Does the presidency moderate the president?
I study the effect of the presidency on presidents by comparing the legislative records of 23 individuals who served in both Congress and the White House. The question is whether someone acts differently when representing the entire nation than he does when representing a single state or congressional district. In this analysis, I am interested in political decisions that may be viewed as the expression of personal preferences subject to the constraints of politics. By observing these individuals' legislative records before and after they became presidents, we can determine whether the presidency moderated their behavior. (1)
It is important to know how the presidency is likely to affect the policy decisions of someone who aspires to be president. Candidates often claim they will think broadly and work with members of the opposite party, but are these empty campaign promises? Is a candidate's record in Congress a reliable indicator of how s/he will act as president? Selecting the president is perhaps the most important decision voters make and this research helps us evaluate how candidates might act in the White House. Additionally, to the extent that one is troubled by the polarization of the presidency, it is vital to properly diagnose the source of the problem: should we blame presidents or the presidency? Restoring centrist presidential leadership may not be a matter of picking the right person for the job, but rather reforming the presidency.
I begin by discussing two contrasting positive theories of presidential representation, noting the evidence in support of each view. Next, I show how these models of representation can be tested by analyzing the legislative records of 23 men who served in both Congress and the White House. Implementing this design requires extensive analysis of roll-call data to estimate the ideal points of presidents and members of Congress on the same scale. Having made these estimates, I find that the presidency effectively moderated presidents' policy preferences for more than a century, but no longer moderates presidents. Rather than moderate presidents, the modern presidency appears to amplify the partisan leanings of the president. The modern presidency compels presidents to assert relatively extreme preferences in the legislative process. I discuss some potential limitations of this analysis and conclude by stating the implication of this research.
The Unique Perspective Thesis
The differences between the national and local constituencies of presidents and members of Congress is thought to compel presidents to act differently than legislators. The unique perspective thesis is broadly advanced in modern analysis of the presidency. According to (now) Supreme Court Justice Elana Kagan (2000, 2335), "[Bfecause the President has a national constituency, he is likely to consider, in setting the direction of administrative policy on an ongoing basis, the preferences of the general public, rather than merely parochial interests." Political scientists have similarly maintained that the president has a uniquely broad perspective on national issues. "Elected by a nationwide constituency," Bond and Smith (2008, 461) observe, "the president tends to see issues from a national perspective." Cronin and Genovese (1998, 198) also emphasize presidential moderation: "Once in office, presidents often bend over backwards in an attempt to minimize the partisan appearance of their actions."
Implicit in the unique perspective thesis is the influence of the median voter. Both presidents and members of Congress seek the median voter's support, but legislators' constituencies are smaller and may be homogeneous with special economic interests. The nation may be evenly divided between two major political parties, but some states and congressional districts overwhelmingly support one party or the other. For presidents, the median voter is the median of median voters, almost by definition more moderate than median voters in districts throughout the country.
In addition to facing a different electorate, presidents play a different role in the legislative process than members of Congress. Members often can vote yea or nay knowing their vote is not decisive while the president's desk is typically the last stop a bill will make. A president does not have as much latitude as a member of Congress to take political positions without worrying about policy consequences. The president may relent to congressional majorities rather than delivering a futile veto message. Historically, Congress has been the creative engine of government; the president, merely an agent of Congress tasked with enforcing laws (Thurber 2013; Shull 2000). According to Selinger (2014, 28), historic era presidents restrained their ambitions to preserve national unity: "The task of steering the ship of state between the partisan extremes and preventing polarizing divisions from escalating into violence was a crucial impetus for action for the antebellum presidents." The president's role discouraged his displaying overt partisanship and intruding upon the legislative process.
According to some presidency scholars, the Framers designed Article II of the U.S. Constitution to insulate the president from the influence of factions and expected the president to enforce laws for the benefit of the entire nation (Nzelibe 2005; Levy 2000, chap. 2). (2) This original vision of the presidency did not rely entirely upon individual largess, but rather attempted to impose enduring constraints on whomever came to occupy the office. The Constitution curbs zealous executives through regular elections, provisions for impeachment, and the checks and balances of other branches of government. (3)
The claim that the president advances broad national interests, rather than the special interests of narrowly defined groups is most passionately asserted by the president himself. The idea that the president operates on a higher plane than ordinary legislators is politically appealing (Rhodes 2014). "Most citizens abhor partisanship and prefer that the president be responsive to the broader public will, rather than simply those responsible for their election" (Wood 2009, 15). In the 2012 presidential election, both President Obama and Governor Romney emphasized the president's unique responsibility to represent the entire country. Obama told a national television audience, "you represent the entire country ... if you want to be president you've got to work for everybody, not just for some" (Walsh 2012). (4) Likewise, Governor Romney emphasized centrist leadership: "[W]e're going to have to have a president who can work across the aisle. I was in a state where my legislature was 87 percent Democrat. I learned how to get along on the other side of the aisle" (National Public Radio 2012).
The Partisan Presidency
Emerging accounts of the presidency call the viability of the unique perspective thesis into question. The partisan model of representation holds that presidents represent their party more than the country as a whole. Rather than bend to the median voter's preferences, partisan presidents may use the national stage to influence mass opinion (Nzelibe 2005; Eshbaugh-Soha and Rottinghaus 2013; Wood 2009; Cohen 2015). In his recent book, The Myth of Presidential Representation, Wood (2009) argues that the president does not adapt his message to public opinion; instead, the president uses executive authority to shift public opinion in support of his policies. The partisan presidency may enable presidents to advance a bolder agenda than members of Congress.
The notion that presidents merely execute laws passed by Congress appears antiquated in the modern political era. The presidency and the political environment in which presidents operate have changed significantly over time. There are at least three reasons the presidency may no longer moderate presidents' political preferences. First, the institutionalization of the presidency enables presidents to vigorously assert their partisan preferences; in particular, presidents can announce their views on legislation early and often. Second, presidents have become the focal point of partisan opposition in Congress. Third, the polarization of American politics potentially undermines the median voter's moderate status. This article may not be well positioned to adjudicate among the possible causes of presidential partisanship, but these are developments that motivate this research.
During Franklin Roosevelt's administration, the executive branch swelled and assumed unprecedented power over foreign and domestic policies (Lowi 1985; King and Ragsdale 1988). Because institutions shape behavior, this dramatic evolution of the presidency may have changed how presidents lead. Institutional power is likely to embolden presidents. Modern presidents, supported by the executive bureaucracy, are policy leaders, rather than glamorous clerks (Neustadt 1991). Modern presidents lead their parties by defining and promoting partisan legislative agendas. Executive authority may make the president less likely to compromise for the sake of passing laws and more likely to assert partisan preferences to lead, rather than be led by, Congress.
The Office of Management and Budget (OMB) plays a particularly significant role in the president's ability to assert partisan preferences early and often in the legislative process. Originally established within the Department of the Treasury to streamline agency appropriations, the OMB was moved into the Executive Office of the President in 1939 (Berman 1979). The OMB regularly issues Statements of Administration Policy to communicate the president's positions on pending legislation clear to member of Congress (Rice 2010). There is ongoing research on how presidents set their legislative priorities (Gelman, Wilkenfeld, and Adler 2015) and whether presidents are able to influence Congress (Edwards 1989; Covington, Wrighton, and Kinney 1995; Steger 1997), but it is clear that modern presidents have the ability to assert their policy views before bills reach their desk and are not limited to announcing legislative priorities in his annual address to Congress (Cohen 2012, chap. 1). According to Wayne (2009, 321), executive position taking is now so frequent that "[t]he absence of a presidential position on pending bills or the failure to communicate one is normally interpreted as a sign of indifference by the administration and as a green light for Congress to work its will.... The White House is not shy about making its opinions known." The modern, institutionalized presidency may, therefore, bolster presidents' capacity to assert their partisan policy preferences; the historic presidency, in contrast, would have resigned presidents to a more passive role in the legislative process. (5)
Because the president's legislative success affects how his party is evaluated, the modern president is often the focal point of partisan opposition in Congress. Legislative success has become something of a zero-sum contest. The opposing party in Congress may vote against the president's legislative positions to advance its own electoral interests (Lebo and O'Geen 2011; Lee 2009, chapt. 4). Democrats in Congress can be expected to oppose a Republican president's legislative agenda and Republicans in Congress a Democratic president's agenda. As a result, President Obama's support (or opposition) to a bill may polarize roll call voting in Congress more than Senator Obama's legislative agenda did. One may wonder whether this dynamic results from a change in the presidency or a change in Congress, but these developments may be viewed as two sides of the same coin. As presidents assert their party's legislative agenda more vigorously and are credited for its success, the opposing party in Congress pushes back against the president's agenda.
Finally, there is a significant debate over the distribution of preferences in the general electorate. As discussed above, a moderate median voter is an implicit assumption of the unique perspective thesis. Some argue that the ideological divide in the general public is increasing and the political center eroding (Abramowitz 2010; Kernell and Rice 2011). (6) Changes in the mass electorate may be particularly significant for presidents in the modern era because the Electoral College does not insulate presidential selection from the popular vote. (7)
The three factors discussed in this section, the institutionalization of the presidency, congressional responses to the president's agenda, and political polarization, challenge the continued vitality of the unique perspective thesis. However, because these are modern political developments, the unique perspective thesis may still be historically accurate. These contrasting accounts of presidential politics may both be useful explanations for different eras of the presidency.
Estimating Ideal Points of Presidents and Members of Congress
Comparing the voting records that individuals compiled in two different offices provides an opportunity to test the competing positive theories of presidential representation. If the president's unique perspective is not merely a rhetorical appeal, this unique perspective should have a measureable effect on individual decision making. If the presidency moderates the office holder, one would expect liberal members of Congress to act less liberally and conservative members less conservatively in the White House. (8) On the other hand, if the presidency accentuates or amplifies the political preferences of the office holder, one would expect senators and representatives on the left to move further left in the White House and those on the right further right. If the institution does not affect individual preferences, the individual's ideal point will not change as a result of moving from Congress to the White House. The competing hypotheses about presidential representation are depicted in Figure 1. (9)
To conduct this analysis, I first need to estimate the ideal points of those who served in both Congress and the White House on a comparable scale. This requires extensive historical research to fill gaps and correct potential biases in prior research. Next, I use these estimates to analyze the effect of the presidency on individual behavior using conventional statistical analysis. In this analysis, ideal points provide a nice summary of individual voting patterns. Because legislative decisions may be influenced by external factors, the ideal points estimated here should not be viewed as measures of subjective personal beliefs. (10) Additionally, one should keep in mind that comparing presidents who served in different eras is problematic because the issues that define party systems change over time. (11)
Over half of U.S. presidents served in Congress before becoming president (and two presidents returned to Congress). Congress, particularly the Senate, has often been viewed as a stepping stone for the presidency. (12) As shown in Table 1, the time between service in Congress and the White House is generally limited.
One can estimate politicians' ideal points from roll call votes. The roll call record of Congress is a particularly useful data source; congressional scholars have analyzed roll call voting in Congress to successfully estimate the ideal points of senators and representatives using a variety of multidimensional scaling methods (Poole and Rosenthal 2007; esp. Clinton, Jackman, and Rivers 2004). While methodological approaches vary, nearly all scholars rely on congressional roll call data with bridging observations to compare presidents' records to those of members of Congress.
The president's position on a roll call vote provides a bridge observation between the executive and legislative branches that is essential to comparing institutions. (13) Past efforts to compare presidents to members of Congress have relied on research by the Congressional Quarterly (CQ) and Swift et al. (2001). Since 1953, the CQ has identified the president's public positions on roll call votes, typically through careful analysis of presidential addresses and OMB Statements of Administration Policy on pending legislation. Swift et al. (2001) analyzed the public papers and addresses of Presidents Washington through H. W. Bush to compile a substantial data set of presidential legislative requests and matched these requests to roll call votes in the House and Senate. This ambitious project has enabled scholars to estimate the ideal points of many nineteenth-century presidents in relation to members of Congress (e.g., McCarty 2009).
Despite substantial progress estimating presidential ideal points, existing measures of presidential preferences are problematic. The estimates are incomplete; when someone casts relatively few votes, it is difficult to pinpoint his/her ideal point. Poole and Rosenthal's (2007, 32-33) DW-NOMINATE estimates included all legislators who cast at least 25 votes in a term of Congress. This generally presents no problem for members of Congress, but the number of votes used to estimate presidents' ideal points can become an issue. The low volume of presidential requests has precluded estimating the ideal points of several members of Congress who became presidents: William Harrison, Millard Fillmore, Andrew Johnson, Rutherford Hayes, and James Garfield.
Additionally, whether presidents announce positions on hotly contested legislative issues or avoid taking controversial stands appears to be a matter of political strategy. Presidents do not take positions on a random sample of votes; they choose their legislative priorities selectively. Some presidents made very few requests to Congress while others frequently made their preferences known. This possibility of selection bias is problematic insofar as it results in some presidents appearing more partisan than otherwise suggested and other presidents seeming more moderate than their records would indicate. According to Treier (2010), Presidents Clinton and G. W. Bush tended to reserve requests for exceptionally controversial issues. As a result, presidents appear more partisan based on their legislative requests than a broader sample of their legislative behavior suggests. Recent research on President Hoover also suggests that the president's request record may be misleading compared to his entire legislative record. (14) Presidents Clinton, G. W. Bush, and Hoover never served in Congress, but these examples suggest broadly sampling presidential legislative activity yields more accurate ideal point estimates than limiting the analysis to votes presidents requested.
To address the problems of incomplete and possibly biased estimates of presidents' ideal points, I identify thousands of new bridging observations. The protocol for coding presidents' positions on roll call votes, detailed in the Appendix, considerably enriches the data available to estimate presidential ideal points. I assume that presidents decide whether to support the yea or nay position on a roll call vote before Congress by comparing the utility of the yea outcome to that of the nay outcome. (15) The president's legislative decisions reflect, in part, his personal preferences, but are also the product of political considerations. As discussed above, the presidency imposes certain constraints on the president but may also enable him to assert his preferences more forcefully. The president's ideal point offers a nice summary of his attempt to maximize utility. Because the president's making a request may change how members of Congress vote, it is important to also consider cases where the president makes a decision after Congress votes without having previously disclosed the administration's position. With respect to the presidents who formerly served in Congress, the CQ and Swift et al. (2001) identified presidential requests on 5,684 congressional roll call votes (1,315 of which were made by LBJ). I am able to code these presidents' positions on 4,501 additional roll call votes. The gains are particularly strong for historic-era presidents.
I use the DW-NOMINATE scaling method development by Poole and Rosenthal (2000; 2007) to recover ideal points from the entire roll call voting record of Congress. This scaling method is particularly useful because it does not require the analyst to identify the ideological content of votes ahead of time; it allows the individual's voting record to speak for itself as much as possible. Additionally, this method estimates ideal points over time and across chambers on the same scale, allowing the analyst to compare legislators and presidents from different time periods. Although I focus on the select group identified in Table 1, I estimate the ideal points of all senators and representatives in order to establish a consistent frame of reference.
Before analyzing the ideal point estimates, let us consider their accuracy and validity. The resulting estimates of presidents' ideal points explain a very high percentage of the legislative records of U.S. presidents. My estimates of presidential ideal points classify, on average, 86.69% of presidential legislative decisions correctly. The successful classification rate for presidential legislative behavior is, in fact, slightly higher than it is for the average member of Congress (85.55%). This means that a high percentage of presidents' requests for legislative action, nominations, treaties, bill signings, and vetoes can be explained in terms of presidential ideal points. I compare my estimates (first dimension only) to other prior estimates, specifically those of Bailey (2007; 2013) and Wood (2009) to assess validity. The current estimates of presidential ideal points correlate highly with Bailey's measure (0.953) as well as (the inverse of) Wood's measure of presidential liberalism (0.664).16 17 Having estimated these ideal points, I analyze whether these ideal points change when a legislator moves into the White House.
Estimating the Effect of the Presidency on Presidents
To analyze whether the presidency moderates the president, let [president.sub.i] represent a president's ideal point as president and [memcong.sub.i] his ideal point as a member of Congress. In its most basic form, the value of [president.sub.i] is simply a linear function of [memcong.sub.i].
[president.sub.i] = a + b[memcong.sub.i] + ei (1)
The moderation hypothesis generates the expectation that the slope, b, will be less than 1; in contrast, the amplification hypothesis gives rise to the expectation that the slope will be greater than 1. Under the null hypothesis, the presidency does not affect the ideal point of its occupant; the slope coefficient equals 1.
I focus on first-dimension estimates of these individuals' ideal points to determine whether the constraints of the office pulled its occupant away from partisan extremes. Following my primary results, I analyze and discuss ideal point change in two dimensions.
Given the small sample size, the opportunity to add other explanatory variables is limited. As discussed above, however, it is important to consider whether the effect of the presidency on individual preferences has changed over time. I evaluate this possibility by interacting [memcong.sub.i] with a time-period indicator, modern. While the dividing line between historic and modern eras of the presidency is subject to debate, FDR is widely considered the first modern president (Lowi et al. 2012, chap. 7; Greenstein 1978). (18) If the effect of the presidency on preferences varies between the historic and modern eras, the slope coefficient [b.sub.2] in the equation below will be significantly different than zero.
[president.sub.i] = a + [b.sub.1] [memcong.sub.i] + [b.sub.2] [memcong.sub.i] [modern.sub.i] + [b.sub.3] [modern.sub.i] + [e.sub.i] (2)
Figure 2 plots the observations, comparing ideal point estimates of members of Congress to those of presidents. Fifteen cases support the moderation hypothesis; seven cases support the amplification hypothesis. One case (Van Buren) does not support either theory of presidential representation. The left/lower and right/upper sides of Figure 2 should not be interpreted as "liberal" and "conservative" because the substantive issues that define American politics have changed significantly over time.
I employ ordinary least squares regression analysis to further evaluate the data. The results of this analysis are reported in Table 2. Model 1 estimates Equation 1; Model 2 incorporates the interaction of institutionalization as specified in Equation 2.
The estimated slope coefficient in Model 1 (0.598) is significantly less than 1. This indicates that the presidency on the whole has tended to moderate presidents. Although the Model 1 intercept (0.027) gives the impression that members of Congress become slightly more conservative as presidents, the intercept term is not statistically different than zero.
Model 2 introduces the possibility that the two contrasting theories of presidential representation described above find support in different time periods by interacting the individuals' ideal point as a member of Congress with an indicator for modern presidents. The interaction term ([b.sub.2]) is significantly different than zero, suggesting that the effect of the presidency interacts with institutionalization. Model 2 indicates that the presidency substantially moderated the preferences of historic-era presidents ([b.sub.1] = 0.424) but has amplified the ideal points of modern-era presidents (the slope in modern era, [b.sub.1] + [b.sub.2] = 1.400, is significantly greater than one [no preference change] at the 95% confidence level). (19) The modern presidency not only fails to moderate the views of those holding the office, but appears to cause the office holder to adopt more extreme views than he held while serving in Congress.
Models 3 and 4 assess the possibility that presidential moderation in the historic era (or amplification in the modern era) reflects changes in Congress rather than effects of the presidency. These models estimate the difference in differences, comparing, for example, Obama's ideal point as senator minus the congressional median (at that time) to his ideal point as president minus the congressional median. (20) This approach to measuring [president.sub.i] and [memcong.sub.i] should control for the relative change in the ideological position of
Congress that occurs over time. (21) The results of this approach are consistent with the results of Models 1 and 2.
Figure 3 plots the observed data points and estimated regression lines (based on Model 2). Despite the small sample size, the figure bears out a dramatic change in the nature of presidential representation from the historic era to the modern era.
During the historic era, the presidency appears to have compelled presidents to adopt a unique perspective on policy issues. The members of Congress who became president during this time period often expressed relatively extreme positions while serving in Congress, but they significantly moderated their preferences while serving as president. During the modern era, however, the evidence suggests that members do not moderate their political preferences after taking the presidential oath. Rather, the modern presidency seems to amplify the partisan preferences that presidents displayed in Congress.
It is also worth noting that the ideal points of congressmen who became president during the historic era were not substantially less extreme than their counterparts in the modern era. (22) Although party activists have played a greater role in nominating presidential candidates since World War II (Caeser 1979; Keech and Matthews 1976; Reiter 1985), members of Congress with relatively extreme preferences have become presidents throughout American history. Historic presidents were not more intrinsically concerned with national interests than are modern presidents. To blame individual presidents for an institutional problem may be misguided and unlikely to restore the style of leadership originally envisioned by the Framers.
Discussion of Potential Limitations
Given the limited sample size, the prospects for adding observations to this analysis deserve mention. (23) Washington, Adams, and Jefferson all served in the Continental Congress (subsequently renamed the Congress of the Confederation), which produced a roll call voting record (Lord 1984). If we assume that delegates to the Continental Congress who subsequently served in the U.S. Congress maintained fixed ideal points that bridge institutions, we can estimate ideal points for the first three presidents as legislators. (24) Obviously, we cannot proceed with the assumption that these presidents maintained fixed ideal points under both governments because that is the proposition to be tested. Although it would be useful to add three observations to this analysis, the Continental Congress and the U.S. presidency are not simply different branches of government, they are entirely different governments. (25)
What about presidents who served as governors? Modern chief executives have increasingly emerged from governor's offices. It may be possible to analyze whether preferences change moving from a Governor's Mansion to the White House. (26) This would improve the present analysis, but is unlikely to alter the central results. The modern-era governors missing from this analysis were not moderate presidents. FDR (-0.406) and Carter (-0.491) rank among the nation's most liberal presidents; Reagan (0.549) and G. W. Bush (0.564) two of its most conservative. While it is conceivable that these presidents were even more liberal or conservative as governors, the available qualitative evidence suggests the presidency may have increased these governors' partisan leanings. (27)
As discussed above, the dividing line between the historic and modern eras of the presidency is disputed. Although FDR is widely seen as the first modern president, some contend that the modern presidency began as early as the late 1800s with the reelection of President Cleveland. Only two presidents in this sample served in the period where scholars dispute the emergence of the modern presidency: McKinley and Harding. I consider them historic-era presidents, but it is not unreasonable to view them as modern figures, particularly if we view either McKinley's predecessor, Cleveland, or Harding's predecessor, Wilson, as the first modern president. If we classify Harding or McKinley as modern presidents, the conclusion that the presidency moderated historic-era presidents is not altered; however, it would appear that the modern presidency does not affect preferences.
Although the issues of the day change from one presidency to the next, this is probably not a significant limitation to this analysis. On average, the time between service in Congress and the White House is only eight years. This analysis does not assume that the issue content of the first dimension is the same for all observations. Here, I am comparing the individuals' records in Congress with their records in the White House (rather than making interpersonal comparisons). Whether the dominant issue is British imperialism, slavery, the New Deal, or national health care, leaders may express relatively extreme or moderate views.
Referring again to Figure 3, James Monroe appears on one extreme as a member of Congress (a staunch Jeffersonian Republican) but near center as president. Twenty-three years separate Monroe's legislative and executive service, the most of any case. I considered whether his case drives the statistical results reported here. If Monroe is omitted from the analysis, the signs and significance of the regression coefficients do not change. (28) Although his singular example is influential, the results do not depend on one case. If I exclude six other cases where more than ten years separate service in Congress and the White House, the results are consistent with Models 1 and 2 (reported in Table 2), but the standard errors of the coefficient estimates increase as the analysis is limited to 16 observations.
I also considered the analysis measuring the distance between ideal points and the point of origin in two-dimensional space with the second dimension appropriately weighted to reflect its substantive significance in roll call voting. (29) Perhaps becoming president moderates the legislators' political attachment to a particular region or on other issues that internally divide parties. If so, we might observe moderation along a secondary, rather than primary dimension of the spatial model. The results are indeterminate in two dimensions and do not clarify the one-dimension results. When preferences are measured as the diagonal distance to the origin, we observe more cases of moderation than amplification (14 compared to 9), but the correlation between preferences as member and as president is not statistically significant. (30)
I also questioned whether the subset of roll call votes used to compare presidents to members of Congress is representative of roll call votes used to measure congressional preferences. Prior works show the enlarged set of bridging observations used to estimate presidents' ideal points here are more representative of the roll call record generally and produced more reliable estimates than do scaling request votes only (Edwards 2014; Treier 2010). While broader samples of presidents' legislative records are more comparable to voting in Congress than narrow samples of request votes, there still arises the issue of whether the difference between the individual's ideal point as president and as a member of Congress is a result of the presidency and not the different agendas of these institutions. Under certain conditions, the majority party in Congress can control what votes reach the floor (Cox and McCubbins 2005). Additionally, research on voting in Congress indicates that roll call votes on procedural matters are often more ideologically divisive than final passage votes (Cox and Poole 2002). (31) I examine the impact of varying agendas by comparing ideal point estimates obtained by scaling all roll call votes (the legislator's agenda) to those obtained by scaling only votes on which one can ascertain the president's position (the president's agenda). If both agendas yield similar ideal point estimates, one might attribute moderation or amplification to the presidency rather than the difference in agenda. (32) Further analysis of this issue is justified, but different agendas do not appear to yield different ideal point estimates in select terms. For example, ideal point estimates of the 112th House obtained from the votes on which one can ascertain President Obama's position correlate to those obtained from all roll call votes at 0.991.
Finally, it is reasonable to consider whether factors such as divided government or House-Senate differences might explain the changes in political behavior analyzed here. During periods of divided government, one might expect Congress to advance more legislation that divides the president from the majority of members of Congress; alternatively, divided government might cause the presidents to express more extreme preferences. The data do not reveal significant differences between presidents who served during periods of divided and unified government. (33) Another interesting consideration is whether there are any differences between representatives who become presidents and senators who become president. Does one group tend to moderate and other tend to amplify? Prior research on legislative behavior suggests House-Senate differences are not enough to change political preferences. (34) Similarly, the data analyzed here do not disclose significant differences between presidents who previously served in the House and in the Senate. (35)
This article analyzes the legislative records of 23 men who have served in both Congress and the White House to answer a fundamental question about the presidency: Does the presidency moderate the political preferences of the president? According to legal scholars, political scientists, and U.S. presidents, the president sees issues differently than members of Congress do and represent Middle America rather than one particular state or legislative district. Although this unique perspective thesis is widely asserted, it has rarely been subjected to empirical analysis.
My research indicates that the presidency effectively moderated U.S. presidents throughout much of American history. Consistent with the unique perspective thesis, senators and representatives who became president before FDR consistently moderated their political views after taking the presidential oath of office. During the modern era of the presidency, which most scholars associate with the growth of the executive branch during the FDR administration, the presidency has had the opposite effect. Legislators who have become president in the modern era have exhibited more extreme political preferences than they did while members of Congress. In other words, the common view that the presidency compels the president to act on behalf of the entire country is probably wrong. For the last eighty years, the opposite appears more likely true: the presidency now encourages presidents to act in a more partisan manner than they did while representing a single state or congressional district.
This research helps voters assess presidential candidates and informs our understanding of the presidency as an enduring institution. Candidate promises to moderate their past positions in order to craft bipartisan legislation and lead from the center if elected should be regarded with skepticism. Rather than compel the president to adopt a unique perspective on national affairs, the presidency now seems more likely to amplify, rather than moderate, the partisan leanings of the president. This research also helps us understand why we are witnessing the polarization of the modern presidency. Legislators who have become president in the modern era are no more extreme than those who became president before World War II. The presidency as an institution, however, is no longer restraining presidents from staking extreme political positions.
Appendix: Data Coding Protocol
I estimate presidential ideal points by coding all presidential positions on roll call votes that can be reliably ascertained. I follow these Coding Rules:
1. Presidential Positions. If the president took a clear position in support or opposition to legislation, his roll call vote mirrors his public position. To implement this coding rule, I incorporated all the presidential positions identified by CQ and Swift el al.
2. Presidential Proposals. The president votes yea on his own nominees and treaties made by him. This Coding Rule extends the method followed by CQ to all terms of Congress.
3. Vetoes. If the president vetoed a bill (including pocket vetoes) without previously coming out in favor of the bill, he votes nay on its final passage. If Congress then votes to override the president's veto, the president votes nay (to sustain his veto).
4. Bill Signings. If the president signs the bill into law without previously coming out against the bill, he votes yea on its final passage.
Treaties generally pass unanimously or with only one or two nay votes. If all legislators vote the same way, a vote does not distinguish among legislators and is not used to estimate ideal points. However, some treaty ratification votes have been heavily contested, however, and may help to identify the president's political preferences. Similarly, votes on treaties, votes to advise and consent to appointments are also unique to the Senate and frequently pass by supermajority votes. However, some nominees are hotly contested and divide the Senate. Again, rather than prejudge a category of votes, I allow the data to inform the analysis.
In addition to nominations and treaties, there are a number of relatively rare roll call votes on presidential reorganization plans, presidential uses of force, and impeachment. Prior to the Supreme Court's finding the legislative veto unconstitutional, presidents had authority to submit reorganization plans for the executive bureaucracy to Congress. Additionally, Congress has occasionally voted to authorize the president's use of force. While no president has conceded this authority to Congress, roll call votes to authorize use of force represent clear cases of Congress expressing preferences on actions supported by the president. Finally, Congress has occasionally voted on Articles of Impeachment and matters related to trial of the president in the Senate. I assume that the president opposes his impeachment and would vote against measurements that advance his impeachment.
If the president signed a bill but added a signing statement to the bill, his bill signing is still coded as voting yea (assuming he did not previously announce opposition to the bill). In general, presidents have not used signing statements to negate substantive legal provisions (like a line-item veto) and courts have not afforded them such force (Ostrander and Sievert 2013 a,b).
Additionally, I coded the few line-item vetoes of bills passed by roll call votes in the 105 Congress as yea votes rather than vetoes. President Clinton did not veto these bills and I determined his line item vetoes left the bulk of the legislation intact. The coding scheme does not allow fractional voting for partial agreement.
If a specific request conflicts with the president's subsequent decision to sign or veto the bill (which occurs rarely), his request prevails for coding purposes. There are instances where the president lobbies for defeat of a bill, but president relents and signs the bill rather than issue a futile veto message. In these cases, the nay request is coded as the president's preference and his subsequent signature, a strategic concession.
Coding for vetoes or bill signings is limited to final passage votes where Congress and the president considered legislation in the same form. Often bills and resolutions are subject to multiple roll call votes. In some cases a vote "to pass" a particular bill is different than the version eventually presented to the president. After one chamber passes a bill, the other chamber may amend it. If the other chamber passes an amended bill, the chamber that initially passed the bill may concur in the other chamber's amendment (a final passage), insist the other chamber pass its version, or convene a conference committee to resolve differences. A bill may take many different paths to the president's desk. Determining which (if any) roll call vote passed a proposal in the form it was received by the president requires careful analysis of legislative history. However, the upside of this work is that it balances the sample of requests presidents chose to make with a substantial number of decisions that Congress compelled them to make.
Although additional votes could reduce the standard errors of presidential ideal point estimates, it is important to analyze clear, expressed preferences. Extrapolating presidential position on a particular amendments or procedural motions based on inference or retrospective information starts the researcher down a slippery slope of subjective judgments.
Abramowitz, Alan I. 2010. The Disappearing Center. Engaged, Citizens, Polarization, and American Democracy. New Haven, CT: Yale University Press.
Aldrich, John H., Calvin C. Jillson, and Rick W. Wilson. 2002. "Why Congress? What the Failure of the Confederation Congress and the Survival of the Federal Congress Tell Us about the New Institutionalism." In Party, Process and Political Change in Congress, eds. Matthew McCubbins and David Brady. Stanford, CA: Stanford University Press, 315-42.
Bailey, Michael A. 2007. "Comparable Preference Estimates across Time and Institutions for the Court, Congress, and Presidency." American Journal of Political Science 51 (3): 433-48. doi: 10.im/j.1540-5907.2007.00260.x
--. 2013. "Is Today's Court the Most Conservative in Sixty Years? Challenges and Opportunities in Measuring Judicial Preferences." The Journal of Politics 75 (3): 821-34.
Bailey, Michael A., and Kelly H. Chang. 2001. "Comparing Presidents, Senators, and Justices: Interinstitutional Preference Estimation." Journal of Law, Economics, and Organization 17 (2): 477-506.
Berman, Larry. 1979. The Office of Management and Budget and the Presidency, 1921-1979. Princeton, NJ: Princeton University Press.
Berry, William D., Richard C. Fording, Evan J. Ringquist, Russell L. Hanson, and Carl E. Klarner. 2010. "Measuring Citizen and Government Ideology in the U.S. States: A Re-appraisal." State Politics & Policy Quarterly 10 (2): 117-35.
Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998. "Measuring Citizen and Government Ideology in the American States, 1960-93." American Journal of Political Science 42 (1): 327-48.
Bond, Jon R., and Kevin B. Smith. 2008. The Promise and Performance of American Democracy. 8th ed. New York: Thompson-Wadsworth.
Burden, Barry C. 2002. "United States Senators as Presidential Candidates." Political Science Quarterly 117 (1): 81-102. doi:10.2307/798095
Caeser, James W. 1979. Presidential Selection: Theory and Development. Princeton, NJ: Princeton University Press.
Carroll, Royce, Jeffrey B. Lewis, James Lo, Keith T. Poole, and Howard Rosenthal. 2009. "Measuring Bias and Uncertainty in DW-NOMINATE Ideal Point Estimates via the Parametric bBootstrap." Political Analysis 17 (3): 261-75.
Clinton, Joshua, Simon Jackman, and Douglas Rivers. 2004. "The Statistical Analysis of Roll Call Data." American Political Science Review 98 (2): 355-70.
Cohen, Jeffrey E. 2011. "Presidents, Polarization, and Divided Government." Presidential Studies Quarterly 41 (3): 504-20. doi:10.1111/j.1741-5705.2011.03884.x
--. 2012. The President's Legislative Policy Agenda, 1789-2002. New York: Cambridge University Press.
--. 2015. Presidential Leadership in Public Opinion. New York: Cambridge University Press.
Covington, Cary R., J. Mark Wrighton, and Rhonda Kinney. 1995. "A 'Presidency-Augmented' Model of Presidential Success on House Roll Call Votes." American Journal of Political Science 39 (4): 1001-24.
Cox, Gary W., and Mathew D. McCubbins. 2005. Setting the Agenda: Responsible Party Government in the US House of Representatives. New York: Cambridge University Press.
Cox, Gary W., and Keith T. Poole. 2002. "On Measuring Partisanship in Roll-Call Voting: The US House of Representatives, 1877-1999." American Journal of Political Science 46 (3): 477-89.
Cronin, Thomas E., and Michael A. Genovese. 1998. The Paradoxes of the American Presidency. 2nd ed. New York: Oxford University Press.
Dougherty, Keith, and Justin Moeller. 2012. "Constitutional Change and American Pivotal Politics." American Politics Research 40 (6): 1092-120.
Edwards, Barry C. 2014. "Putting Hoover on the Map: Was the 31st President a Progressive?" Congress & the Presidency 11 (1): 49-83.
Edwards, George C. 1989. At the Margins: Presidential Leadership of Congress. New Haven, CT: Yale University Press.
Epstein, Lee, Andrew D. Martin, Jeffrey A. Segal, and Chad Westerland. 2007. "The Judicial Common Space." Journal of Law, Economics, and Organization 23 (2): 303-25.
Eshbaugh-Soha, Matthew, and Brandon Rottinghaus. 2013. "Presidential Position Taking and the Puzzle of Representation." Presidential Studies Quarterly 43 (1): 1-15. doi:10.1111/psq.12000
Fisher, Louis. 2012. "Teaching the Presidency: Idealizing a Constitutional Office." PS: Political Science & Politics 45 (1): 17-31. doi:10.1017/S1049096511001673
Gelman, Jeremy, Gilad Wilkenfeld, and E. Scott Adler. 2015. "The Opportunistic President: How US Presidents Determine Their Legislative Programs." Legislative Studies Quarterly 40 (3): 363-90.
Greenstein, Fred I. 1978. "Change and Continuity in the Modern Presidency." In The New American Political System, ed. Anthony King. Washington, DC: American Enterprise Institute, 451-81.
Kagan, Elena. 2000. "Presidential Administration." Harvard Law Review 114: 2245-385.
Keech, William R., and Donald R. Matthews. 1976. The Party's Choice. Washington, DC: Brookings Institution.
Kernell, Samuel, and Laurie L. Rice. 2011. "Cable and the Partisan Polarization of the President's Audience." Presidential Studies Quarterly 41 (4): 693-711. doi:10.1111/j.1741-5705.2011.03910.x
King, Gary, and Lyn Ragsdale. 1988. The Elusive Executive: Discovering Statistical Patterns in the Presidency. Washington, DC: Congressional Quarterly.
Kleinerman, Benjamin A. 2014. "The Constitutional Ambitions of James Madison's Presidency." Presidential Studies Quarterly 44 (1): 6-26. doi:10.1111/psq.12085
Klinghard, Daniel P. 2005. "Grover Cleveland, William McKinley, and the Emergence of the President as Party Leader." Presidential Studies Quarterly 35 (4): 736-60. doi:10.2307/27552726
Krause, George A., and David B. Cohen. 1997. "Presidential Use of Executive Orders, 19531994." American Politics Research 25 (4): 458-81.
Lebo, Matthew J., and Andrew J. O'Geen. 2011. "The President's Role in the Partisan Congressional Arena." The Journal of Politics 73 (3): 718-34.
Lee, Frances E. 2009. Beyond Ideology: Politics, Principles, and Partisanship in the US Senate. Chicago: University of Chicago Press.
Levy, Leonard Williams. 2000. Original Intent and the Framers' Constitution. New York: Rowman & Littlefield.
Lord, Clifford L. 1984. Roll Calls of the Continental Congresses and the Congresses of the Confederation, 1777-1789. Inter-university Consortium for Political and Social Research (ICPSR) [distributor].
Lowi, Theodore J. 1985. The Personal President: Power Invested, Promise Unfulfilled. Ithaca, NY: Cornell University Press.
Lowi, Theodore J., Benjamin Ginsberg, Kenneth A. Shepsle, and Stephen Ansolabehere. 2012. American Government: Power and Purpose. New York: W.W. Norton.
McCarty, Nolan M. 2009. "Presidential Vetoes in the Early Republic: Changing Constitutional Norms or Electoral Reform?" The Journal of Politics 71 (2): 369-84.
McCormick, Richard P. 1982. The Presidential Game: The Origins of American Presidential Politics. New York: Oxford University Press.
MoJo News Team. "Full Transcript of the Mitt Romney Secret Video," Mother Jones, 19 September 2012.
National Public Radio. 2012. "Transcript And Audio: Third Presidential Debate." http://www.npr. org/2012/10/22/163436694/transcript-3rd-obama-romney-presidential-debate (last modified October 22, 2012).
Neustadt, Richard E. 1991. Presidential Power and the Modern Presidents: The Politics of Leadership from Roosevelt to Reagan. New York: Free Press.
Nichols, David K. 2010. The Myth of the Modern Presidency. University Park, PA: Penn State Press.
Nzelibe, Jide. 2005. "The Fable of the Nationalist President and the Parochial Congress." UCLA Law Review 53: 1217-74.
Ostrander, Ian, and Joel Sievert. 2013a. "The Logic of Presidential Signing Statements." Political Research Quarterly 66 (1): 141-53.
--. 2013b. "What's So Sinister about Presidential Signing Statements?" Presidential Studies Quarterly 43 (1): 58-80.
Pika, Joseph A., and John Anthony Maltese. 2013. The Politics of the Presidency. Thousand Oaks, CA: Sage/CQ Press.
Poole, Keith T., and Howard Rosenthal. 2000. Congress: A Political-Economic History of Roll Call Voting. New York: Oxford University Press.
--. 2001. "D-NOMINATE after 10 Years: A Comparative Update to Congress: A Political Economic History of Roll-Call Voting." Legislative Studies Quarterly 26: 5-29.
--. 2007. Ideology and Congress: A Political-Economic History of Roll Call Voting. New York: Transaction Publishers.
Reiter, Howard L. 1985. Selecting the President: The Nominating Process in Transition. Philadelphia: University of Pennsylvania Press.
Rhodes, Jesse H. 2014. "Party Polarization and the Ascendance of Bipartisan Posturing as a Dominant Strategy in Presidential Rhetoric." Presidential Studies Quarterly 44 (1): 120-42. doi: 10.1111/psq.12090
Rice, Laurie L. 2010. "Statements of Power: Presidential Use of Statements of Administration Policy and Signing Statements in the Legislative Process." Presidential Studies Quarterly 40 (4): 686-707.
Rudalevige, Andrew. 2012. "The Contemporary Presidency: Executive Orders and Presidential Unilateralism." Presidential Studies Quarterly 42 (1): 138-60.
Selinger, Jeffrey S. 2014. "Making Sense of Presidential Restraint: Foundational Arrangements and Executive Decision Making before the Civil War." Presidential Studies Quarterly 44 (1): 27-49. doi:10.1111/psq.12086
Shor, Boris, and Nolan McCarty. 2011. "The Ideological Mapping of American Legislatures." American Political Science Review 105 (3): 530-51.
Shull, Steven A. 2000. Presidential-Congressional Relations: Policy and Time Approaches. Ann Arbor: University of Michigan Press.
Silva, Ruth C. 1948. "State Law on the Nomination, Election, and Instruction of Presidential Electors." American Political Science Review 42 (3): 523-29.
Steger, Wayne P. 1997. "Presidential Policy Initiation and the Politics of Agenda Control." Congress & the Presidency: A Journal of Capital Studies 24 (1): 17-36.
--. 2006. "Stepping Stone to the White House or Tombstone on Presidential Ambition: Why Senators Usually Fail When They Run for the White House." American Review of Politics 27: 45-70.
Swift, Elaine K., Robert G. Brookshire, David T. Canon, Evelyn C. Fink, John R. Hibbing, Brian D. Humes, et al. 2001. Database of United States Congressional Historical Statistics, 1789-1989. Ann Arbor, MI: Inter-university Consortium for Political and Social Research.
Thurber, James A. 2013. Rivals for Power: Presidential-Congressional Relations, Fifth Edition. Lanham, MD: Rowman & Littlefield.
Treier, Shawn. 2010. "Where Does the President Stand? Measuring Presidential Ideology." Political Analysis 18 (1): 124-36. doi:10.2307/25791996
Tulis, Jeffrey. 1987. The Rhetorical Presidency. Princeton, NJ: Princeton University Press.
Walsh, Kenneth T. 2012. "Obama: 'You Represent the Entire Country.'" U.S. News and World Report, September 19.
Wayne, Stephen. 2009. "Legislative Skills." In The Oxford Handbook of the American Presidency, eds. George Edwards and Willaim Howell. New York: Oxford University Press, 311-337.
Weinberg, Micah. 2010. "Measuring Governors' Political Orientations Using Words as Data." State Politics & Policy Quarterly 10 (1): 96-109.
Wilkins, Vicky M., and Garry Young. 2002. "The Influence of Governors on Veto Override Attempts: A Test of Pivotal Politics." Legislative Studies Quarterly 27 (4): 557-75.
Wood, B. Dan. 2009. The Myth of Presidential Representation. New York: Cambridge University Press.
(1.) I use the terms "president" or "presidents" to refer to an individual or individuals who have held that office and the term "presidency" in reference to the institution that endures from one president to the next. I use the term "moderate" in a relative sense only. Whether the institution moderates presidents is distinct from asking whether presidents are moderate. It is possible that the president is more moderate as president than he was as a member of Congress, but not moderate compared to the median voter.
(2.) Of course, divining the Framers' intent from the historical record is a difficult task and presidency scholars continue to debate what style of representation the Framers expected the president to exercise (Pika and Maltese 2013, chap. 1; Fisher 2012). The restraint exercised by President Madison provides a compelling case study of how the chief architect of the Constitution expected the president to act (Kleinerman 2014).
(3.) The Electoral College was also designed to prevent demagogues from coming to power. The original plan for electing presidents empowered a relatively small group of electors to select the chief executive (Caeser 1979, chap. 1). In Federalist Paper No. 68, Hamilton argued that it was important to delegate this important task to select, small groups of electors and allow them to deliberate in their home states at different times (rather than allow the impressionable masses to make such an important decision). The presidential selection method outlined in Article II was intended to check fanaticism and factionalism.
(4.) The president's remarks were in response to a candid assessment Governor Romney shared with his campaign contributors: "There are 47 percent of the people who will vote for the president no matter what.... And so my job is not to worry about those people--I'll never convince them that they should take personal responsibility and care for their lives" (MoJo News Team 2012).
(5.) We should take note of another perspective on institutionalization. Institutionalization of the executive branch may constrain and moderate the president's legislative decisions. The president gives up some degree of autonomy and control to gain influence and expertise. As King and Ragsdale (1988, 16) memorably stated: "[T]he president is only one participant in the presidency." Executive initiatives are often written by professional bureaucrats, not the president or members of his inner circle (Rudalevige 2012). Along these lines, Krause and Cohen (1997) find that presidents issued executive orders in an opportunistic fashion during the developmental phases of the presidency, but have been constrained by internal forces in the modern era.
(6.) The changing patterns of polarization of political parties in Congress may similarly affect the president (Cohen 2011).
(7.) The characteristics of the Electoral College that Hamilton found so appealing--its small size, expertise, detachment--are all absent from our modern system of electing presidents. Hundreds of millions of ordinary citizens now participate in presidential elections; presidential campaigns feature passionate, partisan appeals by candidates and culminate in a single day of voting. Although the Electoral College still exists, electors are not afforded any discretion to identify enlightened statesmen (McCormick 1982; Silva 1948).
(8.) A number of these presidents did not actually serve in the White House. "White House" is used here as a figure of speech to counterbalance Congress.
(9.) It may be noted that no model predicts a negative slope coefficient; no extant theory leads one to believe that a member of Congress will invert his preferences after becoming president.
(10.) The theories of presidential representation outlined here do not suggest that the presidency permanently alters the president's personal ideological beliefs. Indeed, the president's personal beliefs are likely firmly set from a lifetime of political socialization; in this analysis, his ideal point as member of Congress is an independent variable that I treat as exogenous. This research focuses on behavior, not beliefs.
(11.) This analysis compares an individual's record as president to his record as member of Congress (although these records were compiled at different times). The labels "liberal" and "conservative" help to illustrate the hypotheses, but should not be used to compare historic and modern presidents.
(12.) For a critique of the belief that Congress is a stepping stone to the presidency, see Burden (2002) and Steger (2006).
(13.) See, e.g., Bailey (2007; 2013) and Bailey and Chang (2001), which use executive and legislative branch positions on Supreme Court cases as bridge observations to compare preferences of all three branches; Epstein et al. (2007), which utilizes "unconstrained confirmed nominees to the Supreme Court" to calibrate judicial ideal points to NOMINATE scores.
(14.) In Hoover's case, the president's legislative requests appear more moderate than his full record of legislative requests, nominations, signings, and vetoes reveals (Edwards 2014). The moderate nature of Hoover's legislative requests is consistent with the relatively passive approach of historic presidents.
(15.) This assumption is admittedly simplistic and omits countless details of the political process. Some may object to these coding rules, particularly Rules Three and Four, based on game theories, which show that presidents will sign bills they actually oppose (to avoid congressional overrides) and veto bills they actually support (in order to force Congress to pass an even more attractive bill). These objections will be discussed as potential limitations to this analysis. It should be noted at this point, however, that presidents frequently make their opposition to a bill known prior to (perhaps reluctantly) signing it into law. In these cases, the coding protocol followed his public position of opposition.
(16.) These measures are not all on the same metric, so I standardize each using its observed mean and standard deviation and compare each series over the years for which it is presently available. Because Wood measured presidential liberalism, I invert his measure so it corresponds to the left-right scheme employed here.
(17.) The correlation of my estimates with Wood's scores is lower than it is with Bailey's score because Wood measured presidential liberalism annually whereas I produce one estimate per president (as does Bailey). The correlation between Bailey's and Wood's estimates is 0.629.
(18.) For an extensive critique of the view that FDR was the first modern president, see Nichols (2010). Some have argued that Grover Cleveland initiated the modern phase of partisan leadership (Klinghard 2005). Others argue that Wilson was the first modern president (e.g., Tulis 1987; Caeser 1979); this debate does not significantly affect this research because Wilson never served in Congress. Harding, Wilson's successor, did serve in Congress, but governed in a manner consistent with historic, rather than modern presidents. Harding's successors, Coolidge and Hoover, also never served in Congress. I consider the impact of different division between modern and historic eras following my main results.
(19.) The standard error of the slope coefficient for modern presidents (b1 1 b2) is 0.227.
(20.) This measurement approach is complicated by the fact that members and presidents serve over multiple terms of Congress and the median can change from one term to the next and/or vary between chambers of Congress. To keep things as simple and straightforward as possible, I estimated the chamber median as the median member of Congress (combining chambers) while the individual served in Congress or the White House.
(21.) One might think the difference between the president/member and the median of his own party would also offer a controlled comparison. The problem is that a president moving toward the median of his party in Congress could be either moderating or amplifying his partisan preferences so the difference from party median is not a reliable measure for this research. Consider, for example, a liberal Democrat (left of his party while in Congress) who moves toward his party after he becomes president. He would be moving to the right (toward both his party as well as the Republican median). A conservative Democrat who moves toward his party would be moving the opposite direction.
(22.) The mean absolute value of ideal points in the historic group is 0.330 compared to 0.284 in the modern group; the difference of these means is not statistically significant.
(23.) Forty-three different men have served as president. Twenty-three are analyzed here. What about the other twenty presidents? Presidents William Harrison and James Garfield served in Congress before being elected, but neither compiled sufficient legislative records in the White House to estimate their ideal points as president. Presidents Washington, Adams, and Jefferson served in the Continental Congress rather than U.S. Congress. Six presidents held no prior elective office: Taylor, Grant, Arthur, Taft, Eisenhower, and Hoover. Finally, nine presidents--Cleveland, T. Roosevelt, Wilson, Coolidge, FDR, Carter, Reagan, Clinton, and G. W. Bush--were governors.
(24.) Washington served in the Continental Congress on behalf of Virginia prior to his appointment as commander in chief of the Continental Army. Washington's legislative service predates the time period covered by Lord (1984) and his roll call votes in the Continental Congress are not recorded. The data set includes 34 votes by John Adams and 76 votes by Thomas Jefferson. Accordingly, scaling the ideal points of these future presidents from their roll call records in the Continental Congress is statistically possible (Aldrich, Jillson, and Wilson 2002; Dougherty and Moeller 2012).
(25.) It is not clear, for example, that the vote of a delegate to the Continental Congress had the same effect as the vote of a senator or representative. The Continental Congress used a state-unit voting system in which the state's vote was determined by a majority vote of that state's delegates. But see Dougherty and Moeller (2012; joint scaling of Continental Congress and U.S. Congress).
(26.) Shor and McCarty (2011) mapped ideal points of state legislators using roll call votes. Although they did not estimate governors' ideal points, it might be possible to do so using governors' legislative decisions, governors' speeches, or an estimate derived from members of their party (Berry et al. 1998; Berry et al. 2010; Weinberg 2010; Wilkins and Young 2002).
(27.) FDR and Carter, for example, both campaigned on their records as fiscally conservative governors. Only if one assumes that, as governors, Reagan and G. W. Bush were far more conservative than any member of Congress who became president, and Clinton, Carter, and FDR, as extreme as Monroe was as senator would one conclude that the modern presidency moderates presidents. A more plausible result of estimating these governors' ideal point would be failing to reject the null hypothesis that the presidency does not change modern presidents' preferences.
(28.) As one might expect, the slope values change if Monroe is omitted; the slope coefficient in Model 1 changes from 0.598 to 0.684, the slope coefficient in Model 2 (not interacted) changes from 0.424 to 0.470.
(29.) The DW-NOMINATE scaling routine estimates a second-dimension weight that maximizes geometric mean probability (GMP) of observed roll call votes with weight of the first dimension set equal to 1 (Carroll et al. 2009, 267-68; Poole and Rosenthal 2001, 25-26). The spatial model used here generates probabilities of voting yea or nay based on the relative positions of actors' ideal points and vote cutting lines. The GMP is the exponential of the average log-likelihood of observed votes; it is a method of weighting some errors more heavily than others in order to estimate a model that makes decisive predictions and avoids gross errors, as opposed to simply maximizing the number of votes correctly classified (Poole and Rosenthal 2007, 38). Because the relative weight of actors' alignment along a second dimension changes the distance between their ideal points and vote cutting lines as well as the GMP of their observed votes. The DW-NOMINATE scaling routine utilized here automatically calculates a second-dimension weight that maximizes GMP. For the present analysis of congressional terms 1 through 112 with the addition of presidential records, the estimated second-dimension weight equals 0.4438.
(30.) Two-dimensional analysis may be problematic given the variation in the subject matter content of the second dimension between the time these individuals served in Congress and served as president. Some presidents (e.g., Monroe, J. Q. Adams, Tyler) reverse signs in the second dimension (all had negative second dimension scores as members of Congress and positive second dimension scores as president). One observes a number of cases that are not explained by any extant theory of presidential representation; these cases are relatively far from the cluster of observations and therefore weigh heavily on statistical analysis.
(31.) If modern presidents avoid divisive procedural votes that members of Congress must make, the difference in agendas would tend to make modern presidents appear less partisan than members of Congress are and biased the analysis against the results reported here.
(32.) This is similar to the issue of whether a subset of votes yields the same matrix of ideal points as the full set of votes. For example, one might analyze whether foreign policy votes yield the same partisan divisions that domestic policy votes do.
(33.) To assess the effect of divided government on presidents, I determine whether each president's party held a majority of seats in both chambers of Congress for most of his administration. So, for example, Truman is considered serving during unified government even though Democrats lost control of Congress for two years of his administration.
(34.) The roll call voting records of members of the House of Representatives who graduated to the U.S. Senate are informative on this point. One might suspect they would moderate their policy preferences somewhat because senators represent an entire state rather than a homogenous district, but their roll call voting records do not support this view. "Contemporary members of Congress," Poole and Rosenthal (2007, 97) found, "do not adapt their positions during their careers but simply enter and maintain a fixed position until they die, retire, or are defeated in their ideological boots."
(35.) If the president served in both the House and Senate, I classified his prior service based on the longer term of service.
Barry Edwards, Ph.D., J.D., is a lecturer in the Department of Political Science at the University of Central Florida. He received his Ph.D. from the University of Georgia, law degree from NYU School of Law, and B.A. from Stanford University. AUTHOR'S NOTE: I would like to thank Keith Poole for his assistance with this research and George Edwards (not related) for his editorial guidance.
Caption: FIGURE 1. Hypotheses about Presidential Representation
Caption: FIGURE 2. Idea! Points as President and Member of Congress
Caption: FIGURE 3. Comparing Model Results for Historic Era and Modern Presidents
TABLE 1 Presidents Who Also Served in Congress Name (State) Representative Senator President Madison (VA) 1789-1797 1809-1817 Monroe (VA) 1790-1794 1817-1825 Adams, J. Q. (MA) 1833-1848 1803-1808 1825-1829 Jackson (TN) 1796-97 1797-98, 1823-25 1829-1837 Van Buren (NY) 1821-1828 1837-1841 Harrison, W. (OH) 1816-1819 1825-1828 1841 (1 mo.) Tyler (VA) 1827-1836 1841-1845 Polk (TN) 1825-1839 1845-1849 Fillmore (NY) 1833-35, 1837-43 1850-1853 Pierce (NH) 1833-1837 1837-1842 1853-1857 Buchanan (PA) 1834-1845 1857-1861 Lincoln (IL) 1847-1849 1861-1865 Johnson, A. (TN) 1843-1853 1857-62, 1875 1865-1869 Hayes (OH) 1865-1867 1877-1881 Garfield (OH) 1863-1881 1881 (7 mo.) Harrison B. (IN) 1881-1887 1889-1893 McKinley (OH) 1877-1891 1897-1901 Harding (OH) 1915-1921 1921-1923 Truman (MO) 1935-1945 1945-1953 Kennedy (MA) 1947-1953 1953-1960 1961-1963 Johnson, L.B. (TX) 1949-1961 1963-1969 Nixon (CA) 1947-1950 1950-1953 1969-1974 Ford (MI) 1949-1973 1974-1977 Bush, H.W. (TX) 1967-1971 1989-1993 Obama (IL) 2005-2008 2009-present Name (State) Yrs. Between Madison (VA) 12 Monroe (VA) 23 Adams, J. Q. (MA) 17, 4 Jackson (TN) 4 Van Buren (NY) 9 Harrison, W. (OH) 13 Tyler (VA) 5 Polk (TN) 6 Fillmore (NY) 7 Pierce (NH) 11 Buchanan (PA) 12 Lincoln (IL) 12 Johnson, A. (TN) 12, 6 Hayes (OH) 10 Garfield (OH) 0 Harrison B. (IN) 2 McKinley (OH) 6 Harding (OH) 0 Truman (MO) 0 Kennedy (MA) 1 Johnson, L.B. (TX) 2 Nixon (CA) 16 Ford (MI) 1 Bush, H.W. (TX) 18 Obama (IL) 1 Note: J. Q. Adams and A. Johnson returned to Congress four and six years, respectively, after their presidencies. TABLE 2 Regression Models of Presidents' Ideal Points Model 1 Model 2 Model 3 Model 4 Ideal Point as MC 0.598 *** 0.424 *** 0.446 ** 0.293 * (0.123) (0.107) (0.128) (0.107) Modern Pres. 0.976 *** 1.085 *** Ideal Point as MC (0.250) (0.278) Modern Presidents 0.095 0.165 (0.079) (0.092) Intercept 0.027 0.017 0.049 0.017 (0.043) (0.042) (0.054) (0.050) N 23 23 23 23 R2 0.531 0.740 0.366 0.661 Adjusted R2 0.508 0.699 0.336 0.607 F Statistic 23.76 18.01 12.15 12.34 p Value for F <0.001 <0.001 0.002 <0.001 Notes: Dependent variable is individual's ideal point as president; in Models 3 and 4, ideal points measured as difference from congressional medians; MC = member of Congress. * p < .05; ** p < .01; *** p < .001.
|Printer friendly Cite/link Email Feedback|
|Publication:||Presidential Studies Quarterly|
|Date:||Mar 1, 2017|
|Previous Article:||Michelle Obama: First Lady, American Rhetor.|
|Next Article:||Risk and Presidential Decision-Making: The Emergence of Foreign Policy Crises.|