Life tables calculated in the classic way on the basis of annual death statistics and age-specific population estimates as of 1 January of each year become unreliable at the oldest ages. Due to the rapid decline of mortality among the oldest old observed in the past few decades and the even swifter increase in the proportion of nonagenarians and centenarians in industrialized countries, it has become more important than ever to improve the accuracy of the measurement of mortality at the oldest ages.
The extinct generation method enables us to do this while using only real data. The method was adapted to be used with cohorts that are not completely extinct, thus making it possible to reconstruct recent mortality after age 90 with an acceptable degree of accuracy.
The trends in life expectancy at 90, 95 or 100 can thus be studied with greater accuracy from the start of the twentieth century. The comparison between the results obtained through traditional methods and those obtained through the extinct generation method shows that life expectancy at those ages was previously overestimated, thus masking its spectacular increase during the last three decades.