Delay discounting
Delay discounting reflects the extent to which rewards are devalued based on their delay in time. It indexes preferences between smaller-immediate and larger-delayed rewards with greater immediate reward preference often construed as a form of impulsivity (Madden and Bickel, 2009). Delay discounting is commonly assessed via a series of dichotomous choices
between immediate and delayed rewards (e.g., $40 today vs. $100 in a month). Although money is the most common reward, DD measures for drugs of abuse and health outcomes have proven useful in addictions research (e.g., Baker et al., 2003; Petry, 2001). An individual’s rate of DD is typically quantified using a hyperbolic discounting function (i.e., k) that reflects the rate at which delayed rewards lose their value (Mazur, 1987). Larger k values reflect steeper discounting and greater impulsivity (see Figure 4.1A). Of note, other model-free methods are also used to quantify DD, including impulsive choice ratios (e.g., Mitchell, 1999) and area under the curve (e.g., Myerson et al., 2001). An overview of the different indices of DD is provided in Table 4.1.
Steep DD is a common clinical characteristic of addiction given that these individuals repeatedly seek smaller-immediate gains from alcohol and drug use over larger-delayed gains (e.g., financial stability, relationships, health).
Across a large number of studies, individuals with addictive disorders have been shown to exhibit more precipitous DD relative to controls. This includes individuals with alcohol use disorders (e.g., Petry, 2001), cocaine dependence (e.g., Coffey et al., 2003), and opiate dependence (Madden et al., 1999). The notable exception to this pattern is cannabis, with prior research reporting no difference in DD between marijuana abusers and controls (Bidwell et al., 2013; Johnson et al., 2010; Mejía-Cruz et al., 2016). Significant correlations between steeper DD and higher addiction severity and greater quantity/frequency of use have been reported in several studies and with several substances (e.g., Kirby and Petry, 2004; MacKillop, Miranda et al., 2010; Petry, 2001).
A meta-analysis synthesized the findings of categorical studies comparing individuals exhibiting addictive behavior and controls (MacKillop et al., 2011). The meta-analytic sample comprised 57 comparisons, with >3,300 participants across six different categories of addictive behaviors (alcohol, tobacco, stimulant, cannabis, opiate, and gambling disorder). Across all studies, a highly significant, medium magnitude effect size (Cohen’s d = .62) difference in DD was found between groups exhibiting addictive behavior and controls. Furthermore, significantly larger effect sizes were found for studies examining individuals meeting clinical criteria for substance use disorder (d = .67) compared to subclinical samples of heavy or recreational users who do not meet clinical criteria (d = .46), and for studies using higher- resolution discounting measures compared to course assessments with only
one or two items. Finally, the meta-analysis also suggested only modest influence of unpublished findings (i.e., publication bias) on the aggregated results. The results of this meta-analysis provide strong evidence that steep DD is a core feature of addictive behavior that is consistent across type of addiction and is meaningfully related to addiction severity.
FIGURE 4.1 Prototypic delay discounting and demand curves. Panel A depicts hyperbolic delay discounting curves illustrating steep (more impulsive) and shallow (less impulsive) discounting in two hypothetical individuals. Panel B depicts demand curves illustrating the individual indices of demand and higher and
lower levels in two hypothetical individuals.
TABLE 4.1 Behavioral economics measures and associated indices of impulsivity and demand
Assessment/Index Description/Definition I. Delay Discounting
Tasks
Behavioral economic measure of impulsivity;
characterizes preferences for smaller immediate rewards over larger delayed rewards
k
Fitted parameter derived from hyperbolic
discounting function that reflects overall rate of discounting of delayed rewards; larger k values reflect steeper discounting (i.e., greater
impulsivity) Impulsive choice ratio
Simple ratio of immediate reward choices to delayed reward choices; larger values reflect greater immediate reward preference (i.e., greater impulsivity).
Area under the curve (AUC)
Model-free index of delay discounting via geometric quantification of area under the
discounting curve; AUC is inversely related to k, such that smaller values reflect steeper
discounting (i.e., greater impulsivity) II. Purchase Tasks (e.g.,
Alcohol Purchase Task;
Marijuana Purchase Task; Cocaine Purchase Task)
Behavioral economic measure of substance demand; characterizes cost–benefit decision- making via estimated consumption at varying levels of price
Intensity Consumption at zero price
Breakpoint First price that suppresses consumption to zero Omax Maximum expenditure across prices
Pmax
Price at which demand curve first becomes elastic;
also price associated with Omax
Elasticity Proportional slope of the demand curve
The majority of prior research examining DD in samples of individuals with addiction has used cross-sectional designs which do not allow for determination of the causal relationship between steep DD and addiction (de Wit, 2009). Retrospective studies have found that steeper DD in adolescence predicts earlier onset of alcohol use disorder symptoms (Dom et al., 2006;
Kollins, 2003). Other studies have found that DD prospectively predicts severity of alcohol use over a six-month period (Fernie et al., 2013).
However, some studies have found that the DD rate may become less impulsive following abstinence. Both ex- and never smokers have been shown to exhibit less steep discounting for monetary rewards relative to current smokers (Bickel et al., 1999). Another study found that discounting rates for delayed health outcomes among ex-smokers were between current and never smokers, although not significantly different from either group (Odum et al., 2002). Finally, a recent longitudinal study found that smokers who maintained abstinence across a 12-month period following a smoking cessation intervention exhibited significant decreases in monetary DD, while no change occurred in individuals who resumed smoking (Secades-Villa et al., 2014). However, DD rates were not affected by a brief alcohol and drug use intervention in college student drinkers (Dennhardt et al., 2015). These studies offer mixed support for reversal of steep DD following abstinence, with the most promising data to date pertaining primarily to nicotine dependence. Further understanding the causal relations between DD and addiction is an important priority and may help clarify the role of DD in preventing the onset of addictive disorders as well as developing more effective treatments (see below).
Demand
A common clinical feature of alcohol and drug addiction is the persistent overvaluation of alcohol and drug rewards relative to other reinforcers (Bickel et al., 2014). This overvaluation is captured via the behavioral economics concept of demand, which reflects how much of a commodity is consumed at a given price. Historically, demand was typically measured via
operant self-administration paradigms (Higgins et al., 1994; Mello and Mendelson, 1965), but recent research in human samples has led to the development of purchase tasks in which participants estimate consumption at varying levels of price. Purchase tasks have been developed to assess demand for alcohol (Murphy and MacKillop, 2006), cigarettes (Jacobs and Bickel, 1999; MacKillop et al., 2008), cocaine (Bruner and Johnson, 2014), and marijuana (Collins et al., 2014). These measures ask participants to report how many units of the commodity (e.g., number of alcoholic drinks, marijuana joints, or hits of cocaine) they would purchase and consume at a range of prices (e.g., “How many drinks would you consume if they cost $4 each?”). Plotting consumption by price yields a demand curve (Figure 4.1B) that can be used to generate five conceptually related indices of demand (Table 4.1). Of note, the axes in Figure 4.1B are reversed from the typical demand curve format initially presented by Marshall (Friedman, 1949;
Marshall, 1898); however, the convention in psychology is to plot price along the x-axis and consumption on the y-axis. Furthermore, in psychological applications of behavioral economics, the elasticity index (Table 4.1) is viewed as a measure of proportional price sensitivity or overall slope of the demand curve (e.g., Hursh and Silberberg, 2008), which differs somewhat from traditional economic definitions of elasticity. Detailed descriptions of equations used to model elasticity from behavioral economics purchase task data are provided in Hursh and Silberberg (2008) and Koffarnus et al. (2015).
Prior research examining demand for addictive substances has focused on two primary methodologies. The first is trait-based demand, reflecting estimated typical level of demand, and the second is state-based demand, reflecting estimated consumption in specific conditions or situations. In the case of trait-based demand, a growing literature has revealed robust associations between elevated demand for alcohol and drugs and quantity/frequency of use and clinical severity. For instance, higher alcohol demand in alcohol purchase tasks has been consistently associated with greater quantity/frequency of alcohol consumption, heavy drinking, and alcohol use disorder severity (Bertholet et al., 2015; MacKillop, Miranda et al., 2010; Murphy and MacKillop, 2006). In the case of drug demand, increased cocaine demand was shown to be correlated with greater cocaine use in cocaine-dependent individuals (Bruner and Johnson, 2014), and elevated marijuana demand has been shown to be related to greater marijuana use and symptoms of marijuana dependence (Aston et al., 2015; Collins et
al., 2014).
The second domain of research in this area has focused on state-based influences on demand. To date, the main focus of this work has been alcohol demand, with a relatively smaller number of studies on tobacco demand (Acker and MacKillop, 2013; MacKillop et al., 2012). In the context of alcohol demand, one finding that has been replicated across a number of studies is that the relative value of alcohol is increased by alcohol-related environmental cues. Specifically, MacKillop, O’Hagen et al. (2010) observed significant increases in several indices of alcohol demand in heavy drinkers following a laboratory alcohol cue exposure. These findings were replicated in a subsequent study (Amlung et al., 2012) that also demonstrated close correspondence between estimated demand on an alcohol purchase task and actual alcohol consumption during a laboratory self-administration protocol.
In another set of studies, negative affect and stress inductions significantly increased alcohol demand in heavy drinkers (Amlung and MacKillop, 2014;
Owens, Ray, and MacKillop, 2015). Finally, Amlung et al. (2015) recently demonstrated that demand for alcohol is dynamically increased by acute alcohol consumption, which may contribute to loss of control over drinking that can occur following intoxication. These findings suggest that state-based indices of demand may complement existing measures of alcohol motivation, such as subjective craving. Importantly, no studies have examined state-based influences on demand for illicit drugs, which is an important target for future research.