The study and plot show the average percentage of people who believe in one of a set of “deranged conspiracy theories”. That’s measuring a different thing from whether more people believe in deranged conspiracy theories. For example, if moon-landing denial and chemtrails are both believed by an average of 15% of people, and the believers in these theories perfectly overlap, that would mean 15% believe in a deranged conspiracy theory, whereas if they’re perfectly disjoint, it would mean 30% do.
Also, though it’s true that the error bars are large, the plotted mean has gone up by about 20% (3pp) from the 2010-2016 period to the 2018-2025 period. I think that should be an update towards more people believing in these conspiracy theories, even if it isn’t statistically significant.
Hmm. Astra tells me that, given a pretty broad prior and the reported estimate, a basic Bayesian calculation gives about a 95% probability that the number went up, and a 30% probability that it went up by more than 5 percentage points (a ~40% increase over the baseline). That seems pretty substantial though yeah perhaps not an “explosion”.
“if moon-landing denial and chemtrails are both believed by an average of 15% of people, and the believers in these theories perfectly overlap, that would mean 15% believe in a deranged conspiracy theory, whereas if they’re perfectly disjoint, it would mean 30% do.”
Fair enough. But that would apply equally to each era, so wouldn’t change the results.
The overlap wouldn’t necessarily be the same in each era. In my toy example, the first possibility (with complete overlap) could describe an earlier era, and the second possibility (with complete disjointness) could describe a later era, in which case the percentage of people who believe in a deranged conspiracy theory would’ve gone from 15% to 30%.
Given that they increased the number of conspiracy theories in the survey from 8 to 21, it seems pretty plausible that something like this could’ve happened (i.e., that the people who believe in the new conspiracies are different from those who believe in the previous ones). For example, say you make a measurement one year, looking only at chemtrails, and find that 15% of people believe in chemtrails. Next year, you expand that to also look at a totally new conspiracy theory—say, the Texas flood conspiracy—and you find that 5% of people are Texas flooders while the number for chemtrails remains at 15%. If the new Texas flood people are different people from the old chemtrails believers, the number of people who believe in any conspiracy has gone up to 20%, but the measured average has gone down to 10%.
And also, if more people believe in any deranged conspiracy theory, but each of these believes in fewer different conspiracy theories, that could fail to show up in this data, even though the number of people who believe in a deranged conspiracy theory would’ve gone up. Though I think omni-conspiratorialism is just as prevalent today as it ever was, if not more so.
To be clear, I don’t think these toy examples are anywhere close to decisive. I think they’re just some reason to be pretty cautious when interpreting the data.
The study and plot show the average percentage of people who believe in one of a set of “deranged conspiracy theories”. That’s measuring a different thing from whether more people believe in deranged conspiracy theories. For example, if moon-landing denial and chemtrails are both believed by an average of 15% of people, and the believers in these theories perfectly overlap, that would mean 15% believe in a deranged conspiracy theory, whereas if they’re perfectly disjoint, it would mean 30% do.
Also, though it’s true that the error bars are large, the plotted mean has gone up by about 20% (3pp) from the 2010-2016 period to the 2018-2025 period. I think that should be an update towards more people believing in these conspiracy theories, even if it isn’t statistically significant.
If it’s not even close to stat sig different (p=0.11 for pre-2010 vs now) and the data is crazy noisy.
Either way, it’s very unlikely to be an “explosion” by any metric.
Hmm. Astra tells me that, given a pretty broad prior and the reported estimate, a basic Bayesian calculation gives about a 95% probability that the number went up, and a 30% probability that it went up by more than 5 percentage points (a ~40% increase over the baseline). That seems pretty substantial though yeah perhaps not an “explosion”.
“if moon-landing denial and chemtrails are both believed by an average of 15% of people, and the believers in these theories perfectly overlap, that would mean 15% believe in a deranged conspiracy theory, whereas if they’re perfectly disjoint, it would mean 30% do.”
Fair enough. But that would apply equally to each era, so wouldn’t change the results.
The overlap wouldn’t necessarily be the same in each era. In my toy example, the first possibility (with complete overlap) could describe an earlier era, and the second possibility (with complete disjointness) could describe a later era, in which case the percentage of people who believe in a deranged conspiracy theory would’ve gone from 15% to 30%.
Given that they increased the number of conspiracy theories in the survey from 8 to 21, it seems pretty plausible that something like this could’ve happened (i.e., that the people who believe in the new conspiracies are different from those who believe in the previous ones). For example, say you make a measurement one year, looking only at chemtrails, and find that 15% of people believe in chemtrails. Next year, you expand that to also look at a totally new conspiracy theory—say, the Texas flood conspiracy—and you find that 5% of people are Texas flooders while the number for chemtrails remains at 15%. If the new Texas flood people are different people from the old chemtrails believers, the number of people who believe in any conspiracy has gone up to 20%, but the measured average has gone down to 10%.
And also, if more people believe in any deranged conspiracy theory, but each of these believes in fewer different conspiracy theories, that could fail to show up in this data, even though the number of people who believe in a deranged conspiracy theory would’ve gone up. Though I think omni-conspiratorialism is just as prevalent today as it ever was, if not more so.
To be clear, I don’t think these toy examples are anywhere close to decisive. I think they’re just some reason to be pretty cautious when interpreting the data.