A More Perfect Union

Following on from my post prior to the election “The Rigged System”, here I will analyze the 2016 results with an eye towards calling out deviations from the proportional ideal. The results are incomplete as votes are still being counted, but we know enough to draw some conclusions.

The big story of course, is the Electoral College bucked the popular vote for the 5^th time in our nation’s history. The well-known data scientists Nate Silver and Nate Cohn expect the final tally to be somewhere between a 1 and 2% margin for Clinton. As of today (11/17), the count on USelectionatlas.org has the margin at 1.01%:

	Votes	%Vote	Electors	%Electors
Trump	61,496,079	46.72%	306	56.9%
Clinton	62,830,751	47.73%	232	43.1%
Margin	-1,334,672	-1.01%	74	13.8%

The bottom line here is an R-bias of 13.8 + 1.0 = 14.8%.

Now let’s look at the House, the people’s chamber. Thankfully, the Cook Political Report has aggregated data on all 435 seats.

	Votes	%Vote	Seats	%Seats
Republican	60,700,356	50.1%	239	55.2%
Democratic	57,104,281	47.1%	194	44.8%
Margin	3,596,075	3.0%	45	10.4%

The bottom line in the House is an R-bias of 10.4 - 3.0 = 7.4%. This is right in line with the 2012 and 2014 biases -- the other House elections in this decade.

For good measure, I looked at the Senate election aggregates as well, which is something I did not consider in my previous post. The data is from USelectionatlas.org. California must be excluded, because they have a different system (a free-for-all multi-party primary and a general election runoff). Their ballot only offered the choice between two Democrats for the Senate. The other state that needs to be excluded is Louisiana since they will be having a runoff election to determine the winner. That said, here are the results aggregated across the remaining 32 Senate races:

	Votes	%Votes	Seats	%Seats
Republican	38,954,023	48.5%	21	65.6%
Democratic	38,216,992	47.5%	11	34.4%
Margin	737,031	0.9%	10	31.3%

The R-bias in the Senate was 31.3 – 0.9 = 30.4%. 

Not only has it shocked the world that Trump won, it is also quite stunning and rare that Republicans have swept the coveted ‘trifecta’. But any suggested mandate should be taken with a grain of salt. The country is much more evenly divided than it would appear. By a significant margin (1% and rising), more people voted for a Democrat in the White House. Similarly, the House elections recorded a significant margin (3%) for Republicans. The Senate elections were almost even (if California had a more normal race).

Side note: It struck me as an odd quirk of our federal system that election data is not readily available. I found out that the raw data that underlies every news site is made available by the Associated Press by subscription. Thanks to the US Election Atlas and the Cook Political Report for their work.

Disclosure: The author of this post has a D-bias.

The Rigged System

The system is NOT rigged, but it does have some bias. And that bias, points in the favor of the Republicans, for the most part. Take the top job for instance. In 2000, it was handed to the candidate with a half million fewer votes, thanks to Florida, a state governed by the candidate's brother whose tally was certified by Katherine Harris.

Bias also extends to Congress, where districts are gerrymandered by state legislatures every 10 years. It is useful to look at the total number of votes received by each party across all House races in a given election year. The table below lists the number of seats and percentage of votes received by each party (source: Wikipedia). The Delta column lists the difference as a percentage of either seats or votes. The last column calculates a measure of bias towards the Republican Party. This number is positive if they received a larger percentage of seats than their percentage of the vote.

Year	D	R	Delta (D-R)	Bias to R
2004	202	232	-7%
	46.8%	49.4%	-2.6%	4.3%
2006	233	202	7%
	52.3%	44.3%	8.0%	0.9%
2008	257	178	18%
	53.2%	42.6%	10.6%	-7.6%
2010	193	242	-11%
	44.9%	51.7%	-6.8%	4.5%
2012	201	234	-8%
	48.8%	47.6%	1.2%	8.8%
2014	188	247	-14%
	45.5%	51.2%	-5.7%	7.9%

In the last six elections, a significant bias for the Republican Party has occurred four times. In one instance there was a significant bias for the Democratic Party and in one instance there was little or no bias.

More analysis is warranted, after all this sample size includes just one presidential and six House elections. But no single election should have such problems anyways. Both of these issues are structural artifacts that are in theory solvable. It is indeed possible to devise a more democratic system, but inertial forces prevent change in a structure that been in place for 240 years.