Archive for September 27th, 2012

Blame the Pollsters too!

The right envisions the election playing out very differently.

If you’ve been reading my “2012 Polls” page, linked above and on the left side of the menu, you know that I constantly up date and report new Presidential and Senate polls as they are released.     So I’ve been swimming in polls, and will keep this up until election day.

Lately the polls are showing a rather pronounced trend towards Obama.    In the swing states that were either tied or leaning slightly, Obama has built sometimes large leads.   Taking the average of recent polls, Obama leads in all the swing states:

Ohio:   Obama + 7
Pennsylvania:  Obama + 9
Florida: Obama + 4
Iowa:  Obama + 4
Colorado:  Obama + 5
Nevada:  Obama + 5
Virginia:   Obama + 4
North Carolina:  Obama + 3

In the national polls it’s the same story:

Pew:  Obama + 8
Gallup Tracking:  Obama + 6
Bloomberg:  Obama + 6
National Journal:  Obama + 7
GWU/Battleground:  Obama + 3
Rasmussen tracking:  tie

The polls have turned dramatically towards Obama in the week since the Romney fundraiser video came out.   Whether its a short term bump or a shifting of the race to one in which Romney emerges as noncompetitive is unclear.   If I were a national pundit worried about my reputation I’d put in all sorts of caveats about how the race is not over.  My gut tells me that absent some kind of major external shock, Obama’s got this thing in the bag.

However, people on the right would argue that I’m living in a fool’s paradise, feeling secure about an election in which Obama is trailing his challenger.   Just as the media got blamed for Romney’s campaign foibles, many now blame the pollsters for creating an illusion that Obama is ahead when he’s not.    One commentator at a different blog thought this was part of an insidious campaign to demoralize Republicans.   Rush Limbaugh claims that the pollsters are in cahoots with Obama’s campaign to try to end the election early because they fear the debates.

A website called unskewedpolls.com, which associates itself with Rush Limbaugh, Laura Ingraham, Thomas Sowell and Dick Morris, says the polls have it all wrong.    In fact, Romney is actually winning this election handily.   Their average has Romney up by 7.8% as of September 20th, the date of their last update  (the chart at the top of this post).

Their argument is simple, and on its face plausible.  Pollsters should weigh party ID just as they weigh other demographic factors in calculating their results.   I admit, I thought they did that.   But I’ve learned that pollsters believe that it would warp the data tremendously and we couldn’t trust the results if party identification was treated like race or ethnicity.

One reason for that is given by Amy Fried in her blog Pollways:

Voter ID trends are fluid, often reflecting the person’s preferences at the moment for one candidate or the other.

In other words, if people are trending towards Obama, it’s likely more people will identify themselves as Democrats.  If Mitt Romney is unpopular, it hurts the Republican brand.

Yet what about Rasmussen?

Rasmussen is a solid pollster.   Yet in this and past election cycles, he’s often shown a clear partisan lean to the Republicans.   This is especially true during the campaign cycle.   Rasmussen does weigh for party identification, and if he’s using his recent August poll numbers as a guide, he has the population as being more Republican than Democratic.  That is in opposition to the trends noted above.

I don’t know how he weighs it – his methodology page simply says:   “After the surveys are completed, the raw data is processed through a weighting program to insure that the sample reflects the overall population in terms of age, race, gender, political party, and other factors.”

There are two problems with that.   First, his methodology is automated, meaning he is focused on landlines which tends to be a more Republican population anyway.   So his survey on voter ID may be skewed.   Second, his data showing more Republicans than Democrats is contrary to most other indicators out there.

But the argument people make for considering party identification isn’t without any merit.   While party identification may fluctuate, most people do not change parties.   How many Republicans or Democrats do you know that veer back and forth?   Yet that makes it even more problematic to think there are more Republicans than Democrats out there.  From exit polls, we can get a fair sense of what party identification in the US is like.

Trends from 1939 to 2009 show both stability and flux, it’s hard to imagine there are really more Republicans now given past trends

The argument by supporters of Rasmussen (and critics of most pollsters) is that 2008 was an anomaly, voters shifted to the Democrats in response to the mideast wars and the popularity of Barack Obama’s campaign.   Now Obama’s shine has faded and the 2010 off year elections make it wrong to use 2008 data to predict turn out – whether by voter identification or even demographics.

Pollsters seem to understand that, and from what I can gather most use a number of ways to get their likely voter sample.  They screen respondents and look at demographic factors from the last two Presidential elections.   It may be that the youth and minorities will be more likely to stay home in 2012, but how much?   The conventional wisdom in the GOP has been that they have an enthusiasm edge, and if Rasmussen is using that to weigh his data (having fewer blacks, youth, Hispanics, etc.) then that could explain his results.

Voter enthusiasm has been rising, especially among Democrats. The idea Obama supporters will sit this one out is unlikely.

Yet all indications are that the enthusiasm gap is shrinking, and at this time Obama is generating enthusiasm while Romney is floundering.    Quinnipiac released polls on Wednesday showing Obama up 12 in Pennsylvania, 9 in Florida and 10 in Ohio.    Politico called these numbers jaw dropping.  Public Policy Polling also showed Obama up 7 in Iowa.  It appears like there has been a real shift towards Obama.

Therein lies the real reason not to trust Rasmussen’s numbers:  Almost every pollster is showing a shift to Obama at the state and national level.     Rasmussen has stayed flat.   To me this suggests that if he is weighing for party identification and makes demographic assumptions that understate Democratic enthusiasm his methodology is getting in the way of recognizing the trend.

These are all pollsters who have been around for quite awhile and have a good track record, including Rasmussen.   Their reputation is on the line.   They’ve been doing this and refining their methodology for years.   Is it really plausible that they all are making the same systemic error, either by conspiracy or chance?   Moreover, why would that “error” suddenly appear in mid-September, after all the polls had shown a close race before hand?  The evidence suggests Rasmussen’s method is flawed – though he could change how he weighs data as the race continues if he concludes that party ID is not favoring the Republicans and that the enthusiasm gap is disappearing.

Blaming the media and trying to find ways to disregard the preponderance of polls are both symptoms of the same disease: a campaign in distress.

UPDATE:   One other point – the same trend is being seen in Senate races around the country.   Democrats have built leads in seats that looked vulnerable, and candidates like Elizabeth Warren and Tammy Baldwin (MA and WI) have gone from being underdogs to favorites.   This reinforces the idea that the election is trending Democratic.

UPDATE 2:  Nate Silver did a thorough analysis of Rasmussen after the 2010 election cycle and found empirical evidence that Rasmussen is biased towards Republicans, and did very poorly.

13 Comments