¹ Mueller, John, “Presidential Popularity from Truman to Johnson,” American Political Science Review, 64 (March 1970), 18–34CrossRefGoogle Scholar.

² Mueller, 20; An Economic Theory of Democracy, ed. Downs, Anthony (New York: Harper and Row, 1957), pp. 55–60Google Scholar. Mueller also notes “minus-sum games” which are described by Wildavsky, Aaron in “The Empty-head Blues: Black Rebellion and White Reaction,” The Public Interest, 11 (Spring 1968), 3–16Google Scholar.

³ In 1956 when respondents were asked in the open-ended SRC interviews what they liked and disliked about Eisenhower, the ratio of personal evaluations to political judgments among positive responses was 2.74 to 1. This compares with ratios of 1.35 and 1.33 to 1 for Johnson in 1964 and Nixon in 1968, respectively. From Stimson, James A., “Public Support for American Presidents: A Cyclical Model,” mimeographed, 1974Google Scholar, Table 2.

⁴ Mueller, 28. Italics in original.

⁵ Hibbs, Douglas A., “Problems of Statistical Estimation and Causal Inference in Dynamic, Time-Series Regression Models,” paper prepared for the annual meeting of the American Political Science Association, Washington, D.C., 1972Google Scholar.

⁶ Stimson, 6, Figures 1, 2. Also see Stimson, James A. and LeGette, Caroline, “Public Support for American Presidents: Does Anything But Time Matter?” paper presented at the annual meeting of the American Political Science Association, Washington, D.C., 1975Google Scholar.

⁷ Stimson, 8.

⁸ Stimson, 11.

⁹ Stimson's final preference to follow Mueller's strategy of letting popularity for each term change at its own pace results in a table similar to Table 1 except the estimates are quadratic, rather than linear. See Stimson's Table 4.

¹⁰ For an interesting discussion on the value of time as a diagnostic variable, see Rao, Potluri and Miller, Roger LeRoy, Applied Econometrics (Belmont, Calif.: Wadswoith, 1971), pp. 99–104Google Scholar.

¹¹ Richard Brody and Benjamin I. Page performed a content analysis of daily news stories and found that Vietnam news had the greatest impact on both Johnson's, and Nixon's, popularity. “The Impact of Events on Presidential Popularity: The Johnson and Nixon Administrations,” in Perspectives on the Presidency, ed. WiMavsky, Aaron (Boston: Little, Brown, 1975), pp. 143–45Google Scholar.

¹² Mueller, 23–24; Mueller only considers cumulative war indices such as the number of U.S. war casualties since the war began. Using cumulative totals rather than each month's increment obviously increases the degree of multicollinearity with time. Later we shall examine noncumulative indices of the war.

¹³ The monthly war casualties are drawn from Figure 2.1 of Mueller, John, War, Presidents and Public Opinion (New York: Wiley, 1973), p. 36Google Scholar.

¹⁴ Mueller reports survey evidence showing, in fact, that the public has generally inaccurate knowledge about war casualties. Mueller, , War, pp. 35–62Google Scholar.

¹⁵ The correlation is for the logarithmic transformation (base 10) of the casualty index. This works marginally better than the untransformed variable suggesting that comparatively small numbers of U.S. casualties during the early stage of the Korean War had a disproportionate negative impact on President Truman's popularity. Because the job performance question was asked intermittently between 1945 and 1949 and because data for some of the other independent variables are missing during this period the analysis will be restricted to Truman's second term commencing in January, 1949.

¹⁶ Both of these indices are drawn from Milstein, Jeffrey, Dynamics of the Vietnam War (Columbus: Ohio State University Press, 1974)Google Scholar, Appendix.

¹⁷ After Watergate became recognized as a serious threat to President Nixon, the Gallup Poll began asking respondents their opinion about Nixon's involvement. Did he participate in the burglary and/or the cover up, just know about it, or “did he have no known knowledge about the bugging and spoke up as soon as he learned about it?” Over the year and a half period during which the question was asked the percent answering “no knowledge” shrunk from 23 percent to 11 percent. The Gallup Opinion Index, September, 1974, p. 16Google Scholar. The percentage believing the president was innocent correlated at .61 with Nixon's popularity.

¹⁸ Katona, George, The Mass Consumption Society (New York: McGraw-Hill, 1964), p. 143Google Scholar.

¹⁹ Mueller, , War, p. 238Google Scholar.

²⁰ Their studies used real income which is not available monthly; see Kramer, Gerald, “Short-term Fluctuations in U.S. Voting Behavior, 1896–1964,” American Political Science Review, 65 (March 1971), 131–43CrossRefGoogle Scholar; Tufte, Edward, “Determinants of Outcomes of Mid-term Congressional Elections,” American Political Science Review, 69 (December 1975), 812–26CrossRefGoogle Scholar; Bloom, Howard S. and Price, H. Douglas, “Voter Response to Short-run Economic Conditions: The Asymmetric Effect of Prosperity and Recession,” American Political Science Review, 69 (12 1975), 1240–54CrossRefGoogle Scholar.

²¹ Three different averages were tested for unemployment and consumer prices: 2, 3, 4; 5, 6, 7; and 11, 12, 13 previous month averages. The six-month moving average was found to work best. This agrees with Henry C. Kenski's report after testing a large number of economic measures. “The Impact of Economic Conditions on Presidential Popularity,” Journal of Politics, 39 (08 1977), 764–73CrossRefGoogle Scholar.

²² This occurs when controlling for Korean War casualties, Early Term effect, and Rally, all of which are examined below. In the final regression equations in Table 5, change in unemployment is omitted for Truman since it fails to approach statistical significance.

²³ Kenski also finds inflation to be a probable cause of popularity. See “Impact of Economic Conditions,” 25–27.

²⁴ A Bureau of Labor Statistics official recalled, “The Council of Economic Advisors felt there was too much good news earlier in the year and too much deterioration later in the year.” Cowan, Edward, “Jobless Rate: Elusive Statistic,” New York Times, January 13, 1978, p. A–11Google Scholar.

²⁵ Polsby, Nelson, Congress and the Presidency (Englewood Cliffs, N.J.: Prentice-Hall, 1964), p. 25Google Scholar. In the second edition he lists the “before-after” effects of international crises on the president's popularity, p. 44, Table 1. See also Mueller, , “Presidential Popularity,” 21Google Scholar.

²⁶ Mueller, , “Presidential Popularity,” 21Google Scholar.

²⁷ Mueller, , “Presidential Popularity,” 22Google Scholar.

²⁸ The three newspapers examined were the Atlanta Constitution, the Chicago Tribune, and the San Francisco Chronicle.

²⁹ Mueller, , “Presidential Popularity,” 22Google Scholar. Technically, it was not necessary for Mueller to begin each term as a rally point. For the second term the count since the last rally point could have continued uninterrupted by the reelection of the president. Mueller's rally variable correlates at –.11 with popularity.

³⁰ Such a procedure is admittedly arbitrary. Several other scoring procedures were tested but without important changes in the relationships. It is interesting that a linear decline in the effect of an event over five months works as well as the learning curve.

³¹ The index is crude in that it picks up whatever is occurring to the president's popularity during the first six months. It ignores the possibility that some “early term” periods will last longer than others. For example, the surge in support at the outset of the second term may be of shorter duration than for the first. The coefficient for Early Term will represent an “average” effect for these two periods, perhaps being too weak for the first and too strong for the second.

³² Sears, David O. and Whitney, Richard E., Political Persuasion (Morristown, N.J.: General Learning, 1973), pp. 12–17Google Scholar.

³³ Paul, I. H., “Impressions of Personality, Authoritarianism, and the Fait Accompli Effect,” Journal of Abnormal and Social Psychology, 53 (November 1956), 341CrossRefGoogle ScholarPubMed. For a summary of the prominent findings in this area of research see Sears, and Whitney, , Political Persuasion, pp. 12–17Google Scholar.

³⁴ Anderson, Lynn R. and Bass, Alan R., “Some Effects of Victory or Defeat Upon Perception of Political Candidates,” Journal of Social Psychology, 73 (October 1967), 227–40CrossRefGoogle ScholarPubMed; Raven, Bertram H. and Gallo, Philip S., “The Effects of Nominating Conventions, Elections, and Reference Group Identification upon the Perception of Political Figures,” Human Relations, 18 (08 1965), 217–29CrossRefGoogle Scholar.

³⁵ Wolfenstein, Martha and Kliman, Gilbert, Children and the Death of a President (Garden City, N.J.: Doubleday, 1965)Google Scholar, and Orlansky, Harold, “Reactions to the Death of President Roosevelt,” Journal of Social Psychology, 26 (11 1947), 225–66CrossRefGoogle ScholarPubMed.

³⁶ Barber, James D., Presidential Character (Englewood Cliffs, N.J.: Prentice-Hall, 1972), p. 5Google Scholar.

³⁷ Stimson, 30–33.

³⁸ Brody and Page view presidential popularity as autoregressive as well and incorporate and lagged variable into their analysis. In fact it is this variable and not their news indices which appears to explain most of the variance in popularity. Their rationale for treating popularity as autoregressive is somewhat different from the one employed here. To them it represents a base which in the absence of intervening events dictates the next month's approval level. Brody and Page do not consider the issue of serial correlation; “The Impact of Events…”

³⁹ Malinvaud, E., Statistical Methods of Econometrics (London: North-Holland, 1970)Google Scholar.

⁴⁰ Unfortunately for our understanding of the Truman and Johnson trends, there are too few observations prior to the start of the wars to obtain an independent estimate of unemployment.

⁴¹ One study of public opinion toward Vietnam War policies has shown that regardless of whether the respondent identified himself as a “hawk” or “dove” any change in policy would be greeted with at least temporary approval. See Verba, Sidneyet al., “Public Opinion and the War in Vietnam,” American Political Science Review, 61 (06 1967), 317–33CrossRefGoogle Scholar.

⁴² The arbitrary measurement of the Early Term and Rally probably reduces the explanatory power of the model and introduces serial correlation. This scoring procedure works as well as any other, but the problem lies in applying the same attrition model to qualitatively different types of events. All international events will obviously not have the same impact on public opinion. Moreover, attrition over subsequent months will vary according to the event's intensity and duration. Some, therefore, should last only a few months while others should perhaps be extended over a longer period. Unfortunately, we currently have little independent basis for weighting events to tap their differential significance. An example of its potential effect on correlations of the residual term is the Nixon Vietnamization speech in November 1969. Even with a high rally score assigned to this event, Nixon's actual popularity was some 12 percentage points greater than estimated. For the next several months the estimated popularity was less than the actual, tending to produce a correlation of the residuals.

⁴³ Mueller, , “Presidential Popularity,” 34Google Scholar.

⁴⁴ The fact that each of the coefficients rests between 0 and +1 informs us that popularity approaches a stable equilibrium. This means that in a constant political environment popularity will gradually move to an equilibrium level probably reflecting in part the partisan division of the electorate. For an example of the interpretation of equilibrium behavior from autoregressive terms, see Spafford, Duff, “A Note on the ‘Equilibrium’ Division of the Vote,” American Political Science Review, 65 (03 1971), 180–83CrossRefGoogle Scholar.

⁴⁵ It is curious that when time is included in the equation for the Nixon administration the coefficients for the other variables–including Watergate–improve: Pop = 1.5(Rally) – 284(Prices) – 19(Watergate) – .6(Early Term) – .04(lag of popularity) – .03(Time) + 68.6. R ² = .89.

⁴⁶ See for example Neustadt, , Presidential Power (New York: Wiley, 1960)Google Scholar, ch. 5; Edwards, George, “Presidential Influence in the House: Presidential Prestige as a Source of Presidential Power,” American Political Science Review, 70 (March 1976), 101–13CrossRefGoogle Scholar; Kernell, Samuel, “The Truman Doctrine Speech: A Case Study of the Dynamics of Presidential Opinion Leadership,” Social Science History, 1 (Fall 1975), 20–44CrossRefGoogle Scholar; and Kernell, Samuel, “Presidential Popularity and Negative Voting: An Alternative Explanation of the Midterm Congressional Decline of the President's Party,” American Political Science Review, 71 (03 1977), 44–66CrossRefGoogle Scholar.

⁴⁷ Hibbs, “Problems of Statistical Estimation.”

⁴⁸ Hibbs.

⁴⁹ Malinvaud, p. 569.