Introduction

Forecasting is not in the first place a technical exercise; rather, it involves thinking methodically about the future. In so doing, we need to strategically leverage our knowledge and expertise of the migration process and assess how our experiences in the past will (or will not) carry over into the future. We have our own theories and models – sometimes intuitive, sometimes explicit – for doing so. In the end, we are obliged to quantify our beliefs about the future – an undertaking that is, by its very nature, technical. In order to make this task of ‘projecting’ a scientific one, a framework that integrates beliefs about the future with empirical evidence about the past is necessary. The Bayesian approach advocated by Wiśniowski and co-authors in chapter 9 of this volume is an example of an appropriate framework for migration forecasting. The Bayesian method is intimately related to forecasting. In fact, Bayesian forecasting is something of a mild redundancy because forecasting is at the core of the Bayesian approach to just about anything (Geweke & Whiteman 2006). Still, its application to the complex undertaking of immigration forecasting is relatively new (for one of the first attempts, see Gorbey, James & Poot 1999). Bijak (2008) also provides a good introduction. The authors have covered new areas with their approaches, and have set the stage for a fruitful new line of thinking about forecasting in migration research. Without an abundance of academic precedents, they sought to invent new ways of dealing with multiple complexities in the application. They have done so successfully, and have tried to document these steps as diligently as possible. Nevertheless, migration forecasting remains a difficult exercise involving many different steps, including a series of Delphi rounds among country experts, statistical modelling, interpretation, adjustments and forecasting.

For all these reasons, this is not just another forecasting exercise, but an innovative new step. Its main innovative qualities can be found in the way expert views and empirical evidence are combined to produce an ‘optimal’ forecasting model. It is an important contribution to the discussion of whether and how to integrate expert opinions about the future into quantitative trend-type models.