The data isn’t exactly new, but it is still fairlyrecent:
[ul]
[li]Republicans think 38 percent of Democrats are gay, lesbian or bisexual (the actual figure is 6 percent.)[/li][li]Democrats think 44 percent of Republicans earn $250,000 a year or more (the actual figure is 2 percent.)[/li][li]Republicans think that 46 percent of Democrats are African-American (the actual figure is 24 percent.)[/li][li]Democrats think that 44 percent of Republicans are 65 years of age or older (the actual figure is 21 percent.)[/li][li]Republicans think that 44 percent of Democrats are members of a labor union (the actual figure is 11 percent.)[/li][/ul]
The study also notes that people who have the keenest interest in politics, tend to have the most distorted viewpoints of both sides.
What’s the methodology here? Ask a bunch of Democrats “What proportion of Republicans do you think are over 65”, and then average all of the answers?
And really, those results (assuming they actually mean anything) are more depressing for what they say about reasoning ability than partisanship. Like, I don’t think that anyone, of any party, actually thinks that 1 in 5 Americans is non-straight… but that’s what “38% of Democrats” would imply.
This old WaPo article talks about Americans thinking 28% of the Federal budget goes to foreign aid, rather than the actual 1%.
The fact that we tend to exaggerate imputed characteristics of the “other team” is not that surprising. The Democratic stereotype of a Republican is a rich, old, white person, so Democrats exaggerate the percentage that they think fit that description. The Republican stereotype of a Democrat is a melange of minority groups (including LBGTQ), low-income (non-white) welfare recipients, and pointy-headed Marxist academics, so they overestimate the percentage with one or more of those characteristics.
One of the reasons that government data gathering and statistical reporting is so important is that actual decision makers, both in the private and public sector, need the real numbers for planning and analysis, not pulled-from-the-posterior guesswork that this kind of survey data highlights as nearly universal.