In this paper, the levels of residential segregation of immigrants in Sweden during the years 1990, 1997, 2005, and 2012 are calculated. This paper applies a novel method for calculating segregation that is multi-scalar and addresses the modifiableareal unit problem (MAUP). The level of segregation is evaluated for each populated location by identifying the population that includes the kappa-nearest neighbours. The share of immigrants in this assessment population is then compared to the share in the reference population that comprises the kappa-nearest neighbours. One of the strengths of this method is the possibility to modify the reference population, thus making it possible to measure the difference in the results due to the size of the reference population. This study demonstrates that the results can considerably differ depending on which reference population is used. Furthermore, this study indicates that using different reference areas can produce completely different trends over time, such as decreasing or increasing segregation. The results demonstrate a general increase in segregation between 1990 and 1997, followed by a more complex pattern from 1997 to 2012. The segregation values are presented for all populated locations in Sweden, and population-weighted means are calculated for the whole of Sweden, in addition to the Stockholm, Malmo, and Gothenburg metropolitan areas.