Monday, 14 April 2014

Day 224: James Heckman: the dismal scientist

By most standards, Mr. James Heckman would have to be considered a star witness for the two defendant tobacco companies who hired him to testify on their behalf at the Montreal tobacco trials.

As economists go, this man is no slouch.

Winner of the Nobel Memorial Prize in Economic Sciences. Distinguished service professor at the University of Chicago. Recipient of a dozen or more honorary doctorates of law or honorary professorships. It's hard for an economist to be more distinguished than that!

And it's a distinction he seems willing to share with asbestos, tobacco and other companies who have found themselves on the defensive in lawsuits.

The 69-year old Mr. Heckman has testified for asbestos companies and has assisted the defence of Philip Morris in a few important large tobacco trials in the United States, (Falise 2000Blue Cross 2001, Department of Justice, 2005). In regulatory issues, also, he has produced opinions to help forestall plain packaging of cigarettes. (United Kingdom, 2012)

And yet, despite this 14-year relationship with tobacco clients (and a billing rate he revealed today to be US$2,300 an hour!) Mr. Heckman says that the industry has never paid for any of his research.

He positively bristled at the suggestion in this year's Surgeon General's Report that he was a "consultant researcher" to the tobacco industry.

(I will leave to others to assess his written direct testimony at the DOJ trial  and the subsequent paper he published on the topic same topic  - Exhibit 21320.5 - against the Surgeon General's conclusion. In making that assesment, one might want to consider that the co-authors of that paper - Mr. Frederick Flyer and Ms. Colleen Loughlin - were also in the courtroom this afternoon. The three are senior mounts in Compass Lexecon's stable of professionals "with testifying experience.")

The Expert Economist and the Straw Man 

Even speaking at a fast clip, and lingering only briefly on his accomplishments, Ms. Glendinning took well over an hour to allow Mr. Heckman to present his background before she asked Justice Riordan to qualify him as an expert in "economics, econometrics and determinants of causality." 

It was then the turn for plaintiff lawyer, Bruce Johnston, to point to reasons why the judge might want to doubt the value of Mr. Heckman's contribution. He immediately honed in on the way that Mr. Heckman's report (Exhibit 21320.1) was aimed at evaluating whether "[plaintiff expert witness on marketing] Dr. Richard Pollay has provided a reliable basis for concluding that a causal relationship exists between tobacco company advertising and aggregate smoking."

The problem, as Mr. Johnston saw it, was that such a topic had not been addressed in Mr. Pollay's report -- a fact that had been confirmed when Mr. Pollay testified last January.

Was this not an example of Mr. Heckman using a Straw Man to misrepresent an argument he was going to rebut?

Justice Riordan looked at the ceiling as the two men tangled over the meaning of straw-man. Mr. Heckman denied that his report was intended to suggest that Mr. Pollay had said there was a causal relationship. "I did not say he did. I said I was asked whether he had provided a reliable basis. I was asked to talk about the issue." 

Not long after, however, Mr. Heckman conceded that Mr. Pollay had never addressed causality. "I completely agree with you." He said his report was not directly linked to Mr. Pollay's except "to the extent that he talks about advertising and I talk about advertising."  

Like a professional wrestling match: a foregone call

Many months ago, another expert witness had been accused of going further in rebuttal than the reports he was challenging. On that occasion, the plaintiffs withdrew much of Robert Proctor's report, in anticipation of having sections struck down by the judge.

For a short while in an otherwise tedious day, we were on the edge of our seats. Would Justice Riordan be asked to apply the same sauce to the gander?  If so, would Mr. Heckman be finished by lunch?

The realpolitik at this stage of the trial demands that there are no more excuses for Appeal Court delays. The result -- no ruling against the defendants -- was thus a foregone conclusion.

So in an I'm-just-going-through-the motions tone, Mr. Johnston objected that that Mr. Heckman's testimony was irrelevant, as it did not respond to any evidence from the plaintiffs. And with a lets-get-this-over-with air, Justice Riordan said that he saw "enough of a connection, and would allow the report to come in."

Once over lightly 

Given Mr. Heckman's stature, one might have expected him to have been given a lengthy exposition, and the opportunity to drill down on some of his findings, as others have. (His main conclusions are appended at the end of this post).

But Ms. Glendinning seemed surprisingly anxious to not let her witness linger too long in front of Justice Riordan. She frequently cut short his answers, or otherwise tried to rush him through her set list of questions. The day adjourned with a full hour left on the clock, during which time her witness would have had a captive audience. So much left unexplained!

The brevity came with a mixture of the overly simple and the overly simplified.

In the morning, Mr. Heckman was asked to explain ideas that by this point in the trial are more than well understood, and which seemed almost insultingly basic. Justice Riordan, who received his own Masters in Economics only four years after James Heckman, certainly does not need to have the Scientific Method nor the role of the OECD explained to him!

James Heckman explains the Scientific Method
to Justice Brian Riordan - Exhibit 21321
In the afternoon, by contrast, the thinking behind Mr. Heckman's conclusions was rapidly skimmed over, where more detailed explanations would have been expected.

A dense table of calculated results was shown, for example, while the witness was merely asked to state his conclusions that there was no real relationship between advertising and smoking prevalence, or between the marketing of "light" cigarettes and the rate at which smokers quit.

I was reminded of the rapid display of the empty magician's sleeve before the dove was produced.

Endogeneity

One of Mr. Heckman's drive-home messages was unmeasured phenomena can lead to mistaken conclusions.

He used the colourful if somewhat dated/distasteful analogy of trying to answer the research question "Does reading Playboy lead to reduced homosexuality?"  One had to consider endogeneity, he said -- or the likelihood that those who read Playboy might already have a predisposition towards heterosexuality.

Endogeneity also played a role when society adopts measures like bans on tobacco advertising. This happens when social attitudes towards smoking were unfavourable and is thus also connected with fewer people smoking.

Smoking rates in Canada were already declining when health measures were adopted he said -- likely the results of an unmeasured social attitude, and not the measures themselves.

James Heckman's view of events in tobacco control
in Canada as tied to smoking rates. 
He applied the same concept at the individual level.

Endogeneity is a factor when considering the child who wears a T-shirt printed with Joe Camel, or who collects a tobacco-branded ashtray. This young person likely has a predisposition towards smoking -- one caused by factors other than industry behaviour.

"There is very little evidence that advertising has an effect – but the price of cigarettes, or whether or not parents are smokers, or the environment and peer effects are major determinants - even the education of the child."  Even genetics could predetermine smoking.

As if to underline that it is defects in the individual that leads to smoking, and not the actions of those who sell cigarettes, he several times characterized smokers as people who adopt "risky behaviours. "Those people who smoke marijuana, who drink, who participate in risky behaviours - they are more likely to also smoke." 

Heterogeneity

As other defence expert witnesses have before him, Mr. Heckman underscored that it was not appropriate to think of smokers as a homogenous class of individuals. 

The differences among people -- heterogeneity -- had to be considered before drawing any conclusions. 
"The 'representative individual' is a fiction. [It] disguises variability in individuals."

Earlier, he had cast smoking in the context of variance of non-cognitive abilities (a major focus of his usual research). It was not measurable brain power that predicted success, he said. Equally or more important were often unmeasured abilities, such as resilience and perseverance. People with these other characteristics did well in a host of ways - including having less chronic disease.

Mr. Heckman seemed to sincerely believe that it was not the way tobacco was provided in the market that led to what he called "the smoking decision". The factors he honed in on were personal strengths, parental background, education and peer values.  

The cross-examination of Mr. Heckman will take place tomorrow. On Wednesday morning, a motion to prevent the plaintiffs from asking for JTI-Macdonald president, Michel Poirier, to testify will be argued. In the afternoon, another industry marketing witness: Mr. David Soberman.


-----------------------------------------------

Some of Mr. Heckman's Conclusions

(Extracts from Exhibit 21320.1, 21320.2)

"[M]y main conclusion is that Dr. Pollay does not provide a reliable basis to support a causal inference between tobacco company advertising and aggregate smoking.

"Dr. Pollay’s analysis is based on a review of tobacco company documents, advertising/marketing theory, and his personal judgment on the impact of various marketing campaigns on consumer behavior. His work does not employ the rigorous analytical methodology necessary to establish the existence of a causal link between tobacco company advertising and the level of smoking in Québec above the level that would have existed absent this advertising. Dr. Pollay’s analysis does not provide reliable empirical support for the conclusion that tobacco company advertising was a causal factor in initiation, quitting or intensity of smoking decisions. As a result, his work does not provide reliable evidence addressing the narrower question of whether tobacco company alleged misconduct caused harm to the class.'

"I also conclude that tobacco companies have economic incentives to advertise, even absent any effect of such advertising on individuals’ decisions to smoke. Namely, I understand that plaintiffs have put forth the argument that a key incentive for tobacco companies to advertise is to increase aggregate cigarette consumption. This conclusion is conjecture without a reliable empirical basis, as firms have multiple economic incentives to advertise."

"My analysis of these Canadian data and other academic research efforts do not support Plaintiffs’ claim that the marketing of light cigarettes reduced cessation rates in Québec."

"I also considered the issue of whether any possible effects of Defendant tobacco companies’ alleged misconduct (including, but not limited to allegations of wrongful lights marketing and marketing to youth) would have differed significantly between class members. In my opinion, substantial heterogeneity among class members in important factors affecting smoking behavior provides support for differential impact. For example, the amount and frequency of smoking light cigarettes and other factors, such as exposure to marketing in Québec varied among class members. In addition, even if the exposure to the alleged misconduct were the same across class members (which it was not), the impact of such exposure would not be uniform across the proposed classes, which constitute a heterogeneous group of individuals who undoubtedly varied substantially with regard to time and place of initiation and continuation of smoking, socioeconomic circumstances, peer influence, genetic makeup and societal conditions such as the extent of public smoking restrictions and predominant attitudes regarding acceptability of smoking. Therefore, there is no reliable basis to conclude that impacts of the alleged misconduct on smoking behavior, even if such
impacts existed, would be common among the class members.