Campaigns and elections

Lessig, Klein, and the Economist on Polarization, Spending, and Gerrymandering

Mar 13 '12

Larry Lessig, Ezra Klein, and an anonymous Economist writer have been debating the relative importance of campaign spending and gerrymandering on partisan polarization.   Unfortunately, the exchange is fairly heavy on conjecture and lighter on evidence.  Because these are topics on which I have written a bit, I thought I might provide a few pieces of the missing data.

  • Gerrymandering is not an important cause of polarization

The most important piece of evidence for this claim is that polarization in the United States Senate follows a very similar trajectory with polarization in the House, and the Senate has not been subject to gerrymandering since the Dakotas were split in the 1890s.  While some have suggested that gerrymandering-induced House polarization generated Senate polarization, the evidence is weak that House polarization causes Senate polarization or that gerrymandering has polarized the House.

I have written extensively on that last point.  The gist of the argument against gerrymandering is that polarization in its modern incarnation is primarily a result of the difference in the way Democrats and Republicans represent otherwise similar districts.  I call this the divergence effect.  For example, the gap between Democrats and Republicans who represent 50-50 partisan districts has grown. Polarization is not the result, as the gerrymandering hypothesis would have it, of Democrats representing increasingly liberal districts and Republicans representing conservative districts.  I call such an effect sorting.  The figures below drawn from my published work illustrate this feature.  The figures plot a common measure of congressional ideology (the DW-NOMINATE score) against the district vote for the Republican presidential candidate (a commonly used proxy for district partisanship).  As one can see, the main difference between the parties in the DW-NOMINATE scale is the gap between the parties at each value of the presidential vote (the distance between the smoothing lines).  This gap reflects divergence. Second, note that between 1970s and the 2000s, the divergence between the two lines has grown markedly.   The sorting effect is not as strong.  There has been an increase in Democrats representing very liberal districts, but this has been primarily confined to urban and majority-minority districts – those that are least susceptible to the partisan gerrymandering that is alleged to have caused polarization.   As Boris Shor and I have recently documented, divergence is a much more important source of polarization than is sorting within state legislatures as well.

More directly, Keith Poole, Howard Rosenthal, and I conducted a very large number of simulations where we constructed congressional districts under neutral districting procedures and predicted the level of polarization for each simulation.  These simulations produce polarization as larger or larger than that we observe in the real data without recourse to gerrymandering.  Once one accounts for the divergence effect and inter-state differences, there is almost nothing left to explain.

  • The association between electoral security and extremism in Congress is surprisingly weak

This fact is also consistent with the figure from 2004 above.  Note that the estimated smoothing lines are reasonably flat.  So on average Republicans from districts with a 50% Bush vote are only slightly more moderate that those with a 70% Bush vote.  Yes, Democrats representing districts with 20% Bush vote are considerably more liberal than those with 50%, but again, the 20% districts are primarily minority districts.    A more reasonable comparison is 30% versus 50% districts where the average difference is quite small.  In sum, the effects of the competitiveness or the partisan composition on a member’s ideological position pale in comparison to the effect of the member’s party.

While these data show that over all extremists are only slight beneficiaries of safer, more partisan seats, Tom Brunell and Justin Buchler provide evidence for a stronger claim that members from competitive seats do not pursue positions closer to those of their constituents than do those from less competitive seats.

  • Extreme Incumbents Do Not Raise More Money from Individuals or Groups

While the patterns may have changed slightly over the past couple of cycles, my work with Poole and Rosenthal did not uncover any substantial relationship between extremity and campaign fundraising.  The figures below are from 2002 and appear in our book.  The data clearly show that members with extreme DW-NOMINATE scores suffer a slight penalty in fundraising.  Even if there is now a stronger correlation between fundraising and extremism, I note simply that Congress was plenty polarized by 2002 without fundraising being an important source of it.  Moreover, in a recent paper, Bertram Johnson (cited by Lessig in Republic, Lost) does find that extreme candidates receive a greater proportion of their funds from individuals and from small donors.  But this paper reconfirms our finding that extremism does not raise the absolute level of contributions from individuals over all.  (And this suggests further evidence that large individual contributors and interest groups appear to shun the extremists.)

Summary: Gerrymandering, polarization, and the excesses of the campaign finance systems are clearly areas of concern for reformers.  But not all bad things gotogether.

(Thanks to Steve Rogers for helping me pull this post together)