A Few Thoughts on the Election and Exit Polls

Whether you're pleased or dejected this morning, I think there's very few of us who aren't stunned by the result in the general election yesterday.  In particular, polling was way off—even exit polls, which are supposed to take the pulse of voters as they leave the booth.  How did they get the result so badly wrong?  (Pre-count models showed Clinton with an average of about 300 electoral votes, and winning about 80 percent of the time.)

I'd guess that there are a number of factors (aside from the conspiracy theories, GOP or Dems):

1. People were embarrassed to admit voting for Trump (i.e., he was viewed as the less respectable candidate), but that shame didn't translate to the actual ballot. That doesn't mean that people voted for Trump on a whim; it just means that they weren't keen on admitting that to someone else, even a pollster they'd never see again.
    
2. Exit polling was not done at all locations, for obvious reasons. So projections were based on a regression analysis that fits estimates to the sampled locations. That regression assumes, among other things, a certain degree of polarization between demographics. It looks like that polarization was even more extreme than expected (which was already significant).
    
3. Trump was simply a higher-variance candidate than the traditional Republican. This strategy makes sense in any contest where you're the underdog (as Trump was for most of the time)—if he were to play a low-risk strategy, he was almost guaranteed to lose. Employing a high-risk strategy increases the probability of a blowout loss, but it also increases the probability of a close win, which is what happened. We're seeing this all the time in sports, where endgame strategies by the trailing team are becoming more aggressive. That increase in variance translated to the polls. Five thirty-eight was very open about this—they pointed out that their model, though predicting a Clinton win, had about three times more variation (by some metric) in it than in past years.

I don't think fraud played any significant role in this election. We're seeing real disquiet with the state of the nation. Whether that disquiet has a basis in fact is immaterial as regards the result of the election.

I may have more to say about the election results themselves, but I'll save that for another post. 

[Most of this post was drawn from a Facebook comment.]

Speaking of the Electoral College

About four years ago, I made a somewhat long-winded post (not by the standards of this blog, I suppose, but generally) about the electoral college, prompted by Bernie visiting my office to ask me about it.  Of course, I had primed him by saying I had something nerdy to say about it, and he's unable to resist that kind of bait.  One of the best things about nerdy posts of this sort is that timeliness is not a big attribute, so here it is, four years on:

Bernie just came into my office because he wanted to hear my spiel on the electoral college. Put aside for the moment the question of whether this indicates he's some kind of pedagogical masochist; what started this was the question of whether voters in a big state like California suffer because their vote is diluted, or are favored because the state's electoral power is so huge. The short answer is that it's mostly the latter, but there are a few interesting wrinkles along the way.

One way to approach the question is to consider how many Missouris (11 electoral votes) it would take to match the electoral power of California (55 electoral votes). The obvious answer is five Missouris, but this assumes that the Missouris all vote as a bloc, as California would (in a presidential election). In general, assuming independent Missouris, this is unlikely. Because some of the Missouris would be likely to cancel others out, the swing power they hold would not be 11 times 5, but 11 times the square root of 5, or about 24 electoral votes. (Remember random walks and square roots?) If all Missouris were independent—and why shouldn't they be?!—it would take 25 Missouris to match one California (in terms of the states' electoral power).

Now, each individual voter in California has less power to impact the state's overall direction, just because there are oodles of people in California—roughly five times as many as in Missouri, to match the disparity in electoral votes. (It's actually a little more than that, but we'll deal with that in a bit.) That means that it takes roughly five Californians to make the same percentage impact on their state's result as one Missourian. Again, that only happens if the Californians vote as a bloc; assuming they vote independently, it would take 25 Californians to equal one Missourian.

So at first blush, it seems that these two effects cancel each other out: California has 25 times as much electoral power as Missouri, but each Missouri voter has 25 times as much individual impact on the Missouri result as a California voter has on the California result. However, there is one additional effect of California's large population: The required swing in close votes in California is smaller, percentage-wise, than it is in Missouri. It's basically the law of large numbers: In any evenly contested election, the outcome probably won't ever be exactly 50-50, but the more populous the state, the closer it will be to 50-50, and the smaller the percentage swing required to change that outcome. This factor is again equal to the square root of 5, and it's what drops out in the final result—that a California voter has a larger impact on the national electoral result than a Missouri voter.

One complicating factor is that the number of electoral votes belonging to a state is not quite proportional to that state's population, not even when rounding is taken into account. The reason is that only the number of Representatives belonging to the state is proportional to the state's population; there are also the Senators, which are two a state. Since there is one electoral vote per Congress member (Representatives and Senators combined), small states have a much higher representation per capita than large states.

The upshot is that the most overall power is held by voters in the largest states, like California, Texas, and New York. Intermediate are voters in states with moderately large populations, such as Ohio or Illinois, as well as the smallest states. The weakest are voters in states like Arizona and Colorado, which are too large to gain much advantage from the "bonus" two electoral votes corresponding to Senators, but are too small to gain advantage from the enormous impact of a large population (and large electoral college representation).

It should be pointed out that the foregoing discussion only applies to votes where each state is contested—a "battleground" or "swing state," in recent election parlance. In practice, the impact a California voter has in the 2008 presidential election is nearly nil, since the state is almost guaranteed to go to Obama. (We'll see if I eat those words. [Obviously, I didn't. —brian]) The necessary swing is way too large for a reasonable number of California voters to overcome. That the predisposition of a state's voters is more than enough to swamp the effect of the largest population in the Union is, in my opinion, an indication that those trying to "fix" the election system (typically by replacing it with direct popular vote) are barking up the wrong tree, often in an irrational attempt to right a wrong—the 2000 Bush victory, say, which went against the popular vote—that ultimately had to do with factors distinctly different from the structure of the electoral college.