A guide to reading polls and preserving your sanity

As one of the resident polling analysts here at Daily Kos, I spend an unhealthy amount of time scouring the internets looking for new polls, or stories about polls, or stories about pollsters, and ... well, you get the idea.

It is no secret, of course, that as the election cycle begins its process of acceleration heading towards November, the volume of polls to digest will challenge the appetite of even the most degenerate elections junkies (which, of course, is why we love you). With that in mind, this Sunday Kos offering serves as a reminder that (a) this is a marathon, not a sprint, and (b) there are some things to remember as you peruse the numbers.

To that first point: sure, I am totally an enabler here, with an update on new polling data five days a week. But it is critical to remember, the polls are going to twist, turn, flip, and dance a million different directions over the next six months or so. In 2008, from April 1st forward, the polling lead between Barack Obama and John McCain in the Gallup tracking poll switched leads 13 times. It simply isn't worth it to get complacent when a poll shows your favored candidate up by ten right now, nor is it worth it to get despondent when a poll shows your favored candidate down by ten right now.

To the second point (the things to remember): the hardcore electoral degenerates here already know this stuff, and can probably augment my list in the comments. But scanning the comments section here and elsewhere, these are the kinds of common bits of polling commentary that can run off the tracks. So what follows is just a smattering of bits of helpful advice to help you look at the polls and maintain your sanity, at the same time. Earlier this week, as it happens, Nate Silver offered a dozen tips for safe and sane poll watching, but I have limited myself to three key points that tend to have special resonance both in our comments sections, and in the most recent public conversations about the 2012 cycle.

1. There are dangers in reading "the internal numbers" too closely.

This admonition is nothing new. Indeed, as the world was falling all around Democratic supporters in the Fall of 2010, I wrote the following:

Conservatives jumped all over polling in both 2006 and 2008, convinced that they had gone under the hood and found that the numbers were all wrong (here is one such example). Indeed, the one recent poll that was decent news for Democrats was immediately assailed from the Right (and noted by Pollster's excellent Mark Blumenthal, who proceeded to execute a devastating takedown of the critique in question).

Could the polls be wrong in 2010? Sure, they could. Pollsters, as always, make assumptions about who will show up at the polls. Those assumptions could be in error. But hanging hopes that the critical mass of polls are in error because of this demographic quirk or that deviation from 2006/2008 is probably a glorified method of "shooting the messenger." Certainly, there's a great temptation to shoot the messengers this cycle--certainly I am guilty of it, and I'd probably be convicted on multiple counts of doing so.

When confronted with a poll that offers a result counter to your expectations (or hopes), it is tempting to dive under the hood and look for a flaw in the sample to explain it away. And certainly, odd polling results could be attributable to unusual characteristics in a sample. But it makes those results unusual. It does not automatically make them wrong. And it's probably a mistake to simply dismiss them out of hand.

Take one of the polling debates this week. Eagle-eyed readers in our Polling Wrap directed me in the comments to an article by Ron Brownstein where he noted that the Gallup tracking poll is assuming a non-white proportion of the electorate roughly similar to that of 2010. An assumption like that, in an election with Barack Obama at the top of the ticket, seems awfully pessimistic.

However, so did assuming that over 40 percent of the electorate would be conservatives in 2010, given that the number was closer to 30 percent in both 2008 AND 2006. But that is exactly what happened, according to the 2010 exit polls.

2. Speaking of conservatives ... Yes, there are a lot more conservatives than liberals in polls. It doesn't matter nearly as much as you think.

Speaking of diving under the hood and looking at the sample makeup, one recurring theme in the comments for the three-plus years I have written for Daily Kos has been "there is no way this poll is right, it has far more conservatives as liberals!"

EVERY poll does, though, and often by a substantial margin. And it almost never matters. The reason is quite simple: ideology, like party ID, is almost always self-reported. And after several decades of watching their ideological point of view used as an epithet, a lot of left-of-center voters simply prefer to self-identify as "moderates."

The proof is in the performance. As I noted last year, a study of the 2009 and 2010 exit polls showed the following:

Even in the ugliness of the 2009-2010 electoral cycle, Democratic candidates carried a majority of the moderate vote in nineteen of the 20 statewide races I tracked. The sole exception was Jon Corzine, who nabbed just 45% of the moderate vote in the 2009 New Jersey gubernatorial election (remember when people thought Chris Christie was a moderate? Fun times).

And, on the House level, even as the world was falling around the Democrats' ears, the exit polls showed that Democrats still carried moderates by a 55-42 margin.

The number to watch is not simply the lib/con gap. It will exist. It is the size of that gap that matters. In the 2006 midterms, it was a 12-point gap (22/34), and the Democrats romped. In the 2010 midterms, it was a 22-point gap (20/42), and the Democrats got stomped.

3. President Obama's approval ratings are not good. Neither are Mitt Romney's favorabilities. Both matter, but neither are terminal.

Here is my favorite "polling analyst" game of dueling banjos in the 2012 election cycle. With damned near every polling release, Republicans confidently predict that the president's approval rating means that he is DOOMED come November. Democrats, meanwhile, counter that Mitt Romney is the least liked major-party nominee in the past few decades, and that it is simply implausible that someone the electorate shows so much antipathy toward can be elected.

Both are right about the current voter mood about both men, and yet the conclusions they draw could easily be errant come November.

President Obama's approval rating is just below the historical standard for re-election: in the 47 percent range. However, as I noted a few weeks ago, there is a recent exemplar for exactly this scenario: 2004, when George W. Bush was under 50 percent in job approval, and still won re-election.

However, for Democrats confident of victory based on Romney's favorability ratings, there is an even more recent exemplar: 2010. One of the more overlooked numbers in the 2010 exit polls is that the voters who showed up on Election Day 2010 actually had a slightly higher opinion of the Democratic Party than they did of the GOP. That didn't stop a seven-point spread in the national House vote, and 62 gained seats for the Republican Party.

So, in other words, both of these numbers are worth watching, but if they stay roughly where they are now, the outcome is more likely to be unpredictable than a slam-dunk win either way.

The simplest tip is to remember that it is still April. It's worth watching movement in the polls, but helpful to remember that there will be ten movements, in opposite directions, down the road. Perspective is hard to keep (and I'll be the first to admit that I am one of the worst at keeping that long view), but it might preserve your sanity in what remains a long run to the finish line.