The truth about Obama’s approval rating

Rasmussen’s Daily Presidential Tracking Poll reports today:

Overall, 42% of voters say they at least somewhat approve of the president’s performance. This matches the lowest approval rating yet measured for President Obama. Fifty-six percent (56%) now disapprove.

This will assuredly set off the usual shouts in the conservative Web about how Obama’s approval rating is “plummeting,” it’s in “free fall”—a “free fall” that has been going on for months yet somehow he’s still falling and hasn’t hit the ground yet.

But in the next paragraph Rasmussen has this clarification:

While the president’s daily rating is at the low end of the range today, a Month-by-Month review of the president’s numbers shows amazing stability. On a full-month basis, the president has been at 46% or 47% approval for nine straight months.

How about that? His approval rating hasn’t been in free fall, it’s not falling at all, but has been quite steady. And now think about all those articles in the conservative Web over the last nine or so months acting as though Obama were in the act of crashing to the earth.

- end of initial entry -

David B. writes:

A year or so ago, I wrote that Obama’s approval rating is unlikely to fall below 42-45% because his nonwhite supporters will never leave him. This is a factor I have never seen media and web conservatives mention.

This is similar to the Tea Partier’s advertising their black members.

LA replies:

42 percent! Your time has come.

Alexis Zarkov writes:

Rasmussen is essentially correct about Obama holding fairly steady in 2010. However as I pointed out in my prior email, starting in May 2010 Obama’s approval rating has been declining at about 1 point per month. Not a lot, but a trend. It’s best to use an aggregate of polls to reduce the fluctuations you get with a single pollster like Rasmussen.

The different pollsters have different kinds of errors. One kind of error is known as the “House Effect.” This is a systematic error or a statistical bias. Rasmussen has a large House Effect for the generic ballot, but not polling for Senate elections, see here. But Nate Silver (a liberal, but fair) says the opposite. Bear in mind that Rasmussen polls likely voters and not registered voters. Since Republicans are more likely to vote than Democrats, this will cause Rasmussen to be systematically different than (say) Gallup. Rasmussen get attacked regularly by the left as a hack for the Republicans, but these attacks are both ignorant and unfair. Nate Silver does not do this and I think he’s fair, but makes some mistakes.

Another kind of error is called “Pollster Introduced Error,” or PIE. This is a random as opposed to a systematic error. An example will help clarify the difference. Your bathroom scale has both a bias and a random error. A one-pound bias would mean an additional pound is always added to the true weight. But if you were to put a fixed weight (like a bowling ball) on the scale every day you would see a small random variation.

A pollster with a small PIE is a good pollster, and provides information. Nate Silver at FiveThirtyEight ranks the pollsters according to their PIE here. Note Rasmussen ranks very high, much better than Gallup. Zogby (interactive) ranks nearly dead last, and his polls provide little information. Ignore him. Thus a good pollster can have a big House Effect.

For all the above reasons I like to look at the Pollster.com graph and make adjustments. If you think all this is a confused mess, it is. We have different pollsters using different methodologies who are in turned evaluated by people using different methodologies in their evaluations. Welcome to the real world where things get messy and you have to sort it out for yourself. I also have my own methods where I can get the win probability, something you generally don’t get from the major pollsters.

Posted by Lawrence Auster at September 04, 2010 06:36 PM | Send