29 Mar 2014 17:52 GMT
I recently produced some research exploring the extent of ethnic separation and integration in Parliamentary constituencies, which was presented at a House of Commons Library Talk on the Census last month. The results are rich and varied — there is more to them than can be adequately represented in summary statistics — and they lend themselves to interactive exploration.
So over the last few weeks I have spent my weekends trying to make the data easier to explore and understand visually. If you want to jump straight in you can begin with these regional maps. The following post explains the research in detail: what the maps show, how they show it, and what the overall picture of ethnic separation looks like in Parliamentary constituencies in England and Wales.
What do the maps represent?
The regional maps show Parliamentary constituencies in England and Wales shaded according to their separation score, which is a measure of the degree of geographical separation between White and non-White ethnic groups in each constituency, based on 2011 Census data. Constituencies are shaded on a linear scale, where the constituency with the highest separation score is shaded black, and the constituency with the lowest is shaded white. All other constituencies are shaded between those two extremes on a blue gradient.
Clicking on a constituency takes you to a map of that constituency broken up into its output areas; these are the smallest areas for which Census data is available. Each output area is shaded to show the percentage of the population in non-White ethnic groups. Here the scale is also linear but the shading is absolute: an area that is 100% non-White is shaded black, while an area that is 100% White is shaded white.
This shading scheme provides the strongest possible contrast between levels of separation in different constituencies at the national level, while making the maps of individual constituencies directly comparable with one another.1
What is separation?
An index of separation (or index of dissimilarity) measures the extent to which different groups within the population of an area live together or separately. To construct an index of separation you need two things. First, you need to be able to divide the population into two mutually exclusive groups. Second, you need information about where people in those groups live at two geographical levels: a larger area, the area in which you are measuring separation, and a collection of smaller areas, which together make up the larger area.
Calculating the separation score involves working out what percentage of the total population of each group living in the larger area lives in each of the smaller areas. The key idea is that if the two groups are perfectly integrated then these percentages will be the same in each smaller area. The separation score is calculated by adding up the absolute differences between the percentages living in each smaller area and dividing the total by the maximum possible total difference (which is 200%). This gives you a value between zero and one, where zero indicates total integration and one indicates total separation.
This sort of index has three weaknesses as a measure of separation. First, it necessarily divides the whole population into just two groups. In this analysis the population has been divided into the White and non-White ethnic groups because the White ethnic group is by far the largest across England and Wales as a whole, so dividing the population in this way provides the broadest comparison of separation between the majority and minority ethnic groups. But it is important to remember that patterns of separation may differ between ethnic groups, and the analysis would produce different results if the population was divided into Asian and non-Asian, or Black and non-Black groups.
Second, while the index measures the relative difference between the proportion of the two groups in each smaller area, it doesn't reflect how these differences are spatially arranged. A high separation score could indicate that one of the groups is located in a single community spanning several smaller areas, or it could indicate that the two groups are arranged in something like a checkerboard pattern, living in a number of exclusive pockets. This is why the constituency level maps are so important, because they let you examine the relationship between the geographical distribution of each group and the separation score.
Finally, the analysis is based purely on people's residential addresses and self-defined ethnicity at the 2011 Census. It doesn't say anything about the extent of social interaction between people in different ethnic groups, or how people think about their ethnic identity.
Patterns of separation
The map below shows all constituencies in England and Wales shaded by separation score.
Looking at the map, certain patterns are immediately apparent. Separation scores at the 2011 Census tended to be lower in the south and the east and higher in the north and the west. This pattern can be clearly seen in the boxplots below, which show the median separation score in each region (the horizontal line), the interquartile range (the coloured boxes), and the range (the vertical grey lines).
Separation scores were typically lower in London than in any other region: 33 of the 50 least separated constituencies were in London. London's relatively high upper bound in the chart above is misleading; it's due to one outlier, which when removed makes London's upper bound lower than that of any other region.
The constituency with the lowest separation score was Sutton and Cheam, where 21.4% of the population was non-White. This was lower than the 40.2% of the population that was non-White in London as a whole, but there were a number of constituencies in London with large non-White populations that also had low separation scores. Croydon North, Harrow West, Walthamstow, and Edmonton were all in the top 20 least separated constituencies and all had a non-White population of greater than 50%.
Going to the other extreme, the most separated constituencies were found in a cluster that straddles the boundary between the North West region and Yorkshire and the Humber. The top five most separated constituencies were: Dewsbury, Burnley, Oldham East and Saddleworth, Oldham West and Royton, and Halifax. In Dewsbury, 21.6% of the population was non-White, a very similar percentage as in the least separated constituency, Sutton and Cheam.
In fact, there is very little apparent relationship between the percentage of the population that is non-White and the index of separation between the White and non-White ethnic groups.
But whatever the combination of factors driving levels of separation, factors leading to greater intergration are found in London, while factors leading to greater separation are found in a cluster of constituencies in the north west of England.
1. When these maps were first published, areas were shaded using a linear HSL scale, where the hue and saturation were held constant and the lightness varied in proportion to the value being represented. However, because of the way humans perceive colour, equal distances in lightness between colours of the same hue are not perceived as equal. The maps have therefore been updated to use a gradient in the HCL colour space, which adjusts for this effect.