Relational model life tables

The relational model life table system in a nutshell

Problem: We have some limited measures of mortality for a particular population.  How can we turn these few measures into a full life table so that we can perform wider demographic analysis of that population?

A completely different approach to deriving model mortality patterns was derived by William Brass in the 1970s. Brass took the view that all human mortality patterns were of essentially the same shape – the shape described in Section 2, Page 2 of this session.

Surveys and censuses can provide mortality estimates, for nations or regions, that are not available from vital registration.  But these estimates are often fragmentary and may themselves be imprecise.  By fragmentary we mean that one survey may provide an estimate of infant and under-5 mortality and another may provide some estimates for mortality at say age 40 or 50, and that's all we have.  How can we produce an abridged life table from birth to age 85 from that fragmentary data?

Brass's concept was that we can fill in the gaps by relating these estimates to a suitable standard life table in a novel but relatively simple way.  The relationship generates two parameters that define the way in which the standard is different from the observed values.  These two parameters allow us to modify or adjust the standard, to produce a new mortality schedule which we accept as the new “fitted” life table that is our best estimate of a life table that fits the observed data whilst keeping a coherent life table pattern.

Note: This session deals predominantly with the classic Brass Relational Model, which is still widely used. There have been several variations or extensions to this model. These are beyond the scope of this session but are mentioned in Sections 14 and 15.