Not surprisingly, a number of scholars pushed back against the Nick Kristof column that I blogged about in my previous post, echoing many of the points that I made (see here and here and here and here and this Instagram and, of course, this wonderful parody). The general consensus (but see here!) is that Kristof doesn’t fully appreciate the growing frequency with which political scientists are using social media, in addition to more traditional news outlets, to make their research much more accessible to journalists, pundits and, for that matter, the general public. This Presidential Power blog is case in point! At the same time, he is perhaps less sensitive to how the professional incentives that motivate political scientists’ choice of research topics and methods differ from those that influence how Kristof’s “public intellectuals” go about their jobs. Most notably, political scientists are engaged in a collective enterprise in which their research must pass muster with their peers through, ideally, a process of blind review. That means the research must be thorough, theoretically explicit and empirically grounded. This process of research and review often takes considerable time – months or years in the case of a book. Moreover, it is not enough to assert that our research helps explain some pattern of behavior or political outcome. It should also be clear about how confident we are that we understand that relationship or outcome. In short, political scientists live in a probabilistic world, in which what we think we know is usually conditional and is often revised in light of additional research.
This means by nature we tend to recoil from the hyperconfident and often more entertaining assertions that are the pundits and bloggers’ stock in trade. Moreover, we are usually better at rejecting pundits’ sweeping assertions than we are in replacing them with our own equally sweeping (and empirically validated we hope!) generalizations. To illustrate, consider what political scientists think they know about the relationship between gerrymandering and partisan polarization in Congress, compared to what I think many (most?) journalists believe. In my last post, while listing several assumptions embedded in conventional media reporting that I believe do not have strong empirical support, I noted that political scientists have not found persuasive evidence that gerrymandering causes polarization. In response, Blake Hounshell (who is Politico Magazine’s deputy editor)* tweeted: “I’m by no means anti-intellectual but am unpersuaded by many of these alleged ‘debunkings’….I mean gerrymandering doesn’t lead to polarization? Defies all logic.”
And, in Blake’s defense, it does make intuitive sense that gerrymandering by state legislatures in the process of redrawing congressional district lines should contribute to partisan polarization. If a state legislature controlled by a party is successful in redrawing district lines to reward fellow partisans running for Congress, then there will be less competitive elections, which lead to a more partisan outcome which translates into a more polarized Congress. For political scientists, however, having an intuitively plausible theory is only the first step in the research process. As it turns out when they look at the empirical evidence, political scientists generally find that gerrymandering of this type contributes very little to the polarization we see in Congress today. Rather than summarize a body of research, a couple of brief counterfactuals will help illustrate the broader findings. First, in states that encompass a single House district, like my home state of Vermont, gerrymandering is obviously not a factor, and yet we find that the voting records of representatives in these states are becoming more ideologically extreme as well. To illustrate, compare the House voting records of Jim Jeffords, who represented Vermont in the House from 1975-89 with that of Bernie Sanders, who took office in 1992! Second, as this graph shows, it is clear that voting in the Senate has become as polarized as the House across the same time period, even though redistricting doesn’t affect Senate representation.
These two examples suggest that gerrymandering may not be the primary cause of partisan polarization. When empirical findings don’t comport with theory, it is makes sense to revisit the assumptions underlying that theory. In this vein, upon further reflection it becomes clear that there are geographic and political limits to the ability and willingness of partisan legislators to draw district lines in ways that contribute to their partisan advantage in the House. First, court rulings mandating that congressional districts must adhere to some admittedly ambiguous standards of contiguity and compactness, in addition to having the same population size, eliminate some of the more extreme shapes these districts might take in the name of partisan advantage. Second, residential patterns show that Democrats are more likely to be bunched up in urban areas, in contrast to the more widely dispersed Republican voters, which again limits how far one can go to advantage one party by drawing partisan districts of equal population size. Third, efforts to maximize a party’s number of likely seats through gerrymandering may also create opposition-party controlled districts that are less competitive and thus that strengthen the incumbency advantage of the opposition legislators. This also means that the majority party can be more susceptible to getting swamped in electoral waves – something party leaders must keep in mind when pushing redistricting plans. For all these reasons (and others), we need to be careful in assuming that gerrymandering is the primary cause, or even an important cause, of political polarization in Congress today.
Note that, if accurate, this line of research suggests that some common reform proposals floated by pundits and bloggers, such as placing “neutral” commissions, judges or even computers in charge of redistricting, are not likely to make a huge dent in the degree of polarization afflicting Congress today. Of course, to say that pundits overstate the impact of gerrymandering on polarization is not to say the two have no relation. Sean Theriault estimates, if I recall correctly, that redistricting may contribute about 10% to the level of polarization in the House since 1973, primarily through the creation of new congressional districts. However, the prevailing view among political scientists is that when it comes to explaining partisan polarization in Congress, other factors are likely much more important.
I should add that I am by no means the first to suggest that gerrymandering is not a significant cause of partisan polarization – by my count it has been the subject of at least a half a dozen blogs posts just in the last year by my professional colleagues, and yet despite these efforts it remains a staple of mainstream punditry. This is another reason why I am skeptical of the benefits that will accrue by heeding Kristof’s call for political scientists to become more active in mainstream punditry. In cases when political scientists have engaged in public discussions, such as trying to debunk efforts to link gerrymandering and polarization, it is not clear we have had much impact. And that may say more about the incentives motivating what journalists report than it does about any failure of political scientists to engage them.
*My apologies – an earlier version of this post misspelled Blake Hounshell’s last name, and didn’t fully identify where he works.