Tag Archives: electoral colllege bias

I have been saying for some time now that if Romney began closing the gap in the national tracking polls, as the political science forecast models suggested would be the case, he would also gain ground in the battleground states.    This is precisely what has happened.  In the table below I show the change in the Rear Clear Politics composite polls in the seven tightest swing states across the last 10 days – that is, from shortly before the presidential debate to today.

As you can see, in his 10-day post-debate polling surge, Romney has gained an average of 4% across these seven battleground states, which collectively total 94 Electoral College votes.  This is a near-uniform surge, and it is consistent with what I have been harping on for so long now – a rising Romney tide will float all states’ polls, more or less.  (Keep in mind that the frequency of polling varies across each state.)   To be clear, there were signs that the race was tightening before the first presidential debate, but that event apparently served as a focusing point that pushed the race more rapidly toward where the forecast models, taken as a whole, suggested it should be.   I don’t expect that the Biden-Ryan debate will have nearly the impact on the state of the race as did the first presidential debate – but then, I didn’t expect the first debate to have quite the impact it did!   Still, if the post-debate instant polls are to be believed, Biden and Ryan fought to a draw.  That certainly was not the collective judgment of those who watched the presidential debate.

My larger point, however, is that I never put much stock in the notion that the Electoral College would serve as some type of firewall that would protect the President from a Romney surge in national polls.  In this regard, several of you have asked whether it is possible that Romney might win the national vote, but lose the Electoral College vote.  Sean Trende has an interesting analysis of that possibility here, and he concludes that while the possibility of such a split is higher this year, it is still exceedingly unlikely for reasons that I have discussed here before: historically, the popular and electoral college votes tend to line up very closely.

As evidence, Trende examines the last 15 presidential elections, and compares the winning candidate’s national popular vote margin of victory with his vote margin in the state that “gave” him his 270^th Electoral College vote – the one that put him over the top, so to speak.  He finds that the difference in vote between the two measures is quite small – .9% on average.

Why is this important? Because Trende is essentially extending my logic regarding the link between national and state-level voting which, in turn, determines the Electoral College results.  I have argued that they tend to trend together.  Trende tries to measure that more directly by estimating how “biased” the Electoral College, which is based on state-level votes, is in any given election. To do so Trende looks at the difference between the national vote margin and the popular vote margin he winning candidate receives in the state that gives him the 270^th vote.  That difference, he says, tells us how much the winning candidate was rewarded (or penalized) by Electoral College.

To follow Trende’s argument, let’s look at the current race and estimate the Electoral College bias, as of today. Romney currently leads in the national vote, according to the RCP composite average, by .7% (in an earlier version of this post I had that number wrong).   If we add up all his strong and leaning states based on polling so far, he is likely to win at least 181 Electoral College votes.   To pick up the additional electoral votes necessary to get to 270, he has to win some combination of the 12 or so battleground states.   Let us assume he wins the ones in which he leads as of today – Missouri, Florida, Colorado, and North Carolina. That gives him an additional 63 Electoral College votes – still 26 votes short of victory.  If we look at the remaining tossup states, he runs closest to Obama in Virginia, where he is down by .4%, New Hampshire at .7% and in Ohio by 1.3%.  Virginia has 13 E.C. votes, New Hampshire has 4, and Ohio 18.  Based on these biggest polling deficits, Ohio is the tipping point state – the one that if Romney wins he will go over 270 votes. Romney has to gain an additional 1.4% nationally to overcome Obama’s lead in Ohio.  Assuming a uniform vote swing, that gain would also give him victory in Virginia and New Hampshire as well and he would clinch the Electoral College.   Put another way, if you compare Romney’s current lead in the RCP national poll – .7% – with Obama’s lead in Ohio – 1.3%, using Trende’s logic the Electoral College, as of today, is biased toward Obama by 2%.   That’s a relatively large bias compared to the average of .9% that Trende finds for the previous 15 presidential elections.  It means that to avoid an Electoral College/popular vote split, Romney must win the popular vote by more than 2% (again, assuming a uniform polling swing between the national and state vote).

Of course, there are a lot of assumptions built into this argument, as Trende quite readily acknowledges, beginning with the idea that changes in national support are felt equally across the states.  More significantly, perhaps, it assumes the race will hold steady at its current configuration for the next four weeks.  However, as my table above indicates, it has been anything but steady in the last 10 days, and there are two more presidential debates to go.  There’s no sign that Romney’s surge has peaked, and he may very well cut further into Obama’s lead in Ohio and other battleground states.  On the other hand, Obama may regain his footing and retake the lead in the national polls, bringing them more in line with the state-level polling and thereby reducing the Electoral College “bias”.

This is all a very speculative exercise, of course – particularly this far out – but it is one way to think about the likelihood of a popular vote/Electoral  College discrepancy in outcomes. Taken as a whole, the political science forecast models project this to be a very tight race.  That certainly increases the probability that there will be a split. Note that Trende’s chart indicates that in 7 of the last 15 elections the Electoral College was biased against the popular vote winner.  Based on current national and swing state polling (remember – this could change), that’s the scenario that appears most likely this year – Romney does better in the national popular vote than he does in Ohio. Remember that in 2004 Bush also underperformed in Ohio, his tipping state, by .4% compared to his national vote total. Gore did so as well in Florida in 2000 by .5% – of course, he also lost the Electoral College vote despite winning the popular vote , while Bush held on to win.   Moreover, Obama over performed in the Electoral College tipping state of Colorado in 2008 relative to his national vote margin by a rather large 1.8%.  If the national vote is as close as the models project, and Obama is able to work similar magic in Ohio this time around by dint of a superior ground game, we could see a split.

But if this suggests the probability of a popular vote/Electoral College discrepancy is perhaps higher this year than in past elections, it still doesn’t mean it is likely to happen.  I still think it more probable that the state-level polls will continue to trend toward the national polls, thus reducing the possibility that we will see the winner of the popular vote lose the Electoral College.  Of course, I haven’t yet discussed an even more exciting scenario – an Electoral College tie!

UPDATE: Romney’s RCP national lead has gone up since I originally wrote this, but he remains behind in Ohio, which further increases the Electoral College “bias” in favor of Obama.  I have to think the Ohio race will see more tightening. Stay tuned.