The PM and the pendulum: Dynamic forecasting of British elections

British Journal of Political Science

Volumn 37, Issue 1, 2007, Pages 71-87

  Lebo, Matthew J ^b   Norpoth, Helmut ^a  

^a STONY BROOK UNIVERSITY   (United States)

^b NONE

Author keywords

[No Author keywords available]

Indexed keywords

EID: 33846063798     PISSN: 00071234     EISSN: 14692112     Source Type: Journal    
DOI: 10.1017/S000712340700004X     Document Type: Article

References (53)

1. See Paul Whiteley, 'Electoral Forecasting from Poll Data: The British Case', British Journal of Political Science, 9 (1979), 219-36;
2. David Sanders, 'Popularity Function Forecasts for the 2005 UK General Election', British Journal of Politics and International Relations, 7 (2005), 174-90.
3. An earlier version of this model proved uncanny with its forecast of a Conservative victory in 1992 when many polls, including exit polls, picked Labour to win. See David Sanders, 'Government Popularity and the Next General Election', Political Quarterly, 62 (1991), 235-61.
4. See Anthony Mughan, 'General Election Forecasting in Britain: A Comparison of Three Simple Models', Electoral Studies, 6 (1987), 195-207;
5. S. Lewis-Beck, R. Nadeau and E. Belanger, 'General Election Forecasts in the United Kingdom: A Political Economy Model', Electoral Studies, 23 (2004), 279-90;
6. Helmut Norpoth, 'Forecasting British Elections: A Dynamic Perspective', Electoral Studies, 23 (2004), 297-305;
7. and Paul F. Whiteley, 'Forecasting Seats from Votes in British General Elections', British Journal of Politics and International Relations, 1 (2005), 165-73.
8. See Whiteley, 'Forecasting Seats from Votes', for this feature as well.
9. A graphic display of this swing of the pendulum, however, uses Big Ben as a backdrop. See Sir Ivor Jennings, The British Constitution, 5th edn (Cambridge: Cambridge University Press, 1966), p. 52.
10. For a recent state-of-the-art volume on this subject, see Han Dorussen and Michaell Tayler, eds, Economic Voting (London: Routledge, 2002);
11. for reviews, see Michael Lewis-Beck and Mary Stegmaier, 'Economic Determinants of Electoral Outcomes', Annual Review of Political Science, 3 (2000), 183-219;
12. and Helmut Norpoth, 'The Economy', in Lawrence LeDuc, Richard Niemi and Pippa Norris, eds, Comparing Democracies (Thousand Oaks, Calif.: Sage Publications, 1996), pp. 299-318.
13. Angus Campbell, Philip E. Converse, Warren E. Miller and Donald E. Stokes, The American Voter (New York: Wiley, 1960), pp. 554-6.
14. Martin Paldam, 'How Robust is the Vote Function?' in Helmut Norpoth, Michael Lewis-Beck and Jean-Dominique Lafay, eds, Economics and Politics: The Calculus of Support (Ann Arbor: University of Michigan Press, 1991), pp. 9-31;
15. Anthony Downs, An Economic Theory of Democracy (New York: Harper and Row, 1957), pp. 55-60;
16. and John E. Mueller, War, Presidents and Public Opinion (New York: Wiley, 1973), chap. 9.
17. George Udny Yule, 'On a Method of Investigating Periodicities in Disturbed Series, with Special Reference to Wolfer's Sunspot Numbers', in Statistical Papers of George Udny Yule (New York: Hafner, 1971), pp. 389-420 (originally published in 1927).
18. Thanks to Box and Jenkins, processes of this sort have become widely familiar as ARIMA models. See George E. P. Box and Gwilym Jenkins, Time Series Analysis: Forecasting and Control (San Francisco: Holden-Day, 1976).
19. The vote data come from F. W. S. Craig, British Electoral Facts 1832-1980 (Chichester: Parliamentary Research Service, 1981), p. 49 (elections 1832-1935),
20. and Anthony King, Britain at the Polls, 2001 (New York: Chatham House, 2002), Appendix, p. 233 (elections 1945-2001).
21. Statistical tests confirm the existence of equilibrium for British elections through the rejection of the null hypothesis of random-walk behaviour (Dickey-Fuller t= -3.6, p < 0.05; Phillips-Perron t= -3.7, p < 0.01). On the concept of electoral equilibrium and a random-walk model of elections, see Donald E. Stokes and Gudmund R. Iversen, 'On the Existence of Forces Restoring Party Competition', in Angus Campbell, Philip E. Converse, Warren E. Miller and Donald E. Stokes (eds.), Elections and the Political Order (New York: John Wiley, 1966), pp. 180-93. For an application to British elections, see Norpoth, 'Forecasting British Elections'.
22. Yule, 'On a Method of Forecasting Periodicities', p. 403.
23. This is the same dynamic that was also identified in American presidential elections and first used in forecasting the 1996 contest. See Helmut Norpoth, 'Is Clinton Doomed? An Early Forecast for 1996', PS: Political Science & Politics, 28 (1995), 201-7; for 2004,
24. see Helmut Norpoth, 'From Primary to General Election: A Forecast of the Presidential Vote', PS: Political Science & Politics, 37 (2004), 737-40.
25. For evidence in British elections, see David Butler and Donald E. Stokes, Political Change in Britain (New York: St Martin's Press, 1969), pp. 220-3.
26. For a classic definition, see Gerald H. Kramer, 'Short-Term Fluctuations in U.S. Voting Behavior' .American Political Science Review, 65 (1971), 131-43;
27. also Morris P. Fiorina, Retrospective Voting in American National Elections (New Haven, Conn.: Yale University Press, 1981).
28. C. A. E. Goodhart and R. J. Bhansali, 'Political Economy', Political Studies, 18 (1970), 43-106;
29. John Hudson, 'Prime Ministerial Popularity in the UK: 1960-81', Political Studies, 32 (1984), 86-97;
30. Helmut Norpoth, Confidence Regained: Economics, Mrs. Thatcher and the British Voter (Ann Arbor: University of Michigan Press, 1992), chap. 8;
31. David Sanders, 'Government Popularity and the Next General Election', Political Quarterly, 62 (1991), 235-61.
32. This is the measure used in the forecast model of Lewis-Beck et al., 'General Election Forecasts in the United Kingdom'.
33. Butler and Stokes, Political Change in Britain, chap. 17.
34. Harold D. Clarke, David Sanders, Marianne C. Stewart and Paul Whiteley, Political Choice in Britain (Oxford and New York: Oxford University Press, 2004), chap. 4;
35. Norpoth, Confidence Regained, chaps. 5-7.
36. Harold D. Clarke and Marianne C. Stewart, 'Economic Evaluations, Prime Ministerial Approval and Governing Party Support in Britain: Rival Models Reconsidered', British Journal of Political Science, 25 (1995), 145-70;
37. Harold D. Clarke and Matthew Lebo, 'Fractional (Co)integration and Governing Party Support in Britain', British Journal of Political Science, 33 (2003), 283-301;
38. and Richard Nadeau, Richard B. Niemi and Timothy Amato, 'Prospective and Comparative or Retrospective and Individual? Party Leaders and Party Support in Great Britain', British Journal of Political Science, 26 (1996), 245-58.
39. Clarke and Lebo, 'Fractional (Co)integration', p. 285.
40. The question typically asks: 'Are you satisfied or dissatisfied with [name] as Prime Minister?' For elections from 1945 to 1997 we relied on Gallup polls. See Anthony King, ed., British Political Opinion 1937-2000: The Gallup Polls (London: Politico, 2001). With that source no longer available, we have turned to MORI, ICM and YouGov polls since then. See www.mori.com/polls/ trends/ and pollingreport.co.uk. Whenever possible, we formed the average from surveys two and three months prior to the date of a general election. For a detailed explanation of how we constructed the measure of prime ministerial approval, see the Appendix.
41. This threshold closely approximates the one for presidential elections in the United States when incumbent presidents are running. Only Harry Truman (in 1948) won re-election with an approval rating below 50 points.
42. A special adjustment had to be made for the 1945 election. With Labour having been part of the wartime government, we discounted the partisan effect of Prime Minister Churchill's satisfaction rating by a factor of 2. His satisfaction rating prior to the 1945 election (87 per cent) thus enters as 43.5 per cent into the analysis.
43. PM satisfaction ratings are inverted at the midpoint of the scale (50), so we subtract 50 from every rating first and then multiply this difference by - 1 for a Labour prime minister. Hence a Labour prime minister with a rating of 58 per cent, would be scored as (58 -50) × -1 = -8, while a Tory prime minister with a 58-point rating would count as (58 -50) = 8. Also recall that ratings are constrained to be below 65 and that the raw satisfaction percentages reported in polls are adjusted for the extent of third-party support. For more details, see the Appendix.
44. Lewis-Beck cites the MAE as 'a rough idea of how much real error the model generates, at least after the fact' (Michael Lewis-Beck, 'Election Forecasting: Principles and Practice', British Journal of Politics and International Relations, 7 (2005), 145-64).
45. See also James E. Campbell, 'Polls and Votes: The Trial-Heat Presidential Election Forecasting Model, Certainty, and Political Campaigns', in James E. Campbell and James C. Garand, eds, Before the Vote: Forecasting American National Elections (Thousand Oaks, Calif: Sage, 2000).
46. This, we must admit, differs from our forecast issued on 24 March 2005, which predicted a Tory vote victory by 3.8 points. The difference is due to a revision of Blair's approval rating and the adjustment factor for third-party strength. The March forecast relied on a single polling organization, whereas the forecast reported here combined the ratings from all we were able to locate, after the fact (MORI, ICM and YouGov). We also used the average of third-party support in these polls to adjust Blair's 'raw' approval rating.
47. The average of polls used in this analysis put Blair's 'raw' approval at 39.7 per cent and the third-party share at 28.7 per cent. That converts to an adjusted rating of 39.7/(1 - 0.287) = 55.7.
48. For a discussion of the causes and changing nature of the bias in electoral proportionality, see: Ron Johnston, Charles Pattie, David Rossiter, Danny Dorling, Iain MacAllister and Helena Tunstall, 'New Labour's Landslide and Electoral Bias : An Exploration of Differences between the 1997 UK General Election Result and the Previous Thirteen', in J. Fisher et al., eds, British Elections and Parties Review 9 (London: Cass, 1999).
49. This differs from our seat forecast issued on 24 March 2005, which predicted a Labour lead of only twenty-six seats. The change is the result of revising the predicted vote lead from 3.8 points in favour of the Tories to one favouring Labour by 1.3 points. Also note that the out-of-sample prediction for 2005 reported in Table 5 is based on the actual vote in 2005, not the predicted vote.
50. Whiteley, 'Forecasting Seats from Votes'.
51. Each probability distribution is defined by the coefficient and standard error estimated in the 1945-2001 models. For example, in each simulation, the coefficient for the PM variable is drawn from a normal distribution with a mean of 0.617 and a standard deviation of 0.058. Autoregressive parameters are constrained between 1 and -1.
52. Based on a series of polls from June 2003-February 2005, we define this distribution as normal with a mean of 37.5 and a standard deviation of 1.5.
53. The parameters of the 1945-2001 SEATS model give us the distribution from which we draw randomly a constant, a coefficient for VOTE, and an autoregressive parameter for our simulations. Note that this procedure uses no information from the 2005 election itself and thereby differs from our 2005 seat forecast based on the actual 2005 vote, which gives Labour a lead of 132 seats (Table 5).