Pollster.com

Articles and Analysis

 

Morning Status Update for 10/14

Topics: Status Update

I delayed this morning's update to include the new Quinnipiac surveys that include pre- and post-debate samples in the battleground states of Colorado, Michigan, Minnesota and Wisconsin. We thus have 21 new statewide surveys logged since this time yesterday.

2008-10-14 daily

While most indicate progress for the Obama/Biden ticket since the last poll by the same pollsters, trends over the last week are less evident. The table above includes 12 new polls that are in states where the pollsters have released two surveys since October 1. Seven (7) of the 12 show net movement in Obama's direction, two in McCain's direction and three show no change at all in the margin.

2008-10-14 trends.jpg

Nonetheless, the new polls continue to improve the Democrats standing on our state trend estimates. In 10 of the 11 states we have been tracking, the new polls increased the Obama margin slightly.

Probably nothing shows how far the Democrats have come in just a few weeks than the states that shifted to our dark blue "strong Democrat" designation over the last 24 hours. The new surveys increased Obama's leads on our trend estimates to nearly 10 points (51.1% to 41.3%) in Michigan, 8 points (51.7% to 43.7%) in Colorado and nearly seven points (51.2% to 44.4%) in Florida, all enough to merit a "strong Democrat" designation. As such, we now show Obama leading by "strong" margins in states with 256 electoral votes, just 14 shy of the necessary 270 (and that does not include the 64 electoral votes classified as "lean" Obama).

Maine, where Obama leads by 7.5 points (50.7% to 43.2%) remains in the "lean category" because pollsters there typically use smaller than average sample sizes.

Virginia remains just barely in the toss-up column despite a 3.8 point margin for Obama over McCain (50.3% to 46.5%), also due mostly to smaller than average sample sizes there.

North Dakota, one state not listed above, shifted into our "toss-up" category on the basis of a new poll yesterday from The Forum and Minnesota State University Moorehead showing Obama with a 2-point advantage over McCain (45% to 43%). A better designation would be "not enough information necessary to classify. We have logged only four polls total in North Dakota since the conventions, and this new poll is the first new data since the Democrats started gaining in mid-September. Moreover, the samples sizes of the few polls are smaller than average. Our trend estimate still shows McCain ahead (47.4% to 43.8%) but not by enough to remain in the "lean Republican" column. Take that "toss-up" designation with a big grain of salt.

Finally, the national margin had narrowed a bit over the last 24 hours, but that is mainly because we started including Gallup's new likely voter results and because of a new tracking poll from Investor's Business Daily/TIPP that shows Obama leading by just two percentage points.

 

Comments
Tominstl:

Shouldn't we toss Missouri inot some shade of Blue given the most recent polls?

____________________

m913:

And should not Minnesota, Wisconsin and Washington be dark blue?

____________________

Read about Sarah's Latest Slam.
____________________


North Dakota and Montana are "Bettors Curiosities" :

These are the Last Prices to buy shares at INTRADE :

Montana (3EVs) .......... Obama 31.0, McCain 73.0
North Dakota (3EVs) ..... Obama 27.0, McCain 75.0

Very remarkable that people risk their money for Obama in these two states. They were assumed 100% secure for McCain a few weeks ago.

Are there any indications of Obama going up in some polls ????

I have been following the Toss Up States behaviour in bets. Everyday I fill the prices in my tables.


Is Obama investing resources there ????

Vicente Duque

____________________

DTM:

Keep in mind Pollster's categorization of states is purely mechanical, meaning it is just a mathematical function of their trend estimates and relevant poll sizes.

____________________

JFactor:

Mathematical models aside (not that they're not useful, they very much are) we all know that Wisconsin, Washington, Maine and Minnesota are "strong DEM" in real life. I suggest that all of you go to RCP and make your own electoral map which you can modify yourself and take into account all the recent polls. It's really a quite handy software.
___________________________________________
http://www.internationalpoliticstoday.com

____________________

Loyal:

Regarding the investor's business daily poll: a spreadsheet and about 15 minutes has convinced me of the following.

IBD defnitely has between 33 and 38% independents and probably has the following breakdown:

D 35%
R 32%
I 33%

which would result in Obama: 44.57%, McCain 42.82% and Undecided 12.61%.

For what it's worth.

Loyal

____________________

DTM:

@JFactor

And to back up your point, Pollster has been hinting recently that they really need a more complex categorization scheme, one which recognizes times when there is not enough current data in a state to make a reliable choice between categories. So, for example, I think it would be fair to say it is likely several states are clinging to "lean" status simply because there hasn't been enough recent polling.

Here, by the way, is where I think something like Nate Silver's "trendline adjustment" comes in handy. Basically, he tries to scientifically use data from the national polls and other state polls to augment the polling available in a given state.

____________________

jme:

@DTM and JFactor

There has been some discussion of pollster's coloring scheme for their national map. However, I would remind everyone that the guiding philosophy of this site (as I understand it) is to clearly and impartially display polling data. Pollster isn't interested in prediction, as Nate Silver most definitely is.

As such, much of the criticism of the current coloring scheme is based on the observation that the colors sometimes don't correspond to the actual polling data for that state in an intuitive way. Not that the coloring doesn't match our expectations for particular states on election day, or based on demographics or recent election history or the national polling.

So I seriously doubt that pollster would implement anything remotely like Nate Silver's stuff (which is really cool, just not what pollster's all about).

Indeed, my complaints about the coloring scheme were aimed at avoiding the mistake that I think JFactor makes: namely that the pollster map colors reflect a prediction about election outcomes rather than simply summary of current polling data.

____________________

johncoz:

@jme

I think you misunderstand what Nate is doing -- it's a projection not a prediction, and there's a big difference. A projection is firmly based on the current data, and does not pretend prognosticate on future trends or events.

____________________

cube:

It seems that Obama is improving in state polls but holding steady in national polls.

The state polls concentrate on battleground states. Could the apparent discrepancy be that Obama is improving in battleground states, where both candidates are focusing resources, but Obama is doing a better job, while Obama is slipping a little in non-battleground states?

____________________

DTM:

@jme

Nate Silver is actually doing a number of different things, and helpfully he reports the results of his different processes. And what I was referring to was not his projection, but rather his trend-adjustment, which actually isn't predictive.

That said, I wasn't necessarily suggesting that pollster adopt his methodology--although if they wanted to do something similar on their own, I would be very interested in the results.

____________________

jme:

@johncoz

No, I understand what Nate is doing; I guess I'd respond by saying that I think you're drawing a (linguistic) distinction (projection vs. prediction) without a difference.

I never meant to imply that his predictions (or projections, whatever) aren't based on data, models etc. Maybe a better word for us to settle on is extrapolation. Nate Silver is taking current data (and not just polling data!) and extrapolating likely election outcomes based on that. Pollster (as I understand it) is not so interested in doing that. They just skip the whole extrapolation part and instead present the data in a useful form.

Both are fine...I would just argue that the world needs both, so I would hope that pollster refrains from engaging in projections (or predictions!).

____________________

jme:

@DTM

Fair enough; I'm not trying to pick a fight about Nate Silver and his stuff, which, again, I really like.

I would point out that his trend adjustment is, technically speaking, part of his projections, since they are a piece of the input for his models. But no matter...

____________________

RossB:

Cube: I think that is a fair assessment. Obama has done well where he has spent the most time. Of course, McCain has improved, but just not as much. Obama spent a lot of time in Iowa, and it is now solid blue. After the primaries, Michigan was a tossup, but he has spent a lot of time there, and turned things around.

One thing I noticed in the CNN instant poll tracking on the debates is how popular both candidates are. Despite a nasty campaign, both candidates are well respected -- their polling numbers go up when they speak. This is why Obama bought the half hour air time. He figures if it is close, he can convince folks at the last minute to vote for him. This is a quite different from the last couple of elections, when negative factors (hatred of Bush over the war, distrust of Kerry over Swift Boats, distrust of Gore over his exaggerations, concern over whether Bush is intelligent enough). In those elections, exposure didn't always help. In this one, I believe it does.

____________________

DTM:

@jme

Sure, Nate's trendline adjustments are a step on the way to his projections. But my point was just that if you are interested only in the problem of sporadic state polling, and think it is a good idea to factor in national polling and polling in other states, you can use just his trendline adjustments and not his projections.

____________________

jme:

@DTM

We agree!

Seriously, though, you're right. Through the magic of the intertubes, we're talking about different things. ;)

Nate's use of national polls to infer trends in states with sparse polling is cool. I'm just perfectly happy to let Nate do that, and let pollster keep doing what it does, which is focus on data/trend visualization. It was a philosophical/aesthetic point, not a judgement on Nate's techniques.

____________________

burrito:

Regarding Gallup:

It appears that Gallup is publishing 3 numbers now (instead of the 2 that were discussed by Mark yesterday). Their headline still takes under consideration the "registered voters" number. However when you click on the link, you can see 3 numbers: "registered voters", "likely voters (expanded)" and "likely voters (traditional)".

Mark discussion went on what numbers to use, either "likely voters I" and "likely voters II" and opted for consider the second one. From what I can see, the numbers for "registered voters" and "likely voters II" are similar, with "likely voters I" narrowing the gap. May I suggest that the 3 numbers are used when posting the poll results and maybe averaging the 3 to obtain a number to be used for the national number calculation?

____________________

sulthernao:

You included the wrong Gallup poll. And you are comparing two different Models in the chart above.

____________________

burrito:

@sulthernao:

"You included the wrong Gallup poll. And you are comparing two different Models in the chart above."

Yes, that is the "likely voter I" instead of the "likely voter II" number. I was trying to explain that in my previous post. Gallup has made things more complicated for sites that collect polling data, as now they (Gallup) have 3 different numbers instead of 1.

____________________

I found this great cartoon using the Pollster.com presidential trend line graph on the Reasons To Be Cheerful, Part 3 blog: dead on!

____________________



Post a comment




Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.

MAP - US, AL, AK, AZ, AR, CA, CO, CT, DE, FL, GA, HI, ID, IL, IN, IA, KS, KY, LA, ME, MD, MA, MI, MN, MS, MO, MT, NE, NV, NH, NJ, NM, NY, NC, ND, OH, OK, OR, PA, RI, SC, SD, TN, TX, UT, VT, VA, WA, WV, WI, WY, PR