Bloomberg's biased polling data

Since July of last year, the Real Clear Politics tracker shows 64 polls measuring the state of the head-to-head general election match between Donald Trump and Hillary Clinton.

And among all these surveys, the one with the largest purported lead for Clinton (+18%) over Trump is a Bloomberg poll released March 23.  While already a bit dated, if you want a textbook example of liberal polling bias against Trump, this one offers a gem.

The Bloomberg story about their flawed poll screams the following: "Trump gets crushed by Clinton In hypothetical matchup; Kasich is only Republican left in field who beats Clinton."

It gets better:

If he emerges as the nominee, the survey also suggests that Trump faces a difficult path to the White House. Hillary Clinton, his likely Democratic competitor, crushes him in a hypothetical general-election match-up. In another troubling sign, Trump is viewed unfavorably by 68 percent of Americans -- well above the 53 percent who feel that way about Clinton... "Trump's numbers are bad and getting worse," said pollster J. Ann Selzer, who oversaw the survey. "A majority of Americans now describe their feelings toward him as very unfavorable. That's a 13-point spike from November 2015." In the process, the Republican front-runner may also be tarnishing his party's brand. Sixty percent of Americans view the GOP unfavorably, easily the highest level recorded in the poll since it was started in September 2009. The Democratic Party, in contrast, is viewed negatively by 43 percent ... just 29 percent of Americans view Trump favorably.

That sounds terrible for Trump.  The results are almost like they were dominantly coming from the mouths of Barack Obama voters.

Oh wait...they were.

In the poll's methodology, the first sign of trouble is at the following question: "Now, I'm going to mention some people and groups in the news recently. For each, please tell me if your feelings are very favorable, mostly favorable, mostly unfavorable, or very unfavorable."  Obama gets a net favorability rating of 57%, whereas Mitt Romney receives the exact opposite – a net unfavorable rating of 58%.

If this is a truly representative poll of the American public, Romney must really be in the doghouse, while Obama is flying with angelic wings.

And then the epiphany comes when the results to the question arrive about whether "[i]n the 2012 general election, did you vote for [Barack Obama for the Democrats] or [Mitt Romney for the Republicans]."  Apparently 44% of respondents voted Obama, while just 30% voted Romney, for a Obama:Romney voting ratio of 1.5.

Of course, in the general election itself, the actual ratio of Obama:Romney voters was 1.08.  In other words, the poll is biased toward Obama voters, and away from Romney voters, by about 40%.

This provides a needed context.  Would we expect a poll heavily biased toward Obama voters to choose Clinton or Trump in a head-to-head race?  And would Obama supporters be likely to view Trump favorably?  Do you think that Obama voters would prefer Trump over Kasich?

This nonsense litters the polling landscape, and much – if not nearly all – of it appears tilted against Trump and the GOP's anti-establishment base, making it impossible to get a good handle on the actual head-to-head situation.