Jesse Richman is a fascist collaborator. Why do I say that? First, Trump and his far right nationalist supporters are fascists. What else do you call such race/nationality based right wing extremism that scapegoats minorities and has no respect for independent institutions of government, a free press, etc.? Second, Jesse Richman and his co-authors, Gulshan Chattha and David Earnest, collaborate with them. They have done so through disingenuous research into voter fraud that finds that large numbers of non-citizens vote in US elections, despite the fact that no one else has come up with any evidence to support that claim. Voter fraud is extremely rare, non-citizen voter fraud even more so.
But any researcher who wants to make a name for themselves can come up with a junk science study to find otherwise. That’s exactly what Richman et al. did, and it worked. The paper was published in the journal Electoral Studies and received national attention. Jesse Richman has since been hired by Trump’s voter fraud commission to duplicate his dubious work for them. He is literally a collaborator with a far right white-nationalist (aka fascist) political movement.
So what makes their bullshit study such bullshit? Start with the sample size. They use an internet survey that had 32,800 respondents in 2008 and 55,400 in 2010. Sounds pretty big right? But non-citizens made up only 339 respondents in 2008 and 489 in 2010, or about 1% in each year. Non-citizens make up 7% of the population of the US. The survey made no attempt to get a representative sample of the US population writ large because the survey was designed for eligible voters, and non-citizens are not allowed to vote[1]. Despite this, the authors claim in their paper that the survey’s sample was “selected to mirror the demographic characteristics of the US population”. They must mean the population of eligible voters, because otherwise that is a lie. Elsewhere they state that “it is impossible to tell for certain whether the non-citizens who responded to the survey were representative of the broader population of non-citizens,” a severe understatement at best.
Of this small unrepresentative group, just under 20% claimed to have been registered to vote or were verified to be on voter registration rolls. The authors could only verify that a minority of that number were registered. Only 3.3% both claimed and were verified to be registered in 2008, and none in 2010. Some self-identified non-citizens claimed to not be registered to vote, but were verified to be registered. The authors leave out any details of the verification and use this finding to assume a higher number of non-citizens were registered than the survey responses and verification process indicated. For example, in 2008 67 non-citizens either reported or were verified to be registered; the authors assume 84 in their estimates.
Of this minority of non-citizen respondents, only a minority voted. In 2008 “71 non-citizens answered a survey question indicating whether they voted, and also had their vote validated. Among these, 56 indicated that they did not vote (but two of these cast a validated vote), while 13 indicated that they voted, of whom five cast a validated vote.” Notice how small the numbers are getting? They are using an un-representative sample size of less than a hundred people to reach conclusions about the voting behavior of millions. This is bullshit social science; that point is hard to understate.
The authors apply population weights so that their teeny tiny sample’s demographic characteristics match that of the non-citizen population as a whole. This is a perfectly reasonable technique when using a large enough sample size. On such a small sample it is unreliable at best. While the average characteristics of their sample may now match non-citizens as a whole it does not mean the distribution of survey responses will match the true figures for the larger population. Basically, their statistical techniques cannot overcome the flaws in their sample.
The authors conclude that “non-voter participation has been large enough to change meaningful election outcomes including Electoral College votes, and congressional elections.” How large? Their adjusted estimate is 6.4% of non-citizens in 2008 and 2.2% in 2010. Again, this is all based on an internet survey where 13 people out of 32,800 in 2008 said they were non-citizens and voted; only five of which could be confirmed. What are the chances that five to 13 people out of 32,800 clicked the wrong box on the citizenship or voting questions? The authors gloss over this and instead publish a highly flawed and prejudicial finding that falls apart upon the slightest scrutiny. A 2015 paper published in the same journal finds that observed levels of response error to the survey Richman et al. use can explain the entirety of their results.
Jesse Richman is now selling these dubious techniques to the Trump administration, where his latest finding that 18,000 non-citizens voted in Kansas was based on a sample size of 37. At best the authors wanted some limelight and were willing to slander a scapegoated minority to get it. Perhaps Jesse Richman and his co-collaborators are not hateful xenophobes and are just incredibly naïve. Perhaps they didn’t know they’d become the academic poster children of voter suppression. Whether bigots or fools they do not deserve the positions they hold at accredited universities. They should have known better than to publish such junk. That they did anyway speaks volumes as to their lack of ethical standards and/or lack of understanding of statistics.
An open letter signed by more than 150 political scientists states that “this paper has been shown to be incorrect”. Jesse Richman has offered numerous responses to the criticism on his blog.
1. With the exception of a few localities in the great state of Maryland, which did not account for the self-identified non-citizen voters found in the survey.
No comments:
Post a Comment