Virtually all of my work here at DK has been in the realm of following the polls and offering polling analysis. It is what I did when I was one of the quarter-millionish diarists roaming the site, and it is what I continued to do when I moved onto the front page in the Spring of 2009.
But, long before that, my main interest in politics was election forecasting. It was a passion that I began when I was a teenaged college sophomore at UCLA, and predicted the 1992 Presidential election. I did fairly well in that initial outing, hitting 49 of 51 contests (I thought that North Carolina and Florida would pad Bill Clinton's landslide margin). In 1998, I even won a contest sponsored by CNN for being one of the few voices in America not predicting the GOP landslide that everyone saw coming.
In the past three cycles, I have come pretty close to the mark, no matter whether the outcome was a good GOP year (2004), or a good Democratic year (2006/2008).
This time around? I will be the first to admit that I don't have a freaking clue.
This is the most unpredictable election in my memory, and, indeed, I can't even think of one that rivals it.
I'm not alone in that assessment. Just as Nate Silver was projecting a 50-seat gain for the GOP in the House, he couched in the caveat that the uncertainty of the races could mean as small a pickup for the Republicans as 20 seats, and as much as 80 seats. Such a wide range of possible outcomes made a few people smirk at the "prediction." But from where I sit, that's a pretty appropriate assessment of the landscape.
In short, there are very few outcomes in nine days that would surprise me. And that is because so little about this election cycle is definitively known.
Consider the tools we usually use to project electoral outcomes. They are, without a doubt, providing as many questions as answers in this most unique electoral cycles.
Usually, an election forecaster makes the bulk of his/her determinations about how an election is going to go based on the polls. But this year, that has proven more difficult to utilize, and for two reasons.
- The Likely Voter Thing: We have been told for years that pollsters use "likely voters" because it provides a more accurate landscape for taking a snapshot of the electorate. After all, what is the use of recording the preferences of voters who aren't, in all probability, going to vote on Election Day, anyway?
While this is inherently logical, there is a problem with that assessment this cycle. There have always been gaps between the outcomes of polls between "likely voters" and the less restrictive screen of "registered voters." Historically, that gap has equated to better Republican performance among likely voters (as Alan Abramowitz noted over at Pollster a while back). But that hasn't always been the case. In the Democratic wave election of 2006, the majority of the polls which offered both LV and RV data had their LV screens slightly favoring the Democrat. What's more (and this could have big implications for 2010): those screens almost universally (75% of the time) overstated the Democratic performance in the race.
Indeed, in the last two election cycles, the "registered voter" screen has been closer to the final outcome more often than the "likely voter" screen. And it wasn't all that close: the RV screen came closer 57% of the time, the LV screen was closer to the truth 38% of the time, and they split the difference 5% of the time.
Another unique feature of the electoral cycle has been the width of the gaps between RV's and LV's. In the 2006 and 2008 cycles, the majority of the gaps between RVs and LVs consisted of gaps ranging from 0 to 2 points. In this election, gaps as wide as 13 points have been reported (an early October CNN poll of Nevada). Of course that has been owed to the most oft-used phrase of the cycle: the enthusiasm gap. More on that later.
- The Pollster Hegemony Thing: Around two months ago, I noted that a majority of the polls in my database for this cycle either emanated from Republican private polling or from Rasmussen (which, of course, could easily be described as Republican public polling). As the trickle became a flood of data (what was, at that time, around 1000 polls is now 2040 and counting), those numbers have changed, but only slightly. Even at this point, 47% of the polls released in this cycle have come from GOP or GOP-sympathetic (read: Rasmussen) sources. By contrast, just 8.5% of the polls have come from Democratic sources.
This can be read two ways. For one thing, in 2006, it was Democratic polls that swamped polls from GOP sources. Read this way, this can be interpreted as a bad sign for Democrats. Republicans are releasing more polls because...well...they can. The data contained in them is data they want to have the public consume. But there is an alternate viewpoint: they can also simply be skewing the reality of the electoral situation. As Nate noted last week, these polls come with a thumb on the scale. If (as has happened this year), twice as many sponsored polls have been come from the GOP side as the Dem side, it will paint a picture that is, perhaps, overly optimistic for the Republicans.
Most of this Republican polling, like the LV/RV gaps I described earlier, makes note of what these pollsters project as an enormous enthusiasm gap. One constancy of polling in this cycle has been the indication that Republicans are more enthused about this election than are Democrats. At this point, that is probably beyond dispute. What is disputable, however, is the corollary that states that, because of the enthusiasm gap, the partisan makeup of the 2010 electorate will be dramatically different. Some pollsters, for example, are anticipating an electorate that is more amenable to the GOP than 1994.
There is some evidence that such audacious predictions about the R/D gap in the 2010 electorate might be unwarranted, however. Early voting statistics have painted a decidedly different picture of the electorate than what we have been sold by the pollsters and the pundit class for most of the cycle:
- In California, site of a critical gubernatorial and Senate election, a study by the Atlas Project shows that the early/absentee vote in 2010 has slightly favored Democrats (43-40). In 2006, it was split almost evenly.
- In Colorado, where Dems are seeking to hold onto a very vulnerable Senate seat and a trio of House seats that are in some semblance of danger, the absentee statistics are similar to where they were in 2008, which was a very good year for the Democrats in Colorado. While Republicans have requested more absentee ballots, the percentage of ballots requested by the Democrats (37%) is actually fractionally better than the percentage sought by Democrats in 2008.
- In Nevada, the voice of Nevada politics (Jon Ralston) tweeted on Saturday that the GOP edge in early voting thus far amounts to less than 150 votes. This is notable, because this is one of the states where the RV/LV gaps were the most immense. If there is a huge enthusiasm gap in Nevada, it has not shown up in early voting. Dem-friendly Clark County cast 67% of the early vote in 2006. In 2010, it is 65%. A gap, but not a big one.
- In Ohio, an analysis (in the name of full disclosure, one done by the Ohio Democratic Party) showed that Democrats have cast 44% of the votes in early and absentee voting, compared to 34% cast by Republicans. Ohio has a toss-up race for Governor, to say nothing of close to a half-dozen potentially perilous House seats for the Democrats.
There are other examples, as well. A study by a diarist at The Swing State Project shows that early voting percentages in Texas are essentially similar in Obama counties and McCain counties, with a spike in counties around Houston, where Democrat Bill White served as Mayor.
What seems conclusive, at least thus far, is that it is difficult to find enormous enthusiasm gaps anywhere. Even in places where early/absentee votes have lagged a bit of 2006 rates, the differences have not been particularly dramatic.
Republicans, however, point out that the early voting numbers aren't the best indicators of turnout, because the Democrats have made that a feature of their GOTV operations to a much greater degree than the GOP. They point to some dramatic differences in early voter numbers from 2008 to 2010 as evidence of the enthusiasm gap, since those are both years that Dems emphasized early voting (and some of those gaps--North Carolina immediately comes to mind--are pretty profound).
But the bottom line is that if Democrats are able to turn out their voters, then the projections in almost every likely voter screen about the GOP turning out in far greater proportions than the Democrats (thus altering the political landscape) would appear to be unjustified. If that's the case, the GOP would need a massive, massive edge with Independents to pull off the kinds of gains that they are breathlessly predicting.
It could happen, of course. And that, frustratingly, is the bottom line. With nine days to go in this election cycle, pretty much anything could happen. Never have there been so many competitive races, so late in the cycle. Never have the traditional predictors of an election been so cloudy.
This is going to be fascinating to watch.
