The Visualisation of Spatial Social Structure


Appendix A lists the algorithm which was developed to create most of the illustrations in this dissertation. The other appendices are included because the tables shown could not easily be derived from any other single source and were all specifically constructed for the purposes of this thesis. Without such lists, and the look-up tables of the relationships between so many different areas, many of the pictures created here could not have been produced.

Appendix B gives a table linking the changing geography of parliamentary constituencies between three incompatible time periods. It represents the consolidation and extension of two previous studies. Appendix C provides the complete list of general election results for these constituencies. It was created from many separate sources. It was meticulously checked by hand. Appendix D gives a list of the changing average house prices in each of the constituencies during most of the 1980s. It was compiled from building society records and a sophisticated translation of the postcode to administrative geographies was made. Appendix E shows how the particularly difficult problems of local geography in Scotland were handled, through several months of monotonous manual work with high scale maps and geographical indexes. Appendix F presents the ward cartogram created for this thesis, in detail. A complete list of the wards is given, along with a dozen insets showing, close up, the actual cartogram covering most of Britain.



Literally hundreds of millions of statistics were used to create the pictures shown here. Most were available from national sources as computer files: the government census office, research council data archive and national information system (NOMIS). Other information was amassed from contacts within the author's department. The cancer registry figures, local election results and building society statistics were collected in this way. Still more numbers were laboriously typed in from paper sources, the 1988 Scottish local election results, for instance. Listed here are those tables which has to be specifically created for this dissertation and for which others may find a use.

Appendix A: Circular Cartogram Algorithm

The algorithm is presented in its Pascal implementation and annotated. The entire code required to produce the cartograms is contained in the next two pages. The final page gives an example of the sort of information the program listed would require, as input, to create a circular cartogram. All that is needed is a list of the places to be included giving their initial centroids in Euclidean space, population, and their topological relationships with their neighbours (indicated by length of a common border). Changing these topological relationships will alter the final image so, for instance, a bridge can be represented as a 10km border in the file to have an influence. The program outputs a list of the places' circle's radii and centroids after an arbitrary number of iterations.

The algorithm could be improved by being able to decide when to finish. However, it is difficult to avoid settling on local solutions too early. It is most effective to be able to see the operation in progress on a computer screen, and to influence the final image through interaction with a pointer. Such procedures are not included in the following listing as they would complicate it and are machine dependant. If a way could be found to avoid having to rebuild the two-dimensional tree structure at every iteration, as it changes very little, that could substantially increase the efficiency of the algorithm. It should be noted that, as given, the code could easy be implemented on a system based on massively parallel architecture. The algorithm is also fast enough to run on a microcomputer with over ten thousand units, as it stands.

Appendix B: Parliamentary Constituencies 1955-1987 Continuity

A list of seven hundred and five parliamentary constituencies was constructed by hand, covering a period containing two substantial boundary changes. In many cases the link over time has had to be between best fitting constituencies. Twenty eight constituencies disappeared between 1970 and 1974, and twenty seven were lost between 1979 and 1983. However, at each change the overall number of constituencies increased (from 630, to 635, and then 650). Sedgefield is a unique example; it existed, as a constituency, between 1955 and 1970 (number 681 in the table), disappeared from 1974 to 1979, but reappeared from 1983 to 1987 (as number 12). Often the name has changed, as with Newcastle North-, from West-, but the actual area and people have remained substantially the same.

Official names and press association numbers are given for each constituency in each period, along with the identification number used in this thesis. Cartograms indexed by this consistent code, are included for the years 1955, 1964, 1970, 1979 and 1987. The changing boundaries of Greater London, the West Midlands, Greater Manchester and central Glasgow conurbation are shown on all the diagrams, as are the Welsh and Scottish borders. These were all carefully added by hand to ensure continuity had been maintained. The constituencies of Northern Ireland are also shown, as they increased in number, and their changing electorates altered the sizes of the circles in population space.

The major two sources for estimating the extent of continuity and implication of parliamentary constituency name changes were Butler D.E. & Kavanagh D. (1974, pp.282-283) for 1970-1974, and BBC/ITN (1983) for 1979-1983. The constituencies are grouped by region.

PANo. stands for "Press Association Number" of the constituency, and can be seen to have had changed after every redistribution of seats.

Appendix C: Parliamentary Constituencies 1955-1987 - Results

A table of results had be constructed and checked for all seven hundred and five constituencies identified in each of the general elections contested between 1955 and 1987 inclusive. The identification number and current name of each constituency is followed by the statistics from the elections of 1955, 1959, 1964, 1966, 1970, February 1974, October 1974, 1979, 1983 and 1987. The BBC/ITN (1983) estimate of what the 1979 result would have been, if it had been fought on the 1983 boundaries, is also included for completeness. For each election the total votes cast for each of the Conservative, Labour, Liberal or Alliance, Nationalist and any other parties (combined) is given, along with the total electorate, listing over forty thousand separate counts of the wishes of almost half a billion electors across four decades.

The figures shown were mostly collated from studies deposited in the Economic and Social Research Council's data archive in Essex. In all cases, except February 1974, two independent versions of each set of election results were available to be compared. The studies used were those of Latham (1982) covering 1955-1979, Curtice (1983) covering 1955-1970, and Payne & Butler (1983) covering October 1974 and 1979. Personal communications from Ron Johnson and Ivor Crewe (via Martin Harrop) provided two versions of each of the 1983 and 1987 results. Inconsistent records were subsequently compared with the paper records of F.W.S. Craig (1971, 1977, 1980, 1981, 1984, 1989), the Nuffield Election Studies Publications (Butler D.E. et al 1955, 1960, 1965, 1966, 1971, 1974, 1975, 1979, 1984, 1988) and the Times Guide to the House of Commons (1987). The final table presented here is thought to be an extremely accurate record.

Where a constituency did not exist at a given election, blanks have been inserted. "" signs are placed where the vote for the other parties was too large to be printed in that column. Each page of the table spans two A4 sheets and is printed at the minimum legible font size, to allow so many numbers to be included.

P.C.A. stands for "Parliamentary Constituency Area" code and can be used to refer from this

table, to those in Appendices B and D.

Appendix D: Average Housing Price by Constituency 1983-1989

A list of the six hundred and thirty-three current mainland parliamentary constituencies is given, with their P.C.A. (Parliamentary Constituency Area) number, sorted by average 1989 housing price, following the average price for the years 1983, 1984, 1985, 1986, 1987 and 1988 which are shown for comparison.

The house prices were derived from data-sets provided by the Nation-wide and Halifax building societies. The average prices for six categories of house, and the number of sales, were calculated for each area (Brunsdon C., Coombes M., Munro M. & Symon P. 1991). From these six measures an average housing price was constructed as a weighted geometric mean based on the share of houses sold in each or the country's postcode sectors against the national distribution.

The figures for postcode sectors for each year were transferred to the administrative ward base. The proportion of individual postcodes in each ward in different postcode sectors was used to derive an average figure for each ward. This produces a slight smoothing effect on the figures. Very rarely no sales would be recorded in a particular postcode sector, but every ward would have a price estimated. Occasionally a figure was unavailable for a certain year, and then a price would be extrapolated (forwards and backwards in time) for that ward. This table was used to calculate local housing price inflation for the detailed pictures included in this work.

For the list given here, the average housing price was translated to the much more coarse constituency base, which is merely an amalgamation of wards. The consistency of the figures given, over time, shows the quality of the sample used, especially at this spatial scale. It should be noted that only the prices of homes sold by the two building societies mentioned above, were included.

Appendix E: ScottishWard to Postcode Sector Look-up Table

Information on local voting, industry and unemployment is often given on a ward basis. But in Scotland "part-postcode-sectors" were employed, rather than wards, in the 1981 census. Thus on the 1981 population cartogram it is these areas which are shown in Scotland. The geography of Scottish electoral wards is not well documented or mapped. A look-up table was painstakingly constructed, using all the available information and cartographic sources.

A set of crude diagrams showing the boundaries of postcode sectors was compared with large scale maps of each area showing ward boundaries of some of the major conurbations. Often obscure place names had to be relied on, each indexed and located manually, before the closest postcode sector could be found and allocated. Sometimes only a rough ward or postcode sector centroid was known. The final coverage is complete. Every ward is connected to at least one part-postcode-sector and visa versa.

The 1987 "National On-line Manpower Information System" number and name of each ward is given, along with the list of 1981 part-postcode-sectors to which it corresponds. In some places, such as the Island areas, several wards are allocated to the same, large sector. Elsewhere the opposite occurs; in the middle of large cites for instance, where few people sleep (and vote) but many work (and thus receive post).

Overall, the fit is not too bad. The table is certainly adequate for the purposes of illustration that it is used for here. The letter "p" after a postcode sector code indicates that the code refers to only that part of the sector lying in the same local government district as the ward to which it corresponds.

Appendix F: Local GovernmentWards, 1981 and 1987

The following twelve pages of figures and the accompanying diagrams show the 1981 ward population cartogram in detail. Each of the 10,444 1981 census wards (and Scottish partpostcode- sectors) is listed with the O.P.C.S. number in bold and corresponding 1987 National Information System ward number (in italics) and name. Many ward names were not reported to the 1981 census authorities, but only uninformative codes given. The link between the 1987 and 1981 ward numbers was largely constructed automatically by linking wards in the same district whose name had not changed. Substantial alterations had occurred to wards in some eighty-two districts including all of Wales (except the Rhonda Valley). All the ward boundaries of Scotland were also (often drastically) revised by 1984. Two sets of boundary changes had been made in several places. High scale maps covering most of the country were consulted to estimate the effects of the changes. Usually a one-to-one link could be made, and that is all that is shown in the list here, to conserve space. More complex situations were also appropriately handled, however.

The insets of the ward cartogram are A4 printer's proofs, of an A0 sized illustration. They were used to check that local continuity had, in general, been retained. They can also be used to identify individual locations in some of the earlier illustrations. Along with the list, which names individual estates and villages, they illustrate the great detail of information within many of the pictures presented in this dissertation. Every spot can actually be named.

The ward cartogram is the most detailed representation of human geography that covers the whole of Britain. Beneath this scale places no longer have names, but are the physical realities of streets, farm houses and tower blocks. Local elections are regularly held at the ward level; unemployment counted every month; industrial structure estimated every few years; and total migration and commuting matrices constructed once a decade. The ward level is often the finest level at which we can study our society by quantitative means. The ward level is also that at which many of the processes of society are acted out. Our wards are our home ground. If we work or shop or go to school "where we live" it is usually in the same ward, or a neighbouring one. To map at the ward level is to map at the human level.


Published by Wiley

Order online: Amazon , iBooks