With less than a week left before Election Day, liberal pundits have been reading the polls in battleground states and don’t like what they see. Though many races are still falling inside the margin of error, Republicans are being given the edge in most of the tossup states. Despite the endless talk about there being no “wave” or this being a “Seinfeld election,” a GOP-controlled Senate next January is likely and their gains in both houses of Congress may well exceed what most observers thought was likely. In response, Democrats are doing what a lot of people do when they don’t like the way things are going: they’re crying foul and claiming the polls are skewed against their party. While there’s a chance they may be proved right, the odds are they’re whistling in the wind. And conservatives should be very familiar with the sensation.
Two years ago as we headed for the presidential vote, most of the polls were telling us that Barack Obama would win a clear if narrow victory in his bid for reelection. Battleground states that Republican nominee Mitt Romney badly needed to build an Electoral College majority were all in play but survey after survey showed him trailing. The response from some conservative pundits was to take a close look at the polls, drill down into the data, and see if the sample was kosher. Most of the major polls were built on a statistical model that seemed to overestimate the number of affiliated Democrats being asked their opinion. The samples invariably showed Democrats turning out in the same numbers as they had in 2008 when Obama swept to the White House on a cloud of hope and change charisma and messianic expectations. It seemed impossible that after four years of an indifferent presidency that Obama could perform the same magic trick again when it came to inspiring the Democratic base and huge numbers of minorities, young voters, and unmarried women to come to the polls. Seen in that light, the pollsters were skewing the sample to favor Obama and the Democrats and shortchanging Romney who might well be even with the president once the totals were adjusted to account for who would really show up and vote.
Unlike many other conservative writers who spent the year convinced there was no way as bad a president as Obama could get reelected, I felt he was more likely to win than not. His historic status as our first African-American president made him a unique political figure and his charms, though lost on me and most other conservatives, kept him popular even after a dismal record in office. But after spending enough time parsing electoral survey samples, I became convinced that wildly inflated estimates for the number of Democrats who would vote had created an exaggerated poll-driven picture of the likely outcome that was boosting Obama’s chances and hurting Romney. Along with others who had worked out the same math, I thought the likely Obama victory predicted by New York Times statistical guru Nate Silver seemed to be based on inaccurate data.
My logic was impeccable and my arguments sound. But there was one problem. I was wrong.
It turned out the pollsters were right to think that the Democratic base would turn out for Obama in 2012 the same way they had in 2008. Silver had rightly understood that the pollsters had accounted for possible changes in the electorate. More minorities and young Democrats would turn out than four years earlier or in the 2010 midterms. The result was that Obama did sweep almost every battleground state, although some were by slender margins. Silver was acclaimed as a genius and those of us who had questioned his figures and those of the pollsters he cited had to admit we were mistaken.
I retell this story not out of nostalgia for a result I still consider unfortunate for the nation but as a cautionary tale for liberals who are spending this week digging the same kind of hole I dug for myself prior to the 2012 vote. Listen to MSNBC at any time of day or read some of the people who now write for The Upshot, the section that replaced Silver’s FiveThirtyEight blog that he pulled from the Times and now operates as his own independent franchise, and you’ll see, hear, or read similar refrains to the ones I was sounding just two years ago. We are told that the sample sizes being used by the polls showing Republicans winning are underestimating Hispanics, women, or Democrats. Adjust the polls for what they think are the real totals and you’ll find that races are tied or with Democrats rather than Republicans leading.
My advice is to Rachel Maddow and Co. is simple: stop digging. Save your breath and start preparing for the worst rather than creating embarrassing sound bites that will come back to haunt you in a week.
As the redoubtable Silver notes today on his site, errors on the scale that conservatives thought possible in 2012 or liberals are alleging today are always possible but not terribly likely. Talk about skewed polls is the last refuge of those in denial about an electoral trend. That was true two years ago and it’s happening today. The temptation to try and “unskew” the polls is obvious and it’s what readers in our bifurcated media want to see. But, as Silver writes, “Usually this doesn’t end well for the unskewers.”
With so many polls out there showing much the same thing about a Republican advantage, the chances that they are all wrong about who will vote (or have already cast ballots in early voting states) are slim. Unskewing seems like it makes sense but it is invariably based more on wishful thinking than sober analysis. Just as conservatives had to eventually accept that pre-election poll estimates of Democratic turnout were right, so, too, will liberals likely have to own up to the fact that today’s expectations about their base’s voting patterns are similarly accurate. Indeed, as Silver writes, it may be that pollsters are underestimating the number of Republicans this year just as they did the same to some degree for Democrats in 2012.
This should not cause us to lose all skepticism about polls. They should be closely examined and probed for possible errors. But such analyses tend to be based on the idea that the candidates you prefer are being shortchanged more than a real suspicion of error. Assuming that the errors will all go one way or that your candidate will catch the breaks is a guarantee that you’ll soon be eating your hat, humble pie, crow, or whatever metaphor you prefer. Ms. Maddow and her friends will soon find that it doesn’t taste any better in their mouths than it did in mine.