Surname Distribution Studies

	Sources/Methods/Tools Preface Spatial Analysis Contemporary Geodemographics 1881 Census
	Of Boundaries and maps e-Maps Further info Printed maps
	The study of English local surnames Guppy's 'local' names

If you came to this page directly, then please access
Modern British Surnames

Brief Checklist

Not an exhaustive list; just the basics

Preface

Before plunging enthusiastically into this topic, the following preface of the case against, might be salutary.

A) Variant or Not?

In the 1960's, and using the GRO birth indexes for 1850, Francis Leeson mapped the distribution of his name and what he regarded as its variants -Lee, Lees, Leigh, Leigh, Lea, Ley, Leese, Leeson, Leason. The resulting plots revealed discrete areas, which remained the same even when compared with a 1960's telephone survey..

A reply was made to this article by Dr Reaney, who criticised the distributions on several fronts:-

1) A plot of the modern spelling does not necessarily equate with the original form or distribution "Lee,Lea, Ley, Lay and Leigh are all one surname. They all go back ultimately to OE leah and both surnames and place-names have a variety of forms; the different modern spellings may be partly due to ME grammar, partly due to the local dialect or simply to mere chance...Parish Registers did not begin until long after surnames becam fixed; they are not necessarily proof of the original distribution."
It would have taken maybe just 1 fertile family to migrate in 1530, to give a false impression of the home of a name. Especially if that name did not appear elsewhere in mediaeval documents.

2) A plot cannot be made comparing a root name and its variants, unless one is totally sure that the supposed variant did derive etymologically from the root name. Reaney points out, that in his opinion, Leese (from OE laes- 'pasture') and Leeson - derived from 'son of Lece'- are not variants of the root names Lee, Lea, Ley, Lay and Leigh.
Reaney has subsequently been criticised for over-reliance on etymology, but I think his general points should be borne in mind by anyone plotting any kind of surname distribution.

B) Surname corruption

George Redmond in his study of Yorkshire surnames has shown the amazing variability of surnames.
"In addition to the obvious variations associated with the distortion of vowel sounds and the confusion when pronouncing consonants, the author draws attention to the remarkably high incidence of elision and truncation, as well as the introduction of so-called prosthetic consonants such as Y, W or S to preface some surnames beginning with a vowel. He also notes that the final consonant of a first name may transfer to the surname, citing Thomas Anderson alias Saunderson and John Nellis alias Ellis."
Book Review in The Escutcheon of - Surnames and Genealogy
Dramatic changes could also occur to the final syllable of surnames. For example -Whithalghe/Whitalk/Whitack and Astmough/Astmall/Asman/Asmond. Surnames such as these seem to have had very little stress on the final syllable - it was left to the listener to decide their own interpretation -often in perpetuity.

If you collect the occurrences of a name from say the Hearth Tax, how do you know that the name is what you think it is? -unless one investigate the genealogy of each bearer.
Surname dictionaries will be of little help, because they tend to ignore local corrupted forms. Surname dictionaries concentrate on the earliest form of a name : surname corruption comes much later

As George Redmonds says, each occurrence of a surname should be treated as being unique.

End of the cautionary preface

Snapshots

Part 1 - Contemporary

Some of the potentially really useful national and comprehensive sources are inaccessible to us - such as the National Health Service Central Register at Southport or the Social Security Central Register at Newcastle upon Tyne.

Plotting by postcode

• The ONS has an introduction to postal geography, which it is advisable to read
• Postcode Statististics (usual resident population) for England and Wales are available in the CD-Rom which accompanies Census 2001: Key statistics for postcode sectors in England and Wales / Office for National Statistics. - London : TSO, 2004. - 0116217499
• There are 124 postcode areas in the UK (120 for EWS, + BT (Belfast), JE (Jersey), GY (Guernsey) + IM (Isle of Man)
• The usual resident population for a postcode area varies from 0.04% to over 3%
• There is a published Atlas of Postcodes which defines boundaries, though the finer ones are subject to change
  Also see an online map of Postcode Areas-
  

The following represents a rough guide to the percentages of the Scotland/Wales/England population in each postcode area and the proportion that are aged under 18. Non-mainland postcodes not yet included are for Belfast, Jersey, Guernsey and the Isle of Man. The figures should be taken as a rough guide.They were compiled prior to the publication of the ONS Census 2001 postcode area figures- but still appear to be in line

Scotland

The population of Scotland at the 2001 Census was 5,062,011.
The population percentage of Scottish postcode areas oF 5,062,011 was about:-

UK Electoral Rolls on CDROM

Pluses

• As near a comprehensive coverage as you will get, listing 42 million names
• Ability to easily extract the number of occurrences of a surname
• Occurrences could be manually mapped using an Atlas of Postcodes

Minuses

• Up to 15% of the possible electorate may not be registered ¹(young people and ethnic minority groups are particularly under-represented)
• Extremely expensive to purchase (something like £3,000 as at 2001)
• Only lists those in the population aged 17 and over

Although the disk is expensive to purchase, there is a fee-based extraction service available from People Finders UK.

UK-INFO Disk

Pluses

• Cheap and cheerful
• Combines both telephone numbers and the Electoral Register
• May be useful just for a casual search alone
• On-line version available

Minuses

• Has been heavily criticised since its data is "of variable vintage" and for the number of missing entries
• It has been estimated that at least 50% of its entries are inaccurate in some way (although the later versions are on first reports much improved in its accuracy)

Up to now, the UK-Info disk could not be recommended for surname distribution analysis, where accuracy in the totality of numbers is so important. The latest disk seems at first to have a much better coverage as a percentage of the population. This is due however to the many duplications in entries caused by Postcode changes. Ensuring that the source is one of 'clean data' is vital in our study.

Telephone Directories

These now come in a variety of formats - Online, Cd-Rom, and printed. However, the telephone directory -whatever its format- suffers from a major proviso -the increasing number of unlisted telephone numbers.

"Although the national average for ex-dir is about 37% the figures do vary enormously between counties, being lowest in northern England and *much* higher in southern England.  So for any surname you will get perhaps 80% listed if they live in a northern county, but less than 50% listed in southern counties, especially East/West Sussex, Hampshire, Surrey, Kent etc.    This imbalance in ex-dir status can be significant in surnames with small numbers, but probably less so with the more common surnames." (John Wynn)
 

The latest online version -PhoneNetUk - is extremely disappointing for our purposes. A regional qualifier is mandatory (under the terms of the licensing authority) , so no national searches are possible. The inclusion of postal codes is erratic, and where they do appear are truncated to the outward code alone.
With the CD, national searches are allowable, but only the first 200 hundred entries are displayed (with full postcode). A tweak is possible to derive statistics of a surname by region, if the number of occurrences exceeds 200. A visit to the local library will probably be required to consult the printed telephone directories.

Colin Rogers has listed the disadvantages of using printed telephone directories:-

• They are not issued simultaneously, and therefore cannot provide a national 'snapshot'
• Boundaries between directory areas are not constant. With the result that an individual's entry may be repeated in more than one directory.

He adds:-

"British Telecom has an Archives and Historical Information Centre at 2-4 Temple Avenue, London EC4Y OHL which is open to the public...it holds an almost complete set of telephone directories from 1879 when the first publically available system was introduced into Great Britain."

Mr Rogers is sceptical about the usefulness of pre-1950 telephone directories for our purposes; the coverage of the population being so small. However, they might be useful as pointers for the study of relatively high frequent names.

National Health Service Central Register [NHSCR]

This database of 60 million names is not available in its entirety - but you can look at an individual frequency. The NHS Central Register is prone to list inflation, and some of the results are surprising, so treat with extreme caution. The whole database does have linguistic possibilities. For a paraphrased potted history of the NHSCR

Survey of Contemporary Surnames

Despite these limitations, a major and significant survey was conducted of the surnames of Britain, using the printed telephone directories 1980-1996. The survey was led by Patrick Hanks and Kate Hardcastle in order to establish those names deemed to be of significance for 'A Dictionary of Surnames' OUP, 1988. The result was 16,000 surnames with a frequency of more than 20 occurrences in any particular directory.
A full listing of the distribution of all the names can be found by following this link

International data sources
The publication of national telephone directories on CD has been used by geneticists to study isonymic rates for individual countries. Onomastic studies based on national datasets are much rarer, but hopefully will increase.

Part 2 - Censuses

1881 Distribution

The 1881 census transcription -despite its known faults- is a marvellous tool for considering the frequency and distribution of names in the late nineteenth century.

The Guild of One-Name Studies has done important work in establishing baselines upon which to commence a study of individual names. The following table of conventions is based on the work of the 1881 Project- co-ordinated by Geoff Riggs

The density is an important indicator. If a surname was evenly distributed it would have a density of 1.
Geoff Riggs shows in his articles that reliance merely on the number of occurrences (s) is a misleading indicator.

For example, below are the 1881 county figures for my own name :-

Surrey has the highest number of absolute numbers, but if one considers the density, then Herefordshire is the leading county, with Berkshire close behind. There is a wide margin to the next county, Wiltshire.
I found this surprising, as a contemporary survey indicates Berkshire as the main county, whilst the IGI favours Worcestershire. The name Dance seems not to have a discrete source, but seems to have arisen independently in several counties from Worcestershire, through Gloucestershire, Wiltshire, Berkshire,Hampshire.

But then, in surname distribution studies, nothing is often clear cut- as the following table of data arrranged by registration district reveals:-

generated with xls2html converter

The distribution is best seen as a map

This map shows the heartland of the Dance surname. Three foci can be discerned:- Marlborough (Wiltshire) Ledbury(Herefordshire) Bradfield/Wokingham (Berkshire) The map was created with Genmap v2 Figures are per 10,000 people

.

Marlborough is the 1881 Dance hotspot.Within the Registration District, the name is located in just 2 parishes:- Marlborough St Mary (174 on map) numbers-35 Preshute- numbers 12 Map created from the HDS Historic Parishes of England and Wales CD

GRO data

One-namers collect the GRO data for their name as a matter of course. An examination of members' pages on the Guild site reveals many excellent examples.

Professor David Hey has built a database from the death registrations for the years 1842-1846 for surnames beginning with the letters A,E,K and R. This has resulted in a computer database of over 220,000 surnames, covering an estimated 12.5 % of the whole set of surnames. He has published his results in his book Family Names and Family History

An examination of the GRO data for my name 1840-45 shows the main counties to be Berkshire, Hampshire, Worcestershire

Part 3

Spatial analysis

-Index numbers

In the section above, an example was given of the Density of a particular surname (my own). This could be applied to every surname in the national database under study, to produce an index value for each name. It is the norm, however, to express Index numbers around base 100. One can either multiply the Significance value (see above) by tha factor, or use the following equation to produce the same result

where

An index number of 200 would indicate that for that surname there are twice as many surname-holders in that area, than one would expect given the total number nationally.
High frequency surnames exhibit a range of index values that is very constricted. For example, in the late 1990's, the surname Smith ranged from a minimum value of 50 to a maximum value of 249.
This should be compared with low frequency names that have ranges 0 - 3,000
At the extreme, some names with very small populations have very high index scores of c9,000

If one looks in which areas (in this case, postcodes) the index values reach a peak, the results seem inconsistent

Why have London Postcodes some of the highest and lowest number of surname peaks? Those who are experienced users of Surname Atlas may have noticed that some surnames seem to display unaccountably heavy concentrations in the Isle of Man or Jersey. This is a distortion that is probably introduced through the large population ranges of geographical areas, as well as the large surname ranges. If you are working with contemporary data, and therefore postcodes, please be aware that postcode area populations vary from 3% (Birmingham) down to 0.04% ( I must do a similar exercise on 1881 Registration district areas).

In the following grid, column b represents a matrix of postcode areas and clusters of similar surnames -large and small. This the top lefthand cell represents frequently occurring names in highly populated postcode areas (the Smiths etc in Birmingham etc); and conversely, the bottom righthand cell represents low frequency surnames in sparsely populated postcode areas (e.g. London EC).
The key represents the standard deviations. Most surnames fall within an irregular but graduated range of 40-420 standard deviations. Those in islands or 'pockets' have much wider ranges; and the standard deviations for low frequency names in low-populated areas are excessive.
In effect a 'cluster' of small names is far more likely to appear of significance than a 'cluster' of large names. The size of the postcode in which the cluster appears can also bolster this bias.

For this reason, the index value ideally needs to be standardised. An equation has been formulated that does this- but is not-as yet- in the public domain.

(This section is based upon elements from an unpublished UCL symposium paper by D Lloyd)

- Mean Separation Distance

This is a measure of how dispersed your name is.

The clearest way that I can think of understanding how to apply the formula is through the following example.

Consider 4 places (parishes, registration districts) A,B,C,D each with holders of your surname numbering 100, 50, 20, 10 repectively. Enter these numbers into the following grid, as well as entering a measurement of the distance of all the other places from each other (noted here as ^dBA,^dCA ,^dDA = distance of B,C, and D from A). This is represented in this case by the third, fourth anf fifth columns below.

Formula 1= The enumerator

This is known as the Total Separation Distance. This figure must now be divided by the Total of Separated Persons cited in the demoninator to result in a Mean Separation Distance of the name. The citiations will for every time the placenumber has been used in the first formula, but this time also including the surname number in place A.

Formula 2= The denominator= Total of Separated Persons

The Mean Separation Distance in this case is 1870/220 = 8.5 km

-Mean Separation Distance of the Place

The above allows you to compare the dispersion of 1 surname with another, but does not give one a national perspective of the relative dispersion of all names by place. To do so, one would have to feed all the MSD's back into the original locations.

For example, in Parish A, has a total population of 90 people, comprising just 3 surnames (p,q,r) with occurrences of 16, 8, 3 and associated Mean Separation Distances (19, 16 and 6).

The MSD of the whole parish would then be calculated as:

p19 + q16 + r12/ total parish population

by substitution

(16 x 19) + (8 x 16) + (3 x 6)/90 = 304 + 128 +18/90 = 450/90= MSD of the parish = 5

This is a daunting exercise for anyone without access to large computing power, but it has been done by the University of Essex for each parish of England and Wales in the 1881 census, and the output plotted onto a map, and published in Local Population Studies no 72 (2004)

This map reveals certain broad belts of low separation distances

• South Lancashire and West Riding of Yorkshire
• Cheshire, north Staffordshire and North-east Derbyshire
• Much of Sussex and south-west Kent

Surname density seems to cut across these areas: except a core area of the South Lancashire seems to fit within the 1st belt. This seems to be an area of low surname density, in which the holders ramified greatly within the same region. As opposed say to North Wales, where the high migration to England co-existed with low surname density.

-Nearest Neighbour Analysis

I have in mind here the comparison of the dispersal of 2 or more names in a modern context. For example:-

• if you wished to compare the spread of Chinese and Indian names in a city suburb.
• Or compare the distribution of 'Catholic' and 'Protestant' surnames in Belfast between the 1970s and now

NNA is a measure of distribution, and not of 'pattern' and does have limitations. The higher the number of points, the higher the reliability of the result; and 30 surname plots would be considered the minimum.

Procedure

1. Measure the distance between each and every surname point, and average the result (D_obs)
2. Derive the density of the points (d) , by dividing the number of surname points, by the area under consideration(e.g. ward, parish, registration district, county)
3. Calculate the expected mean of a random distribution of points (r_e) over this area
   r_e = 1 divided by (2 * square root of d)
4. The nearest neighbour statistic can then be obtained by dividing the Average distance by the mean random of distribution

   Dobs

   /re

The Value of the nearest neighbour statistic (R_n) can range from 0 (extremely clustered) to 2.15 ( an ordered and uniform distribution). A value of 1 would suggest a random distribution

It is now up to the surname analyst to explain the resulting distribution

Be aware that: the above equations are based on 2 assumptions:-

1. The points are located within an infinite area
2. The points are free to locate anywhere within that area

These are severe restrictions in the case of surname study: as quite a few factors come into play - propinquity of kin, economic conditions, lines of transport, geomorphology. And if the area under study changes, then this affects the density of the points.

So this technique may perhaps be useful for comparing the temporal change in a surname distribution within a specified area, such as a parish or registration district - provided its boundaries have not changed in the interim, or for comparing the distribution of 2 surnames in the same area, or jsut perhaps the distribution of a widely-dispersed surname, using the area of England, Scotland, or Wales as a baseline.
But the resulting index number for a surname, does not imply that the distributions are the same as it is "possible for arrangement of points which are very dissimilar to have identical mean nearest-neighbour distances"

-Lorenz Curves

These are plots of cumulative percentage (normally used in economic and social history to plot accumulated wealth).However, they can be used here to plot accumulated name frequency (x axis) against accumulated area (y axis). A surname that aligned on the diagonal (a-d) would have a perfectly even distribution of name-holders such that 10% of the area sample contains 10% of the surname-holders, 50% of its area contains 50% of the name-holder population. Any curve that tends to corner b would lllustrate a name in which a large % of the name-holders are concentrated in a small % of land

a......................................................................................................b

All Lorenz curves should be compared against the diagonal. It is possible -using the Gini coefficient- to measure the area between the diagonal and the curve, as a fraction of the area below the diagonal. A high concentration of surname-holders in a small area would yield a high Gini co-efficient.

Lorenz curves are useful for comparing the differential growth or decline of any two features over time so these curves could be used to compare :-

• the evenness of spread of 1 surname against another: particularly for different time-frames
• the main form of a surname against its known variants

Comprehensive area values can be obtained from the Census Abstracts, and individual volumes of the Victoria County History,
and selected values from this site

 

Comparing Census surname distributions over time

There were 383 changes to the boundaries of registration districts from 1841 to 1911, and almost 20,000 to parishes from 1876 to 1972. "These changes mean it is very difficult to compare one census with its predecessor and make the creation of long run time series of raw data impossible"
This problem might not affect the study of a single name, but would have to be taken into consideration by those who are studying the varying distribution of a class of names.
For example, if one wished to study how the age-distribution of Welsh surnames in a London registration district varied between 1841 to 1901

Gregory and Ell consider possible ways to overcome the problem for census geographers.

Source: Ian Gregory and Paul Ell Breaking the boundaries: geographical approaches to integrating 200 years of the census
Journal of the Royal Statistical Society A 168(2) 2005, pp419-437

Part 4 - Socio-economic

-Geodemographics and surnames

Lots of potential for analysis here- though I am not entirely convinced of the validity of this approach on the microscale

"There is no formal proof and no "theory of geodemographics" either, only the concept that "birds of a feather flock together". All the evidence is empirical..the systems are used simply because they do work.."
R Flowerdew/ B Leventhal- Under the microscope (Market Research Society symposium paper)

"Some of the most persuasive evidence that geodemographic mapping does affect perceptions is the condemnation of this work by other researchers"
D Dorling Mapping p13

Geodemographic schemes use census and private data to create a profile of a neighbourhood. These profiles serve as a likely indication of the area's relative affluence, and the possible life-style of its inhabitants. A classification scheme is used to assign profiles into a hierarchical order
Two well-known geodemographic products are:-

However, Acorn is the more usable to the surname analyst, as the profile assigned to a unit postcode is readily available

A One-namer could obviously tabulate the current overall socio-economic status of their name. Although a minimum number of name-holders would be needed (100+?). If a name is still predominantly located within a specific region, such an analysis would divulge what percentage are rural/urban; associated with town centres, suburbs etc.

• Ideally, the results should be plotted on a map - with an icon/ different coloured spot for each occurrence.
  (Genmap has this facility, though one need to convert postcodes into grid refs)
• Is there a correlation beween the distribution and the neighbourhood type?

This is such a new area, that I am wary that there must be pitfalls in applying a scheme to find the socio-economic value of a surname. And would such a profile have any validity?

Perhaps of more significance would be to define a group of names, and to perform the same profiling. This has been done for the names traditionally associated with one small region (i.e. 'local' surnames). The analysis (of this unpublished academic study) found that 'local' names were more associated with lower status profiles. and neighbourhoods.

This type of approach needs to be repeated for other parts of the country

Portsmouth would be an interesting case-study. It is unique in the UK as being an island city, with a strong-sense of place, and a clannishness associated with long-established families.

-Household Composition and surnames

It is possible to glean further information from the nature of the household entry , by analysing the possible combinations of gender and surnames

Possible household categories

This is a simplified version of the household analysis in Mosaic. The use of such a simplified system does have drawbacks e.g. a brother living with a widowed sister would appear to be a pseudo family.
Given name frequencies could be used to help decide if extended families comprise of parents or offspring (Mosaic classifies your forename into 50 clusters - each with a similar age distribution)

.."it would seem that type of neighbourhood, age and gender represent three items of information which are 'orthogonal', ie complementary to each other in that they operate in three quite independent domains. Given that both gender and age can be inferred from a person's first name with a fair degree of reliability (especially when also using public information such as years at their current address on the electoral roll and the presence and name (if present) of a partner) then it would seem that most behaviours could be predicted for any consumer from their name and address with a fairly high degree of success "
R Webber "father of UK Geodemographics"

and in the USA

"The development process also uncovered a correlation between cluster membership and given names. In the 35 million-name database, there were many names that appeared with unusual frequency in only one cluster. For that reason, all of the clusters were given high-indexing first names, resulting in titles like "Jules & Roz" (affluent and physically active urbanites with children), "Denise" (single mothers on a tight budget), and"Elmer" (very sedentary older men)...
[Certain] people defy categorization and have been lumped into a potpourri group known as the "Omegas." Nearly 9 percent of all U.S. households are Omegas."
Source : J Bickert, 1995

 

Names 'typical' of their age groups

core age
38-44	Michelle, Sharon	Kevin, Gary
44-64	Pamela, Janet	Philip, Brian
65-84	Sylvia, Brenda	Kenneth, Raymond
85+	Hilda, Ethel	Percy, Herbert
Source: 'Geographics,GIS and neighbourhood targeting' Wiley, 2005 p 72

Female names are more fashion-driven than male names. If they are combined with a male partner name, then geodemographers are pretty confident in their estimates of that couple's age-range. This is re-inforced by the length of residency - a statistic that is consistently lower where lower-age groups are involved.

As for surnames -

"...in Scotland the percentages of electors with self-evidently Scottish names is significantly higher among consumers in highland and island communities than among consumers in student areas, defence establishments and areas of high-incomes singles and families in inner areas of Glasgow. Indeed the percentage with Scottish names has proved a more effective indicator than the Census indicator 'speaking Gaelic' in identifying areas with the most traditionally Scottish way of life"
source: R Webber Designing geodemographic classifications to meet contemporary business needs Interactive Marketing 5(3) 2004, p 233-234

 

Example

I have just played around and collected household data for name D in the PO postcode area

	Number	%	National Average	Postcodes involved	Households		Types	Numbers
Wealthy achievers	11	25.6	25.1	5	10		Singles(Young)	6
Urban prosperity	2	4.7	10.7	2	2		Singles (Mature)	17
Comfortably Off	14	32.6	26.6	8	14		Doubles	17
Moderate means	8	18.6	14.5	6	8		3+	2
Hard-Pressed	8	18.6	22.4	6	7

• This name is a broad mid-southern name
• In this Postal area, it is associated with the suburban sprawl, rather than urban heartlands
• Mature residents are to the north of Portsmouth, or Bognor. There is negligible movement to retire to the IOW
• Judging from the forenames, and residential areas, the age-profile is quite high. Many of the 2 person households would appear to be pensioners. The young singles are in a minority

It might be a fruitful exercise to correlate geodemographic status against household composition for a name/class of names

Another possible broader-brush classification scheme at local authority level

Does the above have implications for the philosophy of identity. The standard position is that a name has reference but no meaning i.e. it is a label that refers to one object. If it refers to more than one, then its usage is that of a common noun. Foe example, if I talk about a polar bear, the mind conjures up a class of bear with all its associations, snow, whiteness, polar region. If I say 'John' there is no equivalent class of 'Johns' sharing all the same attributes, into which my one John neatly fits. But with geodemographic clusters, someone with a distinctive name might group into defined socio-economic, lifestyle groups. Not everyone, but a significant number. So are Geodemographic clusters "common nouns" or does the lifestyle communality imply meaning ??? This is in all probability early-morning tosh- I certainly have not thought it through

• Is a Victorian Neighbourhood Classification possible?
• based on Census data, property values, rentals, occupation, hosehold size, composition, age structure?
• Now there's an ESRC funding idea!

Other Spatial Analysis Tools

-Index of concentration
-Location quotient
-Cluster Analysis

Still to be written (sometime):-

• Comparative distribution maps in Colin Rogers and David Hey
• Distribution maps from :Census, GRO Indexes, National Burial Index, Hearth tax, Subsidy Rolls
• John Titterton's Median Radius Technique
• Rex Bottle's Diffusion of English Surnames

If you came to this page directly, then please access
Modern British Surname Studies
Last revised: February 13, 2006.