1

Marriage markets and marital status of South African mothers

By Siv Gustafsson and Seble Worku Abstract This paper studies the effects of local marriage markets in the South African women’s marital decisions. The analysis is motivated by the low proportion of marriages among African South Africans and in particular among those with at least one child. Marriage market indicators have been defined at district levels as availability of economically attractive marriageable men. The analysis is done among women aged 20 to 40 who have already given birth to their first child based on South African census 2001 data. An ordered probit model is fitted with the different marital type ranked from less desirable (never married) to more attractive (married civil). African women enter civil marriages at a lesser rate than White women. The estimation results also suggest that both the quantity and quality of the marriage market matter in women marital choice. Options that are available for women besides marriage make them to be more selective in their choice of spouse and type of marriage.

Key words: Marriage market, sex ratio, marriageable men, ordered probit, African, White.

University of Amsterdam Department of Economics & Tinbergen Institute Roetersstraat 11 1018 WB Amsterdam The Netherlands S.S.Gustafsson@uva.nl S.Yergou@uva.nl JEL classification: D1, J1.

2

1. Introduction

Marriage rates are currently on the decline as couples are increasingly opting for cohabitation and postponement of marriage, or individuals prefer to remain single. Accordingly recent studies have concentrated on the various determinants of marital behavior among women. Links between labour market positions, government policies, the marriage market and cost of children have been noted to have an influence on union formation and fertility decisions (Ermisch 2003, Gustafsson 2001). Other determinants such as the lack of quality in the marriage market and the shortage of marriageable men have been proposed by many researchers (Wilson, 1987). Moreover, in the United States, several studies have focused on the causes of existing racial differences in marriages and family structure and suggestions were made that the differences might be attributed to the legacy of slavery and welfare policies (Frazier 1939, Murray 1984). Table 1 shows marriage patterns according to race in South Africa. Marriage in South Africa is currently almost universal among Whites and Asians but much lower among Africans and Coloured. In 2001 by age 44, close to 87% of White women are married as compared to 58% of African women. Also examining the difference between 1996 and 2001, we observe an increase in the rate of cohabitation as opposed to a decrease in the marriage rate.

Table 1: Women’s marital status by age

African White

Never

married Married

civil Married

customary Living

together Never

married Married

civil Married

customary Living

together Age <15 99.8 0.1 0.1 0.0 99.8 0.1 0.0 0.0 15-19 95.9 0.5 1.4 2.2 96.1 1.8 0.2 1.9 20-24 80.8 2.9 6.5 9.7 68.5 20.7 0.7 10.0 25-29 61.2 11.0 13.1 14.7 28.7 60.6 1.4 9.3 30-34 45.9 22.6 17.4 14.2 14.1 77.8 1.5 6.6 35-39 36.4 31.0 19.8 12.7 9.0 83.7 1.6 5.7 40-44 31.6 36.0 21.5 10.9 7.0 86.6 1.6 4.9 45-49 28.3 38.9 23.5 9.3 5.9 88.3 1.5 4.3 Total Census 2001 77.0 9.5 7.5 6.1 51.9 42.8 0.9 4.4 Total Census 1996 77.7 10.2 8.4 3.7 52.3 44.4 0.7 2.5 Among women aged 20-40 Highest education: None 48.7 13.4 22.8 15.0 50.5 40.8 2.8 5.9 Low 53.3 15.3 16.2 15.2 21.9 67.0 2.0 9.1 Medium 68.1 14.5 8.6 8.9 30.7 60.2 1.3 7.8 High 58.8 29.4 6.8 5.1 29.6 62.3 1.0 7.1 With at least one child 50.6 19.4 16.5 13.5 5.8 86.1 1.8 6.4 Employed 54.5 22.2 10.4 12.9 27.9 62.2 1.2 8.7

Source: Own computation based on Census 2001

3

Among the African population, families are formed later, or not at all. African women increasingly postpone marriage with marriage taking place mostly when women are between 35-39 years old. However among the same age group close to 36% still remain single. The rate of cohabitation is higher among Africans compared to Whites. In addition customary marriage is more common among Africans than Whites. Among African women, the higher the education level of the woman the more likely she would remain single and the less likely she would cohabit. Moreover highly educated women marry in civil marriage and the less educated women would more likely marry in cus tomary marriage. Among African women with at least one child close to 51% are single mothers as compared to only 6% of White women. Among employed African women close to 55% are never married whereas the rest are either married or living with a partner. Similarly among employed White women close to 28% are single whereas the rest are living with a spouse or a partner.

The racial differentials observed in the marriage patterns may be caused by the social conditions resulting from the legacy of apartheid laws. Apartheid rules made it illegal for African men and women to live together as husbands and wives which often resulted in break-up of families after they have been formed or for many family units to never be formed1. Forced to leave the land by its inability to provide any subsistence, the men migrate into the cities as poorly paid laborers. During the long periods of their youthful, sexually active lives, men and women had to live apart. Men would only visit their families during the annual two-week holiday when migrant workers were allowed to visit their homelands and would support their families through remittances. However this income was often not sufficient since African men access to the labour market was limited and confined to low skilled labour which reduces their capacity to support a family. The men become alienated and set up a new life in the towns while their wives wait for months hoping for a letter or money. This forces thus women to supplement their income by taking paid labour. Also African men’s access to education was restricted. Thus the availability of high quality marriageable men is poor.

This paper models women’s decisions in the marriage market as an outcome of marital choice. My hypothesis is that women would prefer civil marriage over all type of relationship once they have a child. The rest of the paper is organized as follows. Section 2 describes the different types of marriage institutions in South

1 The 1952 pass law act enacted to control influx of the African population into urban areas required African men and women over the age of 16 to carry a pass book showing permission to stay in urban areas. The law allows the arrest and removal of any African who is considered idle and undesirable. By leaving the workers families on the land with the assumption that women can feed their families of the produce of the land it is possible to pay lower wages and avoid having to build houses and supply services that are essential to maintain urban populations. Thus the employer obtains a labour supply at less than the wage necessary to maintain the worker’s family, while ensuring the continued supply of labour through annual leave periods that allowed the conception of children. The Pass Laws were repealed in 1986 and that eased legal restrictions on the migration from rural areas to the cities and townships by people searching for work. But the migration also caused a proliferation of "squatter" communities on the periphery of urban centers.

4

Africa and gives empirical evidence on the decline of marriage as well as the profile of individuals who enter the marriage market. Section 3 discusses the theory of marriage in the South African context. Section 4 presents the empirical strategy followed and describes the data set used. Section 5 examines the results and section 6 concludes.

2. Marriages in South Africa In South Africa, different types of marriage habits are practiced depending on the population group of the individuals. Table 2 shows the different type of marriage distribution by population group in South Africa. Civil marriages or also known as common law marriages are similar to those in developed countries. Traditional marriages are also known as customary marriages and include African, Hindu and Muslim marriages and may be polygamous. In 2001, there were 31,382 polygamous men in South Africa. Before 1984 all marriages among Whites and Coloured were in civil marriage and were automatically in community of property whereas Africans and Indians could choose to marry according to civil or customary marriages which were automatically out of community of property.

Table 2: Martial status by population group and gender among men and women aged 15-44

African Coloured Indian White Men Women Men Women Men Women Men Women Married civil/religious 10.6 13.2 30.3 31.9 42.7 48.5 45.4 51.0 Married traditional/customary 9.0 10.9 1.3 1.4 4.8 5.4 0.9 1.1 Polygamous marriage 0.2 0.0 0.1 0.0 0.1 0.0 0.1 0.0 Living together like married partners 8.9 9.6 8.2 9.1 1.8 2.0 5.5 5.8 Never married 70.0 62.6 57.8 52.5 48.3 38.2 44.4 35.3 Widower/widow 0.4 1.7 0.4 1.4 0.3 2.1 0.3 1.0 Separated 0.6 1.0 0.5 0.9 0.5 1.0 0.4 0.6 Divorced 0.4 1.0 1.4 2.8 1.5 2.8 3.0 5.1 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Number (Thousands) 8,483 9,174 976 1,049 288 294 975 1,005 Sex ratio 92.5 93.0 98.0 97.0

Source: Own computation based on Census 2001 However during the rules of apartheid, the various customary marriages were not recognized as legal marriages and thus not regulated by the government. Although marriages rules were different for whites and non-whites, husbands get marital powers upon marriages to their wives in all types of marriages2. Women who are minors cannot own property in their own rights, enter into contracts without the help of their male guardian, or act as guardians of their own children. As shown in

2 The matrimonial Act of 1984 scrapped marital power of a husband over his wife in White civil marriages. The Marriage and Matrimonial Property Amendment Act o f 1988 made African civil marriages automatically in community of property and scrapped marital power of the husband.

5

Table 2, 73% of African men and 42% of White men are never married in 2001. Close to 19% of African men and 49% of White men are married. One reason for this is the different age distribution between the two races as shown in Table 1. In addition, Table 2 indicates that close to 11% of African men and women are married by common law whereas close to 8% of men and 9% of women are married by customary law (Census, 2001). For some South African traditiona list, polygamy is a way of preserving their culture. This practice was in the past mostly confined to members of the royal family and traditional leaders. Also the practice solved the problem of the barren wife that the husband cannot abandon since divorce was not allowed in African customary marriages. According to customary laws, the first wife has to give consent to the second marriage. Also the husband must provide proper living arrangements whereby the wives get their separate quarters. However these rules were rarely observed and men married two wives without the knowledge of the first wife. Subsequently, the first wife often finds herself deprived from access to the marital property. Lobola or dowry, the transfers of cattle/and or money by the husband to the wife’s father validate customary marriages. The settlement of the terms of the lobola is done through elaborate negotiations between the two families. As soon as the lobola price is agreed upon the first installment is delivered and the rest is paid later3. Although some feminist view lobola as oppressive to women, people are still attached to the customs of lobola and children are considered as illegitimate if lobola is not transferred. Thus while lobola payments are not necessary in civil marriages many Africans who marry in civil marriages go through the rites of lobola transactions4. The most famous polygamist is the ex-deputy president Jacob Zuma who was married to 3 wives and paid lobola for a 4th wife in 2002. One of his wives, the Minister of Foreign Affairs Nkosazana Dlamini-Zuma divorced him in 1998, another wife died of a heart attack in 2000 and he is still married to the 3rd wife. His 4th wife to be is Princess Sebentile Dlamini, a niece of Swazi King Mswati III. Mr Zuma has paid 10 cattle to the family of the bride. Also Johannesburg’s City Power chief executive officer, Mr Mogwailane Mohlala, announced in an interview with a Sunday paper in 2003 that he was looking for wife number two. Apparently his wife, who was sitting next to him during the interview, was quite comfortable with this idea as long as she was part of the decision-making process5. A widow friend of mine who was married by customary law to her late husband nearly lost her right to the marital property to her in- laws upon the death of her husband. Before 1998, in customary African marriages no actual wedding ceremony and signing of a marriage contract needs to take place. Lobola payment 3 In 2005, lobola payment can very between R10 000 to R25 000 depending on the wealth of the family. 4 Even with the recognition of customary marriages as of 1998 couples tend to register their marriages as civil but would still practice the rites of lobola transfers. Lawmakers tend to currently look at the best interest of the mother and child when lobola related custody or money transfers become court cases. 5 Dyanti A., 2003, Now it’s wives and mistresses, The Star online edition.

6

is a proof of the marriage occurrence and thus my friend didn’t have a marriage certificate. Thus according to the old law, she had no right to the property of her husband and succession to the property goes through to the male lineage of the family (her eldest son or her husband male relative). The couple bought the property after their marriage while during the marriage the wife invested most of her own income in the improvement of the property. The marriage did not produce any child but her husband had a daughter from a previous relationship. The brother in- law- proclaimed guardianship to both child and widow and occupied the house together with his family. My friend eventually returned to her parent’s house. She is still battling in court to get back at least part of what belongs to her. However in 1998, the Recognition of Customary Marriage Act was passed giving legal recognition to African customary marriages through compulsory registration of the marriages and abolishing marital powers of the husband. Community of property applies automatically unless ante nuptial is signed between the partners and the law directs that entry into customary marriage must be with consent of both partners. This includes widows marrying their brother- in- laws can still occur as long as the widows consent. Even polygamy is permitted to continue as long as the interests of the first wife are protected. The law regulates second and subsequent marriages entered by the man by cutting down on the excessive power that a husband used to have in a customary marriage. Also the law ensures equitable distribution of family property between all wives and children and thus is seen by many as solving the problem of property rights on the death of the husband 6 (Mamashela and Xaba, 2003). Thus this law helped my friend win her case. In addition, the law regulates dissolution of customary marriages and thus ensures the protection of women. However some argue that the true of type of reason why polygamy couldn’t be outlawed was that individual choice between customary and civil type of marriage should be considered as a matter of democracy and freedom of choice in the new South Africa. Others view this choice as a pure matter of personal taste and thus should be replaced by civil marriage. On the other hand, cohabitation in South Africa has no legal bearing in terms of ownership of estate unless the property is registered under both partners’ names. Partners have no automatic right to pensions and other benefits. Either of the parents can act as legal guardians of the children depending on the best interest of the child. This is the situation in which my domestic worker finds herself. She was in an unmarried cohabitation for more than 10 years with the man she calls her “husband” while he is still legally married to his wife and mother of his four children who lives in a rural area. This man has five children with my domestic worker and has guardianship to all the children. Upon recent break up of the relationship, she is fighting to get guardianship of the youngest children who share her surname so that she also gets access to child support grants from the government.

6 The court has to approve a written contract that will regulate the matrimonial property. The court may not grant the application for the second and subsequent wife if it believes that the interests of all parties are not sufficiently safeguarded.

7

It is thus clear that African women have in the past found themselves in very difficult situations while they totally depended on men in terms of laws and customs. There have also been contradictions on who is supposed to take responsibility and decisions and who actually does. Also the possibility for women to enter paid work and access to different forms of government support has recently made an important impact on relationships.

Table 3: Age, marital status and sex ratios

A.Men Britain Sweden South Africa:

African South Africa:

White South Africa:

Total

Age Not

married All Not

married All Not

married All Not

married All Not

married All <15 34.8 19.7 29.1 17.1 44.1 35.5 39.2 20.0 43.8 33.4 15-24 21.8 12.6 18.4 10.9 27.0 22.1 29.7 15.8 27.3 21.2 25-34 16.3 14.7 17.0 12.9 16.1 16.3 13.0 15.4 15.7 16.3 35-44 9.6 15.2 12.1 14.3 7.0 11.8 6.5 15.7 7.0 12.5 45-54 6.2 13.4 9.0 14.7 3.1 7.3 4.6 13.3 3.3 8.1 55-64 4.2 10.7 6.5 13.4 1.4 3.8 3.2 10.3 1.6 4.7 65-74 3.3 8.0 3.7 8.6 0.7 2.0 2.1 6.2 0.9 2.5 >74 3.8 5.7 4.2 8.2 0.5 1.1 1.8 3.3 0.6 1.3 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 B. Women <15 30.7 17.9 26.9 14.7 40.6 32.7 33.4 18.1 40.0 30.8 15-24 18.9 11.8 16.6 9.4 24.8 21.1 23.8 14.4 24.7 20.2 25-34 12.9 14.1 13.8 11.5 14.0 16.2 9.5 15.2 13.6 16.2 35-44 8.2 14.7 9.8 12.8 7.9 12.2 7.0 15.8 7.9 12.9 45-54 5.9 13.0 8.0 13.6 4.7 7.7 6.1 13.4 4.9 8.5 55-64 5.0 10.5 6.7 12.8 3.3 4.8 5.9 10.5 3.6 5.4 65-74 6.4 8.7 5.9 10.1 2.8 3.4 6.5 7.1 3.1 3.8 >74 11.9 9.4 12.2 15.1 1.9 1.9 7.8 5.5 2.3 2.2 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Men (Thou) 14,485 25,574 2,852 4,869 13,591 16,887 1,061 2,080 16,341 21,434 Women (Thou) 15,636 26,786 2,938 5,374 14,902 18,528 1,195 2,212 17,950 23,385 Sex ratio total 92.6 95.5 97.1 90.6 91.2 91.1 88.8 94.0 91.0 91.7 Sex ratio <15 105.0 105.0 105.4 105.4 99.1 99.1 104.1 104.2 99.6 99.6 Sex ratio 15-44 110.3 100.0 114.1 102.1 97.8 92.5 108.5 97.0 98.4 93.0 Sex ratio 45-64 88.4 97.9 102.5 96.7 52.0 81.4 57.2 92.9 52.2 83.8 Sex ratio >65 35.9 72.6 42.3 60.2 23.7 53.0 24.2 70.8 24.6 57.7 Sex ratio 20-40 112.7 99.3 116.2 101.8 100.2 91.8 114.9 96.4 100.9 92.4

Source: Own computation based on Statistics South Africa: Census 2001; Office of the National Statistics UK: Mid-2001 population estimates for England and Wales; Statistics Sweden: Administrative records data, 2001. Table 3 shows the age breakdown of never married men and women comparing South Africa to Sweden and Britain7. Although South Africa has a relatively

For Britain, mid-year of the resident population estimates revised in the light of the local authority population studies are used. For Sweden administrative record data showing the conditions in

8

young population as compared to Sweden and Britain the White population presents an age profile closer to the two European populations. However in Britain 10% of men and in Sweden 12% of men are not yet married by age 44 whereas in South Africa, only 7% of White men in a similar age group are not yet married. A similar distribution is observed among women. Among British and Swedish women aged 74 and above, 12 % are not married because they correspond to widows. This pattern is also observed among White South African women in the same age group. Table 3 also shows that women are in majority for all populations presented in the table, with a total sex ratio varying from 90.6 for Sweden representing the largest shortage of men due to the presence of mostly old widows to 95.5 for Britain. The total sex ratio for Africans in South Africa is 91.1 showing a big shortage of men. This finding is also confirmed by the low sex ratio observed among Africans aged 20 to 40.

3. Theoretical explanation of South African marriage patterns

3.1 Gary Becker’s theory Following Becker (1991), we analyze marriage markets outcomes of staying single, cohabiting without being married, being in a monogamous relationship or polygamous relationship for women or men in terms of gains from marriage output. The marital output includes labour earnings but also results of household work and enjoyment of children. Let us assume identical male and female marriage participants and let a common marriage output be produced within the household and be known with certainty. Let the partners’ marriage output be:

fmmf ZZZ += (1)

where Zm and Zf are male and female shares of marital output within marriage respectively. Men and women would prefer to marry if and only if their utility from marriage exceeds their utility from remaining single:

smm

sff ZZandZZ >> (2)

where Zsf and Zsm are outputs of single women and men outputs respectively. At Zf = Zsf women are indifferent between marrying and being single, at Zm = Zsm men are indifferent between marrying and being single, and at Zmf - Zm < Zmf

-

Zsm men gain from marrying. In the case of monogamy, since Nm < Nf all men marry and some women (Nf - Nm) remain single. In this case all men take all gains from marriage. These women are willing to stay single because their income equals to the income of married women. If Nm > Nf then women take all gain from marriage. Thus the

December 2001 are used. To make the comparison easier, 2001 data has been used for all the countries.

9

model suggests that the proportion of married women is positively related to the ratio of men to women. The above discussion can be extended to polygamous marriages. Assume that the demand for wives increases with polygamy, thus some men are willing to marry an additional wife and offer her:

Zf =MPf(2) = Zmf(2) – Zmf(1) (3) Figure 1: Equilibrium male-female income distribution in monogamous and polygamous marriages Woman’s income Sf S’f MPf(1) = Zmf(1 ) – Zsm e polygamy MPf(2) = Zmf(2 ) - Zmf(1) Zsf emonogamy Nm Nf 2Nm Number of women Source: Becker, 1991 p. 83-85

where Zmf(1) is the output of a household with one man and one woman and Zmf(2) is the output of a household with one man and two women. Equilibrium in an efficient polygamous marriage market doesn’t require that the same number of men and women want to marry, only that the number of women who want to marry equals the demand for wives. Thus at point epolygamy all men and all women marry and some men have two wives. At epolygamy all men and all women receive the same income Zmf(2) - Zmf(1) where all receive the same marginal product of the second wife. Although the number of women is higher than the number of men, the equilibrium income of women is above their single income. Thus women would rather enter polygamous marriages rather than remain single. In sum if all men would have at least n-1 wives and

10

some would have n wives, monogamy would cost each woman the difference between the marginal product of the nth wife and her single income. Thus according to Becker (1991), women are better off in terms of income under polygamy rather than monogamy. Given the inequality in resources between women and men and the shortage of marriageable men, it is better for a woman to be a second wife than remain single. It is even the case that all women benefit if some are second wives because this mechanism washes away the shortage of men. This does not mean that women are happy with these arrangements. If they had more resources they would prefer an egalitarian monogamous marriage. If the women were better off as singles they would not accept marriage. If they were better off as first wives and have an opportunity to have a say in the choice of subsequent wives as the anecdote of Mrs Mohlala whose story suggests that she would not consent to a second wife unless she plays a role in the process. Table 3 above shows that women are in majority for all populations included varying from 91.1 for African South Africans the largest shortage of men to 98.0 in Sweden the smallest shortage of men.

3.2 Past findings on polygamy From Figure 1 it is clear that sex ratio imbalances are necessary for polygamy to occur. Becker argues in a subsequent section that heterogeneity in men may cause the wealthier men to become polygamist even if some men remain single. Grossbard (1980) has researched the heterogeneity in men hypothesis and states the extent of resource inequality among men has a positive effect on the level of polygamy in a society that practice both monogamy and polygamy. This means that women determine the incidence of polygamy in such a society. Also in an earlier paper Grossbard (1976) she uses survey data collected in 1969 on the Maiduguni tribe (Nigeria) and shows that polygamy is a positive function of male income and male education. However she also shows that the more educated the senior wife is the fewer her co-wives are, suggesting that some men substitute marriage to one highly educated wife rather than to a number of uneducated women (Grossbard, 1976). Kanazawa et al (1999) use cross-country comparison data and reach similar conclusions. More specifically, they assert that increasing women’s power should result in more polygamous marriages if men’s resources inequality is large, but should result in more monogamous marriages if men’s inequality is small. Kanazawa et al (1999) also explain institutional change from polygamy to monogamy as a function of resources inequality among men but do not provide empirical evidence to support that. On the other hand, Gould et al (2003) argue that the incidence of monogamy or polygamy is determined by the sources of inequality, not just the level of inequality. In particular, they show that the marriage market is more monogamous if inequality is determined more by differences in human capital versus non- labour income. Their conclusion is similar to Becker’s (1991) where quality of wife and children are valued over quantity, higher quality women have higher prices making polygamy less affordable.

11

Becker’s analysis has been the focus of much research and his arguments have further been extended to provide further theoretical causes for women’s marital behaviour. Some of the hypotheses advanced are a shortage of marriageable men and increased economic independence of women through labour income or welfare income (Wilson 1987, Lichter et all 1992, Wood 1995, Brien 1997, Angrist 2002, Fitzgerald 2003). Furthermore the occurrence of dowries and other marriage endowments are not only a matter of culture but can also be dictated by the socio-economic group of the parents. Other theories are based on habits formation and change of tastes rather than economic arguments. These include living arrangements while growing up such as exposure to non-traditional family type (never married parents, migrant father), family background while growing up and or change in taste and norm which some also associate with urbanization (Jencks, 1992).

3.3 Elaborations and empirical analysis of the sex ratio hypothesis The shortage of economically attractive men “marriageable men” and the shortage of prospective marriage partners due to sex ratio imbalances as well as the Black and White differences in marriage markets have been researched extensively in the US. Wilson (1987) uses current employment as a proxy for a man’s ability to support a family and to construct the “male marriageable pool index”. His major finding that is often quoted is that the decline in the number of Black men who are able to support a family is a major source of the decline in marriage rates among Blacks. Using women’s marital history available from the US National Longitudinal Survey of Youth data Lichter et al (1992) estimate a search model linking quality and quantity of available men to first marriage transition among women aged 18 and above. They construct local market conditions indicators expressed in terms of sex ratios using US census 1980 data (PUMS-D data). They conclude that a shortage in the quantity and quality of available mates in local areas depresses women’s transition to marriage. Also they conclude that local mate availability accounts for larger share of observed racial difference in marriage rates in the US. Angrist (2002) uses US census data from the US 1920 and 1940 censuses on immigrants to study the effects of changing sex ratios among immigrants by ethnicity. More than half of all marriages during the studied period took place within the same ethnic group. His finding are that higher sex ratios are associated with higher marriage rates for both men and women and causes men to marry sooner and also give them incentive to become more attractive. In a similar study done by using the same data set Brien (1997) concludes that ratio of marriageable men to women within specific geographic areas play important roles in the decision to marry. However Wood (1995) findings using US census metropolitan level statistics 1970-1980, suggest that the shrinking pool of high-earning young Black men explains little of the decline in Black marriage. He rather suggests that increased economic opportunities for Black women should be the answer to the question. Lichter et al (1992) oppose this point since their result indicate that economic independence of women is

12

positively associated with entry into marriage. Fitzgerald (2003) studies the link between spouse availability and welfare use among US women using the Current Population Survey data for 1996. He concludes that unavailability of marriageable men increases the use of welfare. However his results further show that for Black women labour market conditions are more important than marriage market conditions for exit from welfare. The available literature also provides important contribution on how to define marriageable men and how to construct marriage market indicators. An important indication is that marriage market variables have to be created in various marriage pools and must include a spouse quality measure. Wilson (1987) uses current employment status of young men to define marriageable men. Lichter et al (1992) use 3 definitions of marriageable men: men who are employed; men who are full-time employed and men whose income is above a certain threshold. Brien (1997) uses 5 definitions of marriageable men by using all men; all employed men; all men employed, in school and who were short-term unemployed; all men who are full-time employed; all men with earnings above a certain amount. He also disaggregates these variables by activity and education. Researchers suggest that measures have to be disaggregated by race and local area since marriage markets are usually segregated by race and spouse searching is generally confined to a local area. However there is uncertainty among researchers on whether to use earnings or employment status to measure spouse quality. Since marriage encourage employment and since the definition of a marriageable man is a man that is able to support his family it is better to use income as a definition of a marriageable man rather than his employment status (Brien,1997; Wilson,1987; Wood, 1995).

3.4 The bride price and dowries theory Bride price is a payment made as a compensation to the bride’s family for the loss of a productive member of the household but also payment by the groom’s family for the services he receives from his wife including the rights to the children she bears for him. Dowry is the exact opposite of the bride price as it consists of a cash transfer from the bride’s family to the groom’s family and is practiced more commonly among Asian countries. In societies that still practice the custom, the size of the dowry is proportional to the groom's social standing. In addition it has been suggested that bride price is the discounted value of part or the wife’s entire share of marriage output and the value of this share must be at least equal to the bride’s possible earnings in alternative households (Becker 1973, Papps 1983). Papps (1983) hypotheses that bride price is the same for all brides of the same type, higher for brides of higher quality and independent of the productivity and wealth of the groom and his family. Using data from a Palestinian village collected in the mid 1920s he evaluates the determinants of bride-price assuming children as public goods. His finding implies that a virgin bride is more than twice worth of a widow bride and grooms would have to pay higher bride prices if they decide to marry outside their village. Tertilt (2002) uses an overlapping

13

generations model with a marriage market to investigate whether differences in the type of marriages that are allowed in a country account for the difference in bride prices value. He shows that monogamy leads to a negative bride price (dowries) whereas polygamy leads to strictly positive bride prices. Tertilt’s other findings include that couples in polygamous marriages have a large age difference whereas those in monogamous marriage have the same age. Under monogamy, daughters select their own husbands whereas under polygamy their marriages are arranged (Tertilt, 2002). Whereas Bergstrom (1994) is in agreement with Becker (1991) on women benefiting more from polygamous marriages rather than monogamous marriages, he argues that they do not benefit from the payment of the bride price money since it is paid to the male relatives of the bride. However, he still maintains women are still better off in polygamous marriages since they do not have to pay dowry. But some authors maintain that such transfers are unrelated to the marriage market and should be regarded as bequests given to sons and daughters at marriage or as some kind of ceremonial or status defining function (Botticini and Siow, 2000;Kressel, 1977).

3.5 Conclusions from earlier theoretical and empirical work The above results generate the conclusion that the decline in marriage rates is due to possible shortage of marriageable men. It is important to measure marriage market indicators in terms of both quantities and qualities of marriageable men disaggregated by local areas in order to capture the effect of marriage markets on marriages rates. For example, 45% of African males under the age of 20 live in rural areas but this figure declines to only 39% by the age of 30. This trend begins to reverse itself after the age of 55, with almost 38% of males returning to their rural homelands by the age of 70 years (Census, 2001). In South Africa, the lower sex ratio among the African population is worsened by the relatively fewer eligible African men. Indeed African men are still confined to lower skill jobs which provide low remuneration or are unemployed. In South Africa, lobola transfer occurs also if the marriage is a civil modern marriage and not only if the marriage is a traditional one. Thus the high unemployment rate or low level of remuneration of the urban poor makes it impossible to raise the lobola money. Hence one can suggest the recent shift to cohabitation to be partly due to the un-affordability of the lobola price. Unfortunately this theory will not be tested in this paper due to lack of data on lobola payment. In addition, the shift from traditional family values towards new social norms has lead to acceptability of non-marital sexual activity. On the other hand, with the advent of democracy in South Africa rules and policies advocating for rights and protections of women have been put in place. Women have increased access to education and the labour market. Also women have universal access to child support and social pension grants by the extension of social benefits coverage 8. Thus another hypothesis is that women have increasingly the freedom of choosing in which kind of relationship they want

8 The grant is known as child support grant and amount in 2005 to R 160 per month per child. It replaces the state maintenance grant which in the past was restricted to the non-African population only.

14

to engage themselves. This paper attempts at identifying the underlying determinants of each type of marriage options that are available in South Africa. Also using marriage market theories, this paper aims at determining whether or not it is the increased independence achieved by women through access to education and the labour market, their social background, the level of urbanization of the area where they live and the un-availability of marriageable men that is responsible for women’s decision into entry in the different forms of marital options. Women forming union may choose from among 4 institutional arrangements: never married or single, cohabit, married traditional or married civil. Some are easy to enter and exit (like cohabitation) and others have better benefit but entry and exit may be difficult, and hence some are desirable and others are less desirable. The most desirable type of relationship for women is assumed to be a monogamous marriage and thus civil marriage would be preferred over traditional marriage since the latter has the potential to be polygamous. In addition, this paper aims at testing the hypothesis that when the sex ratio declines women are more likely to accept less desirable forms of marriages.

4. Empirical approach In this study of South African women’s marriage markets, the focus is on women with at least one child and I assume that marital choices can be ordered from the least attractive 1-never married, 2-unmarried cohabitation, 3-traditiona l marriage to the most attractive 4-civil marriage. Therefore the ordered probit model is used for analysis. With ordered alternative choices, one alternative is similar to those close to it and less similar to those further away. The ordered nature can only be handled by specifying a probit model that accounts for the pattern of similarity and dissimilarity among alternatives. In ordered probit, an underlying score is estimated as a linear function of the independent variables and a set of cut points (Greene, 2003). The probability of observing outcome i corresponds to the probability that the estimated linear function, plus random error is within the range of the cut-points µi-1 estimated for the outcome. Let the underlying response model be described as:

εβ += '* XY (4) where Y* is unobserved, X is a set of explanatory variables that are measurable factors and ε is the residual or unobservable. What we do observe is:

1*1 µ<= YifY (5)

12 *2 µµ ≤<= YifY 23 *3 µµ ≤<= YifY

*4 3 YifY ≤= µ

15

The βs are unknown parameters to be estimated. In order to identify the parameters of the model, the normalization rule that var(ε)=1 is imposed. Thus ε ~IN(0,1). Thus we have the following probabilities:

)'()'()|1(Pr 1 βφβµφ XXXYob −−−== (6) )'()'()|2(Pr 12 βµφβµφ XXXYob −−−== )'()'()|3(Pr 23 βµφβµφ XXXYob −−−==

)'(1)|4(Pr 3 βµφ XXYob −−==

For all the probabilities to be positive we must have 0 < µ1 < µ2 < µ3 . Census data are used for the creation of the local marriage market conditions indicators at a low level of desegregation, local municipalities as well as develop individual socio-economic conditions indicator variables used for the rest of the analysis. Similar to Lichter et al (1992), we make the assumption that current marriage market conditions are strongly correlated to marriage rates. The 2001 census is the second democratic census in South Africa. Data collection was done by face-to-face interviews or self-completion of the questionnaires by respondents. Census 2001 data is currently available at different small area statistics such as local government levels that are also referred to as district councils and metropolitan areas. Currently, there are 6 metropolitan areas and 47 district councils 9. The census data includes information on individuals and households. The complete census data at unit record level is used for analysis. We use a sub-sample of women aged 20 to 40 in 2001 and who have already given birth to their first child because these groups of women have had some kind of relationship with a male partner. The dependent variable is ordered from the least attractive outcome to the most attractive outcome as: 1-never married, 2-cohabiting, 3-traditional marriage, 4-civil marriage. Four factors influencing these choices among adult women are considered. The first factor indicates women’s economic independence. This factor is described by the following variables: income, employment status and education10. A second factor that is used is the background of the woman and is described by the population group of the woman,

9 Administratively, the country is divided into 9 provinces (Western Cape, Eastern Cape, Northern Cape, Free State, KwaZulu-Natal, North West, Gauteng, Mpumalanga, Limpopo. The provinces are in turn divided into 365 magisterial districts. Each magisterial district had a unique name, a unique code number and clearly defined boundaries. The country is divided into 53 district councils including metropolitan areas. Metropolitan areas feature high population density and multiple business districts and industrial areas. These are: City of Cape Town; Ethekwini Municipality (Durban); City of Johannesburg Metropolitan Municipality; Ekurhuleni Metropolitan Municipality (East Rand); Nelson Mandela Metropolitan Municipality (Port Elizabeth); City of Tshwane Metropolitan Municipality (Pretoria). 10 The income levels considered are none: for those with no income; low for income less than or equal R800; medium: between R801-3 200; high: above R3 200. The education levels considered are none: no education; low: grade 6 (standard 4) or below; medium: between grade 7 and grade 12; high: diploma or above.

16

her age, her religion, whether she lives with her parents and whether she is in school. A third factor is geography described by living in a metropolitan area and the local unemployment rate at the district council or metropolitan area level. The DC level unemployment rate is estimated using unit record level census 2001 data11. The metropolitan area include the 6 biggest metropolitan areas in South Africa which consist of the cities of Cape Town, Johannesburg, Tshwane, Nelson Mandela, Ethekwini and Ekurhuleni12. The last factor is a factor indicating the local marriage markets at district council level. The selection of marriageable men indicators is guided by the work of Lichter et al (1992), Wood (1995). Four indicators are used based on the sex ratios of males to females by race, age, employment, earnings and education level to measure the local area supply or deficit of marriageable men. These ratios are calculated as:

∑

∑+

−

+

= 7

3

9

i

ii

i

ii

i

F

MSRP (7)

where SRPi is the age-population group specific sex ratio, Mi is the number of all men for a certain population group and in a certain age group in the local marriage market. Fi has a similar definition13. The three other ratios are calculated by modifying the numerator of the above ratio where Mi becomes the number of employed men or the number of men with education higher than grade 12 (standard 10) or men with income greater than R 800 per month14. Given the small number of interracial marriages that happen in South Africa, it is proper to assume that the marriage market is racially segregated15. It is important to measure marriage market indicators in terms of both qualities and quantities of marriageable men disaggregated by local areas in order to capture the effect of marriage markets on marriage rates. For example, 45% of African males under the age of 20 live in rural areas but this figure declines to only 39% by age 30. This 11The official definition of unemployment rate is used. Derived from a logical series of steps involving a) Worked past 7 days, b) Job although absent, Work category, c) Reason absent from work, d) Acceptance of job, e) Time to start work, f) Work seeking action. 12 See footnote 7 above for previous names of these cities. 13 For a 20-year old African woman, the sex ratio is equal to the number of African men aged 20 to 29 divided by the number of African women aged 17 to 27. 14 In South Africa, there is still no national poverty line. Thus researchers have used different measures of poverty. In 2000 a household income of R800 per month is considered as poor (Stats SA, 2000b). The Human Science Research Council considers an income of R587 per one-household member as poverty income (HSRC, 2004). Currently R2 000 is considered as minimum wage income in the public sector. We use an individual income of R800 per month or above as income suitable to support a family. In October 2001, R 1 was 11.91 Euros. 15 The 1949 mixed marriages act prohibited marriage between persons of different population groups. Also the 1950 immorality act banned sexual relations between Blacks and Whites. The 1950 Group Area Act declares areas for the exclusive use of one particular racial group. It makes it compulsory for people to live in an area designated for their classification group. Although these legislations have today been scraped marriage across racial lines occur rarely.

17

trend begins to reverse itself after the age of 55, with almost 38% of males returning to their rural homelands by age 70 (Census 2001). Means and standard deviations of all variables included in the analysis are presented in Table 4. African women constitute the largest group followed by Coloured women16. The mean age of the women is 30. Among the women included in the data, 70% are unemployed with an average district unemployment rate of 35%. Also 29% live in a big city, 25% live with their parents and 11% of Table 4: Mean and standard deviation of explanatory variables

Variable Description Mean SD,Max,Min Marital choice Never married 0.46 Cohabiting 0.13 Married traditional 0.14 Married civil 0.28 Women economic independence

Highest level of education: None 0.12 Low 0.46 Medium 0.34

Education

High 0.08 Employment status Current employment status: Employed 0.30 Unemployed 0.70

Monthly income: None 0.66 Low 0.29 Medium 0.06

Income

High 0.01 Background of the woman Current age Age in completed years 30.30 (5.7;20;40) Population group Based on self-classification: African 0.82 Coloured 0.10 Indian 0.02 White 0.06 Lives with parents If the woman lives with her parents 0.25 Spouse not in household If the woman does not live with her spouse 0.11 Studying If the woman is currently studying 0.07 Religion Based on broad classification: Christian 0.83 Traditional 0.09 Judaism 0.00 Hinduism 0.01 Islam 0.01 No religion 0.12 Unknown 0.01 Geography Metropolitan area If the woman is currently staying in a big city 0.29 Unemployment level of the area District level unemployment rate 35.25 (11.0;13.8;68.0) Local marriage markets Sex ratio Ratio of men by women 0.73 (0.15;0.3;4.7) Number of employed men Ratio of employed men by women 0.40 (0.19;0.1;4.7) Number of educated men Ratio of men with high school education by

women 0.70 (0.17;0.2;7.0) Number of men with sufficient income

Ratio of men with income beyond R800 by women 0.31 (0.19;0.0;4.3)

Source: Own computation based on Census 2001 16 Census 2001 unadjusted for undercount has been used in this analysis. The undercount level was 17% and was higher among the Whites.

18

the women have husbands who are migrant workers. The majority of women (46%) have low education and an even larger number (66%) have no income at all which makes them less attractive mates. In addition 46% of women in the data are single whereas 28% are married in civil. Local marriage market values vary from 73% to 31% with only 31% of women finding a marriageable man with sufficient income.

5. Results Tables 5A and 5B present mate availability by population group and age. Both tables show that White women have a much better marriage market compared to African women with respect to all mate availability measures and elevated criteria reduces the number of possible marriage candidates. Table 5A depicts existing inequalities in educational attainments as well as income among African and White men and women aged 20 to 40. Among African employed men at age 30, 70% are married whereas 43% of employed men are unmarried. Table 5B presents the mean mate availability ratio by population group and age. The table shows that at age 30 for each African woman there are only 0.628 African men educated at high school level or higher whereas for each White woman of the same age, there are 0.925 White men educated at high school level or higher. Moreover, African women face an even worse marriage market while looking at employed men. At age 30 again, for each African woman there are 0.399 employed African men whereas for each White woman of the same age, there are 0.774 employed White men. However if income is the basis for marriage, African women are further worse off. Table 6 presents results from estimating separate probit models to establish the determinants of type of marital relationship women enter. Preliminary analyses are presented on Table A2 of the Appendix. Model (A) includes all women aged 20 to 40 and is estimated including women economic independence, socio-economic background, geographical location variables and without the marriage market indicators to serve as a benchmark. Results of the models show differences in behaviour between the races to be large and statistically significant. In general, the results show that all three factors (women economic independence, their social background, and the characteristics of the area where they stay) account for a smaller difference in the racial differences in marital status choice. Most explanatory variables are highly significant. Exceptions are Judaism and high monthly income level implying that these aspects of individual background do not contribute for women’s decision to be married in civil marriages.

19

Table 5A: Percentage of mate availability by population group and age

Married men Unmarried men Women Age Total married

Employed With sufficient income

With ≥ high school

Employed With sufficient income

With ≥ high school

Married with ≥ high school

Unmarried with ≥ high school

A. African 20 1.4 30.4 10.8 18.3 9.7 3.8 24.0 16.7 29.3 21 1.9 37.5 14.6 20.9 14.2 6.0 28.7 20.2 33.0 22 2.8 45.3 18.6 22.4 19.4 9.0 33.6 24.9 37.4 23 3.8 50.9 23.7 24.7 24.4 12.6 36.9 28.7 39.8 24 5.3 56.5 28.1 25.7 28.7 15.5 37.7 30.9 40.0 25 7.3 59.0 32.9 29.4 33.0 18.9 38.8 33.5 40.4 26 9.6 62.6 37.3 31.1 36.6 21.4 37.8 34.4 39.2 27 12.2 65.5 41.3 34.2 38.9 23.6 37.6 35.6 38.4 28 15.5 67.6 45.0 35.3 41.4 25.6 36.2 35.4 36.5 29 18.7 68.6 47.0 36.4 42.4 26.6 34.4 34.6 34.9 30 22.2 69.9 50.7 38.4 43.1 27.8 33.0 35.2 33.2 31 26.6 70.8 51.7 35.8 44.1 27.8 29.5 31.4 29.8 32 31.1 71.4 53.1 35.1 45.0 28.9 27.4 30.7 27.1 33 35.7 71.9 54.7 34.7 45.1 28.9 24.9 28.5 25.5 34 39.7 71.9 55.7 34.8 44.4 28.9 23.6 28.9 24.6 35 42.3 71.5 55.8 31.8 43.9 28.0 20.1 25.5 21.3 36 44.8 70.7 54.9 28.9 43.6 27.6 17.8 23.0 19.0 37 48.1 70.1 54.9 27.2 42.9 27.2 16.0 21.4 17.6 38 50.1 69.6 54.5 25.0 42.3 26.1 14.1 19.6 15.8 39 52.1 68.9 54.0 23.5 41.5 25.5 13.1 18.6 15.0 40 53.2 68.6 54.1 22.9 40.9 24.9 12.4 18.4 14.9 Total 21.4 68.7 51.0 30.3 32.8 19.4 30.4 27.4 31.8 B. White 20 2.2 66.6 66.0 56.8 48.1 45.2 75.1 56.9 83.1 21 3.5 73.9 72.2 59.6 55.2 52.5 77.7 66.3 84.8 22 6.6 84.9 81.6 67.8 64.0 61.5 81.1 71.9 86.6 23 11.3 89.2 86.8 73.3 71.1 68.7 81.8 76.5 86.3 24 18.7 90.1 87.6 76.3 77.3 74.9 81.3 79.2 86.0 25 26.9 92.8 90.0 80.3 79.7 77.0 81.9 81.0 85.4 26 36.1 93.3 91.6 82.3 81.8 79.2 81.5 82.3 84.3 27 43.9 93.7 91.6 82.2 82.0 79.5 80.7 82.9 84.2 28 51.0 94.3 92.0 83.4 82.7 80.0 80.6 83.3 82.3 29 58.1 94.3 92.3 82.9 82.7 80.6 78.8 83.2 81.5 30 62.4 94.6 93.1 84.4 81.9 81.0 79.1 83.1 80.4 31 66.6 94.3 92.8 83.8 82.1 80.7 77.9 83.2 79.7 32 70.0 94.3 93.0 82.9 82.0 81.1 77.0 82.1 78.0 33 72.5 93.8 92.5 82.6 80.6 79.8 74.7 81.6 76.9 34 74.2 94.0 92.9 81.5 81.1 79.9 72.8 80.4 76.7 35 75.5 93.4 92.4 80.4 80.8 80.0 73.3 78.6 74.0 36 76.6 93.1 92.1 79.3 80.4 79.6 69.9 77.1 71.9 37 77.9 93.6 92.6 79.0 79.2 78.2 70.4 76.6 71.3 38 77.9 92.4 91.6 79.2 79.5 79.2 70.5 77.2 71.7 39 78.9 92.9 92.0 79.3 78.8 78.3 71.1 77.4 73.0 40 79.3 92.6 92.1 79.3 77.4 77.8 70.3 76.9 72.7 Total 52.0 93.3 92.0 80.9 72.8 70.7 78.1 79.7 81.4 Source: Own computations based on Census 2001

20

Table 5B: Sex ratio of availability of mate by population group and age

African White

Age Sex ratio

Employed men

Men with sufficient income

Men with ≥ high school

Sex ratio

Employed men

Men with sufficient income

Men with ≥ high school

20 0.745 0.203 0.115 0.896 0.863 0.656 0.632 1.078 21 0.738 0.231 0.135 0.839 0.879 0.707 0.685 1.027 22 0.737 0.261 0.159 0.787 0.891 0.747 0.725 0.981 23 0.749 0.293 0.185 0.754 0.895 0.771 0.750 0.950 24 0.751 0.318 0.206 0.724 0.899 0.786 0.767 0.927 25 0.745 0.336 0.224 0.699 0.900 0.796 0.778 0.918 26 0.736 0.348 0.237 0.675 0.888 0.791 0.775 0.909 27 0.737 0.364 0.253 0.660 0.879 0.786 0.772 0.907 28 0.733 0.374 0.263 0.647 0.871 0.781 0.769 0.906 29 0.636 0.377 0.283 0.571 0.758 0.735 0.738 0.996 30 0.732 0.399 0.282 0.628 0.861 0.774 0.765 0.925 31 0.727 0.392 0.285 0.622 0.852 0.766 0.758 0.928 32 0.720 0.392 0.287 0.612 0.845 0.759 0.752 0.936 33 0.723 0.397 0.293 0.612 0.843 0.756 0.750 0.946 34 0.713 0.393 0.291 0.600 0.844 0.756 0.751 0.959 35 0.719 0.395 0.293 0.595 0.844 0.754 0.750 0.966 36 0.714 0.390 0.290 0.589 0.846 0.754 0.751 0.974 37 0.712 0.389 0.290 0.586 0.846 0.751 0.750 0.985 38 0.697 0.376 0.283 0.574 0.840 0.743 0.744 0.990 39 0.631 0.371 0.277 0.566 0.756 0.734 0.737 0.990 40 0.568 0.368 0.275 0.561 0.673 0.724 0.730 0.984

Source: Own computation based on Census 2001 Note: Coloureds and Indians are presented in Appendix Table A1 The results suggest that better educated women have better marriages with very small differences between medium and high education. The coefficient for age is positive and significant but very small throughout the models implying that age is less likely to influence marital choice among women. Population group is positive and significant with Whites and Indians more likely to have good marriages. Employment status of the woman is negative and significant, although influencing the dependent variable weakly; entailing those women who are employed may most likely have the less appealing status. The geographical factors are both significant throughout while only influencing the outcome variable very slightly. Living in a metropolitan area decreases the likelihood of a woman to be married and in particular White women’s likelihood to be married. Moreover women with migrant husbands have better marriages. The effects of all the variables in (A) remain unchanged by the inclusion of the local marriage market variables in (B) to (E). Introduced separately, the local marriage market indicators have a higher positive effect on the type of marital status among women. Moreover the availability of men with sufficient income contributes the highest to civil marriages. However although the introduction of

21

Table 6: Ordered probit model of choice of martial status

All African White Explanatory variables Coef St.err Coef St.err Coef St.err Woman economic independence Education: Low 0.12 0.002 0.12 0.002 0.41 0.028 Medium 0.21 0.002 0.20 0.002 0.61 0.028 High 0.47 0.003 0.47 0.003 0.81 0.028 Employed -0.12 0.007 -0.11 0.008 -0.17 0.026 Monthly income: Low -0.23 0.003 -0.22 0.003 -0.34 0.018 Medium 0.16 0.013 0.23 0.016 -0.16 0.030 High -0.02 0.023 -0.02 0.026 -0.14 0.067 Employed*Low income 0.17 0.025 0.37 0.031 0.03 0.073 Employed*Medium income 0.14 0.015 0.16 0.019 0.09 0.040 Employed*High income 0.13 0.007 0.08 0.008 0.23 0.032 Background of the woman Coloured 0.59 0.003 Indian 1.35 0.007 White 1.33 0.005 Current age 0.05 0.000 0.05 0.000 0.05 0.001 Lives with parents -1.66 0.002 -1.74 0.002 -1.07 0.009 Spouse not in the household 1.20 0.002 1.24 0.002 0.38 0.015 Studying -0.24 0.003 -0.25 0.003 -0.28 0.013 Religion: Traditional 0.22 0.006 0.23 0.007 0.03 0.036 Judaism 0.02 0.024 0.00 0.037 0.03 0.034 Hinduism 0.27 0.009 0.17 0.029 -0.25 0.102 Islam 0.52 0.006 0.53 0.014 -0.20 0.050 No religion -0.12 0.002 -0.10 0.002 -0.37 0.010 Geography Metropolitan area -0.04 0.002 0.00 0.002 -0.21 0.008 Unemployment level of the area 0.00 0.000 0.00 0.000 0.00 0.000 Local marriage markets All men -0.23 0.011 -0.15 0.012 -0.36 0.080 Number of employed men 0.17 0.024 1.16 0.029 1.89 0.371 Number of educated men 0.11 0.006 0.17 0.007 -0.38 0.038 Number of men with sufficient income 0.65 0.019 -0.53 0.024 0.00 0.376 Parameters of the model Cut of point 1 1.47 0.011 1.75 0.014 0.83 0.076 Cut of point 2 1.96 0.011 2.25 0.014 1.33 0.076 Cut of point 3 2.51 0.011 2.91 0.014 1.42 0.076 Number of observations 4,565,785 3,721,535 283,158 Log likelihood -4,350,932.9 -3,669,870.1 -136,556.07LR Chi square 2,707,065.62 1,800,269.39 33,369.81 Likelihood ratio index 23.73% 19.70% 10.89%

Source: Own computation based on Census 2001 Note: Reference categories for the explanatory variables: Education-None; Employment status- Not employed; Monthly income-No income; Population group-African; Religion-Christian. all marriage market indicators simultaneously in the model (model (All) in Table 6) doesn’t change the model much, it has an effect on the sizes of the marriage market coefficients. The shortage of men contributes highly to the single or

22

cohabiting status of women. The availability of men with good income has a positive effect on the shift to marriage. A separate estimation for African and White women is also presented in Table 6 since marriage markets in South Africa are separate for the different population groups. Racial differences in the effects of some variables can be noted. The effects of education more than double up when comparing African and White women. Indeed the results for African women show that highly educated or employed- low income African women would most likely be married in civil marriage. Similarly highly educated or employed-high income White women would more likely be married in civil marriage. However the most striking difference is in the marriage market indicators. First all marriage market indicator variables have a significant contribution in the choice of marital type for both population groups. However the result suggests that African women decision to enter marriage is more dependent on the availability of employed men and to a lesser extent educated men. Yet the availability of employed men is the only marriage market factor influencing civil marriages among White women. Secondly mate availability as well as availability of men with sufficient income at the local marriage market level contribute negatively to civil marriages among both African and White women. On the other hand, availability of educated men contributes positively to civil marriages among African women and negatively among White women. This indicates that the shortage of marriageable men is not only a pure demographic effect of imbalances in the sex ratio but relates to the deficit of quality men. The results indicate that better educated women with more economic resources are more likely to be married, especially among Africans. This finding shows that marriage market factors explain more the racial effects rather than the woman economic independence factor, her social background and geographic factors. Obviously the issue of potential endogeneity arising from some of the explanatory variables (education, income level and employment status) cannot remain unnoticed since it can lead to biased estimates. In this case Greene (2003, p. 715) shows that the endogenous nature of the variables on the right-hand side of the equation can be ignored because it is the log likelihood that is maximized here whereas in linear models sample moments that do not converge to the necessary population parameters are manipulated.

6. Conclusion The recent decline in marriage rates in South Africa creates an opportunity to pose the question why marriage is becoming less common. Recent research in the determinants of marriage rates among women in the US has shown the decline in marriage rates to be due to poor marriage markets as measured by the shortage of “quality marriageable” men. The literature is rich in marriage market measures in quantity as well as quality of marriageable men. A comparison of plain sex ratios between African women and White women in South Africa shows that White

23

women have a better marriage market. The marriage market worsens as the women become more selective, the worst being the availability of African men with sufficient income. Using census data, this paper provides insights into the marital choices among women in South Africa taking into account the different types of marriage markets and the different population groups of the women. The study focuses on women aged 20 to 40 and who have given birth to at least one child. The women have the option between 4 types of marital choices that are assumed to be ordered from the least attractive to the most attractive as: never married, unmarried cohabitation, traditional marriage and civil marriage. The results are consistent with US findings that White women are more likely to marry than their African counterparts. This is due to the unavailability of marriageable African men. Also the estimation results suggest that both the quantity and quality of the marriage market matter in women’s decision of their marital choice. The options that women have outside of the marriage market such as employment and higher education make them search for a better educated and employed mate. On the other hand better education and good financial resources increase the chances of women to be married. References Angrist J., 2002, How do sex ratios affect marriage and labour market? Evidence from America’s second generation, The Quarterly Journal of Economics, Vol. 117, (3) pp 997-1038. Becker G., 1973, A theory of marriage: Part I, The Journal of Political Economy, Vol. 81, No. 4, pp 813-846. Becker G., 1986, 1991, A treatise on the family, Enlarged edition, Cambridge,

MA, Harvard University Press. Bergstrom T., 1994, On the economics of polygyny, University of Michigan. Bernstein H., 1985, For Their Triumphs of their Tears: Women in Apartheid

South Africa, International Defense and Aid Fund, London. Botticini M. and Siow A., 2000, Why dowries?, Boston University. Brien M., 1997, Racial differences in marriage and the role of marriage markets,

The Journal of human Resources, Vol. 32, No. 4 pp. 741-778. Ermisch J., 2003, An Economic Analysis of the Family, Princeton University Press, Princeton and Oxford. Fitzgerald J., 2003, Marriage prospect and welfare use, in Grossbard-Shechtman

S. ed., Marriage and the economy: theory and evidence from advanced industrial countries, 2003, Cambridge University Press.

Frazier F., 1939, The Negro Family in the United States, Chicago, The university of Chicago Press.

Greene W., 2003, Econometric Analysis, 5th edition, Prentice Hall International Edition.

Grossbard A., 1976, An economic analysis of polygyny: the case of Maiduguri, Current Anthropology, Vol. 17, No. 4, pp.701-707.

Grossbard A., 1980, The economics of polygamy, Research in Population Economics, Vol. 2, pp 321-50.

24

Gould E., Moav O. and Simhon A., 2003, The mystery of monogamy, The Hebrew University of Jerusalem, Discussion Paper, No. 9.03.

Gustafsson, S.S., 2001, Optimal Age at Motherhood: Theoretical and Empirical Considerations on Postponement of Maternity in Europe, Journal of Population Economics, 14(2001): 225-247. Human Sciences Research Council (HSRC), 2004, Fact sheet poverty in South

Africa, Fact Sheet No. 1, HSRC, Pretoria. Jencks C., 1992, Rethinking Social Policy: Race, Poverty and Underclass,

Cambridge: Havard University Press. Lichter D., Mclaughlin D., Kephart G., and Landry D., 1992, Race and the retreat

from marriage: a shortage of marriageable men? American Sociological Review, Vol. 57, No. 6, pp 781-799.

Mamashela M. and Xaba T., 2003, The practical implications and effects of the recognition of customary marriages act No. 120 of 1998, University of Natal research report No. 59.

Murray, C., 1984, Losing Ground: American Social Policy, 1950-1980, New York: Basic Books.

Kanazawa S. and Still M., 1999, Why monogamy? Social Forces, Vol. 78, No. 1, pp 25-50.

Kressel G., 1977, Bride-price reconsidered, Current Anthropology, Vol. 18, No. 3, pp 441-458.

Papps I., 1983, The role and determinants of bride-price: the case of a Palestinian village, Current Anthropology, Vol. 24, No.2 pp 203-215.

Statistics South Africa, 1998, The people of South Africa, the population census 1996, Pretoria, South Africa.

Statistics South Africa, 2000, Measuring poverty in South Africa, Pretoria, South Africa.

Tertilt M., 2002, The economics of bride price and dowry: a marriage market analysis, University of Minnesota.

Wilson W., 1987, The truly disadvantaged, Chicago University Press. Wood R., 1995, Marriage rates and marriageable men: a test of the Wilson

hypothesis, The Journal of Human Resources, Vol. 30, No. 1.

25

Appendix

Table A1: Availability of mate measures by population group and age

Coloured Indian

Age Sex ratio

Employed men

Men with sufficient income

High educated men

Sex ratio

Employed men

Men with sufficient income

High educated men

20 0.820 0.459 0.324 0.910 0.879 0.590 0.546 0.953 21 0.823 0.489 0.349 0.856 0.891 0.641 0.598 0.898 22 0.822 0.510 0.371 0.818 0.882 0.665 0.626 0.848 23 0.829 0.530 0.392 0.811 0.881 0.686 0.649 0.828 24 0.832 0.543 0.407 0.817 0.881 0.705 0.671 0.814 25 0.834 0.555 0.421 0.821 0.876 0.711 0.679 0.807 26 0.832 0.560 0.427 0.834 0.869 0.711 0.681 0.808 27 0.830 0.564 0.432 0.845 0.857 0.705 0.677 0.816 28 0.820 0.562 0.434 0.850 0.844 0.697 0.670 0.835 29 0.660 0.473 0.371 0.794 0.717 0.623 0.608 1.089 30 0.797 0.551 0.429 0.868 0.833 0.690 0.667 0.873 31 0.787 0.542 0.423 0.876 0.819 0.677 0.655 0.898 32 0.779 0.536 0.420 0.877 0.813 0.670 0.649 0.922 33 0.770 0.529 0.416 0.873 0.808 0.662 0.642 0.946 34 0.763 0.523 0.413 0.873 0.807 0.659 0.640 0.976 35 0.757 0.517 0.409 0.863 0.803 0.652 0.634 1.003 36 0.749 0.508 0.402 0.851 0.800 0.647 0.628 1.027 37 0.738 0.497 0.396 0.830 0.791 0.635 0.617 1.036 38 0.725 0.484 0.385 0.815 0.782 0.623 0.607 1.046 39 0.659 0.475 0.377 0.796 0.709 0.617 0.601 1.079 40 0.593 0.467 0.373 0.774 0.635 0.608 0.593 1.100

Source: Own computation based on Census 2001

26

Table A2: Ordered probit model of choice of martial status

(A) (B) (C) (D) (E) Explanatory variables Coef St.err Coef St.err Coef St.err Coef St.err Coef St.err Woman economic independence Education: Low 0.13 0.002 0.12 0.002 0.12 0.002 0.13 0.002 0.12 0.002 Medium 0.23 0.002 0.22 0.002 0.22 0.002 0.22 0.002 0.21 0.002 High 0.48 0.003 0.48 0.003 0.47 0.003 0.48 0.003 0.47 0.003 Employed -0.12 0.007 -0.12 0.007 -0.12 0.007 -0.12 0.007 -0.12 0.007 Monthly income: Low -0.24 0.003 -0.23 0.003 -0.23 0.003 -0.24 0.003 -0.23 0.003 Medium 0.17 0.013 0.16 0.013 0.16 0.013 0.16 0.013 0.16 0.013 High -0.02 0.023 -0.02 0.023 -0.02 0.023 -0.02 0.023 -0.02 0.023 Employed*Low income 0.16 0.025 0.17 0.025 0.17 0.025 0.16 0.025 0.17 0.025 Employed*Medium income 0.13 0.015 0.14 0.015 0.14 0.015 0.13 0.015 0.14 0.015 Employed*High income 0.13 0.007 0.13 0.007 0.13 0.007 0.13 0.007 0.13 0.007 Background of the woman Coloured 0.63 0.002 0.67 0.002 0.64 0.002 0.60 0.002 0.63 0.002 Indian 1.62 0.007 1.60 0.007 1.46 0.007 1.55 0.007 1.42 0.007 White 1.67 0.003 1.65 0.003 1.49 0.004 1.60 0.004 1.40 0.004 Current age 0.05 0.000 0.06 0.000 0.05 0.000 0.06 0.000 0.05 0.000 Lives with parents -1.66 0.002 -1.66 0.002 -1.66 0.002 -1.66 0.002 -1.66 0.002 Spouse not in the household 1.19 0.002 1.20 0.002 1.20 0.002 1.19 0.002 1.20 0.002 Studying -0.26 0.003 -0.25 0.003 -0.24 0.003 -0.26 0.003 -0.24 0.003 Religion: Traditional 0.23 0.006 0.23 0.006 0.23 0.006 0.22 0.006 0.23 0.006 Judaism 0.00 0.024 0.01 0.024 0.01 0.024 0.01 0.024 0.02 0.024 Hinduism 0.27 0.009 0.27 0.009 0.27 0.009 0.27 0.009 0.27 0.009 Islam 0.55 0.006 0.56 0.006 0.55 0.006 0.56 0.006 0.53 0.006 No religion -0.11 0.002 -0.11 0.002 -0.11 0.002 -0.12 0.002 -0.11 0.002 Geography Metropolitan area 0.01 0.001 -0.01 0.001 0.00 0.001 0.00 0.001 -0.04 0.001 Unemployment level of the area -0.01 0.000 0.00 0.000 0.00 0.000 -0.01 0.000 0.00 0.000 Local marriage markets All men 0.37 0.006 0.59 0.006 Number of employed men Number of educated men 0.23 0.006 Number of men with sufficient income 0.65 0.006 Parameters of the model Cut of point 1 1.22 0.005 1.71 0.009 1.67 0.007 1.49 0.008 1.49 0.006 Cut of point 2 1.71 0.005 2.20 0.009 2.17 0.007 1.98 0.008 1.99 0.006 Cut of point 3 2.27 0.005 2.75 0.009 2.72 0.007 2.53 0.008 2.54 0.006 Number of observations 4,565,849 4,565,849 4,565,849 4,565,785 4,565,849 Log likelihood -4,356,704.1 -4,354,423.7 -4,352,137.9 -4,355,788.5 -4,351,387 LR Chi square 2,695,689.08 2,700,249.74 2,704,821.41 2,697,354.55 2,706,323.29Likelihood ratio index 23.63% 23.67% 23.71% 23.64% 23.72%

Source: Own computation using Census 2001