Geography and Gridlock in the United States
Jonathan Rodden
Stanford University
Submission for the Hewlett Foundation Workshop:
Solutions to Political Polarization in the United States
October 1, 2013
This essay argues that political polarization in the United States causes inefficiency and policy uncertainty not because of unusually high levels of ideological polarization among voters or legislators that generate large policy swings. Rather, it is a problem of extreme gridlock under a unique form of two-party presidential democracy in which changes from the status quo frequently require defections from opposition legislators. This essay explores the argument that these defections are increasingly difficult to achieve not only because of increased agenda control for the majority party, but because of a profound transformation of American political geography associated with deindustrialization and suburbanization.
Our understanding of this transformation is still limited, as should be the hubris with which we dispense reform advice. This essay advocates a cautious and experimental approach to reform at the state level, focusing on the potential advantages of some form of compulsory voting.
1.What is the problem?
The United States has a rather unique constitution. It is one of the only countries in the world that combines a strict two-party system with a presidential form of government. Its only peers are Venezuela, Ghana, and Sierra Leone. As demonstrated in Figure 1, all other former British colonies with single-member districts and strict two-party systems have a parliamentary form of government. These countries have no need for bipartisanship. Members of the government party almost always vote for the legislative proposals of the executive, and members of the opposition vote against. A bimodal distribution of roll-call voting scores without centrists is neither remarkable nor troubling in a parliamentary system.
Parliamentary / Semi-presidential / PresidentialTwo-party / 12 / 2 / 4
Multi-party / 32 / 19 / 30
Figure 1: Regimes Types and Party Systems
The problem in the United States is that for two-thirds of the years since 1950, the president has not presided over a partisan legislative majority. In order to achieve any change from the status quo, he must assemble the votes not only of his co-partisans, but some members of the opposition party as well. In most other presidential systems, the chief executive is able to assemble multi-party coalitions rather than relying on defections from a single opposition party that controls the legislative agenda.
In what reformers now view as a golden era of bipartisanship, American presidents were able to forge relatively stable relationships with moderaterepresentatives known as blue dogs, boll weevils, and Rockefeller Republicans. We have come to appreciate these “moderate” representatives as the lifeblood of American presidentialismnow that they are gone.
Investors are justifiably concerned with polarization in the United States, but not because Americans are divided into two radically opposed camps of Marxists and Randian conservatives, such that every election is a lottery that might end up with either nationalization of the means of production or a dismantling of the regulatory state. Rather, the current fear is that internecine battles between self-interested groups within parties that lack clear platforms will be resolved in unpredictable ways as part of high-stakes bargaining between the executive and the legislature. Divided government in the presence of partisan control over the legislative agenda produces drama and uncertainty even over votes that are clearly in the public interest, like funding the government or avoiding default.
In short, the United States faces a problem of dysfunctional two-party presidentialism. U.S. parties are typical of what political scientists have learned about parties in presidential systems, especially those in the opposition: they are fractious, undisciplined, and have few incentives to develop realistic alternative government programs outside of presidential campaigns. They have no unifying ideological statement around which members have agreed to compete and govern if they are victorious. Rather, there are two loose groups of factions who cannot make promises about what they might do if elected, because it will inevitably involve complex negotiations with other factions in the party, and then in the likely event of divided government, negotiations with the other party.
Party platforms are ascertained by voters from a cacophony of voices attempting to speak for the party. This cacophony is especially diverse for the party that does not control the presidency. This is nothing new. However, for reasons that will be discussed below, those voices are increasingly polarized, and voters have come to perceive the parties’ platforms as increasingly polarized. The parties are as fractious and incoherent as ever, yet voters perceive them to be highly distinctive, and use the party labels as heuristics that are increasingly difficult for candidates in moderate districts to avoid.
Party leaders have gained greater control over the Congressional agenda, preventing bills that might achieve bipartisan support from reaching the floor of the legislature (Cox and McCubbins 2007, Harbridge 2011), which only reinforces the public perception of the parties as taking polarized positions.
The dominant rhetoric in the reform literature is that Americans are centrists, but elected politicians, in collaboration with interest groups, have pushed the candidates’ platforms and legislative behavior to the extremes through some specific institutional features, including primary elections, redistricting, and agenda control in the legislature. The solution, we are told, is to reform those institutions.
Before accepting these conclusions, however, we must be sure that we have the correct story about the decline of moderate members of Congress. The bulk of this essay suggests that these institutions might be of minor importance relative to deeper shifts in the demographic and geographic support bases of the parties.
Even if the ideological positions of Democrats and Republicans in the electorate as a whole have not moved substantially apart over time, and there are still a number of “centrist” Congressional districts or states, a form of partisan polarization can nevertheless emerge within precisely the districts or states that once gave us “moderate” or ideologically complex representatives like Charlie Wilson or Arlen Specter.
Since World War II, the parties have undergone an astounding process of geographic segmentation, whereby the Democrats have become the party of the inner city, the low-income suburb, and the post-industrial town (in addition to rural pockets of blacks, Latinos, and Native Americans), while the Republicans have become the party of white suburbs and the rural periphery. At the same time, the most rapidly-growing suburban areas are becoming more racially diverse and politically competitive.
As a result of this transformation, a seemingly “moderate” district is often either 1) a small, very Democratic city and its very Republican periphery, or 2) a sprawling, racially heterogeneous suburb of a major metropolis that contains pockets of Democrats and Republicans. Armed with the modern toolkit of micro-targeting and get-out-the-vote, incumbents face incentives to build their electoral strategies around the mobilization of their party’s electoral base rather than appealing to unreliable and unpredictable moderates.
If this depiction is correct, investment in the reform of primaries, campaign finance, or redistricting practices might lead to disappointing results.Responding to incentives, candidates in many districts have invested heavily in a strategy of turnout mobilization in which ideological appeals to moderates are an afterthought. As desirable as they may be, I ignore more radical reforms to the electoral system or the structure of the legislature, and take a realist/incremental approach that advocates state-level experiments with a reform that could undo the entire logic of within-district polarization: compulsory voting.
2. How did the problem emerge?
A central puzzle in the literature on U.S. politics is that most voters appear to be centrists, in spite of the growing polarization in Congressional voting. Relative to voters in other advanced democracies, Democrats and Republicans are not especially far apart in their answers to batteries of policy questions. To the extent that the Democrats and Republicans have meaningfulwritten platforms, they are no further apart on average than platforms of parties in other democracies. Yet American voters (and political scientists)perceive the parties to be further apart than their counterparts in any other advanced democracies. I provide data in support of these assertions in Appendix 1.
What accounts for this polarization in perceptions of party platforms? And why are legislators so unwilling to vote across party lines? These questions must be answered together.
The most obvious possibility is that liberals and conservatives have not only sorted into the ideologically proximate party (Levendusky 2009), but they have also sorted themselves geographically into increasingly homogeneous districts, such that there are simply too few moderate districts left (Bishop 2008). Though it is certainly not a perfect proxy for mean district ideology, it is useful to look at the evolution of presidential votes aggregated to the level of Congressional districts. In order to contrast the era of bipartisan cooperation with the present, figure 2 displays kernel densities of (standardized) Republican vote shares as well as first dimension DW-Nominate scores for the 90th and 110th Congresses. The distribution of presidential votes across districts has changed somewhat: the tails of the distribution have been pulled out slightly, especially on the left, and the density of districts in the middle is now a bit smaller.
Figure 2: Distribution of presidential votes and Nominate scores:
Contrasting the era of bipartisanship with today
However, there is nothing in the distribution of district presidential vote that would explain the increasingly bimodal distribution of Congressional voting. There is still a large density of districts that are evenly divided between Democrats and Republicans. The early post-war period demonstrated that the U.S. presidential system can function with only a small rump of moderate legislators in the opposition party who negotiate with the president, and there are still more than enough “moderate” districts to fulfill that function. The crucial question is why the legislators from these districts will no longer vote across party lines.
The answer may lie in the same demographic trends that are pulling the tails of the district-level distribution outward as they play out within the districts in the middle of the distribution. In short, districts where the presidential vote share is near 50 percent do not necessarily contain a large density of moderates.
Figure 3displays the county-level two-party presidential vote (outside the South) on the horizontal axis, and the logged population density of the county on the horizontal axis. The size of the marker corresponds to the population of the country. Figure 3 displays a fascinating transformation. In the early postwar period, there was a very weak correlation between population density and Democratic voting. The relationship has become steadily stronger in each election, and by 2000 it is stunning.Democratic presidential candidates no longer compete in very rural counties, and Republicans no longer compete in cities.
Figure 3: County-level population density and Democratic presidential vote
Perhaps the more interesting and unheralded parallel relationship, however, is below the county level. Figure 4 displays the precinct-level relationship between population density and the 2008 Obama vote share within Pennsylvania counties.
Figure 4: Population Density and Democratic presidential vote, Pennsylvania precincts
In the non-metropolitan counties, there is a stark geographic segmentation between the sparse rural precincts and locally dense enclaves—for example smaller post-industrial towns, college towns, or even county seats of very rural counties containing clusters of rental housing and public employees.
The relationship is even stronger in the suburban and exurban counties surrounding Philadelphia and Pittsburgh and their post-industrial satellite cities. Democrats dominate the relatively dense communities containing apartment buildings and small, tight older single-family homes, while Republicans dominate the communities with larger single-family houses built in newer developments with lower density.
This relationship holds up virtually everywhere in the United States. Urban America is overwhelmingly Democratic, and very rural places are overwhelmingly Republican. Much of supposedly red “rural” America is a patchwork of very Democratic towns, many of which are losing population as they lose manufacturing jobs, and a largely Republican periphery. Perhaps more importantly, the image of suburbia as homogeneously white and Republican is inaccurate. The suburban and exurban counties that are most rapidly gaining population are becoming far more racially diverse, and contra Bishop (2008), more heterogeneous in their voting behavior. Minorities, especially Latinos, are moving to the locally dense suburban areas where employment opportunities are expanding (Orfield and Luce 2012).
Elsewhere I have pointed out that this pattern creates substantial partisan bias in favor of Republicans (Rodden 2013). The relevant implication here is that the anatomy of a “moderate” district has changed since the golden era of bipartisanship. Before globalization and the collapse of manufacturing, the political differences between non-metropolitan industrial towns and their surrounding rural peripheries were less pronounced. Even in the 1960s, white flight from large cities was incomplete, and the sharp political differences between major cities and their surrounding suburbs had not yet fully emerged (Nall 2012).
Over time, the Democrats became the party of urban and post-industrial America and the Republicans the party of rural America and the outer suburbs. As part of this transformation, national party labels have become increasingly meaningful heuristics for voters in House and Senate elections. The number of House districts that split their votes between presidential and Congressional candidates has been steadily declining. Moreover, when Democrats manage to win in Republican districts (and vice-versa), they face few incentives to emphasize policy, focusing instead on advertising their role in appropriations (Grimmer 2013). Furthermore, a potentially large number of votes that would obtain bipartisan support are suppressed by party leaders. Thus policy debates are dominated by extremists from urban Democratic districts or rural Republican districts, generating the polarized platform perceptions described above.
Increasingly, today’s “moderate” districts are those with relatively even mixtures of highly Democratic communities (like smaller post-industrial or college towns or growing Latino suburbs) and highly Republican exurbs and rural areas. Thus a presidential vote share near 50 percent in a district or state does not imply that it contains a large density of moderates.
Tausanovich and Warshaw (2013) have linked several surveys together so as to obtain large sample sizes on several policy questions in order to characterize the distribution of preferences within districts. McCarty, Rodden, Shor, Tausanovitch, and Warshaw (2013) use these data to show that the most “moderate” districts have the most internally polarized electorates. They go on to show that differences in roll-call votes between Democrats and Republicans in otherwise identical districts are much larger when the districts are internally polarized.
Incumbents in such districts face weak incentives to cultivatemoderate voting records. McCarty et al (2013) argue that a heterogeneous internal distribution of political preferences, combined with uncertainty about turnout, creates uncertainty among candidates about the spatial location of the median voter, weakening their incentives to converge to the district median. Moreover, incumbents must cater not only to the district median, but also to the median of their party primary electorate. As primary electorates are pulled further apart, incumbents with moderate voting records are in increased danger of losing primaries.
As the geographic segmentation of the electorate within districts has increased, so has the technology of geographic and household-level targeting of messages and mobilization activities. When the district electorate is internally polarized, candidates find it increasingly advantageous to invest in activities related to turnout among their supporters rather than attempting to attract unreliable moderates.
3. What can be done?
A number of things have changed during the period from the 1960s to the present. Party leaders have gained agenda control, credible primary challengers have become more common, majority-minority districts have emerged, and campaign funding and strategies have changed. As I argued above, this period has also seen a thorough geographic and demographic transformation that interacts with each of these changes.
We only have one run of history from which to draw inferences, leaving us in a very weak position to attribute causal primacy to one of these factors and recommend institutional change aimed at turning back the clock to the 1960s. Given this morass, it is tempting to recommend wholesale constitutional revisions that would change the basic nature of the game, exiting the constitutional trap of dysfunctional two-party presidentialism. Woodrow Wilson’s proposed solution was to adopt a parliamentary form of government. A related proposal might be to change House and Senate terms to four years and coordinate with presidential elections so as to reduce the prevalence of divided government. Alternatively, the United States might retain its presidential system but double the size of both the Senate and House, adding an upper tier with list proportional representation (a German-style two-tiered system), hoping to achieve a multi-party system that would allow presidents to assemble flexible multi-party coalitions, as happens in virtually all of the world’s presidential systems.
Such radical constitutional changes are extremely unlikely to come to fruition, however, and some of their effects are very difficult to predict. Rather than focus on first-best solutions, I turn attention to a more realistic reform agenda would focus on smaller changes, ideally ones that do not require constitutional amendment. Even better, it is useful to focus on reforms that can be rolled out at the state level in the spirit of experimentation.