Poisoning the statistics

 

                                        The toxicologist.  Photo source: Drew Angerer, Getty Images

Donald Trump’s outrageous comment that immigrants are “poisoning the blood” of America has stirred TV news analysts. On CNN NewsNight Monday, Abby Phillip commented:

“In Iowa, for example, likely Republican caucus-goers say that some of Trump's most authoritarian-inspired solutions and statements make them actually more likely to support him. Sweeping raids, giant camps, mass deportations, well, 50 percent support that. Root out the vermin, 43 percent support that. Immigrants are poisoning the blood of America, 42 percent.

“And you hear it, the adulation of authoritarians. And you hear the adulation that they crave wherever Trump goes.”

Granted: Trump is a cretin. But the question here is whether such remarks lead Republicans to support him. The poll results quoted by Phillip are sketchy. But note that if 42% of likely Republican caucus-goers polled say the “poisoning the blood” remark would make them more likely to support Trump, 58% do not. If all 58% said that they were less likely to support Trump because of the remark, then one can conclude that the typical respondent is less likely to back him. 

Phillip didn’t give data on “don’t know” or “neutral” responses, much less a margin of error, or even which poll it was. (It was the Des Moines Register poll.) . But CNN This Morning gave a few details today (Thursday): 28% of the respondents said they were less likely to support Trump because of the remark, and 29% were basically neutral. I will account for rounding error or for untabulated responses by increasing "neutral" responses to 30%.

We want to know whether a vitriolic remark would make a Republican more likely to support a candidate.  So let's calculate the average response of respondents in a way that measures the respondent's love of vitriol.  Let's assign a 1 to a "more likely to support" response, a -1 to a "less likely" response, and a 0 to "neutral" responses. Now let's calculate the average response. I get 1*.42 - 1*.28 + 0*.3 = .14. This is a positive value. That is, the average respondent was more likely to support Trump because of the remark. And this conclusion is probably statistically significant.  In other words, it is probably true of all likely Republican caucus goers in Iowa, not just those in the sample. I can't be certain of this, because CNN This Morning gave few details about the poll. At minimum, one needs to know the number of respondents; usually the margin of error can be roughly estimated from this. But it is probably safe to say that the typical caucus goer is slightly more likely to support Trump because of the remark...although I would like to know more about how the question was phrased and how the poll was carried out.       

My point is this: When we want to know about the average respondent, we should look at the average response and not just at a particular response.

Not a complicated point. But it sure stumps journalists. On CNN This Morning, Michael Gold, political correspondent for The New York Times, said: "I would also look at the 30 percent of people who said they don't matter [that is, add the neutral responses to the positive responses] because I think those two numbers together are really telling." Gold needs to go get a clue. The negative responses matter, too. Why can't Times reporters do grade-school math?

As CNN demonstrates, statistical misreporting is most severe among TV journalists, because they are more actors than reporters. They strike Authoritative Tones and Glowers.  But the truth is that when it comes to statistics, they have no idea what they are talking about. Come to think of it, maybe that's the reason for the A. T. and G. 

One unfortunate possible consequence of the incorrect reasoning by Phillip, Gold, & Company is that Trumpists may be a lot more inspired to try their own vitriolic remarks to win votes. Judging from this one polling question, the average caucus goer does support vitriol -- but only slightly; not to the extent that Phillip and Gold suggested. The expected value of the response was only .14. A very enthusiastic response to vitriol would have been much closer to 1. 

Here's another way to interpret the result: Of 100 caucus goers, 14 more are more likely to back a candidate because of venom than oppose him because of it. The other 30 are indifferent. Thus venom can boost one's chances of winning among Iowa Republicans, if one's morals are slippery enough. But it would not be safe to build a campaign on vitriol. It is a side issue.  

So let's not overstate the political profit in vitriol. One cretin in the primaries is enough. –Leon Taylor, Baltimore tayloralmaty@gmail.com

Notes

For helpful comments, I thank but do not implicate Annabel Benson, Paul Higgins, and Mark Kennet.  

 

References

CNN NewsNight with Abby Phillip.  CNN.com - Transcripts  December 18, 2023, 22:00 ET.