¹ Crewe, I., ‘How to Win a Landslide Without Really Trying: Why the Conservatives Won in 1983’, p. 159Google Scholar, in Ranney, A., ed., Britain at the Polls 1983 (New York: Duke University Press, 1985), pp. 155–96.Google Scholar

² The literature is voluminous and covers many countries. For a review see: Paldam, M., ‘A Preliminary Survey of the Theories and Findings of Vote and Popularity Functions’, European Journal of Political Science, IX (1981), 181–99.CrossRefGoogle Scholar For more recent references see Whiteley, P., ‘Inflation, Unemployment and Government Popularity’, Electoral Studies, 111 (1984), 25–46.Google Scholar Particularly important work on Britain includes: Goodhart, C. A. E. and Bhansali, R. J., ‘Political Economy’, Political Studies, XVIII (1970), 43–106CrossRefGoogle Scholar; Frey, B. S. and Schneider, F., ‘A Politico-Economic Model of the UK’, Economic Journal, LXXXVIII (1978), 243–53CrossRefGoogle Scholar; Mosley, P., ‘Images of the “Floating Voter”: or the Political Business Cycle Revisited’, Political Studies, XXVI (1978), 375–94CrossRefGoogle Scholar; Alt, J., The Politics of Economic Decline (Cambridge: Cambridge University Press, 1979)Google Scholar; Hibbs, D., ‘Economic Outcomes and Political Support for British Governments Among Occupational Classes’, American Political Science Review, LXXVI (1982), 259–79CrossRefGoogle Scholar; Mosley, P., The Making of Economic Policy: Theory and Evidence from Britain and the United States Since 1945 (Brighton: Wheatsheaf Books, 1984), Chapter 2.Google Scholar

³ Dunleavy, P. and Husbands, C. T., British Democracy at the Crossroads: Voting and Party Competition in the 1980s (London: George Allen & Unwin, 1985).Google Scholar

⁴ Clarke, H. D., Stewart, M. C. and Zuk, G., ‘Politics, Economics and Party Popularity in Britain, 1979–83’, Electoral Studies, V (1986), 123–41.CrossRefGoogle Scholar

⁵ Norpoth, H., ‘War and Government Popularity in Britain’, paper given at the annual meeting of the American Political Science Association, Washington, DC, 08 1986.Google Scholar

⁶ Dunleavy, and Husbands, , British Democracy at the Crossroads, p. 153.Google Scholar

⁷ Dunleavy, and Husbands, , British Democracy at the Crossroads, p. 154.Google Scholar

⁸ Clark, , Stewart, and Zuk, , ‘Politics, Economics and Party Popularity’, pp. 123–30.Google Scholar

⁹ It should be noted that in contrast to the other studies referred to above, Norpoth follows the North American convention of modelling support for the leader of the Government rather than for the Government as a whole. This means that his dependent variable is slightly different from that analysed in other studies, but since Mrs Thatcher's personal rating and that of the Government as a whole are highly correlated (r = 0.91), his findings are of considerable relevance to our discussion.

¹⁰ There are a number of minor technical anomalies evident within each of the studies although we do not consider them to be serious. Specifically: (1) Dunleavy and Husbands report an R ² of 0.87 for their predictive model of Government popularity. Using the same predictor variables – percentage unemployed lagged four months and a Falklands dummy variable – over a slightly larger time period, we obtained an R ² of only 0.78; although the coefficient of b = 15.9 for the Falklands dummy is almost identical to that reported by Dunleavy and Husbands. (2) Dunleavy and Husbands's analysis dates the start of the Falklands dummy from May 1982, even though the conflict dates from 31 March 1982. The fact that government popularity did not experience an upsurge until May 1981 thus gives their Falklands dummy its maximum opportunity to figure significantly in their model of government popularity. If the model is re-estimated, dating the Falklands dummy from April 1982 instead of May, the explanatory statistical power of the Falklands factor is reduced. According to our estimates, the R ² for the Falklands-crisis-began-in-May model is 0.78; for the Falklands-crisisbegan-in-April model, R ² falls to 0.70. (3) Even within Dunleavy and Husbands' own terms of reference, the combination of (a) their possible mis-specification in the timing of the start of the Falklands crisis, together with (b) a small amount of serial correlation in the error term of their two-predictor model is sufficient to produce a mild inflation in their estimate of the magnitude of the Falklands effect. The Falklands-crisis-began-in-May model, estimated with OLS, yields a Falklands factor coefficient of b = 15.9; a Falklands-crisis-began-in-April model estimate using the Cochrane-Orcutt pseudo-GLS technique yields b = 14.4. (4) Clarke et al. include ‘satisfaction’ with Mrs Thatcher as an independent variable in their equation for Government popularity. This is a variable which on theoretical grounds we believe should be omitted from the model. We interpret the high bivariate correlation between satisfaction with the Prime Minister and government popularity as evidence that both variables are a reflection of respondents' general predispositions towards the government. To include satisfaction as a predictor of popularity not only ignores the possibility that popularity influences satisfaction but also appears to be at least partially tautological. (5) Norpoth's estimation technique attempts to explain changes in Mrs Thatcher's poll ratings in terms of deviations about the mean. He justifies this strategy by arguing that party identification explains this average baseload of approval and that what needs to be explained are deviations from it. This argument seems to us inadequate for two reasons. Firstly, all the available evidence (see, for example, Crewe, Ivor and Särlvik, Bo, Decade of Dealignment (Cambridge: Cambridge University Press, 1983))Google Scholar suggests that party identification has been declining in recent years, which presumably means that its effects should be assumed to decline over time rather than to remain constant as Norpoth's chosen strategy implies. Secondly, one of the causes of partisan dealignment might be what is perceived as poor economic performance: if so, then part of the variation in government popularity which might on a priori grounds be attributable to economic explanations is simply assumed by Norpoth to be unrelated to such factors.

¹¹ Fiorina, M., ‘Short and Long-Term Effects of Economic Conditions on Individual Voting Decisions’, in Hibbs, D. and Fassbender, H., eds, Contemporary Political Economy (Amsterdam: North-Holland, 1981), pp. 73–100Google Scholar, estimates a model in which both retrospective and prospective evaluations of government performance figure significantly.

¹² It can, of course, be argued that the differential effects of macroeconomic changes on different social groups may lead to an underestimation of the true significance of those changes when only aggregate data are examined. Hibbs, for example, in ‘Economic Outcomes and Political Support’, has shown that whereas inflation adversely affects the interests of middle-class voters, it need not have such deleterious consequences for the working class. In these circumstances, the effect on aggregate government popularity of an increase in inflation may not appear to be very large because its effects on these two class groups tend to cancel each other out. Again, in the absence of suitable data, we are unable to test for this kind of differential effect.

¹³ For attempts to model expectations from past performance see, for example, Fiorina, , ‘Short and Long-Term Effects’, pp. 93–5Google Scholar; Hibbs, D. and Vasilatos, N., ‘Macroeconomic Performance and Mass Political Support in the United States and Great Britain’Google Scholar, in Hibbs, and Fassbender, , Contemporary Political Economy, pp. 31–47, especially pp. 37–8Google Scholar; for a general discussion of the use of various expectations models in political science and the difficulties of their estimation see Rattinger, H., ‘Econometrics and Arms Races: A Critical Review and Some Extensions’, European Journal of Political Research, IV (1974), 421–39.Google Scholar

¹⁴ Specifically, this is operationalized as the monthly percentage of Gallup poll respondents whose stated intention was to vote Conservative. The source is the monthly Gallup Political Index ‘A’ series (which includes ‘Don't knows’ in the percentage base). The figure of 44 per cent for May 1983 is taken from the poll conducted on 14 May: this date was chosen (from the four polls conducted in May 1983) because it was closest to one month after the date of the April 1983 poll. The figure of 42 per cent for June 1983 was taken from the Gallup poll conducted on 3 June.

¹⁵ The intercorrelations (N = 48) were as follows:

¹⁶ With certain qualifications a similar inference can be drawn in relation to the exchange rate variable: this is discussed on pp. 297–8.

¹⁷ The literature on the mid-term effect is critically reviewed in Hudson, J., ‘The Relationship Between Government Popularity and Approval for the Government's Record in the United Kingdom’, British Journal of Political Science, XV (1985), 165–86.CrossRefGoogle Scholar

¹⁸ The timing of these shifts, of course, is not discernible directly from the graphical representations shown on Table 2. The relevant figures for the transitions from decline to recovery are as follows:*

¹⁹ For a discussion of these criteria, see Sanders, David, Patterns of Political Instability (London: Macmillan, 1981)CrossRefGoogle Scholar, Chap. 8. By ‘robust’ we mean that the parameters of a given model remain stable when each predictor variable in turn is omitted from it. We use significance tests not because we wish to generalize from a random sample to a wider population but because econometric practice has demonstrated that the t-test provides a very useful decision rule for determining whether or not a particular predictor makes a worthwhile contribution to a given model.

²⁰ The R ² of 0.87 reported for this model appears no better than the 0.87 reported by Dunleavy and Husbands for their admittedly more parsimonious ‘unemployment and Falklands factor’ model. As we pointed out earlier, however, on our estimates the unemployment and Falklands factor model only explains 70 per cent of the variance in government popularity (see fn. 10).

It is also worth recording that the model reported in Figure 4 is remarkably robust in terms of its application to different segments of the total time period analysed. Specifically: (a) the residuals from the equation are randomly distributed over time: a correlation between the residuals and time yields r = 0.01; (b) if the series is partitioned into a ‘government popularity downswing’ segment (up to December 1981, when government popularity reached its trough of 18.5 per cent) and an ‘upswing segment’ (from January 1982 to June 1983) the following results are obtained:

²¹ See Campbell, D. T. and Fiske, D. W., ‘Convergent and Discriminant Validation by the Multitrait-Multimethod Matrix’, Psychological Bulletin, LVI (1959), 81–105.CrossRefGoogle Scholar

²² For the mediating effect of newspaper coverage of macroeconomic variables upon voting see Moseley, P., ‘Popularity Functions and the Media: A Pilot Study of the Popular Press’, British Journal of Political Science, XIV (1984), 119–31.Google Scholar For a more general discussion of the media's impact on perceptions see Mosley, , The Making of Economic Policy, Chap. 2.Google Scholar

²³ For a discussion of coverage of the Falklands war see Glasgow University Media Group, War and Peace News (Milton Keynes: Open University Press, 1985), section 1.Google Scholar

²⁴ As noted earlier, government popularity jumped by 11 percentage points between April and May 1982. Protagonists of the ‘Falklands effect’ often fail to mention the fact that there was also an increase of 9 percentage points in April-May 1983 (which we interpret as a simple ‘campaign effect’).

²⁵ Again, the effects of all predictor variables were considered unlagged and lagged at t – 1, t – 2, …, t – 12.

²⁶ Certainly the predicted values from the model Personal Expectations = a + b ₁ consumer spending + b ₂ short-time working + b ₃ interest rates do not exhibit a discontinuity of any sort.

²⁷ This characterization possibly exaggerates Sir Geoffrey's wetness but The Economist (13 March 1982) did describe the 1982 Budget as ‘a slightly wet turn’ (p. 12).

²⁸ There are, moreover, some indications that the 1982 Budget was relatively well received by the general public. A Gallup poll published in the Daily Telegraph of 18 March, for example, reported that 56 per cent of respondents thought the Budget ‘fair’, a remarkably high figure in the circumstances. When the Gallup poll responses to Sir Geoffrey's 1982 budget are considered in comparative context (see Gallup Political Index, 03 1982, pp. 5 and 24Google Scholar) it is clear that in terms of the public's view of the Chancellor's ‘competence’ and ‘fairness’, 1982 was not unusual. In March 1982, 49 per cent of respondents thought Sir Geoffrey was doing a ‘good job’ as Chancellor (compared with an average of 50 per cent for the period 1949–81) and 56 per cent of respondents thought that the Budget was ‘a fair one’ (compared with 52 per cent for 1949–81). However, the comparison which really needs to be made is between the responses to Sir Geoffrey's 1982 budget and the response to his much harsher budget of 1981. In 1981, only 24 per cent of respondents thought Sir Geoffrey was doing ‘a good job’ (1982:49 per cent); in 1981 only 22 per cent thought the Budget ‘fair’ (1982: 56 per cent): and in 1981, fully 61 per cent of those interviewed thought the Budget ‘too tough’ (1982: 27 per cent). While it would be misleading to read too much into the precise magnitudes of these differential responses, the broad shift which they collectively portray certainly lends considerable credence to the view that the 1982 Budget provided a discontinuous shock to expectations.

²⁹ This interpretation is consistent with some individual-level survey data. Heath, Jowell and Curtice found that few of their respondents in a poll held shortly after the 1983 election believed that the Falklands war had been important to their vote. Only 11 per cent said it had been very import ant to them when deciding about voting. While noting the possibility that the Falklands war may have had an unconscious impact, Heath, Jowell and Curtice conclude that the ‘Falklands factor’ was of limited strength by 1983. See Heath, A., Jowell, R. and Curtice, J., How Britain Votes (Oxford: Pergamon Press, 1985), p. 162.Google Scholar

³⁰ If we add an ‘abrupt-temporary Falklands effect’ dummy variable (which takes on the value 1 for May-July 1982 and zero otherwise) to the ‘best equation’ for government popularity shown in Figure 4, then we obtain a coefficient for the Falklands effect of b = 3.40 (t = 2.41) which adds just over 1 per cent to R ². This model indicates that the Falklands affair exerted a small ‘abrupt-temporary’ effect on government popularity of just over 3 per cent for a period of no more than three months.