Demography experts in NZ
A SPEaR-funded workshop in December on demographic modelling and projections at Waikato University's Population Studies Centre updated academics and the policy community on state-of-the-art developments overseas. A follow-up seminar was held in Wellington to a wider audience of officials. Professor Ian Pool reports.
Demography is a very highly applied subject; it was called "Political Arithmetick" in the early days of the Royal Society. Perhaps its most applied form is in the construction and interpretation of models and projections for policy purposes and for national accounts.
Policy makers cannot carry out "experiments" of the sort that are conventional in the laboratory sciences, especially in the social and economic sectors. Thus they turn to models to simulate possible consequences of policy initiatives, and to see how changes in the society or population will affect the achievement of these policies. Fundamental to policy models are demographic methodologies termed projections, macro-simulation modelling and micro-simulation modelling.
Simulation modelling is a rapidly expanding field, particularly in the social and economic areas. There is a need to bring the purveyors of these methodologies (typically social scientists) and users (typically the policy community) up-to-date with emerging techniques and the data sources they draw on.
Demographic projection techniques had taken a similar form, with what is termed cohort-component techniques, for a long time. These methods are arithmetically rather simple but involve large data sets, and detailed and time-consuming computations, albeit that the new generation computers make this far less laborious. Attempts to provide the same results using, say, more parsimonious mathematical or statistical models have proved futile.
But there is a nagging concern. Forecasts in demography, as in related fields, have traditionally given two misimpressions: that they provide exact future numbers, and that the trajectories they portray are deterministic, rather than the balance of probabilities. Agencies attempt to avoid the appearance of determinism by computing a series of different projections with varying assumptions, but recently there has been dissatisfaction expressed about this approach and empirical papers in journals such as Nature have shown the problems that have been engendered.
Their response has been to turn to what are called "stochastic" or "probabilistic" projection techniques. These yield an infinite number of parameters, but the results are then couched in terms of the probability that a trajectory or range of trajectories is likely to occur in the future. So far much of the work has been univariate, particularly on mortality assumptions, but the results are highly significant. The projections of mortality within quite rigidly defined limits (that is, within the 50% level) yield results that show a remarkably wide range for the populations, especially the oldest of the old, estimated by these calculations, and thus have enormous implications for the provision of health and other social services.
As implied above both the purveyors and the consumers of these methods and their empirical outputs - both the scientists and the policy analysts - have to be aware of the methodological problems and how to interpret results.
The workshop provided the opportunity to meet some of the world's leading people in this field and discuss the issues raised, including whether these methods should be used by New Zealand government agencies, and whether we have the data and human capital that would facilitate this. It also gave the opportunity for discussing problems perhaps peculiar to New Zealand, such as those faced when projecting ethnicity.
International speakers at the demography workshop included Shripad Tuljapurkar, Stanford University; Frans Willekens, Netherlands Interdisciplinary Demographic Institute; Nico Keilman, University of Oslo; Heath Booth, The Australian National University; Tom Wilson, Queensland Centre for Population Research; John Bryant, Mahidol University, Thailand; and Michael Rendall, Office of National Statistics (UK) and Rand Corporation (US).
For more information, contact dharma@waikato.ac.nz
