« BACK  |  PRINT

RS

MEMBER DIARY

Are the polls accurate or not?

We are inundated daily with a brain numbing soul wrenching miasma of polling information. With that information comes mountains of punditry that is based entirely on those polls. Even serious and talented people can’t seem to get past those polls even while they point out methodological problems with same.

I enjoy Sean Trende’s analysis as much as anyone’s because he generally lays a strong foundation before building an argument. He is currently on Day 2 of a three day look at the current race for president at Real Clear Politics.

Yesterday he laid out why Barack Obama may win and today he looks at why Mitt Romney may win. 98% of what he says makes sense and is even-handed but I was struck by an inconsistency in today’s piece.

Overall Trende feels that Obama is a slight favorite and I have no quibble with that assessment. What I do have is a problem with is a specific thread of logic behind that opinion.

Below I have pulled several quotes from today’s article. See if you can spot the problem. Here Trende is discussing the history of how challengers perform on election day relative to their poll numbers two weeks after the part conventions.

…no incumbent party has ever held on to the White House while leading by fewer than four points two weeks after its convention; no incumbent president has ever won re-election while leading by fewer than five points….. winning under these circumstances would be unprecedented…

…on average, {the polls}  moved 3.7 points toward the challenger…

….with the exception of 1992 — ………every contest with an incumbent has broken at least three points toward the challenging party from this point in the race through Election Day.

…..George {W} Bush led by 6.8 points as opposed to Barack Obama’s current 2.9 percent. His bounce peaked at 50.4 percent, as opposed to Obama’s 49 percent. If Obama continues to run behind Bush on either metric by similar margins through Election Day, he loses…..

(my emphasis)

Next Trende discusses the method of polling being used.

Much is made of the fact that Romney has never led in the RCP Average. But remember, throughout this cycle, most of the polls were using registered, rather than likely-voter, screens. …….Romney’s poll numbers would have been two-to-three points higher (given the average movement we saw when pollsters activated their likely-voter screens), and Obama’s lead fluctuated between 0.2 points and 3.8 points.

Then he looks at Obama’s job approval numbers in comparison to George W. Bush’s in 2004.

In early September, Bush’s average job approval was 51.4 percent, almost two points higher than Obama’s is today….the strong tendency is that presidents run a few points behind their job approval numbers with the electorate. Bush’s job approval in the RCP Average on Election Day was 49.8 percent, but his job approval with the actual electorate according to the exit polls was 53 percent…..

…. Obama probably has a bit of a higher hurdle to surmount than Bush had. While Republicans typically run ahead of polls of registered voters and adults, Democrats typically run behind them. Because job approval polling contains a mixture of these types of polls, Obama’s job approval with the actual electorate is probably a touch below his average right now….

And finally….

So if the election were held today, President Obama would probably win comfortably.

What am I missing? The only reason anyone thinks Obama is leading today are the polls. If this were 1920 and no polls existed I suspect the presumption would be that the race was close but the challenger has the upper hand. Instead we have a daily inundation of polls that tell us (in aggregate and on their face) that Barack Obama is ahead.

As Trende points out most of the polls to this point have used Registered Voters instead of Likely Voters, so based on history all of those polls have been systematically underestimating Mitt Romney’s share. I really can’t fathom why RV polls are used at all. Its analogous to a football team pretending their opponent has 3 points less than they actually have for the entire game and only facing reality when the two-minute warning is given.

So why does Trende conclude that Obama would win “comfortably” if the election were today?

Even if we take it as absolute fact that Obama is ahead by roughly three points, Trende has shown that, A) No incumbent party (going back to 1968) has ever won with such a narrow lead at this point and, B) on average the challenger makes up enough ground to eclipse a 3% lead.

So on what basis does Trende decide that Obama is a slight favorite?

A shiny red car with a bad engine is still a car with a bad engine. Either the polls are correct or they are not correct. They can’t be both accepted as inaccurate and cited as sole proof that Obama is ahead.

As of today, September 20, this race appears to be a deadlock with the most likely movement to be toward Mitt Romney between now and election day.

(To stay within fair use I have obviously cut out much of Trende’s commentary but I don’t think I’ve done so in way that loses context. By all means check it out, read the whole article and judge for yourself.)

Get Alerts