The 2005 General Election in Scotland
All general elections in Scotland are interesting. Despite the fact that Labour has been the dominant party since the 1960s, the country has developed a complicated four-party system that gives rise to different forms of party competition in different constituencies. The Scottish National Party (SNP) became Labour's principal opponent in the 1970s; geographical concentrations of support have enabled the Liberal Democrats to win significant numbers of seats in recent elections; the Conservative party has been relegated almost to'fringe' status but persists. In 2005, however, the general election in Scotland was even more interesting than usual because it was fought on very substantially revised constituency boundaries.
In the past, electorates in Scottish constituencies have been significantly smaller than those in English seats. In 2001, for example, English constituencies averaged almost 70,000 electors compared with just over 55,000 in Scottish constituencies. As part of the deal that brought about the establishing of the Scottish Parliament in the late 1990s, it was agreed that Scottish Westminster constituencies should be brought more into line as soon as possible. Achieving this necessitated major changes to boundaries. The number of constituencies was reduced from 72 to 59 and only three seats survived unscathed - East Renfrewshire (formerly Eastwood), Orkney and Shetland and Na h-Eileanan An Iar (formerly Western Isles). Only two had relatively minor changes to their boundaries (West Aberdeenshire and Kincardine, and North East Fife). The remaining 54 changed very significantly and in many cases out of all recognition as compared with the old constituencies.
In order to allow analysis of electoral change at constituency level,'notional' 2001 votes for the parties in the new constituencies were calculated and published (see Denver, Rallings and Thrasher 2004). The intention was to produce estimates of how the votes cast for the different parties in 2001 would have been distributed in the new constituencies. Calculating these estimates is not straightforward, however. It has to be assumed that everyone would have voted in the new constituencies exactly as they did in the old, since it is actual 2001 votes that have to be redistributed. It is plain, however, that in many cases, electors would not have voted in the same way had the new boundaries been in place because the tactical situation would have been different. In addition, since local council wards form the building blocks of constituencies and general election votes in wards are not available, local election results have to be used to estimate the electoral effects of different areas moving between constituencies. There are numerous difficulties in extrapolating from local to general elections, however, the most obvious being that the major parties do not contest all wards.
As a preliminary to analysing the 2005 results, therefore, it is worth attempting to assess how realistic the 2001'notional' votes were. Table 1 shows first the standard deviations (which measure spread) of changes in the parties' vote shares across constituencies from the 2001 estimates to the 2005 result and compares these with changes between 1997 and 2001 (when constituencies stayed the same). The figures are encouraging since they are of the same order in the two cases. Secondly, on the assumption that there were relatively similar movements across Scotland between 2001 and 2005 we would expect strong correlations between each party's estimated vote share in the new constituencies in 2001 and its actual share in 2005. As the data in the table show, the correlations are indeed very strong and the coefficients are very similar in size to those measuring the similarity in patterns of support between 1997 and 2001.
Nonetheless, there were a few constituencies where a party's 2005 vote share was significantly smaller or larger than would have been predicted on the basis of the 2001 estimates. These are listed in Table 5 (below) and most can be easily explained. Overall, then, the 2001'notional' vote distributions in constituencies do not seem unreasonable. Indeed, it is worth noting that Na h-Eileanan An Iar produced an unexpectedly bad performances by Labour and the Conservatives and an unexpectedly good one by the SNP while in Orkney and Shetland the Conservatives had a particularly poor result. In both cases there was no change in boundaries at all.
Table 1: Comparing change in parties' shares of votes across constituencies 1997-2001 and 2001-05
Note: The Speaker's seat is excluded from all calculations.
The National Result
Table 2 shows the outcome of the election in Scotland. Turnout rose by about the same amount as it did in Britain as a whole and it seems likely that this was a consequence of the fact that, whereas in 2001 Labour was expected to win easily (and a landslide duly transpired), the polls during the 2005 campaign consistently reported that - at least in terms of votes - the result would be much closer this time. In addition, for a variety of reasons, there may have been an increased determination to vote against the government on the part of some voters.
As elsewhere, the Conservatives had a very small increase in terms of popular support (+0.2 percentage points) but remained firmly in fourth place in Scotland. Nonetheless they retained a toe-hold on representation by gaining a seat (Dumfriesshire, Clydesdale and Tweeddale) that was notionally Labour in 2001. The increase in the Liberal Democrat share of the votes (+6.3) was larger than in England and Wales (+3.6) and took them into second place in Scotland. In addition to holding all the seats assigned to them in the 2001 estimates, the Liberal Democrats added East Dunbartonshire and Inverness, Nairn, Badenoch and Strathspey, taken from Labour. The SNP dropped back in terms of vote share but added two seats to their 2001 notional total - Na h-Eileanan An Iar and Dundee East - both formerly Labour seats. So Labour lost five seats in all as a consequence of a sharp decline (-4.4) in vote share, even although this was a less sharp drop than Labour experienced in the rest of Britain (-5.9). Nonetheless, the operation of the first-past-the-post electoral system ensured that Labour remains heavily over-represented. With just 39% of the votes, the party took 68% of the seats.
Table 2: Share of Votes and Number of Seats Won and Changes from 2001
Note. The Speaker, who was not opposed by the Conservatives or Liberal Democrats in Glasgow North East, is treated as an'other'. The change in the number of seats is from the'notional' 2001 results.
The most successful of the small parties was the Scottish Socialist Party (SSP) with 1.9% of the votes but this was well down on its 2001 share (3.1%). The Green party's 19 candidates got 1.1% of the votes but outpolled the SSP in every constituency (18) where both stood. UKIP, with 22 candidates, made little impression.
Turnout: Constituency Variations
There is little doubt that turnout increased across the board. Even using 2001 estimates - which are liable to error - increases were recorded in 50 constituencies and only marginal decreases in nine. The biggest increase was in East Dunbartonshire (+10.1) which had been transformed into something approaching a three-way marginal and which provides evidence that the electorate noticed - or were informed by the parties about - how the electoral status of some seats had changed. Overall, there was a significant positive correlation (+0.268) between the marginality of constituencies according to the 2001 estimates and the change in turnout. The more marginal the seat, the more turnout increased.
In advance of the election, Labour campaigners feared that some of their supporters might vent their displeasure with the government - especially over Iraq - by not turning out to vote but the aggregate turnout data suggest that these fears were not realised either in Scotland or in the rest of country. Even in Labour's safest Scottish seats (majority over 30%) turnout was up by 1.9% on average (N = 21). There was a slight tendency for turnout to increase more the stronger the Conservatives were in a constituency in 2001 but for the other parties there was no significant association between their previous strength and turnout change. There is also no evidence that any of the major parties was particularly harmed or advantaged by the changes in turnout. In each case the correlation coefficient measuring the association between turnout change and change in a party's vote share fails to reach statistical significance.
Turning to actual turnout in 2005, we encounter a much more predictable and familiar pattern of variation across Scottish constituencies (see Denver and Hands 2004). The lowest turnout was in Glasgow Central (43.8%) - and the next four lowest were also in Glasgow seats - while East Dunbartonshire and East Renfrewshire shared the top spot, both recording 72.1%.
The general pattern of constituency variations in turnout is most effectively described by correlation coefficients which measure the strength of association between turnout and variables indicating the socio-economic characteristics of constituencies and the electoral context. This is done in Table 3. The socioeconomic variables are derived from the 2001 census and they are all (except % private renters) significantly related to turnout. Turnout was higher in constituencies with more professional and managerial households, owner occupiers, people aged 65 and over, people employed in agriculture, people with degrees and Church of Scotland adherents. Turnout was lower the more manual workers, social renters (from the council or a housing association), young voters, persons per hectare, people from ethnic minorities, households without a car and Roman Catholics. In general, then, - and it is nothing new - turnout was higher in more affluent and rural areas and lower in poorer and more urban areas. Like Britain as a whole, Scotland is divided into relatively low turnout and relatively high turnout constituencies and the two are very different in social terms.
Table 3: Bivariate correlations between turnout in 2005 and constituency characteristics
Notes: All coefficients are statistically significant except the one asterisked. Glasgow North East (Speaker's seat) is excluded from all calculations (N=58).
The closeness of the contest in a constituency in the previous election (marginality) is also regularly associated with turnout levels. Parties put greater campaign efforts into more marginal seats (and, these days, virtually ignore those that are either very safe or hopeless for the party concerned) and unsurprisingly these efforts bear fruit in higher turnouts. The 2005 election was no exception. Although the previous marginality of Scottish seats could only be estimated it was strongly related to turnout. On the other hand, while it might be thought that the larger the array of candidates that are available the more people would be inclined to vote (since it would be more difficult to use the excuse that there is no-one to vote for) the better ‘'others' (including the SSP) did in a constituency, the lower was turnout.
While bivariate correlations are interesting and important in themselves, they constitute only a first step in analysis. One problem in interpreting them is that the various measures of the social composition of constituencies are themselves highly inter-correlated. Thus, areas in which a large proportion of households are owner occupiers tend also to have large proportions of professional and managerial workers and small proportions of social renters. Secondly, bivariate analysis provides no clue as to the joint impact of a number of variables or to the impact of any single variable once others are taken into account. Thus, we might want to know how much of the variation in turnout is explained by class and housing tenure together or whether constituency marginality affects turnout once the class composition of the constituency is taken into account.
We can begin to deal with these problems by undertaking multiple regression analysis. This enables us to sort out which variables are the most important influences on turnout, see whether a particular variable remains significant when all others are held constant and evaluate how successfully combinations of variables explain turnout variations. The results of a stepwise regression involving the socio-demographic variables and marginality show that just four variables (the percentages of owner occupiers, those aged 65 and over, manual workers and those belonging to an ethnic minority) account for 77% of the variation in constituency turnouts across Scotland. Each of these has an independent effect on turnout even when scores on the other three are taken into account. It should be emphasised, however, that this does not necessarily mean that the groups mentioned were more or less likely to vote. It is the characteristics of constituencies that is being analysed here, not the behaviour of individuals or groups.
Patterns of Party Support
Changes in party support from 2001 to 2005 were not uniform across Scotland. Table 4 groups constituencies into cities and regions (in order to minimise any errors that may have been made in calculating notional votes in individual constituencies). Labour declined everywhere (although by varying amounts) except in the three South of Scotland constituencies. This was also the only region in which there was a fall in Liberal Democrat support and a significant increase in votes for the Conservatives and it also saw the biggest SNP decline. All of these unusual figures for the South of Scotland can be explained by the new tactical situations in Dumfries and Galloway and Dumfriesshire and also the Liberal Democrat slump in Berwickshire, Roxburgh and Selkirk (probably caused by the retirement of the long-time Liberal Democrat incumbent, Archie Kirkwood). Otherwise, the SNP managed to increase its vote share, against the national trend, in North East Scotland and Dundee while the Liberal Democrat increase in Dunbartonshire (mainly due to the performance of the youngest candidate to be elected, Jo Swinson, in East Dunbartonshire) was dramatic.
Table 4: Changes in vote shares of major parties in cities and regions 2001-5
Note: Glasgow North East is excluded from the calculations for Glasgow.
Changes in party support were even more variable at constituency level. Indeed, as compared with the 2001 estimates all the major parties managed to increase their vote share in some seats while losing support in others. The Conservatives were down in 31 seats, up in 25 and stayed at exactly the same level in two. That is not perhaps surprising given the very small change in their overall vote share but Labour had an improved performance over 2001 in five seats and the SNP in ten while the Liberal Democrats lost ground in five constituencies.
The best and worst performances by each of the parties can be easily identified by a series of simple regression analyses predicting each party's vote share in each constituency on the basis of its estimated share in 2001 and comparing the prediction with the actual 2005 outcome. The best and worst results for each party are given in Table 5.
Most of these are relatively easy to understand. As has been mentioned, the retirement of a popular incumbent probably explains why the Conservatives did better and the Liberal Democrats much worse than would have been expected in Berwickshire, Roxburgh and Selkirk. The dramatically changed tactical situation facing voters in Dumfries and Galloway - the old Galloway and Upper Nithsdale constituency had been a Conservative-SNP marginal whereas the new seat was expected to see a close Labour-Conservative contest - probably explains why Labour did better and the SNP and Liberal Democrats worse than would have been expected. It is possible that a significant proportion of former tactical voters for the SNP in the Galloway part of the seat switched to Labour. Similarly, the 2001 estimates revealed that the Liberal Democrats were in the best position to challenge Labour in East Dunbartonshire and the seat based on Inverness while the Conservatives were best placed in Dumfriesshire, hence the good performances by them and the relatively poor results for other challenging parties. On the other hand, the figures for Ross, Skye and Lochaber - good for the Liberal Democrats, bad for Labour - probably reflect the personal appeal of Charles Kennedy, the party leader.
In the case of Inverclyde, however, it seems likely that, on the basis of local election results, the strength of the Liberal Democrats in the wards that were transferred into the constituency was somewhat overestimated and that of the SNP underestimated.
Table 5: Best and worst results for each party
The figures show by how much the party in question exceeded or fell short of its predicted share of the vote. For the statistically minded, the cases shown are those where the residual (actual minus predicted score) was at least 1.5 standard deviations away from the mean.
Variations in how the parties' support changed at constituency level were partly affected by variations in the electoral context. Table 6 shows the changes in the aggregate shares of the votes according to which two parties were estimated to be in first and second places in 2001. The Conservatives generally did best where they were challenging Labour (+1.3%) but also increased support slightly where they were lying second to the SNP or the Liberal Democrats. They lost ground, however, where Labour and the Liberal Democrats were the leading contenders - not something, of course, that would worry the Conservatives unduly. It was in the latter seats that Labour's losses were heaviest and Liberal Democrat advances strongest. While the Liberal Democrats may have reached something of a ceiling in their contests with the Conservatives (+0.9%), there is no evidence of this where they were challenged by Labour (+10.5%). Although the SNP's vote share declined in each category, some comfort could be taken from the fact that the decline was smaller in seats that they held from the Conservatives (-0.1%) and where they were lying second to Labour (-1.9%).
Table 6: Overall changes in vote shares in different electoral contexts
As with turnout, when we turn from change between elections to variations in absolute levels of support for the parties across constituencies we reach familiar territory in that the patterns are very similar to those evident in previous elections. Correlation coefficients measuring the associations between the shares of the vote obtained by the parties in the constituencies and a standard set of socio-demographic variables are shown in Table 7 and the data contain few surprises. It is noticeable, first, that relative support for the Conservatives and Labour is more clearly structured by the socio-demographic character of constituencies than is support for the Liberal Democrats or, especially, the SNP. Secondly, two variables do not correlate at all with levels of party support - the percentages of young people and of ethnic minority residents. The non-significance of the former is somewhat surprising but the distribution of the ethnic minority population is very heavily skewed - less than 2% of the population in 42 constituencies and greater than 5% in only 4 - so that it is unrelated to levels of party support.
Looking at the parties separately, the Conservatives, as might be expected, had a larger share of the vote in constituencies where there were more professional and managerial workers, owner occupiers, private renters, older voters, people with degrees, people associated with the Church of Scotland and in more rural areas. They performed less well where there were more manual workers, social renters, people belonging to ethnic minorities, households having no car, Roman Catholics and in more urban areas. The pattern of Labour support is almost an exact mirror image of that for the Conservatives. On the other hand, where the coefficients are significant, the pattern of support across constituencies for the Liberal Democrats is similar to the Conservative pattern, although it is interesting to note that the strength of the Church of Scotland is not related to the size of the Liberal Democrat vote. Only four variables are significantly related to SNP strength in constituencies - the party is weaker where there are more professionals and managers, more people with degrees and in more urban areas while being stronger where there are more manual workers.
The extent to which the variables listed in Table 7 are themselves inter-related is illustrated by a series of regression analyses predicting the parties' vote shares in constituencies on the basis of socio-demographic characteristics. Only two variables are significant for the Conservatives (% in agriculture and % professional and managerial) and together they account for 69% of the variation in Conservative vote shares. For Labour, 71% of the variation is accounted for by % social renters, % in agriculture, % owner occupiers and % Roman Catholic. Only one variable is significant in the case of the Liberal Democrats (% social renters) and only 25% of the variation is explained while for the SNP the variation explained is even smaller (20%), although two variables are significant (% professional and managerial and % social renters).
Table 7: Bivariate correlations between party shares of vote in 2005 and constituency characteristics
Note: All coefficients are statistically significant except those asterisked; (N = 58).
It is interesting to note that, even although there are many fewer constituencies involved, variations in Conservative and Labour support across Scotland are less predictable on the basis of the social characteristics of constituencies than is the case across Britain as a whole. The same variables (other than religious denomination) explain 81% of the variation in the Conservative vote and 76% of the variation in the Labour vote across all British constituencies. The reason why the Scottish figures are smaller is because of the complexity of the party system. For a variety of reasons, different parties are in competition in socially similar areas. The electorate is aware of this and hence people of similar social backgrounds vote for different parties in different constituencies.
All four major parties in Scotland could take something positive from the 2005 general election. In somewhat difficult circumstances, Labour remained easily the dominant party in terms of votes and won two-thirds of the seats. The Liberal Democrats had their largest share of the vote since the heyday of the Liberal-SDP Alliance in 1983 and advanced to 11 seats. The SNP vote share fell back but, presumably due to efficiently targeted campaigning, picked up two additional seats. Even the Conservatives inched forward from the nadir of 2001 and ‘'gained' a seat - even if it was not the one that most people expected.
On the other hand, there were also disturbing signs for each party. Labour's vote share was its lowest in Scotland since 1983 and this evidences considerable disenchantment with the government. The Liberal Democrats are now in second place to Labour in 16 seats and might hope at some stage to win three of them - Aberdeen South, Edinburgh South and Edinburgh North and Leith. After that, however, the gaps are large and it is difficult to see where they could go from there. Moreover, the context of the election was ideal for the Liberal Democrats - an unpopular government and a not very popular Opposition - and it may well be that such a propitious situation will not be repeated next time round. For the SNP, the decline in vote share must be worrying. Although the party's first priority is probably the Scottish Parliament elections, wining only 18% of the general election votes does diminish its claim to speak for Scotland. Finally, the Conservatives remain in a pickle. Besides winning one seat, they came second in only 15 and of these only Dumfries and Galloway, Angus and Perth and North Perthshire are within reasonable reach. The road back for Scottish Conservatives at general elections still looks hard and rocky.
Denver, D., Rallings, C. and Thrasher, M. (2004) Media Guide to the New Scottish Westminster Parliamentary Constituencies, LGC elections centre, Plymouth.
Denver, D. and Hands, G. (2004) ‘'Exploring Variations in Turnout: Constituencies and Wards in the Scottish Parliament Elections of 1999 and 2003', British Journal of Politics and International Relations, Vol 6, No. 4, pp. 527-42.
David Denver is Professor of Politics at Lancaster University.
(online 26 May 2006)