PARRY One-Name Study

Modern Distribution in the UK

This page contains a "slightly experimental" attempt at mapping a modern distribution of the Parry surname in the UK. The Guild expects that members attempt to carry out a survey of sources in order to investigate the population distribution of their surname over time. Although the censuses, in theory, provide a suitable source of data for the nineteenth century, obtaining comparable data for dates both prior to, and following, this period, is more problematic. Detailed here is the process that I have gone through in order to try mapping from electoral roll information.

Please note that I am an amateur genealogist, who has had no training in geographical mapping systems and only limited statistical experience. If using any of this in your own research, you are advised to confirm all facts for yourself. If you do notice any errors, or have any comments about these pages, then please contact me.

On this page:

Initial Considerations
The Actual Process
The Maps

Occurrence - point distribution based on Post Towns
Occurrence - area fill per county
Frequency - number of occurrences per 100,000 population in each county (England, Scotland & Wales only)
Parrys per county, as a percentage of all of the Parrys (England and Wales only)
Density (England and Wales only)

Statistics
Known errors and other difficulties
Evaluation and some conclusions on the method
Notes and Bibliography

Parry ONS Home Page

Top

Initial Considerations

Source of the Data

In order to be suitable for plotting a distribution comparable to the earlier censuses, the data should ideally relate to a particular point in time, and should include as much of the population as possible. One of the best sources to meet these requirements, at least in theory, appears to be the electoral roll. This is compiled annually in October and, although it only contains people aged 18 and above, it should include everyone in that category. This means that the totals for the particular surname can be compared to total population figures for the same age grouping, in order to calculate frequencies and densities. (In reality, there is some under-reporting in the rolls due to people not registering. It should also be noted that other researchers, e.g. Rogers¹, and Dance², have discussed the relative merits of various sources and make some criticisms of this source.)

Details from the electoral rolls have been released for a number of years on UKInfo discs and are occasionally made available for free from the 192.com web site³. However, in recent years, people have been able to "opt out" from having their information included so it is necessary to use data from 2002 and earlier, in order to obtain the most complete coverage. As with any database, it is likely that there are mistakes in these listings, such as transcription errors, duplicate entries, etc. But the main problem is how to map the data.

Mapping the Data

One of the difficulties with mapping the data from electoral rolls is the fact that addresses are not recorded in a consistent format. Smaller localities, such as parishes, are not always listed, and county names do not appear for every address. Even if they did, the counties themselves have undergone considerable change over the years and therefore would not be immediately comparable to the historical counties.

The one consistent feature present for each address is the postcode and, in theory, it should be possible to map the data accurately from this, by converting postcodes to their relevant grid references. However, such conversion of postcodes to grid references for mapping is a major business issue and, whilst it is possible to look up a small number of references for free on a variety of web sites, this is impractical for a large database. Unfortunately the cost of either buying the licence to enable the conversion or of getting a database converted in this way, for what effectively is a "one off" map, is prohibitive for most amateur one-namers. It is also debatable, especially with a large database, whether such a level of accuracy is necessary.

As well as the practical issue of how to map from the postcodes, the final format of the maps needs some consideration. Dr David H. Mellor has demonstrated that mapping by postcode area (i.e. the AA portion of the postcode) is one possible method of looking at current surname distributions⁴. However, postcodes are created by the Royal Mail in order to help them deliver the mail efficiently. Although they are increasingly being used by businesses as the basis for geographical mapping, the postcode areas themselves seem to have been designed to have approximately equal populations (or quantity of mail). This results in considerable differences in the size of the area covered ⁵, which means that a sparsely populated area such as Wales is covered by few codes. This is therefore unlikely to be an appropriate method for plotting a surname such as Parry, which has a strong Welsh concentration, since there will be insufficient information in the map to highlight differences across these areas. As work by the Rowlands has indicated⁶, there can be strong regional differences even amongst the most common Welsh names.

It has been suggested that one can draw one's own distribution maps, based on whatever data is being used⁷. This could be an appropriate method for some people and would allow for the changes to the county and registration district boundaries, which have occurred over time, to be indicated where relevant. However, given time constraints (and ability!) I decided this was not a feasible method for me to use.

It is also important to consider how to ensure that the data, and subsequent maps, are comparable to other distribution information available for the surname. If maps and statistics are going to be produced from the earlier censuses on the basis of the historical counties, then it is obviously preferable to attempt some kind of match to that format. Since I already have a copy of the program Genmap⁸, which contains county and registration district boundaries as at the 1881 census, I decided to use this for all of the maps. For the modern distribution, by allocating postcodes, as far as possible, to the relevant historical counties, it should be possible to produce maps that can then be compared to the earlier censuses. Whilst acknowledging that there will be some discrepancies in the figures across the whole series of distribution maps, because of the boundary changes, hopefully, the results will provide a sufficiently accurate picture of the distribution of Parrys at those points of time, to enable some conclusions to be drawn. (In some ways, perhaps mapping like this is preferable to allowing for the changes to county boundaries, which could give the impression of migration between counties, when in fact it is the boundary itself which has moved - the people are still in the same place.)

Variants

At this point in time, I have not carried out sufficient work on the different spellings that occur, in order to clearly identify variants of the Parry surname. The data processed has therefore been restricted to just the one spelling.

Why?

A final point to mention is actually why one should look at the distribution of a surname. Obviously, for anyone researching a particular name, it is of interest just to see where the name occurs, but there is more to the issue than this. By plotting incidence over time it is possible to follow migration patterns. It has also been suggested (e.g. in Rogers¹) that by calculating the proportion of the name in a specific area, in relationship to the proportion of the total population present in that area, one can identify possible places for the surname's origin.

Thus such statistical analyses can lead to the formulation of theories concerning the surname, which can then be investigated more specifically.

Top

The Actual Process

The data was collected from the 2002 electoral roll, via the 192.com site during one of their "free access" periods. An attempt was made initially, with the help of Gordon Adshead, a fellow Guild member, to convert the postcodes to grid references using an old mapping program. Theoretically, Genmap will recognise places on the basis of grid references, but I encountered difficulties with the mapping when no place or county names were included, and with data recognition when the same place was given a variety of references. (Since full postcodes cover between one and eighty addresses, even the postcodes in a small parish are likely to convert to a number of grid references). It therefore became necessary to find a method of converting the postcodes to named places, which could be recognised by Genmap. With over 25,000 data entries, individual lookups of full postcodes was not an option but, as already explained, trying to map the data in terms of the larger postcode areas, would not be sufficiently refined. It was therefore decided to attempt to map the postcodes to their appropriate Post Towns at district level, i.e. the AAXX portion⁹ of each postcode, and to use this locality information to allocate entries to the historical counties.

It was important that each district should only map to one town, otherwise additional entries would be created when the list was matched to the electoral roll data. Using information from a variety of sources¹⁰ such a list was produced of the "outward" district codes, each with an allocated Post Town. [Where a particular Post Town covered several districts, and a more appropriate locality was indicated by the available information, then this was used instead of the Post Town. However, not all Post Towns have been examined to find such alternatives yet.] Matching this list to a second postcode district list, derived from the Census 2001 "Postcode to OA Look-up file"¹¹, ensured that as many postcode districts as possible were included. The final list was then imported into Genmap in order to see which places were not recognised. Although such places could then have been entered into the Genmap Gazetteer, this would need to be done by everyone who used the list. Instead, therefore, unmatched places were amended to the parishes within which that place occurred in 1881, as far as I could identify by using other resources.¹²

Once the postcode district list was fully recognised, it was matched to the extracted details from the electoral roll, in order to allocate a place and historical county to each address. This file was then imported into Genmap in order to produce the first two maps, a point distribution and an area fill per county, derived from absolute counts for each Post Town. [Since I was using Genmap, Yorkshire was divided into the Ridings. However, other programs may require Yorkshire to be treated as one county.]

At this point, some discrepancies were noticed in the conversion, by Genmap, of recognised places into their relevant counties for plotting the area fill¹³. However, since the data for some maps needed to be calculated using various formulas, county totals were instead derived from the original file, (i.e. the electoral roll entries with their allocated Post Towns and historical counties), and subsequent maps plotted using these figures.

In order to compare the distribution of the Parry surname, either to the same name at other times, or to other surnames, the figures need to be standardised in some way. Calculating frequencies and densities relative to the population does this and, since the electoral roll for 2002 would have been compiled in late 2001, it seemed appropriate to use population figures from the 2001 census as the comparison point. Population totals were therefore obtained for those aged 18 and over, in England and Wales, from the census product Key Statistics for Postcode Sectors¹¹. This file was also matched to the list of Post Towns and historical counties, in order to produce approximate population figures for each historical county.

Population figures were also obtained for Scotland from the Scrol web site¹⁴. Although the Scottish figures were used for calculating the frequency of Parrys in each county, they were not included in the density calculation. This is because the figures would not then be comparable to those so far obtained for the earlier censuses. (Data for the earlier censuses in Scotland is available, but I have not yet collected it.)

Other areas, which did appear in the electoral roll file (i.e. N. Ireland, Jersey, and the Isle of Man), were also excluded from the calculations. Although population figures are available for these areas¹⁵, Northern Ireland was excluded because of the difficulty in obtaining census population figures for the postcode districts, as well as the lack of comparable data from earlier censuses. Jersey and the Isle of Man were excluded because they are not part of the UK¹⁶, the subject of this page. (An attempt at producing statistics for these two areas, as well as for elsewhere in the world, will be made at a later date).

Having collected the census figures, this was a suitable point to examine the validity of my method by checking the total population figures per historical county, as allocated by using the postcodes, against the total population figures for each equivalent modern county as given by the official statistics. The results of this are discussed under the Evaluation.

Top

The Maps

Since I have only just started to plot such distribution maps, there is probably some redundancy here. Once I have carried out a similar process for each of the censuses, it will be possible to identify the most informative maps and some of the others here may be removed. The scale at which maps are plotted can make a considerable difference to the picture they show.

Top

Occurrence - point distribution based on Post Towns

Each point represents a particular Post Town, and has been plotted in a size proportional to the number of Parry entries for that Post Town. Even allowing for the clustering of Post Towns in areas of higher population, it can be seen that certain areas do have higher concentrations of Parrys. However, the possible inaccuracy of plotting by just Post Towns should be remembered. (See the Evaluation for an illustration of this, with regard to Herefordshire).

Top

Occurrence - area fill per county

The number of entries allocated to each historical county on the basis of their Post Towns has been totalled by Genmap, and plotted as an area fill. This plot of the actual counts shows a concentration across North Wales and into England, spreading both north through Lancashire and into Yorkshire, as well as south through Staffordshire towards the midlands.

Another area of high incidence is across South Wales, from Glamorganshire, through Monmouthshire and into Gloucestershire. There is also a slightly raised incidence in the south east of England, around London.

Top

Frequency - number of occurrences per 100,000 population in each county (England, Scotland & Wales only), plotted using two different scales

Whilst the absolute counts of surname incidence are interesting and can be used in a time series to examine migration patterns, it is not possible to draw certain conclusions from them alone. One would expect there to be more instances of any particular surname in the areas of higher population. It is therefore necessary to express the incidence in terms of frequency in relation to the population, as shown in the two maps below.

The map on the left is plotted to the same scale as the county fill above, which used absolute counts. On the right, the scale has been changed to show how such a change can affect the picture produced.

Once such frequencies are examined, it becomes apparent that the high incidence in counties such as Lancashire and Glamorganshire are a feature of the higher populations in those areas. Having taken this population into account, although North Wales still features prominently, the other counties in South Wales and the north of England no longer show concentrations of the name. Instead though, the county of Breconshire has become more prominent.

Amending the scale of the map has resulted in some differences across the English counties being highlighted, whilst those in Wales and Scotland have, to some extent, become masked. This indicates some of the importance that needs to be placed on examining the figures, and deciding on scales, and also raises the question of how one decides what differences are significant.

Top

Parrys per county, as a percentage of all of the Parrys (England and Wales only)

Another suggested way of comparing the distribution of surnames is to look at the proportion of the surname in each county, i.e. expressing the number of entries in a county as a percentage of the total number of entries for that surname. This enables easy comparison to other surnames, or to the same surname at a different point in time. The map produced would present a similar picture to that derived from absolute counts. However, it has been included here since the counts included other areas, whereas this one, for just England and Wales, will be comparable to those I intend to produce for the censuses.

[In order to plot the map, percentages were scaled up by a factor of 100.]

Equivalent scale:
Group 1: 0.01% - 0.04%
Group 2: 0.05% - 0.2%
Group 3: 0.21% - 0.75%
Group 4: 0.76% - 1.8%
Group 5: 1.81% - 4.4%
Group 6: 4.41% & over

Top

Density (England and Wales only)

Although frequency figures are more informative than counts for examining distributions, a further measure has been suggested by Eric Banwell (cited in Rogers¹, and in Riggs¹⁷). In this, the proportion of the particular surname in each area is divided by the proportion of the total population in that area. If every area had an equal population, and the occurrences of the surname were evenly spread across all areas, then the results of this "density" calculation would be 1. Results less than 1 therefore indicate that the surname occurs less frequently than would be expected if it was randomly distributed. Results greater than 1 indicate a higher than expected occurrence of the name.

[Again, in order to plot the map, the results were increased by a factor, in this case, of 10]
Equivalent scale:
Group 1: 0.1 - 0.4
Group 2: 0.5 - 1
Group 3: 1.1 - 5
Group 4: 5.1 - 10
Group 5: 10.1 - 24
Group 6: 24.1 & over

It has been suggested that this density measure can indicate the possible origins of a surname - in which case this result looks almost too good to be true!

One can almost imagine the surname originating in Anglesey and, to a lesser extent, Breconshire, and then gradually spreading out across Wales and into England.

But obviously, since Parry is derived from the patronymic system, there will not be one origin. Instead, many families settled on it as their surname, having used the forename Harry. However, it will be interesting to examine the density maps for the surname at other points in time, such as in the censuses, which will be able to be plotted more accurately, in order to see whether a similar picture arises. Certainly, the same result was reflected in the work of the Rowlands⁶.

In which case, it suggests the need for further research in order to consider issues such as:

the popularity of the first name Harry
the incidence and frequency of the surname Harry
the time at which the ap surnames became permanent

Such investigations will lead to other considerations, such as what factors affected them and why, and could shed light on the reasons for the concentrations found in this modern distribution of the Parry surname.

Top

Statistics

Total PARRYs in the 2002 electoral roll extraction (all areas)

27496

REMEMBER - the following tables show allocated totals, i.e. quantities allocated to these areas on the basis of Post Towns. It is apparent from the evaluation that these do not match to the country and county boundaries and that, for example, even the total populations for England and Wales are not exact.

PARRYs in England & Wales		Relevant allocated population in England & Wales		Percentage incidence
	(X)		(Y)	(X*100)/Y
Total PARRYs in England	17860	Allocated Population >17 in England	37986929	0.0470
Total PARRYs in Wales	9131	Allocated Population >17 in Wales	2259754	0.4041
Total PARRYs in England & Wales	26991	Total allocated population >17 in ENG & WLS	40246683	0.0671

Country	County	Code	Allocated Population	PARRY statistics
			Total popn	Popn >17	No. of Parrys		Frequency per 100,000	Percentage Frequency		% of Parrys in ENG & WLS		density in ENG & WLS
				(n)	(s)	Rank	(s*100000)/n	(s*100)/n	Rank	s*100/S	Rank	(s/S)/(n/N)	Rank
ENG	Bedfordshire	BDF	571425	430579	147	45	34	0.0341	40	0.5446	45	0.5091	38
ENG	Berkshire	BRK	831051	638292	262	26	41	0.0410	29	0.9707	26	0.6121	28
ENG	Buckinghamshire	BKM	803235	608447	202	32	33	0.0332	43	0.7484	32	0.4950	41
ENG	Cambridgeshire	CAM	363630	288747	90	49	31	0.0312	48	0.3334	48	0.4648	45
ENG	Cheshire	CHS	1525261	1176987	1837	3	156	0.1561	14	6.8060	3	2.3273	14
ENG	Cornwall	CON	500762	395946	169	36	43	0.0427	28	0.6261	36	0.6364	27
ENG	County Durham	DUR	1443639	1117845	360	20	32	0.0322	44	1.3338	20	0.4802	42
ENG	Cumberland	CUL	304103	238860	71	50	30	0.0297	49	0.2631	49	0.4432	46
ENG	Derbyshire	DBY	882667	685445	314	22	46	0.0458	25	1.1634	22	0.6831	24
ENG	Devon	DEV	1073080	849741	285	25	34	0.0335	41	1.0559	25	0.5001	39
ENG	Dorset	DOR	474976	376185	126	47	33	0.0335	42	0.4668	47	0.4994	40
ENG	Essex	ESS	2657505	2028356	536	17	26	0.0264	58	1.9858	17	0.3940	53
ENG	Gloucestershire	GLS	1302818	1014834	820	11	81	0.0808	18	3.0380	11	1.2048	18
ENG	Hampshire	HAM	1927555	1506839	522	18	35	0.0346	38	1.9340	18	0.5166	36
ENG	Herefordshire	HEF	167317	130859	225	30	172	0.1719	12	0.8336	30	2.5638	12
ENG	Hertfordshire	HRT	1125876	864561	300	24	35	0.0347	37	1.1115	24	0.5174	35
ENG	Huntingdonshire	HUN	142201	107691	19	61	18	0.0176	63	0.0704	55	0.2631	56
ENG	Isle of Wight	IOW	132731	105322	40	55	38	0.0380	33	0.1482	52	0.5663	31
ENG	Kent	KEN	2537715	1946101	576	16	30	0.0296	50	2.1340	16	0.4413	47
ENG	Lancashire	LAN	4835470	3690310	3268	1	89	0.0886	17	12.1077	1	1.3205	17
ENG	Leicestershire	LEI	863941	665332	250	27	38	0.0376	34	0.9262	27	0.5603	32
ENG	Lincolnshire	LIN	929338	722083	167	38	23	0.0231	59	0.6187	38	0.3449	54
ENG	London	LND	26302	22168	8	72	36	0.0361	36	0.0296	56	0.5381	34
ENG	Middlesex	MDX	3436775	2683247	613	15	23	0.0228	60	2.2711	15	0.3407	55
ENG	Norfolk	NFK	757056	602120	169	37	28	0.0281	54	0.6261	36	0.4185	51
ENG	Northamptonshire	NTH	810597	616560	248	28	40	0.0402	30	0.9188	28	0.5998	29
ENG	Northumberland	NBL	785917	618294	167	39	27	0.0270	57	0.6187	38	0.4027	52
ENG	Nottinghamshire	NTT	1080969	840633	304	23	36	0.0362	35	1.1263	23	0.5392	33
ENG	Oxfordshire	OXF	453692	356869	153	42	43	0.0429	27	0.5669	42	0.6393	26
ENG	Rutland	RUT	30012	23119	8	73	35	0.0346	39	0.0296	57	0.5160	37
ENG	Shropshire	SAL	417189	321205	669	13	208	0.2083	11	2.4786	13	3.1057	11
ENG	Somerset	SOM	739775	579932	229	29	39	0.0395	31	0.8484	29	0.5888	30
ENG	Staffordshire	STS	1830163	1407562	839	10	60	0.0596	21	3.1084	10	0.8888	21
ENG	Suffolk	SFK	677998	527123	148	44	28	0.0281	53	0.5483	44	0.4187	50
ENG	Surrey	SRY	2661312	2079450	659	14	32	0.0317	46	2.4416	14	0.4726	44
ENG	Sussex	SSX	1475018	1166572	343	21	29	0.0294	51	1.2708	21	0.4384	48
ENG	Warwickshire	WAR	2048849	1549551	777	12	50	0.0501	22	2.8787	12	0.7477	22
ENG	Westmorland	WES	68370	54962	33	58	60	0.0600	20	0.1223	53	0.8953	20
ENG	Wiltshire	WIL	607599	467909	220	31	47	0.0470	23	0.8151	31	0.7011	23
ENG	Worcestershire	WOR	809501	629862	476	19	76	0.0756	19	1.7636	19	1.1269	19
ENG	Yorks E. Riding	ERY	500993	388107	62	52	16	0.0160	64	0.2297	50	0.2382	57
ENG	Yorks N. Riding	NRY	548361	423307	191	34	45	0.0451	26	0.7076	34	0.6728	25
ENG	Yorks W. Riding	WRY	3688837	2829657	897	9	32	0.0317	45	3.3233	9	0.4727	43
ENG	Yorkshire	YOR	263730	209358	61	53	29	0.0291	52	0.2260	51	0.4345	49
WLS	Anglesey	AGY	66829	52098	916	8	1758	1.7582	1	3.3937	8	26.2172	1
WLS	Brecknockshire	BRE	31488	24780	167	40	674	0.6739	5	0.6187	38	10.0491	5
WLS	Caernarvonshire	CAE	131480	103516	1357	4	1311	1.3109	2	5.0276	4	19.5472	2
WLS	Cardiganshire	CGN	81705	65797	149	43	226	0.2265	9	0.5520	43	3.3767	9
WLS	Carmarthenshire	CMN	172018	134083	189	35	141	0.1410	16	0.7002	35	2.1018	16
WLS	Denbighshire	DEN	215619	168111	1341	5	798	0.7977	4	4.9683	5	11.8944	4
WLS	Flintshire	FLN	192771	148515	1330	6	896	0.8955	3	4.9276	6	13.3534	3
WLS	Glamorgan	GLA	1302965	1002269	2138	2	213	0.2133	10	7.9212	2	3.1808	10
WLS	Merioneth	MER	37437	29473	194	33	658	0.6582	6	0.7188	33	9.8150	6
WLS	Monmouthshire	MON	482955	367728	1028	7	280	0.2796	7	3.8087	7	4.1685	7
WLS	Montgomeryshire	MGY	81148	62894	165	41	262	0.2623	8	0.6113	41	3.9119	8
WLS	Pembrokeshire	PEM	107959	83048	130	46	157	0.1565	13	0.4816	46	2.3341	13
WLS	Radnorshire	RAD	22221	17442	27	59	155	0.1548	15	0.1000	54	2.3082	15
SCT	Aberdeenshire	ABD	383271	303548	40	56	13	0.0132	67
SCT	Angus	ANS	261533	207177	15	63	7	0.0072	77
SCT	Argyllshire	ARL	59874	47164	4	81	8	0.0085	74
SCT	Ayrshire	AYR	361657	281993	15	64	5	0.0053	82
SCT	Banffshire	BAN	43061	33487	0	92	0	0.0000	86
SCT	Berwickshire	BEW	23640	18563	1	89	5	0.0054	81
SCT	Buteshire	BUT	13720	11072	2	85	18	0.0181	62
SCT	Caithness	CAI	26572	20646	1	90	5	0.0048	83
SCT	Clackmannanshire	CLK	49843	38141	8	74	21	0.0210	61
SCT	Dumfriesshire	DFS	95552	74860	9	71	12	0.0120	70
SCT	Dunbartonshire	DNB	115473	89799	10	69	11	0.0111	71
SCT	East Lothian	ELN	65904	50650	16	62	32	0.0316	47
SCT	Fife	FIF	346580	269613	43	54	16	0.0159	65
SCT	Inverness-shire	INV	108816	84725	11	68	13	0.0130	68
SCT	Kincardineshire	KCD	35498	26784	0	93	0	0.0000	87
SCT	Kinross-shire	KRS	10307	7858	3	84	38	0.0382	32
SCT	Kirkcudbrightshire	KKD	22552	18120	5	79	28	0.0276	55
SCT	Lanarkshire	LKS	1415173	1101940	93	48	8	0.0084	75
SCT	Midlothian	MLN	551080	444535	70	51	16	0.0157	66
SCT	Morayshire	MOR	61567	47562	13	67	27	0.0273	56
SCT	Nairnshire	NAI	11193	8662	0	94	0	0.0000	88
SCT	Orkney	OKI	19245	14895	0	95	0	0.0000	89
SCT	Peebleshire	PEE	16387	12781	6	77	47	0.0469	24
SCT	Perthshire	PER	138554	108495	7	76	6	0.0065	78
SCT	Renfrewshire	RFW	257070	200516	4	82	2	0.0020	85
SCT	Ross & Cromarty	ROC	71112	55113	4	83	7	0.0073	76
SCT	Roxburghshire	ROX	41945	33352	2	86	6	0.0060	79
SCT	Selkirkshire	SEL	23624	18555	2	87	11	0.0108	72
SCT	Shetland	SHI	21988	16589	2	88	12	0.0121	69
SCT	Stirlingshire	STI	192212	150539	14	66	9	0.0093	73
SCT	Sutherland	SUT	10646	8480	0	96	0	0.0000	90
SCT	West Lothian	WLN	176841	133823	8	75	6	0.0060	80
SCT	Wigtownshire	WIG	29521	23110	1	91	4	0.0043	84
NIR	County Antrim	ANT			23	60
NIR	County Armagh	ARM			5	80
NIR	County Down	DOW			15	65
NIR	County Londonderry	LDY			6	78
IOM	Isle of Man	IOM			37	57
CHI	Jersey	JSY			10	70

Top

Known errors and other difficulties

Cross border problems

The postcode districts TD12 and TD15, which were allocated to the Post Towns of Cornhill-on-Tweed and Berwick-upon-Tweed respectively (both in Northumberland), also include areas within the Scottish borders and therefore appear on both the England and Scottish census sites, with their relevant populations. When the allocated census file was matched to the electoral roll file, in order to obtain the required county totals, this caused duplication of the Northumberland Parry total, since it appeared within both countries. Since there were no Parrys within the Scottish portion of these postcodes, the problem was easily overcome by deleting the duplicate entry for the Parrys and also by allocating the Scottish population to its correct historical county of Berwickshire. Such adjustments would always be necessary if working at the postcode district level, since it is not possible to subdivide the two postcode districts into their respective English and Scottish portions. (Even at the postcode sector level, this would be a problem, since it appears to be the two sectors TD12 4 and TD15 1 that actually span the border, and are in both country's census figures).

The above problem was easily spotted, because it caused duplication of figures due to the separate sources for the Scottish and English census figures. However, there are likely to be other cases where a postcode district spans a county border. This will have resulted in some quantities being allocated to the wrong county, since the whole district will have been described on the basis of just one locality. One known example of this is Denham, near Uxbridge. This was used as the locality for the postcode district UB9, in preference to the actual Post Town of Uxbridge, since postcode information seemed to indicate it was a better choice. However, Denham is in Buckinghamshire, whereas the other village covered by UB9 is Harefield, which was Middlesex. Using either village, or the Post Town, will result in part of the population for the district being allocated to the wrong county.

Changes to postcodes

The postcode system is regularly updated by Royal Mail, (published updates issued every six months, but some products seem to have monthly updates) with new postcodes being created and old ones occasionally re-used (after a suitable gap). Areas are also re-coded. All of which would make it difficult to use this method of mapping for data covering a range of dates. Different lists would need to be developed for each specific date.

Top

Evaluation and some conclusions on the method

Supporting images for this section have been placed on a separate page here, since they do not specifically relate to the Parrys.

Attempting to map the modern distribution has been a time consuming process, primarily because of the need to develop the initial list of Post Towns, or other appropriate locations, to match to the postcode districts. Once that was prepared, the actual mapping did not take long. However, the time factor is not the main concern with regard to how practical this method is, for mapping a modern distribution.

The main issue with regard to this method is that of accuracy - how accurate is the picture provided by allocating postcode districts to just one place, and then using that place to obtain the equivalent historical county?

To some extent, the answer to this question will depend on the scale at which one wishes to map the data or to develop statistics for. Clearly, the method is not suitable for producing a detailed map of a small area. If one wished to map individuals or families then, since the numbers that could be shown on such a map is probably fairly small, it would be feasible to look up grid references and exact locations for them. There never was any intention of mapping to such precise levels.

What was hoped for was a system, which could be used with a large dataset, in order to provide a reasonable comparison to other distributions of the surname. It is possible to plot the census information on the basis of parish names, as a point distribution, if using Genmap. The most detailed area fill is at the Registration District level. Above this level, values for historical counties can also be plotted as an area fill. Plotting on the basis of Post Towns will clearly be less informative than that of the parish level. However it seems appropriate to consider how close the method of mapping the modern distribution compares to the two levels of area fill mapping.

Initially a map of the Post Towns, and other localities used, was produced [View map]. This shows how plotting using the postcode districts, of which there are almost 3000, does provide a more detailed picture than plotting just by the 124 areas. As one would expect, there are clusters of different Post Towns in the areas of higher population, such as London, and along several parts of the coast.

Another plot of the Post Towns was made, this time showing each point as proportional to the number of districts allocated to that place.[View map]. This shows that, although some areas, such as London, have been able to be subdivided into many different Post Towns others, such as Birmingham, Liverpool, Manchester, and Bristol, have a lot of postcode districts which have all been allocated to the same location.

This in itself might not be a problem - if it was a highly populated city that had developed on a previously unpopulated area, and all of those districts were part of that city, with none of them straddling any historical boundaries. However, there are probably few situations where such conditions would be met. Many modern towns and cities have gradually incorporated their surrounding parishes and villages, which now exist only as localities within the main place. Thus many outlying postcode districts in larger towns and cities might be more appropriately allocated to other places which used to be separate villages.

It is also apparent that many Post Towns cover an area extending some distance from the town itself. This was investigated further, with regard to Herefordshire, which has six Post Towns, covering nine postcode districts. [View Map] Plotting the parishes in Herefordshire as well as the Post Towns, as points, gives some indication of the number of locations which could be mapped by Genmap, but which will all be merged into just the six towns when using this method of mapping. In fact the Hereford Post Town, which represents the postcode districts of HR1, HR2, HR3 and HR4, actually covers almost 200 nameable places, based on the information from postcode-info.co.uk. The other five Post Towns in Herefordshire each cover between fifteen and fifty-five nameable places.

Clearly it would not be possible to produce a "within county" point distribution map comparable to the census information which can be plotted at parish level. But could an area fill map be produced, which was equivalent to the census registration districts? Showing the RD boundaries on the Herefordshire map indicates that the answer is no. [View map] The map shows that, although six out of the seven RDs do contain a Post Town, one does not. This means that any addresses within that area would have been applied to one of the adjoining RDs. An area fill map of registration districts would thus, incorrectly, show some districts as having no entries.

But the issue is not just whether a registration district might have a Post Town within it, but whether the places allocated to that Post Town would have been within that RD. Since the two systems have no relationship between them, this is most unlikely. As can be seen from another plot of the Herefordshire parishes, this time with some indication of the Post Town to which they were allocated, although there is some similarity in the areas, many parishes would appear in the wrong district.[View map]

Such problems of incorrect allocation are not confined to within counties. In addition to the Herefordshire parishes, entries have also been included on the previous map for some of the parishes which are in the adjoining county of Powys (historically Breconshire and Radnorshire) but which have an HR postcode. These addresses would have been incorrectly allocated to the historical county of Herefordshire. It will also be noticed that several of the Herefordshire parishes are not allocated to Post Towns, because they do not appear on the HR postcode list. The concentration of these along the borders with Shropshire, Staffordshire and Worcestershire warranted further investigation, and it can be seen from the following list that many of these parishes would be allocated to those adjoining counties, on the basis of their postcodes:

LD8 - Post Town allocated: Presteigne, RAD
Byton
Kinsham
Knill

NP5 - Post Town allocated: Monmouth, MON
Llanrothal
Welsh Newton

SY7 - Post Town allocated: Lydbury North, SAL
Brampton Bryan
Leintwardine
Lingen

SY8 - Post Town allocated: Ludlow, SAL
Aston
Brimfield
Burrington
Downton
Elton
Leinthall Starkes
Little Hereford
Middleton on the Hill

WR6 - Post Town allocated: Worcester, WOR
Acton Beauchamp
Bishops Frome
Brockhampton (nr. Bromyard)
Evesbatch
Stanford Bishop
Upper Sapey
Whitbourne

WR13 - Post Town allocated: Malvern, WOR
Colwall
Cradley
Mathon (or West Malvern)

WR15 - Post Town allocated: Tenbury Wells, WOR
Stoke Bliss
Wolferlow

And Orleton appears to be covered by both SY8 and WR6, although I suspect that this could be because there are two such parishes. Whatever the reason, certainly this would cause the Orleton in Herefordshire to be allocated to a different county.

With the lower number of counties in comparison to registration districts, one would expect the number of wrong allocations to be less at the county level, but clearly there are postcode districts that straddle many of the county borders. This leads into the final consideration - the extent to which this method of allocating postcodes to Post Towns and historical counties can present an accurate picture of a distribution at a whole county level.

The list on one of the pages of the Kermit Project at Columbia University illustrates a few of the areas that straddle the UK Country borders (i.e. England/Wales and England/Scotland). However, identifying postcode districts which straddle county boundaries within the individual countries will be more problematic, involving examination of each county in the same way as Herefordshire above.

However, since the population figures were derived from the Census 2001, a basic attempt was made to compare the county figures as allocated by postcodes to the actual population figures given by the census. This is not an easy task, given the changes to the counties that have taken place over the years. Some places are no longer part of a county but rather are designated "Urban Areas". The boundaries for other places, such as London, have changed almost beyond recognition. Figures were therefore examined on the basis of the regional totals but even so, considerable differences were found on some regions. [View comparison].

However, the differences in some areas are relatively small and, where an area has clearly defined boundaries such as Anglesey or the Isle of Wight, the mapping by postcode districts is exactly the same as the county boundaries.

This suggests to me that the method does have some potential, but that perhaps further refinement is required.

How might such refinement be achieved? At the moment, the method allocates places to the postcode districts, but the allocation to places could be carried out at sector level (i.e. AAXX X), there being approximately 9,500 sectors in use. However, it would be very time consuming to try to develop such a list from scratch. An alternative would be to take the existing district list and subdivide each district into nine sectors (ie the relevant AAXX along with the numbers 1-9 for the sector code), initially allocating the same location to each sector. Tackling the border areas first, or those districts known to straddle county boundaries, might enable gradual refinement of the list, whilst still allowing it to be used at its current level of accuracy. In some highly populated areas, it might never be necessary to investigate each sector description, if the whole locality was known to be within one historical county, thus saving time.

It has been noted with regard to the Scottish/English border that the two districts straddling the border do so at the sector level. Thus, even this further level of detail will not solve all of the difficulties in providing relatively accurate maps and statistics. It is also obvious from the number of parishes in the Herefordshire example, that each sector could still cover a sizeable area, with a number of different villages. But it would be more accurate than allocating postcodes to places at the district level.

However, one final point to think about with regard to such distributions is perhaps the question of "What makes an area, an area?". Very few of the online distributions available appear to comment on the differences between the historical counties and the census counties (where registration districts straddle county borders resulting in places being allocated to their adjoining county for census purposes.) In some ways, perhaps this does not matter - as long as the total population figures are derived from the same areas, the only "issue" would be slight inaccuracies to the county mapping. All of the statistics would be correct.

So, since the allocation of the total population figures here to the "historical counties" was carried out in the same way as the allocation of the entries from the electoral roll, how different are the results of this method to those of the censuses?

Top

Notes and Bibliography

1 - Colin D Rogers The Surname Detective. Investigating surname distribution in England, 1086-present day, Manchester University Press, 1995.
2 - Modern British Surnames - Philip Dance's very information site. Follow link at the top to the Distribution page.
3 - Ukinfo products and the 192.com Homepage.
4 - David H. Mellor, Surname Mapping by postcode in The Journal of One-Name Studies, (JOONS) Vol. 8, Issue 8.
5 - A map of postcode areas is available on the evox site.
6 - John & Sheila Rowlands, The Surnames of Wales, FFHS (Publications) Ltd, 1996.
7 - By David Hey, cited on Philip Dance's "Modern British Surnames" site, on the page entitled "Distribution".
8 - Genmap is produced by Archer Software.
9 - For a description of the constitution of postcodes see the Post Office page Postcodes explained (or the explanation on the Evox site.)
10 - Some sources for postcode information (several are duplicates of the Post Towns list)

Evox's Post Towns by UK Postcode area
Historical counties from The Association of British Counties.
postcode-info.co.uk
Post Town list on the Data Bureau site
Frank's compulsive guide to postal addresses
The Applegate Directory, which has a postcode area map as well as many address listings.
Brainstorm's UK postcode search service.
London postcodes
London Postcode Districts on Answers.com
Postcodes for the UK
Postcodes in the UK

Another tactic used was to search on Google using the Post Town, followed by the start of the postcode, which often helped identify the relevant towns for each district.
11 - See Census 2001 for information on products available for the 2001 census.
12 - The main sources for identifying possible alternatives to the places not recognised by Genmap were The Phillimore Atlas & Index of parish registers by Cecil Humphery-Smith, and the information on Genuki. Genuki was particularly useful for the districts in Anglesey, where many locations needed to be allocated to their old parishes.
13 - Differences to Genmap when converting from places to county fill:

Prestwick - appears in the Gazetteer as both Ayrshire and Banffshire, but is allocated to Banffshire for the area fill. Since Banffshire and Ayrshire are not adjoining counties, this could be an error in the Gazetteer.
Clerkenwell, which I had allocated to London, appears in the Gazetteer as both London and Middlesex, and is placed in Middlesex for the area fill.
Covent Garden, which I had allocated to Middlesex, appears in the Gazetteer as both London and Middlesex, and is placed in London for the area fill.

14 - Use the Scrol Analyser on Scotland's Census Results OnLine. Having selected demographic for topic, the "Census Area Statistic Sector" as area choice will give figures for postcode sectors (ie AAXX X) within each council area. I combined these into figures for each postcode district (ie AAXX) before allocating the districts into counties on the basis of Post Towns. (In view of the small number of Parrys in Scotland, it probably wasn't worth the effort!)
15 - Census figures for other areas:

Northern Ireland - figures can be obtained for postcodes from the NI Census Site but the full code is needed for individual searches. An alternative would be to carry out a similar process to that described above for Scotland, using the Census Output Areas. In this case, each area would need to be allocated to a postcode sector first. Since there were only 49 Parrys in the electoral roll, with an overall population of 1,233,753 aged over 17 (i.e. frequency = 0.004%), this has not been done.
Isle of Man - no figures appear to be available for the individual postcode districts, but a total can be found from the IM census site.
Jersey - again, no postcode figures but totals can be obtained from the Jersey Census site.
Guernsey - although not mentioned above, since no Parrys were present, census figures for Guernsey can also be obtained from the Guernsey Census Site.

16 - Direct Government's Key Facts about the United Kingdom.
17 - Geoff Riggs, The 1881 Project - British Surname Distribution in The Journal of One-Name Studies, (JOONS) Vol. 6, Number 3 (July 1997).

Top
Parry ONS Home Page