ABC / Washington Post House Ballot Poll Stands Out In More Ways Than One

Photo courtesy of http://historicalromneys.tumblr.com/

You know a poll finding might be a little screwy when the actual pollster releases the information with an explanation or caveat. And that's exactly what happened on Thursday morning when Gregory Holyk and Gary Langer of Langer research wrote this in explaining the substantial Democratic lead on the 2014 generic House ballot question in ABC/Washington Post's new national survey:

While it’s far too early to handicap the 2014 election in any serious way, registered voters currently favor the Democratic candidate over the Republican in their congressional district by 48-40 percent, the largest Democratic midterm advantage since 2006. The party saw steep losses in the 2010 midterms.

The result almost entirely reflects a current Democratic advantage in partisan affiliation. Among registered voters in this survey, 33 percent identify themselves as Democrats, 22 percent as Republicans; most of the rest are independents, and they split evenly in their 2014 preference. Democrats outnumbered Republicans by 6 points in the 2012 presidential election and by 7 points in 2008, but they were even in the last two midterms – meaning the Democratic advantage holds only if their midterm turnout improves dramatically.

What Mr. Langer is saying here is not that their finding is incorrect (though that is certainly possible), but rather that if it is accurate, it would break precedent. And as I've pointed out on this blog before, he's right. A Democratic partisan identification advantage over Republicans of 11 points is just very, very unlikely in a midterm election. In fact, partisan identification in every midterm election since 1990 has ranged between D+3 and R+1, a fairly tight window of difference.

Moreover, a glance at the 2014 House Ballot polling over the last month shows ABC/WaPo is a bit of an outlier, both in terms of party I.D., and the overall topline result:

Because Rasmussen Reports tracks the generic House ballot question once a week, their findings are listed as a monthly average for the sake of the chart.

As you can see, the average of pollsters shows the Democrats with just a 3 point advantage over Republicans on the generic House ballot, not 8 points, and with a 7 point party I.D. advantage, not 11 points. So not only is the Washington Post poll an outlier in the scheme of historical 2nd-term midterm election results, it's an outlier when compared to recent House ballot polling.

But pointing that out apparently gets you labeled a poll truther or "unskewer" by the poll snobs at Huffington Post's Pollster. The NRCC noted all of the things I mentioned above on Thursday in a post meant to call into question ABC/WaPo's House ballot finding. In mocking fashion, Mark Blumenthal, without dealing at all with the merits of the NRCC's arguments with respect to the poll, essentially dismissed their concerns as conspiratorial. It's one thing to arbitrarily change polling data and insist your results are correct (as famed unskewer Dean Chambers did). It's another to simply point out bizarre or unique polling data (as the NRCC was doing).


In light of the poll-snobbery from Blumenthal, and why not...lets "unskew" that interesting ABC/WaPo House ballot finding to see what they would have found, all other findings remaining the same, if party I.D. was the same as the last Presidential race (38% Democrat, 32% Republican, and 29% Independent), or the same as the last midterm race (35% Democrat, 35% Republican, 29% Independent).

So there you have it. Shockingly (look, I can be snarky too!), had the poll NOT found a historical Democratic partisan ID advantage over the Republicans, we would have a much closer race on our hands. When the numbers are reweighted to the 2012 Presidential electorate finding (and remember, midterms electorates tend to favor Republicans more than Democrats as opposed to Presidential electorates), the overall Democratic lead drops to 5 points. When the ABC/WaPo poll is reweighted to the party ID from the last midterm election in 2010, the race is essentially tied, with the Democrats dropping to a one point overall advantage.

On a final note, I couldn't help but notice ABC/WaPo's final survey prior to the 2010 midterm, released on October 28, 2010 (5 days before the election). They had the Democrats leading Republicans on the House ballot 49-44%. The actual result 5 days later? Republicans won a 51-45% popular vote victory, making ABC/WaPo off by a net 11 points. Making them HORRIBLY wrong. Oh, and the party I.D. on that poll? You guessed it. D+10. The party ID on actual election day? D+0.