How Policy Conditions the Impact of Presidential Speeches on Legislative Success n

How Policy Conditions the Impact of Presidential Speeches on Legislative Success n Matthew Eshbaugh-Soha, University of North Texas Objective. Although the impact of the president s rhetoric on public opinion remains unfound, it appears to increase the president s success in Congress. This article argues that instead of moving public opinion, presidential speeches act as informational cues for legislators and holds that the impact of the president s public speeches in Congress is conditional on the salience and complexity of the policy voted on by Congress. Method. I use probit methodology to examine the effect of presidential rhetoric on the likelihood of presidential success on House roll-call votes from 1989 2000. An interactive model assesses the conditioning impact a policy s salience and complexity have on the relationship between presidential rhetoric and legislative success. Results. Presidential rhetoric increases the president s legislative success on votes pertaining to policies that are both salient and complex. Conclusion. Presidential rhetoric matters to the president s relationship with Congress, despite the limited impact it appears to have on public opinion. The topic of presidential speeches and their role in U.S. politics is central to a current debate in the literature: Do they matter to presidential governance? On the one hand are those who argue that presidential rhetoric is an effective and important tool of presidential leadership (e.g., Smith and Smith, 1994; Zarefsky, 1990). It may set the public agenda (Cohen, 1995), alter public opinion (Rottinghaus, 2010), influence media coverage (Peake and Eshbaugh-Soha, 2008), or even change the national conversation on important policy issues or people s presuppositions about government and politics (Hart, 2008:244 45). That rhetoric matters to the presidency is also vital to what Samuel Kernell calls going public. When presidents go public, Kernell (1997) argues, they go over the heads of Congress to enlist public support to achieve their legislative goals. Theoretically, presidential n Direct correspondence to Matthew Eshbaugh-Soha, Department of Political Science, 125 Wooten Hall, 1155 Union Circle #305340, Denton, TX 76203-5017 hmes@unt.edui. Data and coding information are available from the author to those who wish to replicate this study. Previous version presented at Annual Meeting of the American Political Science Association, Chicago IL, 2007. Thanks to the University of North Texas for providing research assistance for this project, John Bond and Richard Fleisher (1990) for making their roll-call vote data available online, and Brandice Canes-Wrone and Scott de Marchi (2002) for providing their data. Thanks to Paul Collins for his helpful suggestions. Many thanks as well to Chris Williams, Ryan Salzman, and Christine Harper for additional data collection. SOCIAL SCIENCE QUARTERLY, Volume 91, Number 2, June 2010 r 2010 by the Southwestern Social Science Association

416 Social Science Quarterly speeches put electoral or public pressure on legislators to support the president s policy positions. On the other hand are those who question the purported effectiveness of presidential rhetoric. Among these, George Edwards s (2003) critique may be the most convincing. His treatise on the president s rhetorical leadership of public opinion, in particular, shows that even two most rhetorically-gifted presidents Reagan and Clinton could not move public opinion on their top policy priorities. Even earlier claims that national addresses increased the president s public job approval ratings (Brace and Hinckley, 1992) appear time-bound (Edwards, 2003:30 32) and an ineffective means to move public opinion for recent presidents (see Welch, 2003). Edwards (1996) has long called into question the claims of many who espouse the benefits of presidential rhetoric, noting that rhetoric appears to have the potential to affect the public s opinion (and, presumably, others as well) but evidence that it does is lacking. If Edwards is correct that presidential speeches do not substantially affect public opinion, then one may surmise that presidential rhetoric has little impact, as well, on the president s legislative success. But this is not the case, according to recent research. Barrett (2004) shows that the average number of remarks per month on a bill increases the likelihood that presidents will be successful, arguing (but not illustrating) that speeches increase presidential success in Congress in the manner that Kernell (1997) envisions. Canes-Wrone (2006) demonstrates that remarks broadcast during nationally televised addresses increase the president s success rate, but her model does not require that presidents move public opinion to be successful, only that they take advantage of existing conditions, and prioritize those issues already popular with the public (see Edwards, 2009). Others circumvent the issue of presidential leadership of public opinion entirely, arguing that presidential rhetoric directly affects the president s yearly legislative success rate, but only for certain policy areas (Eshbaugh-Soha, 2006). No scholarship to date demonstrates a clear link between the president s rhetoric, an impact on public opinion, and a subsequent increase in the president s success rate. This article extends the examination of presidential rhetoric and legislative success in two ways. First, I argue that presidential rhetoric matters to the president s success in Congress, regardless of whether the president moves public opinion to support his policy preferences. Second, what shapes the relationship between presidential speeches and legislative success may not be the explicit movement of public opinion, but the type of policy issue that the president hopes to influence in Congress. Thus, I examine the impact that public policy has on the relationship between presidential speeches and legislative success and argue that different policies offer different incentives for legislators to respond to presidential speeches. Specifically, presidential speeches should increase the likelihood of legislative success only for those policies that are both salient and complex.

Policy, Presidential Speeches, and Legislative Success 417 With this in mind, this article attempts to answer several questions. What impact do the president s public speeches have on the president s success in Congress? Does their effectiveness vary by a policy s salience and complexity or are presidential speeches influential regardless of policy type? I analyze a sample of 460 roll-call votes on which the president took a position between 1989 and 2000 to answer these questions. I argue that the president s public attention to policies before Congress acts as informational and persuasive cues for legislators, leading to an increase in presidential success on roll-call votes even after controlling for other relevant variables such as party control of Congress. I will show that this relationship exists for only salient and complex policies by analyzing roll-call votes categorized by policy type and modeling an interaction between salience, complexity, and presidential rhetoric. This topic is of considerable importance to the study of presidential rhetoric and its effectiveness. Without question, the president s public speeches are a central force in U.S. politics. Presidents are engaged in a permanent campaign to which they devote substantial resources to influence the public, media, and Congress (Kumar, 2007). Although legislators are not always attentive or responsive to presidential speeches, they remain aware of the president s speeches and his policy priorities. If the impact of presidential speeches in Congress varies by policy, then we will have a better understanding of how speeches affect presidential success in Congress and can refine our expectations about when they are most likely to increase it. Moreover, even if we cannot substantiate a clear, quantitative relationship between presidential speeches and public opinion, a measurable impact of rhetoric on the president s legislative success rate is important to explaining why rhetoric matters much to a successful presidency. Presidential Success in Congress Presidential success in Congress is a core topic in the study of U.S. politics, one explained by several key variables. First, presidential success is more likely during the president s honeymoon period (Beckman and Godfrey, 2007; Dominguez, 2005; but see Barrett and Eshbaugh-Soha, 2007) so long as presidents hit the ground running (Pfiffner, 1988). Second, although higher presidential approval ratings marginally increase the likelihood of presidential success, reputation and skills matter less (Neustadt, 1990), and only in a few instances (Edwards, 1989). Finally, party matters. Just as conditions of divided government increase the likelihood that important legislation will fail to pass Congress (Edwards, Barrett, and Peake, 1997), so, too, does party control matter to the president s level of support (Edwards, 1989) and his success (Bond and Fleisher, 1990) on roll-call votes. Presidential speeches also tend to increase the president s legislative success rate, according to a range of studies (Barrett, 2004; Canes-Wrone, 2006; Eshbaugh-Soha, 2006; Kernell, 1997; but see Edwards, 2003).

418 Social Science Quarterly Although this research tells us much about presidential success in Congress, it neglects the impact that public policy may have, even though much scholarship has long held that policy should affect political processes (Lowi, 1972). Members of Congress represent their constituents differently across distinct policy issues (Erikson, 1978; Hill and Hurley, 1999; Miller and Stokes, 1963). Presidents lead public opinion in the short term on civil rights issues, just as they affect public concern for foreign affairs in the long run (Cohen, 1995; Hill, 1998; but see Young and Perkins, 2003). Others have employed large and small categories (equated with complexity by Rudalevige, 2002) to explain presidential influence on domestic policies (Light, 1999; Peterson, 1990) or time and importance to explain the president s policy agendas (Eshbaugh-Soha, 2005). When it comes to presidential success in Congress, though, much research focuses primarily on differences between foreign and domestic policy (see Edwards, 1989). Here, despite evidence to the contrary (Bond and Fleisher, 1990), Canes-Wrone, Howell, and Lewis (2007) illustrate an enduring congressional deference to the president on defense appropriations. Much of Shull s (1983; see also Spitzer, 1983) research also examines policy, but, alas, scholars have not continued to consistently model policy in answering important questions in presidential-congressional relations. Recent scholarship has refocused our attention on the importance of variation in domestic policy, nevertheless, by examining how the salience and complexity of policy affects presidential success in Congress (Canes-Wrone and de Marchi, 2002; see also Gormley, 1986; Rudalevige, 2002). Because it clearly affects other aspects of politics, it is also probable that there is an interactive or conditional relationship between the president s speeches and public policy as it relates to presidential success in Congress. Looking at the coding decisions of past research helps elucidates this point. Both Canes-Wrone (2006) and Barrett (2004) implicitly examine primarily salient policies. Barrett examines important legislation that, by Mayhew s (1991) coding decisions, involves media exposure or salience. Canes-Wrone (2006) studies those policies that public polling organizations have asked survey questions about, another indicator that the issue is salient to the public. By explicitly considering broader variation in policy types, as I do, this article can more definitively determine the extent to which policy matters to the impact that presidential attention to policies on the legislative has on the president s legislative success. It also shifts the focus away from the president s ability to move public opinion (which appears limited), and toward how issues themselves shape presidential leadership of Congress. An Informational Theory of Speeches and Success In light of the difficulties presidents have leading the public, I offer an alternative to the view that presidential speeches increase the president s

Policy, Presidential Speeches, and Legislative Success 419 legislative success rate by first moving public opinion. Theoretically, and in the tradition of Herbert Simon (1976), I rely on a vast and extensive literature that develops rational expectations for decision making. I presume that both presidents and, especially, legislators are boundedly rational actors who have a complex job with numerous decisions to make (Kingdon, 1981; see also Arnold, 1990:60, n. 1; Simon, 1976). Information shortcuts, including committee origin of a bill, colleague vote choices, interest groups, or constituency support (Matthews and Stimson, 1975), help legislators process vast amounts of information and make relatively informed, efficient, and consistent decisions (see Mondak, 1993; Simon, 1976). Another shortcut that legislators may use is information provided through the president s speeches. The president s speeches signal not only the president s policy position, but also his level of commitment, both of which may assist members of Congress as they decide whether to support a bill. After all, the source and presentation of information are both central to whether one may use it in one s decisions (Simon, 1976:164). For speeches to be used as sources of information by legislators, they must be accessible or relevant to the decision at hand. In other words, a cue is only useful to legislators when it helps them achieve their own goals. Legislators not only desire reelection, as Mayhew (1974) teaches us, they also desire good public policy (Fenno, 1978). Absent a veto-proof majority of legislative support, members of Congress require presidential support and his signature to achieve their own policy goals, and for their efforts to become law. Legislators also have limited time and resources to achieve a policy record on which to run and be reelected and so will listen to presidential speeches to ascertain the president s position. They are rational to pursue those bills that they support and that also stand a good chance of passing. As such, legislators will look to accessible presidential cues as a means to make efficient decisions to help in their own goal achievement. Cues sent from a trusted source a president from the legislator s own party, for example matter greatly (Kingdon, 1981:Ch. 6). Persuasion is a two-way street. Legislators need the president to achieve their goals just as presidents need legislators to achieve most of theirs. Of course, the burden of leadership in the modern U.S. system of government is on the president so that it is his persuasive task to convince legislators that what the [president] wants of them is what they ought to do for their sake and on their authority (Neustadt, 1990:30). For three reasons, a sufficient amount of presidential attention is required for speeches to increase the president s legislative success rate. First, the president s speeches must compete with additional cues available to legislators and more attention increases the chances that legislators will rely on the president s information rather than on that provided by another source. Second, more presidential attention increases the ease with which legislators may access the president s preferences and, thus, the efficiency with which legislators can make decisions. Third, others are unlikely to respond to presidential leadership after

420 Social Science Quarterly only the first time they hear the president speak (see Neustadt, 1990:31). These points reveal that the more attention the president devotes to a policy through speeches, the more a legislator knows the president s position on an issue, and given limited time and resources available to a legislator, also increases the probability that he or she will respond to the president s policy preferences or be convinced that what the president wants is what the legislator wants. Hypothetically, this is why more presidential attention to an issue should increase the president s legislative success on that issue. Fortunately for presidents, research shows that they have success influencing the issues that legislators think about. Edwards and Barrett (2000:132) conclude that the president is very successful in obtaining agenda space for his potentially significant legislative proposals. Because presidents tend to devote most of their public speeches to their top priorities (see, e.g., Edwards, 2003), it follows that presidential speeches are partially responsible for shaping the agenda as Edwards and Barrett (2000) find. Edwards and Wood (1999) show, nevertheless, that presidential speeches increase congressional attention only to healthcare, not to education or crime issues. That they find presidential leadership over health-care policy, which tends to be salient, yet complex, supports the importance of policy variation to presidential-congressional relations, a point that I elaborate on next. Policy Variation Having argued that legislators use cues to make decisions and that repeated attention by presidents should increase the chances that legislators will use and be persuaded by the president s speeches, the next question concerns how policy variation conditions legislative responsiveness to presidential speeches. Simply, policy salience and complexity conceptualized next provide different incentives for legislators to respond to presidential leadership. A salient policy is relevant to a large segment of the population and pushes elected officials to address an issue (Gormley, 1986:598). By definition, both presidents and members of Congress will be aware of salient policies because salience demands that elected representatives tackle an issue, given the link between public awareness of an issue and responsiveness (Miller and Stokes, 1963). Complexity concerns that which is technically difficult for the layperson to comprehend (Gormley, 1986), and usually is found in policies on which the public may not have firm positions. As such, the link between legislators and their constituents may be muddled on complex policies, even when salient, because the public does not have enough understanding to send clear signals to their representatives about their policy preferences. The public, because of this, is less likely to influence legislators vote decisions on complex policies. Of course, both legislators and presidents constantly address

Policy, Presidential Speeches, and Legislative Success 421 complex issues, as evidenced by the vast number of roll-call votes on complex policies. Salience encourages legislators to respond to their constituents, given the obvious electoral incentive (Mayhew, 1974), but complexity does not. Yet, legislators are unlikely to receive clear cues from constituents on issues that the public does not comprehend, that is, complex policies. This increases legislators uncertainty about how to vote on complex issues, which encourages them to seek additional information to make efficient vote decisions (see Esterling, 2004). So long as the president expresses his preferences through speeches (given other favorable conditions), an efficient way for legislators to gather additional information about their vote decision is to listen and respond to these speeches. It is without question, after all, that the executive branch possesses an overwhelming advantage in expertise and substantive information (Kingdon, 1981:194) that the president can bring forth in a persuasive effort. Yet, it is not complexity alone that drives legislators responsiveness to the president. Salience and complexity combined provide an optimal condition for presidential influence through speeches. Whereas legislators may respond to constituent preferences on policies that are salient alone, it is likely, given the public s lack of firm positions on complex policies, that legislators require additional information about how they should vote on any complex issue. Thus, legislators consistent with Canes-Wrone and de Marchi s (2002) argument that legislators look to presidential signals when comparable signals from their constituents are unclear are most likely to seek out additional information under conditions of uncertainty when an issue is also important to constituents. It is likely, therefore, that legislators will be most responsive to speeches on policies that are salient, that is, important to the public (and therefore legislators electoral self-interest) and on which the public does not have a clear opinion and is thus unlikely to send clear signals to its representatives (complex issues). Combining the theoretical expectations outlined above with variation by public policy in political processes, therefore, presidential attention to policy is most likely to increase legislative success on policies that are both salient and complex. It is worth noting that policy complexity, while necessary, is not sufficient for presidential speeches to influence legislation. Although speeches should be more likely to increase the president s legislative success on complex issues, presidents have little incentive to publicize their positions on issues that are not also salient (see Eshbaugh-Soha, 2006; Gormley, 1986). Like legislators, presidents have an electoral incentive to express positions on salient policies, and they regularly use their speeches to do so. Yet, presidents have limited time and resources to address those issues that will not benefit them electorally, and so we are unlikely to find that speeches influence nonsalient, yet complex policies. This, in turn, reinforces my argument that speeches will be influential only on policies that are both salient and complex.

422 Social Science Quarterly The Benefits to this Approach The dominant framework within which to view the impact of presidential speeches on legislative success has been Kernell s (1997) going public model. It presupposes that presidents speak to move public opinion and this, in turn, pressures legislators to support the president; a classic definition of presidential power and persuasion. Relying on an alternative framework instead of the coercive going public model put forth by Kernell (1997) may be controversial because much prior research pushes the theoretical expectation, even though there is little research that explicitly supports this indirect mechanism of presidential leadership of Congress. In fact, mounting evidence calls into question the ability of presidents to call the public to arms, from Hart (1987:86), who observes that there is not a positive relationship between increased speechmaking over time and public opinion, to Edwards (2003), who shows that presidents are unable to increase public support for even their top policy priorities. Although Rottinghaus (2010) finds that a major presidential address leads to about a 10 percentage point increase in public support but only in the short-term (at one month), the national address is a rare event and even initial increases in support evaporate as time from a national address increases (see Edwards, 2003:37). No study extends even this short-term influence over public opinion to influence in Congress. Because it espouses an informational theory geared toward legislators vote decisions to explain presidential success in Congress, this article does not expect the president to move or motivate a U.S. public that is generally disinterested in U.S. politics or slow to move by presidential efforts at rhetorical leadership (Edwards, 2003). It is more consistent with many of Canes-Wrone s (2006) theoretical assumptions, which allow for a more flexible and strategic account of presidential leadership. That is, the president is in a strong position to be able to determine his own agenda and can prioritize those issues that stand the best chance to succeed. For Canes- Wrone, this depends on the popularity of a policy. For me, the success of presidential rhetoric depends on policy salience and complexity, which affect legislators decision calculus. Data The data set used in this analysis is a combination of Bond and Fleisher s (1990) roll-call vote data set, Canes-Wrone and de Marchi s (2002) data of presidential approval, roll-call salience, complexity, and presidential success for the House of Representatives, and my own data collection. In effect, I reduced the entire set of House roll-call votes on which the president took a position to 460 by eliminating fiscal or appropriations votes and those votes that do not pertain to enactments of bills, final passage, conference reports, and veto overrides, consistent with Canes-Wrone and de Marchi

Policy, Presidential Speeches, and Legislative Success 423 (2002:497), 1 who reason that it is on only these votes that we can measure salience adequately. Although most newspaper articles addressed enactment of bills directly, amendments and procedural votes were discussed only indirectly and, thus, are not comparable with the other bills. Presidential Success The president s success rate is simply coded 1 if the president was successful on a roll-call vote and 0 if he was not. These data for the House of Representatives were coded initially using the Congressional Quarterly Almanac by Bond and Fleisher (1990) and confirmed in conjunction with data provided by Canes-Wrone and de Marchi (2002). Presidential Attention I collected a measure of presidential attention from the available electronic collection of The Public Papers of the President: The American Reference Library cd-rom, which contains the Public Papers of the Presidents (1989 1994), and Weekly Compilation of Presidential Documents (1995 1998), an online version of the Public Papers of the Presidents (1998 2000). 2 The coders created a set of key words associated with the policy of each roll-call vote 3 and counted the number of paragraphs in a speech (not written communications) in which a key word appeared. We counted mentions that occurred anywhere from one to twelve months before the date of the roll-call vote 4 so that any mention that occurs after a roll-call vote is not included in the count. There are several ways to analyze presidential attention to policy, including a count of the number of speeches related to a roll-call vote (Barrett, 2004). Because not all speeches are equal some are much longer and more detailed than others I account for this by counting and analyzing the total number of paragraphs associated with a roll-call vote. These data average 17 paragraphs per roll-call vote and range from a low of 0 (indeed, 0 is also the modal category) to a maximum of 425 paragraphs devoted to NAFTA in 1993. 1 I add tax bills, foreign policy (including use of force resolutions), and other roll-call votes excluded by Canes-Wrone and de Marchi (2002). Controlling for tax and foreign policy bills (coded mostly as complex) has no substantive impact on the results. 2 Available at hhttp://www.gpoaccess.gov/pubpapers/index.htmli. 3 A key words list is too extensive to include but is available from the author. One example is House roll-call vote number 374 in 1992, HR 4547 Russian Aid. The coders searched aid and Russia n and counted only those paragraphs that deal with the president s position on the bill. They filtered out those paragraphs that may have addressed Russia or foreign aid, generally. 4 Twelve months was selected in part because previous research has done so (Barrett, 2004). Varying this did not make a difference in the results reported below.

424 Social Science Quarterly Veto Threat Another type of signal is a presidential veto threat. My theory suggests that as legislators will look to presidential speeches for information, so, too, will they look to presidents when they threaten to veto. This is a particularly strong, clear, and coercive cue that leaves little doubt of the president s position and commitment to an issue. Presidents should be more successful when they threaten a veto because legislators are likely to make some concessions to the president s position after a veto threat (Cameron, 2000). If legislators do not capitulate, then their efforts to achieve a policy goal will be in vain. Veto threats are coded 1 when the president threatened to veto a bill; 0 otherwise. These data are from statements of administration policy as coded and provided by Kernell (2005). Presidential Approval The measure of the president s job approval rating is taken from Gallup s question: Do you approve or disapprove of the way [the current president] is handling his job as president? It is the measure of approval from the poll immediately prior to the roll-call vote (Canes-Wrone and de Marchi, 2002) and updated by me. In my sample of 196 roll-call votes, average approval rating is 56 percent, ranging from a low of 29 percent in August 1992 to a high of 89 percent for George Herbert Walker Bush in March 1991. Salience and Complexity Salience is coded as media attention to a policy area. This is a content analysis of the number of articles in which a bill was discussed in either a headline or lead paragraph in stories collected from 12 newspapers available in the online newspaper database, newslibrary.com. 5 This measure is very similar to that used by Canes-Wrone and de Marchi (2002) and correlates at r 5 0.86. In part because of this, I borrow their externally 5 To produce a broad measure of salience, I selected at least one newspaper from each region identified by newslibrary.com New England, Middle Atlantic, South Atlantic, East North Central, East South Central, West North Central, West South Central, Mountain, and Pacific and one national newspaper, for a total of 12 newspapers. These newspapers are: Denver Post, Washington Post, Saint Petersburg Times, Lexington Herald-Leader, Detroit Free Press, Star Tribune, Union Leader, Charlotte Observer, Columbus Dispatch, Philadelphia Daily News, Houston Chronicle, and Seattle Times. I searched six months before the date of the roll-call vote and three months after. I could not replicate or extend Canes-Wrone and de Marchi s (2002:496) measure of media attention because, due to an apparent change in the Lexis-Nexis database during the summer of 2007, many newspapers in their sample are no longer available.

Policy, Presidential Speeches, and Legislative Success 425 validated cutoff for what constitutes a salient policy, at 17 stories per roll-call vote. 6 Coding complexity is more subjective than coding salience. Although some suggest counting the number of pages in a legislative statute to assess complexity (Gerber and Teske, 2000), most other research makes reasonable judgment calls (see Nicholson-Crotty, 2009; Gormley, 1986). Among other areas, insurance reform (see also Lupia, 1994) and nuclear energy regulation (see also Kuklinski, Metlay, and Kay, 1982) are both complex because they require technical knowledge to understand. Moral, civil rights, and crime issues, in contrast, tend to be less complex, according to Gormley (1986). Consistent with Gormley (1986), both Carmines and Stimson (1980) and Iyengar (1989) designate race and crime, respectively, as easy or otherwise uncomplicated policy areas. For this article, complexity is generally coded 1 if the roll-call vote involves a regulatory issue or 0 if it concerns a social issue, similar to Canes-Wrone and de Marchi (2002) and consistent with the decision rules in the literature. The idea is that social issues will be more symbolic and easier to understand than regulatory issues that are technically complex. Control Variables Several other variables also explain the president s success in Congress and are controlled for in the following analysis. First, less than a majority of party control in the House should decrease presidential success on a roll-call vote. This variable is coded as the percentage of House of Representative seats that are held by members of the president s party (Canes-Wrone and de Marchi, 2002). Second, the number of positions (logged) on which the president took a position for a year may affect the president s success, with a larger agenda (more positions taken) leading to a lower overall likelihood of success (Rivers and Rose, 1985). Third, there may be differences in success rates across presidents, so I code 1 when Clinton was president and 0 when he was not. This is especially important for an analysis of speeches, since Clinton tended to speak much longer than President Bush. Fourth, presidents should be more successful on veto override votes, given the higher threshold (two-thirds) for defeat. Veto override equals 1 if the roll-call vote concerned a veto override and 0 when it did not. 7 6 Several alternative analyses reveal that there is a range of potential cutoffs for how many articles might be considered salient, from 7 to 20 articles. Using different cutoffs for salience does not alter the central results presented in this article, although it would increase the N reported in Table 2 substantially, to 74, if seven stories were selected. 7 A honeymoon variable (coded 1 for the first year of a president s first term in office) had no substantive bearing on success or other variables in the model, and has been dropped.

426 Social Science Quarterly Findings To begin, Table 1 presents descriptive statistics of my key independent variable the number of paragraphs in presidential speeches by salience and complexity. Accordingly, presidents speak on average most often about issues that are salient and complex. This comports with the idea that presidents should be most influential on these policies and may also be a function of the amount of effort presidents must engage in to be able to explain their positions publicly on complex issues. The standard deviation indicates significant variation in these efforts, nevertheless. Roll-call votes that are salient but not complex reveal a modest average of 31 paragraphs per vote, with a much narrower range than salient and complex issues. In conjunction with the descriptive statistics for salient and complex policies, this suggests that presidents are focused more on pushing policies that are also complex. Presidents averaged only 11 paragraphs on votes not salient but complex, and spoke even less on policies that were neither salient nor complex. 8 To analyze whether the president is successful on a roll-call vote (a dependent variable measured as 1 or 0), the several models use probit analysis. The first model is a baseline model that includes all observations in the sample and is not differentiated by a vote s salience or complexity. As Table 2 demonstrates, presidential speeches do not increase the likelihood that a president will be successful on roll-call votes without differentiating by a policy s salience and complexity. This suggests that previous research may have inadvertently sampled on primarily salient and complex policies (Barrett, 2004; Canes-Wrone, 2006). Party control has a strong and positive impact, as expected, while veto threats, surprisingly, lead to a lower likelihood of success. This seems to contrast with Cameron s (2000) expectation of congressional capitulation after a veto threat and is worthy of additional exploration by future research. It implies, nevertheless, that coercive signals may be less influential than informational cues. More votes suppress the likelihood of success, too, and President Clinton was less likely to be successful on roll-call votes than was President George Herbert Walker Bush. The latter two findings are generally consistent regardless of a roll-call vote s salience or complexity and with Canes-Wrone and de Marchi s (2002) findings. Breaking down roll-call votes by salience and complexity confirms my expectation: the largest and only statistically significant impact of presidential attention is for the salient, complex policy area. The impact of attention on the likelihood of success with variables at their means is 0.01. This is substantively meaningful, too, when converted into a predicted probability. 8 Some have suggested that a more appropriate analysis of policy variation would be to examine the standard domestic-foreign policy dichotomy. An analysis reveals that foreign policy votes are no more likely to be influenced by presidential attention than domestic policy votes.

Policy, Presidential Speeches, and Legislative Success 427 TABLE 1 Descriptive Statistics for the Number of Presidential Paragraphs Mean Paragraphs SD of Paragraphs Salient/complex 49.32 98.89 Salient/not complex 31.42 59.78 Not salient/complex 11.09 45.28 Not salient/not complex 6.45 19.63 Total paragraphs 16.71 53.17 SOURCE: Public Papers of the Presidents, data collected by the author. Increasing the number of paragraphs devoted to a bill by one-half a standard deviation above the mean, or about 49 additional paragraphs devoted to policies that are both salient and complex, increases the probability of presidential success on a roll-call vote by nearly 50 percent. This not only shows that different policies encourage different degrees of legislative responsiveness as I have theorized, it also buttresses the importance of rhetoric to presidential governance broadly. Other variables have expected effects. Party control matters across nearly all roll-call votes. Typically, presidents can count on about a 30 percent increase in the likelihood of success with every four additional seats added to presidential party makeup. Other variables contrast with previous research. Presidential approval ratings have no statistically significant impact in any of the models, which contrasts with Canes-Wrone and de Marchi s (2002) finding that approval ratings increase the likelihood of presidential success in Congress if and only if a roll-call vote is salient and complex. 9 What is more, veto override votes are not significant in any of the models, once again, in contrast with Canes-Wrone and de Marchi (2002). These contrary findings appear to be a function of the much larger sample of roll-call votes used in this article. As an example, presidents lost on over half the veto overrides in my sample, but won a greater percentage of votes in Canes-Wrone and de Marchi s (2002) smaller sample of roll-call votes. Policy also matters in a different way. For those policies that are neither complex nor salient, few variables accurately predict success. Even party control, which matters substantially for all other roll-call votes, is statistically insignificant in this model. Without clear and predictable incentives for involvement by presidents, these roll-call votes defy theoretically driven expectations and appear more random, perhaps a function of the individual policy areas, themselves, that are not explainable by broad policy categories. Kingdon (1981) also observes that cues are not important to legislative decision making when there is general consensus. If one were to demonstrate in 9 These nonfindings are fairly robust and not a function of any collinearity with speeches. Approval fails to reach statistical significance when paragraphs are excluded from the models.

428 Social Science Quarterly TABLE 2 Determinants of Presidential Success in Congress All Votes Salient/Complex Salient/Not Complex Not Salient/Complex Not Salient/Not Complex Presidential attention 0.0007 0.01 n 0.0001 0.002 0.01 (0.001) (0.006) (0.003) (0.002) (0.01) [10.49] Public approval 0.36 1.63 0.08 0.51 1.01 (0.63) (3.09) (1.79) (0.92) (1.34) Party control 11.46 n 12.47 n 31.09 n 14.30 n 5.10 (2.07) (6.21) (7.39) (1.77) (3.88) [10.30] [10.32] [10.78] [10.35] Veto threat 1.02 n 0.08 1.21 n 0.98 n 1.54 n (0.14) (0.47) (0.55) (0.19) (0.34) [ 0.38] [ 0.45] [ 0.37] [ 0.53] Veto override 0.33 0.38 0.52 0.04 0.74 (0.33) (0.65) (0.62) (0.60) (1.07) Clinton 0.95 n 2.74 n 4.02 n 0.99 n 0.21 (0.25) (0.78) (1.28) (0.28) (0.54) [ 0.36] [ 0.78] [ 0.95] [ 0.37] Votes (logged) 0.91 n 3.96 n 4.26 0.58 0.74 (0.39) (1.44) (2.61) (0.51) (1.07) [ 0.07] [ 0.31] Constant 0.07 n 14.31 7.50 2.66 0.11 (2.23) (8.76) (10.51) (2.83) (5.04) Wald chi-square 108.31 n 17.39 n 38.59 n 164.14 n 32.09 n PRE 36.65 13.57 25.79 40.27 32.58 % predicted 69.9 58.48 64.35 71.30 67.61 N 460 50 56 243 107 n po0.05 (one-tailed). NOTE: Robust standard errors are in parentheses. Predicted probabilities, which are one-half standard deviation changes for all variables except dummy variables, which are one-unit changes from 0 to 1, are in brackets. When veto threat or veto override is dropped from a model, it is because each predicts failure perfectly. The modal category is 51.96 percent.

Policy, Presidential Speeches, and Legislative Success 429 future research that roll-call votes that are neither salient nor complex tend not to be conflictual, this would provide additional reasoning behind these results. Nevertheless, these null findings support the general premises of this article: presidential success in Congress and the variables that predict its likelihood vary by policy area, which is particularly pronounced for the key independent variable in this study, presidential attention through public speeches. To address any concerns that the results in Table 2 are simply an artifact of the various sample sizes and separate groupings by policy types, Table 3 presents a series of interactions in two models: one to assess the three-way interaction between speeches that are delivered on policies that are salient and complex and another to examine any two-way relationships with speeches. As I have theorized, there should be a conditional relationship between presidential attention, salience, and complexity accounted for in an interactive model and only this relationship should have a statistically significant impact on success. Table 3 clearly upholds the conditional importance of both salience and complexity to the impact that presidential attention has on success. Only the interaction between all three matters, just as neither salience nor complexity when interacted separately with presidential attention reach conventional levels of statistical significance. Holding variables at their means (or at 1 for dummies) and interpreting based on a half standard deviation increase, as presidents deliver 26 more paragraphs on policies that are both salient and complex, the likelihood of success increases by 10 percent. I remind the reader that I do not theorize a direct relationship between media attention and success, or an indirect relationship between presidential attention, media attention, and success. Although presidential and media attention are clearly related registering a correlation of 0.32 (Pearson s R) in my data set media attention does not independently cause success (or vice-versa) and multicollinearity is not a problem. An analysis that codes salience as newspaper stories that occur only after the roll-call vote and, therefore, after the president ceases to push for passage of a bill produces near identical findings. Moreover, media attention does not significantly affect the likelihood of presidential success in my sample of roll-call votes. Therefore, I reject the notion that presidential attention somehow increases presidential success indirectly through increased media attention and conclude that the impact of salience on success is conditional as I have hypothesized. Finally, these data speak to the president s strategy in giving speeches. Although Peterson (1990) and others maintain convincingly that presidents do not choose to prioritize bills simply because they are likely to be successful, one might surmise that presidents may wish to increase their attention to issues immediately before a vote when victory appears definite. Additional analyses reject this presupposition on two counts. First, a model reveals that the likelihood of success does not increase the number of speeches the president delivers. Most importantly, it does not increase the

430 Social Science Quarterly TABLE 3 Interaction of Presidential Attention, Salience, and Complexity on Success Model 1 Model 2 Presidential attention 0.0002 0.01 n (0.001) (0.003) [ 0.12] Salience 0.16 0.13 (0.18) (0.18) Complexity 0.04 0.05 (0.14) (0.15) Attention * Salience * Complexity 0.01 n (0.004) [10.10] Attention * Salience 0.00007 (0.0004) Attention * Complexity 0.05 (0.15) Public approval 0.004 0.004 (0.01) (0.006) Party control 11.54 n 11.04 n (2.07) (2.05) [10.30] [10.29] Veto threat 1.01 n 1.02 n (0.14) (0.14) [ 0.38] [ 0.39] Veto override 0.33 0.35 (0.33) (0.33) Clinton 0.99 n 0.91 n (0.25) (0.25) [ 0.38] [ 0.35] Votes (logged) 0.94 n 0.94 n (0.39) (0.40) [ 0.07] [ 0.07] Constant 0.09 n 0.28 (2.24) (2.27) Wald chi-square 109.63 n 113.37 n PRE 37.10 36.65 % predicted 69.78 69.57 N 460 460 n po0.05 (one-tailed). NOTE: Robust standard errors are in parentheses. Predicted probabilities, which are one-half standard deviation changes for all variables except dummy variables, which are one-unit changes from 0 to 1, are in brackets. The modal category is 51.96 percent. president s level of attention the month before the roll-call vote. Second, the impact of presidential attention one month before a roll-call vote has a statistically insignificant impact on the president s likelihood of success. It goes without saying that presidents simply do not wait until victory is

Policy, Presidential Speeches, and Legislative Success 431 assured before they advertise their positions a requirement for this counterargument to prove valid but speak months ahead of time, attempting to inform and lead Congress to support their policy positions. Conclusion This article began with a debate about the effectiveness of presidential rhetoric in U.S. politics. Although Edwards (2003) rejects the claim that rhetoric is an effective tool for moving public opinion, scholarship still shows that presidential speeches increase the president s legislative success. We now have additional evidence to support the benefits of presidential rhetoric as a tool of influence, given this article s support for a positive impact of presidential speeches on the president s legislative success rate. Nevertheless, it is not that presidential speeches increase the president s success on all policies before Congress, but that this impact is conditional on a policy s salience and complexity. Presidential attention has negligible effects on roll-call votes when policy is not considered. In many ways, this article also clarifies the findings of previous research that finds rhetoric increases the president s success in Congress. Although they do not do so explicitly, both Barrett (2004) and Canes-Wrone (2006) appear to examine mostly salient policies in their analyses. I explicitly examine both salience and complexity as conditioning the effectiveness of presidential speeches on legislative success and, given the results of this article, conclude that it is not salience alone that matters, but that salience is important only in conjunction with the uncertainty that complex policies introduce to legislators in their vote decisions. This article has theorized, indeed, that legislators look for information shortcuts from the president on complex issues and that are also important to a legislator s electoral selfinterest because the public does not firmly grasp these issues and cannot indicate clearly its preferences to the representatives. It is for this reason that legislators respond to presidential speeches, increasing the president s success on policies that are both salient and complex. These findings hint at an important role for presidential rhetoric that is worthy of additional inquiry. Rhetoric may be a tool presidents can use to frame issues before Congress. Frames are used, after all, to simplify complex issues and make them more accessible to the listener. Although my data are not designed to support this assertion, at the very least, future research should examine the language that presidents use to make their case before Congress. A study of the president s words themselves may help ascertain which types of frames presidents do use on complex and salient issues, and which ones, if any, tend to be most beneficial to the president in his relationship with Congress. This may in turn be relevant to exploring the president s ability to not only influence legislation with his rhetoric, but also affect the national policy agenda through speeches.

432 Social Science Quarterly Future research should pursue two additional lines of inquiry. First, a broader sample that includes roll-call votes from administrations throughout the modern presidency is needed. Time dynamics affect much of what affects presidential success in Congress, including the relative balance of partisanship and public approval ratings (Bond, Fleisher, and Wood, 2003). Analyses of public speeches and policy variation may reveal similar dynamics, that is, that the impact of speeches has varied over time as other institutional and political conditions change. Second, the analysis should be extended to the Senate. The institutional differences between the House and Senate may produce varied results, perhaps with speeches increasing the president s success in the Senate on different policy types or not at all. REFERENCES Arnold, R. Douglas. 1990. The Logic of Congressional Action. New Haven, CT: Yale University Press. Barrett, Andrew. 2004. Gone Public: The Impact of Going Public on Presidential Legislative Success. American Politics Research 32:338 70. Barrett, Andrew, and Matthew Eshbaugh-Soha. 2007. Presidential Success on the Substance of Legislation. Political Research Quarterly 60:100 12. Beckman, Matthew, and Joseph Godfrey. 2007. The Policy Opportunities in Presidential Honeymoons. Political Research Quarterly 60:250 62. Bond, Jon R., and Richard Fleisher. 1990. The President in the Legislative Arena. Chicago, IL: University of Chicago Press. Bond, Jon R., Richard Fleisher, and B. Dan Wood. 2003. The Marginal and Time-Varying Effect of Public Approval on Presidential Success in Congress. Journal of Politics 65:92 110. Brace, Paul, and Barbara Hinckley. 1992. Follow the Leader. New York: Basic Books. Cameron, Charles M. 2000. Veto Bargaining: Presidents and the Politics of Negative Power. Cambridge, UK: Cambridge University Press. Canes-Wrone, Brandice. 2006. Who Leads Whom? Presidents, Policy, and the Public. Chicago, IL: University of Chicago Press. Canes-Wrone, Brandice, and Scott de Marchi. 2002. Presidential Approval and Legislative Success. Journal of Politics 64:491 509. Canes-Wrone, Brandice, William G. Howell, and David E. Lewis. 2007. Toward a Broader Understanding of Presidential Power: A Reevaluation of the Two Presidencies Thesis. Journal of Politics 69:1 16. Carmines, Edwards G., and James A. Stimson. 1980. The Two Faces of Issue Voting. American Political Science Review 74:78 91. Cohen, Jeffrey E. 1995. Presidential Rhetoric and the Public Agenda. American Journal of Political Science 39:87 107. Dominguez, Casey Byrne Knudsen. 2005. Is it a Honeymoon? An Empirical Investigation of the President s First Hundred Days. Congress and the Presidency 32:63 88.