regions of the country. Next, the chapter compares the Russian mortality patterns with Coale-Demeny model life tables, other standard mortality patterns, and other European and U.S. patterns. The chapter ends with a summary and conclusions.
The basic data for administrative units of Russia used for this analysis are numbers of deaths in 1988 and 1989 in Russia by age, sex, cause of death, urban or rural location, and administrative unit. The data are from statistical reports of the State Committee of Statistics of the former Soviet Union for provinces, special districts, and autonomous republics (respectively, oblast, krai, and autonomous republics, hereafter referred to as provinces). Population data are from the 1989 census. For the analysis, the entire data set consists of 292 observations: for each sex, 146 observations cover the urban population of 73 provinces, including Moscow and Leningrad cities; the rural population of 71 provinces; and the total urban and rural population of all of Russia.
For the analysis of spatial variation in mortality levels, we use life expectancies at birth and cause-specific death rates for three causes of death: injuries, cardiovascular disease, and neoplasm. Death rates are standardized by age to the European standard (Waterhouse et al., 1976) for each of four subpopulations: male urban and rural, and female urban and rural (Table 3-1). The underlying provincial life expectancies and age-standardized cause-specific death rates for three causes of death are included in Annex 3- 1. Percentiles for the life expectancies and cause-specific death rates were calculated and quintiles assigned. Quintiles were used to allow comparisons within a province of the rankings in different causes of death. The lowest quintile, 1, represents a situation of low mortality, while the highest, 5, represents high mortality. Quintiles of life expectancy and cause-specific death rates are also given in the annex.
For the analysis of age patterns of mortality, 2-year multiple decrement life tables for 1988-1989 were calculated. These life tables are based on the above data and were constructed by Chiang's (1978) method. 1Thus, the data set consists of 292 life tables, one for each sex and administrative unit. We examine variation in the age-sex profiles by constructing typical profiles through clustering of the provincial mortality profiles. We use a formal approach based on a generalized concept of profile structure developed a number of years ago (Cronbach and Gleser, 1953), which allows the use of cluster analysis to find mortality curves with identical shape (Wunsch, 1984). This approach is discussed more thoroughly in a later section of the chapter.
Given the data quality issues addressed by many of the authors in this volume, a word about the reliability of the data used for this analysis is in order. Certainly, provincial-level data are subject to greater error than national estimates. We note some of these problems with certain provinces in the course of