Immigrant selectivity describes the notion that migrants are not a random sample of the population at origin, but differ in certain traits such as educational attainment from individuals who stay behind. In this article, we move away from group-level descriptions of educational selectivity and measure it as an individual’s relative position in the age- and gender-specific educational distribution of the country of origin. We describe the extent of educational selectivity for a selection of Western European destinations as well as a selection of origin groups ranging from recent refugee to labor migrant populations. By contrasting refugees to labor migrants, we address longstanding assumptions about typical differences in the degree of selectivity between different types of immigrants. According to our findings, there are few and only minor differences between refugee and labor migrants. However, these differences vary; and there are labor migrant groups that score similar or lower on selectivity than do the refugees covered in this study. Selectivity differences between refugees and labor migrants therefore seem less prominent than arguments in the literature suggest. Another key finding is that every origin group is composed of varying proportions of positively and negatively selected individuals. In most cases, the origin groups cover the whole spectrum of selectivity, so that characterizing them as either predominantly positively or negatively selected does not seem adequate. Furthermore, we show that using country-level educational distributions as opposed to sub-national regional-level distributions can lead to inaccurate measurements of educational selectivity. This problem does not occur universally, but only under certain conditions. That is, when high levels of outmigration from sub-national regions in which economic opportunities are considerably above or below the country average, measurement inaccuracy exceeds ignorable levels. In instances where researchers are not able to use sub-national regional measures, we provide them with practical guidance in the form of pre-trained machine-learning tools to assess the direction and the extent of the measurement inaccuracy that results from relying on country-level as opposed to sub-national regional-level educational distributions.