A key variable affecting the public and private health care systems of the U.S. and other countries is the current trend in old-age mortality. At all ages below 100 years, and perhaps at higher ages as well, the chance of dying in any one year of life has fallen in this century. As a result, life expectancy at birth continues to rise. A continuation of this fortunate trend depends primarily on prospects for a sustained decline in mortality rates at very old ages. Will the trend toward higher life expectancy continue in the next century? Clearly, no one can predict the future with certainty. Rather, this project seeks to answer a more limited set of questions: How rapidly has the improvement in old-age mortality taken place? What age ranges have been affected? Have mortality rates fallen at extremely high ages, say above 100 or 105 years? Is there any reason to believe, based on trends in mortality rates at extreme ages, that the upward trend in life expectancy may eventually encounter a biological upper limit? A large portion of the current uncertainty with regard to the real trends in old-age mortality is due to problems of data quality, particularly the problem of age misreporting. A major component of this project will address this issue. Which countries or states have relatively more accurate mortality data? What techniques are available for identifying and perhaps correcting inconsistencies in mortality data? How do inaccuracies in the raw data affect the estimation of mortality trends? Based on data from Japan, France, England and Wales, Sweden, and the United States, this project will concentrate on the accurate measurement of mortality rates among the oldest old. The main purpose is to document the changes in old-age mortality that have taken place over the last century, keeping in mind the theoretical issues and policy implications that motivate an interest in the trends themselves. Unique or innovative features of this project include: a systematic evaluation of data quality and its effects on the estimation of old-age mortality rates; a thorough analysis of techniques for extending the method of extinct generations to non-extinct cohorts; an application of the statistical theory of extreme values to the study of data quality; a thorough documentation of both period and cohort trends in old-age mortality for five countries; a critical investigation of indirect methods for estimating mortality rates at very old ages, in particular the Coale-Kisker procedure; and, finally, calculation of old-age mortality rates for 50 states in the U.S., with particular attention given to the comparative quality of mortality data by state.