Open science, ese! Check an infamous scientific fraud case yourself

I've posted the R and SPSS code and data to conduct the factor analysis on Lewandowsky, et al.’s fraudulent 2012 paper “NASA Faked the Moon Landing—Therefore, (Climate) Science is a Hoax: An Anatomy of the Motivated Rejection of Science”.

Their online study purported to measure belief in various conspiracy theories, asking participants about 14 different conspiracies. Incredibly, the authors reported the same factor loading for all ten items in their main conspiracy factor: .742.

This means that the loadings are false, since they cannot be identical — it's not mathematically possible. The authors never disclose the actual factor loadings.

With my code, you can generate the actual factor loadings. It turns out the conspiracy belief latent variable reported by Lewandowsky, et al. doesn't exist. Few of the items load heavily on any factor, and participants generally didn't endorse the conspiracies. Here are the results:

The paper is false and should be retracted — the authors reported false, identical, and inflated factor loadings.

How could a journal publish a table full of identical item factor loadings? It’s like a correlation matrix full of identical three-digit correlations — impossible. I haven’t found a single case of anyone ever doing this. It’s an error far too obvious to get past reviewers.

Well, the authors inserted this footnote under their table of false loadings: “There were two composite manifest variables for conspiracist ideation. The same loading is shown for all items that entered into a given composite variable.”

That’s not an option. We can’t give the same loading for every item, since they’d necessarily be false and we can’t present false data in a scientific journal article. (“Manifest variable” is a term some foreign researchers use for what Americans call measured or observed variables, or just the item scores. Lewandowsky is Australian.)

That was an attempted Jedi mind trick, to boldly, casually say you’re doing something that you absolutely cannot do and which makes no sense whatsoever. Apparently, it worked on the reviewers.

In their article text, the authors also said: “Responses to items loading on each conspiracist-ideation factor were summed, so that two composite manifest variables were created.”

That’s false, since several of the items didn’t load on a conspiracist ideation factor. As we see above, the moon landing, New World Order, Oklahoma City, and SARS items didn’t load well on any factor. Yet the authors included them in their bizarre “composite manifest variables”. So their statement “responses to items loading on each conspiracist-ideation factor” is false. And note that they never disclose those loadings, which enabled them to conceal the fact that several items, including the one in the title, didn’t load on any factor.

Note also that hardly anyone endorsed most of the conspiracies. So even the .497 loading for the Princess Diana conspiracy is meaningless since only 2.2% of participants endorsed it. Descriptive data trumps inferential statistics — academic researchers and reviewers often get distracted and tricked by inferential statistics, SEM models, factor analysis, etc. without contemplating the descriptive data first. Lewandowsky didn’t report any descriptives in the paper, so readers would have no idea that hardly anyone endorsed these conspiracies.

This was incredibly bold. I’ve never seen or heard of anything like it. A table of identical, false factor loadings just staring right at readers, with a meaningless string of words used as a Jedi mind trick… And never reporting the actual loadings.

Moreover, the titular link between believing that the moon landings and climate science are hoaxes doesn't exist — the authors fabricated it. In the sample of 1,145 participants, only ten (10) participants thought that the moon landings were a hoax — effectively zero — and most of those ten did not believe that climate science was a hoax (though we can't do anything with 10 of 1,145 participants anyway – it wouldn't matter if all ten thought climate science was a hoax).

Thus, there's no link between the two beliefs in this study — there wasn't even any significant data/endorsement for the moon landing belief. The authors even asserted a causal structure for this nonexistent link, with the moon landing hoax belief driving the climate hoax belief: NASA Faked the Moon Landing—Therefore, (Climate) Science is a Hoax.

Note that this false and fraudulent study was widely reported by credulous media at the time, including Scientific American and the Guardian. It drove a narrative that climate skeptics were broadly irrational and believed in all sorts of crazy conspiracies, particularly that the moon landings (plural — there were six) were a hoax. The Association for Psychological Science (APS), didn’t just publish this fraud – they promoted it in their Monitor magazine. No one could be bothered to think critically, to properly review it, or to even look at the data.

“Science” in this era, especially social science, is extremely unreliable. Two reasons stand out:

1. Leftism. Leftist ideology saturates academia and researchers are motivated to advance leftist narratives and power, and to derogate, mock, and even dehumanize non-leftists, particularly conservatives. (Though academics have a binary model of the intellectual and political landscapes, viewing all non-leftists as conservative or right-wing.)

2. There’s little independent thinking and validation going on. A peer-reviewed journal published a table of identical item factor loadings. The reviewers fell for the Jedi mind trick. No one even noticed that the title was false. (Peer reviewers don’t look at the data for studies — no one does, even just to see that it exists.)

We know nothing about the participants of the study. We don’t even know their nationality. Or their sex, age, or any other basic demographics. This is as unprecedented as a table full of identical, false factor loadings. I’ve never encountered a study that didn’t even disclose the sex or age breakdown of their sample.

Lewandowsky was in Australia at the time and this was an online study. He was active in arguing with climate skeptics down under and I assume most participants were in Australia and New Zealand, but we simply don’t know.

Two different data files were posted, but for the same 1,145 participants. All the survey questions used a crude 1-4 scale1 of disagreement to agreement, and one of the files had no ones for one item and no fours for another item. At least one of the files — the one with the missing ones and fours — was altered, possibly both. Summary here.

Something is fundamentally off with the data, since endorsement of the moon landing conspiracy was far too low at 0.8% to be real. In the US, typical rates are 6-8%, and Australia has bounced around 5-10%. The paper’s substantive claims are false, including its very title, but this study wasn’t usable to begin with. Mystery participants, an invalid response scale, poorly worded items, deleted participants, and altered data = there’s nothing we can do with this. As I’ve said before, non-academic organizations like Gallup and Pew operate at much higher standards of scientific rigor than academic researchers.

Joe Duarte grew up in small copper mining towns in Southern Arizona, earned his PhD in social psychology, and focuses on political bias in media and academic research. You can find him on X here and contact him at gravity at protonmail.com.