless the two women are related (Lesthaeghe et al., 1989; Lesthaeghe et al., 1994; Kaufman and Meekers, 1998). Divorce is believed to be lower among patrilineal societies than in matrilineal ones (Burch, 1983; Gluckman, 1950). A patrilineal marriage involves permanent incorporation of a woman in her husband’s lineage and complete transfer of her reproductive power to her husband’s family through payment of bridewealth. Under such circumstances, divorce is more difficult because the bride-wealth would have to be returned and the woman would be separated from her children. In contrast, in a matrilineal marriage, the rights of the woman’s reproductive abilities remain with her lineage so that the duration of the marriage depends on the couple’s individual attributes (Brandon, 1990:151).

THE STUDY CONTEXT

Mozambique’s population of 16 million in 1997 is ethnically diverse. Over twenty different cultural sub-groups of both matrilineal and patrilineal systems of descent can be identified in the country. Using the information on first language, six major ethnic groups were identified: the Tsonga, Sena/Ndau, Lomwe/Chuwabo, Macua, ‘Other’ (a residual of the remaining local languages) and Portuguese/foreigner. The socio-economic character-istics of women belonging to each of these groups are summarised in Table 1. The Portuguese/foreigner is an elite group comprising Mozambicans whose first language is Portuguese and Asian and European descendants and accounts for only 6% of the total population but still comprise about one million people. They are the most educated and urbanised group. Among the local ethnic groups, the Tsonga (who live in the Southern Region) are the most educated and urbanised, followed by the Sena/Ndau of Central Region. The Macua are the least educated and the Lomwe/Chuwabo the least urbanised. There are also ethnic differences in religious affiliation.

Source: 1997 Census

While the Macua are predominately Muslim, the ‘Other’ are mainly Christians. In terms of descent system, the Macua and the Lomwe/Chuwabo are matrilineal whereas the Tsonga and the Sena/Ndau are patrilineal. Finally, the ‘Other’ category is a residual including all the remaining local ethnic groups, some of which are patrilineal and others matrilineal.

Marriage is a difficult concept in African societies because it is a process rather than a discrete event and involves rituals, negotiations and transactions that can stretch over years, making it difficult to say at what point a couple becomes married. There are also more than one type of marriage. In Mozambique, at least four main types of marital union can be identified: Customary/traditional marriage, religious marriage, civil marriage and mutual consent union/cohabitation.

Customary marriage: This is a traditional Mozambican form of marriage and is usually carried out with the consent of both families. It is a process involving a series of stages that can go on for months and even years. The way the process is conducted varies according to ethnic groups and descent system. Among the patrilineal groups (the Tsonga and Sena/Ndau) the process is centred on the payment of bridewealth from the boy’s family to the girl’s family, while among the matrilineal groups (the Lomwe/Chuwabo and Macua) such payments are not required. Under customary marriage, polygyny is permitted without any limit to the number of wives. In fact, among patrilineal ethnic groups polygyny is recommended when the first wife is infertile.

Religious marriage: This is a marriage celebrated in a church or mosque with a ceremony to bless the union of a couple where one or both are members of that religion. Christian marriages are monogamous and among Catholics, who make a large proportion of all Mozambique’s Christians, indissoluble, whereas polygyny up to a maximum of four wives per man and divorce are permitted under the Islamic law on certain specified conditions (Mondal, 1997). This (religious marriage) type of marriage is more expensive than the customary marriage since, apart from bride-wealth payments (patrilineal ethnic groups), wedding rings and modern clothes are also required. The Church or mosque keeps the record of the marriage and a marriage certificate is issued to the couple.

Civil marriage: This is a registered union of a never married/divorced/widowed woman with a never married/divorced /widowed man under the civil law. It is monogamous and indissoluble except by a decree of divorce issued by a court of law. Under Mozambican civil law the minimum age of marriage is 14 years for females and 16 years for males (República de Moçambique, 2001)². Although this type of marriage is monogamous, in practice some men are married by civil law to their first wives and live with one or more additional women. Because civil marriage is expensive, it is more prevalent among urban and more educated segments of the population. It should also be noted that, except for those of Asian and European backgrounds, civil marriage only takes place after traditional requirements (e.g. bride-wealth has been paid, in the case of patrilineal ethnic groups) have been met. Thus, in practice, couples who have had a civil ceremony also had customary or religious (if they are affiliated to a religion) ceremonies at the same time.

Mutual consent union/Cohabitation: this is a relationship where persons of the opposite sex cohabit without going through the formalities of customary, religious or civil marriage. Although a mutual consent union could be a temporary relationship lasting for a few weeks or months, it may be a more permanent tie where the pair lives as husband and wife. In some cases, cohabitation or trial marriage constitutes a stage before the completion of the marriage process, that is, before the social sanctioning of the union. For example, it is a tradition among the Macua that the couple have a ‘period of proof’ of about one year, when they behave as husband and wife, before the marriage process is complete (Martinez, 1989). In contrast, among the Tsonga, for example, the cohabitation usually takes place only after the bride-wealth has been paid in full, except in the case of premarital pregnancy, where cohabitation may precede the completion of bridewealth payment.

Despite the existence of these types of union, neither the 1997 census nor the 1997 DHS (the data sources used in this study) distinguished customary marriage from cohabitation, and religious marriage from civil marriage. Customary marriage and cohabitation were treated as a single category, labelled ‘marital union/living together’, and religious and civil marriage as the ‘married’ category. Thus, in this analysis, women in both ‘marital union/living together’ and ‘married’ categories are treated as ‘married’. The non-distinction between the four types of marriage, especially that between customary marriage and cohabitation, may affect the results. Women of ethnic groups which allow cohabitation before the completion of the marriage process (e.g. Macua) are more likely to have been recorded as being in a marital union than women of ethnic groups where cohabitation is not allowed before the completion of the marriage process (e.g. the Tsonga). This would lead to an overestimation of the proportion currently married and an underestimation of average age at marriage among the former groups and an underestimation of the proportion currently married and an overestimation of average age at marriage among the latter groups.

The Data

The data analysed in this article were taken from Mozambique’s 1997 Population and Housing Census and the 1997 Demographic and Health Survey. The 1997 Census was the second population census to be conducted in the country after independence from Portuguese rule in 1975. The census collected information on current marital status that is useful in the analysis of marriage pattern.

The Mozambique Demographic and Health Survey (MDHS) was carried out between March and July 1997. Its main objective was to provide reliable and current information on fertility and reproductive behaviour, maternal and child health, infant and child mortality, contraceptive knowledge and use, and knowledge and attitudes regarding HIV/AIDS (Gaspar et al., 1998). The 1997 MDHS is a representative, stratified probability sample of women aged 15-49. Households were identified using a two-stage cluster sampling procedure based on census enumeration areas. Three hundred and eighty-eight enumeration areas were selected and 100 to 200 households were selected within each enumeration area. Ninety-eight per cent of all eligible women were successfully interviewed, resulting in a final sample of 8,779 women (Gaspar et al., 1998).

Methods of Analysis

Multivariate analysis techniques are used to examine three aspects of the marriage patterns: age at marriage, polygyny and marital dissolution. The analysis of age at marriage is based on both census and DHS data, whereas those of polygyny and marital dissolution are based only on the DHS data. A complete list of such variables and the methods used for each analysis are presented in Table 2.

Since age at marriage was not collected in the 1997 census, a dichotomous variable indicating whether or not a woman aged 15-19 has ever been married was constructed from the information on current marital status. A random sample of about 15,000 women was drawn from the data set of women aged 15-19 and a logistic regression model (Menard, 1995) was fitted to determine whether ethnic differences in the propensity of ever having been married among women aged 15-19 remained sizeable and statistically significant once other socio-economic factors were held constant. For the DHS data, the dependent variable is the reported age at marriage. Because the age at marriage of women who were never married at the date of the interview, but who will eventually marry some time in the future (censored cases), is unknown, proportional hazard models were used to analyse the data. The proportional hazard model is a life table technique which is suitable for the analysis of censored data and has widely been used in demographic analysis (e.g. Trussell and Bloom, 1983; Halli and Rao, 1992; Le, 1997; Bracher and Santow, 1998; Heaton and Forste, 1998; Hosmer Jr. and Lemeshow, 1999; Kinfu Ashagrea, 2001).

For the analysis of polygyny, the dependent variable is a dichotomous variable indicating whether or not a currently married woman is in a polygynous marriage.

Finally, marital dissolution is measured by a dichotomous variable indicating whether or not a woman had been married more than once at the time of the survey. Because having married more than once does not only measure the frequency of divorce but also the frequency of widowhood, which is related to the level of mortality, the proportion of children ever born who have died is included in the model to allow for differential mortality. Ideally, data on adult mortality are needed to give an indication of the likelihood of widowhood. Since such data are not available, the proportion of children dead has been used as a mortality indicator, as there is normally a fairly close correlation between childhood and adult mortality levels.

The independent variable for the study is ethnicity categorised into Tsonga, Sena/Ndau, Lomwe/Chuwabo, Macua and ‘Other’. Women of Portuguese/foreigner ethnic background were excluded from the analysis because of their diverse composition. To allow for women’s socio-economic differences, many control variables, selected for explicit theoretical reasons and availability in the data sets, were included in the analysis.

RESULTS

Age at Marriage

Table 3 presents the percentage of women never married by selected age groups, the singulate mean age at marriage (SMAM), a measure of age at first marriage computed from the proportions never married by age (Hajnal, 1953), and the median age at first marriage by ethnicity. These measures show that there are substantial ethnic differences in the timing of marriage. The Tsonga have the latest age at marriage, followed by the Sena/Ndau, whereas the Macua have the youngest. It is interesting to see that women of patrilineal descent (Tsonga and Sena/Ndau) marry later than their counterparts of matrilineal descent (Lomwe/Chuwabo and Macua), although the difference between Sena/Ndau and Lomwe/Chuwabo is small. For instance, 73% of Tsonga and 50% of Sena/Ndau women aged 15-19 have never been married, compared with only 48% and 32% of Lomwe/Chuwabo and Macua, respectively.

Table 3: Percentage of Women Who Are Never Married, SMAM and Median Age at First Marriage by Ethnic Affiliation, Mozambique 1997

Source: a= Computed from 1997 Census; b= Computed from the 1997 MDHS

To assess the independent effect of ethnicity on age at first marriage, socio-economic characteristics were controlled for. The results are presented in Table 4. The first two columns are from the 1997 census and indicate the odds ratios that a woman of each ethnic group aged 15-19 has ever been in a union compared with a woman in the reference category (Tsonga). The last two columns are from the 1997 DHS and present the risk of first marriage for each ethnic group relative to the risk for the reference category. The odds ratio and the relative risk for the reference category are equal to unity, and, thus, a value of 1.0 for any of the other categories means that the ethnic group in question has the same first marriage timing as the reference category. A value higher than one indicates greater likelihood/risk of marriage than the reference category, whereas values less than one indicates a lower likelihood/risk of marriage. Models I and III show the gross effects (without controls) and Models II and IV the net effect (with controls).

The results show that the introduction of socio-economic controls did not eliminate ethnic differences in age at marriage. Lomwe/Chuwabo and Macua ethnic backgrounds are significantly associated with younger age at marriage than the Tsonga. Model II shows that the odds that a Lomwe/Chuwabo or a Macua woman aged 15-19 has ever been married is 1.46 times and 2.57 times respectively that of her Tsonga counterpart, while there is no significant difference between Tsonga and Sena/Ndau women in the odds of first marriage. The same, but less striking, pattern of relationships is observed from the DHS analysis (Models III and IV): the risks of first marriage for a Lomwe/Chuwabo and a Macua woman are 36 per cent and 13 per cent respectively higher than for a woman in the reference group (Tsonga).

Regarding the lineage system, the results show that the patrilineal ethnic groups (Tsonga and Sena/Ndau) have a later marriage pattern than the matrilineal ethnic groups (Lomwe/Chuwabo and Macua). This is in contrast with the literature (McDonald, 1985; Lesthaeghe et al., 1994; Kaufman and Meekers, 1998) which suggests later marriage among women in matrilineal than patrilineal societies. However, the early marriage pattern among the Macua is consistent with ethnographic studies (e.g. Conceição, 1960; Lopes, 1960). Conceição (1960:12) reported that Macua girls could marry between age 8 and 10, while, according to Lopes (1960), engagement could happen even earlier, at age 6. However, because Macua marriage is a process rather than a discrete event, these ages may not refer to the beginning of cohabitation but to the beginning of the process leading to it. On the other hand, later marriage among the Tsonga may reflect high male migration to South Africa or Maputo City (McDonald et al., 2000; Arnaldo, 2001), perhaps to gather resources for marriage, which may raise age at marriage for women by creating a shortage of males of marriageable age.

In relation to the other socio-economic factors included in Models II and IV, education and paid employment increase the age at marriage in comparison with no education and non-employment, respectively; while there is no agreement between the results from the census (Model II) and those from the DHS (Model IV) in relation to the effect of self-employment on age at marriage. The census shows a positive effect and the DHS a negative one. The reason for this inconsistency is unclear. But, it is important to note that these results may be misleading because employment status at the time of the census or survey, as used in this analysis, may not be the same as that before marriage.

Model II shows that not having a religion increases the likelihood of first marriage by 30 per cent in comparison to being Catholic, while urban residence and residence in the Southern Region are associated with later first marriage than rural and Northern or Central residences respectively.

The practice of polygyny is fairly common in Mozambique. The DHS found that 28 per cent of currently married women were in polygynous unions. However, there are significant differences among ethnic groups. The highest level of polygyny is observed among the Sena/Ndau (40%) followed by the Tsonga (28%). For the remaining groups the prevalence of polygyny is 22% for the Macua, 20% for the ‘Other’ ethnic group and 15% for the Lomwe/Chuwabo.

The results of the multivariate analysis performed to assess the independent effect of ethnicity on polygyny are presented in Table 5. The table shows odds ratios that a currently married woman of each ethnic group is in a polygynous marriage compared with her counterpart in the reference category (Tsonga) without controls (Model I) and with controls (Model II). Model I shows that there are significant differences in the propensity to be in a polygynous marriage by ethnic affiliation. Sena/Ndau ethnicity increases the likelihood of being in a polygynous marriage by 35 per cent relative to the reference group, whereas being a Lomwe/Chuwabo and a Macua reduces it by 64 and 32 per cent respectively. These differences do not disappear when socio-economic controls are introduced (Model II). In fact, the odds of being polygynous decrease for the Lomwe/Chuwabo and Macua, whereas those for Sena/Ndau decrease and become statistically not different from the Tsonga at the 5% level. In both Model I and II ‘Other’ ethnic background has a negative effect on polygyny but it is only significant at the 5% level in Model II. These findings are consistent with the tendency of patrilineal descent systems to be more polygynous than matrilineal systems. A matrilineal Lomwe/Chuwabo or Macua woman is 74% and 60%, respectively, less likely to marry a polygynist than a woman of the Tsonga ethnic group. On the other hand, there are no significant differences between a patrilineal Tsonga and a patrilineal Sena/Ndau woman in the likelihood of marrying a polygynist. In Ghana, Timæus and Reynar (1998) also found that polygyny was slightly higher among patrilineal populations than matrilineal ones. In a multivariate analysis by Lesthaeghe et al. (1989:314), matrilinearity produced on average a 13-percentage-point diminution in polygyny when compared to patrilinearity, which is consistent with the theoretical proposition that matrilineal systems are less amenable to polygynous household formation because of their tendency towards matrilocal or uxorilocal residence (Lesthaeghe et al., 1989:314).

Table 5: Odds Ratios of Being in a Polygynous Union by Ethnic Background, Married Women 15-49, Mozambique 1997

Note: ® = Reference Category; *p≤0.05; **p≤0.01; ***p≤0.001.
Source: 1997 DHS

Table 5 also shows that women’s secondary education and non-employment are associated with a low likelihood of being in polygynous marriage. Women whose husbands have white-collar occupations are more likely to be polygynous than those whose husbands have other occupations, suggesting, perhaps, that men’s socio-economic status is an important factor in determining their financial ability to be polygynous. Non-Catholic women have higher propensity to enter polygyny than the Catholics, partly reflecting a strong stand against polygyny by the Catholic Church. Rural residence and long marital duration are also associated with a higher likelihood of polygyny.

Marital Dissolution

Table 6 presents two measures of marital dissolution derived from the 1997 census and the 1997 DHS: the proportion of ever married women aged 15-49 who are currently divorced and the proportion of women who had married more than once at the time of the survey. Both measures should be interpreted with caution. The first, which represents a ‘snapshot’ of the proportion currently divorced, is influenced by both marital dissolution and subsequent remarriage. Thus, given that the interval to remarriage after divorce is often culturally determined and so may vary considerably between different groups, the fact that one group has a higher proportion currently divorced than another does not necessarily mean that it has higher divorce rate than the other group. Also, the proportion of women who have been married twice or more may overestimate the level of marital dissolution since two different factors could have caused multiple marriages: remarriage after divorce and remarriage after widowhood. Nevertheless, the census and DHS figures in Table 6 show higher marital instability among the matrilineal ethnic groups (Macua and Lomwe/Chuwabo) than among the patrilineal (Tsonga and Sena/Ndau) and ‘Other’ ethnic groups. For example, more than half of ever married Macua and 27 per cent of Lomwe/Chuwabo women have been married twice or more as compared with less than 21 per cent for women in the remaining ethnic groups.

Table 6: Measures of Marital Dissolution by Ethnicity, Mozambique 1997

Source: Computed from 1997 Census^a and ^bDHS

The independent effect of ethnicity on marital instability was investigated through multivariate analysis, and the results are presented in Table 7. As in the previous analysis, Model I shows the gross effect and Model II the net effect. Model I shows that only being of Sena/Ndau and Macua ethnic backgrounds is significantly associated with higher marital instability than the Tsonga while there is no significant difference between the remaining ethnic groups (Lomwe/Chuwabo and ‘Other’) and the Tsonga. When socio-economic controls are introduced (Model II), the effect of Macua background is reduced but is still significant at the 5% level, and the effect of Lomwe/Chuwabo background increases but is still not significant at the 5% level. The effects of Sena/Ndau and ‘Other’ also decreased and lost their statistical significance at the 5% level. The high instability of Macua marriage has been reported by several ethnographic studies (e.g. Conceição, 1960; Lopes, 1960; Martinez, 1989). According to Conceição (1960), a smallest quarrel was enough reason for a Macua man or woman to divorce his/her partner. For instance, ‘being tired’ was a plausible reason to dissolve a Macua marriage (Lopes, 1960). Although religious adherence is high among the Macua, they are mostly Muslims, who as discussed earlier have high marital instability. Furthermore, the Macua also seem to pay more allegiance to culture than to religion.

Regarding the other socio-economic factors in the Model II, education, employment status, place of residence and number of living children do not have a significant effect on marital dissolution. In contrast, religion, radio ownership, age at marriage, type of marriage, the proportion of children ever born who have died and region are important determinants of marital dissolution. Women who profess a religion other than Catholicism are more likely than the latter to have been married more than once; age at marriage is negatively associated with marital instability. Senior wives in polygynous marriages are less likely than monogamous women to have been married more than once, while, in contrast, junior wives are more likely than senior wives to have experienced a marital dissolution, indicating, perhaps, a strong social feeling against a man’s first marriage being with a divorced woman. This makes divorced women to commonly marry polygynously rather than to an unmarried man. The proportion of children ever born that died has a very strong positive effect on the likelihood that a woman has had multiple marriages. This implies that some marriages were dissolved because the wife was blamed for giving birth to children who died. Marriage is more than three times as unstable in the Northern Region and 10 per cent more unstable in Central Region than in Southern Region, but only the odds ratio for the Northern Region is statistically significant.

Table 7: Odds Ratios of Having Married more than Once by Ethnicity, Women 15-49, Mozambique 1997

Note: ® = Reference Category; *p≤0.05; **p≤0.01; ***p≤0.001.
Source: 1997 DHS

DISCUSSION AND CONCLUSION

This article examined two hypotheses on ethnic differentials in age at marriage, polygyny and marital dissolution. One hypothesis (‘the characteristics hypothesis’) postulates that ethnic differences in marriage patterns are an artefact of socio-economic differences between women of different ethnic backgrounds and that once these differences are accounted for (through statistical controls) group variations in marriage patterns should disappear. The second hypothesis (‘norms/cultural hypothesis’), although acknowledging the effect of socio-economic differences, argues that there are also ethnic differences in marriage patterns in terms of cultural values, norms and practices.

The findings from the study appear to support the ‘norms/cultural hypothesis’. The empirical evidence in Tables 4, 5 and 7 shows that socio-economic controls could not account for all the ethnic differences in timing of marriage, prevalence of polygyny and marital stability. Regarding age at marriage, the findings contradict not only the ‘characteristic hypothesis’ but also the expectation that marriage is earlier among patrilineal than matrilineal descent systems. This study suggests the reverse. The fact that marriage is later among the patrilineal (Tsonga and Sena/Ndau) than the matrilineal (Lomwe/Chuwabo and Macua) ethnic groups, reflects a greater importance given to the bride-wealth payment in the marriage process of patrilineal groups. Since the value of bride-wealth is continuously rising, men in these societies may need more time to gather the necessary resources to meet the costs of marriage than their counterparts in matrilineal ethnic groups where bride-wealth is not required (see e.g. Isiugo-Abanihe, 1994, on the Ibo of Nigeria; Dyson-Hudson and Meekers, 1996, on the Turkana of Kenya). Although this is more likely to affect males than females, girls who do not want to marry a polygynist may be forced to wait for a long time to find a man with enough resources (or for their boyfriends to gather the resources) to meet the costs of marriage. In addition, male migration may also play a role. The Tsonga, who have the latest marriage pattern, are characterised by high male migration to South Africa or to Maputo City (McDonald et al., 2000; Arnaldo, 2001), which may raise female age at marriage by creating a shortage of males of marriageable age.

An early age at marriage for women is usually accompanied by a large gap between female and male ages at first marriage (Lesthaeghe, 1989). Assuming inter-ethnic marriages are rare, the figures presented in Table 8 appear to support this proposition with the age gap at marriage being widest among the early marrying Macua and smallest among the Tsonga. Table 8 also shows that the ratio of males to females of marriageable age is lowest among the Tsonga indicating that the high labour migration that characterises this group leads to shortage of males of marriageable age that may be related to late marriage of females.

Table 8: Singulate Mean Age at Marriage by Sex and Male/Female Ratio by Ethnicity, Mozambique 1997

Source: 1997 Census.

There is, however, a possible ethnic differential bias on current marital status and age at marriage caused by the failure to distinguish cohabitation from customary marriage. Since women of matrilineal descent systems (especially the Macua) cohabit with their partners before the marriage process has been completed, they would be more likely to be recorded as being in union than women of patrilineal descent (Tsonga in particular) where cohabitation usually occurs only after the process has been completed. If this has happened, the Tsonga proportion of women 15-19 currently never married and the odd ratios from the multivariate analysis (Table 4, Models I and II) would be biased upward. However, the consistency between census and DHS results, where more probing was made and living together relationships were recorded and considered as marriages, suggests that such bias did not account for all ethnic differences that emerged from the analysis.

The results with regard to polygyny confirmed the higher propensity of women in patrilineal ethnic groups to enter polygyny than their counterparts in the matrilineal groups. Among the Tsonga and the Sena/Ndau, polygyny is recommended as the best solution for the infertility of the first wife and in such a situation the second wife can even be a younger sister or other relative of the first wife. It is also important to note that, although the matrilineal ethnic groups have a low prevalence of polygyny, there are still considerable differences between them. This analysis has shown that Macua women are 48 per cent more likely to marry a polygynist than their counterparts of Lomwe/Chuwabo ethnic background. This relatively high prevalence of polygyny among the Macua is partly related to the Muslim influence they received from Arab Muslim traders starting from the eighth century (Martinez, 1989; Daniel, 1995). Because they have matrilocal residence, polygynous Macua men have to reside at their senior wives’ family and visit the other wives elsewhere in turns.

In terms of marital dissolution, the findings are consistent with expectations. Matrilineal ethnic groups have a higher probability of experiencing marital dissolution than patrilineal groups, which is consistent with the argument that the absence (or low value) of bride-wealth payments in matrilineal marriage makes divorce easier, while high bridewealth payments and the fact that children belong to the husband’s lineage among the patrilineal ethnic groups make divorce difficult in these societies. Thus, since divorce implies the return of the bridewealth and the separation of the mother from her children, women in patrilineal societies (Tsonga and Sena/Ndau) may be more hesitant to leave unsatisfactory marriages than their counterparts in matrilineal societies. In contrast, in the Macua and Lomwe/Chuwabo marriage, there is no transfer of women’s productive and reproductive powers and the children belong to the women’s lineage so that divorce is easy and does not involve any kind of wealth transfer, since bridewealth payment is not required. The fact that women in matrilineal ethnic groups marry at a younger age than those in patrilineal groups, increases their risk of divorce, mainly due to immaturity and the short duration of search for a suitable partner resulting in poor information about the partner’s characteristics (see e.g. Lehrer and Chiswick, 1993; Tilson and Larsen, 2000). This analysis (results not shown) has shown that marrying before the age of 18 (which is very common, perhaps a norm, among the Macua) increases the likelihood of marital dissolution by 24 per cent compared with marrying at age 18 or later.

Another possible explanation of the matrilineal-patrilineal differentials in marital dissolution may lie on the relationship between infertility, divorce and polygyny. Marriage is unlikely to be stable if the union does not produce children. However, the investigation of this relationship from this data did not find a significant effect of childlessness (the measure used for infertility) and marital dissolution (results not shown). Nevertheless, it is possible that among matrilineal ethnic groups men are more likely to divorce their first wives if they are infertile, while among patrilineal groups, where polygyny can be a clear and an advantageous alternative, men can opt for taking another wife while not divorcing the infertile first wife. However, a clear understanding of how these factors influence each other requires more detailed information on women’s marriage histories, which were not available for this analysis. Nevertheless, these findings are robust and provide a base for future research.

REFERENCES

• Addai, Isaac. 1999a. "Ethnicity and contraceptive use in sub-Saharan Africa: the case of Ghana". Journal of Biosocial Science 31(1):105-120.
• Addai, Isaac. 1999b. "Ethnicity and sexual behavior in Ghana". Social Biology 46(1-2):17-32.
• Addai, Isaac and Frank Trovato. 1999. "Structural assimilation and ethnic fertility in Ghana". Journal of Comparative Family Studies 30(3):409-427.
• Arnaldo, Carlos. 2001. "Provincial differences in age at marriage in Mozambique". Paper presented at the Postgraduate Workshop of the African Studies Association of Australasia and the Pacific (AFSAAP) Annual Conference on Into the 21^st Century: Africa and Australasia. Melbourne, Australia: 3-6 October.
• Benefo, Kofi D., Amy O. Tsui and Joseph de G. Johnson. 1994. "Ethnic differentials in child-spacing ideals and practices in Ghana". Journal of Biosocial Science 26(3):311-326.
• Bracher, Michael and Gigi Santow. 1998. "Economic independence and union formation in Sweden". Population Studies 52(3):275-294.
• Brandon, Anastasia J. 1990. Marriage Dissolution, Remarriage and Childbearing in West Africa: A Comparative Study of Côte d'Ivoire, Ghana and Nigeria. Ph.D Dissertation. Philadelphia: University of Pennsylvania.
• Burch, T.K. 1983. "The impact of forms of families and sexual unions and dissolution of unions on fertility". Pp. 532-561 in R. D. Lee (ed), Determinants of Fertility in Developing Countries. New York: Academic Press.
• Conceição, Lino A. 1960. Os Macuas (Monografia). Lourenço Marques.
• Daniel, António. 1995. Influencia Islâmica no Litoral de Cabo Delgado: O Caso da Educação Islâmica na Ilha do Ibo. Unpublished Honours Thesis. Maputo: Departamento de História, Faculdade de Letras, Universidade Eduardo Mondlane.
• Dyson-Hudson, Rada and D. Meekers. 1996. "The universality of African marriage reconsidered: evidence from Turkana males". Ethnology 35(4):301-320.
• Gaspar, Manuel da Costa, Humberto A. Cossa, Clara Ribeiro dos Santos, Rosa Marlene Manjate and Juan Schoemaker. 1998. Moçambique, Inquérito Demográfico e de Saúde 1997. Calverton, Maryland: Instituto Nacional de Estatística-Macro International Inc.
• Gluckman, Max. 1950. "Kinship and marriage among the Lozi of Northern Rhodesia and the Zulu of Natal". Pp. 166-206 in D. Forde (ed), African Systems of Kinship and Marriage. London: Oxford University Press.
• Goldscheider, Calvin. 1971. Population, Modernization, and Social Structure. Boston: Little, Brown and Company.
• Gould, W.T.S. and M.S. Brown. 1996. "A fertility transition in sub-Saharan Africa?" International Journal of Population Geography 2:1-22.
• Hajnal, John. 1953. "Age at marriage and proportions marrying". Population Studies 7(2):111-136.
• Halli, Shivia S. and V. Vaninadha Rao. 1992. Advanced Techniques of Population Analysis. New York: Plenum Press.
• Heaton, Tim B. and R. Forste. 1998. "Education as policy: the impact of education on marriage, contraception, and fertility in Colombia, Peru, and Bolivia". Social Biology 45(3-4):195-213.
• Hosmer Jr., David W. and Stanley Lemeshow. 1999. Applied Survival Analysis: Regression Modeling of Time to Event Data. New York: John Wiley & Sons, Inc.
• Isiugo-Abanihe, Uche C. 1994. "Consequences of bridewealth changes on nuptiality patterns among the Ibo of Nigeria". Pp. 74-91 in C. Bledsoe and G. Pison (ed), Nuptiality in Sub-Saharan Africa: Contemporary Anthropological and Demographic Perspectives. Oxford: Clarendon Press.
• Kaufman, G. L. and D. Meekers. 1998. "The impact of women's socioeconomic position on marriage patterns in sub-Saharan Africa". Journal of Comparative Family Studies 29(1):101-114.
• Kinfu Ashagrea, Yohannes. 2001. The Quite Revolution: An Analysis of the Change Toward Below-Replacement-Level Fertility in Addis Ababa. Unpublished PhD Thesis. Canberra: Australian National University.
• Kollehlon, Konia T. 1989. "Ethnicity and fertility in Liberia: a test of the minority group status hypothesis". Social Biology 36(1-2):67-81.
• Le, Chap T. 1997. Applied Survival Analysis. New York: John Wiley & Sons, Inc.
• Lehrer, Evelyn L. and Carmel U. Chiswick. 1993. "Religion as a determinant of marital stability". Demography 30(3):385-404.
• Lesthaeghe, Ron. 1989. "Production and reproduction in sub-Saharan Africa: an overview of organizing principles". Pp. 13-59 in R. Lesthaeghe (ed), Reproduction and Social Organization in Sub-Saharan Africa. Berkeley: University of California Press.
• Lesthaeghe, Ron and F. Eelens. 1989. "The components of sub-Saharan reproductive regimes and their social and cultural determinants: empirical evidence". Pp. 60-121 in R. J. Lesthaeghe (ed), Reproduction and Social Organization in Sub-Saharan Africa. Berkeley: University of California Press.
• Lesthaeghe, Ron, G. Kaufmann and Dominique Meekers. 1989. "The nuptiality regimes in sub-Saharan Africa". Pp. 238-337 in R. Lesthaeghe (ed), Reproduction and Social Organization in Sub-Saharan Africa. Berkeley: University of California Press.
• Lesthaeghe, Ron, G. Kaufmann, Dominique Meekers and J. Surkyn. 1994. "Post-partum abstinence, polygyny, and age at marriage: a macro-level analysis of sub-Saharan societies". Pp. 25-54 in C. Bledsoe and G. Pison (ed), Nuptiality in Sub-Saharan Africa: Contemporary Anthropological and Demographic Perspectives. Oxford: Clarendon Press.
• Lopes, António A.P.D. 1960. Notas sobre os usos e costumes dos indígenas do Posto Admnistrativo de Corrane. Lourenço Marques.
• Martinez, Francisco L. 1989. O Povo Macua e a sua Cultura. Lisboa: Instituto de Investigação Científica Tropical.
• McDonald, David A., Lovemore Zinyama, John Gay, Fion de Vletter and Robert Mattes. 2000. "Guess who's coming to dinner: migration from Lesotho, Mozambique and Zimbabwe to South Africa". International Migration Review 34(3):813-841.
• McDonald, Peter. 1985. "Social organization and nuptiality in developing societies". Pp. 87-114 in J. Hobcraft (ed), Reproductive Change in Developing Countries: Insights from the World Fertility Survey. New York: Oxford University Press.
• Menard, Scott. 1995. Applied Logistic Regression Analysis. Sage University Paper series on Quantitative Applications in the Social Sciences, 07-106. Thousand Oaks, CA: Sage.
• Mondal, Sekh Rahim. 1997. "Polygyny and divorce in Muslim society: Controvesy and reality". Man in India 77(4):415-428.
• República de Moçambique. 2001. Projecto de Lei que Altera as Disposições de Código Civil Atinentes às Normas Reguladoras das Relações de Família e do Direito de Família. Maputo: Assembleia da República.
• Tilson, D. and U. Larsen. 2000. "Divorce in Ethiopia: the impact of early marriage and childlessness". Journal of Biosocial Science 32(3):355-372.
• Timæus, Ian M. and A. Reynar. 1998. "Polygynists and their wives in sub-Saharan Africa: an analysis of five Demographic and Health Surveys". Population Studies 52(2):145-162.
• Trussell, J. and D. E. Bloom. 1983. "Estimating the co-variates of age at marriage and first birth". Population Studies 37(3):403-416.
• Van de Walle, E. 1968. "Marriage in African censuses and inquiries". In Demography of Tropical Africa, edited by W. Brass, A.J. Coale, P. Demeny, et al. Princeton, NJ: Princeton University Press.
• Zulu, Eliya M. 2001. "Ethnic variations in observance and rationale for postpartum sexual abstinence in Malawi". Demography 38(4):467-479.

Notes

² The Family Law is still as inherited from the Portuguese administration, but it is currently under review and the legal age at marriage could rise to 18 years for both males and females (República de Moçambique, 2001).

Copyright 2004 - Union for African Population Studies

African Population Studies/Etude de la Population Africaine, Vol. 19, No. 1, April 2004, pp. 143-164

Ethnicity and Marriage Patterns in Mozambique

Carlos Arnaldo¹

Centro de Estudos de População, Universidade Eduardo Mondlane, Caixa Postal 257, Maputo -MOZAMBIQUE

¹ The author would like to thank the Mozambican National Institute of Statistics for providing the data used in this paper, and to Professor Jack Caldwell, Dr David Lucas Dr Heather Booth, and Ms Zitha Mokomane, all of the Australian National University, for their useful comments on earlier versions of the paper.

Code Number: ep04008

ABSTRACT

Marriage in sub-Saharan Africa has been commonly described as early and universal. Yet, its patterns vary across countries and even within countries among different cultural groups. Mozambique is culturally heterogeneous with both patrilineal and matrilineal systems of social organisation represented. Using data from the 1997 Census and 1997 MDHS, this article examines differentials in marriage patterns among five ethnic groups in Mozambique: Tsonga, Sena/Ndau, Lomwe/Chuwabo, Macua and Others, a residual group. Multivariate analyses are used to explore whether the reported ethnic differentials in age at first marriage, polygyny and marital dissolution can be attributed to ethnicity or to other characteristics that distinguish the ethnic groups. The findings are consistent with culture-associated differentials. After controlling for the socio-economic and demographic characteristics of women, the differences remain, with the matrilineal ethnic groups (Macua and Lomwe/Chuwabo) having an earlier age at marriage, lower prevalence of polygyny and higher marital dissolution than the patrilineal ethnic groups (Tsonga and Sena/Ndau).

Marriage in sub-Saharan Africa has been commonly described as early and universal (van de Walle, 1968; Lesthaeghe, 1989) and this situation has partly been blamed for the persistence of high fertility in the region (e.g. Gould and Brown, 1996). However, the region is far from homogenous. Marriage patterns vary across and within countries among different ethnic groups (Lesthaeghe and Eelens, 1989; Lesthaeghe, Kaufmann and Meekers, 1989). Such variations could be due to both cultural and socio-economic factors. This article aims to assess the effect of ‘culture’ (ethnicity) on marriage patterns (age at marriage, polygyny and marital dissolution) in Mozambique. The study seeks to test the hypothesis that ethnic differences in marriage patterns are a reflection of socio-economic differences among the individual members of different ethnic groups rather than ethnicity per se.

While the relative effect of ethnicity on demographic behaviour has been assessed in several countries, including some in sub-Saharan Africa (e.g. Kollehlon, 1989; Benefo, Tsui and Johnson, 1994; Addai, 1999a, b; Addai and Trovato, 1999; Zulu, 2001), in Mozambique such studies are still lacking, although the country has an ethnically heterogeneous population.

THEORETICAL BACKGROUND

Theoretically, group differences in demographic behaviour (especially fertility) have been interpreted in terms of two hypotheses: the ‘character-istics hypothesis’ and the ‘norms/cultural hypothesis’ (Goldscheider, 1971; Addai, 1999a; Addai and Trovato, 1999). Although these hypotheses have mainly been used to explain ethnic differences in fertility (Kollehlon, 1989; Addai and Trovato, 1999), contraceptive use (Addai, 1999a) and the duration of postpartum sexual abstinence (Zulu, 2001), they can also be useful in explaining ethnic differences in marriage patterns.

The characteristics hypothesis assigns ethnic differences in demographic behaviour (marriage patterns in this case) to differing socio-economic characteristics of individual members of different ethnic groups. Inherent in this hypothesis is the assumption that people of different ethnic background become assimilated into the socio-economic structure of the larger society (within the nation) as their socio-economic status improves through increased education, urban residence and modern occupations (Addai and Trovato, 1999). Ethnic affiliation can, therefore, be treated as an indicator of socio-economic status, degree of urbanisation and social mobility of the groups in a particular society (Goldscheider, 1971:273). Marriage pattern differences among ethnic groups that may exist at one point in time are considered temporary phenomena, representing a social or cultural lag, which is often indicative of social and economic backwardness or deprivation (Goldscheider, 1971). Thus, irrespective of their ethnic background, women who possess the same socio-economic characteristics are expected to have similar marriage patterns. Therefore, once differences in socio-economic characteristics are controlled for, ethnic variations in age at marriage, and the propensity to enter polygynous marriages and to experience a marital dissolution should disappear.

An alternative explanation of ethnic differences in marriage patterns is based on the ‘norms/cultural hypothesis’. This hypothesis assigns similar importance to the role of ethnicity/culture in explaining ethnic differences in marriage patterns (Addai, 1999a; Addai and Trovato, 1999). Group norms, ideals and beliefs are assumed to be important factors in determining age at marriage, propensity to enter polygyny and experience of marital dissolution.

One of the important ethnic differences is that between matrilineal and patrilineal systems of descent. Since late marriage is associated with greater independence for women, which is more likely to be found in a matrilineal system than in a patrilineal one (Kaufman and Meekers, 1998), age at marriage is expected to be later among matrilineal ethnic groups than among the patrilineal ones (McDonald, 1985; Lesthaeghe et al., 1989; Kaufman and Meekers, 1998). Matrilineal societies are also expected to have lower prevalence of polygyny than the patrilineal societies because the matrilocal or uxorilocal residence of spouses that characterises the former societies makes it difficult for a man to bring a new wife home to his first wife’s kin unless the two women are related (Lesthaeghe et al., 1989; Lesthaeghe et al., 1994; Kaufman and Meekers, 1998). Divorce is believed to be lower among patrilineal societies than in matrilineal ones (Burch, 1983; Gluckman, 1950). A patrilineal marriage involves permanent incorporation of a woman in her husband’s lineage and complete transfer of her reproductive power to her husband’s family through payment of bridewealth. Under such circumstances, divorce is more difficult because the bride-wealth would have to be returned and the woman would be separated from her children. In contrast, in a matrilineal marriage, the rights of the woman’s reproductive abilities remain with her lineage so that the duration of the marriage depends on the couple’s individual attributes (Brandon, 1990:151).

THE STUDY CONTEXT

Mozambique’s population of 16 million in 1997 is ethnically diverse. Over twenty different cultural sub-groups of both matrilineal and patrilineal systems of descent can be identified in the country. Using the information on first language, six major ethnic groups were identified: the Tsonga, Sena/Ndau, Lomwe/Chuwabo, Macua, ‘Other’ (a residual of the remaining local languages) and Portuguese/foreigner. The socio-economic character-istics of women belonging to each of these groups are summarised in Table 1. The Portuguese/foreigner is an elite group comprising Mozambicans whose first language is Portuguese and Asian and European descendants and accounts for only 6% of the total population but still comprise about one million people. They are the most educated and urbanised group. Among the local ethnic groups, the Tsonga (who live in the Southern Region) are the most educated and urbanised, followed by the Sena/Ndau of Central Region. The Macua are the least educated and the Lomwe/Chuwabo the least urbanised. There are also ethnic differences in religious affiliation.

Source: 1997 Census

While the Macua are predominately Muslim, the ‘Other’ are mainly Christians. In terms of descent system, the Macua and the Lomwe/Chuwabo are matrilineal whereas the Tsonga and the Sena/Ndau are patrilineal. Finally, the ‘Other’ category is a residual including all the remaining local ethnic groups, some of which are patrilineal and others matrilineal.

Marriage is a difficult concept in African societies because it is a process rather than a discrete event and involves rituals, negotiations and transactions that can stretch over years, making it difficult to say at what point a couple becomes married. There are also more than one type of marriage. In Mozambique, at least four main types of marital union can be identified: Customary/traditional marriage, religious marriage, civil marriage and mutual consent union/cohabitation.

Customary marriage: This is a traditional Mozambican form of marriage and is usually carried out with the consent of both families. It is a process involving a series of stages that can go on for months and even years. The way the process is conducted varies according to ethnic groups and descent system. Among the patrilineal groups (the Tsonga and Sena/Ndau) the process is centred on the payment of bridewealth from the boy’s family to the girl’s family, while among the matrilineal groups (the Lomwe/Chuwabo and Macua) such payments are not required. Under customary marriage, polygyny is permitted without any limit to the number of wives. In fact, among patrilineal ethnic groups polygyny is recommended when the first wife is infertile.

Religious marriage: This is a marriage celebrated in a church or mosque with a ceremony to bless the union of a couple where one or both are members of that religion. Christian marriages are monogamous and among Catholics, who make a large proportion of all Mozambique’s Christians, indissoluble, whereas polygyny up to a maximum of four wives per man and divorce are permitted under the Islamic law on certain specified conditions (Mondal, 1997). This (religious marriage) type of marriage is more expensive than the customary marriage since, apart from bride-wealth payments (patrilineal ethnic groups), wedding rings and modern clothes are also required. The Church or mosque keeps the record of the marriage and a marriage certificate is issued to the couple.

Civil marriage: This is a registered union of a never married/divorced/widowed woman with a never married/divorced /widowed man under the civil law. It is monogamous and indissoluble except by a decree of divorce issued by a court of law. Under Mozambican civil law the minimum age of marriage is 14 years for females and 16 years for males (República de Moçambique, 2001)². Although this type of marriage is monogamous, in practice some men are married by civil law to their first wives and live with one or more additional women. Because civil marriage is expensive, it is more prevalent among urban and more educated segments of the population. It should also be noted that, except for those of Asian and European backgrounds, civil marriage only takes place after traditional requirements (e.g. bride-wealth has been paid, in the case of patrilineal ethnic groups) have been met. Thus, in practice, couples who have had a civil ceremony also had customary or religious (if they are affiliated to a religion) ceremonies at the same time.

Mutual consent union/Cohabitation: this is a relationship where persons of the opposite sex cohabit without going through the formalities of customary, religious or civil marriage. Although a mutual consent union could be a temporary relationship lasting for a few weeks or months, it may be a more permanent tie where the pair lives as husband and wife. In some cases, cohabitation or trial marriage constitutes a stage before the completion of the marriage process, that is, before the social sanctioning of the union. For example, it is a tradition among the Macua that the couple have a ‘period of proof’ of about one year, when they behave as husband and wife, before the marriage process is complete (Martinez, 1989). In contrast, among the Tsonga, for example, the cohabitation usually takes place only after the bride-wealth has been paid in full, except in the case of premarital pregnancy, where cohabitation may precede the completion of bridewealth payment.

Despite the existence of these types of union, neither the 1997 census nor the 1997 DHS (the data sources used in this study) distinguished customary marriage from cohabitation, and religious marriage from civil marriage. Customary marriage and cohabitation were treated as a single category, labelled ‘marital union/living together’, and religious and civil marriage as the ‘married’ category. Thus, in this analysis, women in both ‘marital union/living together’ and ‘married’ categories are treated as ‘married’. The non-distinction between the four types of marriage, especially that between customary marriage and cohabitation, may affect the results. Women of ethnic groups which allow cohabitation before the completion of the marriage process (e.g. Macua) are more likely to have been recorded as being in a marital union than women of ethnic groups where cohabitation is not allowed before the completion of the marriage process (e.g. the Tsonga). This would lead to an overestimation of the proportion currently married and an underestimation of average age at marriage among the former groups and an underestimation of the proportion currently married and an overestimation of average age at marriage among the latter groups.

The Data

The data analysed in this article were taken from Mozambique’s 1997 Population and Housing Census and the 1997 Demographic and Health Survey. The 1997 Census was the second population census to be conducted in the country after independence from Portuguese rule in 1975. The census collected information on current marital status that is useful in the analysis of marriage pattern.

The Mozambique Demographic and Health Survey (MDHS) was carried out between March and July 1997. Its main objective was to provide reliable and current information on fertility and reproductive behaviour, maternal and child health, infant and child mortality, contraceptive knowledge and use, and knowledge and attitudes regarding HIV/AIDS (Gaspar et al., 1998). The 1997 MDHS is a representative, stratified probability sample of women aged 15-49. Households were identified using a two-stage cluster sampling procedure based on census enumeration areas. Three hundred and eighty-eight enumeration areas were selected and 100 to 200 households were selected within each enumeration area. Ninety-eight per cent of all eligible women were successfully interviewed, resulting in a final sample of 8,779 women (Gaspar et al., 1998).

Methods of Analysis

Multivariate analysis techniques are used to examine three aspects of the marriage patterns: age at marriage, polygyny and marital dissolution. The analysis of age at marriage is based on both census and DHS data, whereas those of polygyny and marital dissolution are based only on the DHS data. A complete list of such variables and the methods used for each analysis are presented in Table 2.

Since age at marriage was not collected in the 1997 census, a dichotomous variable indicating whether or not a woman aged 15-19 has ever been married was constructed from the information on current marital status. A random sample of about 15,000 women was drawn from the data set of women aged 15-19 and a logistic regression model (Menard, 1995) was fitted to determine whether ethnic differences in the propensity of ever having been married among women aged 15-19 remained sizeable and statistically significant once other socio-economic factors were held constant. For the DHS data, the dependent variable is the reported age at marriage. Because the age at marriage of women who were never married at the date of the interview, but who will eventually marry some time in the future (censored cases), is unknown, proportional hazard models were used to analyse the data. The proportional hazard model is a life table technique which is suitable for the analysis of censored data and has widely been used in demographic analysis (e.g. Trussell and Bloom, 1983; Halli and Rao, 1992; Le, 1997; Bracher and Santow, 1998; Heaton and Forste, 1998; Hosmer Jr. and Lemeshow, 1999; Kinfu Ashagrea, 2001).

For the analysis of polygyny, the dependent variable is a dichotomous variable indicating whether or not a currently married woman is in a polygynous marriage.

Finally, marital dissolution is measured by a dichotomous variable indicating whether or not a woman had been married more than once at the time of the survey. Because having married more than once does not only measure the frequency of divorce but also the frequency of widowhood, which is related to the level of mortality, the proportion of children ever born who have died is included in the model to allow for differential mortality. Ideally, data on adult mortality are needed to give an indication of the likelihood of widowhood. Since such data are not available, the proportion of children dead has been used as a mortality indicator, as there is normally a fairly close correlation between childhood and adult mortality levels.

The independent variable for the study is ethnicity categorised into Tsonga, Sena/Ndau, Lomwe/Chuwabo, Macua and ‘Other’. Women of Portuguese/foreigner ethnic background were excluded from the analysis because of their diverse composition. To allow for women’s socio-economic differences, many control variables, selected for explicit theoretical reasons and availability in the data sets, were included in the analysis.

RESULTS

Age at Marriage

Table 3 presents the percentage of women never married by selected age groups, the singulate mean age at marriage (SMAM), a measure of age at first marriage computed from the proportions never married by age (Hajnal, 1953), and the median age at first marriage by ethnicity. These measures show that there are substantial ethnic differences in the timing of marriage. The Tsonga have the latest age at marriage, followed by the Sena/Ndau, whereas the Macua have the youngest. It is interesting to see that women of patrilineal descent (Tsonga and Sena/Ndau) marry later than their counterparts of matrilineal descent (Lomwe/Chuwabo and Macua), although the difference between Sena/Ndau and Lomwe/Chuwabo is small. For instance, 73% of Tsonga and 50% of Sena/Ndau women aged 15-19 have never been married, compared with only 48% and 32% of Lomwe/Chuwabo and Macua, respectively.

Table 3: Percentage of Women Who Are Never Married, SMAM and Median Age at First Marriage by Ethnic Affiliation, Mozambique 1997

Source: a= Computed from 1997 Census; b= Computed from the 1997 MDHS

To assess the independent effect of ethnicity on age at first marriage, socio-economic characteristics were controlled for. The results are presented in Table 4. The first two columns are from the 1997 census and indicate the odds ratios that a woman of each ethnic group aged 15-19 has ever been in a union compared with a woman in the reference category (Tsonga). The last two columns are from the 1997 DHS and present the risk of first marriage for each ethnic group relative to the risk for the reference category. The odds ratio and the relative risk for the reference category are equal to unity, and, thus, a value of 1.0 for any of the other categories means that the ethnic group in question has the same first marriage timing as the reference category. A value higher than one indicates greater likelihood/risk of marriage than the reference category, whereas values less than one indicates a lower likelihood/risk of marriage. Models I and III show the gross effects (without controls) and Models II and IV the net effect (with controls).

The results show that the introduction of socio-economic controls did not eliminate ethnic differences in age at marriage. Lomwe/Chuwabo and Macua ethnic backgrounds are significantly associated with younger age at marriage than the Tsonga. Model II shows that the odds that a Lomwe/Chuwabo or a Macua woman aged 15-19 has ever been married is 1.46 times and 2.57 times respectively that of her Tsonga counterpart, while there is no significant difference between Tsonga and Sena/Ndau women in the odds of first marriage. The same, but less striking, pattern of relationships is observed from the DHS analysis (Models III and IV): the risks of first marriage for a Lomwe/Chuwabo and a Macua woman are 36 per cent and 13 per cent respectively higher than for a woman in the reference group (Tsonga).

Regarding the lineage system, the results show that the patrilineal ethnic groups (Tsonga and Sena/Ndau) have a later marriage pattern than the matrilineal ethnic groups (Lomwe/Chuwabo and Macua). This is in contrast with the literature (McDonald, 1985; Lesthaeghe et al., 1994; Kaufman and Meekers, 1998) which suggests later marriage among women in matrilineal than patrilineal societies. However, the early marriage pattern among the Macua is consistent with ethnographic studies (e.g. Conceição, 1960; Lopes, 1960). Conceição (1960:12) reported that Macua girls could marry between age 8 and 10, while, according to Lopes (1960), engagement could happen even earlier, at age 6. However, because Macua marriage is a process rather than a discrete event, these ages may not refer to the beginning of cohabitation but to the beginning of the process leading to it. On the other hand, later marriage among the Tsonga may reflect high male migration to South Africa or Maputo City (McDonald et al., 2000; Arnaldo, 2001), perhaps to gather resources for marriage, which may raise age at marriage for women by creating a shortage of males of marriageable age.

In relation to the other socio-economic factors included in Models II and IV, education and paid employment increase the age at marriage in comparison with no education and non-employment, respectively; while there is no agreement between the results from the census (Model II) and those from the DHS (Model IV) in relation to the effect of self-employment on age at marriage. The census shows a positive effect and the DHS a negative one. The reason for this inconsistency is unclear. But, it is important to note that these results may be misleading because employment status at the time of the census or survey, as used in this analysis, may not be the same as that before marriage.

Model II shows that not having a religion increases the likelihood of first marriage by 30 per cent in comparison to being Catholic, while urban residence and residence in the Southern Region are associated with later first marriage than rural and Northern or Central residences respectively.

The practice of polygyny is fairly common in Mozambique. The DHS found that 28 per cent of currently married women were in polygynous unions. However, there are significant differences among ethnic groups. The highest level of polygyny is observed among the Sena/Ndau (40%) followed by the Tsonga (28%). For the remaining groups the prevalence of polygyny is 22% for the Macua, 20% for the ‘Other’ ethnic group and 15% for the Lomwe/Chuwabo.

The results of the multivariate analysis performed to assess the independent effect of ethnicity on polygyny are presented in Table 5. The table shows odds ratios that a currently married woman of each ethnic group is in a polygynous marriage compared with her counterpart in the reference category (Tsonga) without controls (Model I) and with controls (Model II). Model I shows that there are significant differences in the propensity to be in a polygynous marriage by ethnic affiliation. Sena/Ndau ethnicity increases the likelihood of being in a polygynous marriage by 35 per cent relative to the reference group, whereas being a Lomwe/Chuwabo and a Macua reduces it by 64 and 32 per cent respectively. These differences do not disappear when socio-economic controls are introduced (Model II). In fact, the odds of being polygynous decrease for the Lomwe/Chuwabo and Macua, whereas those for Sena/Ndau decrease and become statistically not different from the Tsonga at the 5% level. In both Model I and II ‘Other’ ethnic background has a negative effect on polygyny but it is only significant at the 5% level in Model II. These findings are consistent with the tendency of patrilineal descent systems to be more polygynous than matrilineal systems. A matrilineal Lomwe/Chuwabo or Macua woman is 74% and 60%, respectively, less likely to marry a polygynist than a woman of the Tsonga ethnic group. On the other hand, there are no significant differences between a patrilineal Tsonga and a patrilineal Sena/Ndau woman in the likelihood of marrying a polygynist. In Ghana, Timæus and Reynar (1998) also found that polygyny was slightly higher among patrilineal populations than matrilineal ones. In a multivariate analysis by Lesthaeghe et al. (1989:314), matrilinearity produced on average a 13-percentage-point diminution in polygyny when compared to patrilinearity, which is consistent with the theoretical proposition that matrilineal systems are less amenable to polygynous household formation because of their tendency towards matrilocal or uxorilocal residence (Lesthaeghe et al., 1989:314).

Table 5: Odds Ratios of Being in a Polygynous Union by Ethnic Background, Married Women 15-49, Mozambique 1997

Note: ® = Reference Category; *p≤0.05; **p≤0.01; ***p≤0.001.
Source: 1997 DHS

Table 5 also shows that women’s secondary education and non-employment are associated with a low likelihood of being in polygynous marriage. Women whose husbands have white-collar occupations are more likely to be polygynous than those whose husbands have other occupations, suggesting, perhaps, that men’s socio-economic status is an important factor in determining their financial ability to be polygynous. Non-Catholic women have higher propensity to enter polygyny than the Catholics, partly reflecting a strong stand against polygyny by the Catholic Church. Rural residence and long marital duration are also associated with a higher likelihood of polygyny.

Marital Dissolution

Table 6 presents two measures of marital dissolution derived from the 1997 census and the 1997 DHS: the proportion of ever married women aged 15-49 who are currently divorced and the proportion of women who had married more than once at the time of the survey. Both measures should be interpreted with caution. The first, which represents a ‘snapshot’ of the proportion currently divorced, is influenced by both marital dissolution and subsequent remarriage. Thus, given that the interval to remarriage after divorce is often culturally determined and so may vary considerably between different groups, the fact that one group has a higher proportion currently divorced than another does not necessarily mean that it has higher divorce rate than the other group. Also, the proportion of women who have been married twice or more may overestimate the level of marital dissolution since two different factors could have caused multiple marriages: remarriage after divorce and remarriage after widowhood. Nevertheless, the census and DHS figures in Table 6 show higher marital instability among the matrilineal ethnic groups (Macua and Lomwe/Chuwabo) than among the patrilineal (Tsonga and Sena/Ndau) and ‘Other’ ethnic groups. For example, more than half of ever married Macua and 27 per cent of Lomwe/Chuwabo women have been married twice or more as compared with less than 21 per cent for women in the remaining ethnic groups.

Table 6: Measures of Marital Dissolution by Ethnicity, Mozambique 1997

Source: Computed from 1997 Census^a and ^bDHS

The independent effect of ethnicity on marital instability was investigated through multivariate analysis, and the results are presented in Table 7. As in the previous analysis, Model I shows the gross effect and Model II the net effect. Model I shows that only being of Sena/Ndau and Macua ethnic backgrounds is significantly associated with higher marital instability than the Tsonga while there is no significant difference between the remaining ethnic groups (Lomwe/Chuwabo and ‘Other’) and the Tsonga. When socio-economic controls are introduced (Model II), the effect of Macua background is reduced but is still significant at the 5% level, and the effect of Lomwe/Chuwabo background increases but is still not significant at the 5% level. The effects of Sena/Ndau and ‘Other’ also decreased and lost their statistical significance at the 5% level. The high instability of Macua marriage has been reported by several ethnographic studies (e.g. Conceição, 1960; Lopes, 1960; Martinez, 1989). According to Conceição (1960), a smallest quarrel was enough reason for a Macua man or woman to divorce his/her partner. For instance, ‘being tired’ was a plausible reason to dissolve a Macua marriage (Lopes, 1960). Although religious adherence is high among the Macua, they are mostly Muslims, who as discussed earlier have high marital instability. Furthermore, the Macua also seem to pay more allegiance to culture than to religion.

Regarding the other socio-economic factors in the Model II, education, employment status, place of residence and number of living children do not have a significant effect on marital dissolution. In contrast, religion, radio ownership, age at marriage, type of marriage, the proportion of children ever born who have died and region are important determinants of marital dissolution. Women who profess a religion other than Catholicism are more likely than the latter to have been married more than once; age at marriage is negatively associated with marital instability. Senior wives in polygynous marriages are less likely than monogamous women to have been married more than once, while, in contrast, junior wives are more likely than senior wives to have experienced a marital dissolution, indicating, perhaps, a strong social feeling against a man’s first marriage being with a divorced woman. This makes divorced women to commonly marry polygynously rather than to an unmarried man. The proportion of children ever born that died has a very strong positive effect on the likelihood that a woman has had multiple marriages. This implies that some marriages were dissolved because the wife was blamed for giving birth to children who died. Marriage is more than three times as unstable in the Northern Region and 10 per cent more unstable in Central Region than in Southern Region, but only the odds ratio for the Northern Region is statistically significant.

Table 7: Odds Ratios of Having Married more than Once by Ethnicity, Women 15-49, Mozambique 1997

Note: ® = Reference Category; *p≤0.05; **p≤0.01; ***p≤0.001.
Source: 1997 DHS

DISCUSSION AND CONCLUSION

This article examined two hypotheses on ethnic differentials in age at marriage, polygyny and marital dissolution. One hypothesis (‘the characteristics hypothesis’) postulates that ethnic differences in marriage patterns are an artefact of socio-economic differences between women of different ethnic backgrounds and that once these differences are accounted for (through statistical controls) group variations in marriage patterns should disappear. The second hypothesis (‘norms/cultural hypothesis’), although acknowledging the effect of socio-economic differences, argues that there are also ethnic differences in marriage patterns in terms of cultural values, norms and practices.

The findings from the study appear to support the ‘norms/cultural hypothesis’. The empirical evidence in Tables 4, 5 and 7 shows that socio-economic controls could not account for all the ethnic differences in timing of marriage, prevalence of polygyny and marital stability. Regarding age at marriage, the findings contradict not only the ‘characteristic hypothesis’ but also the expectation that marriage is earlier among patrilineal than matrilineal descent systems. This study suggests the reverse. The fact that marriage is later among the patrilineal (Tsonga and Sena/Ndau) than the matrilineal (Lomwe/Chuwabo and Macua) ethnic groups, reflects a greater importance given to the bride-wealth payment in the marriage process of patrilineal groups. Since the value of bride-wealth is continuously rising, men in these societies may need more time to gather the necessary resources to meet the costs of marriage than their counterparts in matrilineal ethnic groups where bride-wealth is not required (see e.g. Isiugo-Abanihe, 1994, on the Ibo of Nigeria; Dyson-Hudson and Meekers, 1996, on the Turkana of Kenya). Although this is more likely to affect males than females, girls who do not want to marry a polygynist may be forced to wait for a long time to find a man with enough resources (or for their boyfriends to gather the resources) to meet the costs of marriage. In addition, male migration may also play a role. The Tsonga, who have the latest marriage pattern, are characterised by high male migration to South Africa or to Maputo City (McDonald et al., 2000; Arnaldo, 2001), which may raise female age at marriage by creating a shortage of males of marriageable age.

An early age at marriage for women is usually accompanied by a large gap between female and male ages at first marriage (Lesthaeghe, 1989). Assuming inter-ethnic marriages are rare, the figures presented in Table 8 appear to support this proposition with the age gap at marriage being widest among the early marrying Macua and smallest among the Tsonga. Table 8 also shows that the ratio of males to females of marriageable age is lowest among the Tsonga indicating that the high labour migration that characterises this group leads to shortage of males of marriageable age that may be related to late marriage of females.

Table 8: Singulate Mean Age at Marriage by Sex and Male/Female Ratio by Ethnicity, Mozambique 1997

Source: 1997 Census.

There is, however, a possible ethnic differential bias on current marital status and age at marriage caused by the failure to distinguish cohabitation from customary marriage. Since women of matrilineal descent systems (especially the Macua) cohabit with their partners before the marriage process has been completed, they would be more likely to be recorded as being in union than women of patrilineal descent (Tsonga in particular) where cohabitation usually occurs only after the process has been completed. If this has happened, the Tsonga proportion of women 15-19 currently never married and the odd ratios from the multivariate analysis (Table 4, Models I and II) would be biased upward. However, the consistency between census and DHS results, where more probing was made and living together relationships were recorded and considered as marriages, suggests that such bias did not account for all ethnic differences that emerged from the analysis.

The results with regard to polygyny confirmed the higher propensity of women in patrilineal ethnic groups to enter polygyny than their counterparts in the matrilineal groups. Among the Tsonga and the Sena/Ndau, polygyny is recommended as the best solution for the infertility of the first wife and in such a situation the second wife can even be a younger sister or other relative of the first wife. It is also important to note that, although the matrilineal ethnic groups have a low prevalence of polygyny, there are still considerable differences between them. This analysis has shown that Macua women are 48 per cent more likely to marry a polygynist than their counterparts of Lomwe/Chuwabo ethnic background. This relatively high prevalence of polygyny among the Macua is partly related to the Muslim influence they received from Arab Muslim traders starting from the eighth century (Martinez, 1989; Daniel, 1995). Because they have matrilocal residence, polygynous Macua men have to reside at their senior wives’ family and visit the other wives elsewhere in turns.

In terms of marital dissolution, the findings are consistent with expectations. Matrilineal ethnic groups have a higher probability of experiencing marital dissolution than patrilineal groups, which is consistent with the argument that the absence (or low value) of bride-wealth payments in matrilineal marriage makes divorce easier, while high bridewealth payments and the fact that children belong to the husband’s lineage among the patrilineal ethnic groups make divorce difficult in these societies. Thus, since divorce implies the return of the bridewealth and the separation of the mother from her children, women in patrilineal societies (Tsonga and Sena/Ndau) may be more hesitant to leave unsatisfactory marriages than their counterparts in matrilineal societies. In contrast, in the Macua and Lomwe/Chuwabo marriage, there is no transfer of women’s productive and reproductive powers and the children belong to the women’s lineage so that divorce is easy and does not involve any kind of wealth transfer, since bridewealth payment is not required. The fact that women in matrilineal ethnic groups marry at a younger age than those in patrilineal groups, increases their risk of divorce, mainly due to immaturity and the short duration of search for a suitable partner resulting in poor information about the partner’s characteristics (see e.g. Lehrer and Chiswick, 1993; Tilson and Larsen, 2000). This analysis (results not shown) has shown that marrying before the age of 18 (which is very common, perhaps a norm, among the Macua) increases the likelihood of marital dissolution by 24 per cent compared with marrying at age 18 or later.

Another possible explanation of the matrilineal-patrilineal differentials in marital dissolution may lie on the relationship between infertility, divorce and polygyny. Marriage is unlikely to be stable if the union does not produce children. However, the investigation of this relationship from this data did not find a significant effect of childlessness (the measure used for infertility) and marital dissolution (results not shown). Nevertheless, it is possible that among matrilineal ethnic groups men are more likely to divorce their first wives if they are infertile, while among patrilineal groups, where polygyny can be a clear and an advantageous alternative, men can opt for taking another wife while not divorcing the infertile first wife. However, a clear understanding of how these factors influence each other requires more detailed information on women’s marriage histories, which were not available for this analysis. Nevertheless, these findings are robust and provide a base for future research.

REFERENCES

• Addai, Isaac. 1999a. "Ethnicity and contraceptive use in sub-Saharan Africa: the case of Ghana". Journal of Biosocial Science 31(1):105-120.
• Addai, Isaac. 1999b. "Ethnicity and sexual behavior in Ghana". Social Biology 46(1-2):17-32.
• Addai, Isaac and Frank Trovato. 1999. "Structural assimilation and ethnic fertility in Ghana". Journal of Comparative Family Studies 30(3):409-427.
• Arnaldo, Carlos. 2001. "Provincial differences in age at marriage in Mozambique". Paper presented at the Postgraduate Workshop of the African Studies Association of Australasia and the Pacific (AFSAAP) Annual Conference on Into the 21^st Century: Africa and Australasia. Melbourne, Australia: 3-6 October.
• Benefo, Kofi D., Amy O. Tsui and Joseph de G. Johnson. 1994. "Ethnic differentials in child-spacing ideals and practices in Ghana". Journal of Biosocial Science 26(3):311-326.
• Bracher, Michael and Gigi Santow. 1998. "Economic independence and union formation in Sweden". Population Studies 52(3):275-294.
• Brandon, Anastasia J. 1990. Marriage Dissolution, Remarriage and Childbearing in West Africa: A Comparative Study of Côte d'Ivoire, Ghana and Nigeria. Ph.D Dissertation. Philadelphia: University of Pennsylvania.
• Burch, T.K. 1983. "The impact of forms of families and sexual unions and dissolution of unions on fertility". Pp. 532-561 in R. D. Lee (ed), Determinants of Fertility in Developing Countries. New York: Academic Press.
• Conceição, Lino A. 1960. Os Macuas (Monografia). Lourenço Marques.
• Daniel, António. 1995. Influencia Islâmica no Litoral de Cabo Delgado: O Caso da Educação Islâmica na Ilha do Ibo. Unpublished Honours Thesis. Maputo: Departamento de História, Faculdade de Letras, Universidade Eduardo Mondlane.
• Dyson-Hudson, Rada and D. Meekers. 1996. "The universality of African marriage reconsidered: evidence from Turkana males". Ethnology 35(4):301-320.
• Gaspar, Manuel da Costa, Humberto A. Cossa, Clara Ribeiro dos Santos, Rosa Marlene Manjate and Juan Schoemaker. 1998. Moçambique, Inquérito Demográfico e de Saúde 1997. Calverton, Maryland: Instituto Nacional de Estatística-Macro International Inc.
• Gluckman, Max. 1950. "Kinship and marriage among the Lozi of Northern Rhodesia and the Zulu of Natal". Pp. 166-206 in D. Forde (ed), African Systems of Kinship and Marriage. London: Oxford University Press.
• Goldscheider, Calvin. 1971. Population, Modernization, and Social Structure. Boston: Little, Brown and Company.
• Gould, W.T.S. and M.S. Brown. 1996. "A fertility transition in sub-Saharan Africa?" International Journal of Population Geography 2:1-22.
• Hajnal, John. 1953. "Age at marriage and proportions marrying". Population Studies 7(2):111-136.
• Halli, Shivia S. and V. Vaninadha Rao. 1992. Advanced Techniques of Population Analysis. New York: Plenum Press.
• Heaton, Tim B. and R. Forste. 1998. "Education as policy: the impact of education on marriage, contraception, and fertility in Colombia, Peru, and Bolivia". Social Biology 45(3-4):195-213.
• Hosmer Jr., David W. and Stanley Lemeshow. 1999. Applied Survival Analysis: Regression Modeling of Time to Event Data. New York: John Wiley & Sons, Inc.
• Isiugo-Abanihe, Uche C. 1994. "Consequences of bridewealth changes on nuptiality patterns among the Ibo of Nigeria". Pp. 74-91 in C. Bledsoe and G. Pison (ed), Nuptiality in Sub-Saharan Africa: Contemporary Anthropological and Demographic Perspectives. Oxford: Clarendon Press.
• Kaufman, G. L. and D. Meekers. 1998. "The impact of women's socioeconomic position on marriage patterns in sub-Saharan Africa". Journal of Comparative Family Studies 29(1):101-114.
• Kinfu Ashagrea, Yohannes. 2001. The Quite Revolution: An Analysis of the Change Toward Below-Replacement-Level Fertility in Addis Ababa. Unpublished PhD Thesis. Canberra: Australian National University.
• Kollehlon, Konia T. 1989. "Ethnicity and fertility in Liberia: a test of the minority group status hypothesis". Social Biology 36(1-2):67-81.
• Le, Chap T. 1997. Applied Survival Analysis. New York: John Wiley & Sons, Inc.
• Lehrer, Evelyn L. and Carmel U. Chiswick. 1993. "Religion as a determinant of marital stability". Demography 30(3):385-404.
• Lesthaeghe, Ron. 1989. "Production and reproduction in sub-Saharan Africa: an overview of organizing principles". Pp. 13-59 in R. Lesthaeghe (ed), Reproduction and Social Organization in Sub-Saharan Africa. Berkeley: University of California Press.
• Lesthaeghe, Ron and F. Eelens. 1989. "The components of sub-Saharan reproductive regimes and their social and cultural determinants: empirical evidence". Pp. 60-121 in R. J. Lesthaeghe (ed), Reproduction and Social Organization in Sub-Saharan Africa. Berkeley: University of California Press.
• Lesthaeghe, Ron, G. Kaufmann and Dominique Meekers. 1989. "The nuptiality regimes in sub-Saharan Africa". Pp. 238-337 in R. Lesthaeghe (ed), Reproduction and Social Organization in Sub-Saharan Africa. Berkeley: University of California Press.
• Lesthaeghe, Ron, G. Kaufmann, Dominique Meekers and J. Surkyn. 1994. "Post-partum abstinence, polygyny, and age at marriage: a macro-level analysis of sub-Saharan societies". Pp. 25-54 in C. Bledsoe and G. Pison (ed), Nuptiality in Sub-Saharan Africa: Contemporary Anthropological and Demographic Perspectives. Oxford: Clarendon Press.
• Lopes, António A.P.D. 1960. Notas sobre os usos e costumes dos indígenas do Posto Admnistrativo de Corrane. Lourenço Marques.
• Martinez, Francisco L. 1989. O Povo Macua e a sua Cultura. Lisboa: Instituto de Investigação Científica Tropical.
• McDonald, David A., Lovemore Zinyama, John Gay, Fion de Vletter and Robert Mattes. 2000. "Guess who's coming to dinner: migration from Lesotho, Mozambique and Zimbabwe to South Africa". International Migration Review 34(3):813-841.
• McDonald, Peter. 1985. "Social organization and nuptiality in developing societies". Pp. 87-114 in J. Hobcraft (ed), Reproductive Change in Developing Countries: Insights from the World Fertility Survey. New York: Oxford University Press.
• Menard, Scott. 1995. Applied Logistic Regression Analysis. Sage University Paper series on Quantitative Applications in the Social Sciences, 07-106. Thousand Oaks, CA: Sage.
• Mondal, Sekh Rahim. 1997. "Polygyny and divorce in Muslim society: Controvesy and reality". Man in India 77(4):415-428.
• República de Moçambique. 2001. Projecto de Lei que Altera as Disposições de Código Civil Atinentes às Normas Reguladoras das Relações de Família e do Direito de Família. Maputo: Assembleia da República.
• Tilson, D. and U. Larsen. 2000. "Divorce in Ethiopia: the impact of early marriage and childlessness". Journal of Biosocial Science 32(3):355-372.
• Timæus, Ian M. and A. Reynar. 1998. "Polygynists and their wives in sub-Saharan Africa: an analysis of five Demographic and Health Surveys". Population Studies 52(2):145-162.
• Trussell, J. and D. E. Bloom. 1983. "Estimating the co-variates of age at marriage and first birth". Population Studies 37(3):403-416.
• Van de Walle, E. 1968. "Marriage in African censuses and inquiries". In Demography of Tropical Africa, edited by W. Brass, A.J. Coale, P. Demeny, et al. Princeton, NJ: Princeton University Press.
• Zulu, Eliya M. 2001. "Ethnic variations in observance and rationale for postpartum sexual abstinence in Malawi". Demography 38(4):467-479.

Notes

² The Family Law is still as inherited from the Portuguese administration, but it is currently under review and the legal age at marriage could rise to 18 years for both males and females (República de Moçambique, 2001).

Copyright 2004 - Union for African Population Studies

The following images related to this document are available:

Photo images