Surveys: These Marketing Devices Are Accurate, Right?
November 10, 2020
There’s nothing like a sample, a statistical sample, that is. What’s interesting is that the US polls seem to have been reflecting some interesting but marketing-type trends. The bastion of “real journalism”— the UK Daily Mail — published “…We Did a Good Job: Defiant Pollster Nate Silver Rushes to Defend His Profession after Another Systematic Failure of Polls in the Build-Up to an Election.” Bibliophiles will note that I have omitted the tasteful obscenity. I like to avoid using words likely to irritate the really smart software which edits blog posts.
The write up points out:
FiveThirtyEight founder and editor-in-chief Nate Silver hit back at those slamming the website for being so off with their election predictions.
Let’s think about why FiveThirtyEight and other polls seem to have predicted a reality different from the one generated by humanoids marking ballots.
First, there is the sample. Picking people at random is dependent on a number of factors: Sources, selection bias, humanoids who don’t respond, etc.
Second, there are the humanoids themselves. Some people plug in the “answers” which get the poll over with really fast. I lose interest at the first hint of dark patterns which make it tough to know how may questions I have to answer to get the coupon, pat on the head, or the free shopping sack.
Third, there is counting. Yep, humans or machine things can happen.
Fourth, there is analysis. It is remarkable what one can do when counting or doing “analytics.”
The Daily Mail quotes an expert about making polls better:
‘The polling profession needs to reshape and reorganize their questionnaires,’ Luntz [the polling expert] told DailyMail.com. ‘It’s the only way they’ll ever get it right.’
But I keep thinking about the FiveThirtyEight obscenity. Defensive? Eloquent? Subjective? Insightful?
That subjective thing.
Stephen E Arnold, November 10, 2020
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
