FIGURE 5-2 Sex and age pattern of mortality in Cambodia, ten-year probability of dying, 1970-1979. Source: Heuveline (1998a).

age patterns of mortality and the methods built thereon should still be applicable. Model life tables based on data from medium-to-low mortality populations do not seem to provide a very good extrapolation of mortality patterns under higher mortality conditions (Bhat, 1987), however, and the analyst should preferably turn to model age patterns specifically designed for high-mortality populations (Preston et al., 1993).

On the contrary, when violent mortality becomes an important part of overall mortality, the mortality pattern may have little in common with the nearly universal J-pattern and model age patterns cannot be used. Figure 5-2, for instance, presents the age pattern of excess mortality estimated for Cambodian males and females during the 1970s. These age patterns of mortality, especially the male one, bear no resemblance with any model age pattern. This finding is not particularly surprising since wartime life tables are typically excluded from the empirical basis on which the model age patterns are constructed (e.g., Coale and Demeny, 1983). The reason is precisely that the age pattern of war-related deaths differs from that of natural mortality and reflects idiosyncratic conditions at a particular time and place that should not be generalized to other populations.

Existing model age patterns of mortality can not be used to assess and improve the quality of such an unusual age pattern of mortality as shown in Figure 5-2 . But the unusual aspect of this age pattern is also useful because the departure from the original pattern is caused by the rise of a few formerly rare causes of death, typically associated with the mortality