Adult and old age mortality in Latin America: Evaluation, adjustments and a debate over a distinct pattern

Until recent years the study of adult and old age mortality in developing countries has been largely neglected. Particularly in Latin America, a systematic evaluation of the available data and a uniform methodology to adjust registered data have been lacking. A distinct Latin American age pattern of mortality was reflected in the United Nations Model Life Tables which showed that, at a given level of life expectancy at birth, mortality at advanced ages was the lowest among the world\u27s major regions. Competing explanations have been provided to account for the emerging crossovers : a biological phenomenon, commonly referred to as survival of the fittest , or an artifact of bad data. In this dissertation I estimate mortality rates by age and sex based on two censuses and the number of deaths during that interval. Data are reasonably consistent only up to a certain age. At more advanced ages, data are unsuitable for the application of traditional methods. Thus, I develop a strategy consisting of two steps: (1) to identify an age above which data become inconsistent, and (2) to estimate adjusted age-specific mortality rates based on the crude death rate above that age, age-specific growth rates and a new relational model life table system. Although I present evidence that inaccurate or incomplete data produce widespread underestimation of mortality at older ages, the effect of these errors on life expectancy tends to be modest. Simulations of different measurement errors and/or assumptions show that our results are most sensitive to the crude death rate for the open-ended age interval, while the level of growth and the age pattern of mortality play a minor role in determining the overall mortality level. Latin American countries exhibit a distinct age pattern of adult mortality with very low values for the slope of the age-specific death rates when compared to those of more developed countries. Furthermore, such a pattern is determined at the young adult ages, casting serious doubts on the idea that it is simply an artifact of bad data