Referendum 2014 how rural Scotland voted Steven Thomson / October 2014 Research Report
Referendum 2014 how rural Scotland voted Policy Centre Research Report Steven Thomson Senior Agricultural Economist, SRUC October 2014 1
Contents Executive Summary... 3 Introduction... 3 Methodology... 4 Research findings... 4 Referendum Turnout... 4 v voting patterns... 5 Historic regional voting... 6 The rural vote and population characteristics... 6 Conclusions... 9 2
Executive Summary This research report analyses the geography of the results from the 18th September 2014 Scottish independence referendum, specifically looking to see if there were differences between rural and urban areas. Statistical analysis reveals that the rural Local Authorities (LAs) in Scotland were more likely to have voted for the status quo in the 2014 independence referendum than urban LAs. With the exception of the Western Isles, rural areas that voted most strongly against independence in 2014 were also most likely to have voted against home rule in 1979, against devolution in 1997 and against a devolved Scottish Parliament having tax raising powers in 1997. LAs with higher proportions of people that considered themselves as having a Scottish only identity during the 2011 census were more likely to have a low proportion of No voters. There was a greater likelihood of the vote being Yes in areas with higher proportions of younger people (i.e. in the more urban LAs). Likewise, LAs with a higher proportion of older people were more likely to vote No (i.e. the more rural LAs). There was a very strong negative correlation between the proportion of unemployed people in an LA (in 2011) and the proportion of people that voted No. In direct contrast there was a very strong positive correlation between the proportion of economically active people that are self-employed within a LA and the proportion of people voting No. There are characteristics of being rural that have a relationship with the outcome of the referendum. It would appear that rural people voted differently from urban people, and factors such as country of birth, national identity, age and employment status played their part along with a variety of factors personal to each individual. Introduction On the 18th September 2014 a majority of Scottish voters (55.3%) voted No to the question Should Scotland be an independent country? in a historic referendum. To date, there has been no analysis of the rural / urban dimension of the vote. This briefing aims to fill this gap through analysing the geography of the voting results and rural-urban differences, with some analysis of possible contributing factors, i.e. age, employment status, country of birth and national identity. This geographic analysis of voting patterns is placed in context and compared to two other historic Scottish referendums: the 1979 home rule and 1997 devolution (and tax raising option) referendums. 3
Methodology The results from the 1979, 1997 and 2014 referendums were mapped with the rural / urban differences split for the 1997 and 2014 referendums. As referendum votes were announced at Local Authority (LA) level the Randall definition of rurality 1 was used to identify the 14 rural LAs with less than 100 people per square kilometre. Using published results from the 2011 Census 2 at LA level, an Anderson-Darling test of normality was conducted on a number of variables relating to age, national identity, employment status and country of birth. For normally distributed variables, Pearson s correlation co-efficient was used to examine the existence and strength of any relationship, with Spearman s rank correlation co-efficient used where the distribution of a variable did not follow a normal distribution. For both these measures: A negative correlation means as one variable increases the other has a tendency to decrease (-1 means a perfect inverse, straight line, relationship) No correlation means there is no linear relationship between variables (0) A positive correlation means as one variable increases, the other also has a tendency to increase (+1 indicates a perfect positive, straight line, relationship) 3. Whilst a statistically significant correlation reveals that there is a relationship between variables it does not indicate causality, i.e. of voting pattern in this case. Many variables were tested and for brevity only those with statistically significant and strong relationships are discussed in trying to highlight rural / urban referendum voting differences. Research findings This section analyses referendum turnout, rural v urban voting patterns and historic regional voting before moving on to discuss the rural vote and population characteristics including age, employment status, country of birth and national identity. Referendum Turnout The 2014 independence referendum saw the highest voting turnout in modern times at 84., and compares with 63% in the 1979 and 60. in the 1997 referendums. Assessment of the 2014 turnout shows 86.8% in rural LAs, compared to 83. in urban LAs (in 1997 there was very little difference with 61. rural turnout compared to 59.7% urban in 1997). However, the level of 2014 turnout was not found to have any statistically significant correlation to the voting result. 1 Randall, J N, (1985) Economic trends and support to economic activity in rural Scotland", Scottish Economic Bulletin 31 2 http://www.scotlandscensus.gov.uk/ 3 The closer the correlation coefficient is to 1 the stronger the relationship. Evans (1996) suggests that for values of Pearson s coefficient r that the strength of the relationship is: 0 to 0.19 = very weak ; 0.2 to 0.39 = weak ; 0.4 to 0.59 = moderate ; 0.6 to 0.78 = strong ; 0.8 to 1 = very strong. 4
v voting patterns Figure 1 2014 Referendum rural / urban results In the 2014 independence referendum, 5. fewer voters in rural local authorities supported independence at only 41.1% compared to 46.3% in the more populous urban local authorities (as shown in Figure 1). Map 1 shows the regional voting pattern in the 2014 recent referendum, showing high levels of No votes in rural areas. The four LAs with the least support for independence were rural: Orkney (32.8% Yes), Scottish Borders (33. Yes), Dumfries and Galloway (34.3% Yes) and Shetland Islands (36. Yes). The seven LAs with the highest support for independence were all urban ranging from 57.3% in Dundee to 49% in North Ayrshire. Map 1 2014 referendum regional distribution 5
Historic regional voting It is worth noting the similarities in the regional voting patterns in the 2014 referendum, depicted in Map 1, with those of the 1979 and 1997 (including the option for tax raising powers) referendums as shown in Map 2 (note that the scales are different). Whilst the absolute results may differ, it is worth noting that Orkney, Shetland, Dumfries and Galloway, the Borders, Aberdeenshire and Perth & Kinross showed similar trends in the previous referendum. It is particularly noteworthy that, as in 2014, Orkney had the highest level of support for maintenance of the status quo. Map 2 also reveals that the other rural areas that voted most strongly against independence in 2014 were also most likely to have voted against home rule in 1979, against devolution in 1997 and against a devolved Scottish Parliament having tax raising powers in 1997. The cause of this pattern is unknown and could be a combination of socioeconomic, demographic and political factors, from reliance on tourism, EU Structural Funds and Common Agricultural Policies, to fiscal benefits for land owners, political allegiances and the level of migrants in rural areas. It is worth noting the significant change in voting pattern that occurred in the Western Isles between the 1979/1997 and 2014 referendum the causal factors for this change are unknown but may relate to demographic change and economic factors. Map 2 results of 1979 and 1997 referendum by region. (a) 1979 Home Rule (b) 1997 Devolution (c) 1997 Tax varying powers The rural vote and population characteristics National identity, country of birth, age and employment status within LAs were examined to assess if there were any statistically significant correlations with the voting outcome. Figures 2 to 5 show the relationships between LA voting outcomes and these characteristics with rural and urban LAs clearly identified to ease interpretation. Figure 2 shows the relationships between the proportion of people voting No and population density (as a measure of rurality) as well as the proportion of people with different national identities during the 2011 population census. Using Spearman s rank correlation it was found that there was a statistically significant (ρ=0.012), medium strength negative correlation (r= -.438) between population density and the proportion of No votes (for the Yes vote the correlation is the same strength but opposite sign i.e. here it would be r=.438). This means that in rural LAs where population density is lowest (measured at LA level) there was a greater likelihood of a No vote than in higher density urban LAs (remembering we cannot say that there is any causality in this relationship). 6
% British Identity % English Identity Population Density (people/km2) % Scottish Identity Using Pearson s correlation coefficient, Scottish identity was found to have a highly significant (ρ<0.001) strong negative correlation (r= -.638) with the No vote. This means that LAs with higher proportions of people that only considered themselves as Scottish during the 2011 census were considerably more likely to have a low proportion of No voters. Generally speaking, the rural LAs have a tendency to have lower proportions of people with a Scottish only identity. Figure 2 shows a positive association between the proportion of an LA s population with a British only or English only identity and the proportion of No vote. It is clear from these figures that rural LAs tend to have higher proportions of such populations. For British only identity using Pearson s correlation coefficient there was a strong positive correlation (r =.717) that was highly significant (ρ<0.001) whereas with English only identity Spearman s correlation coefficient revealed a highly significant (ρ<0.001) moderate negative correlation (r=.597) with the No vote. This indicates that in LAs with higher proportion of residents with British only or English only identities there was a stronger likelihood of a higher No vote (i.e. many rural LAs). Figure 2 Relationships between the No vote and population density / national identity distinguishing between rural and urban Local Authorities 4,000 75% 3,500 7 3,000 2,500 2,000 1,500 1,000 65% 6 55% 5 500 45% 0 4 1 7% 1 1 5% 8% 3% 1% Similarly to national identity there were relationships between countries of birth and the referendum voting patterns at LA level, as shown in Figure 3. From these graphs it can be seen that there was clearly a positive relationship between LAs with higher proportions of people born in England and the No vote and a negative relationship between the proportion of people born in Scotland and the No vote (i.e. greater support for independence). Statistically, both these relationships were significant (ρ<0.001) with the Pearson correlation co-efficient for born in England and No vote showing a strong positive correlation (r=.650) and born in Scotland and the No vote having a moderate negative correlation (r = -.530). Once again, the figures show rural LAs have a tendency to have lower proportions of Scottish-born residents than urban areas with the converse for English-born residents. 7
% Aged 45-65 % Aged 65+ % Aged 16-24 % Aged 25-44 % Born in England % Born in Scotland Figure 3 Relationships between the No vote and country of birth distinguishing between rural and urban Local Authorities 2 10 18% 1 1 1 1 8% 95% 9 85% 8 75% 7 65% 6 In the referendum much was made about the youth vote, with 16 and 17 year olds afforded a vote for the first time. Figure 4 shows the voting pattern at LA level by proportion of the population in four age groups. At first glance it appears that LAs with a higher proportion of 16-44 year olds (top two graphs) were more likely to vote Yes, whereas there was a greater tendency to vote No in LAs with higher proportions of the population aged over 44 years (bottom two graphs). The age structure of rural v urban LAs is clearly demonstrated in these graphs with higher proportions of older people in rural LAs (and hence higher likelihood of a No vote). Figure 4 Relationships between the No vote and age group distinguishing between rural and urban Local Authorities 18% 33% 1 31% 1 29% 1 27% 1 25% 8% 23% 21% 19% 17% 15% 3 25% 3 2 28% 15% 2 2 1 2 R² = 0.1798 2 5% R² = 0.1539 The Pearson s correlation coefficient between a No vote and each age grouping was statistically significant at the 0.05 level with weak to moderate correlations. For the 16-24 year olds and 25-44 year olds, there were weak negative correlations with the No vote (r= -.369 and r= -.391 respectively) indicating that statistically there was some likelihood of the vote being No in areas with lower proportions of younger people (i.e. the more rural LAs). For the 45-65 year old and over-65 year old categories there were moderate positive 8
% Economically Actie - Part Time Employed % Economically Actie - Unemployed % Economically Actie - Self Employed correlations with the No vote (r=.416 and r.388 respectively) meaning statistically LAs with a higher proportion of older people (i.e. more rural LAs) were likely to vote No. The mean age within a LA also had a statistically significant (ρ<0.018) positive Pearson s correlation with the No vote (r=.417) - i.e., the higher the mean age, the more likely a No vote. Whilst the proportion of economically active people in full time employment did not have a statistically significant relationship with voting results at LA level, there were relationships for self-employed, unemployed and part-time employees as indicated in the graphs in Figure 5. Figure 5 Relationships between the No vote and employment status distinguishing between rural and urban Local Authorities 8% 1 7% 1 1 5% 1 3% 8% 1% 18% 17% 1 15% 1 13% 1 11% 1 There was a very strong negative correlation (r = -.802, ρ<0.001) between the proportion of unemployed people in a LA and the proportion of people voting No. With generally lower levels of unemployment found in rural Scotland, this means that there was a higher likelihood of a No vote in low unemployment rural LAs. In direct contrast there was a very strong positive correlation (r=.767, ρ<0.001) between the proportion of economically active people that are self-employed within a LA and the No vote, and we know that rural Scotland is characterised by high levels of self-employment. A moderate positive correlation (r=.503, ρ=0.003) was also found for part-time employees meaning in LAs with higher levels of parttime employment (many rural LAs) there was a higher propensity to vote No. Conclusions It is clear from this analysis that the rural LAs in Scotland were more likely to have voted for the status quo in the 2014 independence referendum than urban local authorities. Whilst many statistically significant relationships have been found between voting patterns and socio-economic characteristics of LAs we cannot conclude there is any causality between these factors and the voting outcome (i.e. we cannot assume that if the level of selfemployment were to increase there would be a higher No vote). These relationships demonstrate that there are characteristics of being rural that have a relationship with the outcome of the referendum. It would appear that rural people voted differently from urban people, and factors such as country of birth, national identity, age and employment status played their part along with a variety of factors personal to each individual. 9