17
   

Get yer polls, bets, numbers & pretty graphs! Elections 2008

 
 
nimh
 
Reply Wed 12 Dec, 2007 06:36 pm
CENTRALISATION - is the word for today.
  • Topic Stats
  • Top Replies
  • Link to this Topic
Type: Discussion • Score: 17 • Views: 130,006 • Replies: 3,166

 
Cycloptichorn
 
  1  
Reply Wed 12 Dec, 2007 06:37 pm
Re: Get yer polls, bets, numbers & pretty graphs! Electi
nimh wrote:
CENTRALISATION - is the word for today.


Is this open to all? I have a few...

Cycloptichorn
0 Replies
 
ossobuco
 
  1  
Reply Wed 12 Dec, 2007 06:39 pm
Would you mind explaining what you are talking about?
0 Replies
 
nimh
 
  1  
Reply Wed 12 Dec, 2007 06:48 pm
Re: Get yer polls, bets, numbers & pretty graphs! Electi
Cycloptichorn wrote:
Is this open to all? I have a few...

Of course! Absolutelly, go right ahead.
0 Replies
 
nimh
 
  1  
Reply Wed 12 Dec, 2007 06:48 pm
ossobuco wrote:
Would you mind explaining what you are talking about?

Election polls, surveys, bets and speculations, and any attendant tables and pretty graphs! Razz
0 Replies
 
nimh
 
  2  
Reply Wed 12 Dec, 2007 07:53 pm
Excuse me while I copy the context/explanation for the below graphs from earlier posts in different threads..

It's no news that I'm a big fan of pollster.com, and specifically, its lovely and insightful graphs, in which all national and state polls are tracked, trendlines and all.

For example, these ones on the Democratic race in Iowa, the Republican race in Iowa, the Democratic race in New Hampshire, and the the Republican race in New Hampshire.

But sometimes I'm a bit frustrated by them. The pollster.com trendlines are extremely cautious. With good reason, but still. Even when new results come in that diverge from recent ones, mostly the trend line is just made a little more or less steep; actual "bends" or turns in the trend are made only once a three months or so at most.

Again, all for very sound methodological reasons. But I'm following these races and sometimes I see what could just be new trends emerging or turning, and they wont show up in the pollster.com trendlines (yet) because they are too tentative.

And then on the other hand, you have the day-to-day news reports that uncritically zoom in on any one single opinion poll result, and act like it proves some new breakthrough. Like in one week in late November, one news story would conclude "Romney within 4 points of Giuliani", on the basis of a national ARG poll, while another concluded, "Romney down to 8%", based on a Fox News poll.

So, sometimes I want something of a middle way. Hence - new nimh graphics! Smile

What I've been doing is taking the individual poll results (on both Iowa and NH), as registered on pollster.com, and calculating averages for bunches of five to eight polls at a time. So as the primary campaigns have heated up, the intervals have gone from two months to half a month.

The result is a bit more sensitive to short-term trends than the pollster.com trendlines -- but also more vulnerable to the impact of outliers and other statistical "noise".

http://img145.imageshack.us/img145/1046/nhdemsonlyminezk0.png


Now in the image below I've compared the two graphs. On the right is my graph, as shown above but squeezed a bit; on the left the corresponding bit (1 June - 31 December) from the graph on pollster.com.

For those who dont know pollster.com; in its graph, individual (national) polls that come out are represented in dots: each dot represents the result of the candidate in question in one poll, placed on a timeline. On the basis of that data Prof. Franklin plots out the trendlines that you see. It's a regressive trendline (dont ask me) - it's not just a rolling average, but a trendline that is continually adjusted regressively as new polls come in. It's set up in such a way that it will not be impacted much by one or the other divergent individual poll result, it's only solid trends they want to pick up on.

http://img139.imageshack.us/img139/9625/nhdemsboth121307yj9.png
0 Replies
 
nimh
 
  1  
Reply Wed 12 Dec, 2007 08:33 pm
Here's the equivalent images for the Republican race in New Hampshire:

http://img338.imageshack.us/img338/6307/nhrepsonlymine121307de0.png


http://img266.imageshack.us/img266/9989/nhrepsboth121307jb4.png
0 Replies
 
fbaezer
 
  1  
Reply Wed 12 Dec, 2007 09:03 pm
I JUST KNEW Nimh (the lab rat who learn to read) would come with this marvelous thread.

Nice summarization of the ongoing races.

I'll play picky on technicalities:
A HUGE problem of condensing polls is that you tend to combine apples with pears; that is, polls based on different sampling methods and with different type of field work. A few biased polls will ruin the whole "sample".
The best way to analyze polls is to separate the series from the different pollsters. You'll notice that shifts in each serie will tend to be similar, though in some cases with a short "time lag", and that several of them will move like a mirror, while other ones have bigger compasses. In time, you learn which onesw to trust.


Then I go onto momentum analysis:
If we watch the curves, it seems clear that Obama is da man of the moment. He's gaining ground not only on Hillary but also on the other 2 relevant candidates.
As for the Reps, while Romney is still not the favorite, a win in IO, NH will keep him going for a good while. So, if we move into the "second derivative", we notice that the important shift is McCain gaining ground on Giuliani.


Nimh, man, I know I've said this already, but you should work in polling! It's a miserable waste of talent, honestly.
0 Replies
 
nimh
 
  1  
Reply Wed 12 Dec, 2007 10:05 pm
fbaezer wrote:
As for the Reps, while Romney is still not the favorite, a win in IO, NH will keep him going for a good while.

Yes, but his problem is Iowa.. Huckabee has been rising there for months with striking consistency, and has positively gallupped upward the last month or so. He's now in first place, and rapidly increasing his lead.

For a long time, Huckabee's extra support appeared to come overwhelmingly from the undecideds, with none of the other candidates losing great numbers of their own voters, Romney being especially stable. But in the last half a month, both Romney and Giuliani have started to hemorrage support directly to Huckabee as well - and do so in chunks.

Iowa graphs to come tomorrow...

fbaezer wrote:
Nimh, man, I know I've said this already, but you should work in polling! It's a miserable waste of talent, honestly.

Well, thank you Embarrassed .

I am very much just an amateur tho, a hobbyist, a dilettant. So on that count, I'll sure appreciate more of your feedback as we go along the next year here..

And I love my job, but if ever something happens (you never know in the NGO world), I'll consider the alternative career.. :wink: . It would be tricky I'm sure tho, with an education thats been exlusively alpha/gamma, nothing beta in it..
0 Replies
 
nimh
 
  1  
Reply Wed 12 Dec, 2007 10:08 pm
fbaezer wrote:
A HUGE problem of condensing polls is that you tend to combine apples with pears; that is, polls based on different sampling methods and with different type of field work. A few biased polls will ruin the whole "sample".

I know... I know. It's a very imperfect compromise - or concession to my own impatience, rather. There is a reason why pollster.com doesnt do either these timesections like mine or even rolling averages, it's exactly why you say. And yet you cant help wanting to keep an eye out for budding new trends or oscillations that take weeks to properly show up there, without wanting to go on any one poll.. it's tricky.

I was writing something on this very note somewhere else re Romney's lead in New Hampshire. On a TNR thread, a commenter said Romney was "in free fall" there, which is obviously not true, and the blogger, Chris Orr, replied that pollster.com actually has Romney steadily trending upward still. I replied:

Quote:
Good analysis, thank you Chris. Many posts go on just the one latest single poll, leading to excited chatter over what could well be random statistical noise, or in the case of comparisons between pollsters differences springing from varying methodologies, likely voter screens, etc. You look at the overall trendlines, that's the way to go. Enjoyed reading this. [..]

One comment though; you're right to use the pollster.com trendlines, that's a wise approach. But one thing to keep in mind is that those are extremely cautious, for justified reasons. In order to not have them respond to the one or two new polls whose diverging results could well be an outlier, or the consequence of statistical variation or different methodologies, they created the lines in such a way that it takes a fair number of new polls showing a "bend" in the trend for the trendline to be adjusted accordingly.

For example, in the case of Romney and New Hampshire, you're right that he is in no way in any "free fall". But in the eight polls that were done since Nov 25, he scored, on average, 32.6%, which is a percentage point lower than he got in the six polls that were done in the three weeks before. That could suggest his ascent may have been stopped, or even that he has already peaked. The pollster.com trendline doesnt show this yet: his trendline there still points fairly steeply upwards, exactly because it takes more than a few polls to bend or curve it. At most the line will be made a little steeper or less steep as individual results come in.

What is missing so far to substantiate that Romney is curving downward are apples-to-apples comparisons showing him flattening out or even dropping already. The difference between the two averages I mentioned could just be a function of the two samples being made up of polls by different pollsters, who might have different methodologies and likely voter screens. For example, the latest batch includes an MSNBC poll that has him at just 25%, for example, which is far below where all others have him. (Without that one, his average has merely stagnated at exactly 33.7%.)

For what it's worth, when you compare apples to apples, Romney is still going up: in the last Zogby he's up 11% compared to late September, in the last Marist poll he's down 3% compared to early November, in the last Rasmussen he's up 2% compared to early November, and the last ARG has him up 6% compared to late October. So it's not sure whether the stagnating average heralds a turnabout in his numbers - but it might.

December 12, 2007 2:11 PM
0 Replies
 
Cycloptichorn
 
  1  
Reply Thu 13 Dec, 2007 01:10 pm
CNN poll shows Edwards CLOBBERING the Republican candidates.

http://i.a.cnn.net/cnn/2007/images/12/11/tue6ampoll.pdf

http://i188.photobucket.com/albums/z152/UCLABruinKid/Election%20Polls/2007-12-CNN-National-Row.jpg

I just can't figure out why McCain isn't being pumped up more by the Republicans. He's consistently been the closest to the Dems in head to head matchups.

This poll also continues the trend of polls which show both Edwards and Obama as being more electable then Hillary!

Cycloptichorn
0 Replies
 
nimh
 
  1  
Reply Thu 13 Dec, 2007 09:05 pm
Heya Cyclo Smile

These are graphs for the Democratic primary race in Iowa, according to the same recipe as the previous ones:

http://img263.imageshack.us/img263/6352/iademsonlyminekw4.png

http://img527.imageshack.us/img527/8254/iademsbothrv4.png
0 Replies
 
Cycloptichorn
 
  1  
Reply Fri 14 Dec, 2007 12:08 pm
Awesome..

No graph, but:

Quote:

Election 2008: Florida Republican Primary
Florida: Huckabee 27% Romney 23% Giuliani 19%


http://www.rasmussenreports.com/public_content/politics/election_20082/2008_presidential_election/florida/election_2008_florida_republican_primary

Quote:
Those figures reflect a stunning change in the race since November when Rasmussen Reports polling found Giuliani on top with 27% followed by Romney at 19% and Thompson at 16%. Since then, Huckabee has gained 18 percentage points and Romney picked up four points. Giuliani is down eight, Thompson is down seven, and McCain is down four.


Cycloptichorn
0 Replies
 
nimh
 
  1  
Reply Fri 14 Dec, 2007 03:12 pm
On the Obama '08 thread:

Thomas wrote:
georgeob1 wrote:
The polls do measure something, but the thing itself is variable - and at this stage based on hypothesis, not fact. I think people react differently to hypothetical alternatives than to real ones. That is how I rationalize the discount I apply to them now.

I would probably share your skepticism if it was just the polls about the horse race. But Edwards's strong standing in general election horse race polls is consistently affirmed by polls about issues. And that, in my opinion, makes it more than a temporary snapshot.

To see what I mean, let's first ask what priority people are placing on the various issues the parties are currently debating. The general picture you get from the various independent pollsters is that the top three are the war in Iraq, the state of the economy, and healthcare. The next three, not quite as consistently, seem to be terrorism, illegal immigration, and the environment. (Two issues that the Republicans debate a lot among themselves, traditional values and taxes, barely make the top ten.)

Having determined which issues Americans care most about, let's look at the opinions they are holding about those issues.
    [url=http://www.pollingreport.com/iraq.htm][b]The War in Iraq:[/b][/url] The general opinion is [i]very[/i] much on the liberal side: by a 2:1 margin, they oppose the war and disapprove of the way it's being run. By somewhat smaller margins, they think it can't or won't be won, that it was a mistake to start it in the first place, and that the Democrats would handle it better. The war in Iraq is a winner for Democrats against Republicans. Among the Democrats, it strengthens those who took a liberal position on it like Obama, and weakens 'centrists' like Clinton who supported it and refuses to admit that was a mistake. (Edwards is somewhere in the middle.) [b][url=http://www.pollingreport.com/consumer.htm]The State of the Economy[/url] (Also see [url=http://www.pollingreport.com/consumer2.htm]Part 2[/url]):[/b] Throughout the current economic expansion, American have been lukewarm or unhappy about the state of the economy. That's consistent with CBO statistics showing that most of the GDP growth went to the top 10 percent of the income distribution, whereas the bottom 80 percent gained little in real terms (i.e., after correcting for inflation). Gallup finds that by a margin of 2:1, they rate the state of the economy as "only fair" or "poor" rather than "Excellent or good". Asked to identify their major concern about economy, Americans name the cost of healthcare, jobs going overseas, and lack of good-paying jobs. Generally, all this tends to give the Democrats an edge over Republicans. The issue of outsourcing, in particular, gives populists like Edwards and Tankredo an edge over the free traders in both parties, whose position has weakened consistently and sadly during the first seven years of Bush. [b][url=http://www.pollingreport.com/health.htm]Healthcare[/url] (Also see [url=http://www.pollingreport.com/health2.htm]part 2[/url] and [url=http://www.pollingreport.com/health3.htm]part 3[/url]):[/b] Americans are unhappy with the current state of the health care system. They name access and costs as the biggest problems. Although they reject, by narrow margins, the replacement of the current system by a Canadian-style single payer system, Americans believe by a 2:1 margin that the federal government make sure everyone is covered. This, again, should in the hands of Democrats, especially John Edwards, the Democrat who wrote the blueprint of the current generation of Democratic health care plans.
I don't have time to go through the other issues, but the general point is clear: For quite some time now, the position of Americans now has policy priorities that Democrats have been traditionally occupying, bypassing the Republicans' pet issues. Moreover, Americans now hold opinions on most issues that don't just favor Democrats over Republicans, but even favors liberal Democrats over centrist Democrats.

Surprisingly Edwards now voices the opinions of the median American, who has moved far to the left over the last seven years. He is beating the hell out of the Republican candidates, who are out of touch with Americans' new priorities. By contrast, Clinton and Obama, stuck in "the timidity of hope" (Paul Krugman), have a harder time doing the same. (Just as an aside, that's another reason I'm unconvinced by Sozobe's musings about electability.) John Edwards's lead in general election match-ups is much more than a shaky early-poll success.
0 Replies
 
nimh
 
  1  
Reply Fri 14 Dec, 2007 04:11 pm
This is interesting:

Mark Penn, Hillary's chief strategist, analyses the recent poll numbers:

Quote:
What's happening in the Democratic primary for president? A lot less than the headlines would suggest.

To an amateur like me, a lot of what he says seems to make sense.

But then you have to read Jay Cost's unspinning of Penn's article.

He takes on Penn's analysis on a point-for-point basis, and suddenly it all looks very different again.

The Internet, such a delightful thing!
0 Replies
 
ossobuco
 
  1  
Reply Fri 14 Dec, 2007 04:17 pm
ossobuco wrote:
Would you mind explaining what you are talking about?


Aaack, I see I sounded rude, when I was really just confused - re what centralization has to do with, say, the fun/usefulness of polling, graphs, etc. - and said so too abruptly, sorry.
0 Replies
 
Thomas
 
  1  
Reply Fri 14 Dec, 2007 04:48 pm
bookmark
0 Replies
 
nimh
 
  1  
Reply Fri 14 Dec, 2007 04:49 pm
No problem, I was being pretty elliptic myself. Centralisation was as in, bringing it all together in one thread instead of scattered over the place.
0 Replies
 
ossobuco
 
  1  
Reply Fri 14 Dec, 2007 04:50 pm
Ah, sooo, now I get it.
0 Replies
 
nimh
 
  1  
Reply Fri 14 Dec, 2007 04:54 pm
A first analysis:

Senate Sensibilities: A Second Democratic Year in '08?
An Overview by Larry J. Sabato


Summary: At this early stage of the race, and very much depending on who will be the Presidential nominees, the Democrats seem to be positioned to win anything between two and nine Senate seats in '08, which would increase their majority to 53-60 seats.
0 Replies
 
 

Related Topics

 
  1. Forums
  2. » Get yer polls, bets, numbers & pretty graphs! Elections 2008
Copyright © 2020 MadLab, LLC :: Terms of Service :: Privacy Policy :: Page generated in 0.03 seconds on 07/11/2020 at 12:06:08