Abstract
A technique for indirect estimation of mortality originally devised by Courbage and Fargues, was modified and tested. This modification was made, in order to check the fitness of model mortalities for the object population and to find out the best fit model mortality from various families of model life tables. The principle of this modification is very simple. And it does not require any additional assumption. All that is needed is the evalution of the applicability of the original assumption that the rate of under-registration of deaths is nearly constant over certain age. The estimation technique was applied to national data from 35 developing countries for the year around 1975. Nine families of model life tables constructed by Coale & Demeny and the United Nations were applied to each of the object population, and the level of mortality corresponding to the value of PMI (proportional mortality indicator: percentage of deaths at the ages 50 years and over) obtained from death registration data, was determined for each family. By applying these mortality models, expected deaths for the object population were calculated. Then, age-specific death registration rates (R_<(a)>), were derived for each mortality model by simply taking the ratios of registered deaths to expected deaths. And the coefficient of variation (C. V.) of R_<(a)> for ten age groups were calculated. Finally, the level and the pattern of model mortality which gave the least value of C. V. was selected as the best fit model for the object population. The estimated levels of mortality thus obtained as well as the levels obtained for the WEST model were checked in comparison with the independent estimates of the United Nations and the direct estimates. The relations between those estimates and TFR, per capita GNP and literacy rates were also investigated. In a few case, did the level of our estimate caused serious doubt, but in general, they seemed to be in good accord with the other information. Moreover, the geographical pattern of distribution of selected model life table families apppeared to be quite reasonable. The technique is unique in the sense that it does not require any assumption regarding the feature of an object population, and is directly applicable to a non-stable population or even to an open population without any adjustment.
