Tuesday, November 8, 2016

Things to Watch For (and Not to)!

***Just a quick reminder: I will once again be live-blogging the election returns this evening, but this time around I won't be able to start until roughly 9:00pm because of a teaching commitment.***

 Well, it's official. No, not that. Rather, I've officially given up trying to think about anything else today, regardless of relative personal importance. It's true what the pros say about election day: There isn't really all that much to do except sit around, gossip, fret, and wait. Mercifully that wait is once again almost over. In the meantime, thanks to all who've gotten back on the Cinema Democratica bandwagon: yesterday was our biggest single page-view day, ever -- surpassing even the 2008 live blog -- and I doubt any of this would be nearly as much fun without you all. And none of us, apparently, can stand the suspense.

And so as we all await the results of this at-once-wacky-and-yet-maddeningly-stable election, I thought I would take a few moments of excess nervous energy to share some thoughts about what sorts of tea leaves might make the best reading material before the final calls are in. To my thinking, the "pre-available information," if you will, falls into four categories: red herrings, stubborn and quick-to-fall states, partial precinct data, and demographics. Let's have a quick look at all four, with a word or two about how seriously to take each one. 

Red Herrings
Every election cycle the major polling firms all conduct exit polls, and every year the exit polls turn out to be completely useless. There's a simple reason for this, which is that there's no way to adjust exit polling data to reflect the larger voting cohort: Whomever happens to be emerging from the polling place at a given time, and is willing to be interviewed about it, finds his or her way into the data, regardless of whether he or she proportionally reflects the true mood of the state in question. This is called self-selection bias, and it's an instant deal-breaker as to the reliability of the data in question.

Suppose for example that a college -- even one which shall remain nameless -- decides to abandon the time-perfected practice of conducting student opinion surveys in class, on paper, preferring instead to allow students to elect to participate in the opinion-survey practice entirely on their own. Two things will naturally follow: First, those students who have the strongest opinions about the professor or the class will of course be unscientifically over-represented in the results. Second, since the poll itself is anonymous, there will be no way to weight the results in such a way as to better reflect the opinion of those students' non-responding counterparts.

The most famous example of this problem in electoral politics was election day of 2004, in the afternoon of which staffers for John Kerry confidently predicted that he would win -- basing their confidence entirely on the results of exit polling. More generally the problem seems to be (in part) that patterns of voter turnout over the course of the day are non-randomly distributed across party affiliations, and there just isn't time for the polling firms in question to carry out the sorts of labor-intensive adjustments necessary to correct for this phenomenon before getting scooped. Point being, readers would be well advised to take any news about exit polling with about a 250-pound block of salt.

This year a new source of potential information mischief has waded into the murk, in the form of a series of geek-based efforts to pre-project the outcomes of entire states in full view of the public. Cooler heads will refrain from even clicking through, but it's an open question whether the political commentariat can resist the temptation to meta-cover these scientifically dubious efforts, and the fear on the part of many is that any crack in the mainstream media's traditional embargo on this sort of thing could tip the result somewhere. It would be a fool's errand to ask readers not to patronize the geek-off, but don't let the media stories about it (if any) concern you too much either way, because the reporters in question won't know that they're talking about people who don't know what they're talking about.

Stubborn / Quick-Call States
It's utterly shameless self-promotion to say this, but I consider it no small feather in my cap to have correctly anticipated that Indiana's stubborn refusal to be called in 2008 would set the tone for that historic night for Barack Obama and the Democrats. (And entirely by the way, is it just me or does that moment seem like it happened a hell of a lot less than eight years ago?)

This year the two states to watch are New Hampshire and South Carolina. Both of them close relatively early -- South Carolina at 7:00pm and New Hampshire finishes its staggered closing times at 8:00 --  and any combination other than an immediate call for S.C. and a stubborn lack of resolution to New Hampshire is almost certainly the end of the road for The Orange One. I predict exactly the opposite: that, despite the two weeks' worth of solid media narrative about how tight New Hampshire has become, Secretary Clinton will be called the winner at the first possible second, while South Carolina will be this year's Indiana, just hanging there, and hanging there, and hanging there.

Barack Obama eventually won Indiana in 2008, but it was far from necessary and almost certainly won't happen again. The real point is that an surprise un-called state alters the media coverage disproportionately -- precisely because it's such a surprise. I'm sticking my neck out on this one more than usual, but I honestly think that the combination of a sneaky-high percentage of Hispanics in the Palmetto State, together with Mr. Trump's less than flattering words for George W. Bush, will bring the matter at least to within the margin of error, and set much of the tone of the night for the major anchors.

Partial Precincts
I am not popular company with my closest friends when questions of expertise are at issue: it's a blind spot of mine, owing its origins to a childhood in which no one could correctly pronounce my (very, very easy to pronounce) last name, and which was solidified when I chose economics as my field -- an area of human intellect in which every single drunk asshole at the bar has a self-awarded honorary Ph.D. Point being, I don't do well when I know something, and it's right, and other people won't just take my word for it. The most successful people in the world have all sorts of coping mechanisms for this, and I am forty-seven years old and not about to learn any of them.

How this is relevant in the present context, is that every four years I find myself saying the same thing about the partial-precinct reports that scroll across the bottom of the news channels: as an overall barometer of how the two campaigns are doing statewide, they are utterly, sumptuously, radiantly useless. Yes, precinct reports come in at random, in terms of their distribution across a given state, but that's just the thing: The voters themselves are *not* evenly distributed across that same state. The most notorious case of this leading to some distortions in the public's perception of how things are going was in 2000, when the last precincts to report in Florida were (as they always are) from the deep-blue enclaves of Palm Beach, Broward and Dade Counties. And if you've seen Fahrenheit 9/11, then you don't need me to tell you how this fact was manipulated by Karl Rove to create the appearance that Mr. Bush had won the state before he had.

This being said, partial precincts can shed useful light on the state of a state -- provided that they are viewed against the voter splits in previous elections. If, in Duvall County Florida, for example, Hillary Clinton is running well ahead of where Barack Obama ran in that same part of the state in 2008 or 2012, then -- then -- we know something. Not based on the raw numbers from there, which will almost certainly show Trump with a lead.

We're already seeing big indications that the Hispanic portion of this year's vote was badly under-anticipated in the demographic modeling of the major polling firms. Hispanics (and by the way, women too), are turning out in larger numbers than anyone expected, and if this trend continues it will spell a very bad night for the GOP. So as you follow the news -- including here -- pay special attention to any apparent surprises vis-a-vis the makeup of the electorate (if the non-affiliated electorate in Florida continue to include an outsized proportion of Hispanic voters, for example).    

Well, that's it. At this point, all of us have done what we could, and all that's left is to wait, and enjoy the live-blog. Thanks again for following these pages, and if you haven't done so already -- for God's sake, *VOTE*!

Dave O'Gorman
Associate Professor of Economics
Santa Fe College
Gainesville, Florida

No comments: