This could be huge! At the least it's very interesting....
A series of linked discussions from MSNBC's Chuck Todd led 'First Read', from Marc Ambinder, and from Nate Silver's 'fivethirtyeight.com' explore the implications of an ABC / Washington Post poll which shows Obama as having a 3%, 8% or 12% lead depending on how you count (or weight) it.
How can the figures be so different? And what are the implications of these differences? More below the fold...
According to the First Read article, the 3% Obama lead is based on a model of "Likely Voters" (LV). The 8% Obama lead (50-42) is based on a sample of "Registered Voters" (RV). And the 12 lead is based on a sample of all adults.
Here's the First Read quote:
There are a couple of more things worth noting from the Washington Post/ABC poll. One, it appears that a bigger turnout benefits Obama. While the Illinois senator has an eight-point lead among registered voters, his edge is much smaller when you reduce the race to likely voters (49%-46%). (And right now, pollsters will tell you that with older voters leaning McCain these days, any likely voter model is going to favor McCain for now. If Obama moves younger voters as well as many observers assume come October, the likely voter numbers could change). In this poll, when you expand it to include all adults, Obama’s lead is 12 points (51%-39%).
The implications of this seem to me pretty stunning. First of all, the suggestion is that traditional "likely voter" models (which often favor older voters, who support McCain) aren't really valid at the moment.
Sean at fivethirtyeight.com explains it this way:
...when you read in a poll that Obama is further ahead among registered voters than likely voters, be very very skeptical of the conclusions that pollster has made about likely voters. As Pew finds, the age group and partisan breakdown of how people are engaged this time are dramatically different than past patterns. Pollsters are using old assumptions about older voters being more engaged that don't appear to hold water this time around.
This helps to answer the question of "Why isn't Obama winning by more?" He is or he isn't, depending on how you count the voters. Nate and Sean at fivethirtyeight.com espouse the view that for the moment the RV model is the most reliable and that seems reasonable to me. ABC/WaPo apparently agrees with this, too, since they headline their RV results rather than their LV results. In short, they don't find their own LV model to be credible, though they haven't yet changed their model. But other polls may still be using their old LV models without questioning them.
The implication of this is that if registered voters just vote their own numbers, as they seem likely to do, then Obama wins by eight. Cool!
But these numbers also seem to validate emphatically the wisdom and value of 50 State Strategy and the Dean/Obama nexus's unprecedented focus on voter registration. If all the adults in the US voted their hearts (or their pocketbooks), then Obama wins by 12%!
The "all adults" number seems like a target we can shoot for, even if we don't hit it in 2008. Clearly, those numbers suggest the possiblility of a strong and lasting Dem majority for the future. Moreover, when a given poll shows the 2008 election to close, either in the GE or a given state, we should ask what kind of voter model is being used: RV, LV or "all adults." The answers might explain the wide disparity in recent polling.
Clearly, we Democrats have got the right strategy and the right candidate for 2008. And the clear path to victory involves supporting that candidate via voter registration and GOTV on the broadest possible front.