Spatial variability in fertility in Menoufia, Egypt, assessed through the application of remote-sensing and GIS technologies

Fertility in rural areas such as the Governorate of Menoufia in Egypt may be influenced both by spatial factors (including the diffusion of innovations) and by essentially nonspatial factors (such as the availabilityof education for women and the percentage of adult women who are currently married). The nonspatial variables are available directly from censuses but the spatial component requires an accurate location of the villages to which the census data refer and then appropriate decomposition of the data into spatial and nonspatial components. We use IRS satellite imagery to classify the built area in a rural governorate in Egypt and then assign village-level census data to the centroids of those polygons and incorporate the data into a GIS. We then employ measures of global and local spatial statistics to conclude that in 1976 the combination of female illiteracy, proportion married, and spatial clustering accounted for 39% of the variation in fertility in Menoufia. In 1986 those same factors explained 51% of the variation in fertility. In 1976 about one third and in 1986 about half of the explained variability was due to the spatial component ('diffusion') and the other half due to a combination of demographic characteristics. Furthermore, between 1976 and 1986there was a clear north-to-south drift of fertility, with lower fertility being clustered in the north and higher fertility clustered in the south. Introduction In most parts of the world the rural population is concentrated in small but relatively densely settled villages. People do not necessarily live on the land they farm, but rather in a village that represents a reasonably efficient use of the often limited rural resources. This is important from a demographic perspective because it means that the village may provide sources of human capital such as a school or employment opportunities, and it also provides a context within which information is shared that may influence decisions that individuals, couples, or households make about health care, migration, and reproduction. The type and extent of information that is shared may come from mass media such as radio and television but it may also be shared amongst villages because most rural areas in agricultural societies are part of a net­ work of villages. The demographic behavior of villagers is of considerable importance because it is in the rural areas that population growth is highest throughout the world, contributing to redundancy within the rural population itself and encouraging migra­ tion to already overburdened urban areas (for example, see Jones, 1990; Lutz, 1994; Weeks, 1999). An improved understanding of the fertility transitions in developing countries is critical if culturally appropriate and effective reproductive health and family planning policies are to be designed and implemented to improve the welfare of the population. There is now ample evidence that the traditional models of the demographic transition as developed by Thompson (1929), Notestein (1945), and Davis (1945) were overly 696 J R Weeks, M Saad Gadalla, TRashed, J Stanforth, A G Hill simplistic and incomplete. Reformulations began with Davis's theory of demographic change and response (Davis, 1963) but have been driven especially by results of the European Fertility Project at Princeton (Coale, 1973; Watkins, 1991) and by analyses of data from the World Fertility Surveys (Cleland and Wilson, 1987). The models have also been respecified by the disaggregation of the demographic transition into its constituent components of the epidemiological transition (Omran, 1971), the fertility transition (for example, see Chesnais, 1992; McDonald, 1993), and the migration (especially urban) transition (Firebaugh, 1979), with increasing thought being given to the specification of a nuptiality transition (Chesnais, 1992) and/or a more general household transformation or transition. This respecification, which hearkens back to Davis's theory of demographic change and response, allows us to examine each element of the transition. In so doing, the evidence seems to suggest that the fertility transition is neither a simple response to rising individual standards of living nor an automatic response to declining mortality. The picture appears to be more complex in several ways (Reed et al, 1999) and it seems likely that the fertility transition is best understood as a 'blend' of structural factors (exemplified by the supply-demand framework; Easterlin and Crimmins, 1985), diffusion factors (Cleland and Wilson, 1987; Kirk, 1996; Knodel and van de Walle, 1979), and the local context in which reproductive decisions are actually made (Entwisle et al, 1989). In this research we are interested in the extent to which the variation in fertility from one rural village (the local context) to another may be explained by a process of diffusion of behavior from some villages to others, net of the human capital variables such as education that may exist within the village. We lack direct evidence of such spatial diffusion but can infer it from the spatial and temporal patterning of reproduc­ tive behavior. "The diffusionist perspective predicts that if Area A is proximate to areas with relatively low fertility, and Area B is surrounded by areas with high fertility, Area A will have lower fertility than Area BOO (Tolnay, 1995, page 301). The spatial dimension of demographic behavior must thus be measured so that these variables can be evaluated for the role that diffusion may play in fertility behavior. We must first show that proximity matters, and then show that changes occurred over time in a sequence that is consistent with spread or diffusion. Our goal in this paper is to show that proximity matters when it comes to fertility levels, even in a rural area of a developing country. Thus we assess the extent to which the spatial variability in fertility differs from that which can be expected to happen by chance by testing the null hypothesis that fertility is spatially independent. The study site We illustrate this procedure by using data for a rural governorate of Egypt (Menoufia) for 1976 and 1986. Menoufia is one of the twenty-six governorates that form the administrative regions of Egypt, roughly equivalent to states in the United States, although perhaps more analogous to counties in the United Kingdom. Figure 1 illus­ trates the study site. In 1976 the total enumerated population of Menoufia was 1.88 million, accounting for about 5% of the total population (37.7 million) of Egypt that year. By 1986 the population of Menoufia had grown to 2.23 million, representing 4.4% of the total Egyptian population of 51 million in that year. Between these two years, which represent the years of analysis for this study, the population of Menoufia was growing at an annual rate of 1.7%. Despite that speed of growth, Menoufia lost ground (in terms of relative population size) to the even more rapidly growing gover­ norates in upper Egypt. Fertility in Menoufia was high in 1976, and it was still high in 1986, although there has reportedly been some fertility decline since 1986 (the 1996 census data at the local Figure 1. The study site of Menoufia Governorate, Egypt. 697 Spatial variability in fertility in Menoufia, Egypt level had not yet been released by the Egyptian government at the time that this paper was written). For decades Menoufia has been one of the most rural and most densely populated rural areas of Egypt (Gadalla, 1978). It has been, and remains, predominantly agricultural, and the high rate of population growth has increased the redundancy of the rural labor force and encouraged out-migration, to Cairo or to other Arab (especially oil-producing) nations. We chose Menoufia as a study site less for its specific demographic characteristics than for the fact that it has been relatively well studied in nonspatial analyses and thus there are comparative studies by which to judge the spatial analysis produced by our own work. Menoufia does have some advantages for spatial analysis including its essentially flat landscape in the Delta region of the Nile, which means that elevation is not an issue that needs to be dealt with. Partly for this reason, there are numerous villages (312), with an average population per village in 1976 of 5566 and in 1986 of 7175. Data Assuming that we find a spatial component to the variability in fertility, we seek to decompose the contribution to fertility made by sociodemographic and human capital variables such as marital status and education from those contributions made by the clustering of villages with similar or shared characteristics. The first set of factors, which in theory have no spatial component, are part of the broader supply demand framework of the fertility transition (for a review, see Weeks, 1999), whereas the spatial component is more clearly a part of the cultural-diffusionist perspective on the fertility transition. We hypothesize that both sets of factors are important and aim to quantify the importance of each. 698 J R Weeks, M Saad Gadalla, TRashed, J Stanforth, A G Hill