Go to the University of Cumberlands online library. Select an article from the Journal of Management Information Systems. Post a summary of the theory, research design, analysis, and conclusions regarding the article you selected. You must pick a different article than others in the c
Digital Nudging: Numeric and Semantic Priming in E-Commerce Alan R. Dennis a, Lingyao (IVY) Yuan b, Xuan Fengc, Eric Webb d, and Christine J. Hsiehe
operations and Decision Technologies Department Kelley School of Business, Indiana University, Bloomington, Indiana, USA; bDepartment of Information Systems and Business Analytics, Debbie & Jerry Ivy College of Business, Iowa State University, Ames, Iowa, USA; cDivision Management Information Systems Price College of Business, University of Oklahoma, Norman, Oklahoma, USA; dDepartment of Operations, Business Analytics, and Information Systems Carl H. Lindner College of Business, University of Cincinnati, Cincinnati, Ohio, USA; eSan Francisco, California, USA
ABSTRACT Most research on e-commerce has focused on deliberate rational cognition, yet research in psychology and marketing suggests that buying decisions may also be influenced by priming (a form of what Information Systems researchers have called digital nudging). We con- ducted seven experiments to investigate the impact of two types of priming (numeric priming and semantic priming) delivered through what appeared to be advertisements on an e-commerce website. We found that numeric priming had a small but significant effect on consumers’ willingness to pay when the value of the product was unclear, but had no effect when products displayed a manufacturer’s suggested retail price (MSRP) or a fixed selling price. Semantic priming had larger effects on willingness to pay and the effects were significant but smaller in the presence of an MSRP. Thus, the combination of numeric and semantic priming has a larger impact on consumers’ willingness to pay. Taken together, these experiments show that some of the research on numeric priming and semantic priming done in offline settings generalizes to e-commerce settings, but there are important boundary conditions to their effects in e-com- merce that have not been noted in offline settings. In online auctions (e.g., eBay), sellers can influence customers to pay more for products whose value is unclear by displaying products with clearly labelled high prices alongside the products the consumer searched for. However, such tactics will have only minimal effects for auctions of products whose price is known (e.g., those with an MSRP) and no effects on products with clearly listed prices (e.g., Amazon).
KEYWORDS Decision making; anchoring and adjustment; priming; dual process cognition; System 1 cognition; System 2 cognition; digital nudge; online auctions; willingness to pay; pricing; price anchors
Introduction
What affects how much a consumer is willing to pay for a product in an e-commerce marketplace? Much prior research has focused on the rational aspect of consumer buying behavior, so past research suggests that willingness to pay is influenced by consumers deliberately considering pricing information, product value, product image, trust in the seller, website design, available information, and so on [6, 16, 34, 35, 41, 48, 57, 70]. Yet, psychology
CONTACT Alan R. Dennis ardennis@indiana.edu Operations and Decision Technologies Department Kelley School of Business, Indiana University, 1275 E 10th Street, Bloomington, IN 47405, USA
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 2020, VOL. 37, NO. 1, 39–65 https://doi.org/10.1080/07421222.2019.1705505
© 2020 Taylor & Francis Group, LLC
research shows that much of human behavior is guided by nonconscious cognition and can be influenced by seemingly irrational elements in the environment [5, 10, 30, 37, 38].
For example, Nunes and Boatwright [47] set up two booths on a west coast beach boardwalk one Saturday. One booth displayed a single product (a plain sweatshirt) whose price was advertised as $10 or $80. The adjacent booth sold a CD in a single bid auction: the consumer was asked to name a price and if the price met the threshold amount, the consumer purchased the CD. When the unrelated sweatshirt in the adjacent booth was priced at $80, consumers bid significantly more for the CD than when the sweatshirt was priced at $10.
In the 15 years since this study, the irrational nature of consumer buying decisions being affected by the price of a nearby unrelated product has been widely used in research, and reported in the popular press [5, 30]. More than a dozen studies have actually looked more deeply into this phenomenon, most in traditional offline environments [e.g., 2, 30, 32, 69]. The general consensus is that this is a special case of Tversky and Kahneman’s [65] anchoring and adjustment bias called numeric priming, where individuals were asked to produce a numeric value as an initial anchor (which can be easily biased by any number in sight) and then insufficiently adjust this anchor up or down to arrive at a final decision [25].
Research has also identified semantic priming, where the price of a related product can influence a consumer’s willingness to pay [2, 44]. For example, presenting an expensive bicycle on the same catalog page as moderately priced bicycles, increased the amount consumers were willing to pay for a moderately priced bicycle [44]. Semantic priming and numeric priming share common roots but operate through different yet complementary psychological processes [2].
Virtually all of the prior research on numeric and semantic priming in consumer buying decisions has focused on traditional offline settings, not e-commerce settings. There are at least three key differences between e-commerce and traditional settings. First, e-commerce retailers can offer a much larger assortment of products because they are not limited by physical size [30]. Second, many products are sold in name-your-price auction sites such as eBay [21, 72]. Finally, and most importantly, e-commerce retailers have much greater control over how a consumer accesses and interacts with the available products and can strategically design their websites to influence consumer behavior through digital nudges [52, 67].
We conducted a series of seven experiments to investigate the extent to which numeric priming and semantic priming can be used to influence consumer’s willingness to pay in e-commerce. We used advertisements displayed during the buying process as digital nudges to implement the priming. Both numeric and semantic priming influenced buying decisions, but we also identified some clear boundary conditions which limit the general- izability of the findings of research in offline settings to e-commerce settings. In short, the effects of numeric priming are small and limited to only a certain group of products in online auctions, while semantic priming has a small to medium effect sizes across a much wider set of conditions. Thus, we conclude that the widely reported irrational effects of numeric priming [5, 30, 47] have very limited application in e-commerce, while semantic priming [2], which has received less attention, has much wider application as a form of digital nudging [52, 67]. We begin with a theoretical background on priming and its
40 DIGITAL NUDGING.
implementation through digital nudging, and then present each of the seven studies in turn. We close with a discussion and the implications for research and practice.
Theoretical Background
E-commerce has changed the face of business, enabling consumers to purchase goods from both businesses and individual sellers [9, 16]. In this environment, understanding consu- mers’ willingness to pay has become a key to business success [4, 8, 9, 33]. Most prior research is based on rational choice theory, which assumes that an individual acts rationally to balance costs against benefits to maximize personal advantage [17, 51, 53]. The amount a consumer is willing to pay is a function of the consumer’s perceived value of the product [50]. Prior research shows that willingness to pay is influenced by many factors, some related to the product, some related to the seller, and some related to the design of the purchasing environment. These factors are then used as consumers evaluate the perceived costs and benefits and rationally choose the most appropriate product [17, 51, 53]. Information about the product and product reviews influence the perceived value of a product [15, 45, 63], as does prior experience with e-commerce [23]. The design of the e-commerce site itself can also influence willingness to pay [28, 35]. The quality of a business’s e-image also affects the prices received at the auction and individuals’ will- ingness to transact business [16, 29]. Implementing information feedback helps build trust, which is critical in e-commerce [3, 4, 9, 16].
Recent research suggests that we need to look beyond theories of fully informed rational choice in e-commerce buying decisions because consumers have cognitive biases that influence their online buying decisions [52, 67]. One interesting approach is digital nudging, the use of interface design and product selection to influence user buying behavior by applying well known biases in human decision making [52].
In this paper, we explore digital nudging that we implement through the advertise- ments displayed on an e-commerce website. We investigate how digital nudging can employ numeric priming and semantic priming which use well-known biases in the anchoring and adjustment decision process [65]. In this section, we first describe the anchoring and adjustment process, then examine how numeric and semantic priming can influence this process, and finally describe how to implement numeric and semantic priming in an e-commerce setting.
Anchoring and Adjustment
Anchoring and adjustment is a decision making process whose effects are robust and widely observed, even in the face of warnings about it [2, 24, 47, 65]. As the name suggests, people using this process begin by making an initial estimate (the anchor) and then adjust this initial estimate to arrive at a final decision [47, 65]. Because people are susceptible to confirmation bias, they tend to focus on information that supports their initial anchor and disregard information that refutes it [47]. This often results in insuffi- cient adjustment [65] so that the final decision remains close to the initial anchor [25]. Thus the choice of this initial anchor often has undue influence over the final decision.
We would hope that rational decision makers would choose this initial anchor with careful thought. However, there is considerable empirical evidence that this is not the case. Much
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 41
research has shown that decision makers are easily swayed by the presentation of a number prior to a decision — even a number they know is randomly generated [6, 13, 55, 65]. This number biases their initial anchor and thus has strong influence on the final decision.
Recent research in psychology provides the theoretical underpinnings of cognition that explains why this occurs. Researchers have long argued that there are two fundamentally different forms of cognition [26]. Kahneman [38] summarized research in this area and uses the terms System 1 (automatic cognition) and System 2 (deliberate cognition). System 2 is what we mean when we think of a decision maker taking time to invest deliberate and conscious thought to make a decision [64]. In contrast, System 1 runs continuously, and delivers conclusions automatically and involuntarily; “it cannot be turned off” [38, p. 25]. It is the impulsive driver of behavior [64], our intuitive thought process [19, 38]. When presented with a stimulus, our System 1 cognition automatically generates a response in less than one second [38], often as quickly as 300 milliseconds [62]. System 1 runs continuously and supplies these assessments to us, even though they are not asked for [26, 38]. System 2 runs much slower and often adopts the conclusions of System 1 without thought [38]. When we use System 2 to override the initial instinctive reaction of System 1, the results of System 1 often strongly influence the System 2 cognition that follows [19, 38].
Influencing Price Anchoring through Priming
The anchoring and adjustment process plays out in buying decisions [1, 5]. When deciding to buy a product, the consumer often sets an initial anchor price and then adjusts this anchor to arrive at the final amount to pay [2]. This initial anchor price is influenced by relevant external information that the consumer encounters (e.g., reference prices such as prices for substitute products that could be bought instead) [2]. There is also evidence that this initial anchor price is influenced by irrelevant external information that is completely unrelated to the decision (e.g., the price of a sweatshirt when buying a CD [47]).
Two different theoretical mechanisms have been proposed to explain these effects: numeric priming and semantic priming [2, 58]. Priming (whether numeric or semantic) is the presenta- tion of a stimulus designed to influence subsequent cognition or behavior [11]. There are several decades of research showing that priming influences cognition [12, 56]. Priming gets its name because the stimulus is presented first (often via a game or separate experimental task), followed by the focal task of interest; the priming stimulus is designed to “prime” one’s cognition to perform in a specific way. The priming stimulus is often presented supraliminally, such that the individual is consciously aware of the stimulus but not its intent [11]. Individuals are unaware of priming; therefore, they will often deny its effects, even in the face of evidence [11]. Even if individuals are aware of the purpose of priming, it still affects their cognition and behavior [11].
Numeric priming is defined as the presentation of a number prior to a task requiring the individual to make a decision involving a number [68]. Buying decisions in online auctions require consumers to name a price that they are willing to pay, so they may be susceptible to numeric priming. The consumer knows he or she must generate a number to bid in the auction. When the consumer looks at a product in an auction, his or her System 1 cognition attempts to instantly answer the question “How much should I bid?” There are no quick answers to this question, because it depends on the features of the product. However, System 1 will produce an answer in less than a second by quickly finding a number (e.g., the
42 DIGITAL NUDGING.
priming number in working memory [68]), and delivers this as its conclusion. Any reason- able number in working memory [2] or in the visual field is considered, even numbers in the name of the product [22]. This System 1 result often becomes the initial price anchor although the consumer can invoke System 2 cognition to override it. System 2 can (and often does) change this initial price anchor through the adjustment process, but this initial anchor biases the final price determined by System 2 [2], because the adjustment process is often insufficient [25].
Semantic priming is defined as the presentation of a high or low quality product prior to a buying task [2, 44, 58]. Semantic priming is similar to numeric priming, but works via a different theoretical mechanism. The intent of semantic priming is for the priming stimulus to trigger the consumer to engage System 2 to think about features of the stimulus product, with its features entering working memory [2] and influencing System 2’s deliberations. Thus priming with a high quality product leads individuals to think about features associated with high quality, while priming with an low quality product leads to thoughts about features associated with a low price [1, 44]. These product features become salient in working memory and exert stronger influence as the consumer uses System 2 to evaluate a product and considers how much to pay [2, 44, 58]. Features associated with more or less expensive products become the basis on which the product is evaluated [58], which influences willingness to pay [2, 44].
The two theoretical mechanisms are not mutually exclusive. System 1 cognition is more susceptible to numeric priming, while System 2 cognition is more susceptible to semantic priming. Research investigating semantic priming has often included numeric priming because one way to indicate an expensive product is with a high price — a number. There are other ways to indicate an expensive product, such as showing a luxury brand name (e.g., BMW). Likewise, past research examining semantic priming has usually used a related product as the stimulus (i.e., in the same product category such as priming with an expensive bicycle when the task is to buy a bicycle), whereas past research on numeric priming has often used unrelated products.
Priming in the Online Context
Priming has received considerable attention in psychology and marketing, but priming as traditionally studied, is impracticable; in the real world, we cannot force consumers to spend five minutes doing a priming task before they buy a product. Instead, we need to deliver the priming stimulus concurrently with the task. Rather than requiring consumers to focus on the priming stimulus prior to the task, the priming is presented on the same screen at the same time as the task itself [54, 71]. The advantage of this approach is that it is readily applied to e-commerce environments. The task is online and as the user performs the focal task on one part of the screen (i.e., shopping for a product), the priming stimulus appears on an adjacent part of the screen.
We call this form of priming concurrent priming. Concurrent priming may be weaker than traditional a priori priming (with the stimulus presented prior to the task) because there is no guarantee that the user will actually see and process the stimulus; the user may simply ignore it as he or she works on the focal task. Yet, research shows that individuals do not need to consciously process priming stimuli for priming to have an effect [14]. There is also some evidence that the effect of priming is stronger when the priming stimuli
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 43
are presented closer in time to a buying decision [44]; therefore, concurrent priming may be stronger than traditional priming.
We chose to use advertisements as the concurrent priming stimuli. Advertisements are frequently found on e-commerce sites [66], so these advertisements should appear natural to users. Prior research also has used advertisements to deliver priming [71]. We consider concurrent priming a form of digital nudging [52, 67].
In this paper, we investigate the extent to which concurrent priming influences a consumer’s willingness to pay. There has been a substantial amount of research on price anchoring [2, 30, 32], but none of this research has examined priming that is practical to deliver in an online setting, nor examined the relative strengths of the two theoretical mechanisms (numeric priming, semantic priming) believed to influence price anchoring.
Therefore, our goal is to understand the effects and limitations of this form of digital nudging and to investigate the two theoretical mechanisms that have been proposed to explain price anchoring (numeric priming and semantic priming). We conducted a series of seven experiments which are described in the following sections. We begin with a series of experiments using a name-your-price online auction (e.g., eBay), because this has context has long been used by researchers to measure willingness to pay. However, as we explain later, this context may intensify the priming effect, so in two experiments we use a fixed-price e-commerce setting (e.g., Amazon).
Study 1: Numeric and Semantic Priming via Online Advertising
In this first experiment, we examine the effects of both numeric and semantic priming delivered through advertisements for products. This provides an initial test of the effec- tiveness of concurrent priming in influencing online buying behavior and also enables us to examine the relative influence of the two theoretical mechanisms.
Numeric Priming
The first theoretical mechanism is numeric priming. This mechanism theorizes that a number displayed as a visual stimulus influences the amount a consumer is willing to pay. In an online auction, the consumer must decide how much to pay. System 1 cognition is always operating so it will suggest an answer to this question in less a second [38], regardless of whether the consumer wants the answer or not [38]. System 1 quickly searches workingmemory and looks at available visual stimuli for a reasonable number to produce [38].
With concurrent priming as envisioned here, an advertisement for a product is displayed on the screen at the same time that System 1 is asked to produce a number. This advertisement contains a number, a price for a product. System 1 quickly latches onto this number and uses it as it produces a number for how much you should pay. This number, despite its rather arbitrary source, is then often used as an initial anchor. The consumer invokes System 2 cognition and adjusts this number to a more reasonable value, but because adjustment is often insufficient [65], this initial System 1 result often biases the final answer produced by System 2 [2].
One may attempt to deliberately avoid the conclusions from System 1 cognition, but this is difficult because we are usually unaware of them [51]. Likewise, one does not need to focus conscious cognition on stimuli in the visual field for them to affect nonconscious cognition;
44 DIGITAL NUDGING.
as long as the stimuli are in the visual field, they will be processed, even though the consumer may not remember seeing them [14]. Therefore, when an advertisement with a reasonable price is shown on the screen, System 1 recognizes the price and proposes this as an answer.
It does not matter if the advertised price is for a relevant or irrelevant product — for example if the price of a printer is displayed while shopping for a camera — System 1 cognition uses it as an answer. The number must be within the realm of possibility for the price of the product under consideration [61, 68], but the source of the number is irrelevant; the priming effect is as strong whether the number comes from a rational source (e.g., a related product) or something unrelated [61] (e.g., a social security number [5] or subliminal priming [1]).
Semantic Priming
The second theoretical mechanism is semantic priming. This mechanism theorizes that a priming product triggers the consumer to think about features of the product and these features enter working memory [2] and become more accessible as the consumer uses System 2 to deliberate about willingness to pay. Thus the features associated with the priming product are more likely to be used in assessing how much to pay for the focal product. This is in sharp contrast to the numeric priming mechanism, where a number — any number — is the driving force.
With semantic priming, the nature of a high or low quality priming product triggers cognitions about product features that would warrant paying a high or low price [2, 58]. Priming with a high quality camera will increase willingness to pay because the priming product triggers thoughts of features of expensive cameras (e.g., high pixels, low weight) and those features become more accessible in working memory and thus more salient to System 2 as it deliberates [2]. In contrast, priming with an unrelated product such as clothes when buying a camera will have weak effects, if any at all [2], because priming with high quality clothes will trigger thoughts of features associated with clothes not cameras.
It is important to note that much prior research on semantic priming has indicated whether the priming product is high or low quality by showing a price. When a semantic priming treatment includes a number, the treatment also triggers numeric priming. Unfortunately, not all authors have recognized this and many have drawn erroneous conclusions that their studies show the effects of semantic priming alone (when in fact they show the combined effects of both semantic and numeric priming). As we previously noted, it is possible to indicate high or low quality priming products without simulta- neously including a number that triggers numeric priming.
Hypotheses
The two theoretical mechanisms are complementary. Numeric priming primarily affects System 1 and is such that a reasonable number from any stimulus source will have a direct effect: higher priming numbers lead to a higher willingness to pay, lower priming numbers lead to lower willingness to pay; this forms our first fundamental hypotheses (H1). Semantic priming primarily affects System 2 and relies on the stimulus to trigger thoughts about features related to the focal product the consumer is buying which will have a direct effect: higher quality priming products lead to a higher willingness to pay, lower quality
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 45
priming products lead to lower willingness to pay; this forms our second fundamental hypotheses (H2). Thus, our two specific hypotheses for the first study are:
Hypothesis 1a: Individuals’ willingness to pay will be higher when exposed to numeric priming with a higher number than a lower number in an online auction.
Hypothesis 2a: Individuals’ willingness to pay will be higher when exposed to semantic priming with a higher quality product than a lower quality product in an online auction.
Method
Participants Seventy-three undergraduate students taking an introductory business course at a large US public university participated in the study. The average age was 19.5 years; 57 percent were male. Participants received extra credit for participating.
Task We used a single bid auction to assess participants’ willingness to pay. The specific task was an online shopping task modified from Yuan and Dennis [72], which asked partici- pants to imagine themselves as a new student in a new Master of Science in Graphic Design program in the business school. We used a repeated measures design in which participants experienced the priming treatment twice; as a result, they performed the same buying task twice (with two different sets of products). The task stated that to take courses in this program, students needed to purchase two products from the auction website (a camera and a laptop). The instructions provided minimum configuration requirements as well as recommended configurations. We used laptops and professional cameras because both product categories have a similar price range and products in each category have a reasonably wide range of prices.
For each product category (cameras and laptops), participants chose from a selection of five products, one of which was a bargain brand. We used a set of products rather than one single product because it is well established that evaluating a set of products follows a different cognitive process than evaluating a single product [36]. Individuals shopping online usually compare multiple products, so for ecological validity it was critical to present multiple products to the participants. All products in the same category had the same color and appearance to reduce the effects of appearance. Participants were pre- sented with product descriptions, brand descriptions, rating reports, and detailed test results. Product descriptions and brand descriptions were adapted from Amazon.com. Rating reports and detailed test results were adapted from Consumer Reports (see Appendix A). Participants bid on only one product from each category.
Treatment We used a repeated measures design with relatedness of advertised product (related vs. unrelated) as a between-subjects factor and the price of the advertised priming product (high vs. low) as a within-subjects factor. Priming with an unrelated product was a test of numeric priming alone because it presents a number (the price of the unrelated product)
46 DIGITAL NUDGING.
to induce priming. Priming with a related product was a test of the combined effects of numeric priming and semantic priming because the price induces numeric priming and the use of a price of a related product to induce perceptions of high and low quality has often been used for semantic priming.
Participants were randomly assigned to one of the two product relatedness treatments and received both priming price treatments (one for the camera and one for the laptop, with product and treatment order randomized). A within-subjects design for the two advertised prices better controls for individual differences. The advertisement was pro- vided on the right side of the screen (see Figure 1). We wanted the advertisement to appear different on each page, so System 1 would perceive it as a new stimulus, but wanted the stimulus to remain consistent as the participant moved from page to page. The words and price remained constant on every page, but the picture changed from page to page.
The relatedness of the advertised product was manipulated through the similarity of the advertised product to the bidding product. A related product was the same product type as the bidding product, but not one of the specific products that the participants could buy. For example, a camera was used when the subject was buying a camera. An unrelated product was a different product type from the bidding product. We chose a bicycle as the unrelated product because a bicycle is unlikely to trigger thoughts about product features related to cameras or laptops.
The price of the priming product in the advertisement was manipulated. The lower priced product was set at $650. The higher priced product was $950. Both prices were selected to be in the range of reasonable prices for cameras, laptops, and bicycles.
Figure 1. Example of experiment screen in study 1.
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 47
Variables and Measurements The dependent variable was the participant’s willingness to pay for the product, measured as the amount of the bid entered by the participant.
We used product knowledge as a control variable as product knowledge can influence how much a consumer is willing to pay [22, 69]. It was measured with four 7-Likert scale items about laptop knowledge (alpha = .83) and four about camera knowledge (alpha = .83). We found no significant differences in laptop knowledge (F(1,71) = 1.25, p = .268) or camera knowledge (F(1,71) = 3.74, p = .057) between participants assigned to the relevant or irrelevant treatments. There were six other control variables: product order, treatment order, product type (camera or laptop), and whether the product selected was the bargain brand (which could cause a lower bid).
Analysis Hierarchical Linear Modeling (HLM) was used for two reasons [49]. First, HLM controls for statistical issues in repeated measurement designs by incorporating subject-specific effects at level 2, with the treatment-level within-subject observations at level 1. Second, our control variable was different between the two between-subjects treatments. HLM makes it simple to use this as controls at level 1, while traditional repeated measures GLM makes it impossible to have a covariate that is different between treatments. We built a standard two level HLM fixed effects model with robust standard errors using the HLM 6 software:
Level 1 (Priming Price Treatment):
WillingnessToPayij ¼ β0i þ β1i PrimingPriceij � � þ β2i Product Typeij
� �
þ β3i Product Knowledgeij � � þ β4i Bargain Brandij
� �
þ ε (1)
Level 2 (Participant):
β0i ¼ γ00 þ γ01 Relatednessið Þ þ γ02 Product Orderið Þ þ γ03 Treatment Orderið Þ þ � (2)
β1i ¼ γ10 þ γ11 Relatednessið Þ (3)
β2i ¼ γ20 (4)
β3i ¼ γ30 (5)
β4i ¼ γ40 (6)
where i = participant and j = task within participant; ε and ξ are error terms. In this model, Level 1 is the within-subjects priming treatment (a 0-1 variable indicat-
ing high (1) or low (0) priming price), while the Level 2 is the specific participant in the study with the between-subject relatedness treatment (a 0-1 variable indicating related (1)
48 DIGITAL NUDGING.
or unrelated (0) product advertisement). The intercept at Level 1 (β0i) is different for each subject to model subject-specific effects. The control variables that contribute to β0i are at both levels because some controls differ by the priming treatment (product type, product knowledge, and the choice of a bargain brand), and others differ by the subject (product order, and treatment order). The effect of the priming price treatment will be different for priming delivered by related and unrelated product advertisements, so we include relat- edness in the model for the slope on the effects of the advertisement price β1i
Procedure After arriving at the laboratory, subjects were given two minutes to read the task. They were then shown a brief demonstration of the website. Participants were asked to shop for the first product, which was either a laptop or camera (randomly selected), and record their bid amount. They then shopped for the second product and recorded their bid amount. The experiment concluded with a short demographic questionnaire, and parti- cipants were dismissed.
Results and Discussion
Descriptive statistics, including mean and standard deviations as well as correlations among variables are shown in Appendix B. The mean willingness to pay for the low and high price related products were 498.38 and 1061.57, and for the unrelated product were 713.22 and 791.36.
Table 1 shows the statistical results for willingness to pay. The results show that the main effect of priming price (high = 1 or low = 0) on Willingness To Pay is significant (p < 0.001). The interaction between priming price and relatedness of priming product was also significant (p < 0.001). The effect size for the main effect was Cohen’s d = .45 and the interaction was d = .35 (calculated using the HLM coefficients). The effect size for numeric priming (.45) is medium. The effect size for the interaction indicates the incremental effect of semantic priming on top of numeric priming which has a small to medium effect size. The combined effect of both numeric and semantic priming is the combination of the main effect plus the interaction effect which is large (d = .75). We conclude that H1a and H2a were supported.
Table 1. Study 1 Willingness to Pay Results Level 1 variables Level 2 variables Beta p-value
Intercept (β0) Intercept (γ00) 766.32 <0.001 Relevance (γ01) 52.92 0.486 Product Order (γ02) -155.04 0.074 Treatment Order (γ03) -74.49 0.412
Priming Price (β1) Intercept (γ10) 319.87 <0.001 Relatedness (γ11) 474.41 <0.001
Product Type (β2) -107.48 0.125 Product Knowledge (β3) -46.01 0.285 Bargain Brand (β4) -184.78 0.067
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 49
Study 2: Numeric Priming Replication
Study 1 shows that both numeric and semantic priming can influence consumers’ beha- vior, with both producing small to medium effects. Concerns have be raised regarding priming, including multiple failures to replicate well-known results and even research fraud [39]. One simple solution is to replicate priming studies to ensure that that initial results are replicable [39]. We chose to do a theoretical replication, which is more powerful than a literal or exact replication that simply repeats the same study with the same methods [20]. We repeated Study 1, focusing only on numeric priming to ensure the priming effects found in Study 1 are robust. Study 2 is a replication of Study 1, examining only numeric priming (i.e., Hypothesis 1a).
Method
This experiment used the same repeated measures design as Study 1, including the same task, measures, and procedure. The treatments were similar to those in Study 1, but differed in three ways. First, we included only the unrelated product advertisement. Second, the size of the advertisement was reduced so that it occupied about 15 percent of the screen, which is typical of the size of embedded online advertisements on theWeb (see Figure 2). Third, we wanted to see if the effects in Study 1 depended on the specific values of the high and low advertised prices we had chosen; therefore, we used different advertised prices: $600 and $1200.
We added two additional questions to the post-task questionnaire. We asked partici- pants if they had noticed the advertisements, and, if so, to report the prices in the advertisements. A total of 92 percent reported they had seen the advertisements, and of those who did, 82 percent got one price correct and 54 percent got both prices correct. There were no significant differences in the amount bid between those that reported
Figure 2. Example of experiment screen in study 2.
50 DIGITAL NUDGING.
seeing the advertisement or not, nor between those who got the advertisement prices correct and those who did not, which is not surprising [14].
Two hundred and thirty-six business students from the same university participated in the experiment (no student participated in more than one study). We used a larger sample because a power analysis with G*Power [27] indicted that a larger sample was needed to provide power to find a small effect. About 47 percent were male, and ages ranged from 18-26 with an average of 19.5.
Results and Discussion
The descriptive statistics and correlations are in Appendix B. The willingness to pay means for the low and high unrelated products were 734.52 and 799.36. Table 2 shows the statistical results. The results show that the priming price (high = 1 or low = 0) significantly affected Willingness to Pay (p = 0.011). Hypothesis 1a was supported. The effect size was small (d = .12).
The results of this replication match those of the original study. We again conclude that Hypothesis 1a is supported: numeric priming influences the prices bid in online auction sites.
Study 3: Numeric Priming for Non-Auction Products
Studies 1 and 2 show that numeric priming with an advertisement has a small to medium effect on willingness to pay in online auctions. Is this effect generalizable to other forms of e-commerce where the price of the product is set by the seller? That is, our results show that numeric priming affects bidding in online auctions like eBay (where the consumer names the price), but does it affect buying on sites like Amazon (where the seller sets the price)? Will numeric priming induce the consumer to choose a more or less expensive product?
The question here is whether numeric priming will have the same effects when a consumer is asked to select products with specific prices. The shopping task here is very similar to the prior tasks (buy a product); as a result, System 2 cognition should function similarly. However, the question posed to System 1 cognition is very different in this task than the auction tasks in Studies 1 and 2. In auction tasks, once the consumer identifies a product, he or she has to set a price; System 1 must produce a number in answer to the question of how much to pay. In contrast, when buying products with set prices, System 1 is not asked to produce a number; instead is it asked whether or not to buy. Because this question does not require a number as an answer, we theorize that numeric priming will have a minimal effect. Thus, our specific hypothesis is:
Table 2. Study 2 Willingness to Pay Results Level 1 variables Level 2 variables Beta p-value
Intercept (β0) Intercept (γ00) 765.01 <0.001 Product Order (γ01) -7.42 0.867 Treatment Order (γ02) -5.54 0.900
Priming Price (β1) 69.73 0.011 Product Type (β2) -143.85 <0.001 Product Knowledge (β3) -4.96 0.778 Bargain Brand (β4) -100.87 0.104
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 51
Hypothesis 1b: Individuals’ willingness to pay will not be higher when exposed to numeric priming with a higher number than a lower number in set-price e-commerce settings.
Method
This experiment used the same repeated measures design as Study 1’s unrelated advertising treatment, but instead presented prices for the five products the subject could buy. The dependent variable was the price of the product selected. The prices were set to $700, $800, $900, $1000, and $1100, in order to present a range of prices that were reasonable for these products and fell within the general range of willingness to pay of the participants in our prior studies. Analyses of the data from Studies 1 and 2 found that the product descriptions did not affect the amount bid for cameras in Study 1 (F(4,67) = 1.40, p = .245) or Study 2 (F(4,232) = 0.83, p = .505) nor for laptops in Study 1 (F(4,67) = 1.24, p = .303) or Study 2 (F(4,232) = 1.22, p = .346). Thus the product descriptions had no significant effects on the amount bid by our subjects. The same products were used in both the high priced and low priced priming treatments, so assigning prices to products does not influence the outcomes.We assigned the three lowest prices to the three least popular products in the prior studies and the two highest prices to two most popular. We did not include the control for selection of the bargain brand because our outcome variable is product selected and its price, rather than bid amount. One hundred and sixty three undergraduate students from the same participant pool participated in this experiment; no student participated in more than one study. Age ranged from 18 to 31 with a mean of 19.6 years; 60 percent were male.
Results and Discussion
Descriptive statistics and correlations are in Appendix B. The dependent variable was the price of the product selected, as an indicator or willingness to pay. illingness to pay means for the low and high treatments were 966.87 and 968.71. The results (Table 3) show that numeric priming (high = 1 or low = 0) had no significant effect on the price of the product selected (p = 0.854). The effect size is so small as to be meaningless (d = .01). A power analysis showed a power of .80 to detect within-subject differences with a small effect size. We cannot reject Hypothesis 1b: numeric priming does not affect the willingness to pay in a set-price e-commerce setting.
Therefore, the results of this study show an important boundary condition for numeric priming in e-commerce: numeric priming has small but significant effects in online auctions, but no significant effects when the seller sets the price. Numeric priming works in online auctions because the consumer is asked to produce a price and numeric
Table 3. Study 3 willingness to pay results. Level 1 variables Level 2 variables Beta p-value
Intercept (β0) Intercept (γ00) 967.79 <0.001 Product Order (γ01) -11.47 0.285 Treatment Order (γ02) -9.79 0.360
Priming Price (β1) 2.05 0.854 Product Type (β2) 64.79 <0.001 Product Knowledge (β3) 10.83 0.017
52 DIGITAL NUDGING.
priming biases this by suggesting an unrelated number. Numeric priming fails to work in traditional e-commerce sites because the consumer is not asked to produce a number and thus numeric priming has no effect.
Study 4: Numeric Priming with Two Potential Anchors
Study 3 presents an important boundary condition to numeric priming; the presence of a set selling price (e.g., as on Amazon) eliminates the effects of numeric priming. Some products sold in online auctions have a value suggested by the marketplace, often in the form of a Manufacturer’s Suggested Retail Price (MSRP). The MSRP is an external reference price [21, 46] and serves as a critical information point and is often incorporated in the consumers’ decision using deliberate System 2 cognition. Many consumers view the MSRP as the ceiling price for a product, believing that they should not pay more than this price. Thus, the MSRP serves as another anchor for valuation and willingness to pay.
The effects of numeric priming should be weaker when an MSRP is present, because nonconscious System 1 cognition acts upon the MSRP, as well as on the price of the advertised product. System 1 now has two numbers vying consideration (MSRP and the price in the advertisement), so the impact of the advertised product’s price should be less. In this situation, numeric priming should still be active, but its effects should be substan- tially weaker than when only the priming number is provided. Thus:
Hypothesis 1c: Individuals’ willingness to pay will be higher when exposed to numeric priming with a higher number than a lower number in an online auction for products with an MSRP.
Method
This experiment used the same repeated measures design as Study 1’s unrelated product advertising, with two differences. First, each product displayed an MSRP, which ranged from $964 to $999. We used slightly different MSRPs to make the study appear realistic (if all products had the same MSRP it would have appeared suspicious), but we did not include MSRP as a deliberately manipulated variable of interest in the study. Third, the selected product’s MSRP was used as a control variable in the analysis.
Two hundred and thirty two undergraduate students from the same participant pool as Study 3 participated in this experiment; no student participated in more than one study. Age ranged from 18 to 24 with a mean of 19.5 years; 63 percent were male.
Results and Discussion
The descriptive statistics and correlations are in Appendix B. The willingness to pay means for the low and high treatments were 911.44 and 926.11. The results in Table 4 show that priming price (high = 1 or low = 0) had no significant effect on Willingness to Pay (p = 0.955). A power analysis showed a power of .92 to detect within-subject differences with a small effect size. The effect size is meaningless (d = .003). Hypothesis 1c was not supported.
The results of this study show no effect for numeric priming. When two numbers are presented, one relevant to the decision (the MSRP) and one irrelevant (the price of an
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 53
unrelated product), the relevant one dominates the irrelevant one. Thus a second important boundary condition for numeric priming in e-commerce is that it is only effective when no other relevant numbers are available. In other words, if a product has a stated MSRP (or a reserve price), numeric priming has no significant effect.
Study 5: Semantic Priming with Two Potential Anchors
Study 4 shows an important boundary condition to numeric priming: numeric priming has no effect in the presence of a second, more relevant number (e.g., an MSRP). Does the same boundary condition apply to semantic priming or does semantic priming still influence consumer buying in the face of an MRSP? Providing a clear cue to a product’s value (e.g., an MRSP) will likely weaken the influence of semantic priming [40].
System 2 cognition is likely influenced by external reference points for the value of a product [21, 46]. Past research has examined a number of different sources of reference points, such as list prices of related products, prices paid in auctions, and price recommendations such asMSRP [21, 31, 46]. There is also some evidence that presenting high-priced products in a set of moderately priced products increases consumers’ willingness to pay for the moderately priced products [44]. The high-priced products activate thoughts about product features that could warrant higher prices, thus making those product features more accessible in working memory and thus those features become more salient as consumers assess their willingness to pay [44]. The study in [44] used a print catalog and found that products presented on the same catalog page created stronger priming effects than those presented before the catalog was opened.
This study examines semantic priming when an MSRP is provided as an explicit cue to the product’s value. Thus:
Hypothesis 2b: Individuals’ willingness to pay will be higher when exposed to semantic priming with a higher quality product than a lower quality product in an online auction for products with an MSRP.
Method
This experiment used the same repeated measures design as Study 4, but instead advertised only a related product (thus testing the effects of semantic priming combined with numeric priming; Study 4 showed numeric priming had no effect in this situation, and, therefore, this is a test of only semantic priming). As in Study 1, we used a camera when the participants were bidding on cameras and a laptop when the participants were bidding on laptops.
Table 4. Study 4 willingness to pay results. Level 1 variables Level 2 variables Beta p-value
Intercept (β0) Intercept (γ00) 918.78 <0.001 Product Order (γ01) 14.01 0.588 Treatment Order (γ02) 10.86 0.694
Priming Price (β1) 0.72 0.955 Product Type (β2) 13.59 0.507 Product Knowledge (β3) -12.46 0.250 Bargain Brand (β4) 28.93 0.249 MSRP (β5) 0.11 0.895
54 DIGITAL NUDGING.
Thirty-five undergraduate students from the same general pool participated, but none participated in more than one study. A smaller size sample was used because a power analysis indicted that a sample of 35 was enough to provide a power of .80 to detect a small to medium effect size (similar to that in Study 1). Age ranged from 18 to 22 with a mean of 19.5 years; 66 percent were male.
Results and Discussion
The descriptive statistics and correlations are in Appendix B. The willingness to pay means for the low and high treatments were 852.34 and 905.03. Table 5 shows the statistical results. The results show that priming price (high = 1 or low = 0) significantly affected Willingness to Pay (p = 0.019). The effect size was small to medium (d = .36). Hypothesis 2b was supported.
The results show that semantic priming has an effect even in the presence of an MRSP. Priming with products in the same product category as the consumer is buying influences the consumers’ willingness to pay. Even when a product’s MSRP is presented, consumers still bid more when primed with higher priced products. The effects of semantic priming were smaller in the presence of an MSRP (a small to medium effect of d = .36). Study 4 shows that numeric priming has no significant effect in this situation; therefore, we infer that the effect of priming here is solely the effects of semantic priming on System 2 cognition. We note that the interaction effect from Study 1 (which is a similar test of semantic priming) had a similar effect size of d = .35.
Study 6: Semantic Priming for Non-Auction Products
Study 3 showed that numeric priming had no impact on the product selected, identifying an important boundary condition for numeric priming in e-commerce. Numeric priming has no significant effects when the seller sets the price. Does this boundary condition also apply to semantic priming?Will semantic priming influence consumers’ choice of product?We theorize that the same effects found for numeric priming will also apply to semantic priming. Thus:
Hypothesis 2c: Individuals’ willingness to pay will not be higher when exposed to semantic priming with a higher quality product than a lower quality product in set-price e-commerce settings.
Table 5. Study 5 willingness to pay results. Level 1 variables Level 2 variables Beta p-value
Intercept (β0) Intercept (γ00) 878.69 <0.001 Product Order (γ01) 28.79 0.516 Treatment Order (γ02) -0.43 0.992
Priming Price (β1) 51.91 0.019 Product Type (β2) -34.06 0.393 Product Knowledge (β3) -19.55 0.296 Bargain Brand (β4) 48.14 0.171 MSRP (β5) -0.75 0.620
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 55
Method
This experiment follows a similar design as Experiment 3, with three differences. First, a related product was shown in the advertisement. A camera was advertised when participants were asked to select a fixed priced camera while a tablet was shown while choosing a tablet product. Second, the prices of $700, $800, $900, $1000, and $1100 were randomly assigned to the products. This randomization controlled for the potential confound of product descriptions. Third, the display order of the products was rando- mized between subjects.
One hundred and seventy eight undergraduate students from the same participant pool participated in this experiment; no student participated in more than one study. Age ranged from 18 to 31 with a mean of 19.4 years; 58 percent were male.
Results and Discussion
Descriptive statistics and correlations are in Appendix B. The dependent variable was the price of the product selected, as an indicator of willingness to pay. The willingness to pay means for the low and high treatments were 863.48 and 842.90. The results (Table 6) show that the priming price (high = 1 or low = 0) had no significant effect on the price of the product selected (p = 0.116). The effect size is so small as to be meaningless (d = .01). A power analysis showed a power of .93 to detect within-subject differences with a small effect size. We cannot reject Hypothesis 2c: semantic priming does not affect the will- ingness to pay in a set-price e-commerce setting.
Therefore, combining the results of this study and Study 3 shows that when the seller sets the price, both semantics and numeric priming have no significant or meaningful effects on consumers’ choice of product.
Study 7: Semantic Priming across Product Types
Semantic priming is the strongest with the same type of products, but it may spread across product types when with the focal product being purchased shares some general features with the priming product that is a different type of product [2]. Prestige and quality are general features shared by many products that may transfer across product types [2]. Therefore, priming with a product that is known for being high prestige and high quality will induce a consumer to think about prestige and quality across different product types, thus making those features more salient and leading consumers to pay more across a variety of product types.
Table 6. Study 6 willingness to pay results. Level 1 variables Level 2 variables Beta p-value
Intercept (β0) Intercept (γ00) 852.755 <0.001 Product Order (γ01) -9.889 0.532 Treatment Order (γ02) 4.198 0.787
Priming Price (β1) -21.956 0.116 Product Type (β2) -31.107 0.044 Product Knowledge (β3) -6.291 0.392
56 DIGITAL NUDGING.
Past research has tested this theoretical argument by using the same treatments as has been used for numeric priming: priming using a unrelated product and a price (e.g., see Experiment 2 in [2]). Unfortunately, this does not enable us to draw conclusions about the effects of semantic priming separate from the effects of numeric priming. To test this argument, we need to separate semantic priming’s effects from the simultaneous effects of numeric priming; that is we need to remove numeric priming (i.e., remove the priming product’s price). Thus:
Hypothesis 2d: Individuals’ willingness to pay will be higher when exposed to semantic priming with a prestige product than a bargain product (lacking prices) in an online auction.
Method
This experiment used the same design as Study 1. The form of product advertising was changed so that it no longer provided a number (i.e., there was no price for the advertised product). Instead, we chose products that are commonly advertised without prices and would easily be identified by our subjects as a high quality prestige product versus a lower quality bargain product. We selected hotels as the unrelated product and choose two brands familiar to our subject pool: the Ritz Carleton chain and the Motel 6 chain. We conducted a small pilot study with 35 participants and found that they perceived the Ritz Carleton to be more prestigious and higher quality than Motel 6. The words in the advertisement remained constant on every page, but the picture in the middle of advertisement changed from page to page (pictures included rooms, lobby, and exterior of typical hotels in the two chains). The words in the advertisement for the Ritz Carleton promoted hotels that were luxurious and high quality with spacious and lavish rooms, while the words for Motel 6 promoted motels that were simple, affordable, and a great value.
Two hundred and one undergraduate students from the same participant pool parti- cipated in this experiment; no student participated in more than one study. Age ranged from 18 to 28 with a mean of 19.5 years; 51 percent were male.
Results and Discussion
The descriptive statistics and correlations are in Appendix B. The willingness to pay means for the low and high treatments were 745.36 and 718.62. The results in Table 7 show that priming (prestigious = 1) had no significant effect on the price of the product selected (p = 0.809). A power analysis showed a power of .87 to detect within-subject differences with a small effect size. The effect size is so small as to be meaningless (d = .01). Hypothesis 2d is not supported.
Therefore, this study provides no evidence that semantic priming has a significant or mean- ingful effect for unrelated products. Thus, one important boundary condition for semantic priming is that the priming product needs to be related to the product being purchased.
Discussion
Taken together, these seven studies show that consumers’ online buying behaviors can be influenced by numeric priming and semantic priming delivered via advertisements presented during the shopping process. Thus some of the research on the effects of numeric priming and
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 57
semantic priming found in offline environments can be applied to e-commerce environments [30]. However, our research shows important boundary conditions in e-commerce settings. Table 8 provides a summary of the studies and their results.
We investigated the effects of numeric priming delivered by an advertisement contain- ing a price for a product unrelated to the product of interest that was presented beside the products available for purchase. Advertisements containing a high price induced consu- mers to bid more while advertisements containing a lower price led them to bid less. System 1 cognition is highly susceptible to numeric priming [2] because it is designed to instantly suggest an answer to a question. When a consumer needs to determine an initial anchor price, System 1 cognition is easily biased by any number, even one completely unrelated to the product of interest.
Study 1 shows that numeric priming has a small but significant effect in online auctions. Study 2, a replication which used a slightly different experimental design, again shows a significant but small effect for numeric priming. Studies 3 and 4 show important boundary conditions to these effects. Study 3 shows that numeric priming has no significant effects when the seller sets the price (e.g., Amazon). Study 4 shows that numeric priming has no significant effects for an online auction when an MSRP is presented.
We investigated the effects of semantic priming delivered by an advertisement contain- ing a price for a product related to the product of interest that was presented beside the products. Advertisements containing a high price induced consumers to bid more while advertisements containing a lower price led them to bid less. Semantic priming primarily works by triggering System 2 cognition to think about features associated with high or low quality products and these features become more accessible and thus more salient in deciding how much to bid [2].
Study 1 shows that semantic priming has a significant and small to medium effect in online auctions over and above any effects due to numeric priming; the combination of numeric and semantic priming has a large effect size. Study 5 shows that semantic priming has significant effects, albeit a small effect, when an MSRP is presented for the product. Study 6 shows that
Table 7. Study 7 willingness to pay results. Level 1 variables Level 2 variables Beta p-value
Intercept (β0) Intercept (γ00) 731.99 <0.001 Product Order (γ01) -4.28 0.924 Treatment Order (γ02) -25.47 0.542
Priming Treatment (β1) 11.12 0.809 Product Type (β2) -109.05 0.003 Product Knowledge (β3) -14.10 0.469 Bargain Brand (β4) -72.63 0.099
Table 8. Summary of studies and results. Study Type of Priming Context Result
1 Numeric Auction, no MSRP Small effect 2 Numeric Auction, no MSRP Small effect 3 Numeric Set Price No effect 4 Numeric Auction with MSRP No effect 1 Semantic Auction, no MSRP Small to medium effect 5 Semantic Auction with MSRP Small to medium effect 6 Semantic Set Price No effect 7 Semantic, using unrelated Prime Auction No effect
58 DIGITAL NUDGING.
semantic priming has no significant effects when the price is set by the seller (e.g., Amazon). Study 7 shows that semantic priming has no significant effects when the priming is delivered using a product unrelated to the product of interest; its features are too distant.
Thus, taken together, our studies provide evidence to conclude that the findings from offline research [30] can be generalized to online auctions for productswithout a fixed price andwithout anMSRP. Sellers can influence the amount a consumer is willing to pay for products whose value is unclear by subtly changing the design of a website to include price information for other products – even those unrelated to the product of interest. There are many ways to include such price information, and real— or fake— advertising is a simple option.
However, when products have a clear value, indicated by either anMSRP or a fixed price, the findings from offline research cannot be generalized to e-commerce. Sellers are less likely to influence the amount a consumer is willing to pay for products whose value is signaled by a price or MSRP by including price information for unrelated products.
We investigated two forms of priming (numeric and semantic) and used a form of priming in which the priming stimuli were delivered alongside the focal task, called concurrent priming (sometimes called contextual priming in marketing research [54]). Our results show that technology can be intentionally designed to increase the amount individuals bid in online auctions. The effect sizes range from small to medium. The size of the market (more than $75 billion annually through online auctions) is such that even small changes in revenue can have immense impacts on firm revenues. Thus, there may be both positive and negative implications to buyers, sellers, and society as a whole of using online advertising to deliver numeric and semantic priming in online auctions.
These results may not be “good” from an individual’s perspective because they show that for products lacking a fixed price, individual behavior can be influenced by sellers who deliberately design sites to deliver priming. Technology often is designed to help individuals, but technology itself is neutral [42]; it can be designed to help or hurt whomever those with power choose [42]. Many scholars have discussed the myriad of unintended negative consequences of technology such as deterioration of social bonds, technology addiction, increased stress, and privacy risks [7, 43, 59, 60]. Many products are sold without fixed price (e.g., eBay), and if vendors choose use priming to help sell these products, consumers might end up paying more.
These studies suffer from the usual limitations of lab experiments with undergraduate students working on artificial tasks. Student samples are considered to be an appropriate for testing theories about phenomena that expected to hold true across the general population [18], and there is evidence that students buy products online. The task was artificial in that participants were not given real money to actually purchase the product; there were no consequences to making a bid. The alternative would have been to give participants a small amount (e.g., $10) and have them bid on products of low value, which simply substitutes one artificial task for another – people don’t usually buy low value products at auction. Through this study, we want to start the conversation on the priming effect of advertisement on individuals’ willingness to pay. Further research should test our findings in real settings, such as eBay.
Implications for Research
Notwithstanding these limitations, we believe there are three implications for future research. First, our research shows that the conclusions of Nunes and Boatwright [47] on the beach boardwalk two decades ago that have been widely studied in offline commerce [2, 30]
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 59
generalize to e-commerce auctions for products whose value is unclear (e.g., eBay), but do not generalize to e-commerce auctions that provide an MSRP or e-commerce sites with fixed prices (e.g., Amazon). The presence of a price presents a firm boundary condition to numeric priming. More research is needed to examine the effects of numeric priming via advertising. For example, we used prices in advertising as the priming stimulus; would the effects be the same if we presented numbers in other forms, such as product names [22]?
Second, our research shows that semantic priming (advertising with a related product) can influence consumer behavior in online settings, even in the presence of MSRPs. This study also shows the relative strength of the two types of priming. Semantic priming works primarily through System 2 cognition and had small to medium effects, while numeric priming, which works primarily through System 1 cognition had small effects.We conclude that we need to look beyond the straightjacket of rational choice to examine the non-rational nature of e-commerce. We believe this suggests the need for more research on the effects of semantic priming in online auctions and e-commerce in general.
Third, we included no measurement of direct attention to the advertisement (e.g., eye tracking) for both theoretical and practical reasons; priming affects behavior whether it is consciously seen or not (mere presence in the visual field is sufficient) [14] and software designers cannot control where a customer looks (they can only control what is displayed on the screen). Our post-hoc tests in Study 2 showed that most participants noticed the advertisements and about half could correctly remember both advertised prices, although attention and recall did not affect the amount bid. Future research can investigate whether consumer attention influences the effectiveness of either numeric or semantic priming.
Implications for Practice
The results of this study provide insight into consumers’ willingness to pay when they shop online and thus have practical implications as well. First and foremost, the price of a product in an advertisement displayedwhile a consumer is using an online auction to bid on a product with an unclear value affects howmuch they bid, although advertisements for unrelated products have only a small effect. Marketers at online auction sites can use these results to make subtle design changes to influence the bidding decisions of consumers. By advertising higher priced products, especially those in the same category as the product of interest, e-commerce sites can increase the amount that consumers are willing to pay.
Second, this research also has practical implications for consumers. It can be used to alert consumers to the possible effects of priming by advertisements that could induce them to pay more for products. Perhaps in due time, consumers will train themselves to deliberately discount the effects of such advertising, but this is likely to prove difficult.
Conclusion
Most of the past research examining e-commerce has focused on rational decision making, but there have been recent calls to look beyond the rational, for example, digital nudges [52]. Our research shows that digital nudges using numeric priming and semantic priming can influence how much a consumer is willing to pay. Numeric priming delivered through what appear to be advertisements can influence consumers to bid more in online auctions (e.g., eBay) when the underlying value of a product is unclear, but has no effect when an MSRP is present or the
60 DIGITAL NUDGING.
product is offered at a fixed price (e.g., Amazon); thus that the widely reported irrational effects of numeric priming in traditional settings [5, 30, 47] have very limited application in e-com- merce. We found that semantic priming (also delivered through apparent advertisements) had a stronger and more pervasive effect in online auctions (even in the presence of an MRSP), but likewise had no effects for fixed price sales. We believe much more research is needed to better understand how priming and other digital nudges influence e-commerce.
ORCID
Alan R. Dennis http://orcid.org/0000-0002-6439-6134 Lingyao (IVY) Yuan http://orcid.org/0000-0001-9863-3137 Eric Webb http://orcid.org/0000-0003-3535-4936
References
1. Adaval, R.; and Monroe, K.B. Automatic Construction and use of contextual information for product and price evaluations. Journal of Consumer Research, 28(2002), 572–588.
2. Adaval, R.; and Wyer, R.S. Conscious and nonconscious comparisons with price anchors: Effects on willingness to pay for related and unrelated products. Journal of Marketing Research, XLVIII (2011), 355-365.
3. Adomavicius, G.; Curley, S.P.; Gupta, A.; and Sanyal, P. Impact of information feedback in continuous combinatorial auctions: An experimental study of economic performance. MIS Quarterly, 37, 1 (2013), 55–76.
4. Adomavicius, G.; and Gupta, A. Toward comprehensive real-time bidder support in iterative combinatorial auctions. Information Systems Research, 16, 2 (2005), 169–185.
5. Ariely, D.; and Jones, S. Predictably Irrational. New York: HarperCollins, 2008. 6. Ariely, D.; and Simonson, I. Buying, bidding, playing, or competing? Value assessment and decision
dynamics in online auctions. Journal of Consumer Psychology, 13, 1 (2003), 113–123. 7. Ash, J.S.; Berg, M.; and Coiera, E. Some unintended consequences of information technology
in health care: The nature of patient care information system-related errors. Journal of the American Medical Informatics Association, 11, 2 (2004), 104–112.
8. Baker, W.L.; Lin, E.; Marn, M.V.; and Zawada, C.C. Getting prices right on the web. The McKinsey Quarterly, 2(2001), 54–63.
9. Bapna, R.; Goes, P.; and Gupta, A. Analysis and design of business-to-consumer online auctions. Management Science, 49, 1 (2003), 85–101.
10. Bargh, J.A.; and Chartrand, T.L. The unbearable automaticity of being. American Psychologist, 54, 7 (1999), 462.
11. Bargh, J.A.; and Chartrand, T.L. Studying the mind in the middle: A practical guide to priming and automaticity research. In H. Reis and C. Judd (eds.), Handbook of Research Methods in Social Psychology. New York: Cambridge University, 2000.
12. Bargh, J.A.; and Ferguson, M.J. Beyond behaviorism: On the automaticity of higher mental processes. Psychological Bulletin, 126, 6 (2000), 925.
13. Chapman, G.B.; and Johnson, E.J. Incorporating the irrelevant: Anchors in judgments of belief and value. In T. Gilovich, D. Griffin, and D. Kahneman (eds.), Heuristics and Biases: The Psychology of Intuitive Judgment. Cambridge: Cambridge University Press, 2002, pp. 120–138.
14. Chartrand, T.L.; van Baaren, R.B.; and Bargh, J.A. Linking automatic evaluation to mood and information processing style: Consequences for experienced affect, impression formation, and stereotyping. Journal of Experimental Psychology, 135, 1 (2006), 70–77.
15. Chen, X.; Ghate, A.; and Tripathi, A. Dynamic lot-sizing in sequential online retail auctions. European Journal of Operational Research, 215, 1 (2011), 257–267.
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 61
16. Clemons, E.K.; Wilson, J.; Matt, C.; Hess, T.; Ren, F.; Jin, F.; and Koh, N.S. Global differences in online shopping behavior: Understanding factors leading to trust. Journal of Management Information Systems, 33, 4 (2016), 1117–1148.
17. Coleman, J.S.; and Fararo, T.J. Rational Choice Theory. Newbury Park: SAGE Publications, 1992. 18. Compeau, D.; Marcolin, B.; Kelley, H.; and Higgins, C. Generalizability of information
systems research using student subjects — A reflection on our practices and recommenda- tions for future research. Information Systems Research, 23, 4 (2012), 1093–1109.
19. Dennis, A.R.; and Minas, R.K. Security on autopilot: Why current security theories hijack our thinking and lead us astray. DATABASE for Advances in Information Systems, 49, SI (2018), 15–38.
20. Dennis, A.R.; and Valacich, J.S. A replication manifesto. AIS Transactions on Replication Research, 1, 1 (2014), 1–5.
21. Dholakia, U.M.; and Simonson, I. The effect of explicit reference points on consumer choice and online bidding behavior. Marketing Science, 24, 2 (2005), 206–217.
22. Dogerlioglu-Demir, K.; and Kocas, C. Seemingly incidental anchoring: The effect of incidental environmental anchors on consumers’ willingness to pay.Marketing Letters, 26(2015), 607–618.
23. Easley, R.F.; Wood, C.A.; and Barkataki, S. Bidding Patterns, experience, and avoiding the winner’s curse in online auctions. Journal of Management Information Systems, 27, 3 (2010), 241–268.
24. Epley, N.; Akalis, S.; Waytz, A.; and Cacioppo, J.T. Creating social connection through inferential reproduction loneliness and perceived agency in gadgets, gods, and greyhounds. Psychological Science, 19, 2 (2008), 114–120.
25. Epley, N.; and Gilovich, T. The anchoring-and-adjustment heuristic: Why the adjustments are insufficient. Psychological Science, 17, 4 (2006), 311–318.
26. Evans, J.S.B.T. Two minds rationality. Thinking & Reasoning, 20, 2 (2014), 129–146. 27. Faul, F.; Erdfelder, E.; Lang, A.G.; Buchner, A.; and (2007). G*Power 3: A flexible statistical
power analysis program for the social, b., and bio. G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39(2007), 175–191.
28. Goes, P.B.; Karuga, G.G.; and Tripathi, A.K. Understanding willingness-to-pay formation of repeat bidders in sequential online auctions. Information Systems Research, 21, 4 (2010), 907–924.
29. Gregg, D.G.; and Walczak, S. Dressing your online auction business for success: An experi- ment comparing two eBay businesses. MIS Quarterly, 32, 3 (2008), 653–670.
30. Grewal, D.; Janakiraman, R.; Kalyanam, K.; Kannan, P.K.; Ratchford, B.; Song, R.; and Tolerico, S. Strategic online and offline retail pricing: A review and research agenda. Journal of Interactive Marketing, 24(2010), 138–154.
31. Gwebu, K.L.; Wang, J.; Hao, A.W.; and Hu, M.Y. Effects of price recommendations in name-your-own-price auctions. Journal of Electronic Commerce Research, 12, 1 (2012), 61–77.
32. Hinterhuber, H. Violations of rational choice principles in pricing decisions. Industrial Marketing Management, 47(2015), 65–74.
33. Hinz, O.; Hann, I.-H.; and Spann, M. Price discrimination in E-commerce? an examination of dynamic pricing in name-your-own price markets. MIS Quarterly, 35, 1 (2011), 81–98.
34. Hinz, O.; and Spann, M. The impact of information diffusion on bidding behavior in secret reserve price auctions. Information Systems Research, 19, 3 (2008), 351–368.
35. Hong, W.; Thong, J.Y.L.; and Tam, K.Y. The effects of information format and shopping task on consumers’ online shopping behavior: A cognitive fit perspective. Journal of Management Information Systems, 21, 3 (2004), 149–184.
36. Hsee, C.K.; and Blount, S. Preference reversals between joint and separate evaluations of options: A review and theoretical analysis. Psychological Bulletin, 125, 5 (1999), 575–590.
37. Kahneman, D. A perspective on judgment and choice: Mapping bounded rationality. American Psychologist, 58, 9 (2003), 697–720.
38. Kahneman, D. Thinking, Fast and Slow. New York, NY: Macmillan, 2011. 39. Kahneman, D. A Proposal to Deal with Questions about Priming Effects. Nature, 2012. 40. Kamins, M.A.; Dreze, X.; and Folkes, V.S. Effects of seller-supplied prices on buyers’ product
evaluations: Reference prices in an internet auction context. Journal of Consumer Research, 30 (2004), 622–628.
62 DIGITAL NUDGING.
41. Kannan, K.N. Effects of information revelation policies under cost uncertainty. Information Systems Research, 23, 1 (2012), 75–92.
42. Kling, R. Information systems in policymaking: Computer technology and organizational arrangements. Telecommunications Policy, 2, 1 (1978), 22–32.
43. Kraut, R.; Patterson, M.; Lundmark, V.; Kiesler, S.; Mukophadhyay, T.; and Scherlis, W. Internet paradox: A social technology that reduces social involvement and psychological well-being? American Psychologist, 53, 9 (1998), 1017–1031.
44. Krishna, A.; Wagner, M.; and Yoon, C. Effects of extreme-priced products on consumer reservation prices. Journal of Consumer Psychology, 16, 2 (2006), 176–190.
45. Li, X.; and Hitt, L.M. Price effects in online product reviews: An analytical model and empirical analysis. MIS Quarterly, 34, 4 (2010), 809–832.
46. Mazumdar, T.; Raj, S.P.; and Sinha, I. Reference price research: Review and propositions. Journal of Marketing, 69(2005), 84–102.
47. Nunes, J.C.; and Boatwright, P. Incidental prices and their effect on willingness to pay. Journal of Marketing Research, XLI (2004), 457-466.
48. Pennington, R.; Wilcox, H.D.; and Grover, V. The role of system trust in business-to- consumer transactions. Journal of Management Information Systems, 20, 3 (2003), 197–226.
49. Raudenbush, S.W.; and Bryk, A.S. Hierarchical Linear Models: Applications and Data Analysis Methods. Thousand Oaks, CA: Sage, 2002.
50. Rosen, S. Hedonic prices and implicit markets: Product differentiation in pure competition. Journal of Political Economy, 82, 1 (1974), 34–55.
51. Runtian, J. Study on rational assumption in experimental economics. Journal of University of Electronic Science and Technology of China, 28, 6 (1999), 591–595.
52. Schneider, C.; Weinmann, M.; and vom Brocke, J. Digital nudging: Influencing choices by using interface design. Communications of the ACM, 61, 7 (2018), 67–73.
53. Scott, J. Rational choice theory. In G. Browning, A. Halcli, and F. Webster (eds.), Understanding Contemporary Society: Theories of the Present. Thousand Oaks, CA: SAGE Publications, 2000, pp. 126–138.
54. Shen, F.; and Chen, Q. Contextual priming and applicability: Implications for ad attitude and brand evaluations. Journal of Advertising, 36, 1 (2007), 69–80.
55. Simonson, I.; and Drolet, A. Anchoring effects on consumers’ willingness-to-pay and will- ingness-to-accept. Journal of Consumer Research, 31(2004), 681–690.
56. Srull, T.K.; and Wyer, R.S. Category accessibility and social perception: Some implications for the study of person memory and interpersonal judgments. Journal of Personality and Social Psychology, 38, 6 (1980), 841–856.
57. Stern, B.B.; and Stafford, M.R. Individual and social determinants of winning bids in online auctions. Journal of Consumer Behaviour, 5, 1 (2006), 43–55.
58. Strack, M.; and Mussweiler, T. Explaining the enigmatic anchoring effect: Mechanisms of selective assessibility. Journal of Personality and Social Psychology, 73, 3 (1997), 437–446.
59. Tarafdar, M.; Gupta, A.; D Arcy, J.; Turel, O.; and Czerwinski, M. The digital future: Reflecting on the “dark side“ of information technology use. Americas Conference on Information Systems (AMCIS), 2012, pp. 1815–1819.
60. Tarafdar, M.; Tu, Q.; and Ragu-Nathan, T.S. Impact of technostress on end-user satisfaction and performance. Journal of Management Information Systems, 27, 3 (2010), 303–334.
61. Thorsteinson, T.J.; Breier, J.; Atwell, A.; Hamilton, C.; and Privette, M. Anchoring effects on performance judgments.Organizational Behavior andHumanDecision Processes, 107(2008), 29–40.
62. Todorov, A.; and Uleman, J.S. The efficiency of binding spontaneous trait inferences to actors’ faces. Journal of Experimental Social Psychology, 39, 6 (2003), 549–562.
63. Tripathi, A.K.; Nair, S.K.; and Karuga, G.G. Optimal lot sizing policies for sequential online auctions. IEEE Transactions on Knowledge and Data Engineering, 21, 4 (2009), 554–567.
64. Turel, O.; and Qahri-Saremi, H. Problematic use of social networking sites: Antecedents and consequence from a dual-system theory perspective. Journal of Management Information Systems, 33, 4 (2016), 1087–1116.
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 63
65. Tversky, A.; and Kahneman, D. Judgment under uncertainty: Heurstics and biases. Science, 185, 415 (1974), 1124–1131.
66. Wang, X.; and Montgomery, A.L. The effects of advertising on customer retention and the profitability of auctions. Research Showcase @ CMU: Carnegie Mellon University Tepper School of Business, 2003.
67. Weinmann, M.; Schneider, C.; and vom Brocke, J. Digital nudging. Business and Information Systems Engineering, 58(2016), 433–436
68. Wong, K.F.E.; and Kwong, J.Y.Y. Is 7300 m Equal to 7.3 km? Same semantics but different anchoring effects. Organizational Behavior and Human Decision Processes, 82, 2 (2000), 314–333.
69. Wu, C.S.; Cheng, F.F.; and Yen, D.C. The role of Internet buyer’s product familiarity and confidence in anchoring effect. Behaviour & Information Technology, 31, 9 (2012), 829–838.
70. Yen, C.-H.; and Lu, H.-P. Factors influencing online auction repurchase intention. Internet Research, 18, 1 (2008), 7–25.
71. Yoo, C.Y. Unconscious processing of Web advertising: Effects on implicit memory, attitude toward the brand, and consideration set. Journal of Interactive Marketing, 22, 2 (2008), 2–18.
72. Yuan, L.I.; and Dennis, A.R. The happiness premium: The impact of emotion on individuals’ willingness to pay in online auctions. System Sciences (HICSS), 2014 47th Hawaii International Conference on: IEEE, 2014, pp. 3120–3128.
About the Authors
Alan R. Dennis (ardennis@indiana.edu; corresponding author) is Professor of Information Systems and holds the John T. Chambers Chair of Internet Systems in the Kelley School of Business at Indiana University. His research focuses on three main themes: collaboration and social media; digital nudging; and information security. Dr. Dennis has written more than 150 research papers and has won numerous awards for his theoretical and applied research. He has also has published four books. His research has been reported in the press over 500 times, including the Wall Street Journal, The Atlantic, Le Figaro, Chile’s El Mercurio, China Daily, and other media outlets. He is the co-Editor-in-Chief of AIS Transactions on Replication Research and the President of the Association for Information Systems. Dr. Dennis is a Fellow of the Association for Information Systems.
Lingyao (Ivy) Yuan (lyuan@iastate.edu) is an Assistant Professor of Information Systems of College of Business at Iowa State University. She received her Ph.D. from Indiana University. Her research interests include the impact of non-cognitive behavior in decision making, especially the impact of emotion, on computer mediated communication, decision making, and collaboration. Dr. Yuan has conducted research in the fields of virtual reality, electronic commerce, and social media. She has published in Journal of Management Information Systems, Decision Sciences AIS Transactions on Human-Computer Interaction, Group Decision and Negotiation as well as the proceedings of several major conferences.
Xuan Feng (feng@ou.edu) is an assistant professor of Management Information Systems in the Price College of Business at University of Oklahoma. He received his Ph.D. from Indiana University. His research focuses on the use and impact of information technologies across indivi- duals and organizations, especially within healthcare organizations. Dr. Feng’s work has been published in International Journal of Medical Informatics and Journal of Global Information Technology Management.
Eric Webb (webbe3@ucmail.uc.edu) is an Assistant Professor of Operations and Business Analytics in the Lindner College of Business at the University of Cincinnati. He received his Ph.D. from Indiana University. Dr. Webb’s primary research interests include energy operations management and behavioral operations.
64 DIGITAL NUDGING.
Christine J. Hsieh (chsieh@alumni.nd.edu) is a data analyst and researcher, who helps organizations make data-informed decisions. She received her Master’s degree in Business in Information Systems from Indiana University. She has published in Journal of Organizational and End User Computing and has presented her work at several conferences.
JOURNAL OF MANAGEMENT INFORMATION SYSTEMS 65
Copyright of Journal of Management Information Systems is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder’s express written permission. However, users may print, download, or email articles for individual use.
- Abstract
- Introduction
- Theoretical Background
- Anchoring and Adjustment
- Influencing Price Anchoring through Priming
- Priming in the Online Context
- Study 1: Numeric and Semantic Priming via Online Advertising
- Numeric Priming
- Semantic Priming
- Hypotheses
- Method
- Participants
- Task
- Treatment
- Variables and Measurements
- Analysis
- Procedure
- Results and Discussion
- Study 2: Numeric Priming Replication
- Method
- Results and Discussion
- Study 3: Numeric Priming for Non-Auction Products
- Method
- Results and Discussion
- Study 4: Numeric Priming with Two Potential Anchors
- Method
- Results and Discussion
- Study 5: Semantic Priming with Two Potential Anchors
- Method
- Results and Discussion
- Study 6: Semantic Priming for Non-Auction Products
- Method
- Results and Discussion
- Study 7: Semantic Priming across Product Types
- Method
- Results and Discussion
- Discussion
- Implications for Research
- Implications for Practice
- Conclusion
- References
- Notes on contributors
lass has used in previous weeks.