They use different assumptions about the primary electorate.  

is on the move LOL.

In medicine we use a technique called "meta-analysis" in which data collected in multiple studies is -- carefully -- integrated into a single data set.

I wish polling outfits did that.

since last June, National Polling is irrevelant because there is no national primary.

Could it be that HRC was leading in the poll for the two previous days, and obama's bounce of 13 offset that so that we get the +2 avg?

I don't know ~ i rarely see daily nat'l polls ~ most seem to be of the tracking variety

Your last sentence would make a more interesting diary than this one.  I think at this point, I'm not sure how good an argument you could make for it helping Clinton.  We're at the point of "rallying the troops".  If the perception is these next 4 contests can put Obama over the top - something which even Bill made - this could inspire Obama's supporters more than Hillary's.  They can be part of making history, of securing the nomination.  At this point all Hillary can hope to do is live another day.  

And it makes for a good Monday news cycle...

But I think it's a pretty meaningless factor in terms of the state of the race.

In other words, Jonathon, I agree with you entirely but just like to hear myself talk ;-)

I guess maybe they use different voter screens....most likely, one is including more independents than the other.

i HATE these nat'l polls. there is no way one candidate is up 10+ points at this moment in the race.

The first difference I see is that the USA Today / Gallup poll is "adults" instead of LV or even RV.  The lack of a voter screen usually makes for some pretty wild results.

Did anyone else wince when Susan Page referred to "Democrat voters" in the excerpt above? Does anyone else even see what is wrong with that anymore?

Assuming everything else was the same (voter screens, ect.), these two results would not be impossible based on the MOEs indicated.

With a +/- of 3%, you would get the following ranges from each poll. (Polling MOEs are usually based on a 95% confidence interval, meaning that 5% percent of the time, the "true" result would fall outside of the range indicated.)

Gallup/USA Today
  Obama 48-54%
  Clinton 36-42%

Gallup Tracking
  Obama 44-50%
  Clinton 42-48%

There's more overlap in the Obama ranges than in the Clinton ranges but it would be possible for both polls to be "correct."

(And people wonder why I argue you shouldn't use exit polling data to estimate the number of registered Democrats voting for a particular candidate.)

CNN's new poll has Obama moving ahead in Texas.

Hasn't it become clear by now that polls are pretty much useless?  They are all over the map, and frequently dead wrong.  I ignore them.

Remember this, folks. On Super Tuesday, the Gallup tracking poll had Clinton up by 13 points. In fact, Gallup registered a major jump for Clinton that day alone - I believe it was Hillary up 17. Yet, half the nation voted that day (though some much earlier) and gave the candidates 50-50.  It is impossible to square the Gallup tracker with what happened on Tuesday except to say that their likely voter model was completely off, or they just got a terrible sample that day. Since then, I've ignored the Gallup tracker.