Breaking NBC News Poll: Positive Opinions About Trump Grow After Second Debate

[Steve_McGarrett]

The internal polls must be brutal for her.

One of today's leaked Podesta emails confirms it too:

Positive opinions of Trump grow after second debate, NBC/SurveyMonkey poll says

 

[Siete]

nope, email was hacked.

 

[Dalia]

Yes it's depressing, The joker Hillary

will win against Trump mon chouchou Batman...

 

[320 Years of History]

A Survey Monkey online poll. Really? Is that really your idea of a statistically valid thus representationally faithful poll?

Chuck Todd of NBC News says the SurveyMonkey methodology has been "rigorously vetted" and that the polls are "scientifically done." That may be the case, but the accuracy measures for the 2014 election show a different story.

I understand why organizations opt to use Survey Monkey. Having had to perform a statistically valid and reliable survey as part of my master's thesis, and using statistically valid sampling methods to perform audits of Fortune 500 companies, thus knowing quite well how to perform them, I don't understand why people put much stock in Survey Monkey's results.

• Validity and Reliability in Surveys
  
• A Critical Assessment of Online Survey Tools

I suppose it's nice to see someone saying something one wants to hear, but hearing the truth -- whether one likes it or not -- is far more useful for good decision making and sound opinion forming.


Edit:
FWIW, the problem with online polls isn't that the surveys themselves cannot be effectively and properly designed. They can. The problem is that the selection process for respondents injects too much bias regarding who is willing to participate in the survey. Lots of folks, myself included, simply won't bother with taking a statistically valid online poll that's been designed to eliminate/mitigate the bias that inherent to merely asking a question. Such polls have a lot of questions and take a long time to complete because in order to mitigate inherent polling bias (not topical bias), one must ask the same thing slightly differently five to ten times for each topic the poll addresses and one can't very well just ask about just one thing multiple times. The questions for one topic have to separated by enough other questions that respondents are unlikely to perceive they are being asked the same thing over and over.

One example of that kind of bias derives entirely from how people interpret words, notwithstanding that the words'/phrases' meanings should be well understood by everyone. To illustrate, I once wrote that "I was not moved by Donald Trump." Countless people misinterpreted that statement. Now, I've used that phrase all my life and not once until I used it on the Internet had I ever been misconstrued by my audience. That and other inherent challenges in poll design make online polls and the selection of respondents not necessarily representative of the nation at large.