Who Do You Think’s Gonna Win In November...?

[BluePhantom]

Yup, could be. I'm not following the trends in the swing state polls, so I don't know what to expect. Seems to me Romney's biggest cause for hope is the fact that he was relatively close after that ridiculous nomination circus. I'd guess the Dems weren't expecting that.

.

Right now the number of state polls are frustratingly few and so in some states like Iowa, Nevada, Colorado, etc...there's just a whole lot of nothing to look at. But I am tracking what there is to track on my polling analysis thread. Weekly reports for poll averages, trends, state trends, demographics, and likely voter scenarios are all broken down and posted on Wednesday mornings. You are certainly invited to have a look and contribute if you wish. My only request is to keep things as objective as possible when you post. With only one or two exceptions (ahem: NYC and nobrain) everyone has stayed pretty true to that rule and it's made it a pretty nice place to see what's going on from multiple angles.

Cool, thanks. I'm assuming that we'll get more and better data as the election nears. Isn't one problem that not every state has reliable polling data sources, even close to the election?

.

It can be. A lot of it is just the volume and the difference in polling method. The rule is that the closer in you zoom the more the average variance will increase, so national polls tend to be more accurate than state polls, which tend to be more accurate than county polls, etc. Also you have a couple firms that do a shitload of polling (Rasmussen and PPP for example) but most firms only poll what they are hired to do. So we might only get one SurveyUSA poll of say Wisconsin even though SUSA is historically one of the most (if not the most) accurate.

Then of course house biases play in. For example, Rasmussen tends to have a bit of a Republican lean by...oh a point and a half to two points. Most of that is because they do LV polls where everyone else right now is doing RV polls and RV polls have a natural bias of 4-6 points toward the Democratic candidate. PPP on the other hand is affiliated with the DNC and on average their polls have been giving an additional five points to Obama against the average of other pollsters. So it gets real complicated.

PPP for example released an RV poll of Arizona today that shows Romney +7...well if their house bias of five freebies to Obama holds steady that means the spread is really more like Romney +12 and when you add in the RV factor of around four or five points it means Arizona has no chance of going to Obama. But when you do the calculations of averages and trends you really have to take everything, ignore nothing, do the math and let it be.

On my thread I do that, but I also do things like take the RV polls and convert them to LV formats to get an estimate of who the people who are likely to actually go vote support as opposed to just "registered voters" many of whom won't bother to cast a ballot.

That's important because most people will simply look at a poll that says "Candidate A +9" and assume it means Candidate A is actually leading by nine points. Unfortunately it doesn't work that way. In polling 2+2 rarely equals 4. For example on another thread last week I did some basic math and demonstrated that the FOXNews poll on the 15th showing Obama +7 is really more accurately interpreted as a statistical tie.

That's what I attempt to do on my thread I linked you to....not use statistical trickery to show one thing or the other, but to properly apply statistics and mathematics to interpret polling results accurately.