California Center for Population Research
State Views and Local Views of Population: Linking and Comparing Genealogical and Household Register Data in Liaoning, 1749-1909
- Author(s): Campbell, Cameron
- Lee, James Z
- et al.
Most micro-level quantitative studies of Chinese society rely on individual level data from locally produced patrilineal genealogies or from state generated household registers. Stevan Harrell, Ts’ui-jung Liu, Ted Telford, and Zhongwei Zhao and their associates use genealogies to estimate trends in demographic rates over the very long term, especially for southeastern, south central, and southern China. We and Arthur Wolf and his and our collaborators use Qing dynastic (1640-1911) and Japanese colonial (1905-1945) household registers from Liaoning and Taiwan Provinces in northeast and southeast China respectively to examine associations between individual and household characteristics on the one hand and demographic behavior on the other, largely during the last two centuries.
Several studies have identified shortcomings in these sources from internal evidence, comparison with predictions from demographic models, and micro-simulations. Lineage genealogies rarely record wives and daughters, and usually omit sons who die without male offspring, especially if they die during infancy or childhood in the remote past (Harrell 1987; Telford 1990). They rarely provide data on marriage timing and often do not even record any vital data at all (Harrell 2003). Household registers appear to record most wives, but not necessarily most daughters and may even miss some sons. While recording of vital events tends to be complete, some registers appear to underrecord mortality especially among very elderly males (Lee and Campbell 1997). Most recently, micro-simulation revealed that the selectivity against extinct patrilines could distort estimates of demographic rates from genealogies, biasing mortality estimates downward and fertility estimates upward (Zhao 2001). In a caveat, Zhao (2001) noted that other forms of selection could further bias estimates from genealogies, but could not be accounted for in his simulation model.