Introducing duration dependant emigration in the DREAM population projection model

During the recent decade, changes in immigration flows and immigration behaviour are important sources to explain changes in the composition of the resident immigrant population with respect to duration of residence. Considering that demographic behaviour varies considerably with the length of duration, this challenges the baseline assumption of not considering duration of residence when determining future demographic flows. This working paper explains the consequences of allowing forecasted emigration of immigrants, i.e. re-emigration, to depend on duration of residence and investigates whether including this characteristic enhances projection accuracy when facing shifts in immigration structure. The propensity to re-emigrate decreases with duration of residence. Typically emigration probabilities for individuals having immigrated within the last two years lie above the average re-emigration probability, whereas the propensity to re-emigrate lies below average when duration of residence exceeds two years. Using constant emigration probabilities depending on gender, age, origin, and number of years of duration is shown to lead to an increase in the immigrant population compared to the baseline scenario omitting emigration by duration of residence. The implications with respect to projection accuracy are assessed by performing sequential within-data population projections, respectively involving and omitting duration dependant re-emigration. Finding that future shifts in immigration behaviour severely challenge projection accuracy when taking duration of residence into account, it is suggested that duration dependant emigration should be applied with caution.