<i>because I don't see how they can be explained in terms of different voter screens...</i>
I assume you're using loess (or similar) smoothing because you assume that the relationship, or the Rasmussen shift is only piecewise continuous. You've identified three regimes: early cycle, intermediate and post-convention. Rasmussen has explained that his LV screen is <a href="http://www.pollster.com/blogs/so_whats_a_likely_voter_answer.php">tighter when approaching the election:</a>
<blockquote>There's a challenge to defining a likely voter. The process is a little different than in the week before an election for us than it is in two months before an election than it is in a year before an election.</blockquote>
What I would probably due to get on the right track is as follows -- take the limiting cases: early in the cycle and right before the election.
Group the other polls according to their survey samples -- adults/registered voters/LV. Try the scatterplots again for Ras vs survey type, esp Ras vs surveys using adults:
<blockquote>What does it mean in practical terms? Rasmussen Reports and Gallup are the only two polls out there with a daily tracking poll of the President's job approval. If you go back from January 20th on, most of the time you will see that Gallup's reported number is about three or four or five points higher than ours, because these are surveys and there is statistical noise. Sometimes the gap is bigger, sometimes its smaller. In fact there are some days when our number is a little bit higher than Gallup's. But typically, the gap between the adults and the likely voter sample is in the four or five point range.</blockquote>
Essentially, you're looking for this effect in the context of last year's election polling. With respect to the random noise, I would probably double-check that it's truly random --> if not, this should reveal some clues.
When you look at the weeks before the election remember that some of the other pollsters are transitioning to LV models, generally speaking.
I should add that data collected for BLS should be specifically for the employed, so the weighting would exclude children. However, retirees wouldn't be in the data, either. I'm certain the overall skew does not reflect a realistic turnout model for a Democratic caucus.
Let me preface my extended comment by saying that I think it's clear that Obama has momentum in IA right now.
However, this poll is worthless without full crosstabs. The Edwards' numbers are suspect, which prompted me to read the article in full. At the end of it:
Results are weighted so that the sample demographics match Census Current Population Survey parameters for gender, age, education, race, region, and population density. The overall margin of sampling error is plus or minus 3 percentage points for results based on 1,408 registered Iowa voters. Results based on smaller subgroups are subject to larger margins of sampling error.
Mmmmkay ... well the CPS is collected by the Census Dept for BLS. Too lazy to navigate DOL's site for IA-specific demographics, so I went to the state of Iowa's site instead at iowadatacenter.org.
According to the census data, 40% of Iowa's population is 45+ ... this means the poll results were weighted so that 60% of caucus participants are under 45? Well, according to the 2004 entrance poll, it was 32%.
This poll is worthless because the turnout model is beyond unrealistic. I don't believe Edwards is a distant third. Can't comment on Obama vs Clinton. The only thing it tells us -- which I think we already know -- is that Obama does well with under 45.
Here's an article about Obama and the health care effort in Illinois:
When Barack Obama and fellow state lawmakers in Illinois tried to expand healthcare coverage in 2003 with the "Health Care Justice Act," they drew fierce opposition from the insurance industry, which saw it as a back-handed attempt to impose a government-run system.
Over the next 15 months, insurers and their lobbyists found a sympathetic ear in Obama, who amended the bill more to their liking partly because of concerns they raised with him and his aides, according to lobbyists, Senate staff, and Obama's remarks on the Senate floor.
there is no question that within the Rasmussen universe the race has shifted
How so? In the last month the average separation has been 21 pts with a standard deviation of 3. That means anything ranging 15%-27% is within MOE at 95% confidence.
By your logic, Obama's upward drift from 19% (11/9) to 26% for two consecutive days, (11/17 - 11/18) meant that the "race has shifted in the Rasmussen universe," a drop in the difference from 26% to 15%, but as we can clearly see, it was just a fluctuation within the MOE over an 10 day period.
If Obama is averaging around 17% in two weeks, we'd be looking at evidence of a real trend. If we treat each four-day rolling avg as a valid measurement, you'd need a few to discern trend.