Is "Gerrymandering" Responsible for the House Majority?

The 2012 election was, by all accounts, a victory for the Democrats.  President Obama was re-elected in spite of the economic downturn that took place during his first term, and the Democrats picked up seats in both the House and Senate.  However, the GOP maintained its majority in the House of Representatives, 234-201 (assuming the current leaders in the 4 un-called races go on to win).

It was predicted for some time that the House wouldn't switch hands, so many people didn't pay much attention to the races.  But there are some excellent election datasets out there - for instance, the Google Elections data browser (edit: it seems as though the google elections data browser has been taken offline).  And if you browse through the results of the Congressional elections, you might see some surprising things.

Case in point: Florida's 5th District.

Fig. 1: Gerrymandering in Florida?  It seems like Democrats are packed into the 5th District, allowing nearby districts to be won more easily by Republicans.

Florida's 5th District came into its current form during the 2010 redistricting.  It was formed mostly from the former 3rd District, but it also contains some parts of its old neighbors.  The shape of this district strikes me as ridiculous.  To understand why, let's take a step back and talk about "Gerrymandering."  Every 10 years, the United States conducts a census.  Based on the results of the census, states with more people are assigned a greater share of the House of Representatives.  If a state gains or loses representatives after the census, then the state must re-design the boundaries of its congressional districts.  In most states, this process is carried out by the state legislature.  So naturally, whichever party controls the state legislature tries to draw the boundaries to get an advantage.  The term "Gerrymander" dates back to 1812, when Governor Elbridge Gerry redrew a Massachusetts district into a shape resembling a salamander - to benefit his party, of course.

Florida's 5th district is a perfect example of "packing."  It was actually formed as a compromise between the Florida GOP and minority members of the Democratic party - in essence, the boundaries were drawn so that the majority of voters in the district were minorities, leading to more minority representation in Congress.  However, what we see is a very oddly drawn district boundary that contains an overwhelming majority of Democrats.  That district is surrounded by several GOP-leaning ones.  If those Democratic voters were members of neighboring districts, then they may take the majority away from the GOP.  Instead, they are all packed into one district, giving the Democrats only one representative in the House.  For an example of gerrymandering favoring Democrats, check out Illinois' 17th District.  For a list of the 10 most weirdly-shaped districts, go here.

You may recall that the 2010 midterm was a "wave" election for the GOP.  The Republicans gained tons of ground in the House, Senate, and most state legislatures due to the dismay over the economy and opposition to the Affordable Care Act.  It was those legislatures elected in 2010 that went on to re-draw the congressional boundaries based on the census, so the wary reader may suspect some foul play here.

There are still several precincts all over the country which have not returned their results yet.  But the majority of the data is in, and we can start to examine it for patterns.  I've used the election data from the Huffington Post (mostly because if you right click -> view source, you have access to all the numerical results).

Popular vote, electoral house

If we add up the vote totals for each district, here's what we find.  56.0 million votes were cast for Democrats, and 54.9 million were cast for Republicans.  Even though Republicans retained a wide margin in the House, they were actually outvoted 50.5% to 49.5% !  Let's take it a step further, and re-draw the state-wise electoral map.  Below, I've colored each state as though all its representatives were assigned to the party with the most votes.

Fig. 2: Who would win the House of Representatives if its members were assigned in the same way as the Electoral College?  The winner of each state is chosen by adding up all votes for House candidates in that state.

If we divvy up the representatives this way, then the Democrats would have the majority - 225 to 210!

A more equitable way

However, given all the grumbling about the way the electoral college works, I feel that most people would agree that this isn't a fair way to assign representatives.  If we want to figure out whether gerrymandering affected the results, we need to know what the effect of gerrymandering would be.  The idea behind the House of Representatives is that the more population a state has, the more representation they have.  And as a population-based body, the share of representatives in a state from a certain party should match the percentage of votes that party received in that state. 

For example, California makes up about 12% of the population of the United States.  And with 53  out of 435 representatives, California makes up about 12% of the House.  In an unbiased world, if 3 out of 4 California voters voted for Democratic House candidates, then Democrats would win 40 of the 53 seats (roughly 75%).  

Is that the case in this election?  I took each district in each state, and summed their vote totals.  Then I assigned representatives based on the share that each party received.  In other words, if the GOP received 60% of all House votes cast in that state, then I assign them 60% of the seats from that state.  Note that this removes the effect of redistricting - we remove all district boundaries within a state, and assign representatives based on the state-wide vote total.

Here's where things get interesting: in this scenario, the GOP would be in the minority - 215-220.  This is almost exactly equal to the share one would expect based on the national popular vote total (50.5% Dem to 40.5% GOP).

Redistricting rears its ugly head

Here's an example: in Ohio, Democratic candidates received 2,315,250 votes, while Republicans received 2,065,814 (47%).  Ohio has 16 representatives in the House, so one expects the GOP to win 7.5 seats (47%).  However, they ended up winning 12 seats - so the number of excess GOP representatives in Ohio is 4.5.

Let's look at these results by state.  Below, I've plotted the excess number of GOP representatives - this is the actual number of representatives minus the expected number based on the state-wide popular vote share (as explained above).  On the left, this is displayed as an absolute number, and on the right, as a percentage of the total number of representatives in that state (excluding states with less than 4 reps).

We should also pay attention to the redistricting in that state.  9 states underwent redistricting with Republican oversight - those are denoted by a dark red border.  3 states had Democratic oversight (blue border).  5 states had bi-partisan or court oversight (green border).

Fig. 3: The discrepancy between the share of votes cast for a party and the share of seats that party wins (i.e. if 60% of votes are cast for GOP House candidates in a state, the expected share of seats is 60%).  The left figure shows this as a number, so MD:-2 means that the state of Maryland has 2 excess Democratic representatives than expected.  The right figure shows this as a percentage of the total number of seats in that state (excluding states with < 3 seats).  States with a red border underwent redistricting overseen by Republicans in 2010.  Blue border means that Democrats controlled redistricting.  Green border means that there was bi-partisan or court-directed redistricting.

Adding these numbers up, the GOP exceeded their vote share by 21 seats in states where they oversaw redistricting.  Democrats exceeded their share by 7 seats in their states.  Bipartisan states were a wash ( < 1 seat).  So overall, the GOP exceeded their share by 14 seats in states where redistricting occurred. Note that this is not enough to swing the race, as swapping 14 seats would still leave the GOP in control 220-215.

We may also want to consider states that underwent redistricting after the 2000 census.  FL, PA, OH, TX, and MI underwent Republican-controlled redistricting, GA, NC, CT, IN, OK, and CA underwent Democratic-controlled redistricting, and NY, IL, WI, CO, AZ, and NV underwent bi-partisan redistricting.  Analyzing this in detail would deserve its own post, but you can look at the chart above and see where those states stand in 2012.

Can we say with certainty that this discrepancy of 14 seats are due solely to gerrymanding after the 2010 election?  No.  There are discrepancies in many other states that did not undergo redistricting.  Nevertheless, given the fact that all recent partisan redistricting corresponds to discrepancies (Democrat AND Republican), I think it is fair to say that gerrymandering played a major role in the outcome of 2012 House elections.