Monday, December 19, 2011

Polls

One of the better undergrad political science classes I took at MSU focused on polling. There a lot of polls right now with a lot of different samples, in different areas, by different companies, with different results. Polls are educated guesses.

The first thing with a poll to look at is as follows. Is it scientific? If not, it's worthless. The next. Margin of error. Anything over 5% is worthless. Throw it out. If it's got a 3% or 4% margin of error, it's supposed to be considered good. That means if a poll says Gingrich 28, Romney 24, and real numbers are 32-20 or 24-28, that's good. What I mean by real numbers, is also the results from that day. Not January 3rd. If there's movement in that last week, a poll may look way off. It may actually be right - at that time.

Polls when done right, are a snapshot in time of a representative sample of the population covered. A good poll from December 19th, is a sample of the opinions of the population from December 19th. Between now and then, people may change their minds, or the undecideds may all move in one direction, or not. Generally, the larger the race, the better the polls. National is usually closer than statewide, which is closer than congressional. Congressional downward tends to often be quite inaccurate - to the point where state rep/senate races oftentimes don't even use polling. It can be way off. Not always. I was quite skeptical of some 2010 polling in some districts. To the person who told me that we were leading in some of those districts - you called it. I needed to see it to believe it, and I saw it in the actual results. The worst polling however is with issues. Those are almost never right. I've seen issues polled 60-40 in favor end up losing election time on a common basis. I chalk that up often to people lying to pollsters.

The biggest unfortunate aspect with public polls, and we ALL are guilty to some degree here, is with the bandwagon effect. We rely on these polls to shape our voting tactics. That's especially when we want to stop a candidate more than support one. People want to vote for a winner, or they want to stop a frontrunner. I look a whole bunch of polls and look at the momentum, trying to shift through the outliers.

Luckily in Michigan, we AREN'T first. We get to see real results besides opinion polls. Iowa. New Hampshire. South Carolina. Florida. Nevada. Colorado. They will vet candidates and make their decisions. Those are going to shift the polls here, as electability is more apt to be determined.

So when you see a poll, appreciate it for what it is if done right. A snapshot in time of a representative sample of the population covered. From the day the question was asked. That may or may not be the final numbers.

1 comment:

Communications guru said...

I can answer the third question: there is no such thing. They are the extremist fringe of a Republican Party that has taken a sharp right turn. Party’s stand for something, what teabaggers stand for can’t stand up to even the slightest scrutiny.

Regardless, Sen. Stabenow will easily beat Twitter Pete Hoekstra in November.