Gallup released its generic ballot results:
http://www.gallup.com/...
Now, let’s set aside for the purposes of this post the weaknesses of a generic ballot test for national popular vote in an election more than four months away when the result that matters is the outcome of individual districts where incumbents outside of narrowly defined "swing districts" rarely lose.
Instead, I want to point out where knowing more about the data behind the polls actually reveals more about the polls than either the media or the pollsters themselves are willing to share.
One strength of Gallup’s is their increasing willingness to divulge without prompting some of their cross-tabular information. However, today’s numbers also reveal the peril of releasing partial results.
The results:
GOP – 49 (+3)
Dems – 44 (-2)
The cross tabs for type of people voting for Dems:
Dems – 94 (+2)
Indies – 38 (+3)
GOP – 3 (unch)
The crosstabs for enthusiasm about voting:
Dems – 28 (+4)
Indies – 26 (+2)
GOP – 43 (-3)
Total – 33 (+2)
So more people are enthusiastic to vote, but more Democrats and Indies in particular are enthusiastic than before. More Dems and Indies are also inclined to vote for Democrats for Congress. If more Dems and Indies are enthusiastic to vote and want to vote Dems, shouldn’t that move the Dem numbers up?
Nope, not so much. Instead they lose 2 points.
And given that the party breakdown shows no drop off in support for Dems, the only logical explanation is that the universe of likely voters picked up this week by Gallup have trended more GOP-friendly than last week, as that’d be the only way to make sense of the math behind the contradictory cross-tabs trends .
But when you compare that conclusion with the enthusiasm’s shift, you can only conclude from there that there is no meaningful correlation between enthusiasm about voting and likelihood to vote. And if there is no correlation there, then continuing to measure and report on enthusiasm in lieu of likelihood to vote is completely and utterly a waste of time.
When you cherry-pick the cross-tabs to report to the public, you either make people wonder about the rest of the cross-tabs you didn’t report, or you help reveal the utter pointlessness of the ones you do report on. In this case, Gallup helped show that enthusiasm does not matter, and people should stop ordering up polls that measure it.
UPDATED: I plugged the above numbers into a simple Excel spreadsheet. In order to get the Dem vote to drop 2 points even as they increase their support from the self-identified Dems and Independents DOES require a sudden shift in the universe of likely voters.
By my count, 44% of vote achieved with 94% of the D's, 38% of the I's and 3% of the R's (44=0.94D+0.38I+0.03R) requires the LV universe to be 33% D, 32% I, and 35% R. Running the numbers Gallup had last week (46=0.92D+0.35I+0.03R) creates a LV universe of 37%D, 31%I, and 31% R.
So even though Dem enthusiasm rose by 4 points and the GOP dropped by 3, and even though the Dems gained support among Dems and Indies, the GOP sees an overall gain of 5 points on the generic ballot test because the LV universe became 8% more Republican switched from a D+6 to a R+2.
There appears to be little or no statistical correlation between enthusiasm to vote and likelihood to vote. But I'm no Nate Silver, so I can't be completely sure. And as a commenter pointed out, Gallup doesn't weight its LV universe by party affiliation. But that really does make comparing top-sheet numbers from week to week awfully difficult, especially without full disclosure of their cross-tab results.