Article Text
PDF
Abstract
Objective A pressing tobacco policy concern is how to help smokers who have little interest in quitting cigarettes, a group that often suffers severe health consequences. By switching from cigarettes to e-cigarettes, they could obtain nicotine, potentially with less harm. We examined if policy-relevant attributes of cigarettes/e-cigarettes might encourage these smokers to switch to e-cigarettes.
Methods An online survey and discrete choice experiment on a nationally-representative sample of adult smokers in the US who reported low interest in quitting (n=2000). We modelled preference heterogeneity using a latent class, latent variable model. We simulated policies that could encourage switching to e-cigarettes.
Results Participants formed two latent classes: (1) those with very strong preferences for their own cigarettes; and (2) those whose choices were more responsive to policies. The latter group’s choices were only somewhat responsive to menthol cigarette bans and taxes; the former group’s choices were unresponsive.
Conclusions The policies studied seem unlikely to encourage harm reduction for individuals with little interest in quitting smoking.
- electronic nicotine delivery devices
- cessation
- economics
Data availability statement
No data are available.
Statistics from Altmetric.com
Introduction
Cigarette smoking is one of the leading causes of preventable death in the USA, associated with more than 480 000 deaths each year. Yet many individuals who smoke have no interest in quitting cigarettes, and are putting themselves at high risk of lung cancer, other diseases and death.1 The National Health Interview Survey indicates that around 30% of individuals who smoked cigarettes in 2015 were not interested in quitting2 despite the availability of help through clinical cessation treatments (Varenicline, nicotine replacement therapies, counselling services, etc), government tobacco policies (increased taxes, advertising bans, public education campaigns, etc) and smoking self-help tools (books, apps, etc). Helping these individuals quit smoking is difficult, but important, for policymakers.
An alternative strategy to cessation is harm reduction through substitution to electronic cigarettes (e-cigarettes hereafter). E-cigarettes offer individuals who smoke an experience that closely emulates smoking in terms of look, feel and nicotine accessibility, but are potentially less harmful than smoking.3 4 E-cigarettes reduce exposure to many toxicants and carcinogens of combustible tobacco.5 Thus, e-cigarettes may offer an appealing alternative to smoking.4 6 7 Smokers could reduce some of their smoking by e-cigarette use or preferably could substitute completely to e-cigarettes.
Newer pod e-cigarettes may offer a better approach to harm reduction for individuals who smoke and have no interest in quitting cigarettes as compared with older disposable types. Pod-type products typically use salt-base nicotine which absorbs nicotine faster and better approximates the high nicotine levels of combustible cigarettes8 in contrast to older e-cigarettes that use ‘freebase nicotine’.
Public policies could encourage pod e-cigarette use for harm reduction or cessation for individuals not ready to quit smoking. States and municipalities have passed ‘flavour’ bans that prohibit flavoured e-cigarettes and Massachusetts passed a menthol cigarette ban.9 Interestingly, President Biden has stated interest in ‘banning’ menthol cigarettes.10 The Food and Drug Administration (FDA) can also affect the availability of flavours and other attributes as well in e-cigarettes but by a different mechanism. All e-cigarette brands can be evaluated, regulated and taken off the market by the FDA if they do not meet FDA approval based on the impact on public health. Through the FDA’s Pre-Market authority manufacturers must submit marketing applications and receive authorisation, or be taken off the market. The FDA also can, but has not, prohibit nicotine levels to non-addicting levels but cannot eliminate nicotine.
Prices can be increased by higher taxes imposed by multiple levels of the government including local, state and federal, but not the FDA. However, the FDA can indirectly affect prices of cigarettes by imposing product standards and marketing regulations that increase firms’ production costs. Alternatively, federal policies directing government or private health insurance programmes to cover e-cigarette as a cessation or harm-reduction aid could reduce the out-of-pocket price of e-cigarettes for individuals interested in reducing and/or quitting smoking.
This paper provides policy-relevant, empirical findings and fills some gaps. Studies have not yet fully determined: whether and the extent to which individuals not interested in quitting smoking prefer pod over disposable e-cigarettes and pod e-cigarettes over cigarettes; how these preferences vary according to individuals’ characteristics (age, gender, etc); and how their preferences for pod e-cigarette versus cigarettes are influenced by policy-relevant attributes, namely: flavour, nicotine, healthiness and price. This information is crucial for setting FDA and state policies for cigarettes and e-cigarettes to reduce the public health risk of smoking for those not interested in quitting.
Methods
Sampling
The sample contains 2000 individuals ages 35–85 years in the USA who reported ‘little interest in quitting’ cigarettes’. In our study, little interest in quitting was defined as a response of less than 6 (on a scale of 1 to 10) using the Population Assessment of Tobacco and Health (PATH) Survey question, ‘Overall, on a scale from 1 to 10 (1 is not at all interested and 10 is extremely interested), how interested are you in quitting cigarettes?’.11
The sample was matched to quotas derived from wave 3 of the PATH Survey by: Interest-in-quitting status, age (categories 35–54 years, 55–85 years), gender, race (white, non-white), education (any college vs no college) and census region (North-East, North Central/Midwest, West, South). Data were collected by the survey firm Qualtrics from June to September 2020. Qualtrics was unable to fill the full set of quotas when collecting our sample. Thus, to complete the sample within the other quotas, the condition of <6 for quit interest was relaxed to <9. In the final sample, 1892 reported a quit interest of <6; 25 reported 6; 33 reported 7; and 50 reported 8.
Sociodemographic and smoking history data were collected alongside the experiment. Descriptive statistics are provided in table 1. Only 19% of the total sample reported current e-cigarette use, thus indicating an opportunity to encourage smokers to switch to lower harm e-cigarettes.
Descriptive statistics for sample
Discrete choice experiment
The discrete choice experiment (DCE) is a preference elicitation technique used widely in healthcare.12–14 DCEs are increasingly used to examine tobacco markets.15–20 DCEs are particularly useful because currently available secondary data are not suitable for studying new tobacco products, rapidly changing markets and the impact of policies yet to be implemented.
Experimental approach
To determine the extent to which policies could encourage substitution to e-cigarette use, this study examined which policy-relevant attributes of cigarettes and e-cigarettes might encourage individuals with little interest in quitting cigarettes to switch to e-cigarettes. Using choice data from the experiment, we estimated preferences for tobacco products (cigarettes, disposable e-cigarettes and pod e-cigarettes) across key policy-relevant attributes: price, flavour, nicotine level, healthiness and usefulness as cessation aid. Using respondent data on their characteristics, we modelled heterogeneity in preferences across characteristics of smokers using a latent class, latent variable model. Then we simulated impacts of policies that could encourage them to switch to e-cigarettes.
This DCE was designed to elicit preferences for e-cigarettes relative to the individual’s usual cigarette. Respondents answered a series of choice scenarios (figure 1 shows an example choice scenario). In each scenario, respondents chose between their usual cigarette, pod e-cigarettes and disposable e-cigarettes. Using the respondents’ own cigarettes as the comparator embeds realism into the design; respondents consider the e-cigarette options relative to their own cigarette and its attributes (eg, menthol or not) rather than to a generic cigarette.19 20 This feature helps reduce bias from hypothetical responses.20 21 This also allows us to use information from respondents’ real-world behaviour in the choice models (the price they usually pay for cigarettes, their usual cigarette flavour, etc).
Example choice task.
Attributes and levels
The e-cigarettes were described by five attributes: flavours, nicotine level, healthier than cigarettes, helps you quit cigarettes and price. In each choice scenario, ‘levels’ of e-cigarette attributes were experimentally varied; and respondents’ own reported levels of these cigarette attributes were used, based on their responses to survey questions. The products, attributes and experimental levels are presented in table 2 and an example choice scenario is given in figure 1.
Experimental design
The four e-cigarette flavour levels were ‘Tobacco’, ‘Menthol/Mint’, ‘Fruit’ and ‘Sweet’. Collectively, these flavours account for the majority of e-cigarette flavours used in the USA.22 Menthol is the only flavour choice on the market for cigarettes other than tobacco; about 40% of the sample smoked menthol (table 1). Both healthier than cigarettes and helps you quit cigarettes were simple binary attributes whose levels were ‘Yes’ and ‘No’.
The nicotine attribute was defined in terms of nicotine-per-puff to avoid confusion between the differential total amount of nicotine in one cigarette versus one e-cigarette. Nicotine-per-puff of the e-cigarette was compared with the nicotine per puff in the respondent’s usual cigarettes. The four levels were: ‘No nicotine at all’, and ‘Less than’, ‘The same as’, and ‘More than’ ‘your usual cigarette’. This comparative approach avoids confusion about: (a) Absolute amounts of nicotine in cigarettes; (b) different levels of nicotine in e-cigarettes; (c) differences in nicotine absorption between nicotine in cigarettes, salt-base nicotine and freebase nicotine (which typically vary across e-cigarette types). It was used also because often smokers are not aware of the absolute levels of nicotine in their cigarettes.
Price of e-cigarettes was defined to be the price for an amount of e-cigarettes equivalent to 20 cigarettes, a commonly used comparison. Levels, based on market prices,23 were ‘$3.00’, ‘$6.00’, ‘$9.00’ and ‘$12.00’. Price for cigarettes was that of the respondent.
Technical details of the experimental design are reported in the online supplemental appendix.
Supplemental material
Choice modelling
Outcome data are the set of choices made by respondents across the scenarios. Analyses were preregistered before data were collected (https://osf.io/z9gu6/). We built an econometric choice model to analyse the experimental choice data.24 ,25 Utility for tobacco products is 
where individual n derives utility from product i in choice t.
Where preferences for disposable e-cigarettes and pod e-cigarettes relative to smokers’ own cigarettes, are respectively the constants: 
and 
. 
captures any additional smoker preference for menthol cigarettes. 
, 
, 
and 
capture the utility of these attributes. The
β
s are the preferences for the attributes to be estimated. Based on prospect theory and preliminary testing, reference-dependent prices were specified,26 as indicated above. That is, prices for e-cigarettes are defined relative to prices of smokers’ regular cigarettes. Each e-cigarette option is thus either cheaper or more expensive than the individual’s usual cigarette in each choice scenario; parameters ‘cheaper’ and ‘expensive’ are accordingly estimated. The error term, 
follows an iid type I extreme value distribution; we thus use a multinomial logit model in estimation.
This basic framework above is extended in two ways. First, to allow for groupwise behaviours, we used a latent class model with two classes. Second, to control for changes in health behaviours due to COVID-19, we used a latent variable approach. This approach uses reported, COVID-related changes in health behaviours (smoking, drinking, exercise, sleep) to inform a latent variable of health behaviours. This variable indicates if individuals adopted healthier or less healthy behaviours throughout the pandemic. We then use this variable to explain the class membership of each individual. In this way, like other sociodemographic characteristics, the health behaviours are related to the smoking choices in the model. The details of the approach are presented in the online supplemental appendix.
Policy simulations
The estimated choice model is used to simulate the effect of flavour and price-based (eg, taxation) policies on smokers’ product choices. In the USA both policies can be implemented at the federal or state level. Note that e-cigarettes could potentially be subsidised by a mandate on health insurers (private and/or public) as a cessation product. The methods for these simulations are detailed in the online supplemental appendix. The impact of banning menthol in cigarettes and of adjusting the relative prices of e-cigarettes are simulated. E-cigarette flavour bans were not simulated because preferences were negative for these flavours, that is, banning them would not impact cigarette choices.
Promoting data quality
We used several techniques to promote data quality. The experiment was based on smokers’ own cigarettes characteristics which is beneficial for survey realism and data quality. Before the experiment, respondents were given detailed narrative and visual information describing the products, attributes and levels; and an example choice task was provided to familiarise respondents with the task. The survey was piloted on 161 respondents and feedback was used to improve the survey. ‘Forced responses’ prevented respondents from skipping through the survey. Respondents pledged to provide the best answers that they could. A minimum time threshold was used to remove respondents who rushed through.
Results
Choice models
Table 3 presents the results from the latent class choice model. Estimations reveal two classes of individuals, those that we term non-switchers and switchers, respectively. Preferences for products and attributes varied across the classes; and classes were related to socioeconomic and demographic factors as seen in table 3.
Utility function, class membership and diagnostics from latent class, latent variable multinomial logistic (LCLVMNL) model of NIQ smokers’ experimental choices
Non-switchers had very strong preferences for their own cigarettes as seen in the negative and significant estimated coefficients on constant terms for each type of e-cigarette. E-cigarettes’ utility increased when they were cheaper than one’s own cigarette but not when they were more expensive. The negative and significant coefficients on ‘helps you quit’ (ie, ‘used as aid in smoking cessation’), ‘e-cigarettes that were healthier than cigarettes’ and all non-tobacco flavours in e-cigarettes indicate that non-switchers disliked these characteristics for e-cigarettes. However, non-switchers derived utility (positive and significant coefficients) for e-cigarettes with lower (but not the same or higher) nicotine levels than their current cigarettes. Overall, these smokers strongly preferred their own cigarettes to e-cigarettes across most of the attributes.
Switchers did not have particularly strong preferences for their own cigarettes. They were indifferent between their own cigarettes and pod e-cigarettes (non-significant constant term on pod e-cigarette). Yet they had a preference for their own cigarettes as compared with disposables as seen in the negative and significant estimated coefficient on the constant term for disposable e-cigarette. Their utility for e-cigarettes also increased when e-cigarettes were cheaper than one’s own cigarette but not when they were more expensive. The coefficients on ‘helps you quit’ and healthier options were insignificant. However, the coefficient on ‘if cigarettes had no nicotine’ was positive and significant yet the coefficients for less or more nicotine than their regular cigarettes were insignificant. Thus, they preferred nicotine-free e-cigarettes to e-cigarettes with the same nicotine as their regular cigarette. They disliked all non-tobacco flavours in e-cigarettes and preferred cheaper prices.
Non-switchers comprised 68% of the sample, and switchers comprised 32% of the sample. The probability of being in the non-switchers class is positively related to being: older, lower educated, lower income, white and a non-dual user (of cigarettes and e-cigarettes). The difference between the 32% that might switch under different policy scenarios and the 19% that use e-cigarettes now suggests that more smokers might use e-cigarettes with effective policies.
Policy simulations
Table 4 shows the policy simulations. Four scenarios are presented: baseline simulation without and with calibration (matching to real-world data); a ban on menthol cigarettes (relative to the baseline); and an increase in taxes (relative to baseline). We did not estimate impacts of flavour bans on e-cigarettes because these smokers disliked these thus banning them would have no impact on them.
Policy simulations by product type
Simulations indicate that the response of individuals’ choices to a cigarette menthol ban would be a small reduction in cigarette choice shares (0.5%) and small increases in the choice shares for both pod e-cigarettes (2.7%) and disposable e-cigarettes (2.8%). Increasing the relative prices of cigarettes by 50% would result in a reduction in the cigarette choice share (1.0%), a modest increase in the choice-share of both pod e-cigarettes (3.1%), and a more pronounced increase in the choice share of disposable e-cigarettes (11.5%). Note that we simulated choice of product and their product shares (percentages), not the smoking and e-cigarette use rates nor quantities consumed. See notes on table 4.
Validity checks, specification tests and sensitivity analyses
Checks were made as to whether estimated coefficients were in line with theoretical expectations, for example, higher utility for reduced prices. We tested the difference between e-cigarettes being cheaper and more expensive and found that the slope for cheaper was different to the slope for more expensive (rather than assuming linearity in the price sensitivity); equality in the absolute value of the cheaper/expensive coefficients was rejected (p<0.01). This approach also improved the overall fit of the model compared with treating price continuously. Tests of equality were conducted on: disposable versus pod e-cigarettes, fruit versus sweet flavours and pairwise tests for levels of nicotine. In all cases, equality was rejected, supporting our specification.
The model was re-estimated dropping individuals whose quit interest was greater than 5. The results were almost identical. In addition, we constructed a new variable of low interest (1–3 on the scale), moderate quit interest (4–7) and high interest (8) and used this variable in the class membership probability. The results were almost identical to our main analyses. This addresses concern about expanding the inclusion criteria to increase the sample size. Further, at the end of the survey, respondents were asked how carefully they had answered. Most, 1923/2000, selected that they had answered ‘extremely’ carefully. We checked and found that responses other than ‘extremely’ did not impact the estimated parameters.
Discussion
Main findings
We found that individuals selected for little interest in quitting their cigarette smoking fell into two types: those with very strong preferences for their own cigarettes (non-switchers, 68%); and those with weaker preferences (switchers, 32%). Switchers were more open to using e-cigarettes than the others. Yet only 19% of the total sample reported currently using e-cigarettes; that is, some smokers that do not currently use e-cigarettes could, in theory, be encouraged to do so through some policy options.
The strongest finding was that individuals in both groups disliked non-tobacco flavoured e-cigarettes. Switchers’ choices were only modestly responsive to a menthol ban in cigarettes. Both groups disliked the ability to quit by using e-cigarettes and non-switchers were not attracted by better health of e-cigarettes; switchers were indifferent. Both groups preferred lower prices and no nicotine in e-cigarettes; non-switchers also appreciated less nicotine. Switchers were younger, more educated, had higher income, of black or Asian race, and more likely to be dual users of cigarettes and e-cigarettes. Overall choice shares of e-cigarettes remained low even in policy scenarios that favoured them.
Strengths and limitations
This study has several strengths including use of a DCE alongside a survey and provision of policy simulations. These methods allow for the study of policies before they are implemented, allowing determination of the potential efficacy of future policies. Approaches were implemented to obtain a representative sample and ensure high-quality data through survey design. The analyses were preregistered prior to collecting the data. The analyses used advanced models that allowed us to model heterogeneity in choices. The latent class approach was also useful in estimating group types and sizes; and produced information about which characteristics affected group type that otherwise would not have been identified. We were able to control for the impact of the pandemic on health behaviours within the modelling framework. We adjusted to control in part for the impact of COVID-19; we think that these results are more robust for having done so (compared with no control). We also conducted specification tests and sensitivity analyses and compared our empirical findings to known relationships.
Limitations include that outcomes are hypothetical, however, we used methods to enhance choice realism such as asking for choices in comparisons to their own cigarette. We examined percentages who chose cigarette or e-cigarette types but not quantity of smoking or e-cigarette use. We aimed for a sample of those with little interest in quitting but had to relax this criterion somewhat to obtain our full sample; yet we conducted sensitivity tests on this issue. While we sampled on quit interest, we did not sample based on other smoking characteristics, such as e-cigarette use. This may have led to biases in the sample composition and the results. Perhaps another limitation is that the sample does not include many black smokers. Though this is a consequence of our quotas, very few African-American smokers reported low interest in quitting. However, 30% of our sample does report menthol use, meaning the policy predictions are still relevant to the sample. Lastly, we consider these to be short-term, not necessarily long-term, responses.
Our results are difficult to compare to other estimates in the literature for several reasons. First, this study measured price elasticity of choice, which is a different margin of behaviour to price elasticity of demand or price elasticity of participation that have been studied elsewhere.21 27 Second, data sources vary considerably. We use a DCE, though other experimental approaches are available, as are observational data, all of which have been applied.28 29 Third, we sample from a specific subset of smokers, which makes comparisons with other general smokers samples,17 19 or other specific samples (eg, young adults)18 difficult, even when they use choice elasticities. This is the first study to measure price response to e-cigarettes among smokers with little interest in quitting smoking.
Policy implications
Efforts to get individuals uninterested in quitting cigarettes to engage in harm reduction or quitting by switching to the potentially lower-risk alternative of using e-cigarettes might increase the quality and quantity of their lives. However, in this study, flavours were disliked over the full sample so that policies that allow flavours in e-cigarettes would not be productive for this group. Conversely, the results suggest that if the FDA decided to limit flavours to reduce e-cigarette use by youth (the main goal of flavour bans), flavour bans would have little negative impact on the uptake of e-cigarettes by these smokers. Specifically, the proposed banning of menthol cigarettes would not likely induce choosing e-cigarettes instead of cigarettes. These results suggest that higher taxes on cigarettes would have limited effect in promoting switching among individuals uninterested in quitting smoking. Overall, the policies studies had little impact on their choices.
The older individuals who are entrenched in their current smoking practice (non-switchers) may not be persuaded about the benefits of quitting or their ability to switch due to their smoking history.30 31 However, those in the group referred to as ‘switchers’ would be more likely to either continue or add or use e-cigarettes rather than smoke through these policies, but only marginally. Switchers were somewhat responsive to lower e-cigarette prices. This is in keeping with evidence suggesting that providing free e-cigarettes encourages use among people who smoke.20 32 Policies directing government health insurance programmes to cover the costs of e-cigarettes used in quitting and/or harm reduction and mandates for private insurers to do so could have a small benefit to the health of individuals with little interest in quiting smoking.
What this paper adds
What is already known on this subject
Smokers who are not interested in quitting have high health risks from smoking.
Harm reduction can reduce these harms to these smokers.
Evidence suggests some of these smokers are interested in e-cigarettes to reduce their risks.
What important gaps in knowledge exist on this topic
These smokers’ preferences for pod e-cigarettes for harm reduction are unknown.
Their relative preferences for disposable e-cigarettes, pod e-cigarettes and cigarettes are similarly unknown.
The extent that specific public policies can encourage harm reduction via switching from cigarettes to e-cigarettes is not well understood.
What this paper adds
We estimate these smokers’ relative preferences for disposable e-cigarettes, pod e-cigarettes and their own cigarettes, finding two types of smokers; ‘switchers’ and ‘non-switchers’.
Based on these findings, we estimate the impact of a menthol ban and taxes on cigarettes.
We find that neither a menthol ban nor higher taxes on cigarettes are likely to encourage switching to e-cigarettes for either type of smoker.
Data availability statement
No data are available.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by the Yale Human Subjects Committee (HSC), that considered this study exempt from IRB review. Participants gave informed consent to participate in the study before taking part.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors Guarantor: JB; Conception or design of the work: JB, JLS, LMF, SO, SK-S. Acquisition of the data: JB, JLS. Analysis of the data: JB, JLS. Interpretation of data for the work: JB, JLS, LMF, SO, SK-S. Drafting the work: JB, JLS, LMF, SO, SK-S. Final approval of the version to be published: JB, JLS, LMF, SO, SK-S.
Funding Research reported in this publication was supported by grant numbers U54DA036151 from the NIDA and FDA Center for Tobacco Products (CTP) and K12 DA000167 from NIDA.
Disclaimer The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or the Food and Drug Administration. It was also supported by the NIHR Oxford Biomedical Research Centre.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.