Numbers point to narrative driven data manipulation, not "random variance"

In the past two days, I have taken on polling with severely right-skewed samples showing that the president has lost significant ground, with the media driving the narrative that he did so thanks to Americans' anger at rising prices at the pump. Even though Nate Silver, usually an objective statistician, felt compelled to defend his employer by claiming "random variance", the New York Times/CBS News sample was so bad that their polling showed a 41 point drop in public support for contraceptive coverage in employer sponsored health insurance. But then, you would get that if your sample has triple the share of wingnuts as the broader population.

Since then several other polls have been released that confirm my contention that the ABC and NYT polls were based on bogus samples. When some of the other recently released polls are compared to the ABC and NYT polls, this is what it looks like in terms of the president's approval rating.

Here's a short description of these polls.

A Reuters/Ipsos poll (sample: D +12) has President Obama's approval rating rising to 50%, as Americans are shown to be more upbeat about the economy. A Bloomberg poll (sample: D +4) agrees on the upbeat economy and the a rising approval rating for the president (at 48% in the poll), as it also decimates the media narrative about the president's demise tracking to rising gas prices, showing that less than a quarter blame Obama for the pain at the pump, and two-thirds place more responsibility on oil companies and the middle east. A Pew Research Center poll (sample: D +7) also sees the president's approval rising to 50%, while seeing him widen his lead over Mitt Romney to double digits. An NBC News/Wall Street Journal poll (sample: D +7) earlier this month also held the president's approval at 50%, with the president handily defeating all of his potential GOP challengers. I will remind the reader that except for the Reuters poll, all of these samples under-represent the registration advantage for Democrats according to the latest data (D +12).

In fact, gas prices were such a weak narrative that after their own "news" channels peddled the "gas prices put Obama's ratings in crisis", NBC's own Mark Murray was forced to admit that the evidence is "hardly conclusive that gas prices or anything else have taken a toll on the president's numbers."

Why am I telling you all this? Because while statistics is not an exact science, if you're looking at a poll and objectively saying, hey, this doesn't support the recent trendlines and the overall conditions of the factors likely to influence the numbers, there are two possibilities: (a) something changed so drastically over a past given period that the numbers have turned the other way, or (b) the poll you're looking at is bogus. If you maintain objectivity in analyzing the trendlines and the factors, then option (b) is more likely than not.

And that is what happened here, isn't it? The President's numbers have consistently risen because he's been finally getting some much deserved credit for the improving economy. The economy hasn't slowed down. February concluded the best six months for job creation since 2006, consumer confidence is up and at the verge of going back to pre-recession levels, banks are beginning to reduce mortgage principals, credit markets are easing, small businesses are more optimistic, manufacturing is adding jobs for the first time since 1997, and a whole host of other good economic activities are going on around us. Add to that the self-destruction by the Republican party through their attack on women's health and contraception, while an overwhelming portion of the American people back the Affordable Care Act's requirement for it to be covered under insurance. This has especially been driving women voters to the president. In the above-mentioned Bloomberg poll, economic outlook has improved significantly for Americans, with those saying they are better off now outnumbering those saying they are worse for the first time since they started tracking this question in December of 2010.

The only bad economic news - and the media narrative to "explain" the "dip" in the president's support - has been rising gas prices. But did gas prices just start rising since February? Not really. Americans have been through enough presidents and enough ups and downs in prices at the pump to understand, at a very basic level, that the US government, let alone the President of the United States, doesn't control gas prices. Even those who believe the president can do more to affect gas prices (he is actually doing everything he can) see the rest of the economic signs around them - an improving job market, more crowded malls and restaurants, a more confident housing market, and everything else we just discussed above.

Given all of this, it fits no logic that gas prices - while painful - will be such a dampener of the public mood that people would just ignore the rest of the factors - economic and government - and suddenly start hating the president who is responsible for not only keeping us from going into a second Great Depression but beginning to turn the economy around. It fits no logic, that is, except one: the national media wants to push a horserace, "Obama in peril" narrative. They found their excuse in gas prices, something the Republicans want the media to focus on and blame the president for.

So, here is the sequence of events:

Flurry of objectively good economic data on jobs, consumer confidence, manufacturing, and credit easing.
Republicans turn their focus away from the economic good news and go after women's health.
President's numbers rise on both accounts.
Desperate, GOP wants to focus on gas prices and blaming the president for it.
 Suddenly, there are two 'polls' showing the president's numbers plummeting, and it is explained by pundits with rising gas prices.

Coincidence or data manipulation? Employees of one of these polling organizations can certainly keep insisting that it's "random variance," but hopefully Nate Silver will forgive me if I see sample rigging.