Rasmussen daily tracking graph for 10/10/08 - Obama 50, McCain 45 (unchanged)

[DavidDvorkin]

These Rasmussen graphs are all contained on one Web page at http://www.dvorkin.com/rastrack.html

Rasmussen links:
Data in tabular form
Discussion

[TwilightGardener]

1. Good--I'll take it.

[DavidDvorkin]

2. Gallup has been more encouraging, lately

Invigorating, even.

[seasat]

3. Does anyone know why Rasmussen polls have little change?

While the other polling firms, Gallup, Research 2000, Hotline Diego, and GWU Battleground all show swings as the result of events, Rasmussen exhibits little change in either direction. With a few exceptions, their state polls are the same way. They tend to not show large leads or sudden swings even when events like the economy or debates affect the dynamic of the race. I appears that they weight their polling data in such a way that it dampens rapid changes. What methodology does Rasmussen use that keeps their numbers closer to zero and exhibiting low volatility? Do they do this to try and keep their numbers closer to the overall outcome of the race and thus hedge their bets so that their methodology seems better?

[slick8790]

4. For one, they use significantly larger sample sizes than say, GW Battleground or Hotline.

3000 person sample vs a 800 person sample will help control volatility quite a bit. The chances of getting an extreme sample are decreased greatly. Also, I believe Ras weights their polls by party ID, taking whatever raw data they get and weighting it to their monthly rolling averages of party ID. This will also tend to dampen volatility, in case they happen to get a significantly more republican/democratic sample in one poll. As well, Ras and Gallup are 3 day rolling averages, so the previous night's results are averaged in with the results from the two days before that. Therefore, it takes 3 days for a change to be fully realized in a poll.

I would assume they do this basically for the obvious reason, to prevent from volatility. Barring an extreme event (Lehman brothers was one), public opinion tends to move relatively slowly under normal circumstances, and I think Ras is basing their methodology on this assumption.

[seasat]

6. You know, if I was smarter, I'd have read their methodology first before posting.

After your post, I went and read it. Rasmussen does not use live people but are a robo-phone poll. I also noticed that they don't include "other" in their polls. It's either Obama or McCain. Either they don't report or have an option for third party candidates, other, or undecided, while in many of the other polls allow or include those answers. It seems like that inflates McCain's numbers a point or two since other polls seem to indicate that some will choose third party over McCain.

They do seem to have a pretty robust method for party ID, though. According to what I read, they poll about 10,000 to 20,000 people each week just to determine party ID then adjust it for the next weeks polls. They show a peak of about +10 Democratic in May but have dropped it to just above +5 currently.

It just always seems to me that I get excited about a poll showing Obama up by double digits then Ras comes along with a buzz-kill close race number. We get Gallup showing Obama plus 10 when Ras has him declining from +8 earlier in the week to +5 today. Though, realistically, a 5% popular vote win is really a monumental blowout compared to past elections so I probably shouldn't worry about it.

[HughMoran]

5. You can see an average 5-6 point lead here

Which is quite strong IMO. It's not Gallup strong, but it's still quite respectable.