Smoking remains the leading cause of death in the developed world. In recent years, e-cigarettes have
gained traction as an alternative to cigarettes. In response to this, governments are developing a range of
market interventions to control their use. To maximize the impact of interventions, it is necessary to
understand smokers’ behaviors. 
 
A central feature of smokers’ decisions is how the relative risks of cigarettes and e-cigarettes are
perceived. While the harms of combustible cigarettes are well-documented, evidence for e-cigarette harm
is scarcer and inconclusive (Dinakar & O’Connor, 2016). While this type of information is likely to influence
risk perceptions and ultimately impact decision-making, the mechanisms from scientific evidence to actual
choices are complex. Moreover, how risk information is accessed, understood and processed differs,
which may explain observed heterogeneity in risk perceptions (Czoli et al., 2016).
 
Initial evidence suggests that smokers are more likely to have tried e-cigarettes when they perceive them
to be less risky than conventional cigarettes (Viscusi, 2016). In this study, we use a discrete choice
experiment to provide further insights into behavior. First, we model the effect of relative risk perception
on smokers’ utility. We estimate how much more harmful a product has to be before smokers will switch.
Second, we use reference-dependent utility to capture variation in risk perception and model loss
aversion in the health domain. Third, we model the effect of inertia – or status quo bias - on smoking
behavior. Lastly, we use a latent class model to capture group-wise behavior. 
 
A sample of US adult smokers chose between combustible and electronic cigarettes in an online
experiment. Information on risk perceptions, smoking patterns and sociodemographic characteristics was
collected alongside the experiment. This was used to construct measures of relative risk perception and
inertia that were introduced in the utility function. 
 
Our findings indicate that both product preferences and reaction to risk vary by group. For some groups,
there is an underlying preference for combustible cigarettes. Here, these individuals do not switch
products irrespective of their beliefs about the relative risks of these products. In other groups, the
preference for combustible cigarettes is weaker. Here, we are able to identify at what level of relative risk,
on average, smokers will switch products. Introducing inertia reveals that it is not only current dual users
(of combustible cigarettes and e-cigarettes) that are willing to switch; even staunch smokers of
combustible cigarettes will switch if they believe the relative health risks are sufficiently large. Lastly, we
find that smokers are loss averse in the health domain.
 
Our results have important policy implications. It is important to understand how individuals behave as
relative risk perception varies; in particular, how likely smokers are to switch products. It is important to
understand how behavioral biases – loss aversion and inertia - affect behavior for setting policy; for
example, for informing health risk messages on tobacco products. Further, it is important to understand
what kinds of individual behave in certain ways, such that policies can be targeted effectively.