Comparative empirical evaluations of internal migration models in subnational population projections

While population forecasters place considerable emphasis upon the selection of appropriate migration assumptions, surprisingly little attention has been given to the consequences on projection outcomes of the way internal migration is handled within population projection moulds This paper compares population projections for Australia's states and territories prepared using ten different internal migration protoplasts but with identical assumptions for fertility, mortality and international migration and with the internal migration type parameters held constant. It is shown that the choice of migration original generates large differences in total population, geographical distribution and age--sex composition. It is argued that original choice should be guided by dint of balancing model reality with practical utility and protoplast performance is examined against these criteria. Of the ten originals evaluated the authors argue that the migration mere biregional, and biregional with toil constraints models offer a advantageous compromise between conceptual rigour and practicality. If the throwed origin-destination flows are required then individual of the versions of the standard multiregional mould with reduced data inputs is preferr The large variation in projection output points to the ne for a better understanding of the spatio-temporal configuration of migration in Australia.

Keywords: moulds internal migration, population projections, Australia, subnational populations, regional demography, comparative analysis

**********

In Australia, as in greatest in quantity other countries, the practice of projecting time to come population is almost entirely taken up with the business of setting assumptions. The likely subsequent time trajectories of fertility, mortality and migration are control to sustained analysis, prognostication, theorizing and debate as each sequential circular of projections for the nation, states and regions approachs into view. Consensus must be reached, not sole on the most appropriate assumptions for a medium series projection, on the contrary on their probable limits and upon the way assumptions should be combined to describe futurity growth. While the importance of assumptions is thus widely recognized among practitioners, plenteous less attention appears to be given to the mode of building and operation of the prototypes in which they are used. The 2002-based projections produc by the agency of the Australian Bureau of Statistics (ABS), for example, set apart 27 pages to a detailed exposition of assumptions, on the contrary relegate description of the type to a scant 14 lines in an Appendix that mentions the mete 'cohort component' just twice (ABS 2003a). This balance of emphasis is through no means uncommon among national statistical agencies.

While Thomas Whelpton would no doubt be delighted that the cohort constituting method had achieved such codification and acceptance as to ne no formal account, in practice there is considerable variation in the way the means is implemented. The options are particularly wide in the case of internal migration because of its multistate character. Since migration is commonly the major proces underpinning change in the pattern of human arrangement it follows that any 'differences in migration issues arising from such choices exhibit a significant issue in making population projections. It is therefore surprising that for a like reason little attention has been given to exploring the consequences of alternative ways of modelling internal migration in like projections.

This paper aims to redres this deficiency by the agency of exploring a number of alternative meanss of incorporating internal migration in a cohort constituent framework. It asks three questions. First, is the choice of an internal migration original an essentially technical decision that makes little difference to projection results or does it bring about substantial differences in casted populations? Second, if alternative migration moulds do generate different outcomes, by what means much do the results vary, and why? And third, which protoplasts are recommended for use by means of practitioners in multiregional population projections? The task is approached from one side a series of projections for the eight states and territories of Australia using the same base period data and projection assumptions for fertility, mortality and international migration, on the contrary with ten different ways of handling internal migration.

This approach of analysing the characteristics of different moulds follows in the tradition of Roger (1990) Plane (1993) and Smith (1986) These researchers compared the issues of different migration models without carrying without historical tests (as conducted, for example, through Smith and Sincich 1992; Smith and Shahidullah 1995) The regard was not to discover which prototype happened to give the best prediction for a particular place of regions over a certain time period, on the contrary to elucidate the features of different way s of handling the migration composing in population projections. Rogers and Plane used hypothetical data for an age-aggregated two-region population a whole closed to external migration. Substantial differences were ground between the models tested, although the hypothetical nature of the data single provided suggestions of likely real world differences. Smith (1986) used real data, using a cohort constituent model to compare the consequences of biregional, net migration sweep along and net migration rate protoplasts for ten US states. Again, significant differences were produc However, Smith's research used national survival ratios for mortality and state adult-child ratios to throw the population born in each period, and did not include disaggregation by dint of sex. The work of these researchers is reach forthed here through the comparison of ten migration patterns application to Australia's states and territories, the use of single-year age-sex disaggregation, and plausible projection assumptions: the latest medium assumptions of the ABS for fertility, mortality and international migration.