A multilevel modelling approach to the determinants of urban and rural fertility in Bangladesh

When a data set is generated from a country-wide survey, it is important to test it to determine whether or not fertility variations are due to the structure of the data. Generally, small area data are more or less homogeneous in nature. This observation is supported by Garner and Diamond (1988) who pointed out that individuals who share similar characteristics related to fertility behaviour, such as living in the same geographical area, are likely to act in a similar fashion. There may be fertility variations between small areas in Bangladesh, but these have been generally ignored in previous studies (see, for example, Islam and Khan, 1991, 1995; Khan and Raeside, 1994; Amin and others, 1995).