Left-digit pricing effects in a high-resolution examination of hypothetical operant demand for alcohol

Highlights

• Left-digit effects are known to influence consumer decision making.

• There is limited behavioral economic research on the left-digit effect.

• This study shows a high-resolution account of left-digit effects on an alcohol purchase task.

Abstract

Behavioral economic measures have demonstrated marked success in the evaluation of consumer choice. Field-standard operant demand curve analyses provide a valuable model of resource allocation via responses to maintain “free-rate” commodity use or consumption. This demand analysis thereby provides a behavioral complement to consumer science techniques. Despite apparent congruence of operant behavioral economics and consumer science, the left-digit effect represents one area of research predominantly untouched by behavioral economic investigation. Previous efforts have applied the hypothetical purchase task to map the effect of a changing left-digit on subsequent purchase decisions. The current study extends investigation of the phenomenon to responding on the validated Alcohol Purchase Task. Introduction of a high-density price structure revealed evidence of digit sensitivity, wherein demand elasticity was disproportionately affected at and around whole-dollar changes. That responses were influenced by small shifts in pricing implies a possibility for policy-level modulation of alcohol ingestion without need to increase commodity price beyond unit elasticity. Capture of digit preference in a self-report framework speaks to the sensitivity of purchase task methodology to detect small, aberrant changes in consumer product perception. Behavioral economic researchers should consider this extent of sensitivity when interpreting results of hypothetical purchase task investigations.

Introduction

Since the early days of Behaviorism, the translation of psychological principles to marketing has been successful in rendering unique insight concerning consumer behavior (see Link, 1932). John B. Watson produced works and speeches arguing for the synthesis of behavioral science – concepts that embody behavior analysis – with advertising as a means of capturing a greater proportion of the consumer population as buyers for any given product (Kreshel, 1990). The pairing of social science with the body of knowledge already established in sales set the stage for a new strategic perspective that continues to advance market industry (e.g., Curry et al., 2010; Foxall, 1994; see also Foxall, 2016; Hantula and Wells, 2013).

The study of consumer choice from a behavioral perspective has since evolved – in part – to embody a behavioral economic framework (Foxall, 2017). Broadly, behavioral economics is an approach to consumer evaluation that considers the impact of relevant factors (e.g., response cost) on decision making (Thaler, 2015). Distinct traditions of behavioral economics have grown from these origins: a cognitive-perceptual approach made famous by Nobel Laureates like Daniel Kahneman and Richard Thaler (e.g., Kahneman and Tversky, 1979; Thaler, 1999) that focuses on concepts such as mental accounting and bounded rationality, and operant behavioral economics - an approach grounded in operant learning theory and initially theorized by operant scientists (e.g., Foxall, 1990; Green and Kagel, 1987; Hursh, 1984, 2014).

Application of operant behavioral economics and its respective analyses has historically been used to model choice related to consumer health and well-being, as well as outcomes that influence decision making in a range of samples, human and non-human (e.g., Bickel et al., 2016, 1999; Hursh, 1984, 2014; Lamb et al., 2016; Odum, 2011). Operant demand assessment – a method for predicting consumer valuation of a target commodity – models consumer choice via observed (i.e., revealed) or reported (i.e., stated) responses to defend and maintain “free-rate” obtainment of a commodity amidst escalating response requirements (Hursh, 2014; Hursh and Silberberg, 2008). Prototypical demand assays measure consumption across a range of prices and fit to quantitative models to render the demand curve. Resulting curves thereby provide point elasticity estimates across the entire range of potential prices. These curves are logarithmically scaled and characteristically model curvilinear exponential decay, where the left-most portion remains relatively stable (i.e., inelastic demand; < 1 unit change in consumption per 1 unit change in price) and the right-most portion diminishes quickly (with respect to increasing response cost; i.e., elastic demand; > 1 unit change in consumption per 1 unit change in price; Hursh and Roma, 2016; Hursh and Silberberg, 2008; Watson and Holman, 1977).

Traditional operant demand assessments employ work tasks, during which organisms must meet effort requirements to defend consumption of a target commodity (Hursh, 1978). More recent work has advanced the use of the hypothetical purchase task (HPT) with a strictly human sample as means of more rapid – or ethical¹ – collection of consumer data (Jacobs and Bickel, 1999; Petry and Bickel, 1998; see also Roma et al., 2016, 2017). The HPT is a survey-type instrument through which participants report their anticipated resource allocation toward defending access to a target commodity at each of a number of ascending prices. To date, application of hypothetical purchase task methodology extends to a variety of commodities with abuse liability, not limited to alcohol (e.g., Murphy and MacKillop, 2006), cigarettes (e.g., MacKillop et al., 2008), cannabis (e.g., Aston et al., 2015), cocaine (e.g., Bruner and Johnson, 2014), opioid medication (e.g., Schwartz et al., 2019), internet use (e.g., Acuff et al., 2018), pornography (e.g., Mulhauser et al., 2018) and ultraviolet tanning (e.g., Reed et al., 2016). Many of these tasks have been shown to be commodity specific and are therefore sensitive to individual characteristics of the commodity of interest, such as dependence (see Aston and Cassidy, 2019).

Hypothetical demand measures have received a substantial degree of attention in the behavioral economic community for their reliability and validity (Amlung et al., 2012; Murphy and MacKillop, 2006). Specifically, Amlung et al. (2012) analyzed correspondence between hypothetical demand for alcohol and actual alcohol purchasing. Purchases reported on the demand assessment closely modeled choices made when presented outside the hypothetical circumstance, thereby demonstrating adequate predictive ability. Additionally, validity has been analyzed using a one-month follow up such that baseline intensity predicted alcohol consumption and subsequent problems (Murphy et al., 2015; Dennhardt et al., 2015). Indices derived from the APT also correlate with other types of self-report measures of alcohol consumption (MacKillop et al., 2010; Murphy and MacKillop, 2006; Bertholet et al., 2015). For a further review on the assessment of psychometric properties with the APT, see Kaplan et al. (2018). Subsequently, these methods influenced a growing body of literature concerning their implementation and expected outcomes (e.g., Stein et al., 2015). Ongoing investigation underscores the importance of clarity and consistency of task constraints to leverage the weight of possible framing-effects on respondent behavior (Gentile et al., 2012; Kaplan and Reed, 2018; Kaplan et al., 2017; Skidmore and Murphy, 2012; see also Moore, 2010; Urcelay and Miller, 2014; Weatherly, 2014).

One consumer decision influence remaining relatively unexplored is the left-digit effect, or the phenomenon in which disproportionately large magnitude differences in consumption occur proximate to price transition from one whole-dollar amount to the next. Left-digit effect changes are representative of characteristically irrational deviations from expected purchasing decisions, observable when the total change in response cost demonstrates just a small magnitude difference. The effect is hypothesized to occur as a result of greater attending to the left-most digit while at least partially ignoring the subsequent values (Lacetera et al., 2012; Thomas and Morwitz, 2005; Brenner and Brenner, 1982) and has been observed in decision making beyond monetary expense, suggesting some potential extension to other relevant domains (e.g., interpretation of odometer readings; Lacetera et al., 2012).

Marketing and consumer retail research has consistently demonstrated support for the importance of the left-digit effect in purchasing considerations (Gendall et al., 1997; Manning and Sprott, 2009; Schindler and Kibarian, 1996; Stiving and Winer, 1997; Thomas and Morwitz, 2005). The history and theories of left-digit sensitivity have differing origins (e.g., standardization of the British pound after the Civil War led to odd-pricing which became a status symbol of imported goods, compared to domestic goods; see Gendall et al., 1997 for theories) with little agreement; regardless, the effect appears to be pervasive as a market strategy for enhancing consumer perception of saleable goods (Gendall et al., 1997). With such persistence of left-digit-sensitive pricing in everyday sales, left-most price figures appear to exhibit greater control over behavior as values are paired with discounted goods. Such an effect could have strong implications for purchase task methodology, given the reliance upon respondent interpretation of the various media or structures by which price arrays can be arranged and presented (e.g., Reed et al., 2014; Roma et al., 2016). Further, capitalization upon price sensitivity in the global market may have implications for policy-enforced pricing of commodities with high abuse liability (see Hursh and Roma, 2013). The left-digit effect thereby embodies an important target for behavioral economic study – one that lends itself well to simulation in a hypothetical purchase task where prices can be manipulated without market disruption.

In their novel capture of left-digit effects by operant demand methodology, MacKillop et al. (2012) examined points of price sensitivity and insensitivity when consumers simulated purchasing decisions regarding cigarettes. The MacKillop et al. study used a high-density price sequence to yield a high-resolution view of consumer choice. Examination of unit elasticity revealed greater-than-average sensitivity to cigarette price when values escalated to the next whole-dollar amount – significantly fewer cigarettes purchases immediately followed a left-digit shift, despite the relatively small change in price (i.e., $0.20 or less). Such consumer decisions depict changes in reported purchases at rates that exceed the corresponding change in response cost. The authors conclude in favor of the highlighted phenomenon as a tool for use in public policy – carefully engineered price structures could reduce cigarette smoking in the general public, should regulation maintain market price at values just beyond the nearest whole-dollar amount.

Advancing the use of left-digit-sensitive pricing as a tool for socially meaningful change, MacKillop et al. (2014) extended their previous work (2012) by examining the influence of left-digit cigarette price changes in a high-resolution snapshot of self-reported motivation to quit habitual smoking. Participants moved through a high-density array of prices and, much like in a typical HPT, were prompted to imagine each value as reflecting real-world cigarette costs; instead of consumption, respondents reported their prospective desire to quit smoking under said imagined constraints. In addition to the expected positive relation between product pricing and the corresponding probability of smoking cessation, the authors observed – on average – significantly greater changes in reported quit probability following left-digit shifts as compared to the average change across prices.

The purpose of the current study is to further explore the influence of left-digit sensitivity from within a behavioral economic framework. We move to expand the demonstration of this effect to a novel commodity – alcohol – using the previously validated Alcohol Purchase Task (APT; Kaplan and Reed, 2018; Murphy and MacKillop, 2006; Murphy et al., 2009). Given the experimental validity of the APT (see review by Kaplan and Reed, 2018), studying the left-digit effect with alcohol may provide further evidence for the effect within highly dense price-sequences (MacKillop et al., 2012, 2014). Through application of methods similar to those used in past demonstrations, we hypothesized observation of greater than average reductions in alcohol purchasing at or around whole-dollar price changes. As such, we expected irrationally greater sensitivity to price relative to other price changes in a high-resolution view of alcohol demand at values adjacent to each whole-dollar increment.