Tuesday, 28 May 2013

Day 149: The rational decision to smoke

A Winnipeg-born political scientist who has spent much of his professional life examining how voters apply their economic self-interest to the ballot box appeared for the second day at the Montreal Tobacco trials.

Raymond Duch's task at this trial was to present, on behalf of two of the defendant tobacco companies,  his opinion on "the awareness of the Quebec (and Canada) population from 1950 to 1998 of the health risks associated with smoking and of the public's view that smoking can be difficult to quit." 

Yesterday he presented his 180 page report (Exhibit 40062.1) under friendly questions from JTI-Macdonald's lawyer, Doug Mitchell. On the basis of about 35 studies selected from about 300 he considered, Mr. Duch concluded that public awareness about reports linking smoking to disease were "exceptionally high" from the early 1960s, that public beliefs that smoking was harmful became equally high within a few years and that as long as there has been polling on the subject, Quebecers have known that smoking was habit forming.

Today Mr. Duch took his analysis one step further, and introduced a new concept to this trial - rational choice theory as applied to smoking. (A familiar defence framing in U.S. tobacco trials).

High levels of belief that tobacco is harmful means that "the smoking decision is not any less rational than voters' decisions  – they are informed decisions."  

Increasing the numerator makes the numbers fit

Mr. Bourque interpreted these results as
indicating 6 in 10 smokers thought cigarettes
were dangerous, but Mr. Duch said it
meant that 91% were.
Mr. Duch is the second polling specialist to testify as an expert in this trial, and he spared little effort in slagging the efforts of the first, Mr. Christian Bourque of Léger Marketing.

(Mr. Bourque had reviewed results from the private CMA surveys conducted by Imperial Tobacco and produced a much lower estimate of the level of belief that smoking was harmful.)

Today Mr. Duch offered his reinterpretation of Mr. Bourques results to show that they were not so very different from his own. He suggested that all smokers who shared any belief that cigarette smoking was harmful should be grouped together as those who believe and not, as Mr. Bourque had done, to group together all those who thought that some level of smoking was not dangerous as those who don't believe.

Similarly, Mr. Duch felt that smokers' responses to CMA questions about how many cigarettes one could smoke could be reinterpreted to support his conclusions that health hazards were well known.

The cross examination

All but an hour of today was spent in the cross-examination of Mr. Duch -- first by Mr. André Lespérance and later by Mr. Bruce Johnston.

Mr. Lespérance's cross-examination style looks nothing like the TV shows. He has a very gentle affect, and a very calm manner. Even with his arms crossed and staring intently at the witness, as he was today, his voice maintains a kind tone and his questions are phrased in a positive way.

So it was somewhat surprising to see Mr. Duch respond with anger to questions so politely presented and whose content must surely have been expected (and rehearsed) by the ever well-prepared Doug Mitchell. This is the first expert witness that Justice Riordan has had to counsel to "calm down a minute and take a breath."

A view from above

There is a fine balance between a display of expertise that makes a person sound learned, and one that makes them seem unreliably out of touch. Mr. Lespérance successfully invited the witness to cross that line.

With more than a whiff of Faculty Club invective, Mr. Duch had made it clear that he found Mr. Bourque's work did not pass muster. "A classic example of spurious regression ..." "Mr. Bourque is being extremely misleading..." "A careful analyst would simply not do this."  

Mr. Lespérance returned to a few of the areas in Mr. Bourque's report that had drawn criticism. In doing so, he was able to show that the criticisms were misplaced, or blown out of proportion.

One of these was Mr. Bourque choice to replicate a 20-year chart from an Imperial Tobacco report (Exhibit 63) to illustrate the company's own knowledge of its CMA data instead of drawing from the semi-annual survey results.

Mr. Duch criticized the use of
secondary data in Mr. Bourque's report
as an "incomplete analysis"
Mr. Duch said this was not good enough, and contrasted it with his own higher standards.  "There were 34 studies -- not only the [20] identified here. I went back and got every one of the 34 studies -- Why didn't he get all 34 studies?!" ... "I would argue that for example that it is an incomplete analysis of the data that was available." 

Even when Mr. Lespérance forced the witness to agree that all of the years covered by the CMA were covered by this table, and that it had indeed produced the same trend line as his own analysis of public polling data Mr. Duch did not seem to see any wisdom in climbing down from ivory tower. ("I am simply saying that if I were evaluating ..."if I were reviewing this article for a journal...")

"But you do the same exercise and come to the same conclusion," Mr. Lespérance reminded him. "Yes," he admitted "but I conclude much more precisely!"  

Missing studies, Missing data

Most of the studies cited by Mr. Duch in his report were multiple choice (closed ended) questions put to Canadians about whether they had heard that tobacco was linked to lung cancer or whether they believed that cigarette smoking was a health hazard.

Missing from his report were references to any studies that asked smokers to say without prompting (open ended) what they thought the health consequences of smoking were, or to qualify their perceptions of the risk by comparing these with other aspects of daily life.

Some of those studies were among the JTI-Macdonald documents that were shared with the plaintiffs but not, apparently, with their own witness even though he testified he had asked for all relevant material.

One was a 1990 Environics survey of Ontarians, which found that although lung cancer was the most frequently identified health hazard from smoking, it was identified by only 44% of smokers. Only 20% mentioned emphysema or heart disease. (Exhibit 1547.1 and 1547.2) (Mr. Duch said that unprompted questions were "Absolutely not!" a more accurate measure of knowledge.)

Another was a Gallup study commissioned by the Canadian Cancer Society which can be found on the court record of the constitutional challenges to Canada's tobacco laws (in which Mr. Mitchell and his law firm were central players). It shows that as late as 1988, most Canadians though that traffic accidents were responsible for most preventable deaths in Canada. (Exhibit 1548)

Even within the studies which Mr. Duch used for his report there were questions relevant to health beliefs that were not included in his findings.

These included part of an Environics study (Exhibit 40064.24) that compared smoking with other risks but found "Adult Canadians are quite divided and unsure as to whether or not 'smoking is no worse than a lot of other things around us, like chemicals in food and car exhaust fumes.” 

Mr. Duch again appealed to his high methodological standards as a reason for rejecting this data that did not seem to fit his conclusions. "The question wording is inconsistent with all other questions I used."  Besides, he added  "this is a leading question - a badly worded question."

He had similarly not included another response in a study he cited (Exhibit 40064.6), one which asked Canadians to agree to questions like "Air pollution causes more lung disease than cigarette smoking” (70% said it did) or “some cigarettes are safe to smoke” (84% said they were).

Another question he overlooked was in a Goldfarb study he reviewed (Exhibit 40064.51), which asked whether people thought smokers died earlier. (Only 47% said this was true, and 31% said it was false). Again, Mr. Duch appealed to methodological concerns. "I am reluctant to accept that this particular question in and of itself should be the basis for me concluding something about peoples' concerns about smoking and death – only because it is one question." 

Even a study done for the CTMC (by Goldfarb in 1990) which reported on seemingly low levels of unprompted knowledge of health effects drew severe criticisms from Mr. Duch. "People can't orient themselves" with open ended questions, he said. "This does not conform to the conventional format for asking the question." .. They're wrong.” "This is someone who doesn’t know how to analyze data." 

Unanswered questions

By mid afternoon, a pattern of responses by the witness had set in. Survey questions that did not match the binary believe/don't believe pattern of his trend line were discarded. The additional information they might provide was not reliable, or he was not prepared to comment without further study.

He would not engage on many of the issues related to people's knowledge of health harms - like the storied "controversy", the engagement of third party messengers to overcome credibility issues, or the segmentation of the cigarette market according to health concerns.

Mr. Duch had said yesterday that the reason he agreed to undertake this research was his interest in people making decisions that "at least on the face of it, seemed to be irrational." Mr. Johnston wanted to know whether addiction had featured into his "conceptual methodology" for smoking, but Mr. Duch would not go beyond saying that he had found that people knew about addiction.

At the end of the day, Mr. Johnston asked Mr. Duch to reflect on a response from employees at Gallup to the use of their data by defence witnesses in U.S. tobacco litigation. (Exhibit 1239) The pollsters' concerns included the industry's selective use of data, and its view that responses to polling questions about "harm" could be consistently interpreted over a long span of years during which the public understanding of harms was changing.

Mr. Duch seemed unimpressed by their critique. Their paper was "presented in a conference and never peer reviewed or published," he sniffed. "Its authority is limited." 

Information short-cuts

The last question put to the witness came from his own lawyer (the 're-direct'), who invited him to explain why public misconception about the relative death toll from tobacco accidents did not affect his conclusion.

Mr. Duch explained people make decisions about a variety of things in life "using information short cuts. ... we know that people do not engage in exhaustive or complete collection of information."

"People for example do not know all of the information on toxicity of cigarettes in making a decision on smoking or not smoking. The notion that people sit down and assess the specific risks of automobiles or getting in a plane accident or assessing the precise probability of dying from smoking is absurd."

"Once people know that smoking leads to lung cancer – adding specific bits of information is not going to change their decision making calculus."

The trial does not sit next week. When the trial resumes on June 10th, the tobacco companies will call their first witness from Health Canada, Mr. Denis Choinière. The following week, the former Minister of Health, Mr. Marc Lalonde, is expected to testify.