re: #1367 iceweasel

Lots of places. I’m not quoting anyone in particular. This is also based on my own observations of Rasmussen. If you’re into polls, this is how to evaluate them.

If people are really interested, I could dig up critiques of individual Rasmussen polls that will explain how that individual poll is flawed, but as I said, evaluation always has to be done on an individual basis.

There are some things that you need to keep in mind about polling.

1. Polling is not an exact science. Because the polls are written and conducted by humans, no poll will ever be perfect. It is just not possible.

While you can gleen some interesting and sometimes useful information from polling, they are all inherently flawed. All of them.

~

Here is the most import thing to help in understanding how polling companies work.

They are in the business of making money, not in influencing elections. This is true for all of the big houses, whether they are run by people who tend to vote Republican or tend to vote Democrat. The smaller houses can be a little quirky on this, but that is usually attributable to their small size, lack of experience, One-off errors, etc. All of these polling houses are betting their reputations and their future incomes on the accuracy and usability of their results. Faked results, or results that are grossly biased can destroy a polling house.

They won’t knowingly do it.

But they are run by humans, so sometimes they screw up.

2. Polling is a very problematic predictor. A poll is a snapshot in time. It will reveal how a number of respondents answered a set of questions for a given time frame. They can be used to get a good guess as to how people will respond to the same set of questions on another day, or on how they will respond to set stimuli in the future, but each hour that passes from the close of the poll to the target event degrades the prediction validity. In simple terms, shit happens. What people may have thought in one moment may be affected or changed by something that they hear or experience in the next.

3. When it comes to elections, one of the most reliable populations is long term voters. Voters that vote reliably in the small off season elections are a solid bet to vote in the next big election.

New voters, especially young voters, are a terrible bet when it comes to banking on them voting in the next election. They have a long history with understandable reasons for being unreliable.

Of the two, long term voters will give you the best predictor of what will happen in a particular election.

Occasionally, you will have an election like the last one with Obama where the young and new will show up to the polls in larger numbers than they have in the past. These are outlier elections. They are not good for making changes to methodology. What happened in this last election is VERY unlikely to happen again in the next. You will see proof of this in the next congressional races. Barring some extraordinary event between now and the next congressional election, the turn out will be very low for the young and for those who voted for the first time in the last presidential election.

At this stage, in the run up to the next congressional election, any polling house that includes the new voters and the young voters in their results will have crap for numbers. That is a large part of the reason why they will make sure to let people know the population group that they are using, whether it is likely voters are registered voters, or something else. In their own community, they will know what the populations mean to a study. They will judge each others results with the polled population in mind.

Bottom line.

For pollsters, the accuracy of their poll is more important to them then whether or not a particular party or politician wins a given election.