Our results indicate that the notable decline in per capita cigarette consumption observed over the course of California's comprehensive tobacco control programme was mostly from current smokers smoking less. In both periods, 1990–1996 and 1996–2002, this component accounted for a larger share of the decline than that attributed to fewer ever smokers or to smokers quitting. The percentage accounted for by a lower prevalence of ever smokers was the same (~30%) in each period. Finally, in contrast to the 1990–1996 period, when quitting accounted for none of the decline, during 1996–2002, quitting accounted for an appreciable fraction of the per capita consumption decline (22%).
The finding that most of the per capita consumption decline can be attributed to current smokers smoking less is consistent with a recent econometric analysis that attributed changes in consumption between 1996 and 1999 to a decline in consumption among current smokers rather than to a decline in smoking prevalence.24
In our study, reduced consumption was observed in all population subgroups (see Appendix A, table A1). Some of the decline among current smokers may stem from an increasing fraction of current smokers not smoking daily.25
Also, younger cohorts of smokers may never reach the high levels of consumption observed in previous generations. Particularly from 1990–1996, smokers in the youngest age group showed a large decline.
While some smokers may maintain their accustomed nicotine level by smoking fewer cigarettes more intensely (compensation),26
some adjust to a lower nicotine level and therefore become less dependent.27,28
Since less dependent smokers are more likely to quit successfully,29,30,31
interventions that can reduce consumption in the short term may eventually increase quitting in the longer‐term.29,32,33
A longitudinal population study found that smokers who reduced their consumption to < 15 cigarettes/day were more likely to successfully quit later.32
Because quitting is so difficult, it may not be realistic to expect that tobacco control measures (including increased excise taxes) will lead to an immediate drop in smoking prevalence from people quitting.24
Indeed, our results appear to support this; we found a high fraction of the per capita consumption decline was due to reduced consumption among current smokers in the earlier period (1990–1996), with an appreciable fraction from quitting activity only in the second period (1996–2002).
In contrast to smoking cessation, the evidence for improved health from smokers cutting their consumption is not definitive.34,35,36,37,38
One analysis of a large Danish cohort showed no advantage with respect to total mortality, cardiovascular diseases, respiratory diseases or tobacco‐related cancers.36
More recent analyses of this database showed that a pronounced consumption reduction might lower lung cancer risk,37
but there was no effect on myocardial infarction risk.38
The percentage of the per capita decline (about 30% in both periods) attributable to a reduced proportion of ever smokers likely resulted more from a net migration into the state of non‐smokers rather than from fewer people maturing to adulthood as never smokers. A preliminary analysis of total
rather than per capita consumption showed a consumption increase from population growth that was overcome by the reductions due to the other components. The fastest growing segments of the California population are Hispanics and Asians (Appendix A, table A1). Since smoking among California adolescents has declined notably,39
we expect that the ever‐smoking component will account for relatively more of the total decline in future years. From 1990 to 2002, the percentage ever trying a cigarette declined by 70% among 12‐ to 13‐year‐olds, by 53% among 14‐ to 15‐year‐olds, and by 34% among 15‐ to 17‐year‐olds.39
Smoking in the past 30 days among all ages of adolescents (12–17 years) declined 45% over this period.39
A recent decline in young adult smoking was also observed.39
In the second period (1996–2002), quitting among women was responsible for much of the increase in the fraction of the total per capita consumption decline from this source. While other population subgroups also appeared to show per capita consumption declines due to increased quitting (those
65 years of age, non‐Hispanic Whites/African Americans), it is possible that it was the women in these groups that accounted for most of the decline. Results from the evaluation of American Stop Smoking Intervention Study (ASSIST) demonstration project, in which 17 states received National Cancer Institute funding to seed tobacco control efforts, also showed an effect for women but not men in ASSIST states.40
A similar result was also observed in the Minnesota Heart Health Program, another population‐based intervention.41
These population results are in contrast to the widely‐held perception that women have a more difficult time quitting than men.42
Our California results suggest that male smokers in particular appear to be reducing their cigarette consumption rather than quitting.
The use of our decomposition methodology as an evaluation tool is hampered by the times for which data were available. The first California Tobacco Survey (1990) was conducted after the initial excise tax increase ($0.25/pack) that funded California's tobacco control programme took effect in January 1989, so the expected43
decline (fig 1) due to the tax increase was not included in our analyses. Cigarette prices remained relatively stable from 1990–1996, with a small industry‐led decline in 1993, and were stable through 1996.5
During this period, California's comprehensive programme included an active multi‐message anti‐tobacco media campaign and community‐based projects in the earlier years (1990–1993), but programme funding was reduced in 1994–1996.2
Workplace smoking restrictions took effect in 1995. During the second period (beginning in 1996), funding was restored for the media and community‐based programmes. Further, cigarette prices increased sharply due to another excise tax increase ($0.50/pack) and a tobacco‐industry‐led price increase following the Master Settlement Agreement. The inflation‐adjusted increase in cigarette prices was 53% between November 1998 and November 1999.5
Our analyses cannot directly attribute any specific intervention (for example, increased cigarette prices, workplace smoking bans, etc) to specific components of the consumption decline; it can only show the net effect. However, large properly coordinated population surveys before (to establish underlying trends from pre‐existing conditions) and after a specific new intervention could allow the evaluation of their effects on per capita cigarette consumption. For instance, this approach could be used to evaluate the effect of the widespread distributing of free nicotine replacement products to smokers. It would be expected that the quitting component in the decomposition method would account for a relatively larger percentage of the total per capita decline following the intervention than in the previous control period.
What this paper adds
Per capita cigarette consumption has declined notably in California since the start of its comprehensive tobacco control programme. It is unknown what fraction of the decline has occurred from smokers quitting, from having fewer ever smokers in the population, or from current smokers smoking less. If the decline is mostly from smokers quitting, researchers suggest there would be a near‐term health benefit to the population. However, if the decline stems mainly from the other two possible sources, most of the health benefit to the population would occur in the longer term.
The results from our study indicate that the majority of the decline in per capita cigarette consumption was from current smokers smoking less (> 50%, particularly among younger smokers from 1990–1996), followed by a reduced proportion of ever smokers in the population (~30%). Between 1990 and 1996, a negligible fraction of the decline in per capita consumption was from ever smokers quitting, but 22% of the decline between 1996 and 2002 was from this source. Thus, instead of more immediate health benefits from current smokers quitting, benefits are more likely to accrue in the longer term, from there being fewer ever smokers, and from possibly a greater‐proportion of less‐addicted (and perhaps younger) smokers having an increased chance of future successful cessation.
In computing the quit ratio, we used as the numerator all ever smokers indicating that they were not smoking when surveyed. Many recent quitters will relapse to smoking. In fact, at any given point in time, an appreciable fraction of smokers are in the midst of an unsuccessful quit attempt and at least for a while not buying cigarettes. In 1990, 49% of California smokers reported making a quit attempt in the past year, but this percentage increased to 56% in 1996 and to 65% in 2002.5
This increased quitting activity, even if not successful, can decrease overall cigarette sales, and may account for a part of the discrepancy between sales and current smoker self‐reported data (fig 1). We chose to include recent quitters in the quit ratio, so that this measure reflects quitting activity, not successful cessation.
In addition to the issues discussed above, there are other limitations to these findings. Self‐reported cigarette consumption may not be accurate; smokers often report their consumption to the nearest half pack, and they may tend to round down instead of up. This number preference may also in part be responsible for the discrepancy between per capita consumption from sales data and from self‐report (fig 1). Further, low‐consumption smokers may report more accurately, so as per capita consumption declines, the discrepancy between sales and self‐report might decrease.44
However, since the two trends are parallel, this should not seriously affect our results. Survey sampling errors and the use of averages and differences between survey estimates introduces error in addition to the approximation error (assumption of linearity) of our methodology. Nevertheless, it is unlikely that our results are due to severe artefacts introduced from these error sources. Finally, declining response rates may mean that the population of smokers captured in recent years differs from those in the earlier years. However, a recent analysis of both national and state‐level surveys with different degrees of declining responses indicated that smoking prevalence estimates do not appear affected.45
While the most rapid benefit to public health from tobacco control efforts potentially would come from encouraging older smokers to quit,3,46,47
reduced initiation and reduced smoking intensity, particularly among younger smokers, may have greater benefits but in the longer‐term. Preventing adolescents from starting to smoke would eventually lower adult smoking prevalence and thus per capita consumption, but it would take many years for the population health effects to become discernable. If younger cohorts of smokers never reach the consumption levels observed in older cohorts, they should experience better health from reduced exposure to harmful toxins. Also, if less addicted, they should be more able to quit successfully. The evidence is not yet definitive about the health benefits of smokers reducing their consumption, so how this factor will affect future health is unknown.