“How to Lie with Statistics” guy worked for the tobacco industry to mock studies of the risks of smoking statistics

Remember How to Lie With Statistics? It turns out that the author worked for the cigarette companies. John Mashey points to this, from Robert Proctor’s book, “Golden Holocaust: Origins of the Cigarette Catastrophe and the Case for Abolition”:

Darrell Huff, author of the wildly popular (and aptly named) How to Lie With Statistics, was paid to testify before Congress in the 1950s and then again in the 1960s, with the assigned task of ridiculing any notion of a cigarette-disease link. On March 22, 1965, Huff testified at hearings on cigarette labeling and advertising, accusing the recent Surgeon General’s report of myriad failures and “fallacies.” Huff peppered his attack with with amusing asides and anecdotes, lampooning spurious correlations like that between the size of Dutch families and the number of storks nesting on rooftops–which proves not that storks bring babies but rather that people with large families tend to have larger houses (which therefore attract more storks).

This was all a surprise to me, and I suspect to other statisticians as well. For example, Huff’s activities with the cigarette companies are not mentioned on his Wikipedia page (as of 17 Apr 2012), nor are they mentioned in an article on Huff by probabilist J. Michael Steele from 2005.

“How to Lie with Smoking Statistics”

Darrell Huff is best known for How to Lie with Statistics but he wrote or cowrote several other books, including Pictures by Pete (1944), The Dog that Came True (1946), How to Take a Chance (1959), Score: The Strategy of Taking Tests (1961), Cycles in Your Life (1964), How to be the Parent of a Successful Creative Child (1968), Twenty Careers of Tomorrow (1945), and How to Lower Your Food Bills (1963).

It appears that in the late 1960s he was also working on a book called “How to Lie with Smoking Statistics,” which the publisher saw “high likelihood of proceeding into print.”

In November 1965, a letter was sent to Huff as follows:

Here‘s a letter from 1967 where Huff asks the tobacco dudes for another $1500 to keep writing.

And here‘s a letter from mid-1968 from Huff’s publisher, Macmillan:

But publication “as soon as possible” never seems to have occurred.

What happened? In one of these documents, William Kloepfer, vice president for public relations for the Tobacco Institute, wrote of the manuscript, “Frankly, this mass of verbiage needs drastic editing before it will directly address itself to the needs of our industry.”

After glancing at a couple of sections from the draft, I gotta say that William Kloepfer had a point: “mass of verbiage” is a pretty good description! Huff’s book chapter reads like a bad sitcom where the writers were too lazy to put together enough material and they just milk the same couple of jokes over and over again.

In retrospect, I think Huff really dodged a bullet on this one. If “How to Lie with Smoking Statistics” had come out, I expect it would’ve destroyed his reputation—remember, we’re talking 1969 here, that’s five years after the Surgeon General’s report—and taken a big bite out of the later sales and reputation of his 1954 bestseller.

How sincere was Huff? Did he tank his book for strategic reasons?

I wanted to call this post, How to Lie With “How to Lie With Statistics,” but to be fair I have no reason to believe that Huff was lying or intentionally deceiving in his testimony. He may well have simply been misleading himself in analogizing research on the effects of smoking to silly things like studies of storks and babies. And if he was sincere in his views, I can hardly fault him for collecting some money for his efforts.

On the other hand, this document makes me think that Huff may have seen his role as producing talking points in support of a predetermined conclusion. I guess we’ll never know if he really wanted to publish How to Lie with Smoking Statistics. Maybe he intentionally sabotaged it because he sensed it would ruin his reputation, whereas it was possible for him to keep the consulting and testimony under the radar.

15 thoughts on ““How to Lie with Statistics” guy worked for the tobacco industry to mock studies of the risks of smoking statistics

  1. “Maybe he intentionally sabotaged it because he sensed it would ruin his reputation, whereas it was possible for him to keep the consulting and testimony under the radar.”

    But did he return the grant money to the tobacco industry or did he just pocket it? I think failing to produce a product would ruin his reputation as well, although with a smaller community.

    • Maybe Andrew could point back to that being discussed here in 2007?

      Fisher had such strong points, such as reverse dose response (inhaler having higher risks observed than non-inhalers) that he thought Doll and Hill were daft at best.

      When Richard Doll gave a ~2002 Fisher Memorial talk in Oxford he started by saying he was very happy that he was giving the Fisher Memorial talk rather than Fisher giving the Doll Memorial talk.

      He suggested Fisher’s daftness came from the limitation of his skill and insight being to small individual simple studies rather than multiple complex studies. When I suggested that this was inconsistent with Fisher’s early work where multiple repeated studies were often part of the conceptualization (e.g. a good summary being a summary that a later “meta-analysis” would be possible without need of further details – i.e. sufficient.), he replied that “he didn’t think the criticism was beyond reproach, but as someone else had published it, he thought it was OK to repeat it.” (Apparently the hurt feelings had not healed.)

    • 1) Unlike (say climate issues, where physics (with strong conservation laws, orbital predictability, etc) drives trends, overlaid with noise), smoking-disease linkage analyses are heavily dependent on statistical interpretations of data:
      – where one cannot do the definitive lab experiments on humans,
      – where human variability matters (Some people smoke for decades and get away with it)
      – where the actual biological mechanisms of damage by some of the plethora of chemicals in smoke are still not understood

      (Of course, the mouse experiments are strong, as is the overall collection of evidence, i.e., such as fact that about a large fraction of smokers eventually die of related diseases.)

      2) It is is guaranteed that not all statistical studies will be perfect, hence there is always room for some doubt, and as one of the tobacco guys said “Doubt is our product.” Of course, some of us think statistics is our main tool for quantifying uncertainty, but statistics is done by humans.

      3) Given all this, and given the large number of statisticians, is it plausible that the tobacco industry could recruit some statisticians whose views are helpful to the tobacco industry?

      4) Read Proctor’s book.pp.418-458 “Penetrating the universities”

      a) p.436. “hundreds of statisticians have worked for the industry…”

      b) p.436: Fisher:
      “Rumors swirled after Fisher’s death in 1962 that he had either reversed himself on his deathbed or explained away his truculence as opportunism. David Daube, the Oxford biblical scholar, recalls Fisher telling him shortly before his death that his defense of of tobacco was simply “for the money.”(42) (42 is a private communication, unfortunately, but Proctor is a careful scholar, so I tend to credit this).

      c) Section goes on to name dozens more statisticians from the tobacco archives, plus:

      d) pp.439-441 statisticians who have helped tobacco companies by testifying in court: 29 names, of whom some will be familiar.
      One would of course have to go look at the court cases and assess the arguments to make any judgments.

      e) pp.442-443 describe testimonies strikingly akin to those of Wegman of the form “I’m a statistician, not an X.”

      f) p.444 raises question of ethics, the ASA, etc, also lists examples from 26 other disciplines, not just statistics.

      5) Then pp.459-481, “Historians Join the Conspiracy” discusses efforts by members of Proctor’s own discipline.

      • Thanks for this.

        But “42 is a private communication, unfortunately, but Proctor is a careful scholar” is further confounded by Fisher’s habits in being open and candid about what he got wrong.

        Admitting to being daft might have been far worse than just greedy and immoral ;-)

        • Dr. Grinsted:
          From Golden Holocaust…, pp.439-441, here the names and affiliations. Prcctor listed known court cases, I’ve just summarized that with count and range of years.

          But first, note that the “Enriching statistics” section is pp.436-443, within the 40-page “Penetrating the Universities” Chapter within a 737-page book, i.e., this list is a small topic within a complex context. For instance, Proctor recounts various stories in which people’s institutions and/or colleagues had no idea of the consulting work going on.

          Anyone interested in this topic might:

          a) Read the book, which is clearly destined to be one of the definitive works in this turf. Tobacco companies have already started trying to keep this from ever being referenced in court cases.

          b) For verification, try checking a few things in the tobacco archives, formally LTDL – Legacy Tobacco Documents Library. One does have to be careful not to over-interpret the hit counts. Some people are mentioned often, not because tobacco likes them, but for the reverse, and not all people are issued unique names at birth, i.e., one must actually look at some of the documents. For example:
          huff 9,973 documents … of which many are irrelevant
          darrell huff 668
          “darrell huff” 336
          “mr huff” 108, or which some are not Darrell Huff
          IMPORTANT: LTDL gives great behind-the-scenes information, but it more-or-less stops after ~2002.
          The court case info is useful because it is public and continues later.

          c) In any case, given the caveats against over-interpreting, here is the list:

          Edwin Luther Bradley Jr., U of Alabama, 7 (1997-2005)
          Richard C. Clelland, Penn, 1 (1987)
          Eugene P. Ericksen, Temple, 3 (1997-2000)
          Jairus D. Flora, Midwest Research Institute, 2 (2009)
          Bernard G. Greenberg, U North Carolina 1 (1964)
          R. Garrison Harvey, Wecker Associates, 1 (1999) + others
          Maxwell W. Layard, U California, 1 (1990)
          Leo Katz, Michigan State, 1 (1969)
          Lynn R. Lamotte, LSU, 2 (1997-1998)
          Paul Levy, U of I, Chicago Circle 1 (1998)
          Brian P. McCall, U of Minnesota 1(1998)
          James T. McClave, Infotech 5 (1997-2009)
          Daniel L. McGee, Florida State 1 (2010)
          Kenneth Duncan MacRae, U of Surrey 3 (1986-1999)
          M. Laurentius Marais, Wecker Associates 2 (1998-2002)
          Ronald G. Marks, U of Florida 2 (1997-2001)
          Nancy Mathiowetz, U of MD 3 (1998-2001)
          Irwin Miller, Wesleyan U 2 (1969-1987)
          Jacqueline Oler, Drexel U 1 (1996)
          Donald B. Rubin, Harvsrd 10 (1997-2005) see this for discussion.
          Herbert Solomon, Stanford, 2 (1987-1992) {Proctor is @ Stanford, and he mentions other Stanford faculty as well)
          Brice M. Stone, Metrica 2 (1997)
          Larry Tonn, Tonn & Associates 2 (1997-1998)
          Richard Tweedie, Bond U 1 (1990)
          William E. Wecker, Wecker Associates 14 (1997-2005)
          Finis R. Welch, Texas A&M 1 (1997)
          James T. Wittes, Statistics Collaborative, Inc 1 (2005)
          George H. Worm, Clemson U 1 (1997)
          Arnold Zellner, U of Chicago 1 (1997)

          Of 29, 3 have female given names, Lynn might be ambiguous.
          I don’t know the demographics of statisticians well enough to know if the 25-26 of 29 male percentage is interesting or not.

          The high runners are 1) Wecker and 2) Rubin (both of whom Proctor discusses in more detail), plus excerpt of interesting testimony by LaMotte (which is analogous to some elements of Ed Wegman’s Hosue testimony in 2006).

  2. I am a big fan of Huff’s book so I love this post… never knew he did all these other things. A Huff apologist would say (a) this happened roughly around the time the smoking-cancer link was settled so what looked today like scientific misconduct might not have been so foolish in the 1960s; (b) as corporations become the main funders of scientific research, it seems like this may be the norm today rather than the exception.

    • Kaiser:

      Also, you have to remember that Huff was a journalist, not a statistician. I don’t think had any particular expertise in evaluating statistical claims—this isn’t Fred Mosteller we’re talking about. What Huff was good at was telling colorful stories and making fun of people, and that’s what he did for the cigarette companies.

      • As you said, Huff was not a statistician and he clearly did not understand frequentist principles of Neyman-Pearson testing: “there is always the 5 percent chance that no difference at all exists in the direction indicated.” (pages 3-4). That’s just wrong. That said, he makes a very solid point about the Surgeon General’s report over-interpretting the *point estimate* of a mortality ratio of 1.20.

        On the other hand, Fisher’s arguments were sound: see http://www.york.ac.uk/depts/maths/histstat/smoking.htm. It’s easy now to criticize Fisher for his stance because the smoking/lung cancer link is universally accepted. But as Robert Hooke said in his fabulous book, How to Tell the Liars from the Statisticians (now we’ll probably find out that he was in Tobacco’s pocket, too), “the former group [the epidemiologists] turned out to be right, but only because bad logic does not necessarily lead to wrong conclusions.”

        • Hooke: I don’t see any evidence that Hooke was paid by tobacco folks.

          They did like to quote Hooke (and there’s a *different* Robert Hooke who shows up often), but unlike many, where well-crafted searches find numerous documents, there’s no trace of that I could find for him. While absence of evidence is not evidence of absence, the archives have frequent interactions with, for example “darrell huff” or “ronald fisher,” not just quotes.

  3. Pingback: Huff the Magic Dragon « Statistical Modeling, Causal Inference, and Social Science

Comments are closed.