Recently, Sanders et al. have made the intriguing and counter-intuitive argument that the impact of the Falklands war on Conservative popularity was inconsequential. Their analyses raise important theoretical and methodological issues concerning the time-series analysis of party support. This present article contends that the stepwise regression procedures employed by Sanders et al. are misleading, particularly when predictor variables are highly intercorrelated. Box-Jenkins analyses demonstrate that the Falklands strongly influenced Conservative support, net of the effects of macroeconomic conditions and personal economic expectations. The significance of the latter variable in the models confirms Sanders et al.'s argument about the role of subjective economic variables in party popularity functions. Non-economic variables are also relevant, however, and popularity functions that model them correctly will enhance our understanding of both the economics and the politics of party support.
1 See Sanders, David, Ward, Hugh and Marsh, David, ‘Government Popularity and the Falklands War’, British Journal of Political Science, 17 (1987), 281–313.
2 See Crewe, Ivor, ‘How to Win a Landslide Without Really Trying: Why the Conservatives Won in 1983’, in Ranney, Austin, ed., Britain at the Polls 1983 (Durham, NC: Duke University Press, 1985), 155–96; Clarke, Harold D., Stewart, Marianne C. and Zuk, Gary, ‘Politics, Economics and Party Popularity in Britain, 1979–83’, Electoral Studies, 6 (1987), 3–16; Norpoth, Helmut, ‘Guns and Butter and Government Popularity in Britain’, American Political Science Review, 81 (1987), 949–59; Norpoth, Helmut, ‘The Falklands War and Government Popularity in Britain: Rally without Consequences or Surge without Decline’, Electoral Studies, 6 (1987), 3–16; Mishler, William, Hoskin, Marilyn and Fitzgerald, Roy, ‘Hunting the Shark; or Searching for Evidence of the Widely Touted But Highly Elusive Resurgence of Public Support for Conservative Parties in Britain, Canada, and the United States’, in Cooper, Barry, Kornberg, Allan and Mishler, William, eds, The Resurgence of Conservatism in Anglo-American Democracies (Durham, NC: Duke University Press, 1988).
3 Sanders, et al. , ‘Government Popularity’, p. 312.
4 That is, twenty economic variables times thirteen different lags plus four expectation/perception variables and two event variables.
5 Cohen, Jacob and Cohen, Patricia, Applied Multiple Regression Correlation Analysis for the Behavioral Sciences (Hillsdale, NY: Lawrence Erlbaum Associates, 1975), p. 103.
6 See Whiteley, Paul F., ‘Inflation, Unemployment and Government Popularity – Dynamic Models for the United States, Britain and West Germany’, Electoral Studies, 3 (1984), 3–24; Clarke, et al. , ‘Politics, Economics and Party Popularity in Britain’; Mishler, William, Hoskin, Marilyn and Fitzgerald, Roy, ‘British Parties in the Balance: A Time-Series Analysis of Long-Term Trends in Labour and Conservative Support’, British Journal of Political Science, 19 (1989), 211–36.
7 According to the traditional ‘Michigan’ conception, party identification constitutes a stable attitudinal predisposition for most voters and therefore does not manifest significant short-term aggregate change. See Campbell, Angus et al. , The American Voter (New York: Wiley, 1960), p. 127. In the Box-Jenkins models estimated below, government popularity and the economic variables are differenced, so the fact that we are measuring short-term changes in these variables makes it legitimate to exclude party identification from these models. See Whiteley, , ‘Inflation and Unemployment’, pp. 10–11. Similarly, alternative ‘rational choice’ conceptions of party identification such as that proposed by Fiorina imply that voters' partisan predispositions may be omitted from models of party support if party performance evaluations are included in such models. See Fiorina, Morris, Retrospective Voting in American National Elections (New Haven, Conn.: Yale University Press, 1981).
8 See Mosley, Paul, ‘“Popularity Functions” and the Role of the Media: A Pilot Study of the Popular Press’, British Journal of Political Science, 14 (1984), 117–28; Särlvik, Bo and Crewe, Ivor, Decade of Dealignment (Cambridge: Cambridge University Press, 1983), pp. 150–9.
9 See Jowell, Roger and Airey, Colin, British Social Altitudes: The 1984 Report (Aldershot, Hants: Gower, 1985), p. 174.
10 Whiteley, , ‘Inflation and Unemployment’, pp. 3–24; Fitzgerald, Roy and Mishler, William, ‘Through A Glass Darkly: The Dynamics of Public Economic Perceptions in Britain, 1978–87’ (paper presented at the Annual Meeting of the Southern Political Science Association, Atlanta, Georgia, 11 1988).
11 Norpoth, , ‘Guns and Butter’, p. 956.
12 See Hastings, Max and Jenkins, Simon, The Battle for the Falklands (New York: W. W. Norton, 1983), p. 136.
13 Sanders, et al. , ‘Government Popularity’, pp. 283–4, notes 9 and 10.
14 Micro-level analyses show that evaluations of party leader performance in Britain have independent effects on party support. See, e.g., Clarke, et al. , ‘Politics, Economics and Party Popularity’, pp. 133–5. The estimated effects of the Falklands war on evaluations of the Prime Minister's popularity appear in Norpoth, , ‘Guns and Butter’, p. 954; and Norpoth, , ‘The Falklands War’, p. 11.
15 See Norpoth, Helmut, ‘Guns and Butter’, p. 954; and ‘The Falklands War’, p. 11.
16 See, for example, Kinder, Donald R. and Kiewiet, D. Roderick, ‘Sociotropic Politics: The American Case’, British Journal of Political Science, 11 (1981), 129–61; Kiewiet, D. Roderick, Macroeconomics and Micropolitics (Chicago: Chicago University Press, 1983); Monroe, Kristen, Presidential Popularity and the Economy (New York: Praeger, 1984), Chap. 1.
17 Jenkins, Gwilym, Practical Experiences with Modelling and Forecasting Time Series (St. Helier, Jersey: G.J.P. Publications, 1979), p. 191.
18 See Klein, Lawrence R., An Introduction to Econometrics (Englewood Cliffs, NJ: Prentice-Hall, 1962), p. 101.
19 See Maddala, G. S., Econometrics (New York: McGraw-Hill, 1979), pp. 183–90.
20 Clarke, Stewart and Zuk, for example, report that the Falklands war (modelled as an abruptpermanent effect) increased Conservative popularity by approximately 7 points net of controls for inflation, unemployment and several other predictor variables; see Clarke, et al. , ‘Politics, Economics and Party Popularity’, pp. 131–3. Dunleavy and Husbands estimate an abrupt-permanent effect of nearly 17 points; see Dunleavy, Patrick and Husbands, Christopher, British Democracy at the Crossroads (London: Allen and Unwin, 1985), p. 153.
21 Ostrom, Charles W. Jr, Time Series Analysis: Regression Techniques, Sage University Paper Series on Quantitative Applications in the Social Sciences (Beverly Hills, Calif.: Sage Publications, 1978), pp. 25–31.
22 See Box, George E. and Jenkins, Gwilym M., Time Series Analysis: Forecasting and Control, revd. edn. (San Francisco: Holden-Day, 1976); Box, George E. and Tiao, Gregory C., ‘Intervention Analysis with Applications to Economic and Environmental Problems’, Journal of the American Statistical Association, 70 (1975), 70–9. For an introduction to these methods see McCleary, Richard and Hay, Richard A., Applied Time Series Analysis (Beverly Hills, Calif.: Sage, 1980).
23 The main diagnostic test is the Ljung-Box Q statistic, which measures whether or not there is any significant information in the model's residual autocorrelation function over a specified period of lags. See Ljung, Greta and Box, George E. P., ‘On a Measure of Lack of Fit in Time Series Models’, Biometrika, 64 (1978), 297–304.
24 The economic variables are the seasonally adjusted percentage unemployed, corrected for government changes in the series to ensure validity over time, and the retail price index, seasonally adjusted. Personal and general economic expectations are measured the same way as in SWM, i.e. the percentage thinking economic conditions will improve minus the percentage thinking they will deteriorate. However, the data are not exactly the same as those employed by SWM. SWM's economic expectations series was constructed by splicing together responses from two different Gallup series; for the sake of consistency we use one series for the entire period. The two series are very highly correlated (r = 0.93), and show the same trends of increasing voter optimism in the spring of 1982.
25 Preliminary diagnostics indicated that the three economic variables needed to be logged and differenced to ensure variance and mean stationarity. Cross-correlation functions for the economic series showed that they did not suffer from multicollinearity. At the lags at which the variables were employed in the models the correlation (r) between inflation and unemployment was −0.14, whereas those for inflation and personal expectations, and unemployment and personal expectations were 0.05 and −0.01 respectively. Relatedly, the correlations of the estimates for the effects of the economic variables are quite modest (maximum values = 0.33).
26 The calculation is:
See McCleary, and Hay, , Applied Time Series Analysis, p. 174.
27 The assumption that macroeconomic variables influence political support according to a reward-punishment process is virtually ubiquitous in the popularity function literature. See e.g., Monroe, Presidential Popularity, Chaps 1 and 2. The possibility of alternative issue-priority effects in the British case is discussed in Clarke, et al. , ‘Politics, Economics and Party Popularity’, pp. 127–30 and in Mishler, and Fitzgerald, , ‘British Parties in the Balance’.
28 The calculation of the month-to-month effects of the gradual-permanent intervention model is as follows: first month = ω0 second month = ω0(1 + δ); third month = ω0(1 + δ + δ.δ), and so on. The long-term or asymptotic change is: ω0/(1 − δ). Again, because the dependent variable is measured in the log metric, the percentage change in the expected value of the series in a given month is computed by taking the anti-logs of the values produced by these calculations. The predicted percentage increase in the level of the series for a given month is the pre-intervention expected value (i.e., the mean Conservative popularity for the period July 1979 to April 1982 = 29.3 per cent) times the computed percentage change minus 29.3 per cent. See McCleary, and Hay, , Applied Time Series Analysis, p. 184.
29 The calculation of the month-to-month effects of the gradual-temporary intervention model is: δ(n − 1)ω0 + δ(n − 1)ω1, where n is the number of months since the interventions occurred. Conversions of the computed statistics into percentage changes in the expected value of the series and the predicted increase in the level of the series for a given month follows the procedure outlined in fn. 26 and fn. 28.
30 Norpoth, , ‘Guns and Butter’, p. 954. Similarly, analyses of satisfaction with the Prime Minister, Mrs Thatcher, which includes inflation, unemployment, personal expectations and the three Falklands models discussed above, all show that the war had strong effects. As in Norpoth's analyses, the adjustment parameter (δ) for the gradual-temporary model is very large (0.94), indicating that a substantial proportion of the war's impact on public evaluations of Mrs Thatcher's performance remained at the time of the 1983 election.
31 The war had a positive impact on personal (and general) economic perceptions with controls for inflation, unemployment, interest rates, real income, economic shocks of various kinds, election campaigns, and budget debates. See Fitzgerald, and Mishler, , ‘Through a Glass Darkly’.
* Department of Political Science, University of North Texas, Denton; Department of Government and International Studies, University of South Carolina, Columbia; Department of Political Science, University of Arizona, Tucson, respectively.
Email your librarian or administrator to recommend adding this journal to your organisation's collection.
* Views captured on Cambridge Core between <date>. This data will be updated every 24 hours.
Usage data cannot currently be displayed