Transforming the mosaic - The Visualization of Spatial Social Structure - page 132

great deal of information to be lost; local patterns can no longer be seen and
large areas arbitrarily appear uniformly good or bad when the more truthful
picture is very different (Figure 5.1). When aggregation is employed it is often
advantageous if the larger areas are home to similar numbers of people.
To see national patterns or regional or city-sized processes it is sometimes
better not to use national, regional or city-sized spatial units. Rather, show the
eye the finely detailed picture and let the mind decide how much pattern does
or does not exist. Only then can the decision be made whether to smooth the
picture further. Only if you've seen the detail do you know if the aggregation
provides a fair summary. There are also many means other than indiscriminate
geographical amalgamation that can be used to generalise an image.
How, though, do we create these fine images of local change from two
sources based upon small, but differing, areas? One solution is to recognise
that any change in boundaries has only a very local effect (Box 5.1). People are
moved from one side of the line to the other. There is no need to abolish the
line, simply to realise that a few people have been moved. A detailed image,
where nothing but the boundaries has really changed, will simply appear a con-
stant, slightly speckled shade. The eye interprets the fine dithering that will have
been created by misplacement as a colour, not a pattern. The problem has been
reduced away.
It is not always possible to say there is a simple solution to depicting change
when you try to give all people equal representation. Between 1971 and 1981

Box 5.1 Linking the censuses

The 1971 and 1981 census geographies
were linked at the enumeration district
(ED) level. The majority of ED bound-
aries had not changed or were nearly
identical, but in some places substantial
alterations had occurred, for instance
where a new town had been built or an
old estate pulled down. The use of the
'census tracts' designed by OPCS had been found to be far from adequate by
McKee (1989). An alternative, far more flexible solution was devised. Only
enumeration district centroids were known for each set of roughly 130 000
points. Two two-dimensional tree-data structures were built and the closest
1981 district to each 1971 found, and vice versa. Thus every ED in each set
was connected to at least one in the other, but could be connected to any
number, if necessary. Every ED count could then be compared between the
two censuses.