¹ The most recent instance of these tendencies was in Greenwich where the Conservative vote fell from 35 per cent in 1983 to 11 percent in the 1987 by-election, a collapse which bore little relation to national trends but was evidently the product of the polls' depiction of the local electoral situation early in the campaign.

² See, for example, Schwartz, Walter, ‘Bishop Tells Tactical Voters to Hit Tories’, Guardian, 5 06 1987Google Scholar; Seabrook, Jeremy, ‘Why A Tactical Vote is a Vote for the Poor’, Guardian, 18 05 1987Google Scholar; Crewe, Ivor, ‘Why Neil Need Not Lie Down: Tactical Voting and a Small Swing Against the Tories Could Beat Maggie’, Today, 15 05 1987.Google Scholar

³ Linton, Martin, ‘Charting a Path Through the Tactical Voting Forest’, Guardian, 25 05 1987Google Scholar; and also ‘Tactical Voting: How to Unseat the Member for Finchley’, New Statesman, 22 05 1987.Google Scholar See also Shaw, Andrew and Fishman, Nina, ‘TV '87: Tactical Voting Makes Votes Count’ in Crewe, Ivor and Harrop, Martin, eds, Political Communications: The General Election Campaign of 1987 (Cambridge: Cambridge University Press, forthcoming).Google Scholar

⁴ From survey evidence it appeared that the second preferences of Alliance voters were dividing 55/45 per cent in favour of the Conservatives, and that only 20 per cent of them were willing to vote Labour to keep out the Conservative candidate in their constituency. Oakley, Robin, ‘Rise of Tactical Vote in Marginals Could Threaten Many Safe Seats’, The Times, 18 05 1987.Google Scholar

⁵ A large number of constituency polls in marginal seats were conducted by national and local polling organizations during the campaign. Their success in predicting the outcome accurately was mixed to say the least. See Waller, Robert, ‘Constituency Polls in the 1987 General Election’Google Scholar, in Crewe, and Harrop, , eds, Political Communications.Google Scholar On the pitfalls of tactical voting see Hart, Michael, ‘Why Tactical Voting Does Not Always Work’, Sunday Telegraph Election GuideGoogle Scholar; Vallely, Paul, ‘Tactical Voting Fails to Impress the Analysts’, The Times, 11 06 1987.Google Scholar

⁶ Cook, Robin, ‘Beware a Tactical Vote Hijack’, Guardian, 8 05 1987.Google Scholar

⁷ Crewe, Ivor and Särlvik, Bo, Decade of De-alignment: the Conservative Victory of 1979 and Electoral Trends in the 1970s (Cambridge: Cambridge University Press, 1983)Google Scholar; Curtice, John, Jowell, Roger and Heath, Anthony, How Britain Votes (Oxford: Pergamon, 1985)Google Scholar; Himmelweit, H., Humphreys, P., Jaeger, M. and Katz, M., How Voters Decide (London: Academic Press, 1981)Google Scholar; Dunleavy, Patrick and Husbands, Christopher T., British Democracy at the Crossroads: Voting and Party Competition in the 1980s (London: Allen & Unwin, 1985).Google Scholar

⁸ We restrict ourselves to the 633 constituencies in Great Britain, omitting the seventeen seats in Northern Ireland where the major British parties are not contenders.

⁹ The Chow test is an F-test which compares coefficients in several samples with those in a joint sample comprised of all of these; if the restrictions implied by lumping all samples into one are rejected, then the hypothesis that coefficients are the same in each is rejected. The test can be calculated from sums of squared residuals (SSR) as reported in Table 1. See Chow, G. C., ‘Tests of Equality Between Sets of Coefficients in Two Linear Regressions’, Econometrica, 28 (1960), 591–605.CrossRefGoogle Scholar

¹⁰ We use PC-GIVE (Hendry, D. F., PC-GIVE: An Interactive Menu-Driven Econometric Modelling Program (Oxford: Institute of Economics and Statistics, 1986)Google Scholar and RATS (Doan, T. A., and Litterman, R. B., RATS (Regression Analysis of Time Series) Version 2.00 (Minneapolis: VAR Econometrics, 1986.)Google Scholar

¹¹ Regressions below include, of course, only four of the five so as not to produce perfect multi-collinearity with the constant term. The (BBC-classified) regions are 1: North of England; 2: Midlands; 3: South; 4: Scotland; 5: Wales.

¹² We exclude nationalist and other small parties from the sample; i.e. constituencies where one of these parties was first or second do not appear.

¹³ Regressions without the percentage majority variable, reported in an earlier version of this Note, show somewhat lower t-statistics on the tactical dummies: approximately 3.6 for TACTICL and 5.8 for TACTICA.

¹⁴ Standard error of this coefficient estimate = 0.45.

¹⁵ Standard error = 0.32.

¹⁶ Curtice, John and Steed, Michael, Appendix 2 in Butler, David and Kavanagh, Dennis, eds, The British General Election of 1987 (London: Macmillan, 1988).Google Scholar

¹⁷ Curtice, John and Steed, Michael, ‘Electoral Choice and the Production of Government: The Changing Operation of the Electoral System in the United Kingdom Since 1955’, British Journal of Political Science, 12 (1982), 249–98CrossRefGoogle Scholar; Johnston, R. J., ‘Regional Variations in British Voting Trends 1966–1979: Tests of an Ecological Model’, Regional Studies, 15 (1981), 23–32CrossRefGoogle Scholar; Johnston, R. J., The Geography of English Politics (London: Croom Helm, 1985).Google Scholar

¹⁸ See Himmelweit, et al. , How Voters DecideGoogle Scholar; Crewe, and Särlvik, , Decade of DealignmentGoogle Scholar; Heath, et al. , How Britain Votes.Google Scholar