concept
The pollster's practice of adjusting the weight of each demographic group in a sample to match the population. In episode 01 Marvin Barth explains it and argues that when the missing voters are low-trust, reweighting and oversampling can make the bias worse. Start with episode 01, at about 30 minutes.
Across the episodes
01
Enters as the industry's standard repair, which Barth argues cannot reach the missing voters
Barth introduces the standard correction for an unrepresentative sample.
"So what you do is you adjust the weights of those."Marvin, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 29:52
His irony is in "supposedly": the correction fixes demographics, not trust.
"And voila, you have now a supposedly unbiased poll."Marvin, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 30:21
The pollsters' next remedy, which Barth argues only amplifies the high-trust members of each group.
"So what we're really going to do is we're going to try and massively oversample it."Marvin, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 33:46
Grant's summary of the paradox: the fix can be the problem.
"you amplify the error."Grant Williams, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 34:37
Connections
fails onLow-Trust Voters
Weighting cannot reach voters who never respond.