Presidents, polarization, and divided government.

Cohen, Jeffrey E.

For several decades the American political system has been polarizing along partisan, ideological, and issue lines at both the mass and elite levels. One implication of polarization is the disappearance of the political middle: Moderates have almost completely vanished in Congress, for example (Fleisher and Bond 2004), and the political activists stratum, that is, party activists, candidates for office, and office holders, consist almost entirely of consistent liberals in the Democratic Party and consistent conservatives in the Republican-Party. By historical and American standards, polarization has given rise to policy extremism, in the sense of vanishing moderates. (1)

This article investigates the implications of the partisan polarization and extremism of the past two decades on the American presidency. Although the presidency is a key political institution, and often at the center of important political changes and developments, rarely has the literature on polarization considered the presidency. As Layman, Carsey, and Horowitz (2006) state, "the work on growing polarization between the parties in government has focused largely on Congress" (87). (2) The literature on presidential-congressional relations and polarization rarely goes beyond the general point that polarization makes a difficult relationship even more problematic for presidents (Andres 2005; Binder 2003; Edwards and Barrett 2000; Fleisher and Bond 2000a, b; Pomper 2003; Sinclair 1997, 2000, 2002; Theriault 2008; but see Beckmann and Kumar 2010). For instance, almost no attention has been paid to the effects of polarization on presidential policy choice and the implications of such choice on presidential success with Congress. (3)

Cameron (2002) offers one of the few extended discussions of the implications of polarization on the presidency, yet Cameron's essay aims mainly to set an agenda for research. As Cameron argues, polarization touches more than executive-legislative relations but also deeply affects presidential relations with the media, the judiciary, the bureaucracy, and the organization of the White House staff. But, "[p]residential scholars are just beginning to grasp these changes." (Cameron 2002, 647). To date, few have picked up on the research directions set out by Cameron. (4)

This article looks at the implications of polarization on presidential policy choice. The polarization literature argues that the widening gap between the parties should lead to policy extremism as opposed to moderation. Has the presidency, like Congress, also become more extreme, that is, decidedly liberal or conservative, as polarization has increased? I test two competing explanations for extremism in presidential policy, the party activist theory and the congressional context theory. The first theory argues that the reforms in election processes, in particular campaign finance and nominations, increased the power of party activists in the party processes. Consequently, liberals captured the Democratic Party and conservatives the Republican. Presidents, as agents of their party, selected by these newly powerful elements in their party, moved to the policy extremes in the post-reform era.

In contrast, the congressional context theory maintains that policy considerations in part motivate presidents: Presidents care about implemented policy for a variety of reasons detailed below. Two aspects of the congressional context affect presidential policy choice, whether their party controls Congress and the degree of party polarization. Under united government, presidents select policies close to their party center. Under divided government, presidents will moderate their policy positions, being forced to work with the opposition. But the president's ability and/or willingness to work with the opposition during divided government ebbs as polarization between the parties widens. The analysis presented below shows support for the congressional context theory but little for the party activist one.

The next two sections present these theories of presidential policy choice, followed by analysis using data from the early 1950s to the end of the twentieth century. The conclusion puts the findings into perspective and suggests directions for future research.

Party Activists and Presidential Policy Extremism

A large number of studies contend that party activists were important catalysts of the heightened party polarization of the past several decades (Aldrich 1995; Aldrich and Rohde 2001; Fiorina, Abrams, and Pope 2005; Jacobson 2000; King 1997, 2003; Layman, Carsey, and Horowitz 2006; Layman et al. 2010; Saunders and Abramowitz 2004; Shafer 2003). Here we should think of party activists broadly to include those who work in candidate campaigns, financially contribute to candidates and the parties, attend the national conventions, as well as interest groups that seek to influence the nomination and election of candidates for office.

Party activists may stimulate polarization for several reasons. One, they hold relatively extreme policy views compared to voters and rank-and-file members of parties. Two, they play a critical role in selecting the parties' candidates for office, especially in the postreform era of the early to mid-1970s. Three, they provide considerable resources for candidate's primary and general election contests. Thus, in the era of party activist influence, candidates for office and office holders will either resemble the policy extremism of activists or will be policy responsive to their views, repaying activists for their vital electoral support and/or trying to secure their support in the upcoming election.

The reforms of the early to mid-1970s heightened the influence of activists on their respective parties. For example, one consequence of the reforms was to replace party caucuses and other devices with primaries for nominating candidates. As a result, party activists supplanted traditional party leaders in selecting the parties' nominees. Owing to their critical role in the primary and nomination process (Cohen et al. 2008), activists served as gatekeepers, insuring that the party's nominees would reflect the activist strata's policy preferences. This in turn pushed the parties' nominees and office holders toward the policy extremes. (5) As the influence of activists within the parties grew, the Democratic Party moved to the left and the Republican Party to the right.

The reforms of the presidential selection process in the 1970s exposed the presidency to the new-found influence of party activists. Presidents, and presidential aspirants, accommodated to the growing influence of party activists within their parties through selection and/or conversion-adaptation effects. Selection effects mean that as party activists became increasingly important in the nomination process, politically extreme candidates would be more likely to gain the nomination than moderates. Conversion-adaptation effects mean that aspirants for the nomination would move to the political extremes to gain the support of the party activists. Mitt Romney's move to the right in the 2008 Republican presidential nomination contest illustrates this conversion-adaptation process.

The motivation for office, including renomination, link these nomination and election processes to governing, thereby affecting the policy positions that presidents take. Presidents' policy positions in office will affect their ability to be renominated. If they take positions at variance with the preferences of the party activists, they may undermine their chances for the nomination. No president seeking renomination has been denied it since the reforms of the mid-1970s, which may indicate that presidents cater to their party activists. The lukewarm, if not cold reception, that John McCain received in 2008 among conservative Republican Party activists also illustrates the linkage between governing and nomination-election politics. (6) Second-term presidents too have an incentive to cater to these party activists, even though they cannot run for the presidency again. By maintaining their "extreme" policy credentials, they insure some ability to influence who the party selects as their successor, which may be important for the president's legacy, that the party not spurn him so quickly after leaving office.

This party activist perspective leads to the hypothesis that presidents of the party reform era should be more extreme than presidents prior to those reforms. Second, presidents and presidential candidates should become increasing extreme as party activist polarization has grown from the mid-1970s to the present. Layman et al. (2010) find that party activists have increasingly polarized on ideological and policy grounds over the past 30 years. Based on this perspective, we should see presidents becoming more extreme since the party reforms of the 1970s.

One relevant issue is the speed with which the reforms led to activist influence in their parties. In the analysis below, I initially assume relatively immediate effects of the reforms on activist influence. Because the major reforms were implemented before the 1976 election, I first use this as a marker to distinguish the prereform from the postreform era. But data on polarization, for instance, the Poole-Rosenthal DW-Nominate scores for Congress, suggest that the up tick in polarization did not begin in earnest until the late 1970s, and it is not until the mid-1980s that appreciable levels of polarization become evident. Layman et al.'s (2010) data on party activists suggest a steadier increase in polarization across the 1970s, but too, it is not until the mid-1980s that we observe considerable interparty polarization. Thus, the analysis also uses other cutpoints (1980/ 1981 and 1984/1985) to distinguish the pre from the postreform era. But using these later cutpoints also attenuates the connection between the implementation of the party reforms and presidential behavior.

The Congressional Context and Presidential Policy Extremism

The party activist model assumes that presidents cater to the policy preferences of party activists as those activists became increasingly influential within their parties in the postreform era. That model assumes that presidents care primarily about gaining their party's nomination. Instead, let us begin with the alternative assumption that presidents care primarily about public policy. Because Congress is so important to the enactment and implementation of presidential policies, presidents will take the congressional environment into account when establishing their policy positions. Party activists may indirectly affect presidential policy choice through their impact on Congress. Before addressing that indirect linkage, let us turn to the direct impact of Congress on presidential policy choice.

This model begins by assuming that presidents are motivated by policy. Policy motivated presidents want to see at least some of their policy proposals enacted by Congress and implemented. Several factors may lead to this policy motivation. Presidents may view their policy proposal as a solution to a problem, feeling responsible and/or accountable to solving that problem. A president may also see policy as a means to steer the nation in a particular and desired direction. Presidents too may offer policies to repay groups and voters for their support in the past election as well as use policy to attract their support in upcoming elections. And, presidents may view the enactment and implementation of policy as a way to build an historical legacy and to improve their reelection chances. The important point is not the source of the policy as a presidential motivation but the implications of that motivation on presidential policy selection (Light 1999).

First, let us assume that we can locate a president's and a pivotal member of Congress's preferred policy in a unidimensional liberal-conservative policy space. Define the pivot as the member of Congress whose support the president needs for a policy proposal to pass. Considerable controversy in the literature on congressional policy making surrounds the identification of this pivot, whether the pivot is the median member of the chamber (Krehbiel 1998) or the party majority median, or some other member (Smith 2007). We need not enter this debate. As I show below, whether the pivot is the chamber or party median generates the same hypotheses.

Define the president's utility, U, for a policy as a function of two distances: (1) the distance from the president's preferred policy and pivotal member of Congress, either the median from the president's party or the opposition party, whichever party is in the majority and (2) the policy distance between the medians of both parties. Also assume utility declines as distances from preferred positions increase. The president faces an optimization problem, constrained by the preferences of the pivotal member of Congress as well as his party's median member.

Why should the president care about his party median when all it takes is a compromise with the congressional pivot to pass a policy? Disregard instances when a president needs voters from his party for his policy to be enacted. Even when the president does not need his party's support for passage, he will still take into account the preferences of his party. Members of the president's party may rebel if the president cohabits too often with the opposition, perhaps trying to deny him renomination for a second term. The president may also harm his party's reputation with the public if it can not share in the credit for policy enactment, and thus undermine the party's reelection chances. By working with the opposition but not his own party in building policy, the president may also allow the opposition party to gain a reputation for governance, responsibility, and bipartisanship, which may further undermine the president's party in both upcoming congressional and presidential contests. Thus, for several reasons, presidents may take into account their own party in making policy even when he does not need their votes for passage.

[FIGURE 1 OMITTED]

Figure 1 presents several schematics that will help illustrate policy behavior under four conditions, the president's party is in the majority or minority and polarization is either high or low: (1) Majority-Low Polarization, (2) Majority-High Polarization, (3) Minority-Low Polarization, and (4) Minority-High Polarization. As we will see, the policy dynamics for situations 1 and 2 wind up being the same--when the president's party controls Congress, polarization does not factor heavily into policy making.

Figure 1, Panel A presents the case when the president's party is in the majority. On that figure, M represents the pivotal member of Congress, PR is the president's preferred policy position, PP is the president's party policy position, and X the opposition party median. When the president's party is in the majority, by definition, the congressional pivot will be in the president's party. In this case, the president will negotiate with the pivotal member, M, settling on a policy, S1, the midpoint between their positions. If the congressional pivot happens also to be the presidential party median, that is M = PP, then the midpoint between PR and PP (not shown) will be the policy result. Notice that X, the minority party median, does not enter into policy production under majority government in this model, but also notice that by definition X is farther from the president's position than M, a point that will become more important later on when considering policy production under minority government.

In this example, M stands to the left of the president. The president moderates slightly to S1, but if M were to the right of the president, then the president would become more extreme, moving to the right and away from the center. With majority rule, whether the president moderates or becomes more extreme depends upon his position relative to that of his party. The more important point though is presidents need only move a slight distance under majority rule than minority rule. For instance, assume the opposition party controls Congress and the congressional pivot becomes X, instead of M. The midpoint between PR and X will be further to the left of $1; presidents will moderate more under divided than united government, all else being equal.

All else, however, is not equal and the model suggests that presidents take into account the preferences of their party when making policy. Figure 1, Panel B illustrates these dynamics, but now OP stands for the opposition party median. First, consider the case when there is no polarization between the parties, but the opposition party controls Congress. Although unrealistic, this baseline helps illustrate the dynamics involved, especially the interaction between polarization and minority status on policy production. On Figure 1 Panel B, this would mean that OP = PP, that both parties have the same policy preferences. The policy midpoint between the president, PR, and the opposition party median, OP, becomes S2.

Now assume a high degree of party polarization, with the median for the president's party located at PP. The policy midpoint between the president's party and the opposition majority party is S3, to the right of S2, the solution with no polarization. When polarization between the parties exists under minority rule, presidents must take into account the policy preferences of both the majority opposition and the minority presidential party, that is, the president must find some compromise between S2 and S3. On the figure, S4 represents that new policy solution. Notably that solution is to the right of S2, that is, presidents move a shorter distance to $4 than S2. Under minority government, polarization constrains the distance that presidents will move toward the opposition. Presidents moderate less under minority government in the presence of high versus low levels of party polarization. And party polarization is not relevant when the president's party controls Congress as the president only has to negotiate with his own party; polarization between the parties does not affect that negotiation. This model leads to two hypotheses:

H1: Presidents will be more moderate during divided than united government.

H2: Presidential will moderate less during divided government as polarization between the parties increases.

Data

Testing the party activist and congressional contest theories of presidential policy choice requires measuring the degree of moderation versus extremism in presidential policy positions. We also would like data spanning the more polarized present era, as well as times characterized with lower levels of polarization, such as the 1950s-early 1970s. There are three relevant data series available for us to use: Americans for Democratic Action (ADA) scores, Poole and Rosenthal's DW-Nominate scores, and ideal point estimates. Because each has its own recommendations, as well as pitfalls, the analysis below utilizes several measures. (7)

Kenny and Lotfinia (2005), updating Zupan (1992) calculate presidential liberalism by using presidential positions on ADA votes with a series that extends from 1947, the first year that ADA selected key votes, through 2000. There are well-known issues with using ADA roll calls for creating a comparable time series, primarily agenda composition effects, in which the votes/issue across years are not comparable. Due to this problem, Groseclose, Levitt, and Snyder (1999) have developed a method to adjust the ADA scores across time to provide temporal comparability. The analysis below uses these ADA scores, adjusted as Groseclose, Levitt, and Snyder (1999) recommend, following the Kenny-Lotfinia (2005) procedure by averaging the ADA presidential scores for the House and Senate. Finally, Anderson and Habel (2007) bring the conversion scores up to date. (8)

Second, Poole and Rosenthal have calculated DW-Nominate scores for presidents. The DW-Nominate scores are well known and well studied, especially for Congress. DW-Nominate scores for presidents are calculated based on all positions that the president took while in office and thus cannot vary over time for an individual. Hence, at most we would only be able to compare presidents and not whether an individual president's policy position shifts during his time in office.

Third, Michael Bailey (2007) has calculated ideal point estimates for presidents, as well as the House, Senate, and the Supreme Court for 1951-2002. (9) Bailey's ideal point estimates are calculated for each year, allowing us an intraadministration dynamic perspective that the DW-Nominate scores do not allow.

[FIGURE 2 OMITTED]

The analysis below uses the adjusted ADA scores, the DW-Nominate, and the ideal point estimates to insure that the findings are not a result of different measurement strategies. I folded each series at their midpoints to create the three presidential extremism variables. This entails taking the absolute value of each case from the midpoint for the entire series. The natural midpoint for the Bailey ideal point estimates and DW-Nominate scores is "0". For the adjusted ADA scores, I use the actual mean of the series as the midpoint (45.6). (10) Despite differences in underlying assumptions and calculation methodologies, all three measures of presidential extremism are significantly correlated with each other, with correlations ranging from .45-.72. (11)

Figure 2 plots the three series. Because each series employs a different metric, the plot uses their z-scores. The smallest z-score (negative values) indicate moderate presidents, while large z-scores (positive values) indicate extreme presidents. A "0" z-score indicates average presidential extremism in the series. The first notable point is that in general the series travel together, although on occasion we can also see some notable divergences. For instance, the ADA score gives LBJ a moderate score in 1967, but the ideal point estimates suggests strong extremism that year.

Although it is hard to discern visually any distinct temporal trends in these data, each series hints greater extremism for postreform than prereform presidents. Table 1 presents some descriptive statistics for the three series between prereform and postreform presidents using these years to define the two eras: prereform (1951-1976), postreform (1977-2002). In each case, postreform presidents exhibit on average greater extremism than prereform presidents, but each series also indicates within-era variance, with both relatively moderate and extreme presidents serving in each era.

Analysis Issues

Before proceeding to the estimation, diagnostics revealed nonstationarity in the presidential extremism series (Table 2). (12) First differencing each produces stationary series. Also, as some of the independent variables also display nonstationarity, I use a single equation error correction model (ECM). (13) The general form of the single equation ECM model regresses the dependent variable in differenced form on the first order lag in levels of the dependent variable, plus the independent variables in both differenced and first order lags. One advantage of the single equation ECM is that it separates long- and short-term effects, long-term effects denoted by the lagged variables and short-term effects with the differenced form variables. Moreover, the single equation ECM distinguishes the long-term effects of the independent variables from equilibrium effects, indicated by the lagged dependent variable. Finally, single equation ECM allows us to use ordinary least squares.

To test the party activist hypotheses I use a prepost reform dummy, coded 0 for 1951-1976 = 0 and 1 for 1977-2002, and a counter, coded 1951-1976 = 0 and 1977 = 1 ... to 2002 = 26. The dummy tests whether postreform presidents are more extreme than prereform presidents and the counter tests whether extremism among presidents has grown during the postreform era. The estimation only uses the lagged form of these variables. Differenced forms reduce to one case equaling "1" and all other cases equaling "0." Based on the party activist theory, we should expect positive and statistically significant coefficients for the period dummy and counter.

To test the congressional context hypotheses, I use a dummy for divided government, coded "1" if the opposition party controls at least one house and "0" for united government. I measure the degree of polarization with the absolute value of the difference in DW-Nominate scores for the median member of the two parties in each chamber, using the average distance for the two chambers. The interaction of divided government and polarization multiplies these two variables (Friedrich 1982; Brambor, Clark, and Golder 2006). Under united government, the interaction equals "0" but takes the party polarization value with divided government. The congressional context theory suggests a negative effect for divided government, that is, presidents will moderate under divided government. Polarization will be positively associated with extremism, and the interaction term will have a positive sign. Finally, the estimation includes a control for Stimson's public mood variable. As the mood variable ranges from conservative to liberal, I also fold it, using the numerical mean of the series as the midpoint (58.2). (14)

One disadvantage of single equation ECM models is that they can eat up degrees of freedom quite quickly. This may be a problem for the full ECM estimations here where we have 50-52 cases and 11 variables. Second, there is a hint of multicollinearity between the differenced and levels form of the congressional context variables, especially the divided government and interaction variables. Thus, I ran several different estimations, one set using all variables, another set only using the differenced form of the congressional variables, and a third employing only the lagged form of those variables. I did the same for the public mood variable. In all cases, the differenced form of the congressional and mood variables were stronger predictors than the lagged form of those variables. In fact, the lagged forms never get close to being statistically significant, even in estimations with the differenced forms excluded. The significance levels for the differenced variables are stronger without the levels versions of the same variables. In Table 3 I only present the results with the differenced form variables, excluding the lagged variables. I will supply interested readers complete results upon request.

Results

Table 3 presents results of the estimations for each presidential extremism measure. Analysis of residuals finds that all are white noise. Despite the modest number of variables included, each estimation accounts for a considerable amount of the variance in presidential policy extremism, 40% for the ideal point measure, nearly one-half for the adjusted ADA scores, and over 80% in the case of the DW-Nominate measure. The lagged dependent variable is the major factor accounting for the DW-Nominate variance, about 80% by itself. The large coefficient for the lagged DW-Nominate variable also indicates its predictive strength. As a consequence of the lagged dependent variable, the other independent variables do not emerge as strong as they do for the ideal point and ADA measures.

Overall, these results provide little support for the party activist theory. In only one case, ADA scores, is the pre/postreform dummy significant. That coefficient indicates a nearly 10% increase in presidential extremism in the postreform compared to the prereform era. For two of three cases, the postreform counter has the wrong sign but never attains statistical significance.

In contrast, analysis provides more support for the congressional context theory. The divided government variable is properly signed for all three measures and is statistically significant at the .05 level for the ideal point and ADA measures. For the DW-Nominate indicator, the divided government dummy only reaches a significance level of. 10, not quite significant using the conventional .05, but perhaps close enough to warrant some attention. Turning to substantive implications, using the ADA measures, shifting from united to divided government results in presidents being 55 % less extreme, about the difference between the least and most extreme presidents. Divided government has similar effects using the ideal point measure, reducing presidential extremism by 1.46, which again is about the difference between the least and most extreme president. Even for the marginal effect on the DW-Nominate measure, divided government reduces extremism by .34 units or about 40% of the range between the least and most extreme presidents.

Polarization by itself has no significant effect on presidential extreme once controlling for the interaction between divided government and polarization, as predicted. However, the interaction term is significant for the ADA and ideal point measures, while for the DW-Nominate scores, it just misses at p <. 10. For the ADA score, a one standard deviation shift in polarization, about .10, is associated with approximately a 10 point rise in extremism. A more useful comparison is between very low (.47) and very high polarization (.87), capturing the full effects of the rise of polarization the past several decades. Using this comparison, presidents will be 40% more extreme during periods of highest versus lowest polarization, offsetting most of the effect of divided government. Now a president during both high polarization and divided government will only moderate about 15%. Substantively the results are similar when inspecting the ideal point measure, and while not as potent using the DW-nominate measure, the interaction between divided government and high polarization undercuts most of the moderating influence of divided government alone. These results point to strong support for the congressional context theory.

The public mood always affects presidential extremism. Presidents become more extreme as the public does. This finding varies with Wood (2009), who found that presidential rhetorical liberalism does not respond to public liberalism. The difference here may be in the measures of presidential policy. However, despite the statistically significant impact of public mood on presidential extremism, the substantive effects appear marginal, swamped by the congressional context effects reported above. Using the ADA scores, a maximum shift in mood of about 6% leads to approximately a 12% shift in presidential extremism. Rarely does the public mood swing so much from year to year. Only on four of the 50 years here does it move more than 3 points. A one-standard deviation shift in public mood, about 2 points, results in a 4% shift in presidential extreme, a substantively modest impact.

Finally, is the issue of the timing of the reform effects. The pre/postvariables with 1976/1977 as the cutpoints suggest a nearly immediate or speedy effect of the reforms on the influence of party activists, to which presidents also respond quickly. However, there may be a lag between the implementation of the electoral reforms of the 1970s and party activist realization of increased influence. To test for this possibility, I experiment with two alternatives for the pre/postreform dummy and counters by using 1980/1981 and 1984/1985 as the cutpoints. Substituting 1980/1981 cutpoints finds that these new variables perform nearly identically to the original variables, except for minor differences in the size of coefficients and standard errors. For the ideal point and DW-Nominate measures, again the 1984/1985 cutpoint variables work as similarly to the original ones. However, for the adjusted ADA scores, the size and significance of the coefficient grow somewhat (b = .76, SE = .21, t = 3.60, p = 0.001). The counter variable now attains statistical significance, but points in the wrong direction (b = -1.22, SE = .45, t = -2.69, p = 0.010).

Most important for this analysis is that the congressional context variables maintain their impact on presidential extremism in the face of controls for the public mood. In contrast, the party activist perspective receives much less and only scattered support. The congressional context looms large in presidential policy choice, at least with regard to position taking on congressional roll calls.

Conclusions

Presidents, on average, have not become more extreme in their policy positions despite the growing polarization between the parties of the past several decades. One would expect that presidents would exhibit the same behavior patterns of growing extreme as we see in voters, party activists, and members of Congress. As the analysis in this article indicates, it is not that in the age of polarization that presidents are not as policy extreme as members of their own parties, but that presidents were quite extreme in their policy stance in the 1950s and 1960s, when the parties were not so polarized, under specific conditions. Specifically, under united government, presidents, no matter the degree of party polarization, tend to take relatively extreme policy positions. For the most part, John F. Kennedy and LBJ were decidedly liberal and about as far to the left as Ronald Reagan (and perhaps George W. Bush) stood to the right during the age of highly polarized parties.

Divided government motivates presidents to moderate their policy positions in order to achieve policy compromises that result in the production of public policy. It just happens that in the age of highly polarized party, divided government has become more common, perhaps even the norm. Yet party polarization limits the moderating pull of divided government on presidents, as their own parties impose costs on presidents who compromise too often or too much with the opposition party that controls Congress. Thus, presidents of the polarized age, when they faced an opposition Congress, Reagan, Bush I, and Bill Clinton, appear more extreme relative to Dwight Eisenhower, Richard Nixon, and Gerald Ford in the face of opposition control, but when the parties were not so polarized. In a sense, this is a story of the impact of separation of powers. As intended, separation of powers pulls politics toward the center and helps cure the "mischiefs of faction," one of those mischiefs being extremism in government policy.

What about the impact of the party reforms of the 1970s and the rise of activist influence over party nominations? The analysis here detects no direct support for party activist-party reform effects on the presidency. But insofar as activists, and the reforms they helped institute and exploited, pushed the respective parties to the political extremes, then indirectly, through polarizing the parties, activists and their party reforms may have impacted the presidency and policy making. Ironically, insofar as polarization between the parties breeds divided government, then party activists may be perennially disappointed, as divided government mutes their ability to move policy as far to the right or left as they would prefer.

AUTHOR'S NOTE: An earlier version of this paper was presented at the "Going To Extremes: The Fate of the Political Center in America" Conference, Nelson A. Rockefeller Center, Dartmouth College, Hanover, NH, June 19-21, 2008. I want to thank the participants at the conference, Matt Beckmann, Jon Bond, Richard Fleisher, and the reviewers for Presidential Studies Quarterly for their comments on earlier versions of this research.

References

Aldrich, John H. 1995. Why Parties? The Origin and Transformation of Political Parties in America. Chicago: University of Chicago Press.

Aldrich John H., and David W. Rohde. 2001. "The Logic of Conditional Party Government: Revisiting the Electoral Connection." In Congress Reconsidered, 7th ed., eds. Lawrence C. Dodd and Bruce I. Oppenheimer. Washington, DC: CQ Press, 269-92.

Anderson, Sarah, and Philip Habel. 2007. "Revisiting Adjusted ADA Scores for the U.S. Congress, 1947-2006." Manuscript, University of California, Santa Barbara.

Andres, Gary. 2005. "The Contemporary Presidency: Polarization and White House/Legislative Relations: Causes and Consequences of Elite-Level Conflict." Presidential Studies Quarterly 35 (4): 761-70.

Bailey, Michael A. 2007. "Comparable Preference Estimates across Time and Institutions for the Court, Congress and Presidency." American Journal of Political Science 51 (July): 433-48.

Beckmann, Matthew N., and Vimal Kumar. 2010. "Opportunism in Polarization: Presidential Success in the U.S. Senate, 1953-2008." Paper presented at the American Political Science Association meeting, Washington, DC, September 2-5.

Beckmann, Matthew N., and Anthony McGann. 2008. "Navigating the Legislative Divide: Polarization, Presidents, and Policymaking in the US." Journal of Theoretical Politics 20 (April): 201-20.

Binder, Sarah A. 2003. Stalemate: Causes and Consequences of Legislative Gridlock. Washington, DC: Brookings Institution Press.

Bond, Jon R., and Richard Fleisher. 2001. "The Polls: Partisanship and Presidential Performance Evaluations." Presidential Studies Quarterly 31 (3): 529-40.

Brambor, Thomas, William Roberts Clark, and Matt Golder. 2006. "Understanding Interaction Models: Improving Empirical Analyses." Political Analysis 14 (Winter): 63-82.

Cameron, Charles M. 2002. "Studying the Polarized Presidency." Presidential Studies Quarterly 32 (4): 647-63.

Cohen, Jeffrey E. 1997. Presidential Responsiveness and Public Policy-Making: The Public and the Policies That Presidents Choose. Ann Arbor: University of Michigan Press.

Cohen, Marty, David Karol, Hans Noel, and John Zaller. 2008. The Party Decides: Presidential Nominations before and after Reform. Chicago: University of Chicago Press.

De Boer, Suzanne, and Luke Keele. 2008. "Taking Time Seriously." American Journal of Political Science 52 (January): 184-200.

Edwards, III, George C., and Andrew Barrett. 2000. "Presidential Agenda Setting in Congress." In Polarized Politics: Congress and the President in a Partisan Era, eds. Jon R. Bond and Richard Fleisher. Washington, DC: CQ Press, 109-33.

Fiorina, Morris P., Samuel J. Abrams, and Jeremy C. Pope. 2005. Culture War? The Myth of a Polarized America. New York: Pearson Longman.

Fleisher, Richard, and Jon R. Bond. 2000a. "Congress and the President in a Partisan Era." In Polarized Politics: Congress and the President in a Partisan Era, eds. Jon R. Bond and Richard Fleisher. Washington, DC: CQ Press, 1-8.

--. 2000b. "Partisanship and the President's Quest for Votes on the Floor of Congress." In Polarized Politics: Congress and the President in a Partisan Era, eds. Jon R. Bond and Richard Fleisher. Washington, DC: CQ Press, 154-85.

--. 2004. "The Shrinking Middle in the US Congress." British Journal of Political Science 34 (July): 429-51.

Friedrich, Robert J. 1982. "In Defense of Multiplicative Terms in Multiple Regression Equations." American Journal of Political Science 26 (November): 797-833.

Galvin, Daniel J. 2010. Presidential Party Building: Dwight D. Eisenhower to George W. Bush. Princeton, NJ: Princeton University Press.

Groseclose, Tim., Steven D. Levitt, and James M. Snyder. 1999. "Comparing Interest Group Scores across Time and Chambers: Adjusted ADA Scores for the U.S. Congress." American Political Science Review 93 (March), 33-50.

Jacobson, Gary C. 2000. "Party Polarization in National Politics: The Electoral Connection." In Polarized Politics: Congress and the President in a Partisan Era, eds. Jon R. Bond and Richard Fleisher. Washington: CQ Press, 9-30. --. 2007. A Divider, Not a Uniter: George W. Bush and the American People. New York: Longman.

Kenny, Lawrence, and Babak Lotfinia. 2005. "Evidence on the Importance of Spatial Voting Models in Presidential Nominations and Elections." Public Choice 123 (3/4): 439-62.

King, David C. 1997. "The Polarization of American Parties and Mistrust of Government." In Why People Don't Trust Government, eds. Joseph S. Nye, Philip D. Zelikow, and David C. King. Cambridge, MA: Harvard University Press, 155-78.

--. 2003. "Congress, Polarization, and Fidelity to the Median Voter." Working Paper, John E Kennedy School of Government, Harvard University.

Krehbiel, Keith. 1998. Pivotal Politics: A Theory of US Lawmaking. Chicago: University of Chicago Press.

Layman, Geoffrey C., Thomas M. Carsey, and Juliana Menasce Horowitz. 2006. "Party Polarization in American Politics: Characteristics, Causes, and Consequences." Annual Review of Political Science 9 (1): 83-110.

Layman, Geoffrey C., Thomas M. Carsey, John C. Green, Richard Herrera, and Rosalyn Cooperman. 2010. "Activists and Conflict Extension in American Party Politics." American Political Science Review 104 (May): 324-46.

Light, Paul C. 1999. The President's Agenda: Domestic Policy Choice from Kennedy to Clinton, 3rd ed. Baltimore: Johns Hopkins University Press.

Masket, Seth. 2007. "It Takes an Outsider: Extra-legislative Organization and Partisanship in the California Assembly, 1849-2006." American Journal of Political Science 51 (July): 482-97.

Miller, Gary, and Norman Schofield. 2003. "Activists and Partisan Realignment in the United States." American Political Science Review 97 (May): 245-60.

Newman, Brian, and Emerson Siegle. 2010. "Polls and Elections: The Polarized Presidency: Depth and Breadth of Public Partisanship." Presidential Studies Quarterly 40 (2): 342-63.

Pomper, Gerald M. 2003. "Parliamentary Government in the United States: A New Regime for a New Century?" In The State of the Parties, 4th ed., eds. John C. Green and Rick Farmer. Lanham, MD: Rowman & Littlefield, 267-86.

Ragsdale, Lyn. 1998. Vital Statistics on the Presidency, revised ed. Washington, DC: CQ Press.

Saunders, Kyle L., and Alan I. Abramowitz. 2004. "Ideological Realignment and Active Partisans in the American Electorate." American Politics Research 32 (3): 285-309.

Shafer, Byron E. 2003. The Two Majorities and the Puzzle of Modern American Politics. Lawrence: University Press of Kansas.

Sinclair, Barbara. 1997. Unorthodox Lawmaking: New Legislative Processes in the U.S. Congress. Washington, DC: CQ Press.

--. 2000. "Hostile Partners: The President, Congress, and Lawmaking in the Partisan 1990s." In Polarized Politics: Congress and the President in a Partisan Era, eds. Jon R. Bond and Richard Fleisher. Washington, DC: CQ Press, 134-53.

--. 2002. "The Dream Fulfilled? Party Development in Congress, 1950-2000," In Responsible Partisanship? eds. John C. Green and Paul S. Herrnson. Lawrence: University Press of Kansas, 181-200.

Skinner, Richard M. 2008-9. "George W. Bush and the Partisan Presidency." Political Science Quarterly 123 (4): 605-22.

Smith, Steven S. 2007. Party Influence in Congress. New York: Cambridge University Press.

Theriault, Sean M. 2008. Party Polarization in Congress. New York: Cambridge University Press.

Wood, B. Dan. 2009. The Myth of Presidential Representation. New York: Cambridge University Press.

Zupan, Mark A. 1992. "Measuring the Ideological Preferences of U.S. Presidents: A Proposed (Extremely Simple) Method." Public Choice 73 (3): 351-61.

JEFFREY E. COHEN

Fordham University

(1.) In using the term "extreme," I mean politicians and others who are decidedly (very) liberal or conservative, as opposed to being moderate or middle of the road. This differs from equating extremism with radicalism, but policy or political radicalism is rare in American politics.

(2.) There is also a modest literature on polarization in attitudes about the president and presidential-congressional relations, which points to a growing gap in approval of the president among partisans (Bond and Fleisher 2001; Jacobson 2007; Newman and Siegle 2010).

(3.) Skinner (2008-9) argues that presidents since Reagan have become more partisan. Being more partisan may lead to presidential extremism but only if the parties themselves are extreme. A partisan president in a time of moderate parties will also likely be moderate. In a revision of the president-party literature, Galvin (2010) argues that even presidents from Eisenhower through Jimmy Carter, conventionally thought to be disinterested in their parties, extensively engaged in party building.

(4.) For instance, see Beckmann and McGann (2008), which is primarily a formal theory piece, and Beckmann and Kumar (2010), which argues that polarization may help presidential success in Congress, which they empirically demonstrate.

(5.) Several studies document the influence of party activists on the policy stances of candidates since the reforms of the mid-1970s. See Masket (2007) and Miller and Schofield (2003).

(6.) In 1980 many Democrats felt that Carter was too moderate, and liberal stalwart, Senator Edward Kennedy, challenged him for the nomination that year, a suggestion of the implication of a president who bucks the activists in the postreform era.

(7.) There are two other major attempts at coding presidential policy stances, but both rely on rhetoric, not roll calls. First, Cohen (1997) coded each sentence in presidential State of the Union Addresses (SUA) from 1953 through 1989 for liberal-conservative direction. For present purposes there are several limitations to Cohen's measure. One, it ends in 1989, and thus misses the extreme polarization of the post-1989 years. Two, the SUA misses all presidential positions taken throughout the year but not mentioned in the SUA. Second, building upon Cohen, Wood (2009) content analyzes all presidential rhetoric from 1945 to 2005. Wood's series, however, although positively correlated with the three roll-call measures, is only significantly associated with one, the adjusted ADA scores. Presidential rhetoric and roll-call stances may differ for several reasons. Presidents can take many more rhetorical positions and have greater freedom in selecting what to speak about than is the case for roll call measures. But as our concern is with presidential positions vis a vis Congress, the impact of the congressional context on presidential policy positions, the roll-call based measures are more appropriate for our needs than the rhetorical ones.

(8.) Still, caution is in order in using the ADA data. Ragsdale (1998, 410-11) presents her own ADA series for 1960 through 1996, which is in several instances quite discrepant compared to the Zupan-Kenny-Lotfinia (ZKL) scores. For example, in 1970 Ragsdale gives Nixon an ADA score of 50 compared to 16 for ZKL. Again in 1967, Ragsdale gives Lyndon Baines Johnson (LBJ) a 90 to 56 by ZKL, and in 1966, Ragsdale gives LBJ a 90 to 75 by ZKL. Of the years of overlap between the two series, 34, only 13 years have identical scores, usually when the president scored either a 100 or a 0. Although it is unclear as to the source of these discrepancies, it may be that ZKL do not count presidential absences in their calculations, but Ragsdale does.

(9.) http://www9.georgetown.edu/faculty/baileyma/.

(10.) For unadjusted ADA scores the "natural midpoint" would be 50, indicating that a president votes with the ADA 50%; the variable can range from 0% to 100%. Using 50 for the adjusted scores makes little sense as a value in one year is not comparable to the same value in another year. Thus, I use the actual mean to locate the midpoint for creating the folded variable.

(11.) The correlations between the ADA score and the DW-Nominate and ideal point estimates are .72 (p < .001) and .45 (p < .001), and .56 (p < .001) between the DW-Nominate and ideal point estimates.

(12.) The augmented Dickey-Fuller statistic for the ADA series, with one-lag, a trend, and constant is also nonstationary (statistic = -3.35, with a critical value of -4.15).

(13.) On single equation error correction models, see De Boef and Keele (2008). Readers are probably more familiar with ECM models that employ an error correction mechanism variable, which is composed of the first order lagged residuals of regressing the dependent variable on itself, lagged one period, and the independent variables in levels. Although the single equation estimation may consume degrees of freedom, it allows us to determine which levels variables affect values in the dependent variable.

(14.) There is one issue in using the folded mood variable. It is possible for the public to be extreme in the opposite direction from the president. Thus, I also ran the analysis using the unfolded mood variable. The unfolded mood variable never reaches statistical significance but the folded one does, as reported below.

Jeffrey E. Cohen is a professor and chair of the Department Political Science at Fordham University. His most recent book is Going Local: Presidential Leadership in the Post-Broadcast Age.

TABLE 1
Descriptive Statistics for Three Measures of Presidential
Extremism, 1951-2002

Years       Variable       N     Mean    Std.    Min.    Max.

All         Ideal Points   52    0.77    0.41    0.25    1.64
            DW Nom.        52    0.53    0.19    0.17    0.83
            ADA            50   32.30   12.66    2.42   51.02
1951-1976   Ideal Points   26    0.70    0.51    0.25    1.44
            DW-Nom.        26    0.40    0.17    0.17    0.60
            ADA            26   25.30   11.51    2.42   44.57
1977-2002   Ideal Points   26    0.84    0.27    0.42    1.64
            DWNom.         26    0.65    0.10    0.52    0.83
            ADA            24   41.06    7.62   18.77   51.02

Source: See text for details.

For ideal point estimates, see Michael A. Bailey's web page
http://www9.georgetown.edu/faculty/baileyma. For presidential
DW-Nominate scores, see Keith Poole's nominate web page,
http://voteview.com/dwnl.htm. For the Americans for Democratic
Action (ADA) scores, see Kenny-Lotfinia (2005), corrected using
Anderson and Habel (2007).

TABLE 2
Dickey-Fuller Unit Root Tests for Variables Included in the Analysis

Variable                            Levels   Changes

President Ideal Points              -2.51     -7.08
President DW Nominate               -1.41     -7.04
President ADA                       -3.81    -11.81
Divided Government                  -3.10     -6.95
Party Polarization                   1.49     -8.45
Interaction of Divided Government   -2.64     -6.97
  and Party Polarization
Pre/Postreform Dummy                -0.98     -7.07
Reform Era Counter                   8.17     -1.00
Public Mood-folded                  -4.04     -8.94
Critical Value at .01               -3.58     -3.58

Source: see text for details.

TABLE 3
Impact of Party Activists, Congressional Context, and Public Mood
on Presidential Policy Extremism, 1951-2002

                                Presidential Policy Measure

                              Ideal Points       DW-Nominate

                              b                   b
                             (SE)       t        (SE)       t

Dep. Var.-Lag                -.17 *   -1.96    0.91 ***   10.68
                             (.08)             (.09)
Divided-change              -1.46 *   -2.28   -0.34 (+)   -1.54
                             (.64)             (.22)
Polarization-change          1.26       .58    0.55        0.74
                            (2.16)             (.75)
Interaction-change           1.91 *    1.82    0.52 (+)    1.43
                            (1.05)             (.37)
Mood-lag                      .03 *    1.74    0.01 *      2.05
                             (.02)             (.006)
Pre/Postreform                .12      1.10   -0.001      -0.02
Dummy-lag                    (.11)             (.05)
Pre/Postreform               -.01     -1.61    0.0002      0.07
Counter-lag                  (.01)             (.002)
Constant                      .12      1.59    0.05        1.35
                             (.07)             (.04)
[R.sup.2] /Adj. [R.sup.2]     .48       .40    0.85        0.83
N                           50                50
Breusch-Godfrey              2.10       .72    3.58        0.53
  LM-lags 4 / Prob.
  [chi square]

                            Presidential Policy Measure

                               Adjusted ADA

                               b
                              (SE)         t

Dep. Var.-Lag                -0.54 ***   -3.76
                              (.14)
Divided-change              -55.27 *     -2.09
                            (26.43)
Polarization-change          30.86        0.34
                            (90.47)
Interaction-change           99.26 *      2.28
                            (43.62)
Mood-lag                      2.01 *      2.45
                              (.82)
Pre/Postreform                9.72 *      1.90
Dummy-lag                    (5.10)
Pre/Postreform               -0.20       -0.67
Counter-lag                   (.29)
Constant                     13.63        3.48
                             (3.92)
[R.sup.2] /Adj. [R.sup.2]     0.54        0.46
N                            48
Breusch-Godfrey               1.02        0.91
  LM-lags 4 / Prob.
  [chi square]

*** p < .001, ** p < .01, * p < .05, (+) p < .10