Abstract
This paper examines the direct, short-term effect of presidential communications on policy preferences. Using panel studies and post-speech surveys, I demonstrate that following a speech, public opinion changes in the direction of the president. This effect is strongest among people who watch the president. While the number of people who tune to the president may be small, indeed getting smaller over time, it is composed of people who participate in the political process, are more likely to affect it, and therefore are of interest to opinion leaders. The findings reveal that presidents are effective leaders of public opinion.
On October 7, 2002, President George W. Bush addressed the nation on Iraq. In his address, the President discussed the threat from Saddam Hussein to world peace, and made the case for disarming Iraq from weapons of mass destruction and removing Hussein from power. President Bush notified the American public that he “asked Congress to authorize the use of America’s military, if it proves necessary, to enforce U.N. Security Council demands.” According to a Fox News poll the next day, nearly half of Americans watched this nationally televised address. Of those who watched the speech, 69 percent thought President Bush made a convincing case that action must be taken against Iraq. The next morning, members of the House began debating the resolution proposed by the President. On October 10, the resolution passed with impressive majorities in both the House (296 to 133) and Senate (77 to 23). 1
A series of surveys reveals that the public responded favorably to the message of the president by increasing its support for a military action in Iraq. This support, however, was short-lived. Following the speech, public support for taking military action in Iraq to end Saddam Hussein’s rule went up from 62 to 72 percent, but a few days later support went down to 66 percent and by the end of that month it receded back to where it was prior to the speech (62 percent). 2
The speech on Iraq is one of many public addresses President Bush held during his two terms in office and is typical to presidential action. Presidents “go public” often to address the nation on issues they wish to promote. They schedule well-advertised national addresses, hold press conferences, and release numerous public statements. Despite these efforts, extensive research shows that the effect of presidents on policy preferences is limited (see especially Edwards 2003, 2009; Rottinghaus 2010). Presidents face many barriers to opinion leadership, including the shrinking and inattentive audiences, and the hyperpolarized political environment that affects both the media’s news coverage and its audience (Baum and Kernell 1999; Cohen 2008, 2010; Edwards 2003, 2009; Kernell and Rice 2011; Wattenberg 2004; Welch 2000, 2003). But given what we know about the positive effect of elite correspondence on public opinion (Baum and Groeling 2010; Brader 2006; Geer 2006; Vavreck 2009; Zaller 1992), we should expect to see stronger leadership of public opinion from the president—the most dominant actor in American politics.
The problem with most existing work is that it examines long-term changes in public opinion, mainly by measuring overall trends in public opinion or the public response long after a message was delivered. Considering the dynamics of the political environment, however, a long-term effect would be an extraordinary event, which if happens, may have more to do with the intervention of additional actors and voices over the interim between the speech and the survey. Instead of a long-term effect, I suggest that when presidents appeal to the American public, they have a temporary, short-term effect on public opinion. Such effect can serve the president by fending off existing or anticipated political opposition, generating a momentum of public opinion, and targeting specific congressional deliberations or key legislators.
Following this premise, I examine the immediate effect of a speech on public attitudes. The data include surveys run shortly after major national addresses that ask respondents for their views about the president’s message. Using these data, I estimate several multinomial logistic regressions of individual responses and a difference-in-differences model of panel data to assess opinion change. The findings reveal that Americans respond positively to the presidents’ messages and that, holding all else equal, this effect is greatest among people who are attentive to the messages. While the attenaudience may be small, I contend that it is likely to be the focus of the administrations’ interest because this audience is more involved in the political process.
The Direct Effect of “Going Public”
Presidents devote much of their time and effort to their public relations. They make routine public remarks, issue well-crafted national addresses, and appear on radio and television host-shows. Samuel Kernell (1997) proposed that these actions represent the core strategy of governing for modern presidents: Presidents go public to appeal to the root of the representative public; they change public opinion; and that change creates a political gain for the president.
The effect of these speeches on policy positions of Americans, however, is a matter of an ongoing empirical debate. 3 A large body of work focuses on specific case studies of leadership success (Bailey, Sigelman, and Wilcox 2003; Foyle 2004; Kernell 1997; Koch 1998; Rosenblatt 1998; Sloan 1999) and failure (Finbow 1998; Jacobs and Shapiro 1995). More systematic studies of presidential public leadership of policy positions are few and are generally pessimistic about the ability of presidents to affect policy positions. Edwards (2003, 2009) examines public opinion about numerous issues promoted by Presidents Reagan and Clinton, the “Great Communicators,” and finds very little evidence of overall long-term movement in direction of the presidents’ policy positions. The rich data, the systematic analysis and the strength of his conclusion seriously question the incentive in presidential leadership efforts. As Edwards strongly demonstrates, in the long term, presidents have a very limited effect on policy preferences of Americans. Why then do presidents continuously “go public” if it offers them little political gain?
Although their command of the bully pulpit offers presidents a unique position to reach the American public and affect it (Kernell 1997), their efforts do not take place in a political vacuum. Presidential messages occur within a heated political environment where elite messages respond to or initiate public debate that is aligned with existing political divisions (Cameron and Park 2011; Cohen 2010; Skinner 2009). In such an environment, a president occupies one side in a multisided information flow, and the effect of his message depends on the strength of his words, the relative intensity of the opposing message, the extent each side is covered by the media, and how both messages align with existing partisan or ideological divisions (Zaller 1992). In the short term, presidents command the public discourse, draw mass attention, frame the issues in their own terms, and may have a significant effect on public opinion. In the long term, however, the issue positions of the presidents rarely withstand their political opposition. As a consequence of the political debate, the balance of opinion may end up very close to where it was prior to presidential promotion, but this does not necessarily imply that presidents are not effective leaders of public opinion.
Given the stickiness of public opinion, a temporary shift in public opinion is perhaps what presidents can hope for (Page and Shapiro 1992; Rottinghaus 2010). Such effect achieves several goals. First, by communicating their messages, presidents engage in the political debate, increase exposure to their message, and fend off opposing messages or minimize their effects. Presidents take advantage of their position at center stage to address the public about their policy position and increase interest in it (Cohen 1995). They can move first, present a policy position and anticipate a response from the opposition, or they can respond to existing messages. A temporary change in their favor will force others to compete for scarce resources to reach the public and respond to their messages. Even if support for the president’s policy position will not sustain in the long term, the rhetoric of the president affects the public discourse and the level of support. Leaving the issue completely unattended may also have dire consequences of losing control over the issue.
Second, by communicating their messages to the entire public and momentarily increasing public support for their messages, presidents generate a momentum of public support that may have a long-term effect on the structure of mass opinion. Previous research indicates that people tend to support an opinion they believe to be a majority opinion, especially if it is viewed as a winning trend where public support is gaining momentum (Joslyn 1997; Mutz 1998). This effect is moderated by political predisposition, knowledge, and interest. A temporary shift in support can strengthen the position of people sharing the same political commitments with the president and increase the likelihood that people with no prior commitments and low levels of knowledge or interest would side with the president’s message (Joslyn 1997). A temporary change in public opinion can also influence the positions of people who have strong political commitments and high level of knowledge and interest by promoting them mentally to rehearse arguments related to the issue position (Mutz 1998). Such individuals may not increase support for an issue, but their position may crystallize and their commitment to it would increase.
Third, a short-term effect also follows the theoretical view of “going public” as a tool to pressure congressional deliberations, swing key legislative votes, and gain public attention (Kernell 1997). Presidents use their position as the bully pulpit of American politics to reach out to the public hoping to affect current views with the goal of gaining congressional support. Even if the impact of their speeches on public opinion is temporary, it may prove helpful in promoting their policy in Congress (Canes-Wrone 2006; Ragsdale 1984; Rottinghaus 2010). This effect is illustrated in the speech of President Bush about the need to support a military action in Iraq (October 7, 2002), carefully scheduled before congressional deliberations began. As the President alluded to, he appealed to the American public but the goal was Congress. Following the speech, the public responded positively to the speech, increasing its support by ten percentage points. This increase soon receded—indicating no long-term effect of the presidential message—yet the goal of presidential promotion was probably met by the successful bipartisan vote in Congress, which in part was fueled by the temporary support of nearly three out of every four Americans (Foyle 2004).
Focusing on the short-term effect is also methodologically proper because it clearly examines the causal mechanism of presidential public leadership—measuring the action and the public response to it. Public opinion is influenced by domestic and foreign events as well as the dynamics of political circumstances (Page and Shapiro 1992). In order to evaluate correctly how the public responses to a message, we need to minimize the lag between the speech and our measured response—that is, between a presidential speech and the public response to it.
Noting the importance of the short-term effect in analyzing the effect of presidential public statements, Rottinghaus (2010) compares public opinion before and after presidential messages. He demonstrates how countervailing influences in the political process affect the ability of presidents to lead public opinion. By focusing on the short-term effect of presidential messages and by expertly taking into account the range of environmental factors that may influence their effect, Rottinghaus significantly advances our understanding of the relationship between the president, public opinion and the political environment. Yet, in collecting his data, Rottinghaus allows a long lag (up to two years) between the pre- and postsurveys, and thus fails to appreciate fully the short-term effect of presidential rhetoric on public opinion. 4 A long prespeech lag can capture the effect of the events leading to the increase in the public agenda and the consequential presidential interest. A long postspeech lag may capture the effect of the subsequent increased public interest and the political response to the president’s statement. The reciprocal relationship between the attention to policy issue of the mass public and the president can further affect measurements of the effect of presidential activity on public opinion (Hill 1998).
Other studies examine the direct effect of messages of the presidents on public opinion, but do not measure policy preferences. Edwards (2003) compares the change in presidential approval ratings following all national televised addresses of presidents Ronald Reagan to George W. Bush. He finds that presidents rarely succeed in garnering support for themselves. Change in presidential approval however is different from change in policy positions. Considering the deterministic role of political attachments on approval ratings (Erikson et al. 2002), especially at times of strong political polarization, and the range of policies associated with the president, it is unlikely that a president would be able to garner support for themselves by issuing a single speech on one policy. Presidents may, however, be more successful in influencing policy preferences by speaking about the issue and discussing possible alternatives.
Wood (2007) examines the effect of presidential economic rhetoric on public evaluations of the economy and finds persuasive evidence of presidential leadership. Although Wood does not compare attitudes before and after each speech, his monthly aggregated time series allows him to draw strong conclusions about the short-term effect of presidential rhetoric, indicating that presidents are effective short-term movers of public evaluations of the economy. Affecting public evaluations of the economy, however, is different than affecting policy preferences. Americans seek assurance in the president’s handling of the economy, and optimistic accounts of the economy may increase their confidence that the economy is strong. But it does not necessarily mean that they approve of the president’s message.
In sum, the presidents’ key position in government and public affairs gives them an unmatched weight on public opinion. Presidents are powerful leaders of opinion and their messages register strongly in mass opinion, even if the effect is only short-lived. I expect that a systematic analysis of the immediate short-term change of public opinion following presidential messages would reveal a strong effect of presidential addresses on policy positions of the public.
Hypothesis 1: Immediately following a presidential address, public support for the presidents’ policy will increase.
Critics of the “going public” theory point to the decreasing number of Americans who tune to presidential addresses. If only a few listen to the president, how can a president affect public opinion? Indeed, presidents today lack the public attention their predecessors enjoyed during the “golden age” of public television (Baum and Kernell 1999). This change, however, is not unique to presidents. Over the past several decades, the amount of coverage in the traditional news media devoted to public affairs has declined, but presidents still occupy a key position in the news—certainly more than any other political figure (Cohen 2008, 2010; Welch 2003).
Even if presidents reach fewer people today, the audience that listens may still be the audience of interest to presidents–the more active and politically savvy public. During the second half of the twentieth century, Americans have significantly decreased their level of political participation: they go out less to vote, they take part in fewer political activities, and they are less willing to identify with a specific party (Putnam 2000; Zukin et al. 2006). Those who participate, however, have an interest in the political debate and usually tune to presidents when they speak (Welch 2003).
The size of the attentive audience is often very small, and therefore an attitudinal change of this particular group would have a limited effect on public opinion. Yet the attentive audience, rather than the entire public, may be “the root of the representative public,” which elites—and presidents among them—are interested in persuading because of their potential effect on the political process. Put simply, presidents may seldom move the median voter, but they may move attitudes of those who matter most–those who vote and participate in the political process.
Most existing studies of presidential leadership of public opinion make use of general surveys that do not separate between people who watch and people who do not watch the address. Their findings—or nonfindings—are therefore influenced by a significant number of respondents who have no knowledge of the content of the presidents’ speech or of the political debate it engages with. The attitudes of these people may be random or, more problematic, reflect their political predis-positions rather than the effect of the message of the president (Zaller 1992). To overcome this problem, several studies instrument exposure to the presidents’ addresses using experimental research design (Druckman and Holmes 2004; Tedin, Rottinghaus, and Rodgers 2010). Interestingly, these studies are optimistic of the presidents’ ability to influence public opinion, suggesting that people who watch presidential addresses are influenced by them. Applying this to survey analysis, I propose to evaluate the effect presidents have on the attentive audience—those who follow the political debate and watch their addresses—and compare their attitudes to the non-attentive audience.
Affecting the attentive audience may prove difficult because this audience is more informed about politics, and hence is less likely to change attitudes according to incoming information. Yet, there is a large variation among the attentive audience in the strength of their established views. According to Zaller (1992), people who are mildly informed are most likely to change opinion—they are exposed to new considerations but do not develop strong resistance to them. The ability to resist presidential messages, as well as the tendency to be exposed to these messages, is further mediated by levels of education, political commitments and support for the president (Delli Carpini and Keeter 1996; Kernell and Rice 2011; Zaller 1992). After controlling for the special characteristics of the attentive audience, I expect that exposure to a message would increase support for the proposed policy, especially in the short term where alternative views are still not readily available.
Hypothesis 2: Following an address, the attentive audience will be more supportive of the president’s policy than the non-attentive audience.
By examining how a speech affects the public opinion in the short term, and how it affects differently those who watch and those who do not watch an address, I reveal the role presidents play in shaping public opinion. Mainly, I demonstrate that presidents have a short-term effect on public opinion, which is stronger among the attentive, politically involved audience.
Data and Method
To assess the direct effect of a speech on public opinion, I examine surveys that immediately follow a speech and ask respondents whether they watched the address or not. Using the Roper Public Opinion Archive, I searched for surveys from 1981 to 2010 (Presidents Reagan through Obama) that refer to the criteria described: “president,” “watch,” and “speech.” To ensure a policy focus, I included in the data only surveys that asked a policy question relating to the speech. Surveys referring to State of the Union Addresses were not included because these speeches usually lack one clear policy focus (Tedin, Rottinghaus, and Rodgers 2010) and are immediately followed by an institutionalized televised opposition rendering their immediate polarized response. The search resulted in fourteen surveys.
Table 1 lists the surveys included in the data. Six of the surveys are part of panel studies that interview people before a speech and reinterview them shortly after the speech. 5 These data are of particular interest because they can be used for individual-level analysis of attitude change. The eight additional surveys are postspeech surveys that examine the public response to an address of the president. Two of the surveys refer to one speech—Clinton’s 1993 economic address.
Surveys Included in the Data.
Source: iPoll. Search terms: “president,” “watch,” and “speech.”
Together, these surveys correspond to thirteen speeches delivered by five different presidents covering both foreign and domestic policies. Because of the selection process that depends on the availability of surveys rather than the content or type of the speeches, all speeches are major national addresses regarding prominent policy initiatives. These speeches addressed events and policies that the majority of Americans were aware of and opinionated about. They drew strong public interest, and they were likely to have been viewed by a relatively large audience, hence the interest of pollsters in them. Indeed, on average, only 13 percent of respondents in the sample were not able to give a substantive answer to the policy question—a significantly small share of the public.
Focusing on such high-salience speeches would inevitably affect the results, but not necessarily in a predictable way. On one hand, speeches about such major issues are more likely to affect public opinion because Americans expect their president to lead the country on these issues, and therefore take careful note of his position and rationale (Rottinghaus 2010). On the other hand, most Americans are informed about these issues and are likely to already have a clear preference about them. Changing attitudes about such issues is therefore limited (Peake 2001).
Notwithstanding a possible—yet uncertain—bias, our focus should be on these major addresses. Presidents may or may not affect public opinion when addressing a low-salience issue, or when talking about an issue in a more limited event. However, because presidents view their major addresses as tools for reaching out to Americans and signaling a new policy direction (Kumar 2007), the ability of presidents to affect public opinion with these dominant speeches is paramount.
Each survey includes several questions referring to policy positions mentioned in the president’s address. Among the available questions, I used the question that best represented the policy advocated by the president. When choosing the question to analyze in each panel study, I preferred questions that are identical in the pre- and postsurveys. 6
Using the data from the fourteen surveys, I first examine the short-term effect of a speech on public opinion. Taking advantage of the panel studies, I estimate a multinomial logistic regression to assess who changes his or her opinion after a speech and in what direction. Second, I examine how exposure to a speech affects individuals’ support for the advocated policy. I use the panel data to estimate a difference-in-differences model treating exposure to the speech as the treatment and holding constant the non-time-variant variables. I then expand the analysis to additional surveys and estimate a logistic regression model of support for the president’s policy including exposure to the address as a key explanatory variable.
Short-Term Effect of Presidential Speeches
Figure 1 compares the support for the presidents’ policies before and after the six speeches for which panel data are available. Following all but one speech, support for the presidents’ advocated policies has increased. On average, before a president took the stand, 47 percent of respondents supported the policy of the president. After the presidents’ addresses, support climbed to 57 percent.
Increased support following major speeches.
Taking advantage of the panel data, I examine how people change their opinion: direction of change and characteristics of people who change opinion. Each respondent is coded with respect to the direction of opinion change: change from “oppose” to “don’t know” or to “support,” change from “don’t know” to “oppose” or to “support,” change from “support” to “oppose” or to “don’t know,” and no change of opinion. Table 2 summarizes the results.
Attitude Change Following Six Presidential Addresses.
Note: Data were taken from six panel studies by Gallup and CBS. Each column may not reach 100 due to rounding.
The first column indicates that most Americans (74 percent) did not change their preferences following the presidents’ messages. On average, following the six speeches, 18 percent of Americans changed their opinion in the direction of the president’s position. Twelve percent changed from opposing the policy to supporting it; 4 percent changed from “don’t know” to “support”; and 2 percent changed from “oppose” to “don’t know.” In contrast, only 7 percent of Americans changed their opinion against the position of the presidents. Four percent of respondents who supported the policy before the speech opposed it after the speech, 1 percent changed from “don’t know” to “oppose,” and 2 percent changed from “support” to “don’t know.”
Columns 2 through 7 summarize the attitude change following each of the six speeches. The strength of public response varies across speeches. Following all but one speech, the change in support of the president’s position was substantively larger than the change in the other direction. The most significant change is following Clinton’s economic plan in February of 1993, where 25 percent of respondents changed their opinion in the direction of the president, and 6 percent changed in opposition to it.
George W. Bush’s proposal for economic reform during the first days of his presidency also registered well in public opinion. Relatively few (19 percent) Americans changed their preference in response to the president’s call for tax cuts. Yet most of those who did change their view changed in the direction of the president. Fifteen percent of Americans who previously did not have a clear position on this issue or opposed the proposed tax cuts were now supportive of the measure. In contrast, only 4 percent of respondents who previously supported the policy or did not know changed their attitudes to oppose the tax cuts. The most stable attitudes are recorded following President Bush’s address on invading Iraq. Eighty-four percent of respondents did not change their opinion about invasion. Yet the speech came after a long public campaign and it was used to announce the invasion, rather than gain support for it (Foyle 2004).
To examine rigorously the effect of speeches on public opinion, I estimate the determinants of change in support for the president’s policy at the individual level. The dependent variable takes 1 for every individual who changed her opinion to side with the president: if opposed the president’s policy or did not know in the presurvey but supported it in the postsurvey, and if opposed the policy in the presurvey but did not know in the postsurvey. Respondents who did not change their preference are coded 0. Respondents who changed their opinion against the preference of the president or away from supporting it are coded negative one (–1). 7 Because the dependent variable—change of opinion—can take three nonordered values, I estimate a multinomial logit and treat all people who did not change their opinion (coded 0) as the base category. 8
For explanatory factors, I include conventional demographic and political factors—education, race, age, gender, presidential approval, and party identification. All are coded from the presurvey. 9 In addition, I include indicators for each survey to control for particular differences between them (leaving the 2007 speech about Iraq—announcing the surge—as a reference). The results are summarized in Table 3.
Determinants of Change of Policy Preferences Following Presidential Address.
Note: Data are from six panel studies. N = 2,585. Reference category: no change. Standard errors are in parentheses.
p < 0.1, *p < 0.05, **p < 0.01, ***p < 0.001.
Columns 1 and 2 of Table 3 report the coefficient estimates for changing opinion to oppose and to support the president’s position respectively. Approval of the president’s performance in office has a negative effect on policy change—it decreases the odds of changing in opposition of the president but does not increase the odds of changing in support of the president. Identifying with the party of the president has no independent effect on either change in support or in opposition of the president. Independents are less likely to move in the direction of the president.
Consistent with our understanding of political persuasion (Zaller 1992), less educated—high-school dropouts—are more likely to respond to the address and change their views in support of the president compared to people holding a high school diploma (the reference group). Americans who have some higher education, a college degree or more, are less likely to respond favorably. Elderly respondents (above 65) are likely to respond to the president’s message. The coefficient for this group is positive and significant for both outcomes, indicating that they strongly respond to political rhetoric but not necessarily in the direction of the persuasion efforts of the president. In fact, the marginal effect of this variable on both outcomes (
In sum, presidents have a direct effect on public opinion. When they publicly advocate a policy, the public responds with an increased support (hypothesis 1). Consistent with our understanding of public opinion change, this movement of opinion is strongest among the less educated, who are more likely to be influenced by presidential addresses, and elderly, who are more attentive to the political debate. 10 Political predispositions are marginally effective in limiting the opposition to the president’s position, but do not mediate support for it. This, however, may be a consequence of the focus of this analysis on attitude-change rather than policy positions. Arguably, Americans who approve the president and identify with his party are less likely to change their opinion because they already share the same ideological commitments as the president and hence support the policy regardless of the speech. Similarly, people who do not approve of the president and identify with the opposition party are ideologically against the proposed policy. A speech, therefore, only reinforces their opposition. To examine this, I turn to the effect of a speech on the attentive audience—that is, those who watch the address.
The Attentive Audience
Americans who watch a president’s speech are more supportive of the president’s policy than those who do not watch the speech. Table 4 summarizes the overall public opinion about the issue, the opinion of those who watched the speech, and those who did not watch it, and the gap between these two groups. 11 Following several of these speeches, the public was still not willing to support the president’s policy. For example, following the speech of President H. W. Bush’s on the budget agreement, only 36 percent of Americans were supportive of the deficit reduction plan. Similarly, following both of President Clinton’s speeches—on the economic program and the U.S. involvement in Haiti—less than 50 percent of Americans were convinced that the president’s advocated policy was a good one. But, when the attitudes of those who watched and those who did not watch the speech are compared, a clear positive gap is found in all of these speeches. Following the twelve televised addresses, the attentive audience was more supportive of the president’s position than the people who did not watch the speech. In several instances, the difference was by a margin of more than 20 points.
Support for the President’s Policy among Watchers and Nonwatchers.
Those who watch the address, however, may be predisposed to support the president’s message—that is, they watch because they support the message of the president. To unravel the independent effect of watching the presidents’ addresses on support for their policy, I control for competing explanations. First I take advantage of the panel data and estimate a difference-in-differences model that assesses the effect of a treatment—watching the address—while holding all else constant. The difference-in-differences model is summarized in the following equation:
Where δ1 is the individual’s support at the postsurvey, δ2 is the treatment—watched or did not watch the speech, and δ3 is the interaction between the two variables and therefore is the Difference-in-Differences (DD) estimator—support among people who watched the president.
To estimate this model, I use four panel studies that include the attitudes of those who watched the address and those who did not. 12 Errors are clustered by individual respondents. Summarized in Table 5, the coefficients of the difference-in-differences model offer strong support for the second hypothesis that the attentive audience is likely to support the president’s message. 13 Holding all else equal, people who watch an address of the president are nearly 5 percent more likely to support the president’s policy.
The Independent Effect of Watching the President’s Messages (Difference-in-Differences Model).
Note: Standard errors are in parentheses
p< 0.1, *p < 0.05, **p < 0.01, ***p < 0.001.
Second, I expand the scope of this analysis and examine the effect of watching an address on the public response using the more broad set of surveys that follow a presidential address and ask respondents whether they watched the speech or not. To estimate the independent effect of watching an address on public opinion, I analyze eight of the twelve surveys summarized in Table 4: Reagan on economy and Grenada, H. W. Bush on drug control and budget agreement, Clinton on Haiti, W. Bush on Iraq (invasion, March 2003) and immigration, and Obama on health care. Two surveys are dropped because they lack sufficient individual-level data: Clinton on economic program and W. Bush on stem cell research. The two surveys in 2007 asking respondents about Iraq are dropped to avoid overemphasizing the president–public relationship of President George W. Bush or of the war in Iraq. 14
The dependent variable is siding with the president’s position. In the eight surveys included in the data, a total of 3,502 sided with the policy advocated by the president, 1,720 opposed it, and 780 reported that they did not know. Siding with the president’s position is coded 1 when supporting the policy advocated by the president, and 0 if opposed it or did not know. 15
The variable of interest is whether respondents watched the speech or not. Overall, 49 percent of respondents watched the presidents’ speeches. I code watching the speech—even a short segment of it—as 1, and 0 otherwise. 16
Several independent variables that are expected to affect the probability that one would support president’s message are included. First, because people are likely to tune to views that are consistent with their own beliefs (O’Keefe 2002), respondents who have a positive view of the president or are affiliated with his party are more likely to watch the address and support his proposals. 17
Almost 80 percent of respondents who approved the president’s performance as president supported his policy position. In contrast, nearly 70 percent of respondents who did not approve of the president’s job opposed the advocated policy. I expect that this correlation will translate into a strong independent effect of presidential approval on policy preferences. Identification with the president’s party offers additional control for the proximity of the president’s policy preferences with those of the respondent.
I therefore include indicators for presidential approval and party affiliation. Approval is coded 1 when a person approves the president, and 0 otherwise. For party attachments I include an indicator for identifying with the president’s party (1 for identifying with the same party; 0 otherwise) and an indicator for identifying with no party (independents; 1 for independents, and 0 otherwise). 18
Identification with the president’s party offers significant leverage to control for the proximity of the president’s policy preferences with those of the respondent. Hence, I control for party attachments by including an indicator for identifying with the president’s party (1 for identifying with the same party; 0 otherwise) and an indicator for identifying with no party (independents; 1 for independents, and 0 otherwise). 18
Second, because political participation enhances a person’s awareness and interest in public affairs, it is expected that politically active individuals would be more likely to watch the president speak about pressing issues and be opinionated about them (Welch 2000). Unfortunately, the surveys used in this analysis offer very limited information about political participation. In only four of the surveys, respondents were asked if they voted in the previous elections, and even fewer surveys offer any additional information. I, therefore, estimate two models: one with an indicator for voting in the previous elections, and one without.
Third, political knowledge can have a strong effect on policy attitudes and response to presidential addresses. Knowledge influences the extent of persuasion effect (Zaller 1992) and the strength of the policy position an individual holds. While the surveys do not offer instruments to assess political knowledge, research has shown that education can be used as a proxy for political knowledge (Price and Zaller 1993). Less educated individuals are less knowledgeable about political issues. As a consequence, the presidents’ message is likely to have a positive effect on the policy preferences of those individuals. Including education in the model also controls for the possible relationship between watching an address and political knowledge. People who are attentive to speeches of presidents are more likely to be politically informed and knowledgeable about policies compared to a nonattentive audience. By controlling for the influence of education on opinion change, I examine the independent effect of watching the address regardless of education levels.
In addition, I include indicators for the effect of age, race, and gender—all are coded as binary group indicators leaving one group for reference. To control for the possible differences between surveys, I include indicators for each survey, leaving one survey—referring to President Bush’s address on Iraq in March 2003—for reference.
Table 6 summarizes the point estimates generated by the general (column 1) and participation (column 3) models. Taking advantage of the categorical independent variable and the tractable form of the logit model, I calculate the odds ratios of the parameters in the model. For a unit change in each parameter, the odds of support for the president’s position are changed by the factor listed (Long 1997, 79-82). The odds ratios generated by the model for each significant parameter are summarized in columns 2 and 4.
Estimating the Controlled Effect of Exposure to the President’s Message.
Note: Standard errors are in parentheses.
includes only four surveys that asked respondents if they voted in the previous election.
p < 0.1, *p < 0.05, **p < 0.01, ***p < 0.001
People who approve of the president’s handling of his job support the president’s policy. For them, the president’s address rationalizes the positions they already have and gives them an argument they can use when discussing them with their friends. After accounting for this dominant predictor, party affiliation still plays an important and influential role in predicting support for the president’s policy. Americans who are affiliated with the president’s party are 1.78 times more likely to support the policy advocated by the president. The effect of independents is smaller yet significant. An independent is 1.24 more likely to support the president’s advocated policy compared to a person who identifies with the opposition party.
The participation model indicates that voting is not a significant predictor for support of the president’s policy. While the effects of the political variables are elevated in this model, the model is not substantially different from the general model.
Figure 2 illustrates the difference between people who approve and people who do not approve of the president by plotting the probability distributions of these two groups as generated by the general model. The lines represent the density of the population in each probability level. People with a probability between 0 and .5 are likely to oppose the president’s policy; people with a probability between .5 and 1 are likely to support the president’s policy. The closer an individual is to either extreme the more likely he or she will oppose (0) or support (1) the policy.
Probability of supporting for the president’s policy by presidential approval. Source: Probability distribution function based on the full model in Table 6.
The dark line represents the entire population, indicating some bias in favor of supporting the presidents’ policies. The light line indicates that people who approve of a president’s job are more likely to support the policies advocated by this president. The mean support of this group is .77. The gray line demonstrates that there is less homogeneity–more scattered–among people who do not approve the president’s job performance. Their mean of support is. 33
After accounting for political and demographic differences, watching the address is still a significant predictor of siding with the president’s policy. Americans who watch the address are 1.83 times more likely to support the policy advocated by the president. Figure 3 illustrates this point by breaking down the support for the advocated policy by approval of the president and by watching the address. The effect of watching the address is a significant element among people who approve and among people who do not approve of the president. That is, watching the speech positively affects even people who are predisposed to disagree with the president.
Probability of supporting for the president’s policy by presidential approval and watching the president.
The findings demonstrate that by talking to the nation, presidents garner public support for their policy among people who watch their addresses. While this group is relatively small—averaging less than 50 percent of the American public—its importance should not be underestimated. People who watch presidential addresses are also the most active group in the American process: they vote, attend rallies, write to their members of Congress, and are likely to contribute to campaigns. In the four surveys that ask respondents if they voted or not, 40 percent of respondents who did not vote and did not watch the address supported the president’s policy. In contrast, nearly 70 percent of respondents who voted and watched the address supported the advocated policy. The ability to successfully affect the highly attentive and participatory public may be what presidents are hoping for.
Conclusion
Presidents today consistently take a stand and address their policy concerns and policy initiatives to the American people. They use a variety of media resources, they travel outside of Washington, they repeat their messages, and they talk to groups of different sizes and in different settings. Existing studies of presidential leadership of public opinion examine long-term effect of presidential rhetoric and find that presidents have a very limited—if any—effect on the public’s policy positions. Focusing on the short-term effect of presidential communications, I demonstrate that following an address of the president, public opinion moves in favor of the president’s advocated policy, an effect that is strongest among the attentive audience.
The support gained by the speech may quickly decline as the public is exposed to policy alternatives and to critics of the president and his policy preferences (Cohen 2008). Yet any evaluation of political persuasion needs to first measure the effect of the persuasion tool—the public address. By comparing the attitudes of the same people before and after a speech, the data reveal that the tool of public address does not fall on deaf ears. In fact, more often than not, presidents succeed in using this tool to bolster support for their policies. Opinion leaders who oppose the president’s policy now need to use their toolbox to challenge this increase in public support. To some extent, they will succeed. But this is the political game, and researchers need to acknowledge that by assessing first the direct effect, and only then the dynamics of its decline. The underlying conditions to any political persuasion are that people are tuned to the message, and that they respond to it by following the president’s appeal. The evidence here suggests that presidents succeed in doing just that.
Indeed, much can be changed after the speech. President can influence the public agenda, they can present a clear policy preference, but they, thankfully, cannot command the political debate. Once their message is known, it is scrutinized by the media and political rivals (Cameron and Park 2011; Cohen 2008), and these inevitably influence public opinion. The sustainability of the attitude change would, therefore, depend on the persuasiveness of the president’s appeal and on the strength of the opposition. For example, the effect of President Bush’s campaign to persuade the public to support a ground operation in Iraq is hardly comparable to the effect of the campaigns of Presidents Clinton and Obama to reform health care. While the strength of the presidents’ campaigns may have been the same, the opposition to Iraq has been almost a whisper compared to the loud, energetic opposition to health care reform.
In this study, I minimize the effect of these intervening factors by reducing the time lag between a speech and the measured public response, by evaluating panel studies that examine attitude change of the same respondents before and after a speech, and by focusing on people who have been exposed to the speech and using individual-level data to control for their demographic differences. I find that presidents exercise strong leadership of public opinion. The bully pulpit, as the findings reveal, reaches a receptive audience that responds to the presidents’ appeals and changes its opinion.
Footnotes
Acknowledgements
The author thanks Barry Burden, Deven Carlson, John Coleman, Ken Mayer, and Byron Shafer for their helpful comments and suggestions.
Author’s Note
An earlier version of this paper was presented in 2010 at the Annual Meeting of the Midwest Political Science Association, and at the Political Behavior Research Group of the University of Wisconsin-Madison.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.