November 17th, 2012


Geography's consequence at election time.

In the various post mortems of the recent U.S. Election, observers have argued over the degree to which the winner this time around has a 'mandate' ... I'm not going to argue if Obama has such a 'mandate' or not... nor even if he needs one; ultimately I think when someone tries to define "mandate" in this special context they end up saying very little that isn't nonsense. That's not what this post is about.

But in such discussions of a 'mandate' some politicians on the loosing side of the election (Specifically Paul Ryan, but there are others) have counter argued that no such mandate exists, offering as evidence Republican retention of control of the House of Representatives.

The counter-counter argument, of course, is that such retention has less to do with the will of the plurality of the people, and more to do with the idiosyncrasies of how legislative seats are partitioned and awarded.

[Do The Math]

Specifically, as of Nov 7th's counting, the total number of votes cast for Democratic candidates for the House of Representatives was 53,952,240, and the total for Republicans was 53,402,643.

A difference of about 1 or 2 percent, in favor of Democrats, yet it manifests, in the 133th U.S. Congress as a 45.7% to 53.8% difference in representation, in favor of Republicans, about a 10% bias off of the popular vote.

Many have pointed to Republican control of state legislatures and their recent redistricting as an explanation. While I do not doubt that some partisan gerrymandering occurs, that's not all that is going on.

Anyone passingly familiar with U.S. political partisanship knows that Democratic support is concentrated, while Republican support is more diffuse. This goes beyond the typical Urban / Rural divide, or the Big state / Little State divide. Democratic support is highly clustered, both inside and outside of Urban centers. This clustering has specific electoral consequences. Namely, where democrats win, they win big, and all those votes over the +50% threshold are 'wasted'. Clustering means Democratic voters have more wasted votes than Republicans.

Jowei Chen of the University of Michigan and Jonathan Rodden of Stanford University have used the rich data set that is the Florida 2000 election (for all intents and purposes a tie between presidential candidates) to study various methods of district formation. You can read an abstract in PDF form here:

You can also download their code and a host of interesting figures.

Within you see details of how various algorithmic methods, that built districts out of precincts, invariably yield Republican legislative majorities. This bias emerges based on the typical criteria for logical, non gerrymandered districts... all that is required is that they be contiguous. After various simulations performed with altered algorithms and seed conditions, they saw anywhere from a 56% to a %68 Republican legislative advantage emerging from a 50/50 tie vote.

The exact same logic as above yields similar results in elections for state legislators. It is also more or less applicable to every other state, depending on the degree of demographic clustering.

Such bias cannot be overcome by non-partisan, "objective" district allocation. The objective methods themselves yield biased results. Counter-intuitive though it may seem, the only 'fix' (where by fix, I mean making legislative representation match as closely as possible actual voter preference) is a purposeful gerrymandering in the opposite direction.

Why I think this is significant, and troubling.

The above quoted paper talks in terms of Republican and Democratic partisanship, but we can abstract to a more general case. To, say, the interests of those who live in high population densities, versus the interests of those who do not. This dichotomy happens to be currently mapped onto the Republican vs. Democratic political groupings, but there's no reason it need stay that way. Parties and policies may shift, but if we presume that those who live in high population density share interests, and that these interests express themselves politically (which I think is a good assumption) then those voters will always be 'more clustered' than their political opponents, by definition.

In other threads and in other venues I have argued how our various levels of government are selected by means that are less than egalitarian, and incidentally always seem stacked against the interests of urban dwellers. For various historical reasons, they have in-place biases. The obvious bias in Senate and Electoral college representation is typically explained away as an aspect of "State Rights". But the House of Representatives is supposed to be that 'piece' of government most reserved for popular rule. And in any case, "State's Rights" is hardly an apt dismissal for something that biases the makeup of the state legislatures themselves.

So, at ALL levels of government, and in both elected branches, we see not only a distinct bias, but one that runs in the same direction in each instance, privileging the vote of the rural or suburban voter at the expense of the urban. This must affect policy. This must affect our priorities.

If our ideal is that representation should come as close to parity as possible, if my vote ought to, at least in principle, count as much as your vote, then the current system fails our ideal. And it can't be fixed via objective, non-partisan districting, as is shown above. Failing that ideal, it loses legitimacy in the eyes of marginalized voters. That's why I think a move to proportional representation is necessary.