Key features of reproductive behavior in China vary systematically through
space and time. In this article we present an analysis of fertility change in
regional space, using a 1% household sample from China’s 1990 population
census. Elsewhere, we use the same data to analyze reproductive strategizing,
but here we pursue the big picture with a straightforward analysis that takes
reported births as an uncomplicated indicator of fertility.The article has three
objectives: first, to introduce a novel, multilevel spatial model of regionastructure constructed using a geographic information system (GIS); second,
to demonstrate the potential for longitudinal data derived from onetime censuses
to contribute to historical demography in conjunction with regional
analysis; and third, to document the manner in which China’s fertility transition
has unfolded in regional space.We argue that our spatial model, Hierarchical
Regional Space (HRS), effectively captures the spatial structures of
change in socioeconomic status, in family system norms, in the state’s birthplanning
policy and enforcement, and in access to high-tech sex-selective
abortion, which help to explain the observed patterns of demographic transition.
Our analysis begins with a data file consisting of over 12 million records
comprising a 1% sample of China’s 1990 population census. Access to this
data file has been made possible in collaboration with China’s State Statistical
Bureau and the Beijing Institute of Information and Control.The 1% sample
was selected from a master file listing some 1.2 billion returns by geographically
ordered identification number. The full census returns of all persons in
every 100th household are included in the sample. The returns offer a wealth
of data on demographic status, education, and occupation. We manage this
sample of census data, which we refer to as the ChinaS data file, with SAS
statistical software (SAS Institute, Cary, North Carolina).
In constructing our spatial model, we turn to two additional data files
containing data on China’s 2,800 county-level administrative units and over
12,000 cities and towns.These two files, dubbed ChinaA and ChinaT respectively,
are linked to digital maps managed with ArcInfo GIS software (Environmental
Systems Research Institute, Redlands, California). Additional
GIS files describing China’s physiography, hydrography, and transportation
network are also used in our spatial analysis (Henderson et al. 1999). The
Australian Centre for the Asian Spatial Information and Analysis Network
(ASIAN) at Griffith University in Brisbane had primary responsibility for
the development of the spatial data files (Crissman 1997), which have been
compiled at a nominal scale of 1:1,000,000 for an estimated spatial accuracy
of +/-650 meters. By making use of a standardized (albeit undocumented)
spatial coding system embedded in the census record identification numbers,
we have been able to link ChinaS records to the county and settlement levels,
corresponding to our ChinaA and ChinaT data files, and using this linkage
we are able to position sampled households within the HRS model.