Don’t Panic: The NYT Poll is an Outlier

July 19th, 2012



The NYT/CBS poll released today has left many Democrats nervous and pundits wondering if this is the turning of the tide. But neither alarm nor even real interest is warranted: this poll is an outlier and should be viewed with real skepticism.

This kind of thing happens all the time. Polls get weird samples and the results look too strange to be true. Last month, Bloomberg had one showing Obama with a 13-point lead. That was wrong, and this Times poll is almost certainly wrong as well.

Reasons for doubt: Horserace

The NYT/CBS poll has Romney winning the head-to-head by 2 points, within the margin of error. This is unsurprising—no one expects a blowout this year. Real Clear Politics, which takes the average of publicly available media polls and includes this newest survey, finds the race closer than 2 points, but with President Obama winning 46.3% to 44.9%.

Pew’s most recent poll has Obama leading by 7, the same lead he had in May and up 3 points from his June lead of 4. Gallup also has President Obama leading by 2. These figures are all close enough to paint the picture of a tight race, which is what the NYT/CBS survey finds.

However, when we dig into the crosstabs, discrepancies arise. For example, the NYT/CBS poll has Romney winning Independents by 12. Yet Pew’s poll puts this category within the margin error, showing the President up by 1 with Independents. Indeed, Pew’s data over time show a tight race for Independents, not a double-digit blow-out. The same is true of Gallup[1], which has Romney leading by 2 with Independents now, but shows a tight race for the past 2 months.

One final note on the horserace crosstabs: the NYT/CBS poll has President Obama winning women by 5 and Romney winning men by 8. This is not surprising. A gender gap has persisted in national elections for decades. But in Pew’s recent polling, President Obama is leading with women by 13 and Gallup has President Obama up by 10. The 5-point margin for the President in the NYT/CBS survey appears small by contrast.

Reasons for doubt: Favorables

The biggest reason to doubt the findings of the NYT/CBS poll is the favorability numbers for the candidates.

Favorability/Unfavorability Ratings, NYT/CBS

Favorable Unfavorable
Obama 36% 48%
Romney 32% 36%


The 36% favorability for President Obama just is too small to be believed.  Indeed, despite the tepid economic recovery, the President has continued to enjoy high personal favorability ratings, generally above 50% (leading the Romney campaign to focus on Obama as simply over his head rather than an evildoer). Consider this:

  • McClatchy/Marist, July 9-11: Obama favorability 49%-46%
  • CNN/ORC June 28-July 1: Obama favorability 55%-44%
  • Fox June 3-5: Obama favorability 54%-42%

Yes, even that liberal bastion Fox finds President Obama to be widely liked. Real Clear Politics—even including the NYT/CBS results— put President Obama’s favorability average at 49% to 46%. And it’s not just the Obama favorables that are low. So are Romney’s:

  • McClatchy/Marist, July 9-11: Romney favorability 46%-42%
  • CNN/ORC June 28-July 1: Romney favorability 48%-42%
  • Fox June 3-5: Romney favorability 46%-42%

Real Clear Politics—even including the NYT/CBS results— put Romney’s favorability average at a tie, 43% to 43%.

Reasons for doubt: National Security

Although it’s a relatively minor data point, the NYT/CBS findings on who would do a better job handling terrorism and security also stand out. Their poll gives President Obama the edge by 1, 44%-43%. Yet, despite the historic Democratic Party brand on these issues, President Obama generally receives high marks on national security, terrorism, and defense. Pew’s recent poll finds President Obama with a 12 point advantage on who would do a better job defending against terrorist attacks and an 8-point lead on making wise decisions on foreign policy.

In short, the only real takeaway from this poll is that the race is a dead heat. And we knew that already.

[1] For Gallup crosstabs, click on the Demographics tab.