SASI Home| Publications | Data | Maps | SASI Staff
Dorling, D. (1991) The Visualization of Spatial Structure, PhD Thesis, Department of Geography, University of Newcastle upon Tyne

Chapter 4: Honeycomb Structure

Detail cumulates into larger coherent structures; those thousands of tiny windows, when seen at a distance, gray into surfaces to form a whole building. Simplicity of reading derives from the context of detailed and complex information, properly arranged. A most unconventional design strategy is revealed: to clarify, add detail.
[Tufte E.R. 1990 p.37]

4.1 Viewing Society

Britain is an island, a small piece of land upon which, at the end of the twentieth century, over fifty million people live. Why study this single percentage of the world’s population and why study it now? Because it is convenient, it is understandable and it is timely to do so.

The British state is a convenient unit of analysis; consistent national statistics are collected on many subjects at regular intervals about all its constituent elements — often to be mapped54. A single political body is created from areas covering all its territory, and its territory is divided into areas which are themselves subdivided into units of local representation for political election. Its boundary is well defined, and encompasses almost all the movements of people who live within the state borders.

The British state is an understandable social entity. The definitions of people, work, places, political parties and groups are simple. Its history and geography are well documented and easily accessible. It covers too little land area and too many people to divide easily into smaller wholes. The country can be travelled from one end to the other by land, in a day; but only a few of its people can be met in a lifetime. Britain is understandable because we live in it; we make up its social landscape. If we cannot see the nature of our spatial relationship with each other, then what will we be able to see55 (Foley D.L. 1953, Shepherd J., Westaway J. & Lee T. 1974, Bunge W.W. 1975, Evans I.S., Catterall J.W. & Rhind D.W. 1975, Massey D.S., Tedrow L.M. & Stephan G.E. 1980, Census Division, OPCS 1981)?

It is timely to look at the people of Britain now; not only because it has only recently become possible to look at them in the way we are going to, but also because of the moment in history which we will be viewing. After over a century of decline, what makes the British unique is their position, as a people, who have fallen further on the world scale than any other group in recent times. Now they are finally beginning to recognise their position and to take stock of their plight. How better to do it than by painting pictures of the social landscape they live in and make up?

4.2 Who the People Are

The people of Britain are a diverse collection. This small section seeks only to describe the basic geography of their most simple attributes, gender and age for example (Print LXVIII), and in doing so to begin to explore how the detailed mosaic of life in Britain can be uncovered (Figure 11).

The pictures drawn to accompany this chapter are based on population cartograms of one hundred and thirty thousand enumeration districts56. This resolution was chosen as the finest that is possible — closest to the local scale and individual realities of life. Great regional patterns can still be seen in the images, but only where they really exist, not as fabrications of the boundaries chosen. We have collected more information on people in the last twenty years, than over the previous twenty thousand. Is it not surprising that radically new techniques are required to view the social landscape57? Conventional choropleth maps at the level of ten thousand wards have also been included to show how they contrast with the message of the cartograms (Print LXIX).

Gender is the least ambiguous attribute we give people. A picture was originally painted of Britain, where each street block is coloured either black, for over-average proportions of females, or white, for under-average. The picture not only showed the random variation in this statistic, indicated by the speckled nature of the image, but also suggests simple patterns of slight over and under representation. There are more women in the middle of cities and along the South coast. A similar two colour technique is used later to show the distribution of Irish born (see Print LXXI).

To see distinctly the distribution of these proportions, a relative scale of measurement is adopted — above or below the median level of females rather than above or below fifty percent. Thus half the area of Britain, on the population cartogram, is shaded black, the other white. In this example the picture would hardly alter if an absolute scale were used. Absolute scales would require different legends for all pictures. The images would also vary greatly in their levels of saturation, merely because of the use of arbitrary, incomparable measures.

Here, levels above and below the median for Britain, or in some cases groups bounded by quartile levels, are used to shade the images, to treat all variables most simply, consistently and comparatively58. The division into two levels, quartiles and beyond, can be extended until continuous shading is achieved. This has not been used here, as it is difficult to shade, or see, such small areas continuously. More complex colouring schemes are developed later; but, as the images stand, continuous impressions are gained through the dithered patterns created by so many tiny discrete shades.

It is interesting to note that shading a great many small areas by four levels of colour reveals far more information on the complexity of the structure, than would by gained by showing larger areas with more detailed schemes. The interesting divisions are often local, and rarely correspond to administrative boundaries.

The main influence upon the patterns shown by the distribution of gender is age. Women tend to live longer, so where the population is generally older it is likely to contain more women. The distribution of the elderly, in every neighbourhood in Britain can also be shown, in the same way as the distribution of women was depicted59. The two maps could be compared, but shortly we will see how both variables can be shown on a single map, with yet a third variable, the distribution of the young, introduced through the use of colour. Variables, such as age, can then be further subdivided to show yet more detail (or the lack of it — Print LXX).

A traditional means of showing this information is to draw numerous population pyramids, but these fail to convey the distribution of age and sex structure across more than a few large geographical areas (Lawton R. 1968a, Dewdney J.C. 1968, Bureau of the Census 1970, Applied Urbanetics INC 1971, Lycan R. 1980, Warnes A.M. & Law C.M. 1984). Pyramids are, in addition, not easily visually comparable. What profitable use can be made of them is examined in Chapter Eight. It must be stressed here, however, that the use of large areas would naturally dilute the interesting structures by gross aggregation. Shading and colour is the most effective way to see local patterns across the structure of a national population.

4.3 Disparate Origins

Where the people who make up the social landscape of Britain came from in the past is perhaps as well known as it will ever be. Where the people who are alive today originate from is less well studied (but see Pocock D.C.D. 1960, Coates B.E. 1968, Allen J.P. & Turner E.J. 1988, Gibson A. 1988, Diamond I. & Clarke S. 1989, King R. & Shuttleworth I. 1989, Miles R. 1989, Ward R. 1989).

Migration is one of the major themes in this thesis. Static pictures of migration are best provided by looking at the distribution of people across the country who were born in a particular place. Shading every street in Britain by the proportion of its population whose birthplace was in Ireland (North and South) shows the scattering of people who, in the course of their lifetime, flowed from that one island to live in this60. We see immediately how strongly the Irish immigrants are concentrated in particular localities (Print LXXI).

Migration is about mixing. The picture fails if it does not convey the colourful mixtures of people that result from their movement. Colour, resulting from the mixing of light, gives the clearest images of the kaleidoscope of people’s differing origins. Unfortunately not more than three primary colours can be clearly distinguished from a mixture. Red, blue and yellow are the most easily combined and separated by the eye. They can be used to paint the basic picture of people mixing in Britain, representing proportions of those in each street born in the three nations of England, Scotland and Wales respectively (Print LXXII).

The countries naturally contain mostly people born within their boundaries. England is red, Scotland blue and Wales yellow. Mixing takes place between them, thus the Scottish border is purple, the Welsh orange. Disproportionate numbers of Welsh and Scottish people mix in London, where there is a dearth of English born, colouring the capital green (blue and yellow). Closer inspection shows just how intricate the pattern of mixing is. The white areas on the picture of the distribution of British-born are made up of streets where there are shortfalls of all the indigenous nationalities. Here immigrant populations are most densely settled in the social landscape.

As the indigenous population was divided into three, so too can people born overseas be subdivided into broad geographical categories (Print LXXIII). Here we use red for Asian, blue for Irish and yellow for African and Caribbean61. This image is dominated by black areas, with high proportions of all three immigrant groups (black represents the mixing on paper of all three colours, as white represents their absence).

Tints, tinges and trends of colour in the image graphically show how the mixing varies. The East side of the West Midlands is more Asian, the West more Irish in the backgrounds of its people. Those from Africa and the Caribbean settled in greater numbers in the South than the North of Britain, and so on. These are simple pictures; each block of streets (forming an enumeration district) is just a coloured dot, but already reveals, in a picture, details of the diversity of our society which a search of the literature (see Bibliography) and conventional image (Print LXXIV) failed to find.

4.4 Lost Opportunities

Now that we know something of who these people are and where they come from, we want to know what they are doing here. Gaining our information principally from what the British state collects, we are interested in how people are employed. Much of the geographical nature of this is largely determined by the first two questions addressed above. Children do not officially work full-time, the elderly are usually retired, men get more work than women, immigrant areas have often become (if they were not already) those places with the worst prospects of work. So pattern builds upon pattern and we dissect the body of information, before we can rebuild it to a better understood whole.

The simple distribution of the proportion of the population unemployed shows strong connections with aspects of those distributions mentioned above. Employment and unemployment is another major theme to run through this work. If we delve further we can compare the proportions of economically inactive people (mostly housewives, the retired, students and children) sometimes called dependant, with the unemployed and working populations using a three colour scheme (Print LXXV).

Red is used now for the unemployed, blue for those working and yellow for the dependants. The picture can show many variations and a complex geographical pattern. Orange areas are those with high proportions of unemployed and dependant people (the Welsh valleys), green indicates many working and dependant people living in the same places (the Home Counties) and purple shows blocks where high numbers of people are working while many others are simultaneously unemployed (parts of London for instance where there are relatively few dependants). Without this sophistication of colouring we might not have realised that such areas could exist at all.

Traditional maps fail altogether to portray distributions such as unemployment (Print LXXVI). They suggest a massive divide between the north and south by emphasizing the fate of those living in rural areas — Scottish crofters against London stockbrokers. In fact there are more unemployed people, and stronger concentrations of them, in the South than people living in the North East of England. What appear to be black-spots on the maps are great regions of joblessness in population space. Finally, it is the steepness of the slopes between the places of prosperity and poverty that is our concern. The depressed areas abut on the most fortunate62.

When we subdivide those who work (and did work) into what they do as a living, we do not find such close spatial affinity. Again, we are limited to three categories, by the ability of our eyes and flexibility of our imaginations, but three is enough to form a strong impression of the essence of social spatial structure.

4.5 Work, Industry and Home

Occupations can be divided into three broad groups according to how much people are paid. These groups correspond closely with the general nature of the work (Figure 12). The group commanding the highest income are professionals: managers, employers or land-owners, often university educated. The middle section are termed intermediate in this dissertation and include foremen, technicians, skilled labours and those in well-paid white-collar jobs. The lowest paid group are the supervised, made up of unskilled workers, agricultural labours and those poorly-paid in service jobs. Colour blocks of streets, by the proportion of the households whose heads belong to these categories: blue, yellow, and red respectively, and you can see one of the most basic divisions of the social landscape — the geography of class (Print LXXVII, Humpherys G. 1968, Goldthorpe J.H. & Hope K. 1974, Leete R. & Fox J. 1977, Boston G. 1980, Beacham R. 1984, Congdon P. 1989).

The colours do not mix. The blue (professional) lace-work threads its way around the city suburbs, strongest in the South63. The red (supervised) masses mark out the centres of major settlements, while the yellow (intermediate) patches show the distribution of relatively well-paid workers between the two: the coalfields of Wales and the North, for example (this is before most of the miners lost their jobs). London is a city split between the most and least rewarded workers, with little room in between.

Smoothing the picture (see Print CLXXVII) makes it easier to form some generalizations from these pictures. It can be justified, in this case, because the information on occupation is only available from a sample of one tenth of the population. Smoothing evenly over population space on the cartograms averages people with their nearest neighbours. But the technique must be used sparingly if it is not to provide false conclusions. It should also be remembered that it is only the use of the population cartogram which allows the most poorly-paid third of the population to appear in the picture (Print LXXVIII, Figure 13).

There is more to what people do than their general activity and occupation. What are they doing it for? They are doing it to grow things, make things, sell things and think about things. They are doing it for industry. People not only gain employment because of who they are and what they can do (Print LXXIX), but also because of the area of the economy they can do it in. Like everything else, industry is geographically distributed, that distribution being important to the fundamental geology of the social landscape (see Appendix E).

Industry can be divided in many ways, for example into primary, secondary and tertiary, or into public and private service sectors and the remainder. Its distributions can then be painted (see Print XXIII). These are now the distributions of where people work, rather than where they live, and this is an important point I address later in Chapter Six. Instead of mixing colours, a geology type classification has been adopted here, showing which industry has a majority of the workforce in each area. This has the advantage of further possible subdivision into dozens of industrial classifications, using subtle shades of the basic hues, while also showing how other forms of colouring can be gainfully employed. The picture produced shows the clear divisions between the sectors and the strong geographical patterns, which underly many of the images already presented, and those still to be seen.

Finally, the kind of work people do affects the sort of home they will have. Price can indicate the quality of a house, as well as, perhaps, the inflated and depressed states of local markets. It must be remembered that examining housing that is for sale only illustrates the distribution of privately owned provision of homes. Average housing price has been estimated and plotted for wards from a sample of building societies’ sales (Print LXXX). The geographic patterns of yearly inflation and local housing sector structures are investigated in Chapters Five and Eight. For now, the close correspondence, and important differences, between this picture and the others are all presented for consideration64.

4.6 How People Vote

We have seen something of who people in Britain today are, where they come from, what they do, and for which industries and what rewards it is done. The next theme in this thesis is concerned with what they do about it — how they vote. The British state regularly asks its inhabitants for their opinion on its government, through elections of candidates representing political parties standing for particular issues. As the choice is usually only between two or three regular parties this expression is extremely limited.

Geographically such a system produces areas which either support the party, and generally the order of the day, or the main opposition, or a third major alternative65. During the 1980s these were the Conservative, Labour and Liberal parties, which adopted the colours blue, red and yellow to represent themselves. In the early 1980s the Liberal party allied with a newly created Social Democratic party which survived for a decade. Now mostly merged, they are together called the Liberal Democrats. The term Liberal is used wherever possible in this work to avoid confusion. Other parties, such as the Nationalists, Unionists and Greens show interesting, but not particularly influential, geographical distributions.

Study of the human geography of voting is thus well suited to our means of visualization (Prints LXXXI & LXXXII). Cartograms give people the equal representation their votes are worth, and the graduated three colour scheme encompasses most eventualities. Although only the winning candidate holds a seat in parliament, the degree of support and nature of the opposition are also relevant both now and to indicate possible future trends (Prints LXXXIII & LXXXIV, Hollingsworth T.H. 1964, 1966, Roberts M.C. & Rumage K.W. 1965, Prescott J.R.V. 1969, Johnston R.J. 1979, Madgwick P.J. & Balsom D. 1980, Dunleavy P. 1983, King A. 1986b, Swaddle K. & Heath A. 1989, Upton G.J.G. 1991b).

The votes in national (general) elections are only reported for very large areas containing sixty or seventy thousand electors. Whilst being pertinent events, and the only complete record of the people’s (who vote — Print LXXXV) actual wishes for government, the fine detail of local opinion (at which we know everything else about our social landscape) is lost.

Local elections follow a complicated system of timing and are not all simple, one candidate, decisions. They do, however, give us information, at the relatively fine level of about three thousand electors, nationally. County council and Scottish regional elections, however, are based upon a different, very poorly defined, geography. Therefore we must rely upon the results of District elections to see the spatial mosaic (Rowley G. 1965, Clark D.M. 1977, Eagles M. & Erfle S. 1989, Rallings C. & Thrasher M. 1988, 1989, 1990a, 1990b, Bowler S. 1991).

The national picture of voting emphasises the divisions seen earlier in the social landscape66. In general people in areas of high unemployment, recent immigration and older industries vote Labour, whilst the rest of the country is dominated by the Conservative party, closely followed by the Liberals (Print LXXXVI). The geography of political party support accentuates the differences between people in areas, for it is through these parties that people are allowed to register their support or condemnation of the social system within which they are placed.

4.7 The Social Landscape

The social landscape of Britain has now been built up and presented through a dozen colour images. This is the landscape which people know, the landscape which is made up of, and determines, many aspects of their lives. It is the landscape of neighbourhoods, communities, blocks, streets, groups, ghettoes, villages, suburbs, housing estates, life chances and constraints. It is the landscape of age, work, class, immigration and race. It is the landscape of social existence, political power and economic opportunity — the human geography of Britain.

There is, however, much more that could be studied using these methods, adding to the montage. The geography of health — people’s life expectancies, disease and disability; the geography of welfare — the payment of benefits, the provision of services; the geography of privilege — the distribution of ownership of property and shares in industry; the geography of income and wealth.

The fundamental mountains and valleys of the social landscape of Britain would not be gravely altered by the addition of such material67. Here and there a new aspect would come to light, a different impression form. Overall, though, as with the incremental addition of the information already gathered, further insight is far more likely to reinforce our impression of a divided land. We know where most ill-health would be (the old and poor), where the social security benefits would be paid, where the rich would be found and the owners of industry concentrated. As you draw more and more of these pictures, you begin to recognise the same, familiar features over and over again in the social landscape68. So much is so strongly inter-related that it becomes the exceptions — which point to how things can be different — which are worth looking for (White D. 1985, Goddard J.B. & Coombes M.G. 1987, SEEDS 1987, Town and Country Planning Association 1987, Wilsher P. & Cassidy J. 1987, Green A.E. 1988, Owens J.R. & Wade L.L. 1988, Lewis J. & Townsend A. 1989b, Martin R. 1989).

This part of the thesis now moves on to ask: now that we can see the landscape, how is it changing? Then, what movement keeps it alive and what alters it? These are questions which increasingly stretch the limits of the visualization methodology. In the final part of the work the problem of envisioning the history and geography simultaneously will be addressed, leading us to the integration of all this information to create a consolidated image of British social history and geography at the end of the twentieth century. Time and space intermingle. The pictures I have presented up to now show the situation at a single point in time, but those of immigration and birthplace also tell of a different past.

The three themes of migration, work and voting will be integrated as the thesis is developed. Migration — where people come from and where they are going; work, what people do, when they can do it, and voting — what people want, and when, how and where they get it. Human geography must integrate the study of the populations which it first, so finely, spatially separates.



LXVIII The distribution of age and gender in Britain, 1981 (Colour).
LXIX The concentration of age and gender in Britain, 1981 (Colour).
LXX The distribution of children by age in Britain, 1981 (Colour).
LXXI The distribution of Irish bom in Britain, 1981.
LXXII The distribution of British born place of birth, 1981 (Colour).
LXXIII The distribution of Overseas born place of birth, 1981 (Colour).
LXXIV The concentration of Overseas born place of birth, 1981 (Colour).
LXXV The distribution of employment in Britain, 1981 (Colour).
LXXVI The concentration of employment in Britain, 1981 (Colour).
LXXVII The distribution of occupation in Britain, 1981 (Colour).
LXXVIII The concentration of occupation in Britain, 1981 (Colour).
LXXIX The distribution of graduates in Britain, 1981 (Colour).
LXXX The distribution of housing price in Britain, 1983.
LXXXI The distribution of voting in the 1987 British general election (Colour).
LXXXII The map of voting in the 1987 British general election (Colour).
LXXXIII The distribution of first placed parties in the 1987 British general election (Colour).
LXXXIV The distribution of second placed parties in the 1987 British general election (Colour).
LXXXV The distribution of non-voting in the 1987 British general election.
LXXXVI The distribution of voting composition in the 1987 British local elections (Colour).


The small area statistics of the 1971, total population census, consisted of 480 figures for each of 125,476 enumeration districts, some sixty million numbers. Originally stored in eight character wide slots the file was half a giga-byte in size, far too large to be easily stored and repeatedly accessed. A simple form of run length encoding was customised to compress the file and still allow the records of individual enumeration districts to be read instantly. The counts for each cell were stored sequentially as either a run of zeros, half-bytes (0-15), bytes (0-255) or half-words (0-65535). The sophistication of the algorithm was, in deciding whether of not it was profitable to "drop down" an order of magnitude in the form of storage used. This was achieved by looking through the list both forwards and backwards. The following simplified heuristic was employed:

Define: yesterday, today and tomorrow as the magnitude of the previous, present and future cell to be encoded.
Then, if the opportunity to lower the magnitude of storage arises (today<yesterday) continue at the present order while tomorrow>=yesterday.

With a few other caveats, this rule compresses the file to just 5% of its former length: 29,861,010 bytes. The more sparse first section of the 10% population census file containing 368 cells by 125,462 enumeration districts (14 missing) is compressed to a file of just 11,314,567 bytes in size. These figures are better than those achieved by the standard Lempel-Ziv compression algorithm, but more importantly, the file can be read and decoded faster than any other configuration (including the original flat form, due to disc speed restrictions).

Figure 11: Storing the Census

The occupation groupings used in this dissertation are defined by OPCS (1981, pp.24-29). The three combinations chosen (using the New Earnings surveys of 1971 and 1981) were of socio-economic groups:
1: Managers in central and local government.
2: Managers in industry and commerce.
3: Professional workers - self employed.
4: Professional workers - employees.
13: Farmers - employers and managers.
5: Ancillary worker, artists, foremen and supervisors.
8: Foremen and supervisors - manual.
9: Skilled manual workers.
12: Own account workers (other than professional).
14: Farmers - own account.
6: Junior non-manual workers.
7: Personal service workers.
10: Semi-skilled manual workers.
11: Unskilled manual workers.
15: Agricultural workers

The industrial groups were taken from the following amalgamations of 1980 Standard Industrial Classification based codes, referred to as "Broad Industrial groups":

Code  Description  NOMIS Class
1: Agriculture, forestry and fishing
2: Energy and water supply
3: Manufacturing industries
4: Construction
5: Distribution, hotels/catering; repairs
6: Transport/communication, banking, finance
7: Public administration and defence
8: Other service industries

Figure 12: Working Definitions

In some prints in this dissertation the pixel-maps have been smoothed by several passes of a binomial filter. In one dimension it can be written as (1/4, 1/2, 1/4) and dissipates the intensity of a pixel with a value of 1 by the following intensities after the first five passes:

The two dimensional version of this filter is given by the following matrix (after Tobler W.R., 1969):

After approximately ten passes this filter is equivalent to the effect of a normal kernel with variance n/4 (where n is the number of passes). This is one of the simplest and most elegant forms of spatial smoothing. It is also, interestingly, reversible (although this is only practical after one or two passes). Its inverse may presumably also be used to sharpen an image.

Figure 13: Two-dimensional Smoothing


54 [a] Although statistics may be collected to be mapped:

The graphic portrayal of census data has always been a decentralised and in many respects an adhoc affair. After the superb maps produced by Petermann (partly for the government) after the 1841 and 1851 censuses, little ‘official’ mapping was done until that carried out after the 1961 Census, by what is now the Department of the Environment (DOE). A tradition grew up that individual geographers mapped those elements of the census in which they were interested and in 1968 one of the Transactions of the Institute of British Geographers consisted of a set of twelve maps of variables from the 1961 data. This 7-year delay in map availability was very similar to that after the 1841 census. [Rhind D. 1975 p.9]

[b] The practical problems of doing so continue:
The statistician who compiles data about the present aims to provide primary data for constructing a very large number of conceivable comprehensive pictures. Among the mass of figures collected for this purpose are concealed an infinitely large number of latent pictures. To transform these latent pictures into actual ones, to present them in such a way that they even suggest the basic structure of the other images still concealed among the mass of figures — this is one of the most important tasks of the field called census cartography. [Szegö J. 1987 p.149]

55 [a] There are many reasons for mapping census data:
Mapping census data is potentially an extremely effective way of communicating information for tens, hundreds or even thousands of different areas through the medium of one (sometimes small) piece of paper. The extreme example of census mapping is probably the national maps in People in Britain, the census atlas based on 1971 grid-square data: these maps contained between 50,000 and 150,000 areas on each A4 size page. It is important to emphasize at the outset, however, that all such ‘statistical’ maps are a complement to, rather than a substitute for, statistical tables. Moreover, as we shall see, there are many different ways of mapping census data, some of which give misleading results. Mapping, then, can be extremely informative but it can also mislead. [Rhind D. 1983 p.171]

[b] People have been using and mapping census statistics for many years:
For a number of years the American Statistical Association has provided for annual meetings of specialists in the use of census tract statistics. These meetings have brought under one roof investigators and administrators concerned with such matters as “studies of disease, city planning, marketing analysis, labor market studies, civil defence, church planning, studies of juvenile delinquency, housing problems” and “retailing”— to cite a list presented in a recent publication on census tracts (United States Bureau of the Census, 1958, pp.4-5). Political scientists have found use for census tract statistics and methods of urban analysis applied thereto in their studies of voting (e.g., Gosnell and Schmidt, 1936), and even psychologists have found the classification of urban areas by elaborate techniques an interesting exercise (Tryon, 1955). Foley (1953) gives a lengthy list of census tract studies. [Duncan O.D., Cuzzort R.P., & Duncan B. 1961 p.13]

56 [a] There were slight discrepancies between the detailed 1981 census statistics and preliminary report:

Population present — preliminary Report figures. The population present in an area on census night is straightforward to count. The preliminary figures were prepared by the enumerators, collated manually and published within three months of census day. The count for England and Wales was 49.01 million. [Population Statistics Division, OPCS 1983 p.21]

[b] Those who produced enumeration district statistics never assumed they could be mapped nationally:
Thus at the ED level, 10 per cent SAS tables are subject to large errors and will generally need to be aggregated to much higher area levels to ensure small variability in the cell values. The statistics are presented at ED level, primarily to allow flexible aggregation. [Denham J.C. & Rhind D. 1983 p.80]

[c] The 1971 census figures contained some particularly obvious inconsistencies:
The census user can, however, experience problems as a result of adjustment, particularly if he calculates ratios from adjusted figures. He may, for example, find that the percentage figure for a set of categories — age groups or occupation groups — may add up to more or less than 100 per cent of the total population. [Dewdney J.C. 1983 pp.10-11]

57 [a] The essential problem of lack of space in mapping persists:
The main problem in cartography is the counterpart of these very properties. In a diagram, the geographic component AB... only utilizes a single dimension of the plane. The other dimension remains available for transcribing n characteristics. In a map the component AB... constructs a network that utilizes the entire plane, in fact, accounting for the map’s effectiveness. But the y dimension of the plane is no longer available for the representation of the characteristics, so we must choose between two solutions:
-either construct one map per characteristic. In this case the map answers two types of question: Where is a given characteristic? What is there at a given place?
-or superimpose all the characteristics on the same map. But then the question: where is a given characteristic? no longer has a visual answer. Should this question indeed have an answer? This is the basic problem in cartography with n characteristics, that is, “thematic” or more precisely, “polythematic” cartography. [Bertin J. 1981 p.140]

[b] It is the deluge of new digital information which has lead to the recent need for visualization:
While computers are quite adept at the minutiae of computation, the human is far more capable, in dealing with global information, to work as a “Gestalt” recognizer. Hence we may think of the human-computer symbiosis as a process whereby the machine fashions and places mosaics of information (the pixels) so that the human can form an understanding of the vista being worked. It is the human brain that will then visualize the process — “see” the drift of a complex calculation. The computer is used to aid in visualization of anything. Webster’s describes visualization as: 1. formation of mental visual images, or 2. the act or process of interpreting in visual terms or of putting into visible form. [Staudhammer J. 1987 p.24]

[c] The census small area statistics provided a wealth of information, initially incomprehensible:
Quite simply there is far too much information to allow policy-makers, planners, geographers, politicians, schoolchildren, and others interested in census data for a particular area to be able to identify easily patterns of characteristics or features of interest from SAS data without processing and condensing it in some way. [Openshaw S. 1983 p.243]

58 [a] The quartile technique has been adopted by other researchers:
The modified quartile classification was developed only after consideration of the purpose for which the maps would be interpreted. It was anticipated that many different questions about socioeconomic conditions would be addressed to each choropleth map. The potential for answering such questions is maximised when the map pattern is balanced, that is when the area occupied by each symbol is approximately equal. For example, a map with five symbols is balanced when each symbol occupies approximately one-fifth of the pattern. As a map pattern is determined by classification, a balanced map pattern is most likely to be obtained when an equal number of LGAs is allocated to each class. [Massey J.S., O’Shea J.B. & Poliness J.S. 1984 p.286]

[b] It is interesting to note that here we are mapping those things which are said to affort the nature of mapping itself:
To discover these rules, we have to read between the lines of technical procedures or of the map’s topographic content. They are related to values, such as those of ethnicity, politics, religion, or social class, and they are also embedded in the map-producing society at large. [Harley J.B. 1989 p.5]

[c] Divisions of class and race can be seen through location; they are not aspatial as some suggest:
Of course, there are problems other than those raised by the north-south divide facing Mrs Thatcher, the government and the country: other divisions between people of different class, race and sex that are actually more fundamental than those of location. [Lewis J. & Townsend A. 1989 p.4-5]

59 [a] The processes ordering the distribution of the sexes are closely interrelated:
Retirement areas (predominantly coastal areas) are peculiar not only because their proportion of elderly population is high but also because they are associated with high female sex-ratios (Clarke, 1960). This is partially due to the longevity of women and the relatively rich female employment opportunities generated in these areas by tourism and service provision for the elderly. [Kennett S.R. 1983 p.227]

[b] The use of many small units can be repeatedly justified from the errors that manifest when they are not employed:
An odd consequence of the redrawing of county boundaries in 1974 was that Lancashire became an area of concentration of the elderly. [Warnes A.M. & Law C.M. 1984 p.40]

60 [a] Monitoring migration is an age old preoccupation of the British:
Only four pieces of information were collected about each person in the 1841 Census. That one of these was birthplace is indicative of how essential this item was and still is. [Craig J. 1987 p.33]

[b] Even the enumeration district level may be too coarse to see some spatial distributions:
A specific and simple illustration of some of the advantages of microdata can be given by considering the distribution of the Irish-born in Tow Law, County Durham in 1871. They constituted some 10 per cent of the town’s population, distributed amongst the five EDs as shown in table 11.1. while this table reveals some concentration in ED9 and an under-representation in ED10, it does not suggest that the Irish in Tow Law are heavily segregated. However, if we consider the proportion of Irish living in each street, a different pattern emerges, as table 11.2 shows. [Norris P. 1983 p.313]

[c] Here we examine the distribution of all three major groups of immigrants simultaneously:
The emphasis of immigrant community research by British geographers and other social scientists during the past twenty years has been overwhelmingly on the Afro-Caribbean and Asian groups at the expense of those of longer standing and greater numbers, but perhaps of less visibility. [King R. & Shuttleworth I. 1989 p.64]

61 [a] It must be remembered that we are mapping place of birth, not colour of skin:
For example, of the 322 670 persons born in India living in Britain in 1971, between one-fifth and one-third (66 139-104 362) may have been Whites born in India (Peach and Winchester, 1974, p.391). [Peach C. 1982 p.24]

[b] Without mapping we are more prone to make mistakes:
The study of birthplace characteristics identified two major types of immigrant group. The first composed of the Irish, the Other Commonwealth and Other Migrants have similar patterns to the British born. For the latter two types of immigrant, although the differences between regions and between urban zones are small compared to the British born, there are much larger increases in all areas. The second cluster of immigrant types — the Indian sub continent, Africa and West Indies — have markedly different patterns with respect to the British born, especially with respect to intra urban variations. [Spence N., Gillespie A., Goddard J., Kennett S., Pinch S. & Williams A. 1982 p.277-278]

[c] The less aggregation the better:
Further shortcomings exist in census data relating to ethnicity. Dissimilar birthplace groups are frequently aggregated into a single category: for example, all those born in the American New Commonwealth (chiefly the Caribbean) are usually grouped together in the published statistics. More seriously, several cross-tabulations in both 1971 and 1981 SAS group all New Commonwealth-born together. Prandy (1980) has demonstrated that the ‘social distance’ between Asian and West Indian groups living in Britain can be as great as that between either of these groups and the British-born. [Ballard B. & Norris P. 1983 p.105]

[d] In the geography of migration student population should be kept in mind:
A third factor contributing to the large inflow into the South East is that students make up about 15 per cent of all immigrants and London is popular with overseas students as a place of study. The first and third of these factors go some way to explaining the larger than average outflow from the region. Outside the South East, the West Midlands and East Anglia were the most attractive areas for immigrants, relative to their populations. The relatively least attractive place for immigrants were the North of England and Northern Ireland (though it must be remembered that the figures take no account of immigrants from the Republic of Ireland). [Davis N. & Walker C. 1975 p.5]

62 [a] It may, perhaps, be surprising to learn that in the 1980s:
Although the percentage officially unemployed in Greater London is a little smaller than average for Britain the city holds the largest concentration of unemployed in the industrialised world, and the real total is at least 150,000 larger than the total of over 400,000 admitted by the Government. [Townsend P. with Corrigan P. & Kowarzik U. 1987 p.29]

[b] London is clearly a sharply divided city:

Eversley and Begg’s (1985) nation-wide study of deprivation indices for urban areas, undertaken for this research programme, shows that on a wide range of indicators there is a steep gradient in conditions between inner, outer, and fringe areas of London. [Buck N., Gordon I., Young K., Ermish J. & Mills L. 1986 p.12]
[c] Inequality is the crucial ingredient of deprivation:

We shall hold that the most severe deprivation exists where the scores of disadvantage are high, where they affect the largest number of people, and where there is the most crass contrast between these areas and the advantaged periphery. [Begg I. & Eversley D. 1986 p.55]

[d] Other conurbations also exhibit sharp divisions:

The difference between inner Birmingham and the West Midlands southern fringe is 3.80 — the steepest in the country. Less than 10 miles separate some of the worst conditions in the country from some of the best. [Begg I. & Eversley D. 1986 p.75]

63 [a] The constitution and aggregation of classes is a contentious issue:
As BRAVEMAN, 1974, has pointed out, however, ‘The traditional distinctions between “manual” and “white-collar” labour, which are so thoughtlessly and widely used in the literature on this subject, represent echoes of a past situation which has virtually ceased to have any meaning in the modern world of work’ (p.325). [Hamnett C. 1986 p.393]

[b] There are many ways in which people can be grouped:
Nevertheless, we have avoided lumping together both intermediate non-manual workers and skilled manual workers into a ‘new middle-class’. Although some commentators claim to have perceived either the ‘embourgeoisement’ of skilled manual workers through rising incomes, or else the ‘proletarianization’ of the white-collar workers through increased trade union affiliation and ‘militancy’ in the labour market, the bulk of evidence indicates that there are substantial and persistent differences in material rewards, status and life styles between manual and non-manual workers (Roberts et al., 1977; Westergaard and Resler, 1975). We have, nevertheless, avoided the mistake of creating a blanket non-manual category by distinguishing the professional worker from the intermediate clerical strata. [Pinch S. & Williams A. 1983 p.138]

[c] The grouping used here is a similar to one used by Hamnett:
It makes little sense to aggregate such divergent groups and tendencies together and in the analyses which follow, SEGs 12 and 14 are treated separately from SEGs 8 and 9, on the grounds that they have more in common with SEGs 1, 2, 3, 4, 5 and 13 than they do with 8 and 9. Similarly, SEG 6 is analysed together with SEGs 7, 10 and 15 on the grounds of skill levels, renumeration and intercensal comparability. If this is not done, any comparison over time, let alone a sensible and meaningful comparison, is indeed virtually impossible. [Hamnett C. 1986]

[d] Social group affects many aspects of the quality and stability of life:
Insurance coverage varies greatly by socio-economic group (SEG), with 23 per cent of the professional and managerial SEGs benefitting from it, while only 2 per cent of semi- and unskilled manual SEGs are covered; in the 45-64 age group, over 31 per cent of the professional and managerial SEGs are covered. [Curtis S. & Mohan J. 1989 p.187]

[e] You will not find class structure at the city scale:
In only one of the largest cities (Liverpool) did the proportion of semi- and unskilled exceed the national average by more than 4 per cent. If concentrations of the most disadvantaged have occurred as a result of selective decentralization then it would appear to exist at a more localized level within cities. [Goddard J.B. 1983 p.12-13]

64 [a] If we cannot decide which aspect is most important, why not look at several?:
Social class in the sense of status of individuals in the labour market, may today be as well reflected by position in the housing market as by necessarily imprecise occupational labels. [Buck N., Gordon I., Young K., Ermish J. & Mills L. 1986 p.101]

[b] The distribution of housing is intricately connected to many of the other patterns shown here:
The results of this study strongly support the argument (Cheshire, 1979) that inner-city unemployment is not so much a problem of the performance of the city labour market as a whole, but a feature of the other sifting mechanisms in society, mainly the housing market, that concentrate people who are at a competitive disadvantage in society into relatively restricted areas. [Frost M. & Spence N. 1981 p.100]

65 [a] An atlas of British election results was recently criticised as:
The publication is entirely in black and white which is a little disappointing in view of the colour association with the major political parties. All the thematic maps use the area shading technique. This is a conventional technique for such maps but has the weakness that it places greatest emphasis on areas of sparsest population. The effect is doubly unfortunate in that political tendencies often relate closely to population density. Thus supporters of the Labour party might feel aggrieved by the visual diminution of their achievements. A demographic rather than a topographic base would overcome this problem but would almost certainly confuse the general public. Nevertheless simple bar graphs could have been used to complement the maps and avoid any mis-interpretation. [Beard R. 1989 p.172]

[b] There are two basic electoral distributions to consider, who wins, and how people vote:
Finally it should be noted that what polls attempt to measure is the distribution of party support among the electorate. Extrapolating from this to the distribution of seats in the House of Commons is a tricky business which is becoming trickier. In February 1974, for example, the party which won most votes (the Conservatives) did not win most seats. In 1987 ITN seriously underestimated the likely Conservative majority in the House of Commons despite the fact that its exit poll, conducted by Harris, got the Conservative lead over Labour in terms of vote share almost exactly right. [Denver D. 1989 pp.106-107]

[c] Political bias can be in either direction:
The January 1910 election illustrates this situation very clearly. The Irish Nationalists won 82 seats, all but one being located in Ireland. This extreme ‘peripheral’ concentration is reflected in the U-shaped voter proportion distribution (Figure 4(a)) with its very large variance. Thus with 1.9% of the vote the Irish Nationalists were able to secure 12.2% of the seats to enjoy the positive bias of 10.3%. The more recent experience of the Liberals has been a sharp contrast to this situation. [Gudgin G. & Taylor P.J. 1973 p.18]

[d] The local distribution of class is almost identical to that of local voting:
Since 1945, occupational class has been widely seen as the main social basis underlying electoral politics in Britain. A pattern of ‘class alignment’ was clearly apparent in the 1950s and 1960s. [Dunleavy P. 1983 p.32]

[e] The importance of class to voting is widely acknowledged:
The dominant alignment, or cleavage, in British electoral politics is class — employed loosely as a shorthand term for position in the division of labour. [Johnston R.J. 1986 p.574]

[f] The distribution of occupational class is the best predictor for the political composition of an area:
But at the level of explaining why particular areas or constituencies vote the way they do, knowing the mix of occupational classes in the local area continues to be as valuable as ever in explaining or predicting election results. [Dunleavy P. 1983 pp.37-38]

66 [a] Some claim that local elections are too different to be considered along with the national elections:
Because of its significant effects there is little need to justify the analysis of constituency level voting, but the same cannot be said for the study of other levels of voting. [Miller W.L., Raab G. & Britto K. 1974 p.391]

[b] But others state:
‘You can no more take politics out of government than you can keep sex out of procreation.’ [Gyford J., Leach S. and Game C. 1989 p.1]

[c] Constituencies may well represent too high a level of aggregation:
Nevertheless, many remain sceptical that there is indeed a geographical component to the vote or, indeed, to any other kind of behaviour. Some critics of these approaches ground their attack in variants of the ‘ecological facility’. McAllister, in particular, doubts that the processes at work can operate at the levels of aggregation that are sometimes chosen (see the Johnston and McAllister debates in this journal). [Bowler S. 1991 p.92]

[d] Many writers have suggested that constituencies are so large that they blend together the diverse patterns which actually exist:
To the extent that there are processes of political persuasion at work these are much more likely to be at a very local scale rather than that of constituency [Agnew J.A. 1987 p.99]

[e] Constituencies cannot constitute communities:
Despite the fact that both cohesion measures consistently appear to make a statistically significant contribution to the explanation of turnout levels, it must be acknowledged that the magnitude of their impact on turnout is disappointingly small. This may be because constituencies are too large and internally differentiated adequately to represent communities; consequently, communal influences are probably underestimated using constituency level data. [Eagles M. & Erfle S. 1989 p.125]

[f] Local elections provide us with a more precise picture:
Third, it is essential that any analysis should be conducted at the smallest electoral level possible. There is no excuse for using constituency figures if returns are also shown for subdivisions of electorates, or for individual polling booths; the figures related to smaller areas will make it possible to draw the boundaries of voting regions with much greater precision. [Prescott J.R.V. 1969 p.381]

[g] The whole is formed from its parts:
Taken together it is hoped that this book demonstrates the truth in the words that for political behaviour, “It is the local reality that determines the total picture, and not the reverse” (Granata 1980: 512). [Agnew J.A. 1987 p.6]

[h] But these small localities have been largely ignored:
Much work remains to be done. Local elections have received relatively little attention.... [Busteed M.A. 1975 p.54]

67 [a] The image we gain of the social landscape depends far more on how we draw it than on what it contains:
Therefore, rather than think in terms of a simple division between South and periphery, it may be necessary to think of a threefold division into periphery, South and Midlands, and the South East excepting London. Even this division may be too simplistic. [Pinch S. & Williams A. 1983 p.155]

[b] Some of the patterns found here were previously gleaned from more conventional maps:
In terms of geographical patterns there is in 1981 a clear distinction between the places at the two extremes (Figure 7.1A). Most notable is the ring of most privileged LLMAs around London, extending to the South Coast and forming a virtually complete arc on the other flank; the only exception being along both sides of the Thames estuary. There are also significant clusters of better-off LLMAs further westwards along the South Coast and in southern parts of the West Midlands. The prosperity of the South Coast can be gauged in terms of the fact that south of a line between the Severn Estuary and Lincolnshire there are only three representatives of the lowest quintile on this indicator, namely Corby, Spalding and Wisbech, and, of Britain’s bottom 112 places, the South accounts for only 11, all but one of which are located on the margins of the region in East Anglia and the East Midlands (Figure 7.1A). [Champion A.G., Green A.E., Owen D.W., Ellin D.J. & Coombes M.G. 1987 p.91-93]

[c] But the impression presented by conventional mapping often needs to be corrected:
Conclusion: there are an awful lot of poor areas in the South East (especially inner London); and there are a lot of very affluent suburbs around the North’s big industrial towns. Which shouldn’t come as a surprise — but may act as a corrective to the blinkered south-easterner’s view of a vast industrial wasteland north of Oxford. [King A. 1986 p.18]

[d] Diversity and variation are everywhere, yet an order prevails:
In short, a picture of local variation is revealed, here in terms of the advantages enjoyed by some successful northern areas, but more generally reflected in the wide diversity that exists across Britain in many criteria (eg Fothergill & Vincent, 1985). [Goddard J.B. & Coombes M.G. 1987 p.13]

[e] The concentration of deprivation is greatest in the conurbations of population, but in the more populated South of Britain; on a conventional map, this can be overlooked:
Rather, it is claimed that the concept of a North-South divide in Britain is valid despite the existence of local variations because it is demonstrably the case that the concentration of relative socioeconomic deprivation and disadvantage is significantly greater in the ‘North’ than in the ‘South’ (Martin, 1987b, p.573). [Green A. 1988 p.194]

[f] Which divisions matter most, the sharp lines in cities, or the shallow slopes between regions?
And as is well known, the degree of spatial inequality in the socio-economy is scale-dependent; measured differences tend to increase as the geographical coverage of the areal divisions employed decreases. The debate is not just over the existence or significance of local disparities, which can be found everywhere: the issue is also that these local disparities map out and form part of a broader ‘north-south’ geography of socio-economic inequality, and that this regional divide has become an increasingly prominent feature of British society. [Martin R. 1989 p.22]

68 [a] The pattern is most consistent when based upon smal scale populations:

The Victorians were more concerned with patterns found at a smaller spatial scale, and this is also reviving today: as we will see, despite the interest in the north-south divide, some writers see spatial differentiation taking place on a much smaller spatial scale between localities. [Savage M. 1989 p.248]

[b] A picture of “London and the rest” is clearly seen:
In considering the changing social structure of the principal cities of the nation the basic distinction is between London and the rest. Only in the London core does the proportion of the economically active population in managerial and professional occupations (17.7%) and in the intermediate non-manual occupations (22.7%) exceed the national average for this zone. At the other extreme only in the capital does the proportion in the core of unskilled manual occupations fall below the national average. [Spence N., Gillespie A., Goddard J., Kennett S., Pinch S. & Williams A. 1982 p.275]

[c] But it is the divisions within regions which matter to people:
This, then, is the South-South divide. It is a divide which appears in employment opportunities, in wage packets, in each job’s content and potentialities. It reappears in the car park and the bus queue, in the green of the garden and the size of the room. Each part of the divide has its own daily timetable and its own life cycle. [SEEDS 1987 p.10]