A study of the reliability of fetal death statistics in Japan during the first half of the 20^<th> century

Abstract

Historical demographers generally assume that death statistics collected since the end of 19^<th> century are reliable, given that the death-registration system was well established by the 1890s. In this paper, I challenge this assumption by calculating the fetal death statistics in the early 20^<th> century. Prior to the establishment of the death-registration system, the number of unregistered fetal deaths must have decreased before the number of neonatal deaths incorrectly registered as fetal deaths decreased. During this time lag, the number of registered neonatal deaths should have increased. Therefore, the aforementioned assumption can be confirmed by data reflecting these phenomena. First, based on an analysis of fetal and neonatal death rates from 1899 to 1942,I hypothesised that the number of unregistered fetal deaths decreased to the level at which we could analyse annual trends in fetal death rates only after 1910 rather than in the 1890s. Second, to test the above hypothesis, I calculated the number of neonatal deaths that had been registered as fetal deaths since the 1910s and estimated that unregistered neonatal deaths accounted for 16-20% of the registered fetal deaths that occurred in the early 1910s, 11-15% of those that occurred in the late 1910s, 6-10% of those that occurred in the early 1920s, and around 5% of those that occurred in the mid-1920s.