1974 in Latvia, 15 percent in Estonia, and 10 percent in Lithuania, and an increase of 50 percent in Moldova for the previous year. However, for the total Soviet Union and especially for Russia, there is no jump at any one time, just a regular increase during the years 1971-1976. We could argue that in so large a country, new registration procedures would spread slowly, eventually resulting in a general improvement in registration. However, we do not have enough evidence to undertake any correction on this basis.
A second problem consists of age misreporting: infants who die at less than 1 year of age are reported as being older than 1; thus the infant mortality rate is underestimated, while the rate for ages 1-4 is overestimated (Ksenofontova, 1990, quoted by Anderson and Silver, 1990). This problem results in an obvious distortion of the mortality rates at these ages in several Central Asian republics, but it does not apply to the case of Russia.
A third problem with infant mortality data in the region is related to the definition of a live birth. Until very recently, the Russian definition was more restrictive than the World Health Organization (WHO) definition, and this probably resulted in an underregistration of births and early neonatal deaths (see also Kingkade and Arriaga, in this volume). Under the more restrictive Russian definition, children born before 28 weeks of gestation, or weighing less than 1,000 grams or being less than 35 centimeters long, were not supposed to be counted as either live births or infant deaths if they died before the end of their first 7 days of life.
The WHO definition was introduced only in January 1993 in Russia, but it has been applied since 1991 in Latvia and Lithuania and since 1992 in Estonia. We cannot yet verify the impact of this change on Russian statistics, but it is true that it changed the Baltic data rather dramatically. Following the change, infant mortality rates jumped from 13.7 to 17.4 per 1,000 in Latvia and from 10.3 to 16.5 in Lithuania. However, the increase observed for total infant mortality is not due entirely to the rise in mortality within the first week of life, which grew during these years from 6.2 to 9.0 per 1,000 in Latvia, and from 4.8 to 9.4 in Lithuania (Estonian Medical Statistics Bureau, Latvian Medical Statistics Bureau, and Lithuanian Statistics Bureau, 1993). The mean increase in total infant mortality for the two countries is 5.0 points, of which only 3.7 points is associated with deaths during the first week of life, representing the maximum amount of change due to the new rules of registration of early neonatal mortality; the remaining difference is probably attributable to some other problem recently experienced by these countries. Let us say that the real correction due to the new definition is about 3.0 points, which represents an increase of about 50 percent.
Assuming that the impact of the adoption of the WHO definition would be the same for Russia, we can assume a similar increase of about 50 percent. In the 1990s, with mortality during the first week of life accounting for about 50 percent of total Russian infant mortality, the latter would then increase by 25 percent.