Thursday, November 1, 2012

Polls, Predictions and Probabilities...Oh My!

Voting day is nearly here and, not surprisingly, the chatter and speculation as to who will win has intensified each day we get closer to November 6th. Also not surprisingly, each side has expressed confidence that they will be successful.

So where are we? Can anyone say with confidence that they will deliver the goods on Tuesday?

Certainly no one can say with any certainty that they will win. The simple fact is that there is an awful lot of noise in the air relating to polls, predictions and probabilities (...oh my!). 

Let's consider a few points.

1. In a close race, it is imperative that a candidate project confidence and predict victory. And this is a close race. We are not in a race where it is clear who will win and the poor other guy / girl is left to say things like "we'll have to wait for the voters to decide" - something which, if you read between the lines, is more like "please God, let it end."

No, we are in a race in which the polling noise, economic environment and sense of political division all open the door to multiple predictions from state to state. As of this morning, Real Clear Politics showed 11 "toss-up" states - in effect the states that will decide the Presidency.

In such an environment, a candidate needs to make sure their base is sufficiently energized to get out and vote. And, for those few (but potentially decisive) who are waffling or undecided they need to convey a sense of momentum. You will see this from both tickets and their supporters, so see it for what it is.

2. At this stage in the campaign, it sometimes seems like there are as many polls as there are voters. With so many polls coming out, it is important to consider the source, the sample, the methodology and the history of the polling firm.

As noted in earlier posts, Nate Silver's Five Thirty Eight blog at the New York Times is a must read. This is especially true if you have an interest in better understanding polling and how one might view the varying results from the different polls out there.

The model used by Mr. Silver doesn't just average out polls, it weights them by taking into account things like methodology and sample, among others. Knowing whether a polling house is more traditionally Republican or Democratic leaning is a variable. Phone versus on-line samples is a variable.  Consistency in terms of results and accuracy is a variable.

All to say, a poll can be commissioned to suggest many things and its results can be read many ways. Also consider the fact that the real polls which actually drive a candidate's behaviour are the ones we never see. Polling is instructive and indicative, no more. 

3. Building on the subject of polls we have another topic that has come up with more frequency over the past week - particularly in relation to the Silver blog. It is the question of probabilities

Sites like Silver's, Real Clear Politics and others are playing in the world of probabilities. The problem is that people tend to gravitate around the numbers they provide at the expense of the story.

If you only took a cursory glance at Five Thirty Eight you would likely look to the right at the forecasts for vote-share and electoral college votes. If you did so, you would miss a regular and frank explanation as to the rationale behind those numbers.

This rationale does not suggest certainty, it suggests probability. Odds, if you will. For this reason, I find aggregators such as this one helpful in cutting through the polling noise and allowing one to make sense of the landscape.

And yet, blogs like Silver's have come under criticism from the right as being biased towards the Democrats. 

The thing is, probabilities which factor in a host of variables are more likely to be accurate than any one poll. This does not mean that the outcome they suggest as probable is a certaintly; it is just to say it is more probable based on the information available.

My take? The criticism of Silver is a case of shoot the messenger. I would wager that the internal polling of the right is aligned with his probable outcomes.

Given the points above about confidence and projecting momentum, if Silver is correct (which I think he is) and influential on the media narrative (which he certainly is), it is natural that the Romney campaign / supporters would come out against him.


Pity the voter who has yet to make up their mind. Between candidate predictions, a plethora of polls and talk about probabilities there is a lot of noise out there. How one hears it can influence how one votes.

Factor in ads and newspaper editorials and it gets noisier. And tomorrow we have the final jobs report before election day.

All of it makes one wonder: who is more eager for November 6th to come and go - the voter or the candidates?  

No comments:

Post a Comment

Have a comment?

Canadian Blogosphere