15 thoughts on “Body-slam on the sister blog

  1. Doesn’t it take an estimate to beat an estimate? Richman and Earnest have a sample of non-citizens that contains, as far as I can tell from the post, non-citizens with a verified vote. If Richman and Earnest did not extrapolate their population estimate correctly from the sample, then it seems that a proper response is to improve their estimate based on demographics of non-citizens in the CCES and the best available estimates of demographics of the non-citizen population.

    • LJ:

      I haven’t looked at the data myself, but based on the discussion it looks like Richman and Earnest had a sample of people who said they were non-citizens and said they voted, with no check on citizenship and some check on voting, with the check on voting revealing a lot of errors. Meanwhile their noisy data results were accompanied with broad claims about real-world relevance. Had Richman and Earnest been more modest—a lot more modest—about their claims, I’d have had no problem with their post. But, in their post and in their published article, they showed no such modesty, instead making strong claims based on what appear to be 5 respondents (out of a total of 32,800) who were recorded to have voted and who stated in their survey they were noncitizens.

      So, sure, if you want to estimate non-citizens’ political behavior using a survey of citizens, go for it, poststratify and all that. Given that I analyzed the Xbox data I can hardly complain when somebody else analyzes CCES.

      But, no, I don’t agree with your statement that “it takes an estimate to beat an estimate.” I think it’s perfectly reasonable for Ahlquist and Gehlbach to criticize those claims that are so weakly based on data. If Richman and Earnest, or anyone else, want to do better, that’s fine, but in the meantime I don’t like the idea of taking such noisy data and treating it as truth.

      • That’s fine. So maybe we should also consider as too weakly based on data Ahlquist and co-authors’ statement from their voter impersonation list experiment study that: “We find no evidence of widespread voter impersonation, even in the states most contested in the Presidential or statewide campaigns”; the data from that study also appear too noisy to learn much from, given that that the weighted 95% confidence interval for alien abduction rates appeared to be reported in Figure 2 as [-4%, 15%] and the 95% confidence interval in their Figure 1 for voter impersonation in contested states appears to extend from [-30% to 20%].

        http://www.johnahlquist.net/files/AhlquistMayerJackmanELJFinal.pdf

        • LJ:

          I’m particularly disturbed when a low-quality study or poorly-evidenced claim is presented as truth on the sister blog. The Monkey Cage has high visibility, and I’m part of it, so I fight back when I see stuff there that bothers me. A recent example is here.

          I took a quick look at the paper you sent me but only a quick look. I’ve expressed concern before about list experiments so I might indeed be skeptical about the conclusion of that paper. From a quick quick read I noticed this remark in that paper: “Our survey only has so much statistical power and list experiments have their own challenges. We cannot reject the null that the amount of voter impersonation is 0% but nor can we reject the null that the amount of fraudulent voting is 1%. Nevertheless, we can confidently show that voter impersonation is not the widespread, endemic behavior that some have claimed it to be.”

          So in some ways their conclusions are consistent with the data that Richman and Earnest analyzed. The key difference is interpretation, in that Richman and Earnest took 5 respondents out of a total of 32,800, or 13 responses out of 54,000, as being a major concern.

        • re: “We cannot reject the null that the amount of voter impersonation is 0% but nor can we reject the null that the amount of fraudulent voting is 1%. Nevertheless, we can confidently show that voter impersonation is not the widespread, endemic behavior that some have claimed it to be.”

          I have a hard time seeing how this is supported by Ahlquist et al.’s data and analysis. In fact, don’t their confidence intervals include a level of voter impersonation way more widespread than even extreme partisan commentators claim exists?

        • Hi Dean,

          Based on the weighted estimate from Sept 2013 reported in Figure 2, the upper end of the 95% confidence interval for the voter impersonation rate does appear to be 1%; however, the corresponding point estimate is an impossible -10%, and the upper end of the weighted 95% confidence interval for the alien abduction rate appears to be 15%. Footnote 31 in the Ahlquist et al. study indicates that a sample of 130,000 would be necessary to reduce the precision of the voter impersonation rate estimate to 1%; the study had a sample of 3,000.

          Like you indicated, their Figure 1 confidence interval for the vote impersonation rate in contested states appears to end around 20%, and I don’t recall anyone claiming that more than 1 out of every 5 persons has committed voter impersonation.

          Maybe it’s me, but it seems odd that Ahlquist et al. would break down the smaller 1000-person sample in Figure 1 into subsets such as contested states, but not report subset results for the larger 3000-person sample in Figure 2.

        • 1% is also pretty large in terms of number of people committing fraud in the overall population! Certainly larger than seems plausible to me — but also larger than would be needed to generate serious concerns if it were happening.

        • Dean:

          Just to put a sense of scale on this: Ansolabehere and Stewart estimate that millions of votes (1.7% of total ballots cast) “were ‘lost’ because of technological problems, administrative deficiencies, and voter confusion.” As I wrote elsewhere in this thread, it’s still fine for people to study illegal voting, but I’ve seen no good evidence that it’s comparable to miscounted or miscast votes. For Richman and Earnest to pick an election decided by 300 votes and point to illegal voting as the tipping point, that completely misses the point.

        • Just to follow up on the Ahlquist et al. list experiment: The data have been made public, and, from what I can tell, the 3,000-person list experiment produced a weighted estimate of 12 percent of persons listed as registered to vote being contacted to sell their vote around the time of the 2012 U.S. general election (p=0.018).

          Code and more analysis here: http://www.ljzigerell.com/?p=4054

        • We’ve been hearing from the respectable press from years that fraudulent voting can’t be a problem of any proportion, but now we see an example — Al Franken’s election in Minnesota to the provide the 60th vote for ObamaCare — that should have set off alarms because there’s a long history of fellow Democrats accusing Somali Democrats in Minnesota of organized vote fraud and violence against a liberal Jewish Democratic incumbent:

          http://isteve.blogspot.com/2014/02/the-future-of-democratic-party.html

          http://www.unz.com/isteve/the-tribulations-of-phyllis-kahn-continue/

  2. I liked their point about raising participation as best way to minimize the fraud risk. There must be some cognitive dissonance for those who share the understandable apathy about their own vote because, really, what could one vote accomplish, and on the other hand rage at the hint of any anomaly with non-citizen voting.

    Just another example of people coming to the US to do the jobs Americans don’t want to do?

  3. Of course we will not speak of organized disenfranchisement of voters in North Carolina, Wisconsin, Georgia, Indiana, Kansas, Tennessee, Texas, Virginia,… It would be impertinent to speculate on many thousands – tens of thousands? – more disenfranchised voters there will be next Tuesday than fraudulent attempts to vote.

    • Chris:

      All these things are worth studying. I don’t object to people studying the organized disenfranchisement of voters, nor do I object to people studying illegal voting. What I do object to is hyping of weak studies, and I was happy to see Ahlquist and Gehlbach shoot down that weak study that had, unfortunately, appeared earlier in the Monkey Cage.

      • Andrew,
        My comment wasn’t directed at you. I’m just amused* that people will go on about the perils of illegal voting while making no mention of organized disenfranchisement efforts. It takes a special kind of mental gymnastics to pull that off.

        *disgusted

  4. Pingback: Shared Stories from This Week: Oct 31, 2014

Leave a Reply

Your email address will not be published. Required fields are marked *