Simulating the effects of the alternative vote in the 2010 UK general election1

Parliamentary Affairs

Volumn 64, Issue 1, 2011, Pages 5-23

  Sanders, David ^a   Clarke, Harold D ^a   Stewart, Marianne C ^b   Whiteley, Paul ^a  

^a UNIVERSITY OF ESSEX   (United Kingdom)

^b UNIVERSITY OF TEXAS AT DALLAS   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 78651269869     PISSN: 00312290     EISSN: 14602482     Source Type: Journal    
DOI: 10.1093/pa/gsq042     Document Type: Article

References (28)

1. The authors thank three anonymous reviewers for their helpful comments and suggestions. The data presented in this paper were gathered in the 2010 British Election Study which was funded by UK Economic and Social Research Council research grant RES-552-25-0001.
2. There is a very extensive literature on electoral systems and their political consequences. Useful recent contributions include, inter alia, A. Blais, To Keep or To Change First Past the Post, Oxford University Press, 2008;
3. Gary Cox, Making Votes Count, Cambridge University Press, 1997;
4. D. Farrell, Electoral Systems: A Comparative Introduction, Palgravem, 2001; M. Gallagher and P. Mitchell (eds), The Politics of Electoral Systems, Oxford University Press, 2008;
5. P. Norris, Electoral Engineering: Voting Rules and Political Behavior, Cambridge University Press, 2004
6. J. Colomer, 'It's Parties That Choose Electoral Systems (or Duverger's Law Upside Down)', Political Studies, 53, 2005, 1-21;
7. K. Benoit, 'Electoral Laws As Political Consequences: Explaining the Origins and Change of Electoral Institutions', Annual Review of Political Science, 10, 2007, 363-90. Technically, the UK's electoral system for the House of Commons should be referred to as Single Member Plurality (SMP). We prefer to use the term FPTP here as this is the common parlance in UK political discourse.
8. P. Dunleavy and H. Margetts, 'The Impact of UK Electoral Systems', Parliamentary Affairs, 58, 2005, 854-70.
9. In some AV systems, statements of indifference are not permitted. In Australia, for example, ranking of all candidates is compulsory. If this is not done, the ballot is invalid.
10. This said, the ballot papers for AV and SV elections can differ. The SV ballot form is sometimes presented as two columns in which voters mark a cross in each column, rather than being required to provide a numerical rank ordering.
11. See Cox, above note 2, and B. Grofman and S. Feld, 'If You Like the Alternative Vote (a.k.a. the instant runoff), Then You Ought To Know About the Coombs Rule', Electoral Studies, 23, 2004, 641-59.
12. This is not to say that tactical voting ceases to be a possibility under AV. For a discussion of the issues and calculations involved, see Cox, above note 2.
13. Although there is a tendency for AV to be associated wth greater proportionality than FPTP, that tendency is by no means invariant. For example, in US House elections and in Indian Congressional elections-both based on FPTP-disproportionality is significantly lower than in Australian parliamentary elections, where AV is employed. In all 21st century general elections, the Gallagher index of disproportionality for the US averages 2.74 and for India 6.18. The corresponding figure for Australia is 9.41. The equivalent UK figure, under FPTP, is 16.53. We are indebted to one of this article's referees for pointing this out to us. See http://www.tcd.ie/Political-Science/staff/michael- gallagher/ElSystems/index.php.
14. The centrality of first preferences is not so marked for the Coombs variant of AV. Using this AV method, the candidate with the most bottom-rankings is eliminated in the first round-rather than the candidate with the least first preferences. For details, see Grofman and Feld, above note 6.
15. See Gallagher and Mitchell, above note 2.
16. The wording is: 'Do you want the United Kingdom to adopt the "alternative vote" system instead of the current "first past the post" system for electing Members of Parliament to the House of Commons?
17. Most political observers assume this variant is the one most likely to be introduced in the UK.
18. D. Sanders, H. Clarke, M. Stewart and P. Whiteley, 'Does Mode Matter for Modeling Political Choice? Evidence from the 2005 British Election Study', Political Analysis, 15, 2007, 257-85.
19. Northern Ireland traditionally has been excluded from BES surveys. The practice was continued in 2010.
20. See, e.g. H. Clarke, D. Sanders, M. Stewart and P. Whiteley, Performance Politics and the British Voter, Cambridge University Press, 2009.
21. This is slightly less than the 17% tactical voting figure generated by a question asked in the BES postelection survey about people's 'reasons for voting'. The 17% figure is composed of 8% who claimed directly that they voted tactically and a further 9% who declared that they did not vote for their preferred party 'because it stood no chance of winning in my constituency'. The discrepancy between the two figures for overall tactical voting may reflect the fact that people are not always particularly good at explaining their own behaviour.
22. The individual-level analyses weight the data by demographics, media consumption patterns and party identification to ensure that the sample is representative of the British electorate. The Ns reported are all unweighted. In fact alternative weighting schemes make very little difference to the calculated distribution ratios. The table immediately below provides an illustration of the general pattern.
23. An adjustment was made for the reduction in the number of valid votes, reflecting the fact that not all respondents who expressed a first preference also expressed a second preference. This adjustment was based on the percentage difference between these two figures as reported in Table 1: a 5% reduction in England and Scotland and a 3% reduction in Wales.
24. UKIP was included because there were sufficient cases to calculate the requisite third preference distribution ratios; for other minor parties, there were not enough cases to do this.
25. Specific adjustments were made for each of the small number of English constituencies where one of the three major parties was placed fourth or lower in terms of vote share. Note that because the aggregate data file used in the computations reported here did not record votes for Independents and for Respect, the vote shares for these groupings are excluded from the calculations. This means that the simulated results for a small number of constituencies, such as Birmingham Hall Green where these groupings performed relatively well, need to be treated with caution.
26. In Brighton, where the Green candidate was the FPTP winner, Labour had a small AV lead over the Greens at the end of Round 4-athough it fell short of the 50% AV winning threshold. Unfortunately, there were too few cases of fourth preferences for the Greens to enable the iteration process to proceed further. We nonetheless reflect on the Brighton case when discussing the results.
27. Viewed differently, Table 7 shows that in 234 constituencies (204 in England, 23 in Scotland, 7 in Wales) AV was unnecessary because there was an outright Round 1 winner. The second preferences of major party suporters are relevant only at Round 3. These preferences were critical in 271 constituencies (202 in England, 36 in Scotland and 33 in Wales). Of the 202 English seats where major party second prefences produced the AV result, 47 were seats where Labour was third, 22 were seats where the Conservatives were third and 13 where the Liberal Democrats were third. A further breakdown of these 202 seats reveals that: (a) of the 83 English seats where Labour won at Round 3, 10 were constituencies where the Conservatives were third; 73 were from where the LibDems were third; (b) of the 75 English seats the Conservative won at Round 3, 15 were constituencies where Labour was third; 60 were from where the Liberal Democrats were third; (c) of the 44 English seats the Liberal Democrats won at Round 3, 32 were in seats where Labour was third; 12 were constituencies where the Conservatives were third.
28. A coalition of Labour, Liberal Democrats and Plaid Cymru would have produced 324 seats; even with Green support (1 further seat), such a coalition would still have been one short of an overall Commons' majority.