Electoral Commission Northern Ireland

Transcription

1 Electoral Commission Northern Ireland Update Report No. 5 Impact of Individual Voter Registration on Turnout March 2006

2 Contents I Background... 1 Introduction... 1 Our Approach... 4 Structure of the Report... 5 II Geographical Patterns...6 Introduction... 6 The Overall Pattern... 6 The Sub-regional Pattern... 7 The Urban/Rural Pattern Local Variations Concluding Remarks III Socio-Economic and Demographic Patterns Introduction Social Class Deprivation Community Background Housing Tenure and Dwelling Type Demographics Concluding Remarks IV Concluding Remarks...Error! Bookmark not defined. Appendix A: Statistical Tables Appendix B: Turnout and registration rates by Electoral Ward: May 2005 Local Elections. 55

3 I Background Introduction 1.1 This is the fifth 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 The focus of this research report is on geographical patterns in turnout at the elections held in Northern Ireland on 5 May 2005, when two elections were held simultaneously: The Westminster Parliamentary Election, where 18 Members of Parliament were elected from Northern Ireland constituencies; and The Local Government Election, at which 582 Councillors were elected spanning 26 District Councils. 1.3 Each of Northern Ireland s 26 District Council areas is comprised of a number of smaller geographical units called District Electoral Areas (DEAs). There are a total of 101 DEAs. Each DEA is in turn comprised of a set of Electoral Wards. On average, a DEA contains about five Electoral Wards. The total number of Wards is Information on the voter turnout at local Council elections has historically only been published down to DEA level. An innovation in this research report is that turnout analysis is conducted at a sub-dea level. To date electoral analysis of voter turnout has been hampered by a lack of information at Electoral Ward level. This report breaks new ground in compiling estimated turnout rates at Ward level. The purpose of the exercise is to assist the Electoral Commission in its policy remit by shedding light on the geographical and socio-demographic patterns of turnout in Northern Ireland. This new micro level analysis enables, for instance, the areas of extremely low turnout to be identified. Furthermore, the Electoral Commission is able to draw on information on the demographics of turnout black spots and can then tailor their communications campaigns to take account of this new information. 1.5 The context of the research work is very much that of the impact of individual registration requirements on voter participation. These were introduced in the Electoral Fraud Act (2002) as part of a package of measures to help combat perceived electoral fraud. There is considerable interest in assessing what has been the impact of the switch in procedures from household registration to individual registration. The output from this report will assist the Electoral Commission and the Electoral Office for Northern Ireland (EONI) to gauge in a detailed manner the geographical and socio-demographic patterns of turnout in Northern Ireland. It will also discuss the relationship between voter registration rates and turnout rates. Assessing the relationship between the two is clearly crucial to an understanding of voter behaviour and how it is influenced by evolving legislation. 1.6 There has been considerable disquiet in both political and administrative circles at the decline in turnout over the course of recent elections in Northern Ireland given that Northern Ireland has traditionally been viewed as setting the benchmark for turnout in the United Kingdom. Table 1.1 below shows the relevant turnout rates in all parliamentary elections held in Northern Ireland since Page 1

4 Table 1.1: Turnout at Westminster Parliamentary Elections in Northern Ireland since 1974 Year Date Percentage Poll February October May June June April May June May It is of particular concern to note that the turnout at the 5 May 2005 Westminster Parliamentary Election in Northern Ireland was the lowest since the foundation of Northern Ireland in the early years of the 20 th century. It is also noteworthy that the 5 percentage points decline since 2001 in the turnout figure in Northern Ireland went against the trend in the rest of the UK where there was a marginal increase from the 2001 figure (from 59.4% to 61.4%). 1.8 The Westminster and Local Government elections of 5 May 2005 were the first of their kind in Northern Ireland to be conducted since the introduction of individual registration and the requirement placed on voters to produce proof of their identity at the polls. Invariably there will be much speculation concerning the impact of these measures on the turnout. What is factually undeniable is that Northern Ireland bucked the trend in the UK with its sharp decline in overall turnout. The 12 biggest individual UK constituency falls were all in Northern Ireland. 1 The 5% fall in Northern Ireland was not evenly distributed across all 18 constituencies. The largest decline (and indeed the single largest decline in any seat in the UK) was North Belfast which recorded a drop of See Election 2005: Turn-out how many, who and why? The Electoral Commission, p23, Table 7. 2 This excludes the Staffordshire South election which was postponed due to the death of a candidate and where turnout in the eventual election was just 37.2%. Page 2

5 Table 1.2: % Change Northern Ireland Constituencies Westminster Election Seats in rank order of change in percentage turnout Rank Constituency % change 1 Belfast North -9 2 Upper Bann Mid Ulster West Tyrone Newry and Armagh Strangford East Londonderry South Antrim South Down Belfast East Fermanagh & South Tyrone North Down East Antrim Belfast West North Antrim Foyle Lagan Valley Belfast South -2.5 Source: Electoral Commission. 1.9 Turning now to the Local Government arena, Table 1.3 shows the turnout at Local Government Elections in Northern Ireland since All the subsequent elections from that date have been held on the same Ward boundaries. The table highlights the consistency in the decline in turnout from with the drop in the Local Government elections mirroring exactly that in the Westminster election. Table 1.3: Turnout at Local Government Elections, Year Date Percentage Poll May May June May Source: Electoral Office of Northern Ireland (EONI) Page 3

6 1.10 In both 2001 and 2005 the Local Government Elections were held on the same day as the Westminster Elections and the turnout was markedly increased from the 1993 and 1997 figures when elections were held on independent dates. The 2005 Local Government turnout figure of 62.84% was, however, lower than, for instance, the 1981 figure for Local Government Elections (when there was not a simultaneous Westminster election of 66.20%). Research carried out by PwC subsequent to the May elections established that a clear majority of the voting age population preferred the joint elections arrangement. Our Approach 1.11 Based on data provided to the research team by the EONI, the primary focus of our approach was to calculate estimated turnout rates by Electoral Ward in respect of the Local Government elections held on 5 May (A corresponding check was conducted on one sample constituency for the Westminster elections and the results were almost identical.) 1.12 The turnout rate for a particular geographic area is given by the number of votes cast expressed as a percentage of the eligible electorate. The size of the eligible electorate at Ward level is readily available from data published by the EONI. Thus, in order to determine the turnout rate at Ward level it was necessary to estimate the number of votes cast at each Electoral Ward The technique adopted was as follows. In both local and Parliamentary elections, voting takes place at designated locations which have available ballot boxes at which votes are cast. In the May 2005 elections, there were a total of 1,545 ballot boxes The EONI was able to supply, for each of the 1,545 ballot boxes, counts of: The number of electors eligible to cast a vote at that ballot box. The number of ballots actually issued. This is equivalent to the total votes polled at that ballot box From the EONI data it was also possible to identify the Electoral Ward in which the ballot box was physically located. For the most part, the eligible electorate for a ballot box were all within one Electoral Ward only. This was the case for 91 per cent of the 1,545 ballot boxes. In these cases, the ballots issued could be assigned unambiguously to the Ward where the corresponding ballot box was located In the remaining nine per cent of ballot boxes, the eligible electorate was split over two Electoral Wards. As the distribution of the eligible electorate by Electoral Ward was known, the ballots issued for these split ballot boxes were distributed across the Wards in which they were located in proportion to the distribution of the eligible electorate. In that way it was possible to estimate the total number of ballots issued for each of the 582 Wards The number of ballots issued per Ward does not, however, equate exactly with the total votes polled. This is because the ballot box counts did not include postal votes. For the May 2005 local elections, postal votes issued as a proportion of the total electorate averaged out at two per cent, ranging from 0.52 in the Dunmurry Cross DEA to 7.2 per cent in Erne East. In general, postal voting tends to be more prevalent in rural than in urban areas. It was, however, possible to obtain counts of the number of postal votes at DEA level. Within each of the 101 DEAs, the postal vote counts were allocated to the Electoral Ward level in proportion to the distribution of the eligible electorate by Electoral Ward. Page 4

7 1.18 The foregoing procedures enabled the production of estimates of total votes polled by Electoral Ward at the May 2005 local elections that were consistent with the totals published by the EONI for the 101 DEAs and District Councils For the purposes of this report, one final adjustment was made. Local Council elections were not held in the Ballinderry DEA, part of Cookstown District Council. In order to give a complete picture, turnout rates for the Wards within the Ballinderry DEA were calculated from the Parliamentary election, held on the same day in May This probably gives a good estimate of the turnout that would have been observed if local elections had been held in that DEA. As noted above, ballots issued for the Parliamentary election held on the same day in May 2005 were checked against the local election ballots issued for a sample Parliamentary Constituency (North Antrim); the variance between the two counts ranged from per cent to -1.2 per cent, with most values falling in the range ±0.2 per cent Table C.1 in Appendix C below gives the turnout rates by Electoral Ward estimated using the foregoing methodology, alongside the registration rate for that Ward (the eligible electorate as a percentage of persons aged 18 and over). The Table also shows the turnout rate ranking by Electoral Ward. The Electoral Ward with the highest estimated turnout rate is given a rank of 1 (Owenkillew in Omagh, 87 per cent turnout) while the lowest turnout rate is assigned a rank of 582 (Scrabo in Ards, 42 per cent) The turnout rates by Electoral Ward are reported in raw terms, that is, they do not include any adjustment for spoilt votes. However, the inclusion of imputed turnout rates for Wards in Ballinderry means that the turnout rate reported for Northern Ireland in later Sections of this report (63.4 per cent) differs slightly from the published figure for total votes polled as a percentage of the eligible electorate (62.8 per cent) A final point to note is that the variation in turnout rates at ballot box level can be estimated by calculating the number of ballots issued as a percentage of the eligible electorate. This calculation does not include postal votes issued to the relevant eligible electorate and hence will tend to under-estimate the actual turnout rates for the areas encompassed by the ballot boxes Nonetheless, the ballot box calculation can be useful for illustrating the range that can exist in turnout even within an Electoral Ward and is referred to where appropriate. The detailed analysis of all results at ballot box level highlighted a 50 point spread in the turnout with the highest individual ballot box recorded (86 per cent) located in Lower Glenshane and the lowest (36 per cent) in the Conlig Ward of North Down. These remarkable findings, which will be of great interest to all the political parties and all those involved in boosting participation in the democratic process, are explored in greater detail in the following section. Structure of the Report 1.24 Section 2 of this report deals with the geography of the turnout at the May 2005 election and in particular highlights the large disparity in turnout rates at a ballot box and Ward level. This analysis is particularly useful in that it casts serious doubts on what has been the perceived wisdom in Northern Ireland over many years, namely that low turnouts are a prerogative of the more affluent of the eastern seaboard of Northern Ireland Section 3 of the report further explores some of the patterns emerging from the geographical analysis, with a focus on the socio-economic and demographic features of areas with low turnout compared to those with high turnout. The characteristics analysed include deprivation, community background, housing tenure and household type Section 4 presents the concluding remarks. Page 5

8 II Geographical Patterns Introduction 2.1 This Section presents the main findings in relation to geographical patterns of turnout across Northern Ireland. The Section commences with the overall pattern, focusing on a mapping of turnout rates by Electoral Ward. The sub-regional and urban/rural contrasts evident from the overall pattern are then explored in more detail. This is followed by an analysis of local variations in turnout rates. Taken together, the geographical analyses highlight a number of themes which are analysed further in Section 3 below where the socio-economic and demographic profiles of high and low turnout areas are compared and contrasted. The Overall Pattern 2.2 In the May 2005 local elections, the spread in turnout rates at DEA level was 32 percentage points, ranging from 81 per cent in Erne East (Fermanagh) to 49 per cent in the Abbey DEA (North Down). At Electoral Ward level, the spread was even greater, with 45 percentage points separating the Scrabo Ward in Ards (42 per cent turnout) from the 87 per cent turnout estimated for the Owenkillew Ward in Omagh (Figure 2.1. See also Table A2.1 in Appendix A). Per cent Figure 2.1 Turnout by District Council - May 2005 showing minimum and maximum rates at Ward level Belfast Carrickfergus Source: Derived from data supplied by EONI. Lisburn Newtownabbey North Down Antrim Ards Ballymena Banbridge Craigavon Down Larne Ballymoney Coleraine Derry Limavady Moyle Strabane Armagh Cookstown Dungannon Fermanagh Magherafelt Newry and Mourne Omagh 2.3 Clearly, there are very wide disparities in turnout rates at Electoral Ward level. Such variations will inevitably reflect a range of influences. For example, closely fought elections may attract a higher turnout. Political parties may be more or less wellorganised in some areas than in others with regard to getting out the vote. What is of interest in this report is the extent to which there are geographical and socio-demographic patterns that provide indicators for variations in participation in the electoral process. 2.4 As can be seen from Map 2.1, there are very clear geographic patterns in the turnout data. Two patterns in particular stand out: Mean Min Max The east-west divide. The urban-rural contrast. Page 6

9 The Sub-regional Pattern 2.5 For many years commentators have discussed the trend of much higher turnouts in the west of Northern Ireland and in border regions compared to Belfast and the eastern seaboard. Our analysis confirmed this overall picture and it is visually very clear from Map The sub-regional pattern shown in Map 2.1 is summarised in Table 2.1 below. The much higher turnout rates in the West and South are especially striking. In the West and South, the mean turnout rate (73 per cent) was 10 percentage points in excess of the NI average. Indeed, the lowest turnout rate in the West and South (60 per cent in Glebe, Magherafelt Town) was actually in excess of the mean for Outer Belfast (58 per cent). Apart from the West and South, turnout rates below 50 per cent were observed in each of the main subregions. Table 2.1 Turnout by sub-region, May 2005 local elections Per cent of eligible electorate Region 1 : Mean Minimum Maximum % % % Belfast Outer Belfast East North West & South N. Ireland Notes: 1 Based on the European Union s NUTS III classification. 2 Local Council elections were not held in the Ballinderry DEA, Cookstown District Council, in May In order to give a complete picture, turnout rates for the Wards within the Ballinderry DEA were calculated from the Parliamentary election, held on the same day in May Source: Derived from data supplied by EONI. Page 7

10 Project title Map 2.1: Turnout rate of Electorate May 2005 Local Government Elections Page 8

11 Project title Map 2.2: Registration rate of Electorate May 2005 Local Government Elections Page 9

12 2.7 Interestingly, there are rural-urban contrasts within the broad sub-regions listed in Table 2.1 above. This is illustrated by Figure 2.2 which shows the increasing turnout moving from east to west but also, within each sub-region, there is a higher average turnout in rural areas than in urban areas3. This would suggest that there are characteristics of rural areas associated with higher turnout that operate across all sub-regions Figure 2.2 Turnout by region and area type Per cent Urban Mixed Rural 0 Belfast Outer Belfast East North West & South N. Ireland Source: Derived from data supplied by EONI. Area type classifictaions based on NISRA. The Urban/Rural Pattern 2.8 There is a large difference in turnout rates between urban and rural areas. In the May 2005 local elections, the estimated mean turnout rate in urban areas was 60 per cent, compared to 70 per cent in rural areas (see Figure 2.2). Indeed, when turnout rates are ranked by Ward, rural areas are found to account for 84 per cent of the population living in the 20 per cent of Wards with the highest turnout rates. Urban areas predominate in the 20 per cent of Wards with the lowest turnout rates, accounting for 88 per cent of the population living in those Wards (Figure 2.3. See also Table A2.2 in Appendix A). In urban areas, almost one in three electors live in Wards in the bottom 20 per cent ranked by turnout rates. In rural areas, close to half (45 per cent) of electors are in the top 20 per cent of Wards ranked by turnout. 3 The urban/rural classifications shown in Figure 2.2 are based on the NISRA categorisation of Census Output Areas (COAs) within Northern Ireland as being urban or rural. COAs are small sub-units of Wards, typically containing about 100 households. The majority of Wards (89 per cent) are comprised entirely of COAs categorised as urban or rural. The remaining 11 per cent of Wards contain a mix of COAs categorised as urban or rural. Page 10

13 Figure 2.3 Area type and turnout rate ranking 100 Per cent of population Rural Mixed Urban 0 Bottom 20% 21st-79th percentiles Top 20% All areas Turnout rate - ranking Source: NISRA, Census of Population 2001; Turnout rates derived from data supplied by EONI. 2.9 The urban-rural contrast is even more apparent when attention is focused on the tails of the distribution of turnout rates by Electoral Ward, that is, by contrasting the 20 Wards with the highest turnout rates with the 20 Wards where the turnout rate was lowest (see Tables A2.3 and A2.4 in Appendix A). None of the top 20 Wards are urban all of them are in rural areas. There is a strong concentration of very high turnout in the Wards of rural parts of Tyrone and in other rural parts of Mid-Ulster. There are also very clear pockets of extremely high turnout along the border, e.g. in Crossmaglen and Rosslea In contrast, the vast majority of the Wards with the lowest turnout are urban. Furthermore, many of them are in areas scoring highly on key deprivation measures such as the Noble Indicators (Table A2.6). Low turnout is not a North Down phenomenon as often portrayed by the media Contrary to the widely held view that low turnout is concentrated amongst the wealthier areas in Northern Ireland, this new analysis refutes this theory. There is a concentration of very low turnout in some inner-city areas. This is not confined to Belfast, although a cluster of Wards in East Belfast (Woodstock, The Mount and Island) are in the lowest 20 Wards in the whole of the Northern Ireland. Indeed what is striking is the geographic mixture of the Wards with the lowest turnout. There are particularly low turnouts in Wards containing large public housing estates notably Ballysally and Cross Glebe in Coleraine (see Table A2.6). These contrasts indicate the need to consider more closely the sociodemographic aspect to turnout, which is addressed in Section 3 below It is not surprising that the pattern established at a Ward level is replicated at the more micro level of polling stations/ballot boxes. Once again none of the highest turnout ballot boxes are in urban areas. Three of the top 20 turn-outs are in the one rural Ward of Swatragh alone. The geographical pattern is very similar to that of the Ward analysis with very strong showings along the border. Good examples of this in the top 20 ballot boxes are in Silver Bridge at 81.5 per cent and Florencecourt and Kinawley at 82 per cent. There are very strong turnouts across the Magherafelt District Council Area with 5 of the top 20 ballot box turnout rates being located in this Council, a further 5 in Omagh and 6 more in Newry and Mourne. Page 11

14 2.13 Turning to the lowest turnouts at the individual ballot box level, the pattern established in the Wards is again replicated with the results being highly skewed to Wards in urban areas. In addition to the housing estates identified at a Ward level, there were very low turnouts in individual ballot boxes in the Tonagh area of Lisburn and New Mossley in Newtownabbey. The phenomenon of low turnout was not exclusively confined to less well off urban areas. It is noticeable that some of the lowest turnouts are in areas where previous PwC research has identified particularly low levels of electoral registration. Examples of this are in the Loughview Ward of Holywood covering Palace Barracks and in the Rosstulla Ward of Newtownabbey embracing the University of Ulster. Local Variations 2.14 In order to examine some patterns in more detail, the three council areas of Ballymena, Ballymoney and Moyle, which conveniently match the constituency boundary of North Antrim, were chosen for closer inspection. This choice was deliberate as over a series of recent elections North Antrim has been the bell weather seat in respect of turnout with the figure always within a few percentage points of the Northern Ireland average At a Ward level within these three Councils it was feasible to conduct both polling station/ballot box analysis and also detailed Ward analysis. In addition, at the Ward level since the relevant population figures are available, we were able to construct turnout estimates expressed as a percentage of the entire 18+ population residing in the area. These figures are of particular importance since they produce estimates which take account of the prevailing registration rate An interesting finding to emerge from the scrutiny of the ballot box figures was that even within the same Ward there could be very wide variations in turnout. This is perhaps surprising since a priori one might expect similar patterns with respect to voter behaviour across the Electoral Ward. This assumption, however, would over-state the actual homogeneity of many Wards in Northern Ireland Indeed what is apparent from analysis of a number of Wards is that they contain several micro clusters subsumed within the overall Ward total. This often means that very different voter behaviour is masked by the aggregate level of analysis which to date has bedevilled much of the work in this arena in Northern Ireland. One example of how Wards can radically differ internally is provided by the Upper Malone Ward in Belfast where a significant donut of deprivation exists within the Ward boundaries in the form of the Taughmonagh Estate. Given the earlier results which have been presented, it should not, therefore, be surprising that on some occasions there are often disparities in the turnout in Wards where we can inspect evidence from 3 or 4 different ballot boxes, spread across the demographic profile of the Ward The North Antrim example provides clear evidence for this new hypothesis. In the Summerfield Ward of Ballymena Town the 3 ballot boxes produced turnouts ranging from 49 per cent to 64 per cent. One of the most extreme examples of this within-ward variation is highlighted in the Harryville area of Ballymena. There, two ballot boxes covered the whole Ward. The turnout could not have been more different in one ballot box it was 47 per cent and in the other it was 73 per cent. In Dunclug, another Ward in the town, none of the ballot boxes reached a majority of the eligible electorate, ranging from 37 per cent to 50 per cent. In contrast, a ballot box representing the rural village of Dunloy produced one of the highest figures in Northern Ireland with a turnout of 80 per cent. One area which showed more consistency across all the ballot boxes were the Wards of the Glens of Antrim where the turnout was tightly clustered with a spread of 3 or 4 points covering all the areas. Page 12

15 2.19 When the analysis is extended at a Ward level in the North Antrim area to estimating turnout relative to the 18+ population4, some dramatically low figures emerge (see Map 2.3 overleaf). It is estimated that less than 40 per cent of the population aged 18+ voted in the following Wards in Ballymena town Dunclug, Balee, Fair Green and Castle Demesne. None of these Wards fit the identikit profile favoured by many commentators of affluent areas where middle-class voters have switched off from politics. This new information on the identity of very low turnout areas is of vital importance to the ongoing out-reach and educational work of the Electoral Commission. 4 The analysis to date has been on turnout relative to the eligible electorate. Page 13

16 Map 2.3: North Antrim Westminster Constituency - Turnout in Ballymena, Ballymoney & Moyle Councils May 2005 Local Government Elections Page 14

17 2.20 Turning now to the most heavily populated area of Northern Ireland, further analysis was conducted in the Wards of Belfast to ascertain if similar patterns prevailed in a largely urban environment. Once again at an individual ballot box level there was very clear evidence of intra-ward variation. Probably the most remarkable case is in the Ward of Ballymacarratt where that one area produced the highest ballot box figure (70 per cent) and the lowest (42 per cent) across all ballot boxes in East Belfast The geographical patterns also suggest a community background differential (see Table A2.7 and Map 2.3). For example, in West Belfast the highest turnout at the individual ballot box was in Beechmount Ward (73 per cent) and the lowest was in the loyalist Glencairn Estate (43 per cent). The sub-regional pattern discussed above, notably the higher turnout registered by the West and South of NI, also hints at a relationship between community background and turnout. This is further addressed in Section In South Belfast the pattern provides some evidence for a divide in turnout on social class lines; this is indicated with just 44 per cent in a ballot box in the Woodstock area contrasted with 75.9% in Malone and 77.9% in the Musgrave Ward ( on this occasion South Belfast was a hotly-contested parliamentary and local council area- of all 18 parliamentary seats turnout held up best in comparison with 2001) 2.23 In North Belfast there were particularly low turnouts in the large housing estates on the northern outskirts of Belfast, e.g. 45.7% in Rathcoole. The whole area of North Belfast is noticeable for its very low turnout in this election, both in the Local Government arena and in the Westminster poll. No ballot box achieved 70% and the highest recorded was in the Chichester Park Ward off the Antrim Road with 68.7%. Page 15

18 Project title Map2.4: Turnout Rate in the Wards of Belfast District Council (Based on those registered) Page 16

19 Concluding Remarks 2.24 This Section has presented the main findings with respect to the geographical patterns in turnout at the May 2005 local elections. The overall picture is one of a sharp urban-rural contrast overlain on an east-west divide When these overall patterns are further examined, it emerges that areas of high and low turnout embrace a mixture of socio-demographic characteristics. In particular, the areas of low turnout contain a number of inner-city areas with high levels of deprivation and/or with a high share of the population living in social rented housing. But these same areas also contain Wards with relatively low levels of deprivation The uncertain influence of deprivation per se as a factor in low turnout is further highlighted when the 20 Wards with the highest turnout rates are considered. Among these Wards, nine of the 20 are in the 30 per cent most deprived areas of Northern Ireland. Their more distinguishing feature is that the very high turnout Wards are all rural and located west of the Bann. Furthermore, when the Belfast area is examined, the West Belfast Wards exhibit relatively high turnout rates, even though all of these Wards are in the 20 per cent most deprived areas of NI. Indeed, the geographical patterns would suggest that community background, and by extension the nationalist-unionist cleavage, is a factor to consider in analysing variations in turnout rates. For example, 17 of the 20 Wards with the highest turnout rates have a majority Catholic population. In the 20 Wards with the lowest turnout, the exact opposite pattern holds, with 17 of the Wards having a majority Protestant population. Page 17

20 III Socio-Economic and Demographic Patterns Introduction 3.1 This Section focuses on socio-economic and demographic patterns in turnout rates at the May 2005 local elections, under the following headings: Social Class Social class. Deprivation. Community background. Housing tenure and dwelling type. Demographics. 3.2 The variations in the social class profile of Electoral Wards by turnout are summarised in Table 3.1 overleaf using a three-fold classification of Wards: The bottom 20 per cent, that is, the lowest-ranking 116 of the 582 Electoral Wards measured by the turnout rate at the May 2005 local elections. In these Wards, mean turnout was 51.2 per cent, ranging from 41.8 per cent to 55.3 per cent. The highest-ranking 20 per cent of Electoral Wards measured by turnout rates (referred to as the top 20 per cent ). Mean turnout was 76.5 per cent, ranging from 71.4 per cent to 86.8 per cent The remaining Electoral Wards 5. Mean turnout was 63.3 per cent, ranging from 55.3 per cent to 71.2 per cent 3.3 The main finding is that the largest differentials between the profiles for the bottom 20 per cent and the top 20 per cent tend to be found in the middle of the social class spectrum. Most noticeably, small employers/own account workers comprise just six per cent of the population living in the 20 per cent of Wards with the lowest turnout rates compared to 16 per cent in the top 20 per cent. Small employers/own account workers are disproportionately located in rural areas (e.g. the farming community). To the extent that people in this social class have a higher propensity to vote at elections, this would be part of the reason why turnout tends to be higher in rural than in urban areas. 5 Technically, the Wards in the 21 st to 79 th percentiles of the distribution of turnout rates, ranked from highest to lowest. Page 18

21 Table 3.1 Social class composition (aged 16-74, excluding full-time students) by turnout rate ranking: Electoral Wards Turnout All areas Social class: Bottom 20% 21st-79th percentiles Top 20% % % % % Higher managerial/professional Lower managerial/professional Intermediate Small employers/own account Supervisory/technical Semi-routine Routine Never worked/long-term unemployed All Sources: NISRA Census of Population 2001; Turnout rates by Electoral Ward derived from data supplied by EONI. 3.4 Nonetheless, the fact that the representation of, respectively, the managerial/professional and routine occupations does not vary much between the Electoral Wards in the top and bottom 20 per cent would suggest, on the average, a relatively weak relationship between the geography of social class and turnout rates. Deprivation 3.5 In the analysis of geographical patterns in Section 2 above, it was noted that areas of deprivation were to be found in both high- and low-turnout areas, suggesting no clear-cut relationship between deprivation and turnout. This is further supported by the pattern of turnout rates by areas of deprivation shown in Figure 3.1 (see also Table A3.1). As can be seen, the mean turnout rate in the most deprived Wards (62 per cent) was actually slightly in excess of the mean in the least deprived Wards (61 per cent). In fact, the highest turnout rates occurred in the middle-ranking areas by levels of deprivation. This would appear to be an artefact of the urban/rural split in turnout rates. Page 19

22 Figure 3.1 Turnout by deprivation Turnout (%) Least depriv ed 10% Second Third Fourth Fifth Sixth Seventh Eighth Ninth Most depriv ed 10% Deprivation Sources: Derived from EONI data and the NI Measures of Multiple Deprivation. 3.6 As shown in Figure 3.2, turnout rates are, on average, higher in rural areas regardless of the level of deprivation. Indeed, turnout rates tend to be higher in the more deprived rural areas Figure 3.2 Turnout by deprivation and type of area Per cent Urban Mixed Rural Least deprived 10% Second Third Fourth Fifth Sixth Seventh Eighth Ninth Most deprived 10% Source: Derived from data supplied by EONI. Area type based on NISR A. 3.7 The higher turnout rates in rural areas in turn influence the pattern of turnout rates by areas of deprivation shown in Figure 3.1 above. This is because the rural population share tends to be higher in the middle-ranking areas by levels of deprivation (Figure 3.3). Combined with the higher turnout rates in rural areas, this is why turnout by deprivation bulges in the middle of the distribution in Figure 3.1 above. Page 20

23 Figure 3.3 Rural/urban split by deprivation 100 Turnout (%) Rural Mixed Urban 0 Least deprived 10% Second Third Fourth Fifth Sixth Seventh Eighth Ninth Most deprived 10% All Sources: Derived from EONI data and the NI Measures of Multiple Deprivation. 3.8 Overall, it would appear that deprivation is not correlated with variations in turnout rates. Though, as could be seen in the analysis in Section 2 above, the geographical pattern in turnout rates is complex and the relationship with deprivation is one that will be further examined below in the context of other possible influences on turnout. Community Background 3.9 When the 582 Electoral Wards are categorised according to whether the local population is majority Protestant or Catholic, the mean turnout rate in majority Catholic areas is 68 per cent, almost eight percentage points higher than the mean of 61.5 per cent in majority Protestant areas This apparent correlation between turnout and community background is even more sharply illustrated when turnout rates are compared with the Catholic population share. As can be seen from Figure 3.4, there is an almost linear positive relationship between turnout rates and the Catholic population share (see also Table A3.2). In areas where the Catholic population share is less than one in ten, the average turnout rate was 58 per cent. But in those areas where Catholics account for 90 per cent or more of the Ward population, turnout averaged 68 per cent One consequence of the correlation between the Catholic share and the turnout rate is that, in the top 20 per cent of Wards ranked by turnout rates, the Catholic population share is well in excess of the overall NI share, 66 per cent compared to 44 per cent (Figure 3.5. See also Table A3.3). The corollary of this is that, in the bottom 20 per cent of Wards ranked by turnout rates, the Protestant population share is well in excess of its overall NI share, 74 per cent compared to 53 per cent. Page 21

24 Figure 3.4 Turnout by Catholic population share Per cent %-10% 20%-30% 40%-50% 60%-70% 80%-90% 10%-20% 30%-40% 50%-60% 70%-80% 90%-100% Catholic population share Figure 3.5 Community background and turnout rate ranking 100 Per cent of population None Other Catholic Protestant 0 Bottom 20% 21st-79th percentiles Top 20% All areas Turnout rate - ranking Source: NISRA, Census of Population 2001; Turnout rates derived from data supplied by EONI The foregoing contrasts in turnout rates by community background of Electoral Wards would seem to suggest a higher turnout propensity amongst Catholics at election time. Further evidence in that regard is provided by the fact that the turnout in majority Catholic areas tends to be higher than in majority Protestant areas even after controlling for levels of deprivation (Figure 3.6). Though, the pattern in turnout rates by deprivation and community background provides some evidence to indicate that, in majority Protestant areas, turnout tends to be lower in the more deprived areas. By contrast, there is no obvious relation between deprivation and turnout in majority Catholic areas. This is consistent with the analysis of geographic patterns in Section 2 above and points to a more nuanced influence of deprivation than Figure 3.1 might suggest. Page 22

25 Figure 3.6 Turnout by areas of deprivation and whether ward population is majority Protestant or Catholic Turnout (% of eligible electorate) Least depriv ed 10% Second Thi rd Fourth Fifth Sixth Seventh Eighth Ninth Most depriv ed 10% Majority Protestant Majority Catholic Sources: Turnout rates derived from data supplied by EONI; Population and c ommunity bac kground from NISRA Two further points can be made in relation to community background. First, the correlation shown in Figures 3.4 and 3.5 above is not dependent on the urban-rural pattern. Turnout rates are higher in rural areas regardless of the Catholic population share (Table A3.2). In addition, the Catholic population is slightly more likely to reside in rural areas (see table A3.4), but the difference is not great and certainly not sufficient to account for the patterns shown above Second, the differences in turnout rates by community background are not uniformly distributed by region. In particular, the differential in turnout rates is always smaller in the sub-regions than for Northern Ireland as a whole; the differences by sub-region range from -0.9 to +6.6, compared to a Northern Ireland-wide average of +7.2 in favour of Catholic majority areas. Figure 3.7 Turnout rates by sub-region and whether Ward population is majority Protestant or Catholic Per cent Majority Protestant Majority Catholic Belfast Outer Belfast East West & South North N. Ireland Sources: Turnout rates derived from data supplied by EONI; Population data from NISRA, Census of Population Page 23

26 3.15 This is because the sub-regions containing higher proportions of Catholic majority areas are also those in which turnout is above average, notably the West and South. The main disparities by sub-region are in Belfast and the North of Northern Ireland. But the mean turnout in majority Protestant areas in the West and South is actually slightly in excess of the mean in majority Catholic areas. In short, the geographical pattern in community background population shares is not perfectly correlated with differences between Wards in turnout rates. Housing Tenure and Dwelling Type 3.16 One of the themes that emerged from the geographical analysis in Section 2 above is that turnout appears to be lower in areas with higher shares of social rented sector housing 6 ; this was a noteworthy feature of the 20 Wards with the lowest turnout rates (see Table A3.6). This is confirmed by the analysis shown in Figure 3.8 (see also Table A3.5). Almost one in three households in the bottom 20 per cent of Wards ranked by turnout are in social rented housing, compared with just 13 per cent in the top 20 per cent. In the population as a whole, slightly over one in five households (21 per cent) live in social rented housing. 100 Figure 3.8 Tenure and turnout rate ranking Per cent of households Other rented Private landlord Social rented Owneroccupied 0 Bottom 20% Top 20% 21st-79th percentiles Turnout rate - ranking All areas Source: NISRA, Census of Population 2001; Turnout rates derived from data supplied by EONI Household tenure patterns vary strongly between urban and rural areas. In rural areas, almost four in five households (79 per cent) are owner-occupiers, compared with 64 per cent in urban areas (see Table A3.6). Similarly, social rented accommodation is more prevalent in urban (26 per cent) than in rural areas (12 per cent). Thus, the geographical pattern by tenure type is one of the factors associated with the urban-rural divide in turnout. 6 Rented from the Northern Ireland Housing Executive or a Housing Association. Page 24

27 3.18 Tenure patterns are also associated with variations by dwelling type. Whereas the majority of the social rented sector live in terraced housing (48 per cent) and flats (23 per cent), owner-occupiers are much more likely to reside in detached (47 per cent) and semidetached (30 per cent) houses (see Table A3.7). Reflecting their different land-use intensities, there is also a strong urban-rural contrast between dwelling types. Urban areas contain high proportions of terraces and flats whereas detached houses are the norm in rural areas (see Table A3.8) As with tenure type, the geographical pattern by dwelling type is also correlated with turnout rates. Households living in the more urban-centred flats and terraced dwellings are much more likely to live in the bottom 20 per cent of Wards by turnout whereas households in detached dwellings comprise a majority of those in the top 20 per cent of Wards by turnout (Figure 3.9. See also Table A3.9). 100 Figure 3.9 Dwelling type and turnout rate ranking Per cent of population Flats, etc Terraced Semidetached Detached 0 Bottom 20% Top 20% 21st-79th percentiles Turnout rate - ranking All areas Source: NISRA, Census of Population 2001; Turnout rates derived from data supplied by EONI The fact that there is a correlation between dwelling type and turnout helps to describe the geography of voter turnout, but does not of course tell us why it should be the case, for example, that high proportions of detached dwellings should be positively correlated with voter turnout. Nonetheless, this does suggest that turnout rates are also likely to vary with demographic factors such as household type. Demographics 3.21 It is well known that younger people, especially those in the age bracket 18-24, are less likely to register to vote than are other age groups in the population. Nonetheless, the population age profile of areas in the bottom 20 per cent by turnout is remarkably similar to the age profile for the Wards in the top 20 per cent of election turnout (Table 3.2). Overall, geographical variations in population age composition are not correlated with the pattern in voter turnout rates. Page 25

28 Table 3.2 Population age profile by turnout rate ranking: Electoral Wards Turnout All areas Per cent aged: Bottom 20% 21st-79th percentiles Top 20% % % % % All Sources: NISRA Census of Population 2001; Turnout rates by Electoral Ward derived from data supplied by EONI Apart from age, other demographic factors of interest include marital status, household type, household size and migration/population mobility. Each of these factors has an urban-rural dimension, and they are each correlated to varying degrees with the geography of voter turnout Turning first to marital status, there is a slight correlation with voter turnout (Table 3.3). The proportion aged 16+ who are married are more likely to live in high turnout areas than in low turnout areas. The profile for people who are separated, divorced or widowed shows the opposite pattern, as they have a higher share of those living in low turnout areas compared to their share of the population living in high turnout areas. The proportion of single people does not vary with the geography of voter turnout. It can also be noted that the voter turnout pattern shown in Table 3.3 is about in line with what might be expected from the urban-rural pattern in marital status (see Table A3.10). Table 3.3 Marital status profile by turnout rate ranking: Electoral Wards Turnout All areas Marital status (% aged 16+): Bottom 20% 21st-79th percentiles Top 20% % % % % Single Married/remarried Separated/divorced/widowed All Sources: NISRA Census of Population 2001; Turnout rates by Electoral Ward derived from data supplied by EONI. Page 26

29 3.24 The geography of household types is correlated with voter turnout in a similar manner as with marital status. In particular, couples with dependent children are more likely to be located in high voter turnout areas than in the bottom 20 per cent of turnover (Figure See also Table A3.11). Households comprised of one person or all students are somewhat more likely to live in low turnout areas, as are lone parents. The proportion of pensioner households does not vary with voter turnout. 100 Figure 3.10 Household type and turnout rate ranking Per cent of households One person, all student and all other Lone parent Couples, nondependent children Couples, dependent children Couples, no children Pensioner 0 Bottom 20% Top 20% 21st-79th percentiles Turnout rate - ranking All areas Source: NISRA, Census of Population 2001; Turnout rates derived from data supplied by EONI Three points can be made regarding the picture for household type. First, the profile for the low turnout areas is more diverse than for the high turnout areas. Second, the profile for the low turnout areas is very similar to what would be found in urban areas while the high turnout profile is typical of rural areas (compare Tables A3.11 and A3.12) Third, the relatively high weighting of couples in the high turnout profile means that these areas can also be expected to exhibit above-average numbers of persons per household. This is precisely what the data show. In the high turnout areas, the number of persons aged 18+ per household averages out at 2.15 compared to 1.78 in the low turnout areas (Figure 3.11). Indeed, average household size is more highly correlated with voter turnout than any of the other socio-economic and demographic factors discussed in this Section 7. Furthermore, there is a household size gradient within each of the broad types of areas. That is, in each type of area, the average household size of Wards in the top 20 per cent is greater than those in the bottom 20 per cent. 7 The simple correlation between the average number of persons per household aged 18+ and the voter turnout rate is Page 27

30 Figure 3.11 Average household size by type of area and turnout rate ranking 2.5 Household size Bottom 20% 21st-79th percentiles Top 20% 0.0 Urban Mixed Rural All Type of area Sources: NISR A Census of Population 2001; Turnout rates deriv ed from data supplied by EONI The final demographic factor discussed is population mobility or migration. Not unexpectedly, the higher the rate of migration, whether in- or out-, the more likely is it that an area will be in the bottom 20 per cent of voter turnout Wards (Table 3.4). This is a further factor that can be associated with the urban-rural divide in voter turnout, as population mobility rates tend to be much higher in urban than in rural areas (see Table A3.13). Table 3.4 Migration 1 (% of population) by turnout rate ranking: Electoral Wards Turnout All areas Bottom 20% 21st-79th percentiles Top 20% % % % % In-migration Out-migration Total movements Stayers Persons changing address in the year previous the Census date March Sources: NISRA Census of Population 2001; Turnout rates by Electoral Ward derived from data supplied by EONI. Page 28

31 Concluding Remarks 3.28 This Section has examined socio-economic and demographic patterns in turnout rates at the May 2005 local elections. An important point to note is that such area-based profiling provides further detail on the types of areas where turnout is likely to be low or high. But it should be clearly understood that, while they are indicative, area-based profiles do not provide unambiguous evidence on how people behave within the areas under study. For example, areas with high proportions of households comprised of couples are more likely to exhibit high turnout rates. But it does not necessarily follow that electors living in such households are more likely to vote than those living in other household types Bearing that caveat in mind, the indicators analysed in this Section can broadly be grouped into three sets. First, there are factors that are not well correlated, if at all, with turnout rates. On the average, the age profile of an Electoral Ward would not appear to provide to be a useful predictor of the turnout rate. Social class and deprivation would also appear to fall into this category, albeit there are important nuances within each of these indicators, as discussed further below Second, there are a set of indicators that are correlated with turnout and which can be associated with the urban-rural divide in voter turnout discussed in Section 2 above. These factors include the small employers/own account workers social class category, housing tenure, dwelling type, marital status, household type and population mobility. Taken together, these factors point to the more mixed and diverse profile of the population in urban areas Third, there are factors that are correlated with turnout and which exert an influence over and above the urban-rural divide. The average number of persons per household aged 18+ falls into this category, though it is plausible to argue that this may reflect an additional effect due to the interaction between factors that influence household size, such as household type and marital status Community background also falls into the category of factors associated with higher turnout over and above any correlation with the urban-rural divide. Specifically, areas with a Catholic majority tend to have higher turnout rates than do areas with a Protestant majority. This holds across the urban-rural divide and differences in deprivation levels between areas. The most noteworthy exception to the pattern was that in the South and West region there was, on the average, parity in turnout rates between areas classified by community background The analysis in this Section pointed to the possibility of a turnout gradient by level of deprivation in Protestant areas, which is absent in Catholic areas. It is useful therefore to conclude this Section by briefly re-visiting the local variations discussed in Section 2 above. Starting with the 20 very low and 20 very high turnout Wards, Table 3.5 below presents a selection of socio-economic and demographic characteristics. The main point of note is that areas of deprivation occur in about equal proportions in both the highest and lowest-ranking Wards. But in the lowest-ranking Wards, the areas of highest deprivation mostly (seven of the nine Wards) have a Protestant majority while in the highest-ranking Wards they all have Catholic majorities (see Tables A2.5 and A2.6). 8 This is the ecological fallacy problem associated with the analysis of geographical variations in socio-economic and demographic indicators. For example, classifying areas as more or less deprived is useful in targeting resources, but not all poor people live in deprived areas and not all people living in deprived areas are poor. Page 29

32 Table 3.5 Selected characteristics of the 20 Wards with highest turnout and the 20 Wards with lowest turnout at May 2005 local elections Rural 30% most deprived Majority Protestant Social rented 40%+ % % % % 20 highest turnout Wards 20/20 9/20 3/20 0/20 20 lowest turnout Wards 1/20 11/20 17/20 10/ Further evidence for a deprivation effect in Protestant majority areas can be seen by focusing on Belfast and North Down Councils. These are interesting in the first instance because they are both highly urbanised. In addition, North Down is overwhelmingly Protestant and, while it is relatively affluent, there are some Wards that fall within the 30 per cent most deprived. Clearly, within Belfast, deprivation does not affect turnout in majority Catholic areas (Figure See also Tables A.2.7 and A3.14 and Maps B.2- B.3) 9. By contrast, there would appear to be a deprivation effect in Protestant majority Wards. This is clear in Belfast In North Down, there is only one Ward (out of 25) in the bottom 30 per cent by deprivation. Seventeen of North Down s 25 Wards are in the 30 per cent least deprived. The interesting point about these least-deprived Wards is that voter turnout in May 2005 was not much less than in the least-deprived Protestant majority Wards of Belfast (see Map 3.1 for ward level turn-out in North Down) 9 A similar pattern can be seen in the Derry City Council Wards see Table A3.15 and Map B.4. Page 30

33 Figure 3.10 Turnout by deprivation and community background: Electoral Wards in Belfast and North Down Councils BELFAST Majority Catholic Catholic, 30% most deprived Middle-ranking deprivation 30% least deprived Majority Protestant 30% most deprived Middle-ranking deprivation 30% least deprived NORTH DOWN 30% most deprived Middle-ranking deprivation 30% least deprived Turnout (%) Sources: Turnout rates derived from data supplied by EONI; Community bac kground and depriv ation from N ISRA. Page 31

34 Project title Map 3.1 Turn-out in the Wards of North Down Council May 2005 Page 32