Hacker News new | past | comments | ask | show | jobs | submit login
What Went Wrong with Polling? Some Early Theories (nytimes.com)
3 points by jbegley on Nov 10, 2020 | hide | past | favorite | 4 comments



Oddly enough, I remember reading several articles from people like Nate Silver and other pollsters assuring me that their models had compensated for "Shy Trump Supporters". Some pundits even argued that "Shy Trump Supporters" are a myth.

I personally received something that looked like it was from a pollster in an unsolicited text message. Did I answer it? Of course not. I get inundated with spam calls and texts constantly. I'm also weary of answering any pollster because I have no clue how my responses are recorded and if such responses are being used in an advertising profile.

I wonder if pollsters have some way of compensating for that in their model? Because it looks like their methods are just to try and adjust sampling rates for other demographics rather than changing their questions and outreach methods.


I have a tendency to consider any individual to whom -ster is a valid colloquial suffix to be someone on whom I won't put any effort into honest communication with.

Admittedly, that list is mobster, pollster, and fraudster at the moment, and prone to change on further validation, but as a general guideline it has served alright.

Note: The rule is more accurately codified as <noun>-ster or <verb>-ster.



Convenient, but not a slam dunk. You can't really count most of those since the formula is <complete-word>-ster.

Give me a sec to see what can get filtered out and what exceptions I'd have to accept.

Hmmmm. Most problematic would be minister Adjuster Sister Tester

Then again, those fail the full standalone word criterion.




Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: