“Don’t trust your lying eyes.”
That’s basically what a recent piece from The Week said.
Ryan Cooper took up the Electoral College under the headline “Which states got screwed worst by the Electoral College in 2016?”, and stated flatly that the state that was most disenfranchised by the Electoral College was … *drumroll* … FLORIDA. And after Florida, it was North Carolina, Pennsylvania and Ohio in the 2nd to 4th spots. Meanwhile, Wyoming, Vermont and DC were identified as the most over-represented.
*pause while heads explode*
Top-10 and bottom-10 states in terms of voting population per electoral vote.
Now, I’m being a bit unfair to Ryan Cooper, because in the context of his full article (which I do suggest that you read), it’s clear that this list is presenting only one facet of disenfranchisement under the Electoral College. Most reasonable people aren’t silly enough to think that the voice of a voter in Florida isn’t being heard in our presidential elections, or that Wyoming is deciding our elections ... and Mr. Cooper is one of the reasonable ones.
He goes on to describe other problems with the Electoral College, such as the impacts of the winner-take-all system of allocating votes within a state, and the fact that this system makes the only states that truly count those that have approximate parity between the parties. He forgets to bring up the natural sort of gerrymandering that occurs through demographic sorting, in which some groups get clustered into a few states where their votes become increasingly redundant, leaving other groups to have more impact elsewhere. But still, he sums it up nicely:
Taken together, we can see why the Electoral College is a garbage way to elect the president. It weights residents of small states much, much more heavily than those of large states. It allows losers to win, potentially by a very large margin. And it directs most presidential campaign attention to the handful of states that have roughly equal numbers of Republicans and Democrats.
In the end, however, what becomes clear is that Mr. Cooper doesn’t have an appropriate metric for the full distorting impact of our system for voting for president. Furthermore, it points to a broader problem: most of us don’t really know how to evaluate or describe the shortcomings of the Electoral College adequately, other than knowing that it’s kinda complicated and kinda broken and some people are valued more than others.
To describe the overall problems with the electoral college, a better metric exists: Voter Power Index (VPI)
FiveThirtyEight’s System
VPI is a concept that has been in use for years by Nate Silver’s team at FiveThirtyEight.com. Essentially, they describe it as “the relative likelihood that an individual voter in a state will determine the Electoral College winner”, based on their election models. The ability of a voter to change the outcome of the election is, after all, the bottom line when people think about a vote mattering.
Vote Power Index (VPI) top-10 and bottom-10, from FiveThirtyEight.com
The VPI table at FiveThirtyEight shows us a much more sensible picture. The states at the top read as a veritable Who’s-Who of swing states that get inundated with campaign dollars and media attention. The lone unexpected exception is New Mexico at the top spot, which might represent a limitation in FiveThirtyEight’s modeling (hey, no one’s perfect), but which makes more sense when one considers that it was modeled to be the 12th most-likely state to cast the deciding vote (seems reasonable), but has a smaller population than all the 11 states above it (increasing its per-voter value).
Most people probably ignored the VPI when looking at the election odds on FiveThirtyEight. However, it tells us something profound about the election: The power of individual voters in the states at the top of the list was considered to be in excess of 20x greater than the power of individual voters in states at the bottom of the list. 20x is a minimum, because rounding error prevents us from seeing just how weak the influence of a voter in DC or Oklahoma actually is (hint: it’s probably a fraction of even the power of a voter from Idaho).
To restate, if we held a national popular vote for president, everyone’s votes would be counted equally, but under the Electoral College, a persuasive argument can be made that one person’s vote is worth greater than 48 times another person’s vote! At a minimum!!
A Simpler VPI
The trouble with FiveThirtyEight’s VPI values (other than that they should be presented with an extra decimal place or two) is that they’re the product of a highly-complex proprietary election model that most people can’t hope to recreate on their own. That makes the concept less useful for further analysis and general discussion.
So, I have developed an alternative that can be applied quickly and easily, and can even be adapted to other types of elections (for example, congressional elections within a state):
VPI = {# of electors}/{margin of victory, in total votes}
The margin of victory can be either an expected margin (like FiveThirtyEight would produce), or the real margin. Essentially, it says that the value of an additional vote in a given state is equal to the size of the electoral prize divided by how many votes it would take to change the outcome. For example, Michigan has 16 electoral votes and was decided by about 10,000 votes, so the VPI for Michigan would be 0.0016. Indiana has 11 electoral votes and was decided by about 550,000 votes, so the VPI for Indiana would be 0.00002. We could then say that Michigan voters therefore have about 80x the power of Indiana voters. To make the numbers easier to digest, I personally scale them all up so that the median voter has a value of 1.
Another way of looking at these calculations is like this: let’s say you’ve got a lot of money with which you plan on buying votes at a set price, with the goal of amassing as many electors as possible (to keep it simple, we’ll stay away from the magic number of 270 and pretend that every additional elector is equally valuable, from 1 to 538). In a perfect market, the price of a vote would scale exactly with VPI. You could get 55 electors from California by buying 4.6 million votes at $1 apiece, or you could use the same money to buy 115,000 votes from Florida , where the VPI is 20x higher, at $20 apiece, get 29 electors, and repeat with a similar state.
It also means this: if you took the votes representing the margin of victory from a blue state with a VPI of 1 and redistributed them in red states with a VPI of 10, you could get 10x the number of electors out of the deal.
The Results
I’ve run the simple VPI calculations on the 2016 election results, as well as a calculation that represents my personal estimate of Predicted VPI, which is essentially a back-of-the-envelope estimate of what the average VPI would be if we held a new election tomorrow with generic candidates, taking into account the uncertainty in each state’s vote.
The Predicted VPI is used to smooth out the peaks that come from occasional votes ending up unusually close to a tie, and to account for variation from election to election. Essentially, a calculation for Florida in 2000 would have put its VPI hundreds of times larger than almost every other state, due to how close the election was. Florida is a toss-up, but it’s more reasonable to say that it’s more likely to have a margin somewhere around 1% than 0.01%.
2016 Election Results, 2016 VPI and prediction of VPI for future elections, sorted by Predicted VPI
For the technically-minded: the predicted margin above was a simple weighted average of the % margins of the 2012 and 2016 results, with 2016 given twice the weight of 2012. The standard deviation of the predicted margin is the root-sum-of-squares of 2% (used as a variability baseline) and half the % change from 2012 to 2016. The predicted VPI was calculated through averaging VPI across a Gaussian distribution of the voting margin, using 2016 turnout for total number of votes cast.
Conclusions
What we see in the table above should be shocking: New Hampshire and Florida have about 50x the voting power of Washington DC (and that doesn’t include the poor residents of DC not having any senators!). In this sense, my calculations match those of FiveThirtyEight rather closely.
As expected, California, New York, Massachusetts and Maryland are also right down there at the bottom of the list, together with Oklahoma, Kentucky and Alabama. If not for the Evan McMullin effect, Utah would be down there as well. Each of those states has less than half the voting power of the median VPI voter (who, in this calculation, resides in South Carolina).
Meanwhile, each of the traditional swing states (with the exception of Colorado being replaced by Maine and Minnesota) commands over 3x the power of the median voter, with Nevada and Pennsylvania joining Florida and New Hampshire with a gigantic 10+ times the median VPI!
Furthermore, it’s crystal clear that the great liberal intellectual cities are killing liberal politics nationally by sucking up all the minorities and young liberal college-educated voters and packing them into a select few states where their votes are redundant. Washington DC by itself could have sent out enough Democratic voters to flip the results in all of Pennsylvania, Michigan, Wisconsin and Florida, gaining their 75 electoral votes and still keeping its own 3, but instead Hillary won by a ridiculous 84% and almost all of those voters, many originally hailing from those exact battleground states, had their votes wasted. Voters in New York City, San Francisco, Los Angeles, Boston, Chicago and Seattle are all likewise getting screwed by the current system.
In the end, there were over 28 million votes cast in blue states with a VPI less than 0.5, while there were only 5.5 million votes cast in red states with a VPI less than 0.5 (it would be 6.6 million with Utah). Essentially, we won in the states that matter least by a combined total of about 6 million votes.
I don’t know how we fix that, beyond switching to a National Popular Vote. Hopefully, the Voter Power Index can help us make the case for that change, possibly by showing the people of Arkansas, Tennessee, West Virginia and Idaho that their vote is worth half that of the average American under our current system, and less than 1/10 that of people in Ohio or Arizona … but that they’ll be on a par with everyone else if they join us in election reform.
Lastly, I suggest that others use my VPI formulation to analyze gerrymandering in states such as North Carolina, Ohio, Pennsylvania and Michigan … I think if we can put numbers on the problem, we may be able to reach more people and get more traction in undoing some of the damage done to our states’ districting processes.
Thank you for taking the time to read this!
Update: Also, thank you to the DKos team for putting this post in the Community Spotlight! It’s truly an honor :)
