Modeling the rally event in presidential politics: Toward conceptual and methodological clarity

Abstract

This research provides conceptual clarity both to the modeling of rally events in presidential politics by introducing the stimulus-cognitive response model and to the measurement of rally events by presenting the steps necessary for proper specification and empirical analysis using the time-series method. The utility of stimulus-cognitive response modeling in conjunction with time-series is demonstrated by an examination of the attempted assassination of President Ronald Reagan. Several of the specific findings are counterintuitive and have implications for the presidency-congress literature.

Notes

1 See, e.g., Mueller (1970, p.19), Cronin and Genovese (1998, pp. 75–85), Edwards and Wayne (1985, pp. 109–110), Eisenstein and Witting (2000), Ragsdale (1993, pp. 1943–1948) and Stimson (1976). The inevitability of a downward trend in presidential approval is questioned, though (see Ostrom & Smith, 1992; also Behr, 2003).

2 The meaningfulness of Mueller's inclusion of the criterion “international in nature” may be questioned. Being ‘international in nature’ certainly is not a necessary condition for a dramatic change in presidential approval. While it is true that a distinction may be made, although somewhat blurred, between domestic and international issues, it is also true that any distinction made between a public opinion rally stemming from a domestic event and one stemming from and international event is artificial—a rally remains a rally despite the classification of the stimulus. In addition, purported “domestic” crises, and presidents’ attention to, or management of, such crises, can have international implications and, therefore, be in fact international in nature. According to Bennett (1980, p. 352), “the need for public support applies to the President's action in the world … other nations have struck hard bargains with Presidents whose opinion levels could not stand another failure.”

3 In a more recent analysis, Edwards and Gallup (1990) use Mueller's criteria for identifying international crises, but refine the criteria by stating that the measure of presidential approval must raise more than 5% to compensate for possible polling error. Hugick and Gallup (1991) adjust slightly Mueller's criteria and state that the rise must exceed 3%. MacKuen (1983) utilizes the World Almanac to identify events. Likewise, Brace and Hinckley (1992) define events as those that appear in the summaries of three well-known almanacs. Ostrom and Simon (1989) define events as those items that receive front-page coverage in the New York Times. As noted early by Brody and Page (1975, p. 138), “the selection of events to correspond with opinion changes is necessarily subjective.”

4 Nimmo (1978, pp. 184–185) distinguishes between happenings, “all those things that go on in the world whether we are aware of them or not,” and events, those happenings that are non-routine in that people collectively think and discuss them. The attempted assassination of Reagan appears each day on the front page of the New York Times from March 31 to April 12. The following month only four articles appear. For the ratio of positive to negative news stories see Cook and Ragsdale (1998, p. 324).

5 Zaller (1998, pp. 185–186) notes that following an event or scandal the public is quick to return to the “bottom line politics” of policy issues of political substance. Rallies may have “sharp” increases but also “precipitous” drops (Hugick & Gallup, 1991, p. 15).

6 It is of interest to note that the logic of floor and ceiling effects for trend data may temper the statement of purely theoretical expectations contained in a hypothesis. Construction of this hypothesis, for example, is dependent, in part, upon knowing that Republican identifiers have fairly high approval prior to the event. If Republican approval is low, then the hypothesis may have to be stated much differently. This is instructive since most social science trend data have a floor and many have, in addition, a ceiling.

7 Yet Callaghan and Virtanen (1993) find that both Republicans and Democrats rally to support the president and the duration holds for the same period. Independents evidenced both the largest boost and the shortest slowest decay.

8 A test for a relationship using regression, especially that of a conditional relationship, necessarily requires a number of occurrences that are qualitatively similar. In fact, one of the theoretical stumbling blocks of earlier research is the classification of dissimilar events into artificial categories (see, e.g., Lee, 1977; Mueller, 1973). The wounding of a president in the attempted assassination is an uncommon occurrence. There are several assassination attempts on President Ford's life, but he is not injured. With the exception of Eisenhower's October 1955 heart attack, there are few other instances when the health of the president is in serious question. Simply, there are few, if any, occurrences that are qualitatively similar and, therefore, regression may not be the best approach. However, given a single event, the utility of time-series is found in that it allows for an empirical analysis.

9 The theoretical concerns advanced by Edwards revolve around research that commits an ecological fallacy by comparing trends in aggregate employment data with presidential approval.

10 Neither the raw or the differenced ACF and PACFs indicate seasonality.

11 The Q statistics are significant at the 95% confidence interval indicating that the null hypothesis is rejected and the residuals are indistinguishable from white noise.

12 The time-series data sets for Republican, Independent, and Democratic identifiers are slightly different since they do not incorporate the ABC-Washington Post Survey data (the poll does not distinguish partisanship). Similar to the aggregated series, these additional analyses follow the same logic, methods of specification, and construction of an integrated poll with a 2-week unit of analysis.

13 Perhaps this is due to Independents’ disenchantment with the political system as a whole and feelings that governmental institutions are not capable of responding effectively to major problems (see Wattenberg, 1991). Keith, Magleby, Nelson, Westlye, and Wolfinger (1992) show that Independents are far from homogeneous. Most Independents identify as either Republican-leaning or Democratic-leaning and behave as partisans. They find that political alienation is related only with “pure” Independents, a small portion of all those identifying as Independents. The authors caution that an analysis, such as this one, that fails to disaggregate the “otherwise disparate responses by the three varieties of Independents may aggregate into a misleadingly neutral pattern” (p. 173).

14 See, e.g., Bond and Fleisher (1990, pp. 176–195), Cohen (2000, pp. 162–165), Edwards (1989), Kellner (1990, p. 136), King and Ragsdale (1988, pp. 25–26), Light (1991), Neustadt (1960), Ostrom and Simon (1985), Peterson (1990, p. 135), Pfiffner (1988, pp. 137–139), Rivers and Rose (1985), Shull and Shaw (2004, p. 590), and Zukin and Carter (1982, p. 208).

15 It is clear that the pollsters recognized the wounding of Reagan as a significant occurrence, one that could possibly have measurable effects on the popularity of the president. In late April, less than 4 weeks after the assassination event, the Harris Survey queered respondents, “After the assassination attempt, did your opinion of Ronald Reagan as a person change for the better, for the worse or did it stay about the same?” While 77% of respondents stated that their opinion did not change since the event, 20% reported that their opinion had in fact changed for the better. If we take this question at face value, we undoubtedly may surmise that the event had a significant positive effect on Reagan's popularity.

16 In March 1981, the Reagan administration made the first of two proposals to reduce the benefits of Social Security. The popularity of the federal Social Security program made any suggestion of cuts a politically explosive issue. By the time the president's proposal reached Congress in the middle of May 1981, it received little support from either Democrats or Republicans. The Republican controlled Senate, for example, unanimously rejected the administration's proposal and handed Reagan his largest policy defeat since he took office (Nathan, 1983, p. 56; see also Edwards, 1996, pp. 210–212; Killerman, 1984, pp. 221–229). Cohen (1997) calls Reagan's handling of Social Security in the first half of 1981 “the benefit reduction debacle” (p. 212) which “generated great political heat” (p. 214). Reagan soon abandoned support for these cuts. For a discussion of the bipartisan coalitions during this period see (Greenstein, 1983, pp. 16–18; Light, 1991, p. 248; Pfiffner, 1994, pp. 164–165).