Free Markets, Free People

Daily Archives: September 17, 2012

About all those polls… (Updated)

If you listen to the podcast, you may have noticed that, over the past couple of weeks, we’ve talked a lot about polling, and why Obama is doing so well. We’re not the only ones. A lot of people are wondering why Obama is polling well when the things are so bad. One of the criticisms I’m seeing about a lot of the polls is that they skew so heavily democratic. Except for Rasmussen, almost all of the polls coming out seem to have larger numbers of Democrats than one would expect. They have been as high as a D+11% advantage in the population.

This is seen by some as proof that the pollsters are skewing the respondent population towards Democrats. I’m not impressed by the argument, because most pollsters don’t actually try and set up a likely voter model for the poll. Instead, the poll is a sample of usually between 1,000 and 1,500 randomly selected voters. The Democratic advantage in this poll, therefore, is not an artifact of the selection method, but is actually the result of what the respondents identify themselves as. If you call 1,000 people, and 380 of them say they’re Democrats, then that’s the sample.

The poll, then, reports what the respondents say. It’s not the result of selecting a particular number of Democrats or Republicans. That’s a vitally important distinction, because voter identification changes over time. The poll reports what voters say their party affiliation is, but a voter may say he’s a Democrat this week, and a Republican or Independent two weeks from now.

So, the key here, it seems to me, is to look at a set of polls from a particular pollster and see if the party affiliation varies widely from poll to poll. If it does, then there’s probably a problem with their methodology. You might see a shift in party affiliation over time, but the change between consecutive individual polls should probably be fairly small. But in general, if a pollster uses the same methodology for every poll, and is not explicitly looking to create a voter response model, then the results are probably fairly accurate, and show small movements–if they occur–to party identification from poll to poll.

What I’m hearing from a lot of conservatives this week is the idea that the polls are horribly skewed, as if there’s some industry-wide conspiracy to make Obama look good. That doesn’t seem very likely, especially since nearly every pollster uses a bipartisan polling team, i.e. one Democrat and one Republican. So, what I’m hearing from conservatives sounds like the response Democrats made in the 2004 election, when John Kerry was polling badly. Then, as now, there was this feeling that the polls were horribly wrong, and their candidate wasn’t actually losing. But the losing candidate was, in fact, losing.

So if the polls are off, then it must be the result of either a gross, industry-wide incompetence that is causing them all to use a faulty methodology, or a gross, industry-wide conspiracy–between both Republican and Democrat pollsters–to push a pro-Obama narrative. The alternative is that the polls aren’t off, and within a 3% or so margin of error, are reporting accurately what the electorate is saying. The latter seems to me to be far more likely.

Now, as to why so many voters are identifying as Democrats, I don’t have a clue. But consider this: pretty much everyone knows Bill Clinton is smarmy liar, and if he could run for a 3rd term…he’d win.

Also, consider that everyone remembers the Bill Clinton presidency as a time of economic growth and balanced budgets. They remember the end of Bush’s two terms as a time of complete economic collapse. The underlying reasons don’t matter, because most voters neither understand nor care. It may be that voters simply trust Obama more on the economy than they do Romney, because they fear a return to economic collapse. Maybe they think Obama has done as well as could be done. But simply dismissing that with a "the polls can’t be right" explanation is just whistling past the graveyard.

UPDATE: More here, including this graphic.

A21mt5KCYAAWTIb

Now, let’s split this out and look at correlation:

polls

That’s a pretty weak correlation. Look at the blue diamonds for the Obama lead. What is that, a bell curve? Seriously?

No, unless the poll makes a specific effort to model a voter turnout, and specifically samples for a given percentage of R-D-I, then the poll is just telling you what the respondents are telling the pollsters. They may tell them something different next week or next month, but the R-D-I sample is simply a result of respondent self-identification.

~
Dale Franks
Google+ Profile
Twitter Feed

Ever have one of those days?

Well I’m having one of those days.  Nothing horrific.  Nothing tragic.  Nothing to worry about.  Just one of those days.

So, other than this, I’ll likely not be saying much more today.

Well, except a reminder to join us on Facebook for QandO plus (which I also haven’t had time for today).

Later.

~McQ

Twitter: McQandO

Facebook: QandO

Economic Statistics for 17 Sep 12

Today’s only economic release is the Empire State manufacturing index, which fell sharply to -10.41 vice -5.85 in August. New orders plunged to -14.03 down 12 points since June. Unfilled orders have been falling for months now, and stand at -14.89. On the other hand, the 6-month expectation for new orders is up to 17.02, well into positive territory. Whether those expectations will be met remains to be seen. Meanwhile, the district’s manufacturing sector continues to bump along the bottom.

~
Dale Franks
Google+ Profile
Twitter Feed