Sorry. And the part where they intentionally misquote? And misrepresent the models? They lied, plain and simple
Hajnal et al coded one state's turnout as 0, which dramatically skewed their results --- so that the point estimate of voter ID laws was to increase turnout. This was implausible, so they just assumed that the sign went the other way and reported this point estimate negative.
The part where they lie in the response is directly quoting *half a sentence* in which the first part negates the part they quoted.
I haven't thought about this in a few months but this second part still baffles me. There's just no way they could get away with it...except for the obscuritanist trolls on this website.
Thanks. Unbelievably bad practices.
They also misinterpret an estimated positive total effect of voter ID on turnout among minorities as a negative effect because the race interaction is negative (but smaller than the coefficient on the lower-order term).
they also lie about the models that were present in the original paper. And they claim to run a placebo check that actually isn't a placebo check.These aren't lies and it does not rise to the level of research misconduct. ZH did not code the data as 0, as is claimed. They were just not careful enough to notice this issue with the data. They misinterpreted their results, but this does not qualify as lying. It was just poorly executed research that was inaccurately reported. The key to research misconduct is intent. Sloppy, yes. Intentionally falsifying research - no.
We do have examples of bona fide research misconduct in our discipline, and nothing happens in those cases. So, the general point is accurate.