CIAO DATE: 6/00
Domestic and International Determinants of Presidential Rallies
Athanasios Hristoulas
Patrick James
Jean Sebastien Rioux
International Studies Association
41st Annual Convention
Los Angeles, CA
March 14-18, 2000
* We are grateful to Walter Dean Burnham, Robbie H. Hart, Zeev Maoz, Evan Ringquist Wayne Steger and Duncan Wood for helpful commentaries, and Monica Lugo for research assistance.
How does the foreign policy activity of a state impact upon its domestic politics? 1 James and Rioux (1998) attempted to answer that question in the context of the US' experience with Foreign Policy Crises (FPCs). The present work builds on their findings by examining the relative importance of domestic and international determinants of presidential popularity. It does this by expanding the testing procedure to include multiple measures of conflict involvement from different datasets and also by including novel and never before examined measures of economic performance. The investigation begins with a brief discussion of the literature on linkage politics, especially recent theoretical developments. The next step is to develop propositions about the international factors that are most likely to impact upon domestic politics. Variables are operationalized and monthly data pertaining to the United States from 1953 to 1994 are examined. The findings suggest that domestic economic factors have the greatest impact on presidential popularity. The study culminates in suggestions for further research.
Background: Linking Domestic and International Politics
In recent years, students of international politics have returned to what is known as "linkage politics": how domestic politics can affect the foreign policy of a state. Consider, for example, the so-called scapegoat or diversionary hypothesis (James 1987; Levy 1989; Morgan and Bickers 1992; Levy and Valiki 1992; Moore and Lanoue 1998), for which recent studies have produced mostly encouraging results (Ostrom and Job 1986; James and Oneal 1991; Brace and Hinckley 1992; Morgan and Bickers 1992; James and Hristoulas 1994; DeRouen 1995; Raber 1998; Morgan and Anderson 1999; Prins 1999; Scott, and Carter 1999; Heldt 1999). Domestic factors such as the economy and presidential popularity seem to affect the likelihood of dispute or crisis involvement by US presidents.
Another linkage-oriented approach focuses on how the international political environment affects politics within states; in others words, the causal effects also are unidirectional but presumed to go from outside the state to inside (Gourevitch 1978: 882; 1986). International conflict certainly can have dramatic effects on domestic politics (Mueller 1994: 76; Bueno de Mesquita, Siverson, and Woller 1992; Bueno de Mesquita and Siverson 1995). Furthermore, evidence suggests that the leaders of democratic states are aware of the political side-effects created by foreign policy actions (Lamare 1991: 8; Siverson 1995: 484-486; Bennett and Stam 1996: 253).
Finally, one study uses a simultaneous equation model to asses an overall network of effects among politics, economics, and the use of force in crises (DeRouen 1995). The results are generally consistent with the diversionary model; force is more likely when a crisis is severe (i.e., high Soviet involvement, etc.), the US is not otherwise involved in an ongoing conflagration, and public presidential approval is relatively low.
In sum, preliminary evidence seems to suggest that a relationship does exist between domestic social, political and economic conditions and international behavior. And the relationship seems to work both ways. Under certain conditions, the foreign policy behavior of a state affects its domestic conditions and vice-versa.
This study seeks to combine the relevant conditions in a more comprehensive model that focuses on both FPCs and Militarized Interstate Disputes (MIDs). More specifically, the present investigation seeks a better understanding of the ways in which foreign and domestic events together affect the popularity of a US president.
Foreign Affairs and Presidential Popularity: What are Rally Events?
A "rally" is a foreign policy event that is believed to have a positive impact on the standing (more specifically, popularity) of a US president. Beyond that, however, agreement on what constitutes such an event is tentative at best. The natural starting point is Mueller (1973: 209): "a rally point must be associated with an event which (1) is international and (2) involves the United States and particularly the president directly; and it must be (3) specific, dramatic, and sharply focused". Fleisher and Bond (1988: 750, 754-755, 758), however, find that greater support in Congress for the president on foreign policy issues emerges only for Republican presidents and, even then, the source is restricted to liberal Democrats. Since the concern in the present study is with support for the president among the voting public, the above-noted result does not directly contradict the standard practice of focusing on international events in evaluating rally effects.
More recently, Brody (1991) argues that international crises are special moments in public opinion. Brody adds a qualification to Mueller's list of rally prerequisites. He argues that an international crisis is likely to lead to an increase in presidential approval when "the administration has a virtual monopoly of information about the situation [and] opposition political leaders tend to refrain from comment or to make cautiously supportive statements" (Brody 1991: 64). Opposition spokespersons are motivated to alter their normal stance by an "unknown mix of patriotism and outrage at the threat to the country" (Brody 1991: 64). In other words successful rally events are associated with bipartisan support for the president's actions and are presented thus in the media. Using a case study methodology, Brody finds strong support for the hypothesized relationship.
Focusing on Great Britain, two related studies examine the impact of the Falklands War of 1982 on the popularity of the Thatcher government. Crewe (1985), for example, observes that between the time of the Argentine invasion on March 31, 1982 and the British victory of June 14, Conservative support in the opinion polls rose by some 15 percentage points. Crewe concludes that " in the space of three months, public opinion and party politics had been transformed". Dunleavy and Husbands (1985) concur that the Falklands War substantially increased government popularity. Their model -- which contains only two independent variables, unemployment and a Falklands dummy variable --accounts for 87 per cent of the variance in government popularity between September 1979 and April 1983. Dunleavy and Husbands also agree that the Falklands led to about a 15 per cent increase in approval for Thatcher's Conservatives.
Therefore, many scholars concur that involvement in foreign conflict can lead to a rally effect 2 , although the evidence is limited primarily to states at the apex of the international system. However, there still is no standard operationalization of this concept. Studies assessing the rally effect of international events focus on a range of international events; examples include "uses of force" (Brace and Hinckley 1992) or MIDs and FPCs (Brecher and Wilkenfeld 1997). What these studies do not answer is whether or not factors such as (i) the type of adversary; (ii) the level of hostility; (iii) costs associated with the conflict; or finally (iv) different types of conflict, make a difference. Therefore, multiple measures of conflict involvement that take into account the above-mentioned factors seem warranted.
Studies of rally effects are limited further by the existing approach toward measurement of executive approval. For example, Ostrom and Job (1986), Morgan and Bickers (1992) and James and Hristoulas (1994), among others, look only at the quarter-year time interval. It would seem that any resulting effect could be produced by confounding factors that occur in those three month increments. Furthermore, how should the various approval ratings be averaged for this time frame? Thus a shorter interval is desirable, if only to avoid the averaging problem.
Domestic Factors and the Popularity of a President: The Economy
Brody (1991, 91) argues that "because the president is the nation's chief policy-maker, it would be reasonable and just if the state of the economy figured prominently in the public's evaluation of presidential job performance". Although the literature (see, for example, Goodhart and Bhansali 1970; Kernell 1978; Hibbs 1982; MacKuen 1983; and Sanders, et al 1987; Fordham 1998) suggests that objective macro-economic indicators are important, Brody (1991) argues that these type of indicators may affect the individual voter less than "subjective judgments about individual economic prospects for the future". Brody (1991: 96) proposes two models for explaining how government popularity is affected by economic indicators. The first model, referred to as 'direct effects', proposes that voters are sensitive to economic performance and judge the president -- as chief economic policy-maker -- according to how they are affected personally. The second approach -- referred to as the spectator model -- proposes that direct personal experience is much less important: "Irrespective of personal circumstances and experiences, a failing economy will be reflected in negative presidential job ratings" (Brody 1991: 91). Brody finds that the first model better predicts presidential popularity.
These results suggest that the standard measure used in examining the impact of economic factors on presidential popularity, namely, the Misery Index, may not be well-suited to the purposes at hand. The Index is operationalized as the sum of the unemployment inflation rates for a given month, weighted by the percentage of the general US public identifying the economy as the worst domestic problem in monthly polls (Ostrom and Job 1986). Thus, the Index does not directly measure public concerns with respect to the economy. Although the inclusion of the percentage of the general public identifying the economy as the worst domestic problem begins to capture public "mood" or sentiment with respect to the economy, the objective indicators of inflation and unemployment have been found to be less effective in previous studies. Thus, as suggested by Brody (1991), a more accurate assessment of the impact of economic conditions on public attitudes also should examine subjective economic factors independently from the more objective, macro indicators.
What is Expected
Contrary to the recent pattern of findings, we believe that more complete data and a shorter time frame will reveal that involvement in conflict, in general, does not produce significant or long-term effects on presidential standing in the US. Put another way, any rally effect that appears will be minimal and short-lived (Oneal and Bryan 1995). Thus, any President contemplating the use of international conflict to substantially boost his domestic standing will be disappointed. Public opinion, most notably with regard to presidential approval in the US, is hypothesized to be much more a function of domestic political and economic conditions.
Costs of involvement in international conflict, however, should tell a different story. Along with material damage, casualties represent costs that the public can understand readily. As the costs resulting from involvement in conflict increase in terms of lives and material, the public is expected to react proportionately, consistent with diminishing marginal effects (Ostrom and Job 1986). Reactions to these specific factors could range from perceptions, such as deterioration in executive standing, to behavior, including protests and even violence. From a leader's point of view, higher costs of involvement in international conflict are expected to have an unfavorable and rapid domestic impact. By contrast, benefits, such as economic rents from control over an adversary's territory, are experienced gradually and do not affect short-term public perceptions of a leader's job performance.
Two basic hypotheses emerge from the preceding discussion. The first hypothesis is consistent with both the realist and linkage perspectives (i.e., connection of domestic and foreign events):
H1: Higher costs of involvement in an international crisis will have an unfavorable impact on domestic politics, assessed in terms of the incumbent leaders point of view.
Moreover, as suggested by previous research, economic factors seem to have an important impact on the popularity of a president. Therefore, a second hypothesis, which emphasizes the importance of domestic economic conditions on presidential popularity, is included in the analysis:
H2: Worsening economic conditions will have an unfavorable impact on domestic politics, assessed in terms of the incumbent leaders point of view.
The data analysis that follows is an extension of James and Rioux (1998), which focused on the United States. The US continues to be the central concern because of the desire to compare the present results with those of James and Rioux (1998), which dealt exclusively with FPCs.
Data and Measurement
The Dependent Variables: Presidential Popularity
The unit of analysis is monthly observations. Monthly data on conflict activity, costs of involvement in international conflict, economic conditions and domestic politics have been gathered for the United States from 1953 through 1992, inclusive. This design and time frame which follows James and Rioux (1998), represents two improvements over previous studies: (1) it allows monthly rather than quarterly analysis, as in Ostrom and Job (1986), James and Oneal (1991) and James and Hristoulas (1994); and (2) the time span is somewhat longer than that of previous studies, which generally covered 1953 to 1985 (e.g., Oneal and Bryan 1995). 3 Although data for some variables are available over a longer period, most notably those concerning crises and militarized disputes, the analysis requires complete data on US presidential approval. Monthly data are available from the Gallup organization only since 1953 (Edwards with Gallup 1990; Gallup 1991-1994). Each dependent variable now will be described.
Four components of executive support are used. The first is the overall level of public support, APPROVAL, measured by the percentage of respondents who approve of the president's performance (Edwards with Gallup 1990). The second, third, and fourth measures of support are derived by looking at PARTISAN, INDEPENDENT and OPPOSITION APPROVAL levels. The polling question used by the Gallup Organization has not changed since 1953 and reads as follows: "Do you approve or disapprove of the way [president's name] is handling his job as president?" Gallup also asks respondents to give their party affiliation. Since neither the phrasing of the polling question nor the two-party structure have changed since 1953, this measure of presidential approval is robust over time.
These alternative measures capture each president's popularity within his ruling coalition (Morgan and Bickers 1992) -- meaning those party members who might donate money, advice, or otherwise help in reelection and among independents and persons of the president's opposition party. These measures are important for two reasons.
First, along the lines of a "cybernetic" model (Ostrom and Job 1986), the case could be made that the president may have an incentive to closely monitor the approval of his natural base of support more than the public-at-large, which includes many who would not re-elect the president no matter how strong his performance in office. Evidence exists that retaining the presumably natural coalition is most important to presidential support in the House of Representatives, which is consistent with the idea that partisans should be monitored very closely (Fleisher and Bond 1983: 753). Also, it can be argued that presidents have an incentive to monitor their performance among independent voters as well, if they believe independents somehow represent a segment of "median voters" that can be captured into the ruling coalition (Downs 1957). Thus, by examining all the different components of executive approval, potential differences among levels of support can be isolated and analyzed, offering yet another improvement over other studies that only examine aggregate approval levels.
Second, theoretically interesting questions may arise in comparing reactions of the general public and the president's partisans, as well as independent voters and members of the opposition party, if differences are observed. Research on presidential popularity among partisans and the public in general as related to congressional voting suggests that such differences will not emerge. Bond and Fleisher (1984: 300-301) and Bond, Fleisher and Northrup (1988: 59, 61) report that the impact of presidential popularity is marginal for both voters overall and partisans. 4
Independent Variables: Foreign Policy Crises, Militarized Interstate Disputes and Economic Conditions
Twelve independent variables appear in the regression model. Eight are relevant directly to testing the hypotheses; the other two control for fluctuations in presidential approval that are common to all administrations.
The first two independent variables represent general controls on presidential approval. The first, PRESIDENT, is a dichotomous variable. Its purpose is to control for the idiosyncratic differences among the mean level of popular support for each president. Each US president comes into office with his own personality, agenda, and rapport with the public and media environment, and accordingly, has a different "natural" mean level of support. This difference must be filtered (Kernell 1986: 180). Brace and Hinckley (1992: 30-31) find that this technique is useful to "separate the effects common to all administrations and the circumstances that vary among administrations" in order to find meaningful statistical effects from the other independent variables in the model.
The second independent variable, MONTH, is included to control for a structural component of presidential approval that is common to all administrations. Brace and Hinckley (1992: 23-24) label this as the "cycle of deflating expectation," or more simply, the "decay of support." This is an observable decline in general approval over the first 30 months, which picks up after the 32nd month, until the election in the 48th month of each administration. This common cycle is not a monotonic decay of support over the four years of office, because it does pick up during the election year. It is interesting to note that this decay in support, followed by a rise in the final year, also is present in the second administration of those presidents who served two terms during the time period covered by this study (i.e., Eisenhower, Nixon, and Reagan). The variable introduced for each month of the presidential administration controls for this structural feature of presidential approval. 5 Thus, to recap, the first two independent variables appear as controls for the known structural features of US presidential approval.
The third and fourth independent variables examine the state of the economy. The subjective measure of Public Economic EXPECTATIONS of the future is included in the analysis because, as suggested, it is critically important to control for difficult economic times. These factors are known to greatly affect domestic support for the president while being unrelated to international events per se (James and Oneal 1991). Collected by the U.S. Department of Commerce (1996), the variable asks respondents to indicate whether their economic expectations of their immediate future are expected to be (1) the same; (2) worse; or (3) better. This variable is believed superior to previous studies that look only at inflation and unemployment (or some combination thereof) because it is a measure of perceived economic conditions. Theoretically, the public should turn against the president when it believes that economic times are bad, regardless of actual conditions. The MISERY INDEX is included for the purposes of comparison and also in order to link the present investigation to previous studies such as James and Rioux (1998). 6
The next eight independent variables, which pertain to the international level, are most central to this study. CRISIS is a dichotomous variable that depicts the entry of the United States into a crisis as defined by the International Crisis Behavior (ICB) Project: A foreign policy crisis, that is, "a crisis viewed from the perspective of an individual state leadership, is a situation with three [individually] necessary and [collectively] sufficient conditions deriving from a change in a state's external or internal environment. All three are perceptions held by the highest level decision-makers of the actors concerned: a threat to basic values, along with the awareness of a finite time for response to the external value threat, and a high probability of involvement in military hostilities" (Brecher and Wilkenfeld 1997: 3, emphasis in the original). This class of events is more inclusive than uses of force or interstate wars. CRISIS takes a value of one when the US enters into a crisis in a given month and as long as it remains a crisis actor, and zero otherwise. Since the domestic-level variables described earlier control for other possible influences on the president's level of approval, the direct impact of crisis can be assessed.
The other independent variables include USSR ACTIVITY and MAJOR REGION. These two variables are explained in detail by Oneal and Bryan (1995: 392). Each is significant in accounting for the prominence of coverage given by the New York Times for a given foreign policy crisis, which in turn affects the magnitude of any rally that might occur. USSR ACTIVITY is the USSR's (later Russias) level of involvement in the crisis and MAJOR REGION identifies the location of the crisis as either "major" (Central America, Caribbean, Europe, Middle East, North Africa) or minor (Asia, Oceania, South America, Sub-Saharan Africa), from the point of view of the US. 7 As expressed by H1, none of these factors is expected to have a significant impact on either domestic politics in general or presidential approval in particular.
Finally, the costs of involvement in international conflict for the United States are represented by two variables: Casualties experienced, CASUALTIES, and amount of force exerted, FORCE. CASUALTIES corresponds to the human cost of the United States' involvement in armed conflict, which results from escalation of a crisis. As opposed to some kind of economic index, this particular measurement is used because material costs are so difficult to assess. Furthermore, the material losses are expected to carry less weight than the known costs expressed as the number of casualties (which, consistent with H2, are expected to have a negative domestic impact). The variable is based on cumulative battle deaths for the US, logarithmically transformed, consistent with a diminishing marginal effect (Ostrom and Job 1986).
The final independent variable, FORCE, is the level of force used by the US and has a range of four values: (0) no use of force; (1) minor use of force; (2) either one major force component or a strategic nuclear unit is used; and (3) two or three major force components are used, plus at least one strategic nuclear unit (James and Oneal 1991: 316; Blechman and Kaplan 1978: 50-51).
Three variables from the MIDs dataset (compiled by Gochman and Maoz 1996) also are included in the analysis for the purposes of comparison. The first, coded dicotomously, recognizes US dispute involvement. Months where the US was involved in a MID are coded one and all others zero. The second MID variable, FATALITIES, serves as a measure of the number of deaths experienced by US armed forces during a dispute. It is coded as follows: (0) no deaths, (1) 1 to 25; (2) 26 to 100; (3) 101 to 250; (4) 251 to 500; (5) 501 to 999; (6) greater than 999. The third and final MID variable is level of HOSTILITY which is coded as (1) no militarized action; (2) threat to use force; (3) display of force; (4) use of force; (5) war. Once again, we expect that only FATALITIES and HOSTILITY will have an effect on presidential popularity.
Model Specification and Estimation
With an operational set of variables two similar models can be specified. The equations are as follows, with the dependent variable assuming the four alternatives discussed earlier:
FPC Model [1]
APPROVAL-Xit =
MID Model [2]
APPROVAL-Bit=
Where
APPROVAL-Xit= percentage of survey respondents who approve of the way president is handing his job at time t; models are run for GENERAL, PARTISAN, INDEPENDENT, and OPPOSITION approval levels.
PRESIDENTi = intercept change attributable to president i (Brace and Hinckley 1992);
MONTHt = intercept change attributable to month t;
MISERYt = effect of the economy for president i at time t;
EXPECTATIONSt = effect of economic expectations for president i at time t;
CRISISit = effect of a crisis for president i at month t;
MAJOR REGIONj = Region of Crisis j;
USSR INVOLVEMENTj = USSR / Russia's level of involvement in crisis j;
CASUALTIESit = number of casualties for president i at month t;
FORCEit = effect of the level of force used for president i at month t;
MIDSj = effect of MID involvement for president i at month t;
FATALITIESit = effect of the number of fatalities for president i at month t;
HOSTILITYit = effect of the level of hostility used for president i at month t;
eit = error term.
As noted earlier, only casualties and force (similarly fatalities and hostility for the MID model) are anticipated to affect the domestic polity to any degree in the FPC model. In each instance, because of the costs involved, the impact should be negative. As levels of war involvement and use of force increase, presidential approval should decline. The fact of a crisis itself, and the disruption it represents at the international level, are expected to produce insignificant effects on the domestic polity.
Given the time series data for presidential approval, the models are estimated using the Cochrane-Orcutt autoregression technique. This procedure corrects for the presence of a first-order autocorrelation process (AR-1) that was detected through a Durbin-Watson test and examinations of the ACF and PACFs for the residuals of the four measures of approval levels. The Cochrane-Orcutt technique is useful because it automatically estimates and fits a rho value in the regression model; moreover, interpretation of the estimated coefficient is identical to Ordinary Least Squares (OLS). The final Durbin-Watson d-statistic and the rho value will be reported for each model.
Finally, the respective dichotomous variables represent categories that are not meant to be measured on a ratio scale, yet are important in assessing differences among groups or categories. These variables do not pose serious methodological problems when a model is correctly specified (Hardy 1993).
Do FPCs and MIDs Affect Domestic Politics?
The results of obtained for the two respective models are presented in Tables 1 and 2. Table 1 reports the findings for the FPC model and Table 2 represents the MID model. Two separate models have been estimated in order to independently test the effect of using data sets. Different models were tested even though regression diagnostics did not reveal multicollinearity among the two different sets of variables. The relative absence of multicollinearity is quite surprising and further emphasizes the point that previous findings were extremely dependent on the type of data set being used.
Multicollinearity did not present itself as a problem with respect to the economic variables. The R-square between the Misery Index and Economic Expectations is 0.56, which suggests that the variables are related, but are measuring somewhat different things. The other independent variables also do not pose problems either: (a) the US does not always use force in a crisis (or high levels of hostility in a MID); (b) crises do not always occur in major regions; and (c) the USSR does not become active in all US foreign policy crises. So among these variables, collinearity varies from 0.50 to 0.68 and is not high enough to affect the models. 8
Table 1 Presidential Approval: Foreign Policy Crisis Model
Variables | General Approval | Partisan Approval | Independents | Opposition |
Crisis | 3.20** | 4.98*** | 3.04 | 2.91* |
(-2.14) | (-3.09) | (-1.38) | (-1.48) | |
Major Region | -0.54 | -1.48 | -1.8 | -1.56 |
(-0.52) | (-1.32) | (-1.17) | (-1.14) | |
USSR Activity | -0.33 | -0.48* | -0.53 | -0.27) |
(-1.22) | (-1.65) | (-1.31) | (-0.76) | |
Casualties | 0.00 | -0.00*** | -0.00* | 0.00 |
(-0.01) | (-2.67) | (-1.50) | (-1.06) | |
Force | -1.20** | -1.50*** | -0.71 | -0.82 |
(-2.15) | (-2.47) | (-0.86) | (-1.12) | |
Misery Index | -3.50*** | -2.20*** | 2.87*** | 4.16*** |
(-4.14) | (-2.99) | (-2.80) | (-4.09) | |
Expectations | 4.92*** | 3.91*** | 6.50*** | 5.01*** |
(-3.17) | (-3.57) | (-4.2) | (-2.94) | |
Eisenhower 2 | -14.86*** | -4.43* | -8.11 | -16.74*** |
(-3.74) | (-1.66) | (-2.14)** | (-3.90) | |
Kennedy | -13.96*** | -10.66*** | -10.60*** | -22.06*** |
(-2.79) | (-3.51) | (-2.45) | (-4.31) | |
Johnson | -19.64*** | -21.60*** | -23.32*** | -21.40*** |
(-3.97) | (-7.69) | (-5.82) | (-4.43) | |
Nixon 1 | -16.63*** | -4.63* | -10.40** | -12.99*** |
(-3.10) | (-1.51) | (-2.37) | (-2.47) | |
Nixon 2 | -40.59*** | -20.70*** | -28.69*** | -39.43*** |
(-6.34) | (-5.05) | (-4.94) | (-5.99) | |
Ford | -0.58 | -5.66 | 5.42 | 4.11 |
(-0.09) | (-1.34) | (-0.91) | (-0.61) | |
Carter | -5.30 | -19.16*** | -7.35 | -6.25 |
(-0.87) | (-4.86) | (-1.32) | (-0.98) |
NOTES FOR TABLE 1
The coefficients for the economic control variables will not be discussed in great detail, except to point out that the effect of these conditions (Misery Index and Economic Expectations) on presidential approval is significant and strong. For every unit of increase in the Misery Index, general presidential approval falls by about 3.50%. The effect is more extreme for opposition party voters where approval declines 4.16%. Given the fact that opposition voters have a greater propensity to dislike the President in comparison to partisan and independent voters, this finding is plausible.
Economic Expectations tells a slightly different story. Once again the coefficients are strong and in the expected direction (General Approval, 4.92%; Partisan, 3.91%; Independents 6.50% and Opposition, 5.01%); however, independents not opposition voters respond in the most extreme manner. It is unclear why this should be the case. Predictably, however, partisan voters punish the president the least when their economic expectations are not perceived as positively. Either way, the evidence strongly supports the notion that economic conditions, both objective and subjective, are important determinants of presidential popularity. Thus, H2 is strongly supported.
As stated previously, the theoretically relevant variables used to evaluate a possible rally effect are as follows: CRISIS, MAJOR REGION and USSR ACTIVITY; and the two variables representing the costs of involvement in crisis, and CASUALTIES and FORCE.
In TABLE 1 the effect of CRISIS is positive and significant for the voters in general, the president's partisans, and marginally significant for the opposition. In other words, the model predicts a slight rally effect for these groups ranging from about 3 to 5%. The rally effect among the independents is not significant. The other variables pertaining to crisis involvement generally do not produce any statistically significant coefficients.
Interestingly, MAJOR REGION and especially USSR Activity produce negative coefficients for all groups. One might have expected the American public to rally during a threat to the security in an area of vital US interest, or at least against the Soviets. These results may indicate a certain conservatism on the part of all voters; when the USSR becomes involved or when a crisis takes place in an area of vital US interest, they are more wary of potential escalation and their support is more volatile. However, it is best not to speculate too much on these results, since the coefficients are not particularly strong. The most reasonable conclusion is that a slight rally effect exists when looking at crises in general.
The two variables pertaining to costs of involvement tell a slightly different story. The variable CASUALTIES produces two significant coefficients for the partisans and independents. Thus, this particular operationalization of costs of crisis activity did not produce the anticipated results across the board. However, the coefficient for the variable FORCE is negative and significant for the general population and partisans, thus counteracting any potential rally effect of being involved in a 'safe' crisis. In other words, with crisis escalation, a possible rally effect dissipates, thus indicating a certain conservatism with regard to use of force. It is interesting that this coefficient is negative for every group, and highest for the president's partisans. This result lends mild support to H1; higher potential costs of crisis activity measured by the use of force can have a negative impact on approval, although it is not significant for all classes of voters.
In sum, one finding here is that when we examine the effect of international conflict activity on changes in domestic politics, what we probably are measuring is the effect of executive actions, and not necessarily the effect of an abstract international event -- a crisis -- on domestic politics. This is entirely consistent with the findings of Oneal and Bryan (1995), and we can claim some support for H1, but contingent to US presidential response in a crisis. It may not be the crisis per se that rallies the public, but executive action. Perceptions of competence are likely to be crucial in explaining the level of approval.
Table 2 Presidential Approval: Dispute Model
Variables | General Approval | Partisan Approval | Independents | Opposition |
Disputes | 0.05 | -0.08 | 0.27 | -0.11 |
(-0.31) | (-0.39) | (-0.09) | (-0.50) | |
Fatalities | 0.00 | -0.34* | -0.15 | -0.28 |
(-0.03) | (-1.77) | (-0.60) | (-1.31) | |
Hostility | 0.24 | 0.55*** | 0.28 | -0.02 |
(-1.39) | (-2.69) | (-1.01) | (-0.08) | |
Misery Index | -3.81*** | -2.20*** | -2.77*** | -4.03*** |
(-4.31) | (-2.75) | (-2.56) | (-3.59) | |
Expectations | 4.55*** | 3.50*** | 5.94*** | 6.20*** |
(-2.68) | (-2.73) | (-3.45) | (-2.82) | |
Eisenhower 2 | -20.43*** | -4.93 | -6.92 | -29.21*** |
(-3.64) | (-1.06) | (-1.11) | (-4.27) | |
Kennedy | -21.86*** | -11.77** | -9.64 | -42.05 |
(-2.95) | (-2.16) | (-1.32) | (-4.49) | |
Johnson | -26.40*** | -24.00*** | -22.70*** | -31.60*** |
(-3.45) | (-4.63) | (-3.30) | (-3.26) | |
Nixon 1 | -26.43*** | -8.03 | -10.90* | -29.29*** |
(-3.19) | (-1.5) | (-1.48) | (-2.75) | |
Nixon 2 | -52.88*** | -23.06*** | -29.90*** | -58.84 |
(-5.41) | (-3.51) | (-3.40) | (-4.79) | |
Ford | -10.15 | -6.5 | 5.28 | -7.32 |
(-1.06) | (-1.06) | (-0.61) | (-0.61) | |
Carter | -9.46 | -20.45*** | -7.08 | -11.06 |
(-1.07) | (-3.34) | (-0.86) | (-0.99) | |
Reagan 1 | 0.07 | 5.82 | 4.34 | -8.62 |
(-0.01) | (-0.93) | (-0.51) | (-0.74) | |
Reagan 2 | -10.14 | 2.25 | -2.9 | -20.87* |
(-1.03) | -0.39 | (-0.38) | (-1.70) | |
Bush | -31.19*** | -12.88** | -13.65* | -41.29*** |
(-2.82) | (-1.93) | (-1.52) | (-2.95) | |
Constant | 90.27 | 95.08 | 78.16 | 82.62 |
(-10.72) | (-16.1) | (-9.88) | (-7.72) | |
R _ ( Adj. R _ ) | .42/(.31) | .42/(.32) | .33/(.21) | .41/(.30) |
Durbin Watson | 2.09 | 2.16 | 2.1 | 2.19 |
Rho | 0.89 | 0.73 | 0.73 | 0.89 |
Residual d. f. | 333 | 339 | 339 | 339 |
We re-ran the models replacing the five CRISIS variables with the three MID-based variables. As expected, the economic indicators continue to perform well. However, MID involvement, FATALITIES and HOSTILITY generally do not produce significant coefficients. Looking first at MIDs, there appears to be absolutely no rally effect. This lends support to Brodys (1991) argument that a rally event must be dramatic in nature. Less intense situations, many of which are included in the MID data, are therefore more likely to be overlooked by the domestic public. FATALITIES produces one mildly significant coefficient for partisans in the anticipated direction. All other coefficients are in the expected direction, but are not significant.
An increased level of HOSTILITY has a slight positive and significant impact with respect to partisans. In other words, an increase in the level of hostility leads to a slight increase in presidential popularity. Coefficients are in the expected direction with respect to general approval and independents but are not significant. Predictably, greater hostility leads to a decrease in popularity in the case of opposition voters. However, not much should made of this since the coefficient is so weak. In sum, dispute involvement does not appear to have any tangible impact on domestic politics unless people begin to get killed. This also lends mild support to H1, a conclusion that should be tempered by the fact that the coefficients are small and in general not significant.
Table 3 Presidential Approval: Lagged Foreign Policy Crisis Model
Variables | General Approval | Partisan Approval | Independents | Opposition |
Crisis | -2.25* | -2.49* | -1.66 | -4.49** |
(-1.45) | (-1.50) | (-0.72) | (-2.05) | |
Major Region | 1.09 | 1.54 | -1.18 | 0.81 |
(-0.99) | (-1.31) | (-0.73) | (-0.52) | |
USSR Activity | 0.44* | 0.03 | 0.49 | 0.93** |
(-1.52) | (-0.09) | (-1.17) | (-2.34) | |
Casualties | 0.00 | -0.00** | -0.00* | 0.00 |
(-0.27) | (-2.27) | (-1.61) | (-0.99) | |
Force | 1.20** | 1.84*** | 1.10 | 1.61** |
(-2.16) | (-2.93) | (-1.27) | (-1.95) | |
Misery Index | -1.91** | -0.83 | -0.97 | -1.98** |
(-2.27) | (-1.14) | (-0.94) | (-1.98) | |
Expectations | 6.44*** | 5.09*** | 8.97*** | 6.62*** |
(-4.49) | (-4.96) | (-6.10) | (-4.59) | |
Constant | 76.86 | 95.77 | 76.24 | 62.52 |
(-15.55) | (-28.62) | (-16.03) | (-13.46) | |
R _ (Adj. R _ ) | .30/(.19) | .54(.46) | .38/(.27) | .37/(.26) |
Durbin Watson | 2.04 | 2.08 | 2.1 | 2.05 |
Rho | 0.82 | 0.64 | 0.66 | 0.67 |
Residual d. f. | 420 | 366 | 366 | 366 |
We re-ran both models with a one-month lag. Tables 3 and 4 report the findings using the CRISIS and MID based indicators, respectively. Predictably, if a crisis lasts longer than one month, support for a president begins to drop and dramatically --- in the case of opposition voters, i.e., 4,49%, suggesting that any rally effect is short-lived and that a president should begin to worry about domestic political standing during protracted crises. The other interesting finding concerns the use of force by a president. In this instance, coefficients are positive for all voters, with the coefficient for partisans being highly significant. Thus, as the use of force increases, so does presidential popularity. This stands in sharp contrast to the findings for the zero lag table, where FORCE produced negative coefficients. Speculation might suggest that the possibility of the use of force by a president initially scares public opinion, but once desensitized to its existence, the public rallies behind the president in his time of need. This conclusion is further supported by the findings for Soviet involvement. Once again, the coefficients are positive, suggesting a slightly delayed rally effect in the case of severe or intense crises.
Table 4 Presidential Approval: Lagged Disputes Model
Variables | General Approval | Partisan Approval | Independents | Opposition |
Disputes | 0.21 | 0.32* | 0.41 | 0.29 |
(-1.11) | (-1.48) | (-1.4) | (-1.02) | |
Fatalities | 0.39** | 0.53*** | 0.25 | 0.53** |
(-2.12) | (-2.58) | (-0.93) | (-2.02) | |
Hostility | -0.12 | -0.13 | -0.01 | 0.03 |
(-0.62) | (-0.61) | (-0.04) | (-0.10) | |
Misery Index | -2.28*** | -0.87 | -1.07 | -2.33** |
(-2.60) | (-1.10) | (-0.97) | (-2.23) | |
Expectations | 6.32*** | 4.91*** | 8.30*** | 6.12*** |
(-3.96) | (-4.21) | (-5.00) | (-3.89) | |
Constant | 80.78 | 98.82 | 81.46 | 65.55 |
(10.37) | (18.12) | (10.54) | (8.95) | |
R _ (Adj. R _ ) | .30/(.18) | .48/(.38) | .33/(.20) | .34/(.21) |
Durbin Watson | 2.07 | 2.1 | 2.08 | 2.06 |
Rho | 0.85 | 0.67 | 0.7 | 0.69 |
Residual d. f. | 368 | 333 | 333 | 333 |
Table 4 presents the lagged model of presidential approval for MIDs. Dispute involvement produces positive coefficients for all groups. However, the effect is extremely small, ranging from 0.21 to 0.41%. Thus, to conclude that MID involvement leads to an increase in presidential popularity after one month would not be warranted. Strangely, an increase in fatalities leads to a slight increase in presidential approval. This result appears consistent with the force-related findings form the lagged FPC model.
One final note with respect to the lagged economic variables seems warranted. The MISERY index produces coefficients in the expected direction for all groups, with the findings for the general public and independents being significant. In comparison, all the coefficients for EXPECTATIONS are strong and highly significant. (See Tables 3 and 4.) This strongly supports the notion that subjective measures of the state of the economy have a stronger and longer lasting impact on presidential approval.
To summarize the results, a small rally effect in the 3-to-4 percent range appears when the US is involved in a crisis. However, if force is used, then any rally effect disappears and the president may even suffer a loss at the polls if the use of force is great enough. Use of a different method and an extended data set confirms earlier findings that rally effects linked to foreign policy activity are small and linked to adversarial superpower involvement. This probably would come as disappointing news to presidents who expect otherwise (Gallup with Edwards 1990; Brace and Hinkley 1992; Oneal and Bryan 1995). Also, in some of the models, interesting differences in approval levels were detected among the four electoral sub-groups. We will not develop extensive explanations of any of the (slight) differences observed, but this study has shown that different parts of the US electorate respond differently to uses of force, and this should provide a point of departure for developing cross-national versions of this type of investigation, the lesson being that the public is not a unified actor.
Conclusions and Suggestions for Further Research
This study demonstrates that, after controlling for idiosyncratic factors, domestic economic conditions are stronger that the international events in explaining variation in US presidential approval. The potential rally events uncovered generally are mitigated by other factors, such as the location, level of violence and Soviet involvement. Furthermore, international rally events are short-lived. Therefore, US presidents cannot expect to boost their approval levels easily by becoming involved in a foreign crisis.
The superior performance of the ICB data over MID probably is due to differences in coding. In future research it will be interesting to probe these measurement issues. Future research also should focus on two other things. First, as noted in James and Rioux (1998), crossnational data should be collected to evaluate differences in public reactions to international rally events. A useful starting point would be to gather data from UK and France, two other democratic states that often have been involved in international conflict. It also would be useful to undertake more detailed case studies to discover which crises and disputes (if any) are associated with significant rally effects and why.
In sum, we are coming closer to identifying several key variables that influence, at least on a short-term basis, the popularity of US presidents. The publics perception of domestic economic conditions certainly ranks as one of the most important factors. With respect to international conflict, FPCs do not boost presidential popularity more than three or four percent in general and this is mitigated by the conditions enumerated above. Therefore, in this respect, if the dog is tempted to wag, it could end up losing its tail.
References
Bennett, D. Scott, and Allan C. Stam III. 1996. "The Duration of Interstate Wars, 1816-1985." American Political Science Review 90: 239-257.
Blechman, Barry, and Stephen Kaplan. 1978. Force Without War. Washington, DC: Brookings.
Bond, Jon R, Richard Fleisher and Michael Northrup. 1988 "Public opinion and Presidential Support". Annals of the American Academy of Political and Social Science 449: 47-63.
Brace, Paul and Barbara Hinckley. 1992. Follow the Leader: The President and
Public Opinion Polls. NY: Basic Books.
Brecher, Michael, and Jonathan Wilkenfeld. 1997. A Study of Crisis. Ann Arbor: University of Michigan Press.
Bond, Jon R., and Richard Fleisher. 1984. "Presidential Popularity and Congressional Voting: A Re-examination of Public Opinion as a Source of Influence in Congress." Western Political Quarterly 37: 291-306.
Brody, Richard A., 1991. Assesing the president: the Media, Elite, Opinion and Political Support. Stanford: Stanford University Press.
Bueno de Mesquita, Bruce, Randolph M. Siverson and Gary Woller. 1992. "War and the Fate of Regimes: A Comparative Analysis." American Political Science Review 86: 638-46.
Bueno de Mesquita, Bruce, and Randolph M. Siverson. 1995. "War and the Survival of Political Leaders: A Comparative Study of Regime Types and Political Accountability." American Political Science Review 89: 498-512.
Crew I, 1985. "How to Win a Landslide Without Really Trying: Why the conservatives won in 1983". In Britain at the Polls 1983, ed. A Ranney. New York: Duke University Press.
DeRouen, Jr., Karl R. 1995. "The Indirect Link: Politics, the Economy, and the Use of
Force." Journal of Conflict Resolution 39: 671-695.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: HarperCollins.
Dunleavy, Patrick and Cristopher T. Husbands. 1985. British Democracy at the Crossroads: Voting and Party Competition in the 1980s. London: George Allen & Unwin.
Edwards, George C., with Alec M. Gallup. 1990. Presidential Approval: A Sourcebook. Baltimore: The Johns Hopkins University Press.
Fleisher, Richard, and Jon R. Bond. 1988. "Are There Two Presidencies? Yes, But Only for Republicans." Journal of Politics 50: 747-767.
Fordham, Benjamin. 1998. "The Politics of Threat Perception and the Use of
Force: A Political Economy Model of U.S. Uses of Force, 1949-1994." International Studies Quarterly 42: 567-590.
Gallup, George C, Jr. 1953-1994. The Gallup Poll. (Yearly editions).
Goodhart, C. And R Bhansali. 1970. "Political Economy". Political Studies 18: 43-106.
Gourevitch, Peter. 1978. "The Second Image Reversed: The International Sources of Domestic Politics." International Organization 32: 881-911.
Hardy, Melissa. 1993. Regression With Dummy Variables. Newbury Park, CA: Sage.
Heldt, Birger. 1999. "Domestic Politics, Absolute Deprivation, and Militarization of Interstate Territorial Disputes, 1950-1990." Paper presented at the Annual Meeting of the International Studies Association, Washington, DC.
Hibbs, Douglas A, 1982. "The Dynamics of Political Support for American Presidents Among Occupational and Partisan Groups." American Journal of Political Science 26: 312-323.
James, Patrick. 1987. "Conflict and Cohesion: A Review of the Literature and Recommendations for Future Research." Cooperation and Conflict 22: 21-33.
James, Patrick, Eric Solberg and Murray Wolfson. 1999 "An Identified Systemic Model of the Democracy-Peace Nexus", Defence and Peace Economics 10: 1-37.
James, Patrick, and Sebastien Rioux. 1998. "International Cirses and Linkage Politics: The experience of US." Political Research Quarterly 51: 781-812
James, Patrick, and Athanasios Hristoulas. 1994. "Domestic Politics and Foreign Policy: Evaluating a Model of Crisis Activity for the United States." Journal of Politics 56: 327-348.
James, Patrick, and John R. Oneal. 1991. "The Influence of Domestic and International Politics on the President's Use of Force." Journal of Conflict Resolution 35: 307-332.
Kernell, S. 1978. "Explaining Presidential Popularity". American Political Science Review 72: 506-522.
Kernell, Samuel. 1986 Going Public: New Strategies of Presidential Leadership. Washington, DC: Congressional Quarterly, Inc.
Lamare, James W. 1991. "International Intervention and Public Support: America During the Reagan Years." In International Crisis and Domestic Politics: Major Political Conflicts in the 1980s, ed. James W. Lamare. New York: Praeger.
Levy, Jack S. 1989. "The Diversionary Theory of War: A Critique." In Handbook of War Studies, ed. Manus I. Midlarsky. Boston: Unwin Hyman.
Levy, Jack S. and, Lily I. Vakili. 1992. "Diversionary Action by Authoritarian Regimes: Argentina in the Falklands/Malvinas Case." In The Internationalization of Communal Strife , ed. Manus I. Midlarsky. London: Routledge.
Lewis-Beck, Michael S. 1980. Applied Regression. Newbury Park, CA: Sage.
Lewis-Beck, Michael S., and Tom W. Rice. 1982. "Presidential Popularity and Presidential Vote." Public Opinion Quarterly 46: 534-537.
MacKuen, M.B.1983. "Political Drama, Economic Conditions and the Dynamics of Presidential Popularity." American Journal of Political Science. 27: 167-192
Meernik, James, and Peter Waterman. 1996. "The Myth of the Diversionary Use of Force by American Presidents." Political Research Quarterly 49: 573-590.
Moore, Will H., and David J. Lanoue. 1998. "Do Executives Gamble for Resurrection? Evidence from Britain." Paper presented at the Annual Meeting of the Peace Science Society, New Brunswick, NJ.
Morgan, J. Clifton and Christopher J. Anderson. 1999. "Domestic Support and Diversionary External Conflict in Great Britain, 1950-1992." Journal of Politics. v 61, 3: 799-814.
Morgan, T. Clifton, and Kenneth Bickers. 1992. "Domestic Discontent and the External Use of Force." Journal of Conflict Resolution 36: 25-52.
Mueller, John E. 1994. Policy and Opinion in the Gulf War. Chicago: University of Chicago Press.
Mueller, John. 1973. War, Presidents and Public Opinion. New York: Wiley.
Norpoth, Helmut. 1987. "Guns and Butter and Government Popularity in Britain." American Political Science Review 81: 949-959.
Oneal, John R., and Anna Lillian Bryan. 1995. "The Rally 'Round the Flag Effect in U.S. Foreign Policy Crises, 1950-1985." Political Behavior 17: 379-401.
Ostrom, Charles W., and Dennis Simon. 1985. "Promise and Performance: A Dynamic Model of Presidential Popularity." American Political Science Review 79: 334-38.
Ostrom, Charles W., and Brian Job. 1986. "The President and the Political Use of Force." American Political Science Review 80: 541-566.
Prins, Brandon. 1999. "Systemic Threat, Domestic Politics and the Resolution of Interstate Conflicts." Paper presented at the Annual Meeting of the International Studies Association, Washington, DC.
Putnam, Robert D. 1988. "Diplomacy and Domestic Politics: The Logic of Two-Level Games." International Organization 42: 427-460.
Raber II, Donald R. 1998. "The Role of Economic and Political Factors in Presidential Uses of Force, 1948-1984." Paper presented at the Annual Meeting of the American Political Science Association.
Rivers, Douglas, and Nancy L. Rose. 1985. "Passing the President's Program: Public Opinion and Presidential Influence in Congress." American Journal of Political Science 29: 183-196.
Russett, Bruce M. 1990. "Economic Decline, Electoral Pressure, and the Initiation of Interstate Conflict." In Prisoners of War?: Nation-States in the Modern Era, eds. Charles S. Gochman and Alan Ned Sabrosky. Lexington: Lexington Books.
Sanders, David, et. al. 1987. "Government Popularity and the Falklands War: A Reassessment." British Journal of Political Science 17: 281-313.
Scott, James M., and Ralph G. Carter. 1999. "Acting on the Hill: Congressional Assertiveness in U.S. Foreign Policy." Paper presented at the Annual Meeting of the International Studies Association, Washington, DC.
Siverson, Randolph M. 1995. "Democracies and War Participation: In Defense of the Institutional Constraints Argument." European Journal of International Relations 1: 481-489.
Steger, Wayne P. 1999. "Pro-action, Reaction, and Inertia in Presidential Foreign Policy Choice." Paper presented at the Annual Meeting of the International Studies Association (Midwest).
U.S. Department of Commerce, Bureau of Economic Analysis. URL //www.stat-usa.gov/stat-usa.html, January 21-25, 1996.
Endnotes
Note 1: Examples of studies with focus, broadly conceived, include Ostrom and Simon (1985), Norpoth (1987), Putnam (1988), Bueno de Mesquita and Silverson (1995), Oneal and Bryan (1995) and James, Solberg and Wolfson (1999). Back.
Note 2: A related study by Steger (1999) tracks presidential foreign policy decision-making in response to factors ranging from prior conduct (i.e., inertia) to actions by other states. Back.
Note 3: Meernik and Waterman (1996: 574-577) argue against the use of time periods rather than events as the unit of analysis; for a response see James and Rioux (1998). Back.
Note 4: These results raise an obvious question: If neither partisan, independent, nor general approval appear to affect presidential succes in governing, why study the determinants of popularity at all? For justification of this line of inquiry see Rivers and Rose (1985: 193) and Lewis-Beck and Rice (1982: 535). Back.
Note 5: Brace and Hinckley (1992) provide a complete discussion of this observed phenomenon; because the decay is not monotonic, a "counter variable" method is not used. Back.
Note 6: The data for this variable have been gathered from the Bureau of Labor Statistics and the Gallup Organization on a monthly basis from January 1953 to December 1994. The operationalization of the Misery Index is validated by another macroeconomic fact: unemployment and inflation tend to work against each other. Low unemployment carries inflationary tendencies and high unemployment is deflationary, so it is the joint effect that produces the public's 'misery'. Back.
Note 7: When regressed on each other, the three variables do not produce evidence of multicollinearity with R-squares ranging form 0.13 to 0.62. See Lewis-Beck (1980) for an explanation of the procedure used to assess multicollinearity. Back.
Note 8: See James and Rioux (1998) on approaches toward the problem of multicollinearity in the context of this research design. Back.