Throughout each of the four epochs of innovation, behavioral economics
methods have provided assessments of the abuse liability of existing and emerging tobacco products. Defined broadly, abuse liability describes the degree to which a substance is likely to produce persistent, problematic patterns of use (Food and Drug Administration, 2010). More specifically, abuse liability is a multivariate construct drawing from decades of research in the natural, physical, and social sciences, where the degree of abuse liability depends on a number of factors. These factors, among others, include a tobacco product’s subjective effects, speed of onset of those effects, the product’s reinforcing value, and the likelihood it would produce physical or psychological dependence (Balster, 1991; Carter et al., 2009; Henningfield and Keenan, 1993; Stitzer and de Wit, 1998). Additional product constituents may also act synergistically with nicotine to increase abuse liability beyond that of nicotine alone (e.g., acetaldehyde and monoamine oxidase inhibitors in tobacco smoke; Guillem et al., 2005; Talhout et al., 2007). Moreover, the potential influence of environmental factors in determining abuse liability, such as social acceptance, product marketing, and ease of purchase, has also been recognized (Carter and Griffiths, 2009; O’Brien, 1996; World Health Organization, 2008).
Each component of abuse liability may be assessed using distinct methodology (Carter et al., 2009; Carter and Griffiths, 2009). For example, pharmaco-kinetic and -dynamic factors, such as the magnitude and speed of onset of drug effects, may be tested using biological sampling and self-report measures; the likelihood of physical dependence may be examined through the assessment of withdrawal following acute or chronic drug administration;
and reinforcing value may be examined through simulated or real drug self- administration procedures. Assessment of self-administration provides the broadest understanding of abuse liability, as decisions to self-administer a drug (assuming sufficient prior experience) are influenced by all additional factors in the construct (e.g., speed of onset and magnitude of effects).
Own-price elasticity of demand
Drug self-administration can be uniquely understood with behavioral economics. The basis of this unique understanding is through assessment of a drug’s own-price elasticity of demand, or sensitivity of consumption of a commodity (e.g., cigarettes) to increases in unit price. Unit price, a cost–
benefit ratio, is broadly defined as the behavioral or monetary cost required to obtain one unit of a commodity (DeGrandpre, Bickel, Hughes et al., 1993;
Hursh, 1980). Thus, the degree to which consumption of a given commodity is sensitive to changes in its own price (i.e., own-price elasticity) provides an index of that commodity’s value, with lower levels of elasticity representing a greater willingness to defend consumption of that commodity against increasing price (Hursh and Roma, 2013; Hursh and Silberberg, 2008; Hursh and Winger, 1995). When applied to drugs of abuse, measures of elasticity reflect one component of abuse liability (Carter et al., 2009).
Own-price elasticity of demand is typically quantified using methods of least squares regression. Historically, linear demand models have been used to estimate demand elasticity in the behavioral economics laboratory (see Hursh et al., 1989 for a description of the linear model). However, an exponential model of demand is now more frequently used, which describes the nonlinear relationship between consumption and price evident at the individual-subject level (Hursh and Silberberg, 2008):
where Q is consumption, P is price, k is span of consumption in log10 units, and Q0 and α are free parameters describing demand intensity (i.e., consumption at minimal price) and demand elasticity, respectively. Here, we note that the Q0 and α parameters may be interpreted independently of one another.
Although linear and nonlinear equations are useful in quantifying demand elasticity, we have recently proposed and validated a modified version of Equation 1 that remediates problems posed when participants contribute zero values in consumption (e.g., at the highest prices). Because zero values cannot be log transformed, application of Equation 1 to demand data requires either elimination of zeroes or replacement of zeroes with nominal values (e.g., 0.01); both of these strategies, however, can drastically affect demand estimates depending on the number of obtained zero values or the precise nominal values used (see Koffarnus, Franck, et al., 2015). Use of the following modified equation, however, eliminates these problems (Koffarnus, Franck, et al., 2015):
in which all parameters are otherwise identical to those described for Equation 1; however, the equation appears as an exponent of 10, removing the need for log transformation and therefore allowing analysis of unaltered consumption values, including zeros.
An additional analytical challenge comes when participants occasionally contribute consumption data that are not systematically affected by price (i.e., either invariant, highly variable, or no consumption across prices). In some cases, such data may be an important property of the commodity or population under investigation; however, in others, such data are likely produced by extraneous variability not germane to research aims (e.g., inattention or failure to understand task instructions) and may obscure effects of the variable(s) under investigation. We recently proposed and validated a method for identifying such data, assuming only that consumption of a commodity should decrease monotonically with increasing price (Stein et al., 2015). Use of this method, as well as related guidelines for handling nonsystematic data, may serve to enhance accuracy of demand estimates and therefore improve the ability to detect effects of clinical treatments and other variables that influence demand for tobacco products.
Cross-price elasticity of demand
In contrast to own-price elasticity, measures of cross-price elasticity describe the interaction between the reductions in consumption of one price- manipulated product (e.g., cigarettes) and the consumption of one or more constant-priced alternative products (e.g., e-cigarettes). The economic relationship between these products may be categorized along a continuum in which the constant-priced alternative products serve as either: (1) substitutes, at one end of the continuum (i.e., consumption of a constant-priced alternative product increases to compensate for reduction in consumption of a price-manipulated product), (2) independents (i.e., consumption of the constant-priced alternative product does not change), or (3) complements, at the other end of the continuum (i.e., consumption of both the constant-priced alternative product and the price-manipulated product decrease).
Like own-price elasticity, cross-price elasticity of demand has been historically quantified using linear models (Allison, 1983; DeGrandpre et al., 1994); however, reflecting its nonlinear form, cross-price elasticity may be
quantified by fitting consumption curves for constant-priced alternative commodities using the exponential equation (Hursh and Roma, 2013, 2015):
in which the free parameter Qalone is asymptotic consumption of a constant- priced, alternative commodity when the price (P) of a price-manipulated commodity is infinitely high; I is the y-axis span of consumption in log10 units, describing the degree of interaction between commodities; and b is sensitivity of alternative product consumption to reductions in consumption of the price-manipulated commodity. Here, negative values of the I term indicate substitution between commodities and positive values indicate complementarity.
In the following sections, we review historical and contemporary methods used to assess the behavioral economics demand measures outlined above.
We organize these sections according to the four epochs of methods innovation, highlighting important findings from each epoch that have contributed to our understanding of tobacco regulatory science.