Waldmann’s thoughts on voting districts
by Robert Waldmann
I think this is the best post I have read so far on the topic by Kathleen Geier. Just go read it before reading my comment below.
Of course I have my usual comments:
Thank you Kathleen Geier for your excellent post based on good shoe leather (OK leather ear to the phone) reporting. I should also thank the political scientists who have been doing their job while I pretend to do it commenting on blogs (I’m supposed to be an economist).
But (of course there is a but) in those comments I have been stressing that bipartisan and court-drawn maps systematically cause Republicans to win a larger fraction of seats than votes. The reason is that a bipartisan and court (except sometimes the Supremes) principal is that the House should look like the country and not like the White plus some brown and no black at all Senate. There is an absolutely deliberate effort to create majority minority districts. Those districts are overwhelmingly Democratic districts. The intent isn’t partisan, but the effect is to cause Republicans to be over represented. I add that I support the effort to make the House look like the country and not like the Senate.
Also, while I see why incumbency helps the party with more seats I do not at all find an explanation for why it explains the gap between popular vote and seats won. The advantages of incumbency discussed in the post should cause the party with more incumbents to get more votes, yet they are discussed as if they affect seats won but not vote totals. To explain the gap, one has to argue that incumbents generally win by narrow margins or lose by wide margins.
Here I think the problem may be with the reported vote totals. Some states don’t report votes in races where there is only one candidate on the ballot. Incumbents often get on the order of 99% of votes or more. That’s a lot of wasted votes which would cause the party of incumbents to have a lot of votes per seat won. But only if they are counted.
Even if the votes are counted, they may be few. Some people (definitely including me) are irritated when invited to vote for the only candidate on the ballot (if I wanted to live in the USSR I would have moved there back when it existed). For decades I chose Michael Capuano over write in, but I’m sure lots of other people left the oval empty out of irritation (this year I actually got to vote in a contested congressional election for the first time since hmmm voting for O’Neill over Abt in 1982 IIRC). I think some correction for races with only one candidate on the ballot is needed.
Finally I don’t get this bit about fewer people in red districts. What geographic boundaries are being considered ? I can see how the requirement that each state have 1 representative could in theory create a small population at large district. However, I think it doesn’t as even Wyoming is not an unusually small district. I don’t see why representation of low population states would be rounded up more often than down. If the boundaries are, say, county boundaries, then current redistricting does not respect them (I vote in the same county and a different district). I tend to suspect that the random districting which respects geographic boundaries doesn’t correspond to anything but a silly modeling assumption.
Polite conclusion about how this is a great post and political scientists are awesome.
cross posted with Robert’s Stochastic Thoughts
“one has to argue that incumbents generally win by narrow margins or lose by wide margins”
Alternatively, Republican incumbents winning by smaller margins than Democratic incumbents would produce results totally consistent with the data. Crossover vote for incumbents being asymmetric would not change that incumbency is a key factor.
(This is a thought experiment, not backed up by reference to data.)
Let’s ignore incumbency.
How about the Brown/Coakley special election:
http://www.boston.com/news/special/politics/2010/senate/results.html
In that election, Brown didn’t carry any town with more than 74% of the vote, while Coakley polled over 74% in 17 cities and towns. In cities and towns where each candidate won, the winner’s votes % (ignoring the third party candidate) was on average 62.71% for Brown and 63.02% for Coakley.
http://oi46.tinypic.com/35n4zlf.jpg
In 2012, the results show a similar sort of shape, Warren polled over 74% in many towns, and Brown didn’t poll even as high as in 2010, but Brown’s average victory was 58.66% (of the two parties), and Warren was 63.68%:
http://i45.tinypic.com/qsjo6s.jpg
I thought I had some sort of point but I lost it. I imagine it is something along the lines of “if you want to win you have to get more votes.”
Goodwin:
I do the same thing in writing longer posts. Duh!
“If you want to win,” you have to win more of the states; especially those with a higher density of population as they carry more of the electorial votes than sparsely populated states.
“If you are among those people stuck in endless traffic jams from Boston to Richmond and from San Diego to Santa Barbara, you may find this hard to believe, but only 4 percent of all the land in the contiguous United States is urbanized, and only 5.5 percent is developed.” http://www.smithsonianmag.com/people-places/presence-oct06.html#ixzz2CiCsJNrR
It is no surprise there are more people concentrated in small spaces.
As I understand it, the way it works is as follows. Let’s consider a hypothetical rectangular state with one large city and suburbs with half the population, and some rural areas that need to be divided into 4 congressional seats. Well you might have one district that contains many of the people in the city (i.e. half the total city+suburbs population) and then three seats that contain 2/3 rural rural and 1/3 suburban voters. Such an arrangement, which mostly results from geography and the desire to create contiguous districts, would skew the representation towards more republican/rural voters.
The numbers I’ve seen are that Democratic house reps got 50.5% of the vote but 46% of the reps, which is a big difference in terms of passing bills, but small enough that it could certainly be the result of geographical bias.
Each party has over 140 seats where it’s candidate won with 60.0% or more of the vote. Democrats, however, have 79 seats where they received at least 70.0% of the vote while Republicans have just 40. As Ms. Grier suggests, many are minority majority districts.
Some of these seats in areas like NY City or parts of Los Angeles might be hard to re-draw. Not so a lot of others. New Jersey’s two minority majority districts give it a 6-6 split despite Obama’s 58% vote share state wide. The Hispanic district in particular, meanders over much of the eastern part of the state.
The mark of a deliberate gerrymander is the seat with a 55-59% winning vote. Nobody draws a 51-49 district to gerrymander. Republicans hold a 57-26 edge in these districts. Pair the 37 seat edge in 55-59 districts with the 39 seat deficit in 70% plus districts and the match would show a very slight Democratic edge.
Normal districting might reduce that to a small Republican edge but one that is 12-15 seats smaller than the current totals.
As for the tendency to re-elect, 78 House members either chose to retire, lost a primary or were defeated on election day.
Btw, Obama carried the Bronx with Romney getting an incredible 8% of the vote. Proof that the city-rural divide is real.
@Goodwin I know the feeling
@Kowalski. Thanks for the valuable information.
@aaron. Uh oh the reply below is combative. I don’t mean to be but won’t rewrite. Just skip it if you don’t like the tone I often have.
I have a problem with the word “mostly” in “Such an arrangement, which mostly results from geography and the desire to create contiguous districts,”
Do you mean roughly 60% or rougly 80% by mostly ? The result follows not at all (0.00%) from insisting on contiguous districts. Contiguous districts could be made starting with a 4 corners point right in the middle of the city where one district has the north east quadrant etc. Then at the city limits the borders could bend so that each district would have the same population. All 4 districts would include city center, city perifery and suburbs. the lines can be drawn so all contain exurban and rural areas too.
It is just not true that Republican over representation is a result of insistence on contiguous districts.
As to geography. Well huh I guess that depends on what you mean by geography. If geography means putting urban voters with urban voters, then the Republican slant follows. But the difference between making urban districts and Republican partisan gerrymandering is, I assert, a distinction without a difference.
Cambridge MA is not a large city but it is divided into more than one congressional district. My representative changed (Capuano to Markey). My voting address didn’t.