Can a single model for forecasting migration and assessing its impact be appropriate for different European countries? The simple answer is no. Due to different time horizons, migration studies have to deal with significant uncertainty resulting from policy changes and unforeseen events. However, we can use established statistical and econometric methods to improve the analysis for a range of different situations. In a recent report, we introduced a new approach to examining the effects of migration and forecasting migration for a selection of European countries through a macroeconomic lens.
At the macroeconomic level, the continuing differences between Western and Eastern Europe as well as the size and direction of migration processes hint at why a single migration model is not able to consider all possible circumstances. Even just looking within Western Europe, the disparities in terms of GDP and migration introduce too much complexity for one model. High migration levels in Italy and Spain are not corresponding to the macroeconomic foundations, with lower-than-Western-European-average GDP per capita coinciding with high net immigration. In Eastern European countries, countries with positive net immigration have higher real GDP per capita than those with negative net immigration. To help study migration as a macroeconomic process, we, therefore, assess different groups of countries separately, based on their macroeconomic and migration profiles.
Moving to the micro level, there is also an important distinction between migrants who can be classified as high- or low-skilled. The analysis of their relative labour force participation rates offers insights that are applicable to different countries, which macroeconomic data alone are unable to show. Nevertheless, by using macroeconomic data based on the analysis of a range of possible models, we can also make some general statements that hold across different country groupings. One example is that an increase in net migration is expansionary for the economy – and vice versa. Net tax revenues, wages and salaries tend to increase with net migration and so do employment rates. A more detailed analysis is offered in the report.
In terms of forecasting, we confirmed earlier findings that migration is generally difficult to explain and predict. This especially holds for net migration, which is less predictable than either immigration or emigration. In the short-term horizon, the prediction errors for immigration and emigration were also reasonably well-calibrated, more so than for net immigration. Highest uncertainty levels were observed for migration patterns of Eastern European countries, particularly those with changing migration profiles. For longer-range migration scenarios, we found that after a decade or so, the uncertainty bounds were too wide to be useful, which pointed to the need to use approximate solutions instead.
One example of the types of questions that can be explored with macro-level models of migration dynamics is the role of job automation. Here, we employ theoretical models which use micro-foundations to describe migrant decision making, to answer questions such as: ‘for countries who are net receivers of low-skill migrants, how does the automation of jobs, or replacement of jobs by robots, affect migration flows?’ The answers depend on the degree to which migrant workers and robots are substitutes and the strength of the reaction of the demand for labour to automation. This is just one example demonstrating the importance of these models so that we can better account for uncertain shocks and the many complexities of migration modelling and forecasting.