Skip to content

A very short statistical consulting story

I received the following email:

Professor Gelman,

My firm represents ** (Defendant) in a case pending in the U.S. District Court for the District of **. This case concerns [a topic in political science that you have written about].

I’ve reviewed your background and think that your research and interests, in particular your statistical background, may offer a valuable perspective in this matter.

I’ve attached a report drafted by Plaintiffs’ expert, **. The Plaintiffs have submitted this report in support of their Motion for Preliminary Injunction. Our response to the Plaintiffs’ Motion is due on **. This is the same date by which we would need to submit any rebuttal expert reports.

Do you have a few moments when I could discuss this case in further detail with you? If so, please let me know when I could give you a call and the best number to reach you.

Thank you,

I replied:

Hi—I took a look at **’s report and it looks pretty good. So I don’t know that you should be contesting it. He seems to have done a solid analysis.

That was an easy consulting job—actually, not a job at all, as I declined the opportunity to take this one on. People send me so many bad analyses to look at; it’s refreshing when they send me solid work for a change.

P.S. This all happened a year ago and appeared just now because of a combination of usual blog delay and bumping due to one of our coronavirus posts this spring.


  1. Zhou Fang says:

    Thanks for the PS. For a moment I thought you missed the Minnesota/Michigan thing, the Benford thing, or the “A-B is negatively correlated with B!” thing, amongst all the other pseudo-statistical BS that’s flying around lately.

    • Andrew says:


      Some of these came up in blog comments during the past few weeks. I didn’t want to post on them in part because I hate to even dignify them with a discussion. I’m sure that some of the proponents of those data analyses sincerely believe them, but the bigger picture seems to be that these analyses are part of a strategy by political operatives of “flooding the zone” with crappy arguments of all kind, on the hope that something will stick or that there will be a general lessening of respect for the democratic process. So, while I’m glad that people are refuting these arguments as they come in, I don’t want to play the game of taking each one seriously.

    • paul alper says:

      For those or you who missed “the Minnesota/Michigan thing” go to

      “But as the Powerline blog first reported, the affidavit made a major mistake. Its data wasn’t actually from Michigan; it was from Minnesota.”

      The confusion was due to the similarity of the abbreviations, MN and MI, and the total incompetence of a Trumpista security consultant.

      “The text [of the affidavit] states that, in Wayne County, ’25 of those 47 precincts/townships show 100% turnout.’ But it then lists 25 precincts from Minnesota.”
      Furthermore, “in fact they are from rural Minnesota — i.e., Trump country.”

    • jimmy says:

      what are the Benford thing and the “A-B is negatively correlated with B!” thing?

      • Zhou Fang says:

        Violating Andrew’s “don’t want to dignify them with a discussion thing”…

        The Benford’s law thing is a claim that Biden’s county vote totals violate Benford’s law (i.e. that thing about the first digits being more often 1 than any other), ignoring that Benford’s law requires the values of interest span multiple orders of magnitude, whereas they looking at a number of counties with about 1000 voters each so Biden getting 50-80% of the vote in each county means the first digit of his vote totals would usually be 5-8.

        The A-B is negatively correlated with B thing is a claim that the existence of a negative correlation between county proportions of “People who voted straight ticket GOP/People who voted straight ticket” and “(People who voted Trump/People who who voted Biden or Trump) minus (People who voted straight ticket GOP/People who voted straight ticket)” shows votes stolen from Trump, when it obviously follows from basic correlation theory.

  2. Matěj G. says:

    Are you at liberty to share any information at all about the analysis? It might be instructive to see why you found it good. (Especially among the heaps of bad science that appear so frequently here.)

    • Andrew says:


      It was a causal inference based on a combination of experimental and observational data. Nothing fancy, just some reasonable inferences and some reasonable discussion of the relevance of those inferences to the problem at hand.

  3. jonathan says:

    Reminds me of a young lawyer story. Client came in. They had engaged in large tax fraud but it was incredibly hard to unravel what they did. The IRS has to share their work, just like any prosecution, so we had this giant mess of documents. Near as I could figure, they’d maybe put in $25k and generated into the millions of deductions, not only from illegal shelters but by not even paying the money into those. We had a meeting with the agent, who looked like he’d been jerking off in his car because his shirt was literally poking out of his pants. He was scary brilliant. Walked into the appellate division. Met the guy with the empty desk except for a little US flag in a holder. He offered to settle for $25k. The brilliant dude had transferred to international investigations and no one could figure out the file. We went back to the office and the partner put his feet up and said, ‘what should we bill them?’ It was a large amount, multiples of the settlement offer, but none of it came down to me. They saved huge sums and possible jail. But then we had a client come in and they showed us they’d not paid withholding. The advice was like yours but it was ‘go to them and ask for mercy’ because there was no way out. We didnt bill them.

Leave a Reply