> "In a recent YouGov survey, 70% of US Millennials said they'd vote for socialism."
This isn't related to generational prognostication but it's really dumb that Nick Clegg, a former British politician, doesn't stop and give this claim a basic reality check. Is Zuckerberg getting his money's worth from Clegg?
It's sort of an open secret in the polling community that panel polls have serious balance problems when it comes to political questions like this. Even the best polling firms are affected by extreme levels of what is called "volunteering bias" which they have no way to remove. The underlying panels are made up of people who aren't representative of the actual population in some key respects and so pollsters compete on the skill of their models in weighting the responses. For their bread and butter surveys (brand recognition etc) it's assumed that the bias doesn't matter, but if it does the problem is fixed with models that are calibrated for a relatively small set of tasks, like predicting elections. When polling firms ask socio-political questions for which they don't have sufficient remodelling ability they frequently yield extreme results, and this is one of them.
People are just way too willing to repeat claims by polling firms even when they are mad as a box of frogs. I wrote an essay about this problem a few years ago, using data from Pew Research and some social science studies to show how this problem can lead to extremely implausible claims about what people think (it was about a YouGov poll) [1]. Pew is one of the better/more self-reflective pollsters and have done some great research into the volunteering bias problem.
A few years later the problem was demonstrated again when the BBC blithely reported that a quarter of the British population thought COVID was a hoax, that millions of people had participated in marches against 15 Minute Cities and obscure conspiracy theory magazine "The Light" was by far Britain's top selling news magazine with over 3.4 million people helping to distribute it (i.e. a distribution network six times larger than the Labour Party) [2]. The source was a poll by Savanta laundered through King's College London [3]. Despite three different orgs being involved nobody sanity checked these numbers, and only KCL apologised (and even then only in a half-hearted way). The BBC doubled down even (Marianna Spring needless to say).
tl;dr take polling results on social questions with a big dose of salt.
This isn't related to generational prognostication but it's really dumb that Nick Clegg, a former British politician, doesn't stop and give this claim a basic reality check. Is Zuckerberg getting his money's worth from Clegg?
It's sort of an open secret in the polling community that panel polls have serious balance problems when it comes to political questions like this. Even the best polling firms are affected by extreme levels of what is called "volunteering bias" which they have no way to remove. The underlying panels are made up of people who aren't representative of the actual population in some key respects and so pollsters compete on the skill of their models in weighting the responses. For their bread and butter surveys (brand recognition etc) it's assumed that the bias doesn't matter, but if it does the problem is fixed with models that are calibrated for a relatively small set of tasks, like predicting elections. When polling firms ask socio-political questions for which they don't have sufficient remodelling ability they frequently yield extreme results, and this is one of them.
People are just way too willing to repeat claims by polling firms even when they are mad as a box of frogs. I wrote an essay about this problem a few years ago, using data from Pew Research and some social science studies to show how this problem can lead to extremely implausible claims about what people think (it was about a YouGov poll) [1]. Pew is one of the better/more self-reflective pollsters and have done some great research into the volunteering bias problem.
A few years later the problem was demonstrated again when the BBC blithely reported that a quarter of the British population thought COVID was a hoax, that millions of people had participated in marches against 15 Minute Cities and obscure conspiracy theory magazine "The Light" was by far Britain's top selling news magazine with over 3.4 million people helping to distribute it (i.e. a distribution network six times larger than the Labour Party) [2]. The source was a poll by Savanta laundered through King's College London [3]. Despite three different orgs being involved nobody sanity checked these numbers, and only KCL apologised (and even then only in a half-hearted way). The BBC doubled down even (Marianna Spring needless to say).
tl;dr take polling results on social questions with a big dose of salt.
[1] https://dailysceptic.org/do-online-opinion-polls-overestimat...
[2] https://morningstaronline.co.uk/article/f/bbc-verify-falls-i...
[3] https://www.kcl.ac.uk/policy-institute/assets/conspiracy-bel...