The impact of product information and trials on demand for smokeless tobacco and cigarettes: Evidence from experimental auctions

Abstract

Introduction

Epidemiological and toxicological evidence suggests lower risk of smokeless tobacco (ST) products compared to cigarettes. Less is known, however, about consumer perceptions and use of novel forms of ST, including snus and dissolvable tobacco.

Methods

In this study, we conducted in-person experimental auctions in Buffalo, NY, Columbia, SC, and Selinsgrove, PA with 571 smokers to test the impact of information and product trials on smokers’ preferences. Auctions were conducted between November 2010-November 2011.

Results

We found no evidence of an impact of product trials on demand in our auctions. Anti-ST information increased demand for cigarettes when presented alone, but when presented with Pro-ST information it decreased demand for cigarettes. It did not decrease demand for ST products. Anti-smoking information increased demand for ST products, but did not affect cigarette demand.

Conclusions

These findings suggest that credible and effective communications about tobacco harm reduction should reinforce the negative effects of smoking.

INTRODUCTION

Beginning in the mid 2000’s, cigarette manufacturers such as Philip Morris and RJ Reynolds began to market novel smokeless tobacco (ST) products of varying forms, often as brand extensions from their existing cigarette products. Examples of such products include Camel Snus, Marlboro Snus, Camel Dissolvables, and Marlboro Sticks. Epidemiological and toxicological evidence suggests that exclusive use of ST products can be substantially less harmful than cigarette smoking (Stratton et al. 2001, Tobacco Products Scientific Advisory Committee 2012, Levy et al. 2004, Levy et al. 2006, Royal College of Physicians and Tobacco Advisory Group of the Royal College of Physicians 2007). Nevertheless, traditional ST is associated with heart disease, stroke, and some cancers (particularly oral and pancreas). ST is also addictive (Critchley and Unal 2003, 2004; Henley et al. 2005, International Agency for Research on Cancer 2007), and the long-term health risks of the novel ST products are as yet unclear.

Smokers generally respond favorably to products described as less hazardous alternatives to smoking (Timberlake 2009 and Heavner et al. 2009), but interest may not reliably foreshadow use. Evidence from ST marketing suggests that use of novel ST products is relatively low (Biener et al. 2011). Whether or not a smoker will try ST may be influenced by the type of information available to aid this decision—in the public health community, there are competing calls for pro- and anti-ST messaging (Rodu and Godshall 2006, Phillips and Heavner 2009, Kozlowski 2006, Zeller and Hatsukami 2009, and Tomar et al. 2009). Whereas expert consensus suggests that these products are likely to represent less than 10% of the risks of cigarettes (Levy et al. 2006), many smokers view them as equally or more harmful (O’Connor et al. 2007, Borland et al. 2011, Tomar and Hatsukami 2007). For example, Borland et al. (2011) found that only 17% of smokers believed ST to be less harmful than cigarettes, with no significant change in this belief from 2002 to 2009. Some have suggested that smokers be provided with credible, scientifically accurate relative risk information to encourage their use of these less harmful alternatives (Kozlowski and Edwards 2005, Ferguson et al. 2011).

Information isn’t the only suggested method for increasing ST adoption. Smokers’ willingness to continue to buy a product may be influenced by the sensory characteristics of the product, which can only be ascertained through trial. RJ Reynolds has used free and reduced-cost sampling (via coupons) as a means to increase purchase of Camel Snus, for example. Product-specific concerns such as aesthetics, consumer subjective responses, and nicotine delivery may play a role in the low adoption of ST among smokers (Sami et al., 2012). Other data suggest that brief trials of products (the opportunity to taste a new product) may familiarize smokers with novel nicotine delivery methods, encouraging further sampling and perhaps increasing likelihood of adoption of a suitable product (Borland et al., 2011; O’Connor et al 2011; Schneider 2011). The tobacco industry did make a modest effort to provide free trials in the US with snus, although the effectiveness of their strategy is debatable (Rogers et al. 2010).

A web survey study design revealed that, when briefly exposed to advertisements, a nationwide panel of participants preferred nicotine lozenges over ST products. Out of a list of three alternative products (Commit Lozenges, Camel Snus, and Camel Dissolvables), lozenges were most often nominated as the product participants were most willing to try (O’Connor et al. 2012). Product demand and cross price elasticity from the same study showed that participants were willing to substitute all three of the alternative products for cigarettes as cigarette price increased, albeit weakly (O’Connor et al., in press). Viewing advertisements seemed to have little impact on willingness to try the product as a whole; however, it seems plausible that gravitation towards lozenges may stem from prior “quit smoking” marketing and familiarity. Outside of this study, little work has been done to assess smokers’ demand for ST products and in particular how this demand is affected by product trials and information.

In the study described here, we conducted in-person experimental auctions to test the impact of information and product trials on smokers’ preferences. The study was designed to address both the cognitive (e.g., beliefs about health risks) and sensory aspects of product evaluation to determine which plays the greater role in demand for novel ST products (Rees et al. 2009). Experimental auctions, which are a behavioral economic mechanism, allow estimation of demand in a more-controlled setting than self-reported, hypothetical intentions to purchase products, and, as a result, provide more accurate assessments of preferences. These have been used extensively to examine issues related to food demand (for a review, see Lusk and Shogren, 2007). Recently, experimental auctions been used to assess US smokers’ demand for low and no nicotine cigarettes (Monchuk et al. 2007), demand for PREP cigarettes (Rousu, Nonnemaker, and Farrelly 2011), as well as demand among adult smokers in Mexico (Thrasher et al. 2007) and the US (Thrasher et al. 2011, Rousu and Thrasher 2012) for cigarettes with text-only warning labels, pictorial warnings labels, and “plain” packaging (i.e., elimination of brand colors and logos from packs). Our goal in the current study is to examine the extent to which providing information and product trials affects smokers’ demand for novel ST products and cigarettes, as well as to provide evidence on whether information or product trials could prompt smokers to use ST.

METHODS

Participant recruitment and sample size

The study protocol was approved by the IRBs at the University of South Carolina, Roswell Park Cancer Institute and Geisinger Medical Center. Participants were recruited by radio ads, newspaper ads, and flyers in Buffalo NY, Columbia SC, and Selinsgrove PA. These sites were chosen because they differ substantially in prevalence of ST use (NY=2.3%; PA=4.4%; SC=3.6%) (Centers for Disease Control and Prevention 2013), cigarette taxes (NY=$4.35; PA=$1.60; SC=$0.57), and clean indoor air policies (NY=comprehensive smokefree policies; PA= exemption for bars; SC=exemptions for bars), as well as because they differed from one another in terms of racial diversity and urbanicity. They were also chosen in part for convenience, as study authors were from these locations, making oversight and administration of the experiments easier. Eligible study participants were 18 and older, currently smoked, were not currently using nicotine replacement products, and had no major medical issues that would warrant exclusion. Participants were paid $50 for their participation, and sessions usually lasted about an hour. Auctions were conducted with eight to sixteen participants at a time, and a total of 571 smokers participated between November 2010-November 2011.

The products

Participants bid on three cigarette alternatives: Camel Snus, Ariva Dissolvable Tobacco, and Nicorette Mini-lozenge (see Figure 1). These products were chosen to cover three distinct product styles (pouched ST, dissolvable tobacco, and medicinal nicotine, respectively), and which were all available in the open market. Participants also bid on Marlboro brand cigarettes, either the red, menthol, gold (light), or menthol gold (light) variety, depending on their individual preference.

Figure 1.

The smokeless tobacco products up for auction

Experimental conditions

We sought to assess demand for the three products, relative to cigarettes, under alternative treatments. Treatment assignment was at the group level – all participants at a given auction session received the same treatment to facilitate the auction protocol. The treatments fell into two broad classes: information (5 conditions) and product trial (3 conditions). The treatments were as follows:

1. Participants received no information and weren’t offered a trial of ST. (N=62)

2. Participants received only pro-ST information (N=60)

3. Participants received only anti-ST information (N=67)

4. Participants received both pro-ST and anti-ST information (N=61)

5. Participants received anti-cigarette information (N=60)

6. Participants received pro-ST and anti-cigarette information (N=65)

7. Participants were offered a trial of Camel Snus (N=64)

8. Participants were offered a trial of Nicorette (N=67)

9. Participants were offered a trial of Ariva Lozenges (N=65)

Table 1 shows the number of participants in each of nine treatments and the distribution of participants per treatment in each site. The information was presented in three separate brochure formats of similar style and tone, based on peer-reviewed literature and created by the study investigators (see Appendix A-Appendix C). The first featured anti-smoking information (emphasizing the harms of smoking), the second presented anti-oral nicotine information (emphasizing the harms of nicotine and ST products), and the last included pro-oral nicotine information (emphasizing the lower relative risks compared to cigarettes). Participants were asked to read the information provided in silence, to allow each participant to process the information with minimal influence from other study participants. If participants were in treatment 4 and 6 and received two sources of information, both were given simultaneously. After five minutes, participants were instructed to place their information brochures to the side and the practice auction began.

Table 1.

Sample sizes and sample characteristics, overall and by treatment group.⁺

	Overall (N=571)	Control (N=62)	Only pro-st (N=60)	Only anti-st (N=67)	Both pro- and anti- ST (N=61)	Anti- cig (N=60)	Anti-cig & pro-ST (N=65)	Snus trial (N=64)	Nicorette trial (N=67)	Ariva trial (N=65)
Observations – NY*	209	24	24	25	21	22	23	21	24	25
Observations – SC	195	22	19	25	21	21	20	23	25	19
Observations – PA	167	16	17	17	19	17	22	20	18	21
Race_white**	64%	66%	62%	57%	70%	68%	65%	67%	66%	60%
Race_black	27%	31%	22%	37%	18%	20%	31%	23%	27%	32%
Race_other	8%	3%	17%	6%	11%	12%	5%	9%	7%	8%
Age – under 30	37%	27%	42%	26%	39%	42%	48%	36%	39%	51%
Age – 30 to 50	40%	35%	38%	52%	38%	35%	42%	41%	36%	38%
Age – over 50	23%	37%	20%	31%	23%	23%	11%	23%	25%	11%
Female	40%	39%	37%	43%	34%	38%	31%	41%	51%	45%
Income_below 30K	53%	56%	47%	54%	52%	52%	45%	63%	49%	58%
Income – between 30K−60K	17%	19%	5%	7%	10%	23%	17%	9%	10%	11%
Income – over 60K	11%	10%	23%	21%	11%	18%	17%	11%	13%	17%
Income – chose not to reveal	19%	15%	25%	18%	26%	7%	20%	17%	27%	14%
Participant is moderately or very worried about future quality of life	60%	61%	72%	64%	49%	58%	52%	59%	64%	58%
Participant has used smokeless tobacco in the past	45%	39%	47%	43%	30%	42%	55%	50%	43%	38%

⁺Auctions conducted from November 2010-November 2011.

For the free trial treatments, participants were offered the opportunity to try either Snus, Ariva, or Nicorette, depending on assigned condition (See Table 1 and Figure 1). We asked participants if they would like to try the product but did not require trial to continue in the study. We asked for participants who tried the products to keep their reactions to themselves until after the experiment ended. The monitor reminded participants of this before, during, and after the product trial. For analytic purposes, we distinguished those who declined the trial offer from those who accepted it (since the non-triers would bias the effect toward the null). While participants placed a bid on each product, only one product was randomly chosen as the binding product (i.e., the product that would actually be auctioned). This product was selected after all bidding by a random draw to ensure participants would not decrease their bids due to thinking they might win more than one tobacco product (Rousu, Beach, and Corrigan 2008). However, by randomly choosing one of the products through a draw, participants were forced to take each auction round seriously, as they potentially could have won any of the four products, depending on which product was drawn to be auctioned.

Experimental design

Data were collected using the random n^th price auction mechanism (Shogren et al. 2001 or Rousu et al. 2007), in which participants are initially given enough money to compensate for their time and to provide them with more than enough money to pay the “clearing” price for the product of interest. Participants are told, both with written instructions and with an oral description, that this auction is different from other auctions in that they can only bid once (on a product) and it is in their best interest to submit a bid equal to the full price they would pay for the product. Great care is taken to explain the procedures to participants so they understand and time is allowed for questions. After the questions, participants take a short quiz on the auction mechanism. The auction moderator went through the answers for all participants in the group to ensure everybody understood how the auction worked. See appendix D and appendix E for the written instructions participants received on the auction mechanism along with the quiz. Once all bids are collected, bids are sorted from highest to lowest. After eliminating the highest bid, one bid is then selected randomly. The (n-1) participants that bid more than this randomly selected price purchase the product, paying the price of the selected n^th highest bid; the participants who bid less than (and the participant whose bid is equal to) the n^th highest bid do not purchase the product. In this mechanism, a participant will not pay more than their submitted bid for the product. This auction is “demand revealing” in that it is in a participant’s best interest to bid his or her true value (demand) for the product because the amount the auction winners pay is determined by another subject’s bid, not their bid. Someone who bids higher than her true value for the product could end up paying more than that true value, while someone who bids lower than her true value may miss out on a profitable purchase if the randomly selected binding price is less than her true value but higher than the bid she submitted.

Procedures

After participants arrived and signed a consent form, they filled out a brief survey on smoking behavior (Step 1). Next (Step 2), participants received a detailed explanation of the auction mechanism (both orally and in writing), with an emphasis that it was in their best interest to bid their true value for the products. Next, participants participated in a practice auction for two candy bars (Step 3) that demonstrated the real procedures by having participants place bids for different candy bars in different rounds, including random selection of the binding product. In Step 4, participants received their randomly assigned treatment (information, trial, or nothing). Then (Step 5), participants placed separate, private bids on each of the four products. There were not enough groups to adequately randomize the order of bidding within and across treatments, so participants in all treatment always placed their first bid on the Snus, their second bid on the Ariva, their third bid on the Nicorette, and their final bid on the cigarettes. After all four bids were submitted, a random draw was conducted to determine which product was the binding product, followed by a random draw to determine the n^th price (Step 6). This determined who won products, which product, and how much the winners would pay. Finally (Step 7), participants filled out a post-auction questionnaire, winners exchanged money for their product, and the experiment ended.

This type of experimental mechanism has been shown to have credibility in predicting consumer choices in the marketplace (i.e., have external validity). Chang, Lusk, and Norwood (2009) tested both hypothetical and non-hypothetical mechanisms and found that a non-hypothetical experiment similar to what we are proposing outperformed hypothetical mechanisms and did a good job of predicting retail sales. Ding et al. (2005) showed that bids from experimental auctions predicted non-hypothetical choices in external environments.

ANALYSIS

To examine the possible impact of information treatments, product trials, and participant characteristics on demand for ST, we estimated random effects regression models with the following equation:

$$BIDit={\alpha }_i+\delta ’{\mathrm{L}}_i+\beta ’X_i+\gamma ’{\mathrm{C}}_i+{∊}_{it}.$$ (1)

In equation 1, BIDit is participant i’s bid for the ST product t (where t=Snus, Ariva, or Nicorette), α_i is a random effects intercept term; L _i is a vector that represents which ST product the participant was bidding upon and δ’ is the associated coefficient vector; X _i is a vector that represents which information or product trial treatment participant i received and β’ is the associated coefficient vector; C_i is a vector that represents the demographic and smoking-related characteristics of participant i and γ’ is the associated coefficient vector; and ε_it is the error term. We ran a multivariate model, containing all variables of interest, as well as a series of ‘unadjusted’ models. One ’unadjusted’ model contains only the experimental treatment dummy variables (shown in Table 2), and the remaining models include these the treatment dummy varaiables along with each individual demographic or smoking related characteristic included in the models separately and one at a time (e.g. age or race).

Table 2.

Mean bids across cities and treatment groups⁺

	Bid_SNUS	BID_ARIVA	BID_NICORETTE	BID_CIGARETTES
Overall (N=571)*	$1.26	$1.58	$2.09	$4.12
Mean Bids across Cities
Buffalo (N=209)**	$1.86	$2.24	$3.06	$5.34
Columbia (N=195)	$0.84	$1.08	$1.51	$2.72
Selinsgrove (N=167)	$1.00	$1.35	$1.55	$4.22
Mean Bids across Treatment Groups
Control Group (N=62)	$0.72	$1.44	$2.04	$4.09
Only Pro-ST info (N=60)	$1.59	$1.69	$1.94	$4.40
Only Anti-ST info (N=67)	$1.30	$1.92	$2.97	$4.77
Both pro- and anti- ST info(N=61)	$1.52	$1.60	$1.76	$3.15
Anti-smoking info (N=60)	$1.85	$2.06	$3.00	$4.17
Anti-smoking and pro- ST info (N=65)	$1.12	$1.35	$1.60	$4.24
Snus trial (N=64)	$0.60	$0.91	$1.54	$3.30
Nicorette trial (N=67)	$1.55	$1.80	$1.90	$4.16
Ariva trial (N=65)	$1.13	$1.48	$2.03	$4.76

^*Differences between bids for snus and avira, snus and Nicorette, ariva and Nicorette, and all ST products vs. cigarettes are statistically significant at the 1% level using a t-test.

^**Bids for all products are higher in Buffalo, NY than the other locations. The differences in bids are statistically significant at the 1% level using a t-test.

⁺Auctions conducted from November 2010-November 2011.

We estimated a Tobit model to examine the demand for cigarettes. We used a Tobit model because it can correctly handle bids that are censored at zero (Greene, 2000). The Tobit model was estimated in the following way:

$$BI{D}_i=\alpha +\beta ’X_i+\gamma ’{\mathrm{C}}_i+{∊}_i$$ (2)

where BIDi is participant i’s bid for the cigarettes, α is an intercept term; X _i is a vector that represents the demographic, background, and smoking-related characteristics of participant i, and β’ is the associated coefficient vector; C_i is a vector that represents the treatments that participant i received and γ’ is the associated coefficient vector; and ε_i is the error term. Given that cigarettes and ST both contain nicotine and could be substitutes for one another, a comparison of the impact of ST information and product trials on demand will help in understanding the relationship between demand for cigarettes and ST.

RESULTS

Table 1 shows some demographic and background information of our participants. Most participants (64%) were white, 27% black, and the remainder indicated another racial group. Thirty-seven percent of our sample was under 30 years of age, with 40% between 30 and 50 years of age and 23% over 50 years old. Over half of our sample (53%) reported a household income below $30,000 per year. Forty percent of our sample is female, while 60% is male. The breakdown by treatment shows that our random allocation of treatment conditions to groups of participants resulted in similar demographic characteristics across all the treatment groups and the control group.

Table 2 shows the mean bids across cities for each of the three ST products and for the cigarettes along with the mean bids for each of the nine treatment groups. The bids for ST products and cigarettes were higher in New York (NY) than in either South Carolina (SC) or Pennsylvania (PA). There are also some other differences in bids across groups. For example, participants in the control group and the group that received the snus trial bid less for snus than in the pro-ST information group. It is worth noting that in the free trial groups, approximately 65% of participants tried the Nicorette Lozenges and the Ariva Dissolvables, while only 44% tried the Camel Snus. The differences between the percentage who tried Nicorette or Ariva relative to Snus was statistically significant at the 1% level using t-tests.

Table 3 presents the results of the random effects regression models for ST products.¹ Bids for all ST products are between $1.03-$1.21 lower in SC and PA relative to NY (p<0.01) depending on the model.² Anti-smoking information had a positive and statistically significant effect (p<0.5 or p<0.01 depending on model) on bids for ST products, indicating that receiving anti-smoking information caused smokers to increase their bid by between $0.90-$0.97. No other information had a statistically significant impact on bids.

Table 3.

Random Effects Regression Model regressing bid for all three smokeless tobacco products. Dependent variable = Bid for smokeless tobacco product. Standard error in parentheses. N=1713 (571*3). ⁺

	Dependent Variable=Bid for ST
	Unadjusted	Multivariate
Intercept	0.02** (0.28)	1.28** (0.48)
Bid was for Ariva	0.32** (0.12)	0.32** (0.12)
Bid was for Nicorette	0.83** (0.12)	0.83** (0.12)
Treatment that received only pro-ST information	0.32 (0.39)	0.28 (0.38)
Treatment that received only anti-ST information	0.66 (0.38)	0.64 (0.37)
Treatment that received only anti- smoking info	0.90* (0.39)	0.97** (0.38)
Treatment that received both pro-ST and anti-ST info	0.23 (0.38)	0.30 (0.38)
Treatment that received both pro-ST and anti-smoking information	−0.04 (0.38)	0.01 (0.37)
In treatment that was offered Snus and tried Snus	−0.18 (0.48)	−0.18 (0.48)
In treatment that was offered Ariva and tried Ariva	0.29 (0.42)	0.20 (0.42)
In treatment that was offered Nicorette and tried Nicorette	0.03 (0.42)	0.10 (0.42)
In treatment that was offered Snus but rejected Snus	−0.54 (0.45)	−0.49 (0.44)
In treatment that was offered Ariva but rejected Ariva	−0.16 (0.54)	0.04 (0.53)
In treatment that was offered Nicorette but rejected Nicorette	0.85 (0.51)	0.82 (0.50)
Age	0.01 (0.01)	0.00 (0.01)
Female	0.17 (0.18)	0.15 (0.20)
Income – Below $30,000	0.20 (0.19)	0.13 (0.19)
Income – Between $30,000-$60,000	−0.46 (0.31)	−0.23 (0.31)
Race_black	0.37* (0.21)	0.18 (0.23)
Race_other	0.19 (0.33)	0.20 (0.33)
Participant is from South Carolina	−1.21** (0.21)	−1.16** (0.22)
Participant is from Pennsylvania	−1.07** (0.21)	−1.03** (0.23)
Participant has used ST at some point in past	−0.08 (0.18)	0.27 (0.20)
Participant is worried quality of life may be lower because of smoking	0.06 (0.18)	0.07 (0.18)

^**p< 0.01

^*p< 0.05

⁺Auctions conducted from November 2010-November 2011. N=571*3

Table 4 examines the bids for cigarettes. In both model specifications, we find that anti-smokeless tobacco information increases bids for cigarettes, from between $0.91-$0.94 in adjusted and unadjusted models, respectively. Those who receive both pro- and anti-ST information bid less for cigarettes, although this information is only statistically significant in the multivariate model. Those who were offered but rejected snus bid less for cigarettes, and we find evidence in the unconditional model that those who were offered but rejected Nicorette bid more for cigarettes. Participants from SC and PA both bid less for cigarettes than participants from NY. Participants who are worried their quality of life may be lower because of smoking bid $0.37 less for cigarettes, but this was statistically significant only in the multivariate model.

Table 4.

TOBIT MODEL: Dep variable = Bid for cigarettes. Std error in parentheses. ⁺

	Unadjusted	Multivariate
Intercept	4.09** (0.29)	5.72** (0.46)
Treatment that received only pro-ST information	0.73 (0.43)	0.44 (0.39)
Treatment that received only anti-ST information	0.94* (0.41)	0.91* (0.36)
Treatment that received only anti-smoking info	0.15 (0.42)	0.08 (0.37)
Treatment that received both pro-ST and anti-ST info	−0.75 (0.42)	−0.84* (0.37)
Treatment that received both pro-ST and anti-smoking information	0.22 (0.41)	−0.03 (0.37)
Tried_Snus	−0.14 (0.52)	−0.44 (0.47)
Tried_Ariva	1.14* (0.45)	0.70 (0.41)
Tried_Nicorette	0.02 (0.47)	0.27 (0.42)
Offered but rejected Snus	−1.10* (0.49)	−0.94* (0.43)
Offered but rejected Ariva	−0.29 (0.58)	−0.11 (0.52)
Offered but rejected Nicorette	1.14* (0.57)	0.85 (0.51)
Age	0.00 (0.01)	−0.01 (0.01)
Female	0.26 (0.20)	0.10 (0.20)
Race_black	0.17 (0.23)	0.25 (0.23)
Race_other	0.38 (0.35)	0.37 (0.32)
Participant is from South Carolina	−2.46** (0.21)	−2.60** (0.22)
Participant is from Pennsylvania	−1.04** (0.22)	−1.09** (0.23)
Participant has used ST at some point in past	−0.01 (0.20)	0.38 (0.20)
Participant is worried quality of life may be lower because of smoking	−0.31 (0.20)	−0.37* (0.18)

^**p< 0.01

^*p< 0.05

⁺Auctions conducted from November 2010-November 2011. N=571

DISCUSSION

Our results provide evidence that to increase demand for ST products, providing smokers with anti-smoking information is more effective than providing information about the benefits of ST. Indeed, neither pro-ST information nor anti-ST information appeared to affect bids for ST products.

These results suggest that anti-smoking information campaigns may be necessary to encourage smokers to switch to less hazardous alternatives. Nevertheless, our study indicated that anti-smoking information did not affect cigarette demand. Indeed, when examining demand for cigarettes, providing anti-ST information alone increased bids for cigarettes. However, providing both anti-ST information and pro-ST information simultaneously decreased bids for cigarettes. Receiving Anti-ST information alone may make using ST as a substitute for smoking relatively less attractive in the absence of reminders about the risks of smoking. This result should be interpreted in light of current warning label policy in the US, where ST products have larger, more prominent warnings than cigarette packs (Waxman, 2009). Tobacco industry litigation to delay or even halt implementation of graphic warning labels on cigarettes in the US likely increases the demand for cigarettes, relative to the already low demand for ST products. Research is sorely needed on messaging that most effectively conveys the relative risk of different tobacco products.

Our results also show that demand (as estimated by bids) for all three ST products was consistently lower among smokers in SC and PA than in NY. Because cigarette prices are higher in NY, standard economic theory predicts that demand for a substitute of cigarettes, like ST, would be greater in NY than in SC and PA, where price is lower (e.g., see Mankiw 2012). Our results provide evidence that smokers in NY may be more likely to use (or perhaps switch to) ST than smokers in SC and PA, and higher cigarette prices achieved through taxes or substantially lowering the price of ST relative to cigarettes may motivate ST demand.

There are several limitations with our study worth noting. First, our experimental auctions were not conducted with a nationally representative sample, so our results may not be generalizable to the public at large. That being said, we ran auctions in three locations from different regions. Compared to many surveys and experiments, our sample was comprised of a larger percentage of smokers from minority and lower-income groups, where smoking is more heavily concentrated. Therefore our results may be more generalizable than studies with samples that underrepresent these groups. Second, participants may have been unfamiliar with the ST products, resulting in generally low bids; however, the auction method nevertheless represents their demand, even if demand may change with greater familiarity. Third, our experiment did not randomize the order of the bids for products. This shortcoming may impact overall bids but since this was applied to all information treatments, any disparities in bids would be differenced away when comparing across information or product trial treatments. Fourth, practical considerations influenced our decision not test all possible treatment combinations (e.g., both anti-smoking and anti-ST information; information plus trial). Research into how consumers behave under these treatments would be useful, however. Fifth, our experimental auctions assessed the impact of information given as print brochures. Participants might behave different when receiving information via other media outlets (TV, radio, etc.) than from print brochures, as TV has been found to be a particularly potent medium for smoking cessation campaigns (National Cancer Institute, 2008). Sixth, we are capturing demand at one moment in time. If information or product trials has an impact that changes over time, we don’t capture this impact. Finally, our sample of 571 is large for experimental auctions, but still may not have been adequately powered to find all statistically significant differences. This could especially be true in the product trial groups, as a large percentage of participants chose not to try the ST products.

Unlike surveys and focus groups, participants in the experimental auction make decisions that have true financial impact (List and Gallet 2001). In other words, auction winners pay for and receive the product, just as they would in the marketplace. Although this method does not assess cognitive impact or provide psychological explanations for differences in demand, it captures a behavioral outcome (i.e., purchasing the product) that may be considered more proximal to desired behavioral impact than self-reported psychosocial indicators, such as intention to purchase or an expression of willingness to pay that does not have financial consequence. This method offers the additional advantage of allowing greater experimental control over transaction conditions than studies of naturally occurring market transactions.

Our nonhypothetical results from experimental auctions to estimate the demand for ST products under alternative information and product trial treatments should be quite useful, especially given the dearth of data on smokers’ perceptions of ST products. We find that anti-smoking information was most effective at increasing demand for ST products and is likely to be a critical component of efforts to increase ST demand among smokers in real life. Regulators and public health professionals who are interested in promoting ST use as a reduced risk strategy for smokers should consider raising cigarette prices and reminding smokers about smoking-related health risks, perhaps through the implementation of graphic warning labels and hard-hitting media campaigns.

• Little is known about consumer perceptions of novel smokeless tobacco (ST).

• We conducted experimental auctions of ST and cigarettes with 571 smokers.

• Anti-smoking information increased demand for ST products, but not cigarettes.

• Anti-ST information increased demand for cigarettes, but not ST products.