2.1.29 The counties with the greatest increase in the population count falls into two types .sx One is the residential county adjacent to a large urban centre :sx for example , Buckinghamshire , Berkshire , and West Sussex in relation to London ; and Hereford & Worcester and Shropshire in relation to the West Midlands .sx The second type is made up of remoter , mainly rural districts , including much of the South West ( Cornwall , Dorset , Somerset , Wiltshire , and Devon ) and East Anglia ( Cambridgeshire , Norfolk , and Suffolk) .sx A ranking of all counties by the proportionate change in their population count during 1981-91 is given in Table 5 .sx 2.1.30 The district and county patterns make it clear why the most rapidly growing regions in 1981-91 were East Anglia , the South West , and the East Midlands ( see Table A) .sx These are the three regions of England which do not include a metropolitan county .sx Furthermore , the fastest growing region - East Anglia - lacks a city in the large cities category .sx ( It is the metropolitan counties and large cities that experienced the biggest fall in population counts) .sx Wales might seem to be an exception to this , in that although it also contains no metropolitan county its count increased by less than 1/2 per cent over the decade .sx But industrial South Wales has some similarities to a metropolitan county .sx Population density .sx 2.1.30 The population density of England & Wales as a whole , as measured by the counts of people present on Census night 1991 , was just over 3 persons per hectare .sx But there is considerable variation over the country , and England ( 3 1/2 persons per hectare ) is almost three times as densely populated as Wales .sx The most densely populated regions in 1991 were the North West and South East , with 8 1/2 and 6 persons per hectare , respectively .sx No English region is as sparsely populated as Wales overall , the least densely populated regions being East Anglia , the South West , and the North , all with fewer than 2 persons per hectare .sx 2.1.32 The metropolitan counties ( including Inner and Outer London ) were over eight times as densely populated as the non-metropolitan counties , with Inner and Outer London the most densely populated .sx Of the four most sparsely populated counties in England & Wales , three were in Wales ( Powys , Dyfed , and Gwynedd ) , the other being Northumberland .sx 2.1.33 Despite a fall in the population count in the 1980s , Islington and Kensington & Chelsea remained the most densely populated London Boroughs , containing over 100 persons per hectare .sx Outside London , the greatest densities were in Portsmouth and Blackpool with over 40 persons per hectare .sx At the other extreme , there were 64 districts with densities below 1 person per hectare , predominantly in Wales , the North , East Anglia , and the South West .sx 2.1.34 In contrast , population density in terms of the count of persons present per household space was remarkably uniform across all areas of England and Wales , ranging from 2.5 persons per household space in Oxfordshire to 1.9 persons per household space in Inner London .sx ( The count of household spaces includes all private living accommodation , whether occupied on census night or not .sx ) .sx 2.1.35 At local government district level , the range of population densities is only a little wider :sx 2.6 persons per household space ( Tamforth ) to 1.4 persons per household space ( City of London) .sx 2.2 1991 Census preliminary counts compared with population estimates .sx 2.2.1 Each year , OPCS produces mid-year estimates of the resident population of England & Wales , of which the mid-1990 estimates are the latest available at national level .sx The current series of estimates is based on the 1981 Census results .sx They have been updated annually by adding a year to the age of the population , adding births , subtracting deaths , and making allowance for migration .sx Though the 1991 Census will be used as the starting point for similar estimates through the 1990s , the preliminary Census results presented here have not been , and cannot be used for this purpose because of differences in coverage described below .sx It will be important , in due course , to compare estimates rolled forward from 1981 with new results based on the 1991 Census to test the quality of the estimating procedure , and if necessary make adjustments .sx This cannot be cone properly until full Census results for the resident population are available .sx The preliminary counts represent only the population counted as present on Census night .sx 2.2.2 There are differences in the treatment of students and armed forces between the Census and the population estimates .sx The estimates count students at their term-time addresses , and armed forces at their stationed addresses , but the Census question on term-time address of students will allow figures for students to be produced on both bases .sx The OPCS population estimates also allow for Census under-enumeration , the extent of which will not be established until the relevant results from the Census Validation Survey are available in Summer 1992 .sx Comparison for England & Wales .sx 2.2.3 For the whole of England & Wales , the preliminary Census count is a little lower ( by about 2 per cent ) than the 1981-based estimate , after making corrections for definitional differences between the two figures .sx The estimated difference is about 1 million people .sx This estimate will change when a comparison with the census count of residents is made .sx 2.2.4 Data on births and deaths are considered to be reliable ; but the international migration figures , based mainly on data from the sample based International Passenger Survey , are less accurate .sx However , errors in the estimation of migration flows are not expected to be able to account for more than about 100,000 people .sx Other components of the difference between estimates and Census results include :sx the number of households where nobody could be contacted ( believed to be around 140,000 when the preliminary counts were compiled ) , and a change in the balance between visitors to England & Wales and absent residents since the 1981 Census .sx This change is estimated around -300,000 , but this involves a substantial margin of uncertainty .sx It is also likely that under-enumeration in the 1991 Census was greater than in the 1981 Census , when around 1/4 million people were omitted from the present population .sx Early indications also suggest that not everyone present on census night in hospitals , hotels , and other communal establishments was counted .sx The size of the under-enumeration will be estimated when the results of the Census Validation Survey are available .sx Comparison for districts .sx 2.2.5 For sub-national areas , the discrepancies between rolled forward estimates and the Census preliminary counts are proportionately larger .sx This is because there is no satisfactory method of bringing the two sets of figures on to a comparable base for small areas .sx Because of this difficulty , the particular analysis presented in Table E of the 1981 Census Preliminary Report is not repeated here .sx Instead , to give an early indication of the level of comparability between the Census and the estimates , Table E below shows a comparison of intercensal changes calculated from the 1981 and extrapolated 1991 estimates , and the increases and decreases recorded in 1981 and 1991 Census preliminary counts ( with allowance in both cases for boundary changes) .sx 2.2.6 The table shows that for 319 of 403 districts the direction of population change is the same when calculated from both sources , but only 141 districts fall in the same broad band .sx For most districts , the increase of the estimated resident population is greater than that of the preliminary count of the population present ( generally between 2 and 3 per cent) .sx This shows up in Table E , where the figures for most districts lie below the diagonal row for cells running from top left to bottom right of the table .sx As with the comparison at national level , the reasons for this are thought to include :sx a greater number of 'no-contacts' in the 1991 Census compared with 1981 ; .sx a changed visitor/absent resident balance compared with 1981 ( this cannot be corrected for local areas ) ; and an increased amount of under-enumeration ( eg missed households and persons missed in enumerated households ) compared with the 1981 Census .sx All these make the 1981 count larger relative to that of 1991 .sx 2.2.7 A few districts show significantly different changes when the Census counts are compared with the estimates .sx The Census counts for those districts growing much faster according to the Census Cambridge , and the Isles of Scilly , will be affected by the presence of students and visitors respectively - factors which can be expected to blur the assessment of change in the resident population since 1981 .sx The districts with substantial deviations in the other directions are Richmondshire and Purbeck which contain considerable armed forces populations , the Vale of Glamorgan , and Hackney .sx 3 Future publications .sx Population present on Census night .sx 3.1 Following this Preliminary Report , census reports and tables will be produced by analysis of the data keyed into the computer directly from forms .sx Census forms received too late for the Preliminary Report will be included in the final figures .sx In 1981 , the overall difference between the final figure for the population present on Census night and the preliminary figure was 0.3 per cent , part of which was due to a processing error which inflated the final counts .sx Resident population .sx 3.2 The main census results will be based not on the population present on Census night , but on the population resident in each area .sx This is because central government , local and health authorities , and others generally plan the provision of services and the allocation of resources for the resident population .sx The resident population will exclude visitors to an area , and include residents who happened to be absent from the area on census night .sx The count of residents will also include imputed details for households where nobody could be contacted , but where people were believed to live .sx ( An article explaining how this is being done was published in the Summer 1991 issue of Population Trends - No .sx 64 ) .sx 3.3 Census questions can be divided into those to which the responses are relatively easy to process and those which are more difficult , and therefore more expensive , to process .sx In general , this division determines whether the question is fully processed ( a 100 per cent item ) or whether it is processed only for a ten per cent sample of the household forms and a ten per cent sample of persons returned on the forms for communal establishments ( a 10 per cent item ) .sx A detailed list of the questions is given in Appendix IV , which also contains a list of the topics for which special volumes will be published .sx 3.4 The topic volumes will be preceded by statistics for individual local areas where these are sufficiently large to maintain the confidentiality of personal census data :sx bullet County Monitors , pamphlets containing about 100 key tabulated counts published with accompanying commentary , as an introduction to the local results ; .sx Local Base Statistics , a set of 20,000 tabulated counts , covering all Census topics , released in machine readable form for all areas down to ward level ; .sx bullet Small Area Statistics , a selection of 9.000 tabulated counts from the Local Base Statistics , released in machine readable form for all areas down to enumeration district level ; and .sx County Reports , printed volumes for each county containing the Local Base Statistics for each district within the county , and for the county as a whole , with accompanying commentary .sx Appendix IV also contains details of the publication timetable for the main sets of local statistics .sx Evaluating the results .sx 3.5 Final results from the census will inevitably still contain some inaccuracies , mainly form the following :sx failure to identify all the living accommodation in an area ; .sx incorrect classification of accommodation as vacant , and hence failure to include the occupants in the census ; .sx failure to identify all the households in a building ; and .sx incorrect information supplied by people filling in forms , perhaps through misunderstanding of the form .sx 3.6 The Census Validation Survey will quantify these inaccuracies .sx Field work for the survey has already been carried out in a sample of enumeration districts .sx In this voluntary survey , addresses in the sampled enumeration districts were listed for comparison with the enumerators' records ; the accuracy of response for a sample of 6,000 households was checked by interview ; and the number of households at each sampled address was investigated .sx