Electoral Commission Northern Ireland

Transcription

1 Electoral Commission Northern Ireland Update Report No. 6 Components of Change in the Canvass May 2006

2 Contents I Background... 3 Introduction... 3 Approach... 7 II Change in the Electoral Register September to March Introduction The April Reinstated The December Reinstated New from Canvass : Summary Rolling Registration Concluding Remarks III Canvass Non-response: Geographical Patterns Introduction The December Reinstated Overall Non-response The March 2006 Register Concluding Remarks IV Concluding Remarks Appendix A Additional Material Page 2

3 I Background Introduction 1.1 This is the sixth report in an ongoing desk research programme to assist the Electoral Commission in meeting its statutory duty to keep under review and, from time to time, submit reports to government on electoral law and practice in Northern Ireland with respect to the Electoral Fraud (Northern Ireland) Act 2002 ( the 2002 Act ) 1. The primary focus of this report is the canvass, which culminated in the publication of a new Electoral Register in December. 1.2 Prior to the 2002 Act, electoral registration in Northern Ireland had been on a household basis. Individual registration was introduced on foot of the 2002 Act in order to combat widely-held perceptions of fraud and abuse, thereby enhancing confidence in the electoral process. Under individual registration, persons who are qualified to be on the Register must provide three personal identifiers, that is, date of birth, signature and National Insurance number (NINO). These requirements have been important in ensuring the accuracy of the Electoral Register. Under the new system, it has proven more challenging to achieve a comprehensive Register, as measured by the proportion of the population aged 18+ who are registered to vote. 1.3 Presently, the Electoral Register is compiled afresh on an annual basis, following a canvass undertaken by the Electoral Office for Northern Ireland (EONI). The new Register is published on the first of December following the canvass period. The first annual canvass under individual registration was held in the autumn of 2002 (see Box A overleaf). This first Register, published on 1 December 2002, contained 1,072,346 entries representing 86 per cent of the population aged 18 at that time. 1.4 However, the number of entries on the Register declined following the 2003 and 2004 canvasses. In December 2004, after the third canvass under individual registration, the number of entries stood at 1,047,601. The registration rate of 82 per cent was three percentage points below the outcome from the first canvass under individual registration. 1.5 As pointed out in previous update reports, the process of compiling the Electoral Register afresh each year had the effect of introducing a downward trend in the Register. The requirement to produce three personal identifiers places the onus to register firmly on the individual elector. But, as in any canvass, there will inevitably be an element of nonresponse, leading to the observed downward trend in the first three canvasses. 1 See Box A.1 in Appendix A for a list of previous reports and the topics covered to date. 3

4 Box A Individual Registration: Chronology Period Event Electoral Register Sept 02-Nov 02 First canvass under individual registration Dec 02 First Electoral Register based on individual registration 1, Sept 03-Nov 03 Second canvass suspended for NI Assembly Elections 1,097,558 (Sept 03) Nov 03 NI Assembly Elections Feb 04 Second Electoral Register 1,069,160 May 04 European Parliament Elections 1,076,937 Sept 04-Nov 04 Third canvass 1,075,439 (Sept 04) Nov 04 Minister announces ending of annual canvass and temporary introduction of reinstatement Dec 04 Third Register 1,047,601 April 05 70,000 Electors 1,148,487 May 05 Westminster and local Council elections 1,148,010 Aug 05 Government consultation paper setting out proposals for change Sept 05-Nov 05 Fourth canvass 1,145,540 (Sept 05) Dec 05 Fourth Register 1,157,053 4

5 1.6 In August 2004, the Government responded to the declining numbers on the Electoral Register by announcing the following measures: The intention to end the annual canvass, and move towards a system of continuous registration. The temporary introduction of reinstatement. 1.7 The Government intends to end the annual canvass in order to: Increase the number of citizens registering to vote while maintaining the current high level of accuracy in the register, in particular by lifting the burden on electors and refocusing resources on more effectively targeting those not registered. 1.8 The reasons for the temporary use of reinstatement were stated by the Minister as follows, in the context of the upcoming May Parliamentary and local Council elections: The Government is determined to ensure that as many people as possible both secure and use their right to vote. Our goal is an electoral register that is both as accurate and as comprehensive as possible. 1.9 Thus, in April, a new Electoral Register was compiled that included the names of 70,283 individuals who had been on the Electoral Register as of September 2004 but who had not re-registered either during the October canvass or in the period of rolling registration following the publication of the post-canvass Register on 1 December. These 70,283 individuals represent the April The impact of the April reinstatement was examined in detail in the third Update Report (June ) and also in the fourth Update Report, which focused on the Government s proposals for change, as set out in a consultation paper published in August For the canvass, those individuals who were in April but who did not re-register were removed from the Electoral Register. Reinstatement has, however, been retained for the purpose of compiling the December Register. Thus, those who had been on the Electoral Register at the start of the canvass period, in September, but who did not re-register, have been onto the December Register. As with the April, their names will be removed from the Electoral Register if they fail to re-register over the course of the 2006 canvass. 5

6 1.12 In the Government s proposals, the last annual canvass will be held in Reinstatement will not be used for the final 2006 canvass. This is to ensure that, going forward, the Electoral Register is free of any inaccuracies that might arise due to the use of reinstatement e.g. persons being on the Register with an incorrect address entry because they have moved within Northern Ireland since last registering, or incorrectly included because they have moved abroad. Hence, the number of entries on the Electoral Register in December 2006 will be strongly influenced by the number of those who were in December and who subsequently respond to the 2006 canvass In that context, the behaviour of the April in response to the canvass is of considerable interest. In particular, the proportion of the April who chose to re-register during the canvass ought to provide a guide to the likely response of those who were in December The number of individuals who were in compiling the December Register over 95,000 - is also important to consider. These individuals were precisely because they did not respond to the canvass. The December are therefore of interest in predicting the potential level of non-response to the final 2006 canvass. This in turn will also be an important ingredient affecting the size of the baseline Electoral Register moving forward to continuous registration In light of the above, the objectives of this report are to examine the canvass with respect to the following: The April, distinguishing between those who reregistered and those who were removed from the Electoral Register following the canvass. Those who had been on the September Register and were in December because they did not respond to the canvass. Patterns of canvass non-response across Northern Ireland, with particular reference to any correlation between canvass nonresponse and relative deprivation. The relative size and patterns in other components of change, including the inflow of new electors who had not been on the precanvass Electoral Register. 6

7 Approach The April. Those who had actively registered, either at the 2004 canvass or via rolling registration from December 2004 to September The main focus of interest in each of these sub-groups is the extent and pattern of canvass non-response, that is, the proportion who did not apply to be included on the December Register. In the case of the April, those who did not respond to the October-November canvass were removed from the Electoral Register. Those who had been on the pre-existing Register, but did not respond to the canvass, were ( the December ). Figure 1.1 The components of change in the Canvass September December March 2006 Reinstated Removed April 70,823 Reregistered 1,067,980 Did not reregister The framework used for meeting the objectives of this report is to analyse the components of change in the canvass, as set out in Figure 1.1. The pre-existing September Register comprised two distinct subgroups: Reregistered at canvass same address Different address VAP new from 05 canvass Attainers at 1 Dec 05 - Remaining Reregistered Rolling Reg 7

8 1.18 As noted in previous reports, it is often the case that there is a rush of late returns which are not reflected in the Register published at the beginning of December following the canvass. In order to get a more accurate picture of the extent and pattern of non-response, it is therefore necessary to take account of the effect of such late returns, as shown in Figure The remaining components of change due to the canvass, also shown in Figure 1.1, are as follows: New electors of voting age (18+) who had not been on the precanvass September Register, perhaps because they had not previously bothered to register or because they had been living abroad. Attainers, that is, those aged 17 on or before 1 December, but who will turn 18 sometime over the life of the December Register An important point to reflect is that it has been possible, for the first time, to implement the components of change approach to the analysis of the canvass at Northern Ireland, parliamentary Constituency and Electoral Ward levels due to the detailed level of information supplied by the EONI. Specifically, information was supplied by Electoral Ward, for the following: The April, distinguishing those who had reregistered and those who had been removed. The December. These are currently being tracked by the EONI and information was made available on the number of late returns through February The number of electors who had been on the September Register, not including the April, who re-registered at a different address. This gives an insight into patterns of movers, which will be important in implementing the proposed continuous registration process. The number of new electors, who had not previously been on the September Register We would like to take this opportunity to thank the EONI for their cooperation and the effort that they have made to supply the data required for this report The next Section of this report looks at the components of change at Northern Ireland and Parliamentary Constituency levels. Each of the components shown in Figure 1.1 above is discussed in turn. 8

9 1.23 Section 3 of the report examines the components of change at Electoral Ward level. There is a particular focus on non-response and whether this is correlated with relative deprivation, as measured by the NISRA Measures of Multiple Deprivation (Noble, ) Section 4 of the report presents the conclusions, with a particular emphasis on the implications for the final annual canvass, which will commence this coming autumn. 9

10 II Change in the Electoral Register September to March 2006 Introduction 2.1 This Section presents the analysis at Northern Ireland and Parliamentary Constituency level of the change in the Electoral Register in the period from September through December to March The Section falls into two main parts. The first and main part focuses specifically on the canvass period, commencing with a brief overview on the conduct of the canvass before documenting the change in the Electoral Register from September to December under the following headings: The April. The December. New from canvass. 2.2 The second part of this Section examines the behaviour of the Register in the period of rolling registration from December to March This is important to reflect in analysing canvass non-response since, as noted in Section 1 above, the size of the non-response rate in the canvass period tends to be reduced as a consequence of late returns. These do not materialise on the Register until after 1 December. Focusing solely on the canvass period would, however, lead to an over-estimate of the true canvass non-response rate. 2.3 The canvass comprised a number of stages, including: A post-out of forms to everyone who was on the pre-existing register. This was supplemented by two forms being sent to every household where the EONI believed that no-one was registered. Their current assessment is that some 110,000 households fall into this category, based on Valuation and Lands Agency (VLA) data. A full in-person canvass exercise, which commenced in September. The "on the doorstep" exercise was followed-up by a postal reminder in October. A special targeting campaign on the April. 10

11 The April Reinstated 2.4 The April reinstatement comprised a little over 70,000 individuals who had been on the September 2004 Register, but who had not responded to the 2004 canvass. As can be seen from Figure 2.1, these individuals were drawn from across the age spectrum, but with a distinct skew towards the younger age bands, especially those aged (see also table A2.1, Appendix A). The younger age profile of the is not at all unexpected; such persons tend to be more mobile and hence can be more difficult to reach through a canvass exercise of the kind outlined above. A point to note is that the age profile shown in Figure 2.1 is the first time that such information has been made available from the Register. It would be tremendously beneficial to have such information in respect of the full Register. It is clear from Figure 2.1 that canvass non-response is higher in the younger age groups and this is likely to be an important consideration in targeting efforts to enhance the comprehensiveness of the Register following the switch to continuous registration. 30 Figure 2.1 April : Age composition 25 Per cent Reinstated April 05 Population 5 0 Under Age band Sources: EONI; NISRA Mid-year Population Estimates In total, the original 70,000+ April comprised a little over six per cent of the September Electoral Register (Figure 2.3. See also Table A2.2). At Parliamentary Constituency level, the proportion ranged from over nine per cent in Belfast North to four per cent in Fermanagh and South Tyrone. Broadly, the April were somewhat more highly concentrated in urban than in rural constituencies (see Table A2.3). 2.6 By September, the April had fallen to around 60,000 via a process of natural wastage and an activist approach by EONI whereby they removed from the Register those individuals where there was evidence that they were definitely not living at the address to which the polling card was sent. 11

12 Figure 2.2 The April Per cent of September Register Belfast East Belfast North Belfast South Belfast West East Antrim East Londonderry Fermanagh & S. Tyrone Foyle Lagan Valley Mid Ulster Newry & Armagh North Antrim North Down South Antrim South Down Strangford Upper Bann West Tyrone N. Ireland Removed Re-registered Per cent Source: EONI. 2.7 In the present context, the key point of interest regarding the April is the proportion who re-registered over the course of the canvass. As shown in Figure 2.3 (see also Table A2.2), slightly over half of the April re-registered at the canvass. The reregistration rate ranged from a low of 43 per cent, in Foyle, to a high of 56 per cent in Lagan Valley. Though, Foyle apart, there was not a great deal of variation around the NI average in the re-registration rate of the April. As can be seen from Figure 2.2 above, the overall magnitude of the effect on the Electoral Register owed more to variations in canvass non-response than to the dispersion in the re-registration rate. Figure 2.3 The April : Re-registered at Canvass Belfast East Belfast North Belfast South Belfast West East Antrim East Londonderry Fermanagh & S. Tyrone Foyle Lagan Valley Mid Ulster Newry & Armagh North Antrim North Down South Antrim South Down Strangford Upper Bann West Tyrone N. Ireland Source: EONI Per cent re-registering 12

13 The December Reinstated 2.8 Focusing on the sub-group of the September Register who had not been in the April, the EONI data show that 86.1 per cent re-registered at the same address, five per cent re-registered at a different address and 8.9 per cent did not re-register (Figure 2.4. See also Appendix A, table A2.4). Those who did not re-register comprise the December, of which there were 95,120 individuals. Figure 2.4 Response to canvass Base = Previously registered, excluding April Belfast East Belfast North Belfast South Belfast West East Antrim East Londonderry Fermanagh & South Tyrone Foyle Lagan Valley Mid Ulster Newry & Armagh North Antrim North Down South Antrim South Down Strangford Upper Bann West Tyrone N. Ireland Re-registered - same address Different address Did not reregister Per cent 2.9 As can be seen from Figure 2.5, the non-response rate (excluding the April ) ranged from 6.9 per cent (North Antrim) to 14.5 per cent (West Belfast), a spread of over seven percentage points. Figure 2.5 Did not re-register at canvass and had not been in April 05 Belfast East Belfast North Belfast South Belfast West East Antrim East Londonderry Fermanagh & South Tyrone Foyle Lagan Valley Mid Ulster Newry & Armagh North Antrim North Down South Antrim South Down Strangford Upper Bann West Tyrone N. Ireland Source: EONI Per cent 13

14 2.10 What is most striking about the picture shown in Figure 2.5 is that canvass non-response rates fall into two distinct groups by Parliamentary Constituency: The more heavily urbanised Constituencies, of Belfast and Foyle, with the largest non-response rates. The remaining Constituencies, both rural and outside Belfast but within the Greater Belfast Urban Area, where non-response rates were relatively tightly bunched around 7-8 per cent The geographic patterns of canvass non-response are further examined in Section 3 below, particularly the relationship with deprivation. Clearly, there is an urban-rural divide. It is not possible from the available data to say if this is because urban areas are inherently more difficult to canvass, for example, due to having a more mobile population. This is a plausible explanation. But the range in non-response rates is large and would certainly suggest that, for the 2006 canvass, the particular difficulties of canvassing in urban areas should be identified and steps taken to mitigate the problems. New from Canvass 2.12 One of the benefits of a canvass is that this provides an opportunity to attract new entrants to the Electoral Register. There are a number of potential sources, such as persons re-locating from abroad; those who had not previously been registered; and attainers. For Northern Ireland as a whole, the EONI data indicate that about 45,000 persons aged 18+ registered in the wake of the canvass, who had not been on the September Register. They represented 3.5 per cent of the population aged 18+, or about 30 per cent of voting age persons who had not previously been registered Considered by Parliamentary Constituency, the inflow of new entrants aged 18+ ranged from 2.5 per cent in Mid-Ulster to a little over four per cent in North Belfast (Figure 2.6). Though the pattern is not especially pronounced, the inflow of new voting age persons tended to be higher in the Belfast Constituencies and those surrounding Belfast, perhaps reflecting the more mobile population in these areas Attainers comprise the second major source of new entrants from the canvass process. Though, at just over 8,000, the number of attainers who responded to the canvass was relatively modest. Overall, just under 30 per cent of persons aged 17 were registered, with fewer than 25 per cent in the Belfast Constituencies ranging to 38 per cent in Fermanagh and South Tyrone (Figure 2.7). Attracting attainers onto the Register is clearly difficult, even with a door-step canvass. 14

15 Figure 2.6 New aged 18+ from canvass - per cent of population Belfast East Belfast North Belfast South Belfast West East Antrim East Londonderry Fermanagh & S. Tyrone Foyle Lagan Valley Mid Ulster Newry & Armagh North Antrim North Down South Antrim South Down Strangford Upper Bann West Tyrone N. Ireland Sources: EONI; NISR A Mid-year population estimates Per cent Belfast East Belfast North Belfast South Belfast West East Antrim East Londonderry Fermanagh & S. Tyrone Foyle Lagan Valley Mid Ulster Newry & Armagh North Antrim North Down South Antrim South Down Strangford Upper Bann West Tyrone N. Ireland Figure 2.7 Attainers as % of population aged 17 Sources: EONI; NISRA Mid-year population estimates Per cent : Summary 2.15 The foregoing components of change in the canvass period from September to December are summarised at Northern Ireland level in Figure 2.8. This shows both the total flows and the percentage effect, relative to the September starting position. The components by Parliamentary Constituency are shown in Appendix A, Table A

16 2.16 Obviously, re-registration is the main component of change, amounting to 88.5 per cent of the September Register (that is, persons who reregistered at the same or a different address and had not been plus the April who re-registered). Figure 2.8 : Changes to the Register September December Reinstated 34,470 Removed April 70,823 35,813 Reregistered % of Sept 3.0% 3.1% 95,120 Did not reregister - 8.4% 1,067, , ,932 Reregistered at canvass same address 80.8% 52,928 Different address 4.6% 53,241 45,211 VAP new from 05 canvass 4.0% 8,030 Attainers at 1 Dec % 2.17 The April who were removed from the Register represented a reduction of three per cent on the September baseline, though this is somewhat of an over-statement as some of the April had been removed prior to the canvass. 16

17 2.18 The use of the reinstatement mechanism means that the components of change analysis is less straightforward than would otherwise be the case. The use of reinstatement means that the 8.4 per cent who did not respond to the canvass do not, strictly speaking, represent a deduction from the Register. Nonetheless, it can be noted that the December were in excess of the total number of new entrants (the voting age population and attainers) by a margin of 3.7 percentage points. This gives an indication of the negative canvass effect that would have been observed in the absence of reinstatement 2 Rolling Registration 2.19 In previous Update reports it has been noted that rolling registration tends to boost the Electoral Register through two main mechanisms: the late returns and looming election effects. The latter effect has been present in varying degrees in each of the previous registration cycles (see Box A above), but has been absent thus far in the present cycle In the present electoral cycle, the late return effect has manifested itself in re-registrations by persons who had been in December. Thus, by February 2006, over 12,000 of the 95,120 December had re-registered by filling in the appropriate form (Figure 2.9). Figure 2.9 The change in the Electoral Register: September to March Though, the April somewhat clouds the issue, as some portion of these could well have made their way back onto the Register in any event during the canvass. 17

18 September The canvass December Rolling registration March 2006 % of Sept 1,113,803 (incl. Apr 05 ) Reinstated Dec 05 95,120 Reregistered (incl. previously Apr 05 ) 83,075 12,045 Remaining Reregistered Rolling Reg 7.3% 1.1% 88.6% Attainers at 1 Dec 05 Attainers at 1 Dec % VAP new VAP new 4.0% -3.0% Apr 05 removed Adds from Rolling Reg Deletions 0.6% -0.4% 2.21 At Northern Ireland level, therefore, one in eight of those who had been at December had re-registered. The range by Parliamentary Constituency is shown in Figure 2.10 (see also Appendix A, Table A2.6). As can be seen, there was no tendency for the Constituencies with the highest reinstatement rates, such as West Belfast and Foyle, to gain from the late returns by the December. Indeed, re-registration by the December was relatively low in both these Constituencies. 18

19 Figure 2.10 The December : Re-registering post-canvass: Per cent of the December Belfast East Belfast North Belfast South Belfast West East Antrim East Londonderry Fermanagh & S. Tyrone Foyle Lagan Valley Mid Ulster Newry & Armagh North Antrim North Down South Antrim South Down Strangford Upper Bann West Tyrone N. Ireland Source: EONI Per cent re-registering 2.22 One consequence of the variations by Parliamentary Constituency in reregistrations by the December is that, through February 2006, the overall pattern of non-response rates remained largely unchanged. The highest non-response rates were still observed in the more urbanised Constituencies, especially West Belfast and Foyle (Figure 2.11). Figure 2.11 Overall non-response rates, through February 2006 Per cent of September Register Belfast East Belfast North Belfast South Belfast West East Antrim East Londonderry Fermanagh & S. Tyrone Foyle Lagan Valley Mid Ulster Newry & Armagh North Antrim North Down South Antrim South Down Strangford Upper Bann West Tyrone N. Ireland April 05 December Per cent Source: EONI A final point to note from the summary of changes through to March 2006 shown in Figure 2.9 above is the relatively low rolling registration addition and deletion rates. These are lower than in previous years because, as noted above, the late returns effect has been subsumed by re-registrations of those who were in December. 19

20 2.24 But the comparatively low levels of rolling registration activity also mean that the Electoral Register has been relatively static in the period since December. Based on previous experience, in the absence of some form of stimulus such as a looming election, this is likely to remain the case over the remainder of the current registration cycle, through to September Concluding Remarks 2.25 As on previous occasions, non-response was a feature of the canvass. Including the April, the total non-response rate amounted to over 10 per cent of the September Register, from which the canvass was drawn Due to the use of reinstatement, the effect of canvass non-response is not immediately apparent from the published December Electoral Register. Indeed, for the first time since the introduction of individual registration, the Electoral Register has increased in the wake of the annual canvass, rather than decreasing as had happened previously. This can be seen both at Northern Ireland level (Figure 2.12) and, to varying degrees, in 17 of the 18 Parliamentary Constituencies, the single exception being Foyle (see Figures 2.13a-2.13b). 1,200 Figure 2.12 The trend in the Electoral Register 1,150 Thousands 1,100 Registered Excl Reinstated 1,050 1,000 Dec-02 Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Mar-03 Sep-03 Mar-04 Sep-04 Mar-05 Sep-05 Mar-06 Source: EONI 2.27 A different picture emerges if the Register is shown with the excluded. As can be seen from Figure 2.12, when that is done, the negative canvass effect manifests itself once again. Nonetheless, an important finding to emerge from the canvass was that half of those who had been in April re-registered at the canvass It is not possible to say if this would have happened anyway i.e. whether these individuals would have re-registered even if they had not been and regardless of any targeting efforts by the EONI. 20

21 2.29 Whatever the reasons for the 50 per cent re-registration rate, it may well be reasonable to project a comparable re-registration rate for the December in the forthcoming 2006 canvass. This is very important, as the re-registration rate of the December will have a significant effect on the size of the Electoral Register following the 2006 canvass, when reinstatement will not be used again If, in fact, it is reasonable to extrapolate from the experience with the reregistration rate of the April, this would mean that the true size of the Electoral Register is somewhere between the upper (published Register) and lower (excluding all of the ) points shown in Figure 2.12 for March 2006, that is, around 1.12m Figure 2.13a The Electoral Register: Belfast Constituencies Thousands Dec-02 Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Mar-03 Sep-03 Mar-04 Sep-04 Mar-05 Sep-05 Mar-06 East North South West Source: EONI Thousands 80 Figure 2.13b The Electoral Register: Antrim & South Down East Antrim North Antrim South Down South Antrim Source: EONI Dec-02 Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Mar-03 Sep-03 Mar-04 Sep-04 Mar-05 Sep-05 Mar-06 21

22 80 Figure 2.13c The Electoral Register: Mid-NI and Rest of East 75 Thousands Lagan Valley Mid Ulster South Down Strangford Upper Bann Dec-02 Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Mar-03 Sep-03 Mar-04 Sep-04 Mar-05 Sep-05 Mar-06 Source: EONI 75 Figure 2.13d The Electoral Register: Border and West 70 Thousands E. Londonderry Ferm & S. Tyrone Foyle Newry & Armagh W. Tyrone Dec-02 Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Mar-03 Sep-03 Mar-04 Sep-04 Mar-05 Sep-05 Mar-06 Source: EONI 22

23 III Canvass Non-response: Geographical Patterns Introduction 3.1 This Section considers geographical patterns, at Electoral Ward level, in canvass non-response. As the April reinstatement effect has been discussed in detail in Update Report No 3, the analysis of canvass nonresponse focuses primarily on the December. The approach taken is to provide a descriptive overview of the range in canvass non-response rates by Ward and then to provide an analysis with respect to three profile variables, that is: Deprivation. The relation with electoral registration rates is an ongoing concern of the Commission. Type of area, as the urban-rural effect was evident in the Parliamentary Constituency patterns discussed in Section 2 above. Households in social housing. While this is correlated with deprivation, to an extent, and also area type, social housing is often clustered in estates and can have a distinct geographical profile. Also, in previous Update reports, tenure has been found to have an independent effect on registration rates. 3.2 Prior to drawing conclusions, the Section considers the geographical pattern in registration rates as of March The December Reinstated 3.3 As discussed in the Section 2, the December reflect the extent of canvass non-response amongst those on the September Register who had not been in April. As can be seen from Figure 3.1, the canvass non-response rate (computed for this purpose from the number who did not re-register at the canvass, expressed as a percentage of the September Electoral Register, excluding the April ), showed a considerable degree of dispersion by Electoral Ward, ranging from 26 per cent (Creggan Central, Foyle) to as low as two per cent (Craigywarren, North Antrim). 3.4 A second noteworthy feature in the range of non-response rates is that there is a pronounced tail in the distribution, whereby the very highest non-response rates are concentrated in a relatively few Wards. As it transpires, this tail has a very distinct spatial expression. One way of illustrating this is to examine the 20 Wards with the highest non-response rates, and contrast their characteristics with the 20 Wards with the lowest non-response rates. 23

24 Figure 3.1 Non-response to canvass Excluding April Per cent Electoral Wards Source: EONI 3.5 Turning first to the 20 Wards with the highest non-response rates, the following characteristics can be noted (see Table 3.1): Deprivation. Sixteen of the 20 Wards are amongst the 10 per cent most deprived areas in Northern Ireland, based on the Noble Measures of Multiple Deprivation. Three of the remaining four Wards are in the eighth or seventh decile ranking of most deprived areas, where 10 indicates the most deprived 10 per cent and a decile ranking of one indicates the 10 per cent least deprived. Type of area. Seventeen of the 20 Wards are entirely urban. The remaining three are mixed urban/rural, outer estates of Belfast or Derry. Social housing. In Northern Ireland as a whole, the proportion of households living in social housing is in the region of 20 per cent. All but two of the 20 Wards have a much higher proportion in social housing. Community background. All but one of the Wards (Shaftesbury) has a majority Catholic population. Parliamentary Constituency. Half of the 20 Wards are in Foyle with a further three in Belfast West. It will be recalled from Section 2 that these are the Constituencies where canvass non-response was well in excess of the NI average. 3.6 The picture that emerges is therefore one of a clustering of higher nonresponse rates in the more deprived areas, with a preponderance of households living in social housing in urban areas or estates on the periphery of Belfast or Derry. 24

25 Table 3.1 The 20 Wards with the highest non-response rates to canvass: Per cent of September Register, excluding April Parliamentary Constituency DEA Ward Nonresponse Deprivation Type of area Social housing % Decile % Foyle Cityside Creggan Central Urban 64.3 Belfast West Dunmurry Cross Collin Glen Mixed urban / rural 77.9 Foyle Northland Crevagh Mixed urban / rural 46.1 Foyle Cityside Brandywell Urban 64.2 Belfast South Laganbank Botanic Urban 19.9 Foyle Shantallow Shantallow West Urban 53.7 Foyle Shantallow Carn Hill Urban 41.2 Belfast South Laganbank Shaftesbury Urban 65.9 Foyle Northland Strand Urban 30.5 Belfast North Oldpark New Lodge Urban 70.4 Foyle Northland Springtown Urban 21.2 Foyle Shantallow Culmore Urban 33.5 West Tyrone Mourne Ballycolman Urban 52.8 Belfast West Upper Falls Glencolin Mixed urban / rural 40.6 Belfast East Pottinger Ballymacarrett Urban 66.0 Belfast West Lower Falls Whiterock Urban 60.2 West Tyrone Omagh Town Lisanelly Urban 44.6 Foyle Cityside Creggan South Urban 65.0 West Tyrone Mourne East Urban 57.3 Foyle Cityside Westland Urban 36.8 Sources: EONI; NISRA. 25

26 Table 3.2 The 20 Wards with the lowest non-response rates to canvass: Per cent of September Register, excluding April Parliamentary Constituency DEA Ward Nonresponse Deprivation Type of area Social housing % Decile % Mid Ulster Magherafelt Town Knockcloghrim Mixed urban / rural 4.1 Mid Ulster Sperrin Lower Glenshane Rural 4.6 East Londonderry South Antrim Limavady Town Antrim South East Rathbrady Urban 16.0 Parkgate Rural 4.2 Lagan Valley Downshire Hillsborough Rural 7.7 South Antrim Antrim North West Shilvodan Rural 1.0 South Antrim Ballyclare Doagh Rural 8.9 Fermanagh & South Tyrone Erne North Kesh Ederney and Lack Rural 13.3 North Antrim Ballymena North Ballyloughan Urban 1.8 Mid Ulster Ballinderry Coagh Rural 11.6 Fermanagh & South Tyrone Erne East Brookeborough Rural 7.2 South Down Knockiveagh Bannside Rural 5.1 East Antrim Coast Road Glenarm Rural 12.4 East Antrim Kilroot Boneybefore Urban 9.4 Strangford Ards West Lisbane Mixed urban / rural 4.0 Newry & Armagh The Orchard Hockley Rural 4.0 North Antrim Bann Valley The Vow Rural 11.8 Fermanagh & South Tyrone Enniskillen Tempo Rural 9.8 West Tyrone Derg Clare Rural 11.3 North Antrim Braid Craigywarren Rural 4.2 Sources: EONI; NISRA. 26

27 3.7 The 20 Wards with the lowest non-response rates show a much different picture (Table 3.2). All but two are in rural areas 3. The two mixed urban/rural Wards are in less deprived areas (decile ranks of 2 4 ) on the edge of District towns. None of the 20 Wards with the lowest non-response rates have a social housing share in excess of the NI average (20 per cent). Only one Wards (Clare in West Tyrone) counts as a relatively deprived area (decile ranking of nine). Thirteen are in the 30 per cent least deprived areas of Northern Ireland (decile ranking of three or less). 3.8 The Wards in which non-response rates were lowest therefore present a very different profile to those in which non-response was highest. The patterns suggested by Tables 3.1 and 3.2 are, in fact, statistically valid. As shown in Table 3.3, the non-response rates by deciles of deprivation range from an average of 5.5 per cent in the least deprived 10 per cent of Wards to 13 per cent in the most deprived areas 5. Reflecting their higher nonresponse rate, the ten per cent most deprived areas contain 19 per cent of the December. 3.9 Consistent with the preceding qualitative analysis of the tails of the distribution of non-response rates, it can also be seen that, the larger the share of households in social housing, the greater was the average nonresponse rate The statistical analysis also shows the average non-response rate in urban areas (nine per cent) to have been higher than that in rural areas (six per cent). It would, however, appear that there is a degree of interaction between area type and relative deprivation in canvass non-response rates. As shown in Figure 3.3 below, within urban and mixed urban/rural areas, the greater the relative deprivation measure, the higher was canvass nonresponse, on average. In rural areas, however, there is no relationship between relative deprivation and canvass non-response (Figure 3.4). For example, it can be seen from Figure 3.4 that the spread in non-response rates amongst rural Wards in the eighth and ninth deciles of relative deprivation was very similar to that in the fourth through seventh deciles. That is, canvass non-response was especially pronounced in areas that were both urban and relatively deprived. For whatever reason, this was not the case in rural areas. 3 For comparison purposes, note that 30 per cent of the NI population live in rural areas, with 58 per cent in urban areas and 12 per cent in mixed urban/rural areas. 4 That is, they are among the ten per cent of Wards that next to the least deprived decile. Relatively, these are not areas of deprivation. 5 The relationship between non-response and deprivation is statistically significant. 6 The relationship between tenure type and canvass non-response is statistically significant. 27

28 Table 3.3 The December : Did not respond to canvass (adjusted for late returns through February 2006) Per cent of September, excluding April All 7.8 Deprivation (Deciles) Least deprived 10% Most deprived 10% 13.2 Type of area Urban 9.0 Mixed urban/rural 7.6 Rural 5.8 Social housing Less than 5% 5.9 5%-14.9% %-29.9% % Sources: EONI; NISRA. % 3.11 A final point to note from Figures 3.2 and 3.3 is that non-response rates varied considerably within each decile of deprivation. This was especially noticeable in the more deprived urban areas and also more widely across rural Wards. This would in turn suggest that, looking to the future, there may be opportunities for more effective canvassing aimed at producing lower, and more evenly spread, non-response rates. 28

29 Figure 3.2 Non-response at the canvass: Urban/mixed Per cent of September base, excluding April Per cent Deciles of Deprivation - lowest to highest Sources: EONI; NISRA Figure 3.3 Non-response at the canvass: Rural Per cent of September base, excluding April Per cent Deciles of Deprivation - lowest to highest Sources: EONI; NISRA. Overall Non-response 3.12 Prior to considering the registration rate implications of canvass nonresponse, it is useful to briefly discuss the overall non-response rate in the canvass, including the April. When expressed relative to the baseline September Register, the April who were removed from the Register comprised a higher proportion in the more deprived areas than in the less deprived areas (Table 3.5). While the contrasts by type of area and social housing were less pronounced than was the case for the December, this nonetheless serves to reinforce the finding with respect to the relationship between canvass nonresponse and relative deprivation. 29

30 Table 3.4 Overall non-response: Per cent of September Register April December 1 Combined % % % All Deprivation (Deciles) Least deprived 10% Most deprived 10% Type of area Urban Mixed urban/rural Rural Social housing Less than 5% %-14.9% %-29.9% % Adjusted for late returns through February Sources: EONI; NISRA Thus, as can be seen from Table 3.4, the combined non-response rate ranged from 7.8 per cent in the least deprived 10 per cent of Wards to 16.8 per cent in the 10 per cent most deprived Wards, a spread of nine percentage points. There was no strong relationship between deprivation and new from canvass (see Table A3.1 in Appendix A). Hence, the main factor shaping the geographic pattern in registration rates was canvass non-response. 30

31 The March 2006 Register 3.14 When calculated on the basis of the published Electoral Register, there was relatively little change in registration rates pre- and post the canvass, either for Northern Ireland as a whole or in terms of the geographic profile variables considered in this Section. Thus, the average NI-wide registration rate in March 2006 stood at 89.3 per cent, a little over half a percentage point up on the September rate (see Table 3.5). For the most part, a similar pattern held by deciles of deprivation. On average, the 10 per cent of Wards in the second decile of deprivation showed an increase of 1.3 percentage points. In the most deprived 10 per cent of Wards, there was a decrease, on average, but this was less than half a percentage point Of course, the apparent stability in the registration rates based on the published Electoral Register reflects the stabilising effect of the use of reinstatement as a mechanism for managing canvass non-response. Excluding the, the calculated registration rates show a considerably greater degree of dispersion across the profile variables shown in Table When the are excluded, the impact of canvass non-response is immediately obvious. By deciles of deprivation, the adjusted registration rates vary from 88 per cent in the second decile to 75 per cent in the 10 per cent most deprived areas, a spread of 13 percentage points. By contrast, in the published March 2006 Register, the spread in registration rates by deciles of deprivation was eight percentage points (from 93 per cent in the second decile to 85 per cent in the most deprived 10 per cent) As discussed in Section 2 above, the true registration rate is likely to lie somewhere between the published rate and the lower bound rate when the December are excluded. Some of the December may have moved abroad. Others may have temporarily declined to register themselves, for whatever reason The true underlying registration rate will not be revealed until the outcome of the 2006 canvass, when the December will have the opportunity to re-register or be removed from the Register. It is, of course, a moot point as to whether the proportion of the April who were removed from the Register serves also as a predictor for the December. Viewed in that context, it is interesting to note that the proportion of the April who were removed from the Register was higher in the more deprived areas (see Figure 3.4). In the ten per cent most deprived Wards, 56 per cent of the April were removed from the Register following the canvass, compared to per cent in the least deprived areas. Extrapolating that pattern to the 2006 canvass would suggest that the spread in registration rates by areas of deprivation will be stretched further in December 2006, compared to the pattern shown in Table 3.5 for March 2006 based on the published Register. 31

32 Table 3.5 Registration rates (per cent of population aged 18+) September March 2006 Registered Difference Excluding 1 % % pps % All Deprivation (Deciles) Least deprived 10% Most deprived 10% Type of area Urban Mixed urban/rural Rural Social housing Less than 5% %-14.9% %-29.9% % Adjusted for late returns through February Sources: EONI; NISRA. 32

33 Figure 3.4 The April : Per cent removed from the Register Per cent Deciles of deprivation - least to most Source: EONI. Concluding Remarks 3.19 This Section has analysed geographic patterns of non-response to the canvass, focusing primarily on those who had been on the September Register and who were in December because they did not respond to the canvass. On that measure, nonresponse rates varied widely by Electoral Ward, from as high as 25 per cent to as low as two per cent Furthermore, canvass non-response exhibited a distinct geographical pattern. In particular, the highest non-response rates tended to be concentrated in the most deprived urban areas, especially those with a high share of households in social housing. Non-response was generally lower in rural areas, regardless of the relative level of deprivation The use of reinstatement provides a temporary protection for the Register from the full effects of canvass non-response. Thus, there was little change in registration rates based on the published Register when the precanvass September Register is compared to the post-canvass Register measured as at March 2006, to allow for late returns (Figure 3.5) While registration rates based on the published Register tend to be lower in the more deprived areas, the gradient in registration rates is rather steeper when the December are excluded so as to show the registration rate effects of canvass non-response (Figure 3.5). 33

34 3.23 Of course, some portion of those who did not register themselves during the canvass (and hence were ) will re-register at the 2006 canvass. However, based on the experience with the April, the proportion of the December re-registering at the 2006 canvass may well be lower in the more deprived areas. This raises the possibility of a widening gap in registration rates in the most deprived areas leading in to the new system of continuous registration which is proposed for the period beyond the 2006 canvass. 95 Figure 3.5 Registration rates by deciles of deprivation 90 Per cent Sept 05 March 06 Excl Reinstated Deciles of deprivation - least to most Sources: EONI; NISRA. 34

35 IV Concluding Remarks 4.1 This Report has examined the extent and pattern of responses to the canvass. The following are the main findings: One in two of the April re-registered at the canvass. Geographically, the proportion re-registering was lower in the more deprived areas. Re-registration by the April added three percentage points to the change in the Register from September to December. After allowing for late returns through to February 2006, those who had been on the September Register and were in December because they did not respond to the canvass comprised 7.3 per cent of the pre-canvass September Register. Canvass non-response varied considerably by Electoral Ward, from almost one in four to just two per cent. The general pattern was one of higher non-response rates in the most deprived urban areas with an above-average share of households living in social housing. In rural areas, there was no obvious correlation between canvass non-response and relative deprivation. Within these areas, however, there were still substantial variations in canvass nonresponse. The widely varying patterns of canvass non-response across Electoral Wards would suggest that, looking to the future, there may be opportunities for more effective canvassing aimed at producing lower, and more evenly spread, non-response rates. These patterns should be considered further in drawing lessons for the conduct of the 2006 canvass, especially as this will be critical in setting the baseline for the shift to continuous registration. New electors who had not been on the pre-canvass Electoral Register added 4.6 per cent to the change in the Register from September to December Amongst these new electors, attainers added under one per cent to the change in the Register. Overall, the number of attainers registering during the canvass amounted to a little under 30 per cent of the population aged 17. Again, this would suggest that there is scope for improvement in the attraction of new attainers. 35