Several authors have proposed different methods for splitting a difference between two life expectancies into its components by age and cause of death (Andreev. 1982: Pollard. 1982, 1990; Pressat, 1985; Valkovics, 1984).

France and England were chosen because reconstituted series of death by cause were available for both countries. In France, the 1925-1991 series are available, with deaths classified according to the detailed list of the ICD-9 (Meslé and Vallin, 1993b). We first reclassified the deaths into the Soviet classification and then applied the same groupings as for Russia. For England, the exercise was a little more hazardous because continuous series were available only for deaths classified according to the Basic Tabulation List of ICD-9 (Meslé and Vallin, 1993a), much less precise than the detailed one. Because of no obvious correspondence between the Soviet and ICD items, it was not possible to reclassify these data into the Soviet classification. We therefore directly grouped the English deaths into categories with the same medical definition as those defined for Russia and France.

The sharp decrease in "respiratory diseases" observed in England in 1984 is due to a change in the interpretation of the WHO rules for the identification of underlying cause. Until 1984, in many cases, the immediate cause was wrongly considered the underlying one. For this reason, items such as "pneumonia" or "pulmonary congestion" were used instead of the item corresponding to the underlying cause. The change resulted in a more or less important symmetrical increase in almost every other item of the classification (this phenomenon, for example, appears clearly for cerebrovascular diseases).