¹ Studies have considered the use of objective indicators vs. subjective perceptions (e.g., Sanders, David, Ward, Hugh and Marsh, David (with Tony Fletcher), ‘Governmental Popularity and the Falklands War: A Reassessment’, British Journal of Political Science, 17 (1987), 281–313)CrossRefGoogle Scholar; egocentric vs. sociotropic judgements (e.g., Kiewiet, D. Roderick, Macroeconomics and Micropolitics (Chicago: University of Chicago Press, 1983))Google Scholar; retrospective vs. prospective time frames (e.g., Clarke, Harold D. and Stewart, Marianne C., ‘Economic Evaluations, Prime Ministerial Approval, and Governing Party Support: Rival Models Reconsidered’, British Journal of Political Science, 25 (1995), 145–70CrossRefGoogle Scholar; MacKuen, Michael B., Erikson, Robert S. and Stimson, James A., ‘Peasants or Bankers? The American Electorate and the US Economy’, American Political Science Review, 86 (1992), 597–611)CrossRefGoogle Scholar; and direct vs. mediated assessments (e.g., Lewis-Beck, Michael S., ‘Comparative Economic Voting: Britain, France, Germany, Italy’, American Journal of Political Science, 30 (1986), 315–46).CrossRefGoogle Scholar

² See Hudson, John, ‘Prime Ministerial Popularity in the UK: 1960–81’, Political Studies, 32 (1984), 86–97CrossRefGoogle Scholar; Mishler, William, Hoskin, Marilyn and Fitzgerald, Roy, ‘British Parties in the Balance: A Time Series Analysis of Long-Term Trends in Public Support for Major Parties, British Journal of Political Science, 19 (1989), 211–36CrossRefGoogle Scholar; Clarke, Harold D. and Whiteley, Paul, ‘Perceptions of Macroeconomic Performance, Government Support and Conservative Party Strategy, 1983–1987’, European Journal of Political Research, 18 (1990), 97–120CrossRefGoogle Scholar; Lafay, Jean-Dominique, ‘Political Dyarchy and Popularity Functions: Lessons from the 1988 French Experience’, in Norpoth, Helmut, Lewis-Beck, Michael S. and Lafay, Jean-Dominique, eds, Economics and Politics (Ann Arbor: University of Michigan Press, 1991)Google Scholar; Nadeau, Richard, Niemi, Richard G. and Amato, Timothy, ‘Expectations and Preferences in British General Elections’, American Political Science Review, 88 (1994), 371–83CrossRefGoogle Scholar; Lanoue, David J. and Headrick, Barbara, ‘Prime Ministers, Parties and the Public’, Public Opinion Quarterly, 58 (1994), 191–209.CrossRefGoogle Scholar

³ Clarke and Stewart, ‘Economic Evaluations, Prime Ministerial Approval, and Governing Party Support’, p. 156.

⁴ Bean, Clive and Mughan, Anthony, ‘Leadership Effects in Parliamentary Elections in Australia and Great Britain’, American Political Science Review, 83 (1989), 1165–80, at p. 1167.CrossRefGoogle Scholar

⁵ Nadeau, Richard and Biais, André, ‘Economic Conditions, Leader Evaluations and Election Outcomes in Canada’, Canadian Public Policy, 21 (1995), 212–18.CrossRefGoogle Scholar

⁶ Clarke, Harold D., Mishler, William and Whiteley, Paul, ‘Recapturing the Falklands: Models of Conservative Popularity, 1979–1983’, British Journal of Political Science, 20 (1990), 63–81, p. 68CrossRefGoogle Scholar; Bean and Mughan, ‘Leadership Effects in Parliamentary Elections’: the relationship between partisanship and perceptions of leader qualities ‘is nowhere near strong enough either overall or in individual cases to show perceptions of leaders' qualities to be artifacts of partisanship, or vice versa’, p. 1169; Stewart, Marianne C. and Clarke, Harold D., ‘The (Un)Importance of Party Leaders: Leader Images and Party Choice in the 1987 British Election’, Journal of Politics, 54 (1992), 447–70CrossRefGoogle Scholar, at p. 467: ‘multivariate analyses demonstrate that leader images, net of pre-campaign party identification, had large effects and outperformed economic evaluations and issue concerns for Labour and Conservative voting’; Clarke and Whiteley, ‘Perceptions of Macroeconomic Performance, Government Support and Conservative Party Strategy’, p. 111: ‘changes in the leader performance evaluation cause, but are not themselves caused by changes in Conservative [party] popularity’; IvorCrewe and Anthony King, ‘Did Major Win? Did Kinnock Lose? Leadership Effects in the 1992 British General Election’ (paper presented at the Elections, Public Opinion, and Parties (EPOP) Annual Conference, Lancaster, 1993), p. 17: ‘the ranking of leaders in 1992 adds…to our ability to predict whether or not someone would vote Conservative (as distinct from any other party) in 1992 once their vote and their partisan preferences in 1987 have been taken into account’; Lanoue and Headrick, ‘Prime Ministers, Parties and the Public’, p. 197 (after having run a Granger causality test for the 1953–87 period): ‘PM popularity is the independent variable and government lead is the dependent variable’.

⁷ See also Crewe, Ivor and King, Anthony, ‘Are British Elections Becoming More Presidential?’, in Jennings, M. Kent, ed.. Elections at Home and Abroad: Essays in Honor of Warren E. Miller (Ann Arbor: University of Michigan Press, 1994).Google Scholar

⁸ Key, V. O. Jr, The Responsible Electorate (Cambridge, Mass.: Harvard University Press, 1966).CrossRefGoogle Scholar

⁹ Downs, Anthony, An Economic Theory of Democracy (New York: Harper and Row, 1957).Google Scholar

¹⁰ Sanders, David, ‘Government Popularity and the Next General Election’, Political Quarterly, 62 (1991), 235–61CrossRefGoogle Scholar; Sanders, , ‘Why the Conservatives Won – Again’, in King, Anthony et al. , eds., Britain at the Polls 1992 (Chatham, NJ: Chatham House, 1993)Google Scholar; MacKuen et al. ‘Peasants or Bankers?’

¹¹ Miller, Arthur H. and Wattenberg, Martin P., ‘Throwing the Rascals Out: Policy and Performance Evaluations of Presidential Candidates, 1952–1980’, American Political Science Review, 79 (1985), 359–72.CrossRefGoogle Scholar

¹² Mishler et al., ‘British Parties in the Balance’, p. 229; see also Hudson, ‘Prime Ministerial Popularity in the UK’, p. 96.

¹³ Mishler et al., ‘British Parties in the Balance’.

¹⁴ This might again stimulate the question of whether the leadership and vote intention questions are essentially measuring the same thing. We think not, inasmuch as the leadership questions explicitly refer to the candidates – and probably call to mind personal characteristics as much as anything – while the vote decision also includes judgements about partisanship and issues. In addition to material cited in fn. 6, further evidence that voters distinguish between the government and the prime minister is found in a Gallup poll in March 1992, in which the approval rates were 30 per cent for the government and nearly 50 per cent for Major. The same poll shows that voters also distinguish between a leader's personality and policies. Whereas 63 percent of the voters said they liked Major's personality, only 41 per cent said they liked his policies.

¹⁵ ‘Are you satisfied with [name] as Prime Minister?’ ‘Do you think [name] is or is not proving to be a good leader of the Labour Party (Liberal Democrats, Social Democratic Party)?’ The questions about other leaders are not as purely Keysian as is that for the prime minister. They are still Keysian, however, because they lack any comparative basis and do not have a forward-looking orientation. That approval or satisfaction questions (especially for incumbents) are basically retrospective while questions asking how well candidates (especially challengers) would do are mainly prospective has been demonstrated by Miller and Wattenberg, ‘Throwing the Rascals Out’, p. 361. They also demonstrate that candidate assessments are made on a comparative basis and that they involve both retrospective and prospective judgements. These findings buttress our confidence that approval and ‘best prime minister’ questions refer to different time frames and that the latter are more appropriate for the evaluation of alternative candidates.

¹⁶ The ‘best prime minister’ question has been asked almost every month since June 1984. Our sample thus starts with this observation. It ends with the last observation before the general election of 1992, thus covering ninety-four months. Eight cases are missing; one, November 1990, is also missing for the approval question and has been deleted. The other missing cases are: 1987: 8,9; 1988: 5–8, and 1989: 12. To cope with this problem, we performed every statistical analysis using only the complete cases (n = 86), with an estimate for the missing cases based on a linear interpolation. The results are nearly identical; we report only the latter. Data for both measures were taken from the Gallup Political Index.

¹⁷ On the desirability of specifying separate equations for Labour and Alliance, see Sanders, David and Price, Simon, ‘Party Support and Economic Perceptions in the UK, 1979–87: A Two-Level Approach’, in Broughton, David, Farrell, David, Denver, David and Railings, Colin, eds, British Elections and Ponies Yearbook, 1994 (London: Frank Cass, 1995), pp. 45–72.Google Scholar For findings with the government-lead variable, see Pissarides, C. A., ‘British Government Popularity and Economic Performance’, Economic Journal, 90 (1980), 569–81CrossRefGoogle Scholar; Nadeau et al. ‘Expectations and Preferences in British General Elections’; and Lanoue and Headrick, ‘Prime Ministers, Parties and the Public’. Our interest in having multiple equations comes in part from the fact that Keysian and Downsian indicators, though different, are strongly correlated. The bivariate correlations between our two indicators are 0.83, 0.58, 0.90, and 0.62, respectively, in the models associated with the four dependent variables. Those high correlations are not surprising; they parallel the links between retrospective and prospective perceptions of economic performance (MacKuen et al., ‘Peasants or Bankers?’, p. 599; Clarke and Stewart, ‘Economic Evaluations, Prime Ministerial Approval, and Governing Party Support’, p. 3). These strong links have not prevented scholars from demonstrating that prospective and retrospective judgements on the economy have different origins and distinct effects. See Sanders, David, Marsh, David and Ward, Hugh, ‘The Electoral Impact of Press Coverage of the British Economy, 1979–87’, British Journal of Political Science, 23 ( 1993), 175–210CrossRefGoogle Scholar; Sanders and Price, ‘Party Support and Economic Perceptions’, and MacKuen et al., ‘Peasants or Bankers?’

¹⁸ Among others, see Clarke and Stewart, ‘Economic Evaluations, Prime Ministerial Approval and the Governing Party’; Clarke and Whiteley, ‘Perceptions of Macroeconomic Performance’; Clarke et al. ‘Recapturing the Falklands’; Mishler et al., ‘British Parties in the Balance’; Nadeau et al., ‘Expectations and Preferences in British General Elections’; Helmut Norpoth, ‘The Popularity of the Thatcher Government: A Matter of War and Economy’, in Norpoth, et al. , Economics and Politics, and Confidence Regained: Economics, Mrs. Thatcher, and the British Voter (Ann Arbor: University of Michigan Press, 1992)CrossRefGoogle Scholar; Sanders, ‘Government Popularity and the Next General Election’, and ‘Why the Conservatives Won’; Sanders and Price, ‘Party Support and Economic Perceptions’; Sanders et al., ‘Governmental Popularity and the Falklands War’; Sanders et al., ‘The Electoral Impact of Press Coverage’.

¹⁹ Miller, W. L. and Mackie, W. M., ‘The Electoral Cycle and the Asymmetry of Government and Opposition Popularity’, Political Studies, 21 (1973), 263–79; Sanders and Price, ‘Party Support and Economic Perceptions’.CrossRefGoogle Scholar

²⁰ In the period under study, the percentage intending to vote Conservative, Labour, or Liberal Democratic corresponds, respectively, to the percentage intending to vote for the government, the official opposition, or a centre party. Similarly, the percentage selecting or approving of the Conservative, Labour, or Liberal Democratic leader(s) for the BESTPM or APPLEADERS variables corresponds, respectively, to the percentage selecting or approving the current prime minister, the leader of the official opposition, or the leader(s) of a centre party.

²¹ Because it slightly improved the performance of these equations, a second lag of the dependent variable is also included in the Labour and the Liberal Democrats cases (Equations 2 and 4).

²² Beck, ‘The Economy and Presidential Approval’; Norpoth, ‘The Popularity of the Thatcher Government’.

²³ Anderson, Christopher, ‘Party Systems, Governing Parties, and the Dynamics of Mass Support in Western Democracies: Britain and Germany, 1960–1990’ (paper presented at the Annual Meeting of the American Political Science Association, Washington, DC, 1993)Google Scholar; see also MacKuen et al., ‘Peasants or Bankers’.

²⁴ Other lag structures did not improve the performance of the model. This model is based on the conventional assumption that voters respond to measured data, which means that even if these data contain measurement errors, the standard least squares techniques are valid (see Johnston, J., Econometric Methods (London: McGraw-Hill, 1984), p. 430).Google Scholar

²⁵ ‘How do you think the financial situation of your household will change in the next twelve months? (Get a lot better, get a little better, stay the same, get a little worse, get a lot worse)’. Following conventional practices, we operationalized this variable as the percentage difference between the first two categories (get a lot or a little better) and the fourth and fifth categories (get a little or a lot worse). Sanders et al., ‘Governmental Popularity and the Falklands War’, and ‘The Electoral Impact of Press Coverage of the British Economy’; Sanders, ‘Government Popularity and the Next General Election’, and ‘Why the Conservatives Won’. The data are from the Monthly Gallup Polls.

²⁶ Anderson, ‘Party Systems, Governing Parties, and the Dynamics of Mass Support’, p. 10. On the impact of unemployment and inflation on support for parties in Britain, see Pissarides, ‘British Government Popularity and Economic Performance’; Mishler et al., ‘British Parties in the Balance’; Clarke et al., ‘Recapturing the Falklands’; Clarke and Whiteley, ‘Perceptions of Macroeconomic Performance’; Price, Simon and Sanders, David, ‘Modeling Government Popularity in Postwar Britain: A Methodological Example’, American Journal of Political Science, 38 (1993), 317–34.CrossRefGoogle Scholar Noting that some authors recently specified popularity functions for the United Kingdom excluding the objective economic indicators (see Sanders, ‘Why the Conservatives Won’; Clarke and Stewart, ‘Economic Evaluations, Prime Ministerial Approval, and Governing Party Support’), we also estimated the various equations described below excluding the unemployment and inflation rates. None of the substantive conclusions in this article about the leaders variables were altered by these exclusions. The same holds true when unemployment is measured in terms of change. Data on unemployment and inflation were taken from the OECD Main Economic Indicators: Historical Statistics (various years).

²⁷ Sanders et al., ‘Governmental Popularity and the Falklands War’, and ‘Macroeconomics, the Falklands War’, and Sanders, ‘The Electoral Impact of Press Coverage of the British Economy’; Clarke et al., ‘Recapturing the Falklands’; Sanders, ‘Governmental Popularity and the Next General Election’, and ‘Why the Conservatives Won’; Nadeau et al., ‘Expectations and Preferences in British General Elections’. We also considered ‘rival models’ based on other subjective perceptions of the economy (general retrospections, personal retrospections, general expectations; see Clarke and Stewart, ‘Economic Evaluation, Prime Ministerial Approval, and Governing Party Support’). None of the substantive conclusions in this article about the leaders variables were altered in the context of these alternative models.

²⁸ Sanders, ‘Why the Conservatives Won’. The problems in estimating a model of party support for the centre parties is complex given the special characteristics of the ‘alliance’ among them during the period under study. To cope with these problems, we proceeded in the following fashion: (1) support for the centre parties was established by adding the percentage intending to vote for any party in the ‘alliance’ (Liberals, Social Democrats, Alliance; Greens and other parties were thus excluded), (2) the approval and ‘best prime minister’ variables were constructed by averaging the percentage approving or selecting any of the leaders belonging to the alliance.

²⁹ A poll tax variable was not significant, perhaps because we have included leader variables and the tax was largely seen as Thatcher's tax. Clarke and Stewart, ‘Economic Evaluations, Prime Ministerial Approval and Governing Party Support’, Table 6, found a poll tax variable to be only marginally significant.

³⁰ A number of studies evaluating aggregate party support have raised questions about non-stationarity in the data, which imposes certain constraints in die specification of the empirical model (see Hyelleberg, S. and Mizon, G. E., ‘Cointegration and Error Correction Mechanisms’, Economic Journal, 99 (1989), 113–25CrossRefGoogle Scholar; see also Clarke and Stewart, ‘Economic Evaluations, Prime Ministerial Approval, and Governing Party Support’, p. 154). There are two central issues. First, could we be estimating a ‘spurious regression’? Secondly, what is the appropriate distribution theory for inferential purposes? With respect to the first problem, our regression results were not plagued by problems associated with ‘spurious regression’, the salient features of which are (relatively) high R² statistics, low Durbin–Watson statistics and significant t and F ratios. With respect to the second, we believe that variables that are specified as rates (i.e., unemployment, inflation, popular support) belong in a model in this form as they are intrinsically stationary variables (see Cochrane, John, ‘Comment’, in Blanchard, Oliver and Fischer, Stanley, eds, NBER Macroeconomics Annual (Cambridge, Mass.: MIT Press, 1991)).Google Scholar That is, opinion, unemployment and inflation do not behave as if they have no fixed mean, as do financial series; nor do they have a strong secular growth component as do GNP and price levels. In this context, least squares regression on a level specification is perfectly acceptable and standard econometric theory applies. We thus stick to a standard model specified in levels that shows no evidence of being plagued by any of the standard problems associated with nonstationary data. Furthermore, most econometricians would argue that the ability to ascertain whether a particular data series possesses a ‘unit root’ or I(l) driving process from a short series observed over eight years is problematic since such a series will miss all of the ‘mean reverting’ behaviour in the data. A more detailed discussion of these issues in the context of our model is available from the authors.

³¹ We used here the following five tests: (1) the Ramsay's RESET test, which in this particular case tests for a quadratic component in the conditional mean; (2) the Lagrange Multiplier tests for serially-correlated errors, which here performed a joint test for first- through fourth-order dependence; (3) the ‘White’ test utilizing the squared residuals and linear and quadratic terms of the regression to test for a general form of heteroscedasticity; (4) the ARCH test for conditional heteroscedasticity; and (5) a test for the normality of the residuals, which is a joint test of the skewness and kurtosis coefficients. Apart from two minor (and inconsequential) departures – the ARCH test in the Labour equation and the RESET test in the Liberal Democrat one (significant at 0.10 but not at 0.05) – the specifications used here handily passed these tests. We can thus conclude that, in terms of the standard battery of diagnostics generally applied to single-equation regressions using time-series data, the models used seem well-specified. (On diagnostic testing procedures to check for the general adequacy of model specification, see Spanos, Aris, Statistical Foundations of Econometric Modelling (Cambridge: Cambridge University Press, 1966)Google Scholar and Granato, Jim, ‘An Agenda for Econometric Model Building’, Political Analysis, 3 (1991), 123–54.)CrossRefGoogle Scholar Also, as the dependent variables are bounded, we re-estimated the various equations using a logistic procedure, as in Price and Sanders, ‘Modeling Government Popularity in Postwar Britain’. (One of the variables, which had negative values, had to be rescaled first.) This made very little difference to the observed results.

³² In the first stage we regressed ‘best prime minister’ on a series of ‘instruments’ at least one of which was excluded from the second-stage regression. The residuals from this regression were then included as an additional explanatory variable in the popularity model. A t-test on the residuals then serves as an exogeneity test. If the t ratio on this coefficient falls below the appropriate critical value, then we may regard the preselected variable as exogenous. In all regressions the coefficient on the residuals fell well below the critical value from the t-distribution. On the exogeneity of the approval variable to party support, see Clarke and Stewart, ‘Economic Evaluations, Prime Ministerial Approval and Governing Party Support’, p. 163.

³³ The coefficient and the t statistic of the approval rate variable in a regression excluding the ‘best prime minister’ variable are: 0.17 (0.05). The adjusted R² is 0.81.

³⁴ The coefficients and t statistics for approval rates in separate regressions were, respectively: 0.26 (0.05), 0.28 (0.06) and 0.05 (0.03) for the three dependent variables in Table 2. The adjusted R² for these regressions are: 0.82, 0.80 and 0.87.

³⁵ A leader's past performance is important, but as an indirect influence in that it constitutes a piece of information, among other pieces, that helps voters to assess potential leaders' future performances. For example, ‘Kramer…assumes that voters make use of retrospective economic evaluations in predicting the future behavior of political parties and candidates, much as Downs…suggests’ (Lockerbie, Brad, ‘Changes in Party Identification: The Role of Prospective Economic Evaluations’, American Politics Quarterly, 17 (1989), p. 293).CrossRefGoogle Scholar

³⁶ Ansolabehere, Stephen, Behr, Roy and lyengar, Shanto, The Media Game (New York: Macmillan, 1993), p. 57.Google Scholar

³⁷ Negrine, Ralph, Politics and the Mass Media in Britain (London: Routledge, 1989).CrossRefGoogle Scholar

³⁸ The finding that voters compare leaders even if they are not explicitly invited to do so gives further support to the notion that voters think in comparative terms. As Crewe and King, ‘Did Major Win? Did Kinnock Lose?’, p. 6, put it: ‘Although the BES questions did not ask respondents to choose between the leaders, it seems clear that Kinnock benefited personally from the comparison with Thatcher in 1987 and suffered personally from the subsequent comparison with Major.’

³⁹ MacKuen et al., ‘Peasants or Bankers?’

⁴⁰ Campbell, Angus, Gurin, Gerald and Miller, Warren E., The Voter Decides (Evanston, Ill: Row, Peterson, 1954).Google Scholar

⁴¹ MacKuen et al., ‘Peasants or Bankers?’ pp. 606–7.