2. Overview of factors which may influence divorce 4. Results

3. Methods

The data of this study came from five villages of Teknaf where ICDDR,B maintained a demographic surveillance system (DSS) during 1976-89. It is at the south-eastern tip of Bangladesh bordering Myanmar. People are mostly illiterate, conservative Muslim. Men are employed in agriculture and fishing in the nearby Naf river and the Bay of Bengal while women are engaged in household chores. Teknaf has higher fertility and infant mortality than Matlab [Rahman et al. 1985-1986, Shaikh et al. 1984 and 1985].

Teknaf-DSS recorded vital events, migrations, marriages and marital dissolution by means of monthly household visits. Marriage refers to a union, which is established by civil or religious procedures while divorce is dissolution of marital union through civil, religious or social procedures. Most marriages and divorces are not registered with the government civil registration system. A total of 1762 marriages were recorded by Teknaf-DSS during 1982-83; they were followed up in the period 1982-1988 to determine changes in marital status as well as out-migration and death of one or both of the spouses.

The marriage form contained information on marriage related events as well as on previous marital status of the groom and the bride and their ages at current marriage, years of schooling, occupations and kinship. Polygyny was determined by considering information on the groom's marital status at the time of the current marriage (i.e., in 1982-83). The groom's prior marital status was categorized into never married, divorced, widowed or polygynous (already married and cohabiting with one or more wives in addition to current wife). The bride's marital status prior to the current marriage was categorized into never married, divorced or widowed.

Information on live births born to the 1982-83 marriage cohort during the first five years of marriage was obtained by matching marriage records with birth records of 1982-88. Newly married women were categorized into a group which had a live birth in the preceding six months and another group which had not.

The Teknaf Household Socioeconomic Census in 1982 provided information on socioeconomic status. Ownership of land was coded 1 if the family owned more than 0.30 acre of land and 0 if they had less. Ownership of fishing boat, fishing net, betel leaf garden, betel nut garden or receipt of remittances from abroad by groom's family were coded 1, otherwise it was coded as 0. An index of the number of economic assets owned by the groom's family was used to determine household economic status. Groom's years of schooling was grouped into none or some. Since nearly all brides were illiterate, (4.9 per cent of brides had some schooling), bride's education was not included in the analysis.

Data Analysis: During the five-year follow-up period of 1762 Muslim marriages recorded in 1982-83, 40 marriages were disrupted due to death of one spouse, and another 28 marriages due to emigration of both partners. Marriages, which were lost to the follow-up were excluded from the analysis subsequent to their emigration or death of spouse. Marriage duration was determined in segments of six months. The crude effects on risk of divorce of the predisposing variables (duration of marriage, bride's and groom's previous marital status, groom's education and household wealth, bride's fertility status and kinship between partners) were estimated by calculating divorce rates per 100 marriage-years (or per 200 six-months segments).

The net effects of the independent variables on the risk of divorce of marriages were estimated using a discrete-time hazard logistic model. All the variables except groom's age at marriage, were made binary, and so was the dependent variable - divorced/intact marriage. The initial logistic model contained all the variables and the interactions between duration of marriage and spouses' previous marital status, education and household assets. Interactions of insignificant coefficients were eliminated one by one if the difference in likelihood ratio statistics between the two nested models (the latter is the subset of the former) was not significant (chi-square<0.1). The final logistic model contained all the variables and the significant interactions only.


2. Overview of factors which may influence divorce 4. Results

Determinants of Divorce in a Traditional Muslim Community in Bangladesh
Nurul Alam, Sajal K. Saha, Jeroen K. van Ginneken
© 2000 Max-Planck-Gesellschaft ISSN 1435-9871