This article deals with a long-standing unsolved issue in Japanese historical demography. This is the issue of birth underregistration in shumon aratame-cho (SAC) which has been the main data source in the field for the last thirty years or so. The issue of the birth underregistration prevents researchers from accurately estimating infant mortality as well as fertility, and makes it very difficult to compare these indices with those of other societies. To solve this problem has become an urgent issue in Japanese historical demography not only because infant mortality is one of the best indices to know the standard of living of society, but also because fertility recently has become the most active area of research in historical demography in Europe and the U.S. To solve the problem of birth underregistration in the SAC, this article treats infant mortality and fertility separately. The first part of the article discusses a method to estimate infant mortality from the SAC. By using data of the early vital statistics of the Japanese government as well as information from the SAC, I tried to find out an empirical relationship between infant mortality and SAC first-year-death-rates (SFDRs), the only index related to infant death which can be calculated from the SAC, and used this relationship to estimate infant mortality. An application of this method to the SAC of a village in Northeastern Japan yielded a reasonable result for infant mortality of Tokugawa peasants, and thus proved its validity. This method can easily be applied to other regions of Japan since the method is reasonably straightforward and does not require much computation. The second part of the article examines the level of birth underregistration in the SAC through microsimulation. The main portion of the model used in the simulation is based primarily on the studies by Bongaarts and Potter, and those by the Institute of Population Problems, Ministry of Public Health, but I modified the model to handle the problem more properly. The simulation result reveals that the level of the birth underregistration is positively associated with infant mortality; and with the level of infant mortality in Northeastern Japan during the Tokugawa period (in the neighborhood of 180 to 230 per thousand), the level of the birth underregistration is likely to have ranged between 14 and 18 per cent of the actual number of births. It is well known that infant mortality of peasants in Northeastern Japan was generally higher than in Southwestern Japan. If we take 150 to 230 per thousand as a reasonable estimate for infant mortality of Tokugawa Japan as a whole, the level will range between 12 and 18 per cent. Recent studies in Japanese historical demography often assume the level of birth underregistration in the SAC to be 20 per cent. But the simulation result indicates this assumption is somewhat overestimated. Some demographers caution us that the level of birth underegistration in the SAC may vary according to SAC compilation dates. This is because Tokugawa Japan, or pre-industrial society in general, is characterized by having the clear seasonality of births. The microsimulation is used to examine whether the difference in SAC compilation dates could have any serious effect on the level of the birth underregistration. The answer is negative: the effect is only minor; the difference in SAC compilation dates can hardly make any difference in estimated fertility in a measurable way. Finally, by using the microsimulation, the article experimentally examines variances in fertility of small cohorts of different sizes. Since historical demographers are not always as fortunate as to be able to find 100-year unbroken records of SAC, but rather they often encounter SAC of a much shorter period, the result of the simulation gives a measure of variance in fertility of a small population and helps us deal with fragmentary SAC.
View full abstract