Between 1964 and 1991, per capita cigarette consumption among adults (aged 15 and older) in Australia fell by 33.3 percent, from 2,740 grams to 1,827.1 For 13 years of this period (1964–76), consumption remained virtually unchanged. However from 1977, the year following a national ban on direct forms of radio and television cigarette advertising, the average annual fall has been 2.2 percent. And since 1983, the year when the first of a series of large-scale mass media campaigns commenced,2 the fall increased still further to a mean 2.5 percent per year.

How do we explain this internationally acclaimed success story in modern public health? Is it reasonable to nominate, as implied above, particular policy or intervention landmarks such as advertising bans or large government campaigns as being more explanatory of the change than other less prominent and tangible but perhaps more pervasive influences barely deserving to be described as “interventions”? How can decisions be made about the success or failure of particular programs and policies intended to further this trend? What limitations are there on the ability of quantitative research methods to address these questions?

This paper explores these questions through consideration of some awkward problems raised by a case study. I argue that the ambition to attribute specific preventive or cessation effects to particular tobacco control interventions is highly problematic in any situation characterised by the interplay of continuous, uncontrolled, unmeasured and sometimes unmeasurable variables intended to influence consumption. Such an ambition reflects a reductionist epistemology that is largely incapable of illuminating the complex nature of how it is that individuals either fail to take up smoking, quit, or reduce their consumption – when aggregated, the three determinants of falling per capita tobacco consumption. I will argue that attempts at attributing causal effects to specific interventions in situations characterised by multiple interventions and influences are invariably fraught with highly questionable assumptions that serve the interests of those requiring simplistic, quantified explanations of what are in fact highly complex phenomena. These explanations have more to do with the contexts in which attempts at explanation take place, and with the politics of program funding, than they do with any dedication to a full account of the changing phenomenon of smoking in cultures such as Australia.

​A day in the life of an Australian smoker

Consider a recent day in the life of an Australian smoker, John. As he wakes, John listens to a news item concerning a government decision to end all remaining forms of tobacco advertising.3 Since the mid-1970s, he has woken to many similar announcements concerning various forms of tobacco advertising. These have occasioned discussions at his office and in social gatherings, where smoking has become a common and sometimes highly charged topic of conversation. Newspapers have also been thick with news about smoking. In 1988 for example, he might have read up to 1,600 separate items in newspapers alone, of which only 17 percent would have delivered even a vaguely comforting message.4

As John smokes at the breakfast table, his two children playfully chorus their usual anti-smoking slogans: “Smokers suck! But we get half the muck!” and “Kiss a non-smoker . . . taste the difference.” It seem there have been dozens of these taunts over the years. Undoubtedly they have picked them up from school, where he knows that they are regularly given lessons about the health consequences of smoking.5

His wife, who doesn’t smoke, has, like a lot of people, slowly turned into someone who actively dislikes smoking. She has recently begun seriously talking to him about whether he might go outside when he wants to smoke. In making this request, it seems to John that she is not really being overzealous.

On the way to the train station, he stops to buy a new pack. When he proffers $4.55 for his usual pack of 30s, the shop assistant reminds him that they have gone up by 20 cents a pack following the latest federal budget. John has calculated that by smoking a pack a day, he is spending $1,734 a year on cigarettes – the price of a ten-day holiday in a luxury hotel in Bali.

Boarding the train, he ponders that here is yet another place he can’t smoke. Public transport went smoke-free in 1976, joined in 1990 by all domestic flights in Australia, and in 1992 by a ban on smoking on an increasing number of some international routes and even a total ban inside Australian air terminals.

As he reads the morning newspaper, he notices how many of the “share accommodation” classified advertisements specify that only non-smokers need apply. Of 335 advertisements that day, 42 percent include this requirement6 – a higher rate than any other quality sought by advertisers. He also notices that every government job advertisement, and a not inconsiderable number of private sector ads, states that “a smoke-free workplace is company policy”. Browsing through the lonely-hearts advertisements he is again struck by how many of these people seek a partner who is a non-smoker. And then he is confronted by a full-page advertisement from a life-insurance company offering substantially reduced rates for non-smokers.7

Arriving at work, John stubs out what will be his last cigarette until lunchtime. In 1988, his office introduced a total smoking ban. Since then, there has been a virtual stampede throughout the Australian business world to do the same. By 1992, 58 percent of the top 139 companies in one state had total bans, with over half of these having introduced them in the previous 12 months.8 A successful civil suit by a worker whose asthma was severely aggravated by passive smoking in her workplace seems certain to hasten this process.9 The ban at work has certainly forced down John’s own consumption. It has been estimated that the average smoking office-worker has reduced daily consumption by around 25 percent10 and that the workplace bans will cost the Australian tobacco industry $73 million in lost sales a year.11

At lunchtime, John goes with some colleagues to a nearby Pizza Hut restaurant. The entire chain has recently gone smoke-free, reflecting an overwhelming community demand for smoke-free dining.12 He then passes a street sign warning him that he could be fined for discarding his cigarette butt in the street13 – the non-biodegradability of butts makes them a major pollution problem, especially in a city where stormwater runs into the picturesque harbour around which the city is built. Being environmentally conscious, he feels awkward about his usual throw-away method of disposing of butts.

Home that evening, John relaxes in front of TV, where on the news he hears a report linking smoking with yet another dreaded disease – leukaemia.14 “Is there anything that smoking doesn’t cause?” he thinks to himself, reflecting on all the news reports he has heard about the subject over the years. Being a sports fan, he zaps his TV between channels showing the national soccer and basketball competitions. And there it is again: anti-smoking sponsorship messages on the sidelines and even on the players’ clothing. And then to put the icing on the cake, a gory government advertisement showing how much black tar a smoker will inhale in a year is shown several times during commercial breaks.

The next day, John decides that he will finally quit. Over the next 12 months, he makes three or four unsuccessful attempts,15 one inspired by a brief warning given to him by his doctor, and another being a period when he uses over-the-counter nicotine gum after prompting from his pharmacist. Eighteen months after his initial decision, he smokes what will be his last cigarette. In doing so, he joins approximately 3.8 million Australian adults who identify themselves as former smokers.16

Shortly after he finally stops smoking, he is phoned by a researcher working on the evaluation of a government media campaign to encourage people to quit smoking. John joins those respondents who say that they have seen the campaign, who strongly agree that it made them think about quitting, and who respond (unprompted) that “health reasons”, “social unacceptability” and “cost” are the three main reasons they have stopped smoking.17

The researchers subsequently write a scientific article in which they claim that their state-wide media campaign is probably the factor responsible for the state’s higher than average quit-rate. This claim is based on extrapolations made from the aggregated sample of recent quitters like John.

​Discussion

How do we explain John’s decision to quit? What do we make of a community cessation rate extrapolated from data including John’s responses and its partial attribution to the government campaign? And what should we make of John’s own account of why he quit? In the evaluative research literature of tobacco control, such questions are seldom asked and even more seldom thoroughly pursued. Where questions of attribution are assessed, it is usual that the influence of a particular variable such as advertising or a health education campaign is examined using standard pre-post or intervention-control group designs. Occasionally, a limited number of potential confounders, like price changes, are incorporated into such studies. Control areas are seldom if ever matched with intervention areas for anything remotely like the range of variables described in the above scenario. Essentially qualitative variables such as tobacco advertising are conveniently homogenised into measurable units such as cost, as if all advertising campaigns could be considered of equal impact.18 Such an assumption would be news indeed to many in the advertising industry, who know too well how many of their efforts seem to make little difference to brand sales.

Yet from the foregoing scenario, it is obvious that in the life of every smoker, there has been a plethora of interventions, campaigns and “influences” to which they have been exposed over many years. From John’s perspective, these have not passed in front of him in any neat, sequential order, or in any way that would allow him to reliably quantify their respective influence on his gradually changing perception of his own smoking and the evolution of his decision to quit. Indeed, at many times in John’s recent smoking history, it would have been quite impossible to isolate and quantify the effects of any one of up to a dozen concurrent variables. Quantitative evaluative research processes avoid the methodological imbroglios that are inherent in accepting the reality of the dynamic interplay of the sort of factors described in the scenario above. There are at least four outstanding explanations for this.

​1. Reductionist epistemology

Evaluative research in tobacco control is located almost entirely within the scientific tradition. This tradition assumes a reductionist epistemology whereby the task of science is to discover and quantify the exact relationship between variables. Any difficulties in assessing these relationships are assumed to lie with the imprecision of the methods used to assess them, and not with the very conception of the nature of how it is that a complex behaviour like smoking changes throughout a population or an individual’s lifetime. The ambition to exactly quantify the assumed relationship is seen as a task worthy of pursuit, whereas consideration of the gestalt of how various cultural, economic, organisational and educational factors actually combine to influence smoking behaviour is viewed as messy and unscientific. The only manageable truths in this tradition are those that are simple and uncomplicated: advertising bans and price rises reduce aggregate demand, education programs decrease the incidence of uptake, and so on. The messy gestalt is entangled in the explanatory gossamer of a myriad of experiences, conversations, memories and exposures to interventions, but researchers bearing reductionist precepts and methods wear the equivalent of boxing gloves in their attempt to unravel these delicate threads.

​2. The explanatory privileging of recent factors

It is not just single factors, but also recent factors that are privileged by reductionist explanations. The view seems to be that the effects of recent interventions and policy changes could be expected to be less confounded by the intrusion of other influences than policies and events enacted further in the past. This assumption is fuelled by attributions often given by individuals when they nominate specific events (e.g. recent illness or symptoms, the death of a relative, an intense period of haranguing from their children, a straw-that-broke-the-camel’s-back price-rise) as “why” they quit. Such explanations may well represent accurate and heartfelt perceptions of the precipitating factors that prompted quitting, but reveal little of the complex historical precursors that may well have been necessary to predispose individuals to quit when finally subject to the precipitating event. For example, it may be the case that demand sensitivity to price rises is dependent on a widespread acceptance of the tobacco–disease nexus. Respect for the importance of such plausible predisposing factors is rare in evaluative studies about smoking control and was not raised in a recent expert consensus report on research priorities in tobacco pricing and taxation.19

​3. Concern for policy tractable factors

In recent years, concern has been increasingly expressed that research should concentrate on better understanding how to influence so-called policy tractable factors that influence smoking. These are factors that are amenable to manipulation by government policies and include price, advertising, packaging, laws and regulations on smoking in public places, and school curricula. They stand in contrast to factors also said to be relevant to smoking that include age, sex and cultural proscriptions on smoking, social class, occupation, income, school performance, and smoking by parents, peers, siblings and workmates. None of these is as directly or even at all amenable to influence through government policy.

Pragmatic considerations of “What can we directly influence?”, schooled from an “upstream” preventive analysis,20 have directed research attention onto the role of precise factors such as price or large public information campaigns. Again, the problems arising from the reductionism involved here tend to be overlooked in the fervour to produce “action-oriented” research that can be fed into policy and political processes.

​4. Relationship of evaluation to funding

Health promotion campaigns that involve relatively large sums of money are generally subject to intensive scrutiny bred from the competitive funding climates in which they operate. Unlike “passive” preventive strategies like price controls and advertising restrictions, which require little or no money to implement, health promotion campaigns are continually called upon to justify their allocation of funds. Evaluation of the “effects” of funded health-education campaigns against smoking are thus partly inspired by a concern to be able to show that an intervention is effective or, better, cost-effective. Such considerations produce a highly selective orientation to evaluation driven by a priori concerns to assess interventions deemed worthy of evaluation, rather than an attitude toward explanation of the quitting process that is open to the possibility of a thoroughly “messy” account like the scenario above.

In many cases, these interventions have been organised, developed and run by the very people who either conduct or commission their evaluation. Often these people are employed on “soft” project funding which may cease should a political or administrative perception develop that the interventions “don’t work”. Such considerations raise more tangible concerns about the impartiality of the explanatory process.

​Confounding run amok

Individual platforms of comprehensive tobacco control policy are seldom implemented by governments in isolation from others. Exceptions like Sudan, which banned tobacco advertising but has done virtually nothing else, simply prove the rule: that when a government is committed enough to introduce (say) bans on smoking on aircraft, it will have done this in a spirit of wanting to reduce the burden of death and illness caused by tobacco and accordingly will be predisposed to introduce other policies with similar intent. In practice this has meant that nearly all countries where evaluation studies of tobacco control policies and programs have been undertaken have been characterised by the coalescence of a multitude of these factors, much in the manner described in the scenario.

Many of these factors will be introduced in an ad hoc, opportunistic way rather in any way remotely analogous to the timed and controlled drip-feeding of therapeutics in laboratory or clinical trials. Politicians and tobacco control advocates understandably have little to no regard for the violation of the sanctity of control groups, areas or periods so coveted by researchers hoping to conduct a neat study unconfounded by unexpected influences. Instead, they will have their noses constantly in the political wind for opportunities to engage in media advocacy, to lobby for price rises and further restrictions on advertising and so on. In large countries like the USA and Canada, where federal, state, provincial and local governments have jurisdiction over different elements of tobacco-control policy, it is often the case that at any given time, quite complex different configurations of tobacco-control activity will be being played out in different parts of the country. Some of these events will be newsworthy and picked up by national media networks, which will amplify a local issue into a national concern, thus further corrupting pristine research designs. Most evaluative studies simply pretend that all this does not occur and that the independent variables (policies and interventions) they are evaluating constitute the only players in the landscape.

​Conclusions

What does this analysis suggest for the future of evaluation of tobacco-control policies and programs? The sort of methodological problems discussed above should not induce an evaluative paralysis in tobacco control researchers. They should not inspire any abandonment of the evaluation of outcomes in tobacco control nor any shying away from the challenges of the attribution problem. Continuing debate about ways of sampling and controlling for differing “micro climates” of influence and intervention between areas, states and nations will be very welcome. As well, though, a more open recognition of the limitations of reductionist thinking in considering the causes of declining tobacco-use throughout populations could redirect researchers into considering the potential of qualitative methods as important adjuncts in the explanatory process.

Mark Twain wrote that if your only tool is a hammer, then all your problems come to look like nails. And so it has largely been with the dominant explanatory paradigms in smoking-control research. Social scientists have long argued for multiple methods or triangulation in studying complex human phenomena.21 Triangulated research can involve the use of different investigators, theories and methods to study the one phenomenon with the assumption that the weaknesses in each single method will be compensated for by the counter-balancing strengths of another. This is not to argue that triangulation can ever produce a single “true” reality beyond the frameworks and interpretations provided by each research approach.22 Data collection methods and interpretive approaches drawn from ethnomethodology,23 oral history and discourse analysis24 hold promise as ways of rendering complex social processes like the natural history of smoking cessation more transparent. The products of such parallel research would doubtless be examples of what the anthropologist Clifford Geertz has called “thick description”.25 They would also frequently be culturally and historically specific: understanding the process and motivations of smoking cessation among septuagenarian men in cardiac wards in Cairo will throw up radically different insights than those provided by research into why Australian teenage girls in the age of Madonna are smoking more than their male counterparts.26 Such hermeneutic characteristics would doubtless perplex and frustrate some number-bound readers hoping for simple, replicable truths and axioms about successful tobacco control. But just as it would be facile to attempt to describe the meaning of Da Vinci’s Mona Lisa only in terms of paint and brush strokes, so too are tunnel-visioned truths tethered to the short leashes of quantifiable explanatory variables equally as frustrating.

The degree of analytic sophistication possessed by most politicians and funding bureaucrats will rarely require any venturing into the complexities of the attribution problem. Such people invariably want two-paragraph answers to questions like “Do these school programs work?” or “Will banning advertising reduce demand?” They are slaves to entrenched, simplified decision-making processes that conspire against answers predicated on any honest admission of the highly intertwined nature of the relationships involved.

International tobacco control agencies and expert groups, in their wisdom, have long called for comprehensive policies to turn the public tide against tobacco.27 They have also been dismissive of efforts by the tobacco industry to attribute population-wide trends in tobacco consumption to the presence or absence of single variables (e.g. the industry’s frequent insistence that the absence of tobacco advertising and the high smoking rates in the former Soviet Union proved that advertising bans did not reduce demand).28 The rationale for comprehensive policies lies not in any belief that the individual platforms of such policy (advertising bans, price increases, clean indoor air policy, mandated school health education, strong health warnings and so on) simply have incremental, additive effects on demand. Rather, it lies in the recognition that each of these platforms are nurtured by the others, creating a synergism which produces slides in demands apparent in countries including Canada, New Zealand and Singapore. Probing the dynamics of this synergism using the wider range of research and analytic methods proposed above is long overdue in the professional literature of tobacco control. Tobacco Control, the BMJ’s new specialist journal in this area, will welcome reports and analysis that attempt to do this.