Unlikely Voter

Four polls came out taken on the eve of the passage of the Patient Protection and Affordable Care Act (PPACA). Quinnipiac, Bloomberg, CBS, and CNN all produced similar but not quite the same numbers. How do we average them in a way that makes sense?

Gary Langer at ABC has more details on that bad poll by Harris.

It’s worse than I thought.

“Lies, damn lies, and statistics.” “You can use a poll to prove anything.” We all hear the lines like these, which reflect the popular view of statistical analysis and opinion polling. Nobody believes it. It’s all made up, says the conventional wisdom, or at least doctored to a degree that it can’t be trusted.

It’s not actually so bad at all, but it’s hard to make that case when people like the Daily Beast’s John Avlon undermine the field with examples like these.

I’m sure Lord Pollington is horrified at such informal language being used on this site, but it’s accurate. Right now my efforts are focused on making the national projections possible, and pretty in fact.

Here’s a peek at the basic map I will use, based on those electoral maps published by the National Atlas. What election it was originally made for is obvious.

In the possible rematch in Michigan’s 7th Congressional District, there seems to be no independent polling yet. By request I looked for polls on this race, but for the budding 2010 matchup of Tim Walberg and Mark Schauer all I find is one internal poll.

Walberg’s is good for him showing 46 R/37 D/5.6 MoE for a 78% win rate for the Republican. I expect real polling only after the primary for a race like this.

Contact me and I’ll give your race a look. Primary, general, Democrat, Republican, it doesn’t matter.

Unlikely Voter is still in a bit of a holding pattern until the 2010 general election begins in earnest, so I’d love to hear suggestions for what I should be looking at right now.

On the heels of a Rasmussen poll suggesting Democrats should worry about Senator Barbara Boxer’s re-election chances, comes worse news from Field for the three term incumbent.

The raw numbers: Campbell 44/Boxer 43, Fiorina 44/Boxer 45, DeVore 41/Boxer 45, 3.7% Margin of Error. My model’s win percentages: DeVore 29%, Fiorina 44%, and Campbell 55%.

There are number of tools that go into the production of Unlikely Voter now and going forward. In the interests of allowing skeptical observers the ability to try to reproduce my work, I will share them with you. Not that I’m going to give away the store, but I’m ready to give strong and clear hints.

The California Republican primary race to determine Barbara Boxer’s challenger is getting heated. And while all of us in the state have our biases and preferences, here’s what my cold, hard math says about Rasmussen’s latest poll.

While playing around with primary polling lets me practice how I look at an individual poll, it does nothing for my work on aggregating polls and creating national projections.

So if I can get the historical election week polling data I expect to start doing some backtesting over the spring and early summer. I’ll come up with models and test them by “projecting” past House elections, Senate elections, and Presidential elections.

Whichever model works best I will use to project the 2010 midterms. Additionally I’ll learn which pollsters are better than others, which will only improve my projections.

Today I use a poll of a primary race between two Democrats to illustrate my initial model of a horserace election poll. Until the general election season begins in earnest, the primaries are all we have and I intend to use them as a warmup.

I wanted to see how my single-poll model looks when I apply it to the history of a race, and selected the Florida Republican Senate primary between former Speaker Marco Rubio and Governor Charlie Crist.

Rasmussen Reports polled all three likely California Senate matchups of Republicans against Senator Barbara Boxer, but only one candidate’s electability is routinely called into question: Assemblyman Chuck DeVore’s. Just how unelectable is he, according to Rasmussen? This poll serves as a fine example of how my individual poll analysis is going.