Harmonising geographies for analyses of residential segregation: an example using the 1km2 cells of the European grid for the city of Barcelona

Juan Galeano, Centre d’Estudis Demogràfics (CED), UAB
Albert Sabater, University of St Andrews

To determine the trajectory of residential segregation over time we need a series of consistent population estimates. However, a time-series of estimates is hard to calculate for a variety of reasons, most notably due to alterations to the geographical boundaries for which data are disseminated. Therefore, unless a consistent geographical approach with time series data are taken, it is difficult to know whether changing trends are taking place or whether observed changes are simply an artefact of a boundary change. In Spain, this problem is particularly acute for small geographies such as census tracts, which are constantly affected by electoral changes. In this paper, we provide evidence of the potential of harmonising time-series data on a consistent geographical basis using the 1km2 cells of the European grid as the target geography for national and international comparisons. Using data from the Municipal Register of Inhabitants (or population register), we present an implementation for analyses of residential segregation for the City of Barcelona. In doing so, we address our main research question –can consistent time-series be used as a way to improve comparability of residential segregation of foreign-born populations over time and space? Our results suggest that making the census tracts consistent over time is crucial for the interpretation of change in segregation, as indices can be altered and misleading when data directly from the population register are used. In this context, the results highlight that an increase in segregation can be purely artefactual, reflecting solely small area boundary changes between 2002 and 2013. After harmonising time-series data on a consistent geographical basis a decrease in residential segregation is observed for most groups, with alterations on the index values that can be greater than the impact of changes over time.