Saturday, June 07, 2008

"All Politics is Local" Spoke Tip O'Neill

WE'VE MOVED! Democratic Convention Watch is now at http://www.DemocraticConventionWatch.com

We can look at elections in terms of national popular votes, or Electoral College votes. We can look at which demographic voted in which way. But no matter how you slice it, ALL elections come down to what each individual does when he/she is in the voting booth. While some people vote against their own self-interest, in the end a vote for most people is “Which candidate do I believe in? Which will do the best by me and my family and my beliefs?”

I am pleased and humbled that the DCW guys are allowing me to come to you as part of this most auspicious of blogs. What will forever be THE historic source for Superdelegate tracking. My plan is to write mainly about the down-ticket races. I will be putting out “Sunday with the Senators” to keep you apprised as to what is happening with the Senate races. First, tomorrow, framework for which races are in play, why they are in play, and where you can go to help with the election of the progressive Democratic candidates (and why you should….). In addition, I will be writing on the competitive House races. For both sets of races, I’ll have poll numbers, and will cite all relevant sources. So let’s talk a little about the polls.

A lot of us read the polls to see who is ahead in a given race. But polls also report what matters to us as voters. So, we know that the major current American concerns are the economy and the war in Iraq. That might change between now and November, and it may not. By knowing what issues are most important to “We the People” allows candidates to frame their campaigns in ways that respond to our concerns.

But polls can also fool you if you only look at part of the information. Currently, the polls "say" that Obama and McCain are equally matched in the numbers. But that's not exactly correct, and in certain ways, it doesn't matter. For better or worse, the popular vote is not the metric used to determine who wins the Presidential election. (If you don't believe me, ask Al Gore.) So, a "National Poll" tell us nothing, since we don't know how the states will go based on the Electoral College. DailyKos has a great breakdown of the regional numbers -- and basically, these numbers show that McCain is holding the south, and nothing else. Remember, the devil is in the details.

Let's consider what the polls tell us, and what they don't tell us. There are a number of great web sites that compare the polls and point out which are polling close to how individual primaries turned out, and some that missed the mark by a mile (or, in polling parlance, "20 points").

Polls are normally conducted by telephone. Read: land line telephone. Therefore, people who have cell-only service are undercounted. In addition, if you're one of the people who doesn't answer your phone unless you know the name on caller ID, which is a whole bunch of people, you don't get polled. In addition, some polls call only registered voters, or only voters who have voted in the last several elections, or just Americans, whether they are registered to vote or not.
There is also the issue of something called "push" -- which relates to how committed the poll is to "pushing" the respondent to get to a decision. (And is different from a “push poll” which is an ugly political tactic.) Do they accept "undecided" as an answer? Do they allow for the inclusion of 3rd party names and write-ins? Do they include leaners? Do they ascertain whether you might vote, you will probably vote, or whether only death would keep you from voting.

We'll be seeing a lot of polling numbers over the next several months. Since so many more polls are automated than ever before, that is computer calls, they are cheaper to do. So more people will do them. And remember that in almost all cases someone PAYS to get a pollster to run a poll. So, when you look at where a poll is coming from, it pays to keep in mind who is a Republican pollster, who is Democratic, who is providing internals to the campaigns, and who is being paid by which special interest group. While it should not affect the outcome in that people paying for polls WANT the most accurate data possible, it does affect the way questions are written, who is polled, how many people are polled, which cross-tabs are broken out, and therefore, indirectly, the outcome.

So starting tomorrow, “Sunday with the Senators” will set out the polling numbers for the competitive races; to the extent they are available. And while that information is important, we’ll also look at local issues, and context. Is the Presidential race important? More than most in American history. But the down-ticket races matter too.

3 comments:

tmess2 said...

While people with cell-phones are undercounted, most polls do "adjust" the data to make sure that all demographic groups are appropriately represented. While sometimes these adjustments can hide trends that should be recognized (i.e. changes in party identification, union membership), they generally assure the accuracy of the poll result.

At the present time, I am not sure that cell-phone use represents a politically significant demographic. I would love, however, for some pollsters doing exit polls this fall to include a question on cell phone usage to see if it is a significant category.

ac62 said...

tmess2

And yet, if cell phone users include a much larger percentage of the more "high tech savvy" citizens (and I think they do) then their omission from political polls could be quite significant in terms of excluding those who are a part of the new digital age voting bloc.

I very much agree with your point about the value of exit poll results which look at the otherwise hidden voting trends of this newer demographic.

Vicki said...

Thanks, I'm looking forward to an intelligent analysis of polls, which is what this sounds like it will be (or try to be). I have a long history with statistics in a variety of fields, and it's easy to default back to what Mark Twain said - and more difficult to take those numbers and try to tease some truth out of them.