“…For this purpose, we estimate the following cross‐sectional log–log regression model: where ln( represents the natural log of the above‐described NMR8 for the period of 2007–2014,9 represents a comprehensive set of county characteristics in the baseline year 2007, and ε i is a random error term. This approach is similar to that of previous studies estimating the general appeal of regions for migrants (Buch et al, 2014) or the regional distribution of foreign employment (Lehmann & Nagl, 2019) in Germany. However, unlike the former authors, we are particularly interested in the diverging locational preferences between the above‐described subgroups of migrants and between the external and internal migration of foreigners, and unlike the latter authors, we are interested in flows and their contribution to population development instead of the static distribution of foreigners.…”