Sunday, July 27, 2008

An Update on the Rasmussen "Leaners" and a Look at How They Affect the Electoral College

We documented earlier this week that there may be a problem with Rasmussen's decision to report two different polling results with each release and have everyone use the newer and different, "with leaners" data when reporting it secondhand or using them in an electoral college analysis. Comparing a poll from Georgia in June that had no inclusion of leaners with a poll from July that did is problematic at the very least. Below is an updated look at the polls that Rasmussen has released since the "with leaners" distinction began appearing in press releases following July 4.

Rasmussen Polls Since w/Leaners Distinction was Added (7/9/08)*
StateDatew/o Leaners
w/Leaners
Change
Undecideds Drop
Missouri
7/7+5
+50
-10
New Jersey
7/7+5
+3+2
-4
Illinois
7/8+13
+11+2
-6**
North Dakota
7/80
+1+1
-7**
Wisconsin
7/8+13
+10+3-6
Louisiana
7/9+20
+19+1
-2
South Dakota
7/9+4
+40
-4
Washington7/9+9
+8+1-6
Iowa
7/10+10
+100
-9
Michigan
7/10+8
+80
-5
Minnesota
7/10+18
+17+1
0
Kansas
7/14+20
+23+3
-9
North Carolina
7/15+3
+30
-5
Oregon
7/15+9
+90
-5
Nevada
7/16+2
+20
-5
Virginia
7/160
+1
+1
-6
Alaska7/17+5+5
0
-7
Arkansas7/17+10
+13+3-4
Georgia7/17+9
+11+2
+1
Maine
7/17+10
+8+2
-2
Colorado
7/21+7
+3+4-13
Ohio
7/21+6
+10+4
-7
Florida
7/22+1
+2+1-7
Minnesota
7/22+12
+13+1--***
New Hampshire
7/23+6
+4+2-5
Pennsylvania
7/23+5
+6+1-8
New Mexico
7/24+5
+6+1-6
California
7/24+12
+10+2-6
Avg. Change+1.00-5.7
*The "with leaners" distinction was added to reports that were released beginning on 7/9/08. The date on which these polls were conducted (The ones that these releases were based on) stretches back to 7/7/08.
**Rasmussen has only conducted one poll in these states. Therefore, the difference was taken from between the with and without leaner numbers within the same poll in these cases.

***Previous poll had been taken after "with leaners" change had been made.

Overall then, there is still a bias toward McCain in cases where "with leaners" data is replacing w the data that does not include them. As I've said, this isn't going to hae major ramifications at the outset, but over time, once a series of polls have been conducted, that collective difference could have a huge impact on what FHQ and others are doing with their electoral college analyses. In the only instance where a state has been polled by Rasmussen more than once since the switch, the results cancelled each other out. The first poll gave McCain a one point edge with leaners while the poll that followed two weeks later saw Obama with the same advantage. Such a result likely won't be the norm, however. In Ohio, for example, using one version or the other shifted the Buckeye state's 20 electoral votes from Obama to McCain and back again. And though that is the only case of a change thus far, that list is likely to grow in the future when new polling data is released.
[Click Map to Enlarge]

So looking at the map, we see that using the without leaners data -- the type that was used prior to July 9 -- shifts the map in Obama's direction with the electoral vote tally reverting to the 298-240 margin that had been in place for most of July until the new Ohio poll surfaced last week. That four point change in Ohio triggered just a 0.34 point change in our weighted average, but that is all that is necessary to move 20 vital electoral votes from one candidate's column to the other's.

It could be said that all were arguing over is whether Ohio is a toss up favoring McCain or Obama when the real point is that the Buckeye state is a toss up. Period. It doesn't really favor anyone. That's a valid point. However, there are differences in other states -- obviously -- but one poll has not been enough to affect change in those cases. Further polling, though, may bring about that change.

Note: I don't know how regularly I'm going to update these figures, but I have built them into my spreadsheet and I'm tracking the differences. When and if any changes occur -- major or minor -- I will bring them to the attention to our readers.


Recent Posts:
The Electoral College Map (7/27/08)

Guam: Oh Well, So Much for Frontloading the General Election

A Game of Vice Presidential Chicken

2 comments:

Unknown said...

A minor note, since you mentioned Georgia.

Georgia is a special case. It was polled twice in June. Why? If you look at the toplines for the second poll, you'll see "other" was no longer a choice, but "Barr" was. First of all, this left Nader voters nowhere to go. But secondly, the number sure looks as if there were leaners included: 4% combined for Barr and undecided.

The July poll for Georgia shows 5% for Barr without leaners and 1% with.

My speculation is that the second June poll was done to test out some new methodology. After being given the choices "1 for McCain, 2 for Obama, 3 for Barr, 4 for not sure," people who answered 3 or 4 were probably asked something like: "If you had to pick, would you choose McCain (1) or Obama (2)?" The system (remember, this is automated) still allows them to pick 3 or 4, but the follow-up question doesn't tell them that. That would explain why Barr drops when leaners are included, but doesn't go to 0.

So my guess is that the second June Georgia poll was the test of the new methodology, and that it became the "with leaners" methodology in all state polls starting a few weeks later.

This is all speculation on my part, and it's just one state, so I'm not recommending that any special treatment be given to that second June Georgia poll. If Rasmussen was messing around on one poll once, so be it.

But I do think it's interesting.

Anonymous said...

That is interesting.

And here I thought I was just pulling Georgia out of a hat because of proximity. I'm glad I remembered that Alabama's July poll was before the switch because I had it in there originally. Then I opted for Georgia.

Lucky me. Good stuff, Scott.