on party polarization in Congress

Beginning in the mid-1970s, congressional politics became much more divisive. More Democrats staked out consistently liberal positions, and more Republicans supported wholly conservative ones. Pundits, the press, and politicians themselves often use the term ‘polarization’ to describe contemporary American politics. But how do we know, scientifically, that politics has become more polarized over the past thirty years?

The fact that the members of Congress form overlapping cohorts suggests a possible method to investigate this question. The conservative Rick Santorum recently held the Pennsylvania Senate seat once held by the moderate John Heinz. How can we claim that Santorum was more conservative than Heinz? Heinz served with Arlen Specter, who also served with Santorum. Heinz and Specter had very similar voting records. In contrast, Specter voted with the liberals (or Democrats) much more frequently than Santorum did. Observations like this one form the basis for measuring increases in polarization. Thus replacements like that of Heinz by Santorum in Pennsylvania, or of Ervin by Edwards in North Carolina, are the symptoms of polarization.

By adapting the standard dichotomous-choice model, we have created a procedure for measuring ideology–DW NOMINATE. We have used it to analyze all of the roll-call votes in the first 108 Congresses (1789–2004), a project we discuss in greater detail in our book Ideology and Congress. This procedure is based on a simple geometric model of voting behavior. Each legislator is represented by one point, and each roll call is represented by two points–one for ‘yea’ and one for ‘nay.’ These points form a spatial map (maps available at voteview.ucsd.edu and voteworld.berkeley.edu).

Jordan Ellenberg, in “Growing Apart: The Mathematical Evidence for Congress’ Growing Polarization,” a Slate article about our research, likened this map to a road map. While a spreadsheet that tabulates the distances between every pair of sizable cities in the United States contains the same information as the corresponding road map, the spreadsheet, unlike the road map, does not give us an idea of what the United States looks like. Likewise, a spatial map formed from roll calls allows us to visualize the political world of a legislature in a way that a mere spreadsheet with the data cannot. The closeness of two legislators on the map shows how similar their voting records are, and the distribution of all the legislators displays what the dimensions are.

The number of dimensions needed to represent a legislature is usually small, because legislators typically decide how to vote on the basis of their positions on a small number of underlying evaluative, or basic, dimensions. In recent U.S. Congresses, we can easily predict how a ‘liberal’ or a ‘conservative’ will vote on most issues. In fact, just two dimensions can account for almost all of the fourteen million choices of the twelve thousand members who have served in Congress. And one dimension alone suffices in most cases, except in two periods, roughly 1829–1851 and 1937–1970, when race-related issues introduced a second dimension. The two brief periods where the spatial model fails are the Era of Good Feelings, when there was a one-party system, and the 32nd Congress (1851–1853), when the Compromise of 1850 unraveled. In these periods, there is a poor fit, even when ten or more dimensions are used. Voting is chaotic.

The first dimension centers on the role of government in the economy. The second dimension differentiates the members by region mainly over race and civil rights, although in the latter part of the nineteenth century it picked up regional differences on bimetallism and the free coinage of silver as well.

Our current political-party system emerged at the end of Reconstruction in 1879. Republicans have always anchored the conservative end of the first–the liberal-conservative–dimension. They moderated to the left beginning in the early 1900s through the 1960s but then turned back to the right in the early 1970s. Before the late 1920s the Southern Democrats anchored the liberal end of this dimension. After World War I they drifted to the right, until the early 1970s when they reversed course and began moving back to the left. During this whole span of time the Northern Democrats have remained fairly stable on the left side of the spectrum.

The split in the Democratic Party occurred during the latter part of the New Deal when, in the wake of the 1936 elections, Northern Democrats heavily outnumbered Southern Democrats in Congress. Many of the programs initiated during the subsequent Second New Deal were not to the liking of the South. Voting on minimum wages in 1937 and 1938, followed by voting during World War II on the poll tax and on voting rights in the armed forces, helped to splinter the Democratic Party into two highly distinct regional wings. Voting in Congress became two-dimensional in order to differentiate Northerners from Southerners on civil-rights and related votes.

With the passage of the 1964 Civil Rights Act, the 1965 Voting Rights Act, and the 1968 Open Housing Act, this second dimension slowly declined in importance and is now almost totally absent. Issues that are related to race– affirmative action, welfare, Medicaid, subsidized housing, the earned income tax credit, etc.–are now questions of redistribution. Voting on these issues largely takes place now along the liberal-conservative dimension; and the old split in the Democratic Party between North and South has largely disappeared. Voting in Congress is now almost purely one-dimensional–the liberal-conservative dimension accounts for about 93 percent of nonunanimous roll-call voting choices in the 109th House and Senate.

At the same time that voting has become one-dimensional again, the two parties have become increasingly polarized. We use the difference in party means to measure polarization. Polarization declined in both chambers from roughly the beginning of the twentieth century until World War II. This was the consequence of the Republicans moderating and the Southern Democrats becoming more conservative. Polarization was then fairly stable until the mid-1970s. Over the past thirty years, however, polarization has increased sharply despite rhetoric about ‘compromise,’ ‘working together,’ and ‘compassionate conservatism.’

This polarization has been mainly the result of the Republican move to the right during the past thirty years. Conservative Republicans have replaced more moderate Republicans outside the South, and moderate and conservative Democrats in the South. The effect has been a rightward movement on the liberal-conservative dimension of the Republican Party as a whole. As the number of Southern Democrats declined sharply, the reduced Democratic contingent has become much more liberal.

The polarization trend is essentially the same in both the House and Senate despite the differences in the institutions and their apportionment. The correlation between the two series in general is 0.93. The 1879–2006 correlation between the House and Senate Republican liberal-conservative means is 0.94, and for the Democrats the correlation is 0.84. These numbers indicate that the forces driving changes in polarization are broader events (like the Great Depression and wars) and ideas (like a vigorous antigovernment conservatism) that arise outside of Congress.

Indeed, our research has ruled out a number of institutional forces widely believed to cause polarization. For example, primaries existed when polarization was decreasing. There would not have been an institutional barrier to prevent a Ned Lamont from challenging a Joe Lieberman in the 1960s. Gerrymandering is also not a factor, since polarization is being driven by the sharply different ways in which Republicans and Democrats are representing the same or similar constituencies, and not by how the nation is sorted into Republican and Democratic districts.

The modern trend to polarization began with the breakdown of the three-party system. For almost fifty years, from the late 1930s to the early 1980s, the United States had three political parties: the Northern Democrats, the Southern Democrats, and the Republicans. Depending on the issue at hand, each of the three parties could easily form coalitions with one of the others against the third. The Northern and Southern Democrats united to organize the House and Senate, thereby seizing the spoils due the ‘majority’ party. The Northern Democrats and Republicans combined to pass the 1964 Civil Rights Act and the 1965 Voting Rights Act; and the ‘conservative coalition’ of Republicans and Southern Democrats unified to block liberal economic (and, in the 1970s, social) policies.

The demise of this system started with the assassination of President Kennedy. President Johnson was able to do what Kennedy was unable to do: push fundamental civil-rights legislation through Congress. President Johnson’s 1964 landslide victory over an ‘extremist’ Barry Goldwater followed, producing a liberal Northern Democratic congressional majority for the first time since 1936. This destabilized the Democratic coalition. Democrats in the 89th Congress no longer required Southern support to pass many of the expansive federal programs that are so much a part of our current political debate. These programs, along with other redistributive programs initiated by the federal courts (mandatory school busing being the most conspicuous), led to a polarizing backlash in the late 1960s and early 1970s. Watergate blunted the effects of this backlash, though, and the Republican Party did not fully recover its footing until the 1980 elections.

But the old Southern Democratic Party, in effect, disintegrated, and with it the second dimension of congressional voting. Race has been drawn into the first dimension, as race-related issues increasingly became questions of redistribution. The end result is that the Democrat and Republican parties have become more homogeneous. The moderates are gone, leaving us with a polarized, one-dimensional Congress.

Whether we have finally reached the peak of the modern polarization trend remains to be seen. If anything, the most recent Congresses (1987–2006) mark an acceleration of polarization. In the Senate polarization has continued unabated through the 109th Congress (2005–2006), and the House has shown no sign of moderating. The leaders then and now reflect this trend to polarization: Nancy Pelosi and John Boehner are more extreme and far less bipartisan than Tip O’Neill and Gerald Ford were thirty years ago. But perhaps the center will hold. Certainly the agreement by the ‘Gang of Fourteen’ to forestall the ‘nuclear option’ in the Senate in 2005 is a promising sign. We have some hope that over the next decade polarization may finally plateau and start to decline.