Polls Apart

Media is quick to quote polls in which Donald Trump is trailing Hillary Clinton.
That’s what they want, so they are working hard to promote that meme.

Sierra Rayne at Americanthinker.com had a few interesting things to say about the reliability of these polls:

Stepping out of reality into the rapidly expanding landfill of polling data biased against Donald Trump reveals what a disaster is taking place when it comes to surveying the public’s actual opinion — rather than the desired liberal narrative — about the 2016 general election.

Public Policy Polling (PPP) has released yet another state poll on the Trump versus Hillary Clinton match-up, this time for Virginia. Yesterday’s article examined a range of PPP’s state-level polling data in the Trump v. Clinton cage match, revealing some apparently serious liberal bias. The Virginia poll just adds to the concerns.

The proverbial “tell” in these types of data sets is how respondents answered the question regarding their presidential vote in 2012. If the poll is representative of the public, the relative percentages of Obama 2012 versus Romney 2012 voters surveyed should approximate — within reason — how the state in question actually voted in 2012. But if there are significant deviations between the poll’s composition and the 2012 results, the cause either needs to be fully explained by the pollster, or we default to the assumption of a bias.

PPP’s Virginia poll, representing “one of the most important swing states in the country,” claims the following:

The Presidential race in Virginia is pretty tight. Hillary Clinton leads Donald Trump 42-39, with Libertarian Gary Johnson at 6% and Green Party candidate Jill Stein at 2%. In a head to head contest Clinton’s lead remains 3 points at 48/45. Clinton’s benefiting from Democrats in Virginia (83/8) being more unified around her than Republicans (76/5) are around Trump. But with independents Trump’s up 42/29.

When asked who they voted for in the last presidential election, 50% of respondents said Obama while just 41% said Romney, for a 9% Obama (read: liberal) edge. But Obama only won the state by 3.9% in 2012, meaning there appears to be at least a 5% liberal bias in the survey composition.

Remove that bias favoring the Democratic candidate, and Clinton’s lead disappears — leaving Trump likely leading by 2% or more in Virginia, depending on the potential presence of other compounding biases in the poll.

Then there is the latest edition of Reuter’s Polling Explorer from June 14, supposedly showing Clinton up by 8.5% over Trump, 39.1% to 30.6%. But of the 1,481 respondents, 642 (43.3%) are Democrats, 493 (33.3%) are Republicans, and 206 (13.9%) are Independents, with 138 (9.3%) “members of another party.” A 10% bias of Democrats over Republicans is 9% above the past two-month average of actual party affiliations. Remove that liberal bias, and now the race is a statistical tie.

Even worse, when asked who they voted for in 2012, 582 (39.3%) said Obama and just 355 (24.0%) said Romney. Thus, since the national results in 2012 only had Obama ahead of Romney in the popular vote by 3.9%, we conclude this suggests a 11.4% liberal bias in the survey composition. Based on this built-in bias, it appears Trump may actually be ahead of Clinton by nearly 3% at the national level once the bias is corrected for.

Finally, there is a Reuters/Ipsos poll released Thursday — and conducted June 11-15 — that claims Clinton is up 9% over Trump in the head-to-head. No demographic data was released with the poll, which should raise a red flag. With what information we can tease out of the information provided, significant flaws are apparent.

Based on the data for “All Adult Americans” surveyed on issues such as the major problems facing the nation, as well as Obama’s approval rating and whether the nation is headed on the right track or not, it is clear that the ratio of Democrats:Republicans in the poll was a remarkably high 2:1! For interested readers, the general math behind such calculations is described in my previous article.

Moving on to the head-to-head match-up among registered voters, solving the available polling data with a 3-equation system solver reveals that the Trump versus Clinton poll appears to be comprised of about 46% Democrats and 36% Republicans, for a 10% Democrat bias. Yet again, remove the clear bias, and Clinton’s lead is gone.

This isn’t surprising, given the source. Serious concerns have been raised previously over the possible political motivations behind liberal bias in Ipsos polling.

Overall, in all polls seen to date at the state or national levels, systematic liberal bias is clear. In some cases, Democrats are being polled at apparent 2:1 ratios over Republicans, and in all situations, once the polling bias is removed, so is any Clinton lead.

Best to avoid looking at any polls at this point. They are not accurate.

... Leave a Reply