Multi-state projection models
The 'states' in the multi-state projection model that has been discussed up to this point are regions. They could instead be any other exhaustive set of mutually-exclusive categories into which the total population can be divided. For example, in a marital status projection, the set of states into which the population is divided is single, married and divorced and widowed: every individual belongs to one of these statuses and nobody can legally hold two or more of them at the same time. In this model, death is referred to as an ‘absorbing state’ as there are no flows from it back into the other states, while the various marital statuses are referred to as transient states.
Multistate representation of the population according to marital status
As well as being of interest in its own right, projection of the population by marital status may improve the prediction of future births if a large proportion of women of childbearing age are unmarried and fertility varies greatly by marital status.
Similarly, the International Institute for Applied Systems Analysis in Austria has devoted considerable resources to producing global population projections by educational attainment using multi-state methods (Lutz and Samir KC 2010, Samir KC et al. 2010). Apart from the intrinsic importance of education for global development, the main justification for this is that educational attainment has a major influence on both fertility and mortality and, therefore, disaggregating the population in this way should produce more accurate forecasts.
One difficulty with marital status projections that does not arise with regional or educational projections is that the model should be implemented in a way that maintains consistency between the forecasts for the men and women. For example, in a population in which marriage is always monogamous, the same number of men and women should marry each year. Equally, the same number of men and women should divorce. Moreover, the number of deaths of married men should equal the number of newly widowed women, and vice versa.