Shu-Hong Zhu, Cancer Center, Mail Code 0905, University of California, San Diego, La Jolla, CA 92093-0905, USA. E-mail: email@example.com
Aims We examine the phenomenon of low-frequency smoking (non-daily smoking or smoking ≤ 5 cigarettes daily) among California Latinos and address its implications for addiction theory and population tobacco control.
Design, setting and participants Data gathered in 2001 and 2003 through the California Health Interview Survey (CHIS), the largest general health survey in California. The present study focused on Latino current smokers (n = 1254 for CHIS 2001; n = 946 for CHIS 2003).
Measurement Latino smokers reporting either non-daily smoking or smoking ≤5 cigarettes daily were identified and grouped into one category: low-frequency smokers.
Findings Weighted by population parameters, more than 70% of Latino smokers in California were found to be low-frequency smokers [70.7% (CI = 67.2%, 73.9%) in 2001 and 70.8% (CI = 67.1%−74.2%) in 2003]. This high proportion cut across all demographic dimensions in both surveys, suggesting pervasiveness and reliability of this phenomenon. Proportions for non-daily smokers and low-rate daily smokers were 48.6% and 22.1% in 2001 and 54.9% and 15.9% in 2003. In both surveys, more than 80% of non-daily smokers consumed ≤ 5 cigarettes on their smoking days.
Conclusions The fact that most Latino smokers are low-frequency smokers calls for a new theoretical framework—beyond withdrawal-based theories—to account for the prevalence of this behavior on the population level. It also calls into question the harm-reduction approach as a tobacco control strategy for California Latino populations. Strategies emphasizing that every cigarette can hurt, and encouraging complete cessation, seem more fitting for this group of smokers.
Low-frequency smokers, meaning non-daily smokers and low-rate daily smokers [consuming ≤5 cigarettes per day (cpd)], have received increasing attention in recent years [1–5]. Before the 1990s, most population surveys of tobacco use in the United States did not ask smokers if they smoked daily because it was thought that every established smoker smoked daily. This reflects the influence of the classical view of addiction, which assumes that tobacco use is driven by the need for regular nicotine intake to avoid withdrawal symptoms [6,7]. Given that the half-life of nicotine is approximately 2 hours , it might seem difficult to conceive of established smokers not smoking daily. However, later studies of light smokers have found that a substantial proportion of smokers do not smoke daily [2,4,5,9]. Moreover, even some daily smokers do not consume enough cigarettes to maintain a consistent plasma nicotine level [10,11]. Daily low-rate smokers, or chippers, who consume less than six cpd , do not smoke sufficient amounts to maintain effective nicotine regulation [10,11], nor do they evidence withdrawal symptoms . Non-daily smokers, or occasional smokers, regularly go without cigarettes for a whole day. On average, they smoke only about 15 days per month [2,5,14]. The behavior of these low-frequency smokers clearly does not fit the withdrawal-based addiction models.
Studies have also found that the proportion of low-frequency smokers varies significantly across groups. Latinos who smoke, for example, are much more likely to be non-daily smokers than are smokers of any other ethnic groups [5,15,16]. Although no specific study has compared proportions of low-rate daily smokers by ethnic group, Latino populations generally have a high proportion of light smokers, defined variously as smoking fewer than 10 or 15 cpd [7,17,18]. Furthermore, studies have found that the proportion of low-frequency smokers in the established smoking population varies over time [14,19]. When there is a strong tobacco control program, for example, more heavy smokers become non-daily and low-rate smokers than the other way around [14,20]. This shift runs contrary to the trajectory assumed for an addictive behavior, which generally exhibits an increased consumption of the addictive substance over time due to the development of tolerance [6,7].
The present study focuses on low-frequency smoking behavior of Latino smokers in California. Given that Latinos are more likely to be low-frequency smokers, and given that California has conducted one of the nation's strongest and longest-running tobacco control programs , it might be expected that the California Latino population would have a larger than average proportion of low-frequency smokers. Because the behavior of low-frequency smokers (defined here as non-daily smoking or daily smoking of ≤5 cpd) poses the greatest challenge to the withdrawal-based theories [7,10,12], the present study focuses on these smokers rather than on light smokers in general (<15 cpd) [7,17]. We use population survey data to examine this low-frequency phenomenon, which is more likely to be observed among the general smoking population than among participants in clinical studies.
There is also practical value in identifying the proportion of low-frequency smokers in the current smoking population. A significant change in smoking consumption in the current smoking population implies a reduction of health risk, because the disease burden of smoking is dose-responsive to the consumption level . Moreover, longitudinal studies have shown that low-frequency smokers are more likely to succeed in their attempts to quit [14,23]. Thus, any changes in the proportion of low-frequency smokers in the current smoking population can have implications for developing population tobacco control strategies.
The present study uses data from the California Health Interview Surveys (CHIS) of 2001 and 2003 [24,25]. The CHIS was chosen for two reasons. First, it provides the largest survey sample of smokers representative of the Latino smoking population in California. Secondly, CHIS is a general health survey. It has been suggested that some non-daily smokers, due to their infrequent smoking behavior, may not identify themselves as smokers in order to decline participation in smoking-specific surveys ; therefore, we employed data from a general health survey in the hope of capturing more information about non-daily smokers. The use of data from two surveys (CHIS 2001 and 2003) provides a reliability check of results.
The California Health Interview Survey (CHIS) is the largest general health survey in California. One randomly selected adult was surveyed in each sampled household, using a two-stage, geographically stratified, random-digit-dial (RDD) sample design [24,25]. CHIS 2001 surveyed 57 848 adults, and Latinos (both English-speaking and Spanish-speaking) made up 21.4% of the sample (10 559). CHIS 2003 surveyed 42 044 adults, and Latinos made up 20.9% of the sample (8770). The overall screener completion rates were 59.2% for CHIS 2001 and 55.9% for CHIS 2003, and the extended interview completion rates were 63.7% for CHIS 2001 and 60% for CHIS 2003. Thus, the overall weighted response rates were 37.7% for CHIS 2001 and 33.5% for CHIS 2003, comparable to other health surveys such as the Behavioral Risk Factor Surveillance Survey [27,28]. Demographic data from the 2001 CHIS showed comparability to 2000 US Census data .
In both surveys, 11–12% of adult interviews were conducted in a language other than English. For Latino participants who preferred Spanish, all written and spoken materials were translated into that language. Translation analysis was conducted, and multi-language letters were utilized with both English and Spanish versions. Spanish bilingual interviewers were trained and monitored by Spanish-speaking supervisors.
To our knowledge, CHIS is the largest survey that provides smoking data on Latino smokers in California. The Latino sample size in CHIS is bigger than those in the ongoing California Tobacco Surveys  and than those in other national surveys—such as the Current Population Survey: Tobacco Use Supplement —that include California samples. The present study focuses only on Latino respondents who reported current smoking. The effective sample sizes for the present study, the Latinos who identified themselves as current smokers, were 1254 for CHIS 2001 and 946 for CHIS 2003.
There are three tobacco questions embedded in the first quarter of the CHIS, in a section on health behaviors following general health questions. CHIS 2001 asked the following smoking-related questions: ‘Altogether, have you ever smoked at least 100 or more cigarettes in your entire life-time?’; ‘Do you now smoke cigarettes every day, some days, or not at all?’; and for all current smokers ‘In the past 30 days, when you smoked, how many cigarettes did you smoke per day?’. CHIS 2003 asked the same smoking-related questions, except that the question assessing the number of cpd was revised. For daily smokers, CHIS 2003 asked, ‘On the average, how many cigarettes do you now smoke a day?’. For non-daily smokers, it asked: ‘In the past 30 days, when you smoked, how many cigarettes did you smoke per day (on the days you smoked)?’.
Current smokers were defined as ‘those who have smoked 100 cigarettes in their life-time’ and are currently smoking. Among these current smokers, non-daily smokers were defined as ‘those who smoke on some days’ based on their response to the question mentioned above; low rate daily smokers as ‘daily smokers who smoke five or fewer cpd’. The low-frequency smokers are the sum of the non-daily smokers and low-rate daily smokers.
The category ‘Latino’ was defined as anyone who answered ‘yes’ to the question: ‘Are you Latino or Hispanic?’. Also included in this category were those who answered ‘no’ to the above question but later, when presented with a list of ethnic groups and asked, ‘Of these [groups], which do you most identify with?’ identified themselves as Mexican/Mexicano, Mexican American, Chicano, Salvadoran, Guatemalan, Costa Rican, Honduran, Nicaraguan, Panamanian, Puerto Rican or Cuban.
Household income was derived by asking the household's total annual income from all sources before taxes. The public-use data set for CHIS anchored this information to federal poverty level (FPL) guidelines, which divided family income into four poverty levels: at or below FPL; 100–199% FPL; 200–299% FPL; and greater than or equal to 300% FPL. Dominant language was defined as ‘the language spoken at home’. If the respondents used Spanish only at home, they were coded as Spanish-only. If they used both English and Spanish, or English and other languages, they were coded as English-speaking.
The proportion of non-daily and low-rate daily smokers was calculated separately. They were also combined into the one category of low-frequency smokers. The confidence intervals were calculated only for the low-frequency smokers as one group and are presented in the table. All calculations (averages, rates and percentages in histograms) are weighted by 2000 US Census data. All computations were performed with SAS 9.1  and confidence intervals were estimated using SUDAAN 9.1 .
The proportions of non-daily smokers and low-rate daily smokers among Latino current smokers in California are presented in Table 1. In 2001, close to half (48.6%) were non-daily smokers and 22.1% were low-rate daily smokers. In 2003, the proportion of non-daily smokers increased to 54.9% and the percentage of low-rate daily smokers dropped to 15.9%. The overall proportion of low-frequency smokers remained around 71% (70.7% in 2001 and 70.8% in 2003).
Table 1. Distribution of consumption levels by demographics for Latino smokers in California.
Daily low-rate %
Combined low-frequency % (95% CI)
Daily low-rate %
Combined low-frequency % (95% CI)
CHIS = California Health Interview Survey; CI = confidence intervals; FPL = Federal Poverty Level.
≤ 12 years
> 12 years
≥ 300% FPL
Table 1 also presents the population-weighted distributions of California Latino low-frequency smokers along the basic demographic dimensions, plus the participant's dominant language. It is clear that their tendency toward low-frequency smoking cuts across almost all these demographic dimensions. There are variations, but few reach statistical significance. Females appear more likely than males to be low-frequency smokers, as are younger smokers compared to older ones. However, few comparisons reach the conventional significance level, as indicated by the overlapping confidence intervals. For the combined category of low-frequency (i.e. non-daily + low-rate daily), all entries are greater than 50%.
Figure 1 presents the number of cpd consumed by California Latino non-daily smokers on the days that they do smoke. The distribution is positively skewed for both the 2001 and the 2003 CHIS. In 2001, the mode was 2 cpd, whereas the mean cpd on smoking days was 4.6 (SD = 0.57) and the median was 2. In 2003, the mode was 1 cpd; the mean cpd on smoking days was 4.5 (SD = 0.41), and the median was 3. Thus, non-daily smokers, on average, can be considered low-rate smokers on the days they smoke. More precisely, Fig. 1 shows that 85.0% of the non-daily smokers in 2001 and 80.1% of non-daily in 2003 could be considered low-rate smokers (i.e. smoking ≤5 cpd) on their smoking days. The rest of them smoked more than 5 cpd, including a few reporting more than 40 cpd.
For low-rate daily smokers, the mean cpd was 3.2 (SD = 0.09) and 3.4 (SD = 0.13) in 2001 and 2003, respectively, and the median was 3 in both 2001 and 2003. The distribution of cpd for low-rate smokers was not as skewed as that for non-daily smokers; thus it is not presented in the figures. It should also be noted here that for the minority of California Latino daily smokers who smoke more than 5 cpd (a group that is not the focus of this study), the mean cpd was 14.0 (SD = 0.50) and 12.3 (SD = 0.47) for the years 2001 and 2003, respectively.
This study found that over 70% of Latino current smokers in California are low-frequency smokers: they either do not smoke daily or smoke only ≤5 cpd. This confirms previous studies that have found that many Latinos smoke infrequently [2,5,15,16], but 70% may be the highest proportion that has ever been reported in the literature. One reason for such a high proportion could be that the present study employs data from a general health survey, rather than a tobacco-specific survey. The three tobacco questions were embedded in the first quarter of the CHIS survey in a section on health behaviors following general health questions. This could have made the non-daily smokers more likely to identify themselves as smokers than they would have been in a tobacco-specific survey . A second reason for the high proportion reported in this study could be that the study was conducted in California, where the consumption level of smokers has been declining since 1990 due to a strong tobacco control program . There may be other reasons as well, but the virtually identical overall proportion of low-frequency smoking obtained in both 2001 and 2003 and the consistent pattern of low-frequency smoking across demographic dimensions in both surveys suggest that this is a genuine phenomenon.
Among these low-frequency smokers, about a quarter are low-rate daily smokers, and the rest—the majority—are non-daily smokers. As mentioned, laboratory studies of low-rate daily smokers have demonstrated that they do not smoke enough cigarettes to maintain a consistent plasma nicotine level [10,33], nor do they exhibit symptoms of nicotine dependence if they have smoked this way for some time [12,13]. Few laboratory-based studies have examined non-daily smokers in the same manner. However, non-daily smokers, by definition, can go without smoking for at least a whole day. Even on the days when they smoke, most of them do not consume many cigarettes. Figure 1 shows that the majority (> 80%) of California Latino non-daily smokers were low-rate smokers on their smoking days, and the mean cpd for all the non-daily smokers in both 2001 and 2002 was less than five. Their low level of consumption on smoking days, added to the fact that they do not even smoke daily, suggests that they do not smoke to maintain consistent plasma nicotine levels. It is also quite likely that they do not experience classic withdrawal symptoms on days when they do not smoke.
The fact that 70% of California Latino smokers are low-frequency smokers indicates that we need to go beyond withdrawal-based addiction models to find explanations for their smoking behaviors. These data, based on the CHIS, do not allow detailed formulation of an alternative model. However, they do provide clear constraints for any possible models. Any theory of smoking behavior will have to account for the prevalence of low-frequency smoking among Latino smokers because for them, it is the majority behavior. In practical terms, any model that leaves the behavior of the majority unaccounted for will not be a very useful tool for public health efforts to control tobacco use for this population.
A natural question is whether there are biological factors that equip Latino smokers to maintain a low-frequency habit. A search of the literature found no study reporting biological evidence or suggesting biological explanations for Latinos' tendency toward low-frequency smoking. In contrast, studies have found slower nicotine metabolism among Asians  and slower cotinine metabolism among African Americans , two groups that actually include fewer low-frequency smokers than Latinos [5,15]. Explanations, for now at least, must be sought on a psychosocial level.
One possible explanation is that the proportion of low-frequency smokers among Latinos is so high that it has passed a critical point, making it relatively easy for smokers to maintain a low-frequency habit. On a cognitive level, the frequency of smoking among Latino current smokers is so low that the expectation, even among the smokers themselves, may be that in most situations the norm does not include smoking. In addition, many of the non-daily smokers may not even label themselves as smokers due to their low smoking frequency. All these would lower the probability that smokers would think of smoking while engaged in most daily activities [36,37] On a cue–response level, Latino smokers may be less likely to be exposed to smoking cues, including seeing others smoking [38,39] Even if they frequently encounter members of their social group who do smoke, they are not likely to see these individuals actually smoking, given the friends' likelihood of being low-frequency smokers themselves. On a behavioral level, it is even possible that many of the non-daily smokers do not carry their cigarettes with them due to their low-frequency smoking habit. Thus, even if they develop an urge to smoke on some unexpected occasion, they may find that they do not have ready access to cigarettes [40,41]. All these could work to keep the likelihood of smoking low among established Latino smokers, and each of these conjectures will need to be tested empirically.
The psychosocial mechanism outlined above, if it is shown to be correct, could be operative in smokers of any ethnic background, although it might work more effectively in groups such as Latinos, whose culture is more collectivistic. In any case, the advantage of studying Latinos is that the proportion of low-frequency smokers is so high in this group that it will be easiest to test the validity of these conjectures with them.
The results reported here already have some important implications for public health practice in places such as California. Given that most Latino smokers are low-frequency smokers, the intervention message for this group should emphasize that even a single cigarette can harm one's health [42–44]. The much-debated harm reduction approach, which presumes that smokers have so much difficulty in quitting that they would be better served by switching to less harmful nicotine delivery systems such as smokeless tobacco or medicinal nicotine, seems quite inappropriate for this group. In fact, the high proportion of low-frequency smokers helps to explain why Latino smokers are more likely than smokers of other groups to think that they can quit easily whenever they want  and why they are less likely than other groups to seek formal treatment when quitting [46,47]. The data from the present study suggest that a campaign emphasizing the difficulty of quitting, with the aim of urging Latino smokers to use individual, formal treatment, is likely to affect only a minority of them (the heavier smokers). This is not to say that we need not encourage Latino smokers to seek help in quitting, but the intervention should capitalize on the social forces that appear to operate on Latino smokers. For these smokers, the campaign would focus on the social undesirability of smoking, including the importance of protecting children and other family members from secondhand smoke and setting good examples for children. Such a campaign has, in fact, been waged in California for many years . Studies have shown that Latinos respond to such social messages more quickly than other groups, evidenced by their higher proportions of home restrictions on smoking than any other groups [18,49,50]. Such a focus on social undesirability could also lead more heavy smokers to become low-frequency smokers [14,51] Because low-frequency smokers are more likely to make quit attempts than are heavier smokers, a higher proportion of low-frequency smokers can be expected to lead to a higher rate of quit attempts , which can lead to a higher cessation rate [14,52]. In fact, studies have already shown that Latinos are more likely to make quit attempts than are other ethnic groups, despite being least likely to receive a physician's advice to do so or to use pharmacological aids . Longitudinal studies based on random samples of smoking populations have also reported that the greater number of quit attempts among Latinos has indeed led to higher cessation rates for this group .
To some extent, therefore, the fact that most Latino smokers in California smoke infrequently suggests that the approach taken by the California Tobacco Control Program , emphasizing social norm change regarding smoking, is working to reduce health disparities related to tobacco use, at least for the Latino group. One example is the California effort to reduce health damage from secondhand smoke. The effort started with a law banning smoking at work-sites state-wide [21,48]. Simultaneously, a state-wide media campaign emphasized the harm of secondhand smoke. Researchers soon noticed that more Californians were banning smoking in their homes voluntarily, apparently inspired in part by the legal bans at work-sites and by associated state tobacco control activities . Somewhat surprisingly, Latinos are adopting the home ban idea faster than are any other ethnic populations in California [49,50], a practice that might contribute to their elevated likelihood of being low-frequency smokers [14,18,49,54]. Data of this type often do not fit the prevailing assumptions about health behavior of ethnic minorities or about how to conduct public health campaigns to reduce disparities. It is hoped that results such as those reported in the present study will inspire new research on ethnic minority groups, leading to new ways to reduce health disparities associated with tobacco use.
A number of strengths and limitations should be borne in mind when considering the results of the present study. A primary strength of the study is the utilization of a large-scale, general health survey administered on two separate occasions, providing evidence for the reliability of the results. As with other large surveys of its kind, this study was limited by the self-report nature of smoking information, which may be inaccurate. However, even if some smokers are underestimating the number of cigarettes they smoke [55,56] the proportion of low-frequency smokers in this group would still be very high. Also, CHIS has limited information about acculturation (the only relevant information available in both CHIS 2001 and 2003 is the language spoken at home), which does not allow more detailed analysis of how acculturation might have affected the smoking behavior. Additionally, the CHIS data are California-focused, and the majority of Latinos in California are of Mexican origin. Given the great heterogeneity among Latinos, future studies could explore whether or not patterns of low smoking frequency hold true for Cuban Americans, Puerto Ricans, South Americans and Central Americans in other parts of the United States. The present study is also limited by the fact that the CHIS did not assess how many days per month the non-daily smokers smoke, which makes it impossible to calculate the total number of cigarettes consumed per month by non-daily smokers. However, many studies, including those examining Latinos in California, have found that non-daily smokers generally smoke on about half the days in any given month [2,5,14]. There is little reason to expect that non-daily smokers in this study smoke more than 15 days per month on average. When such information becomes available, future studies could compare the monthly total number of cigarettes smoked by non-daily smokers and the monthly total smoked by low-rate daily smokers, providing a more complete picture of low-frequency smoking behavior.
In sum, this study shows that low-frequency smoking is the majority behavior (> 70%) for Latino smokers in California. This fact calls for a new theoretical framework beyond withdrawal-based theories to account for the prevalence of this behavior on the population level. Coupled with the fact that there is no known genetic explanation for Latinos' low-frequency smoking, these findings suggest that psychosocial factors are the main force driving the behavior reported here. A better-developed theory is needed not only for improving our understanding of smoking behavior, but also for guiding intervention efforts to reduce tobacco use on the population level.
This work was supported by a supplemental grant to the UCSD Cancer Center from the National Cancer Institute: Grant 5 P30 CA 23100-22S4 (Dr Zhu), which supports the utilization of large data sets on minority smokers that have already been collected. We are grateful for the data that are made available for public use by the California Health Interview Survey research team. Dr Baezconde-Garbanati is supported by the National Cancer Institute—TTURC NIH CA-98-029. The authors wish to thank Quyen Nguyen for assistance with the literature review and Chris Anderson, Sharon Cummins and Gary Tedeschi for comments on earlier drafts.