Estimativas indiretas de expectativa de vida em idades avançadas no Brasil e suas regiões

Abstract

With the increase in survival levels at older ages, it becomes increasingly important to know, evaluate and solve problems in the measurement of adult mortality. Mortality functions are mainly affected by two types of errors: underreporting of deaths and errors in the declaration of age. In Brazil, we still know little about the biases caused by age declaration errors, which tend to overestimate survival levels at older ages. Since this type of error affects data sources differently, the objective of this dissertation is to examine the consistency of death and population data in Brazil and its regions, as well as to test alternatives for adjusting the distribution of deaths by age, in order to estimate the impact of age misstatement on life expectancy in adulthood. To achieve these objectives, different methodologies were used. Initially, the levels and patterns of mortality calculated based exclusively on the population distribution by age (Preston, Bennett 1983) and the distribution of deaths by age (Bennett, Horiuchi 1981) were compared. Then, through the methodology of intercensus cohorts, an attempt was made to assess the level of inconsistency between the two sources in different age groups. Finally, correction matrices of death certificates prepared with data from two other countries were used – Costa-Rica (Dechter, Preston, 1991) based on the study by Ortega and García (1986) and the United States (Preston et al, 1998) – to redistribute deaths from the Midwest, Northeast and North regions, regions with higher levels of inconsistencies. The results indicated relatively lower mortality levels in older ages, whatever the source used, suggesting the presence of errors in the declaration of age. Furthermore, errors in death and population data seem to have different magnitude and direction, which ends up causing the underestimation of conventional mortality rates at older ages. Corrections made from data from two other countries resulted in life expectancies at adulthood lower than directly estimated for both men and women. Although it is difficult to generate specific correction matrices for Brazil, due to the lack of information necessary for its construction, it is important that new studies in this area are conducted.