Forecasting age patterns of international migration: adapting and extending the Lee-Carter Model to different data types and time series

James Raymer, Australian National University
Arkadiusz Wisniowski, University of Southampton

Migration is often considered the most difficult demographic component to forecast. The reasons are due to the vast heterogeneity of migrants, the lack of a clear observable event and the levels are influenced strongly by economics and politics, which are also difficult to forecast. Because of these reasons, there has not been much success acheived in migration forecasting. Instead, national statistical offices tend to rely on very simple assumptions regarding net migration based only on a few recent years or that with future net migration set to zero. Age and sex patterns are either ignored or kept fixed over time. This is unfortunate because migration is often the most influential component of population growth, especially in developed societies. In this paper, we adapt and extend the Lee-Carter (1992) model to demonstrate its usefulness for forecasting age-specific immigration and emigration for the United Kingdom, Sweden and Australia. These three countries all have time series of migration but at different measurements and levels of quality. The results demonstrate the promise of this approach as well as directions for further research.