Tuesday, June 20, 2017

At it again?

Reuters-Ipsos presidential approval polling results from last week:


Yikes, looks bad. They all do, excepting Rasmussen--the outfit that nailed the popular vote margin perfectly in 2016--which is mediocre rather than just plain bad for Trump.

R-I's sample raises suspicions, though. The partisan breakdown among those surveyed:

Democrat -- 45%
Republican -- 33%
Independent/other -- 21%

Compare that to the 2016 exit poll results:


Exit polls showed Democrats with a +3 advantage in November, but R-I's approval poll shows a +12 Democrat advantage, just as it's inaccurate pre-election polls regularly did. In fairness, the approval rating poll is of the general public, not of registered or even likely voters (though the latter was the case during 2016 when R-I was similarly overstating Democrat numbers).

I imagine Reuters' tech guys going about their work honestly, making the data--complete with fully customizable cross-tabs--fed to them by the organization's pollsters publicly accessible. The more ideological mass of the news service's managers and executives, meanwhile, are unaware of how exposed this leaves them.

Don't they know transparency is their achilles' heel? Here's how CBS, which gives Trump a net disapproval rating of 21, explains its methodology:
The poll employed a random digit dial methodology. For the landline sample, a respondent was randomly selected from all adults in the household. For the cell sample, interviews were conducted with the person who answered the phone.

Interviews were conducted in English and Spanish using live interviewers. The data have been weighted to reflect U.S. Census figures on demographic variables.
Nothing on the respective samples sizes by partisan affiliation. Reveal it and any fudging is easily detectable. Keep it under wraps, though, and...

Wait a minute. Random dialing methodology. Interviews conducted in Spanish. Data weighted to census figures. Are respondents even asked about citizenship status?

I wonder how many of these approval rating results--not just from CBS, but in general--include illegal immigrant and other non-citizen responses.

1 comment:

Random Dude on the Internet said...

Pollsters continue to lose credibility. The country supposedly vehemently hates Donald Trump but yet Democrats have yet to win a single special election so far this year despite Democrats trying to make these special elections a referendum on Trump's presidency. The polls were so skewed with the Georgia special election that Nate Silver looked like a fool yet again. If Trump is doing as bad as Reuters is implying, Jon Ossoff and others would have won their elections. Even if you were to take the polls at face value, that would mean that people dislike Trump but not so much that they want to vote against him. Unfortunately for the polling industry, people are just not buying it.