Dynamic pricing in e-commerce
—Which purchase scenarios do consumers prefer?
Measuring consumers’ preference with dynamic pricing
under various online purchase scenario assumptions
By Tzuyun Hsu
MSc Marketing Intelligence, Faculty of Economics and Business University of Groningen
PO Box 800, 9700 AV Groningen (NL)
Supervised by Arnd Vomberg Co-assessed by Felix Eggers
Abstract
Dynamic pricing has become more common with the increased prevalence of online marketing, especially for online retailers and the industries of electronic consumer goods. Although it is a great strategy on profit maximization, the negative consumers’ reactions raise an important question about the viability of dynamic pricing over time. Understanding the preference of consumers is helpful for online retailers to avoid price unfairness perception by the use of dynamic pricing. As a result, the purpose of this paper is to identify consumers’ preference of dynamic pricing in purchase scenarios based on the price dynamics and online retailer characteristics in the Dutch market. Choice-based conjoint analysis is used to mimic consumers’ purchase behaviour and to capture consumers preferences on dynamic pricing purchase scenarios. Three main findings were provided in this paper, including (1) consumers are not significantly against dynamic pricing in the dynamic pricing purchase scenarios ;(2) consumers’ preference on dynamic pricing scenarios differs a lot individually based on gender or which segments he/she belongs to ;(3) consumers have significant preference on online retail brands and strongly on their general price level. The paper later provides plausible dynamic pricing strategies in the discussion stage.
Acknowledgement
In 2013, I started as a student of Accountancy at Soochow University in Taiwan. In these four years, I became familiar with analytical thinking in Financial Management, Accounting & Controlling. Upon finishing my Bachelor’s degree, I didn’t feel connected to having a career as an auditor or a financial manager. As a result, to pursue a broader career in the future, I started MSc. Marketing in the track of Marketing Intelligence at the University of Groningen in 2018.
Studying in Marketing Intelligence is one of the best choices I have ever made in my life. The program strikes a perfect balance between business, economics and data analytics in my opinion and provides very solid knowledge in Marketing and Data Analytics. I enjoyed the courses provided in the program. Although the process required a great deal of hard work and was rather suffering and stressing to be honest, what I have learned in the year are beyond my imagination. Now, almost at the end of my student life, I acknowledge all the efforts have paid off and I am glad to reach one of the most important chapters as my student career with this piece of work.
I would like to take this opportunity to express my appreciation here officially to the people who have helped me during my master thesis. I would like to thank Dr. Arnd Vomberg as my first supervisor for his professional input has encouraged me to progress further on my thesis. I would also like to thank Dr. Felix Eggers for his professional suggestions on my conjoint design. In addition, I would also like to thank the University of Groningen for the satisfying and valuable education. Last but not least, I would like to thank my family and friends for being supportive unconditionally through the journey of my Master’s. All in all, I am glad the journey has almost come to an end and I hope to become a stronger and more professional person. Thank you all.
Sincerely, Tzuyun Hsu
1. Introduction
Amazon.com was found that changing a microwave oven price nine times in one day, with the price ranging between $744.46 to $871.49 (Angwin and Mattioli 2012); Delta Airlines charged frequent flyers with higher prices for the same class tickets if they identified themselves on the company website (Sanburn 2012). Dynamic pricing has been applied in the airline and hotel industries. Research about the impact of dynamic pricing indicates that the adoption of dynamic pricing can increase firms’ overall profits (Elmaghraby and KesKinocak 2003). In recent decades, the use of Dynamic pricing has become more common with the increased prevalence of internet marketing, especially for online retailers and the industries of electronic consumer goods, entertainment services and sports (Haws and Bearden 2006; Li, Hardesty and Craig 2017).
Even though it is proved that dynamic pricing leads to an increase of profitability up to 25% (Garbarino and Lee 2003; Petro 2015), there is no rose without a thorn. It is also proven by behavioural price researchers that the negative effects of dynamic pricing on price-disadvantaged consumers’ perceived fairness, trust and repurchase intention (Garbarino and Maxwell 2010; Grewal et al. 2004; Haws et al. 2006). Once consumers notice the prices changes and perceive it as unfair, the negative effect with the customer-based price differences reveal the resistance of purchase. Amazon.com once priced the same DVD movies differently to consumers based on their online profiles and previous purchasing behaviours (Monroe 2003; Grewal et al. 2004 ). However, when customers found out about Amazon’s dynamic pricing strategy, their complaints against the company soon filled the chat boards. Amazon had to publicly claim that they would no longer use dynamic pricing (Streitfeld 2000). Therefore, online retailers such as Amazon, hold great interest in knowing the valuation of their platform not only on shopping and website user experiences but also on their pricing policies, especially when it comes to dynamic pricing. According to the retailer brand equity framework proposed by Anselmsson, Burt and Tunca (2017), pricing policy plays a determinant role on consumers’ choice of shopping platforms. Thereby, it is crucial for online retailers to understand the preference of dynamic pricing strategies before or during the adoption of dynamic pricing to find the sweet spot that keeps the balance between profit maximization and avoiding unfairness perception from consumers and to increase the competitiveness of the shopping platforms with effective pricing strategies.
analysis and a field experiment to investigate consumers’ preferences regarding several pricing programs, including dynamic and static ones. Similarly, Schlereth, Skiera and Schulz (2018) examined consumers’ preference on dynamic pricing plans by a choice-based conjoint analysis and conduct a Hierarchical Bayes extended covariate logit estimation to link consumers’ preferences to the antecedents from the proposed conceptual framework at the individual level. Lastly, they measured respondents’ probability of switching from a time-invariant pricing plan to a time-variant one. On the other hand, Solgaard and Hansen (2003) measured the store choice decision of supermarket shoppers by both standard logit model and Hierarchical Bayes model. Although there are cases of preference measurements about dynamic electricity pricing and grocery retailers separately, there are few papers examine consumers’ preference on dynamic pricing for online retailers in e-commerce. Thereby, this paper aims to bridge the gap by empirical analyzing consumers’ preference in dynamic pricing under purchase assumptions in e-commerce.Purchase simulation is applied in the paper by adding online retailer characteristics into the analysis and adopting a choice-based conjoint model to mimic consumers’ behaviour of buying consumer electronics in a questionnaire. Consumers’ preference is estimated at three different levels which are respectively the aggregate, segment and individual levels.
2. Literature review
2.1 Dynamic pricing
Properly implementing dynamic pricing can improve firms’ revenues and profits by up to 8% and 25% respectively (Sahay 2007). Firms are able to enhance profitability by reducing the costs of changing prices, implementing real-time price tests, or customizing prices dynamically depending on demand and buyers’ shopping preferences and patterns (Kannan and Kopalle 2001). In recent decades, the increase in the adoption of dynamic pricing can be attributed to the increased availability of demand data, the ease of changing prices due to advanced technologies, and the availability of decision-support tools for demand data analysis and customized pricing (Elmaghraby and Keskinocak 2003 ). Dynamic pricing is also being discussed as one method of demand-side management (DSM) in electricity systems ( Dütschke et al. 2013). Based on the rationales of pricing policies, dynamic pricing in electricity markets are divided into two categories, real-time pricing and personalized pricing (Yaghmaee, Kouhi and Garcia 2016). The previous oneis mostly dependent on the grid's dynamic load behaviour and adjusts real-time prices based on grid's real-time demand level dynamically. The latter one not only makes use of the grid's real-time consumption data but also considers consumption levels of each customer and defines real-time prices individually. According to the definition of dynamic pricing from Haws et al. (2006), it is a pricing strategy in which prices vary over time, consumers and/or circumstances in the purpose of profit maximization. Applying the same rationales of assortments in electricity markets to the definition, dynamic pricing in online retail industries can also be distinguished as time-based pricing and personal pricing, depending on the base is either customer-specific data or time. This paper partially adopted the same definition of dynamic pricing from Haws et al. (2006) and focused more on time-based dynamic pricing.
result showed that the mean benevolence trust is significantly lower due to the dynamic pricing events which are given heavier weight online retail brands are expected to have a significant effect on consumers’ choice of substantially in the format of overall trust and leads to a marginal decrease in overall trust. Consequently, the adoption of dynamic pricing may lead to negative consequences to the seller, including consumer boycotts (Goldman 1994), civil action(Kaufmann Ortmeyer and Smith 1991), or lower sales (Grover 1994). Consumers’ actual reactions to cope with the price unfairness perception depend on the degree of how buyers feel disadvantaged(Xia, Monroe and Cox 2004). On the other hand, perceived unfairness is asymmetrically more severe when the inequality is to the buyer’s disadvantages than when it is to the buyer’s advantage ( Ordóñez, Connolly, and Coughlan 2000). Martins (1995) also finds that the perceived fairness effect of a comparable other buyer paying less is stronger than when the comparable other pays more. For the sellers, the negative influence driven by price unfairness perception on dynamic pricing harms the long-term profitability. Therefore, it raised a question about the viability of dynamic pricing over time.
As most of the research exclusively asked respondents about price fairness, Schlereth et al. (2018) considered not only price fairness but economic antecedents as covariates in the process of examining consumers’ preference on dynamic pricing plans in electricity markets. They proposed a conceptual framework of antecedents of preferences for dynamic pricing plans in electricity markets and it consists of price fairness consideration and economic considerations. The results showed that economic considerations have a stronger effect on the choice of dynamic pricing plan than price fairness considerations in electricity markets. With the highest direct impact on consumers’ preference on any time-variant pricing plan, price consciousness is the most important economic antecedent and the predictor of whether a consumer would be willing to abandon the time-variant pricing plan. The results raised an important question about the role of price consciousness relative to consumers’ preference on dynamic pricing scenarios in e-commerce.
2.2 Preference measurement
attributes,i.e., a combination of attribute levels. The main concept of preference measurement is to capture how much each component value itself (i.e., attribute level) and hereinafter generating quantifiable outcomes. Such valuation results are produced in a utility function which translates the specific characteristics of a product/service into the perceived preferences of consumers. In addition, the function is capable of predicting purchase decisions in various conditions (e.g., product modifications, pricing scenarios). Based on the predictions, the decision-makers could identify the attributes and characteristics with the most impact of consumers, or alternatively to measure price elasticities as well. Except for pricing and product management, preference measurement itself can evaluate the potential a brand extension into other product categories or a brand’s monetary value. Furthermore, it is also applied to predictions about the diffusion of innovations which usually combine the use of the Bass models or market segmentation based on consumers’ preferences (Eggers et al. 2011) . According to Eggers et al. (2011), methods of preference measurement can be categorized into three groups that are compositional, decompositional and hybrid approaches. Among the methods, conjoint analysis as a decompositional approach, has become the most popular method to measure consumers’ preference structure. Conjoint methods can be distinguished by the way the respondent evaluates the examined product. Traditional conjoint methods use a ranking or rating procedure whereas choice-based conjoint methods ask consumers to repeatedly choose their most preferred product/ service among multiple alternatives across the choice sets. However, rating is not the typical behavior that a person usually has in the process of purchase decisions whereas choosing the most preferred product/service is the natural manifestations of buying a specific product and it is intuitive for respondents to accomplish. Based on the characteristics, choice-based conjoint methods are generally more suitable for forming purchase scenarios in this paper.
increases the likelihood of switching from static pricing plans to time-variant pricing plans. On the other hand, Solgaard et al. (2003) observed a significant variance estimate about price level (at the 10% level or beyond) in households’ preference on grocery retailer choice which indicated the extent of heterogeneity among households insensitivity to price level by estimating consumers’ preference at the household level with a Hierarchical Bayes model. Sinha, Ashill and Gazley (2008) measured brand equity in a product category at the individual level by a preference measurement about consumers’ purchase decision and then decomposing the brand attribute into five brand equity assets sub-components, including brand awareness, brand personality, trust, pride and perceived quality. The result showed a large extent of heterogeneity on consumers’ preference of the brand assets sub-components, indicated by the means and the standard deviation of the brand equity assets variables. According to the heterogeneity of consumers preference on dynamic electricity pricing, grocery retail stores and product brands in previous studies, it is expected that consumers’ preference on dynamic pricing scenarios in e-commerce similarly differs across consumers or segments. In such a case, estimations at the segment and individual level seem to be necessary in this paper to understand the full picture of consumers’ preference on purchase scenarios under dynamic pricing.
Key attributes
Price changing frequency and price variation
It is assumed that a consumer’s choice of a preferred dynamic pricing purchase scenario is based on the perceived utility that s/he derives from the purchase scenario. In the study by Dütschke et al. (2013),
pay differently depending on the time of use and/or on the current load at the household level (Dütschke et al. 2013). The perception of price complexity leads to a decrease in depth of information processing. In such a case, the likelihood that consumers apply heuristic processing on tariff evaluation increases which decreases consumers’ thought confidence and increases perceived bill amount and ends up resulting in lower behavioural intention (Layer et al. 2017). Prior pricing research also proved that relying on heuristics can substantially distort consumer price evaluations (Morwitz et al., 1998). That could be one of the reasons why consumers still refuse time-variant pricing plans which allows them to realize lower prices if they are willing to reduce demand during peak times in electricity markets. Thus, lower pricing changing frequency and price variation are more preferred by the consumers in general. In addition, price is generally found to have a negative effect on purchase intentions (Lichtenstein, Ridgway, and Netemeyer 1993). As a result, it is expected in this paper that consumers prefer purchase scenarios with lower perceived price complexity in e-commerce. That is, lower price changing frequency and smaller size of price variation are expected to be preferred by consumers.
H1: Lower price changing frequency has a positive effect on consumers’ preference
H2: Smaller size of price variation has a positive effect on consumers’ preference
Online retail brand
retailers’ credibility in non-contractible aspects of the product and service bundle, such as shipping reliability and as a signal, or a bond for consumers to identify retailers with higher service quality.
Other than service quality, Sheth, Newman and Gross(1991) argued that brand choice is influenced by factors not just functional attributes.Specifically, trust and pride are present the part of brand equity about the organizational associations in a product category, e.g., television industries (Sinha et al. 2008). As a set of an asset representing brand equity, the brand equity assets are proposed by Aaker (1991; 1996a) to measure the “incremental utility” associated with a brand name that is not captured from functional attributes and the added value endowed by the brand ( Kamakura and Russell 1993 ; Farquhar 1989). The empirical results by Sinha et al. (2008) showed that except for brand personality, all the other brand equity assets have an effect on consumers’ purchase decisions and that organisational associations, measured by the two variables, pride and trust, play an important role in the development of brand equity. Although perceived quality has the highest impact on consumers’ choice among all the brand asset attributes, the importance of organizational associations is indicated by the fact that the combined effect of pride and trust is more than the effect of perceived quality. Thus, it is concluded from the result that organizational associations have a strong effect on brand equity in a product category. However, measuring retail brand equity based on the framework for product brands may not present the overall picture of how retailer equity influences consumers’ purchase decision-making. Anselmsson, Burt and Tunca (2017) presented a retailer brand equity framework, extending the conceptual brand resonance model presented by Keller (2001) by incorporating retailer-specific image dimensions such as customer service, pricing policy, physical store, and retailer trust which help us to understand and explain how retailer equity can build customer loyalty.The conceptual domain of the extended scale consists of seven dimensions, namely awareness, product quality, customer service, pricing policy, retailer trust, physical store, and loyalty. In this paper, as three other attributes - general price level, price changing frequency and price variation captures customers’ perceived price perceptions with the pricing policy dimension in the retail brand equity framework, the online retailer attribute evaluates all the other six dimensions at an overall point of view. According to the papers, it is expected that online retail brands similarly have a significant effect on consumers’ choice of dynamic pricing purchase scenarios as product brands.
General price level
When it comes to preference of dynamic pricing, price consciousness, as the most important antecedents, had the highest direct impact on preferences for any time-variant pricing plans among all the proposed antecedents in the electricity market (Schlereth et al. 2018). However, the dynamic pricing policy is not the only element that triggers consumers’ price consciousness in online purchase scenarios and could possibly influence consumers’ online purchase decisions. The price level of online retailers could also trigger consumers’ price consciousness in dynamic pricing scenarios and be decisive to consumers’ online purchase decision-making. In the paper about consumers’ grocery retail choice, the price level is included as one of the variables to measure consumers’ decision-making on grocery retail stores format and it is proved to be an important driver (Solgaard et al. 2003). Accordingly, as online retailers are also included in the dynamic pricing purchase scenarios, it is reasonable to include the price levels of online retailers in the paper to capture consumers’ price consciousness based on the retailer characteristics in dynamic pricing purchase scenarios in e-commerce and are expected to have a significant effect on consumers’ choice of dynamic pricing purchasing scenarios in this paper. Proposed as the antecedents of preference for dynamic pricing plans in electricity markets, economic considerations including price consciousness, have a stronger effect on the choice of a time-variant plan than price fairness consideration in electricity markets(Schlereth et al. 2018). In addition, in terms of grocery retail stores, price apparently plays a decisive or even dominant role and it is a much more important role than in the positioning of products and brands (Solgaard et al. 2003). In a study about consumers’ perceptions of grocery retail chains, Solgaard (2000) observed that although discount grocery retail chains are rated very poorly compared to other supermarket formats on a whole range of store values except one, good prices (in all 21 value aspects were rated), they are still growing and gaining market share over time. Furthermore, store price perceptions are central to retailer brand equity, mainly because independence of other associations, customers take prices into consideration before making a purchase from a retailer (Ailawadi and Keller 2004). It seems that grocery retail store choice is primarily motivated by utility considerations rather than by hedonic considerations and that grocery shopping is a functional activity, where consumers’ perception of price plays a major role.
preference on dynamic pricing plans in electricity markets, the concept of decomposing dynamic pricing purchase scenarios for conducting preference measurement essentially requires respondents’ cognitive efforts to roughly compute the approximate final purchase price of each purchase scenario which refers to a high level of perceived price complexity for consumers. Attributes that are directly related to the final purchase prices as economic considerations with lower levels of perceived price complexity would have stronger effects on consumers’ choice of dynamic pricing purchase scenarios than those which are not directly linked to, as they may trigger consumers’ evaluation based on simple heuristics. Obviously, the attributes presenting the dynamicity of price require more cognitive efforts, compared to the price level or the retail brand. In addition, it is also indicated that price has a larger effect on consumers’ choice than brand assets and brand personality (Sinha et al. 2008). Therefore, it is expected in this paper that the price levels of the retailers not only have a significant effect but play a decisive role in consumers’ choice of dynamic pricing purchase scenarios in e-commerce. As price is generally found to have a negative effect on purchase intentions (Lichtenstein et al. 1993), it is also expected that the lower general price level has a positive effect on consumers’ preference.
H4: Lower general price level has a positive effect on consumers’ preference
Moderator : gender
Gender may potentially moderate the relation between the attributes and consumers’ choice of dynamic pricing purchase scenarios. It is indicated that gender can affect online repurchase intention as a moderator between relational benefits (i.e. product quality and e-service quality) and perceived value (Fang, Wen, George, and Prybutok 2016) . At the same time, it is revealed that males have more functional buying attitudes, holding stronger utilitarian values that emphasize efficiency and effectiveness in offline shopping whereas their focus on functionality becomes more pronounced in an online shopping environment (Dittmar, Long and Meek 2004). That is, men are also more outcome-focused and placing higher value on efficiency, compared to women ( Mattila, Grandey and Fisk 2003). According to social role theory, men are also more willing to take risks than women, because men are socially expected to engage in individual riskier behaviour ( Walsh, Evanschitzky and Wunderlich
of differences in thinking or purchase behaviors, it is proven by Bhaskar et al. (2017) that there is no difference in the gender-wise reaction towards dynamic pricing. That is, males and females all feel offended by the dynamic pricing strategy. Based on the different outcomes, it raised an important question about consumers’ choice of dynamic pricing purchase scenarios which includes not only the preference of price dynamics but also online retail brands and the price level.
3. Methodology
Understanding the preference of consumers in advance could be helpful for online retailers to avoid price unfairness perception from the use of dynamic pricing. As a result, conjoint analysis is made in understanding Dutch consumers’ preferences on dynamic pricing with several ‘choices’ of a combination of attributes and also possibly to identify plausible dynamic pricing strategy in the discussion stage.
Choice-based conjoint analysis
Conjoint analysis is one of the most popular methods within preference measurements. Choice-based conjoint (CBC) is the most popular conjoint among several product concepts which use choices as the dependent variable to mimic consumers’ behaviour when they are making purchases(Eggers and Sattler, et al. 2018 ). A dual response choice-based conjoint analysis(DR-CBC) is adopted in the report to measure consumers’ preference on different dynamic pricing scenarios in the Dutch market. The estimated utility of a dynamic pricing purchase scenario is based on the assumption that a consumer’s choice reflects their highest utility (Green and Rao 1971). The dual response is to provide participants with a no-choice option as a separate question under each choice set. It is discussed that the implication of the dual response is to observe the preferred alternative even if it is not acceptable to purchase and the increased salience of the no-choice option leads to more realistic predictions of adoption shares (Wlömert and Eggers 2016). In general, measuring consumers’ preferences by conjoint analysis can be composed into five steps followed by a study design, a choice design, choice elicitation, estimation, analysis and interpretation (Eggers and Sattler 2011). The steps are replicated in this study to capture the preferences.
3.1 Study design
determines the patterns of the price dynamicity. Another core attribute related to dynamic pricing is
Rate:price spread(i.e., the prices per kWh) which refers to the variation of the price rates based on the current price rate and determines the size of price changes. Based on the previous design, price changing frequency and price variation rate are proposed as the central attributes to present the price dynamicity of e-commerce in this study. Choice-based conjoint analysis is used not only to mimic consumer purchase behaviours but also to simulate consumers’ comparing behavior on different shopping platforms in the study. In addition, a pricing scheme incorporates several attributes, each of which may vary resulting in a large number of possible combinations. However, dynamic pricing is an abstract concept that consumers are likely not to be aware of how it actually works in essence. Thus, it is necessary to restrict the analysis to a relatively small number of attributes to enable participants to understand what is expected from them. Therefore, dynamic pricing purchase scenarios are decomposed into 4 attributes which are Online retailer, General price level of the platform, Price changing frequency, Price variation in percentage. The study design is presented in Table 1. Consumer electronics are chosen to be the focal products in the survey scenarios and the current study to facilitate measuring consumers’ preference for dynamic pricing. Consumer electronics are commonly being sold online across different online shopping platforms in the Netherlands. Compared to other categories of products(e.g., clothing, grocery), Dutch consumers are relatively more familiar with the application of dynamic pricing on consumer electronics as there are relatively more consumer electronics online retailers are currently adopting dynamic pricing, such as Mediamarkt, Amazon.
Online retailers
The first attribute is Online retailers The purpose of the attribute is to capture consumers’ brand preferences of online shopping platforms. Amazon, MediaMarkt and bol.com are three well-known online retailers in the Netherlands. Amazon is one of the largest e-commerce destinations in the US and Europe in 2019 (Ecommerce News Europe 2019). It is known for changing price several times a day according to market demand with the advanced algorithms. Without a Dutch version site until 2019, Amazon used to
Netherlands. Bol.com does not release any public statements about using dynamic pricing. Thus , it is included in the design not just for the largest sales in the Netherlands but also for the purpose of evaluating consumers’ preference on online retailers without using dynamic pricing. Meanwhile, as the study is focusing on consumers’ preferences of dynamic pricing on consumer electronics in the Dutch market, three of the online retailers are chosen to present the attribute.
General price levels
To measure consumers’ preferences on dynamic pricing, the final purchase price is intuitive to be an influential variable. However, the price could be too dominant compared to other attributes and it may lead to the difficulty of capturing consumers’ preference on dynamic pricing. Therefore, the general price level is used to indirectly present the price effect based on the retailer characteristics. The purpose of the attribute is to reflect the general pricing level of consumer electronics from each retailer. Consumers anchor on a reference price that is a weighted average of the lowest and the most recent price (Nasiry and Popescu 2011). Based on the concept, the reference price level here is similarly designed as the weighted average price level of consumer electronics in the market. The attribute represents the general price level of consumer electronics in each shopping platform when it compared to the weighted average price level in the market. For example, US retailer Best Buy is well-known for a lower price level on consumer electronics than other competitors. On the other hand, It was found ( Brynjolfsson and Smith 2000) that Internet retailers’ price adjustments over time are up to 100 times smaller than conventional retailers’ price adjustments — presumably reflecting lower menu costs in Internet channels. Furthermore, it is common that a popular product has the same prices across different online shops or only has a small difference in prices due to the competition of online retail. As a result, the general price level difference in the study is in a range of 20% between three attribute levels which are respectfully 10% higher than the average, same as the average and 10% lower than the average to reflect the competitive pricing online.
Price changing frequency
by the transient inconsistency in Amazon’s infrastructure, rather than actual price changes by sellers. In addition, Amazon states that it can take 15 minutes for all the systems to converge to the new price. As a result, the attribute levels of Price changing frequency are designed as every 2 hours, 6 hours and 12 hours to present the real price changes by sellers.
Price variation in percentage
The last attribute is Price variation in percentage. It refers to how much the price changes per time, based on the previous latest price. The purpose of the attribute is to capture consumers’ sensitivity to the size of the price change in percentage. Among 4 attributes, price variation in percentage is the only attribute that is directly related to the number of the final purchase price. To present the dynamicity and uncertainty of the price changes, the attribute levels(change ±5%, ±10%, ±25%) are designed with “±” to show the possibilities that price could either increase or decrease. For example, the newly generated price of a product could be 10% higher or lower than the previous price.
Table 1. Attribute levels of the conjoint experiment design
Level 1 Level 2 Level 3
Online retailer Amazon Mediamarkt bol.com
General price level of the retailer 10% higher than average 10% lower than average Same as the average Price changing frequency
Every 2 hours Every 6 hours Every 12 hours
Price variation in percentage Change by ±5% per time Change by ±10% per time Change by ±25% per time
3.2 Choice design
quickly becomes hard to handle in an experimental survey (cite) the fatigue effect for the respondents. Thus, a full factorial is only required if all main effects and all potential interaction effects should be estimated(Eggers and Sattler, et al. 2018 ). Thus, the conjoint experiment is designed as a fractional factorial analysis (i.e., only a part of the attribute level combinations are used across the choice sets). The choice design follows the four requirements of balance, orthogonality, minimal overlap and utility balance (Eggers and Sattler 2011). Those requirements are achieved by a computerized search due to the great complexity of optimization. In the conjoint experiment, there are a total of 8 choice sets in the survey. As the concept of decomposing dynamic pricing into attributes is relatively abstract and not intuitive, each choice set is designed to contain only two alternatives to avoid the fatigue effect.
3.3 Choice elicitation
The choice sets from the optimal design are incorporated in a survey which is handed to random participants. During the survey ( Appendix 1 ), a brief explanation of an online shopping scenario and the attributes is introduced to the participants to make sure the participants understand each alternative before they proceed to the choice sets. In addition, the CBC analysis also consists of no-choice options that are below every choice set (Appendix 1). The no-choice options enable the participants to show their preferred alternative even if all the alternatives are not acceptable to purchase.
3.4 Data
Date collectionThe data used in the analysis was collected through a survey from Preference Lab (the software used to collect the data in this paper). The survey was designed with questions for respondents’ demographic information and questions for conjoint analysis part. It was distributed for two weeks, from 12th to 26th December 2019. The requirement of the survey is that the participant is currently living in the Netherlands or has once lived in the Netherlands for more than six months. Most respondents were approached via social media. Studies about conjoint analysis for commercial purposes generally have between 100 to 1000 respondents. Analysis such as the one here with lower number of attributes generally requires between 100 to 150 participants (Cattin and Wittink 2006).
Data exploration
or exiting the survey. The first two responses were removed before the analysis due to the change of demographic questions and attributes. The average time to complete the survey is 4.9 minutes with a median of 3.9 minutes. Incomplete rate of the survey(73.2%) was high, with 306 respondents. For those respondents who did not finish the survey, most of them stopped at the beginning questions, such as introduction or asking demographics . There were only 132 respondents who answered the CBC questions and only 20 of them did not finish it all. A plausible reason could be going through multiple choice scenarios exhaust the respondent’s attention. A total of 54.46% of the respondents were female and 43.75% were male while 1.79% of the respondents prefer not to tell their gender. All the respondents were between 18-44 years old, 66.96% of them are between 18-25 years old and 25.89% of them are between 26-34 years old. Nearly half of the respondents (48.21%) hold a bachelor’s degree. 40.18% of respondents hold a master degree or above. On the other hand, as the survey was distributed mostly among students, 63.68% of the respondents have no income(36%) or monthly income after tax under 1000 euro(27.68%). Only nearly 10% of the respondents have a monthly income above 2500 euro after tax. Meanwhile, as the analysis is focusing on the preferences of Dutch consumers on Dutch online shopping platforms, the survey is only distributed to people who are currently living in the Netherlands or had stayed in the Netherlands for over six months. Therefore, most of the respondents (86.61%) are currently living in the Netherlands. However, the majority of the respondents(62.5%) are internationals, only 37.5% of the respondents are Dutch. A possible reason could be that there are a lot of international students studying in the Netherlands.
Figure 1. Correlation matrix of the attributes
3.5 Estimation
which is corresponding to the case for this study. Therefore, Hierarchical Bayes analysis estimates consumers’ preference at an individual level. Equation(1) and (2) are shown below:
(1)ni
ni
ni
U
= V
+ ε
• Uni : overall utility of consumer n for dynamic pricing scenario i
• Vni : systematic utility of consumer n for dynamic pricing scenario i , rational utility • ε ni : error component of consumer n for dynamic pricing scenario i , error term
(2)
• : number of attributes (1=Online retailers, 2 = general price level of the retailer, 3= Price changingk
frequency, 4=Price variation in percentage)
4. Results
Before analyzing the results, there are assumptions to be checked before interpreting the preference(Eggers and Sattler 2011). First, there are no external effects, and all the attributes that influence the buyer’s decision on dynamic pricing scenarios are included in the study. Marketing efforts are equally effective in different scenarios and across online retailers. Buyers are aware of dynamic pricing. In the survey, all alternatives are equally available. Lastly, there are no switching costs between dynamic pricing scenarios.
4.1 Aggregate level
p>0.05 ), in such order. Surprisingly, both price changing frequency and price variation in percentage, presenting dynamic pricing in the purchase scenario, does not significantly influence consumers’ purchase decisions. Overall, among all the options of purchase scenarios, the general price level of the retailer that is 10% lower and the same as the average are widely preferred.
Table 2. The utility estimates of Model 1 Attributes Estimates Std. Error
4.2 Attribute Importance
Table 3. Attribute importance of part-worth model
Attribute Importance
Online retailer 9.60%
General price level of the retailer 85.08%***
Price changing frequency 3.76%
Price variation in percentage 1.55%
Model 2 is combined with the quadratic effect of price variation in percentage and the interactions between the general price level of the retailer, price variation in percentage and gender. The previous interaction is hypothesized in H3. The model shows better explanatory power with a significant increase (p=0.03008) with adjusted Pseudo R square closer to 0.3 which indicates a better model, compared to the first model.
Table 4. The utility estimates of Model 2
Table 4 shows the attributes and their hypothesized effect in conjoint analysis. According to the result, there is a significant interaction effect( β =0.402619,p<0.05*) between gender and 10% higher general price level of the retailer( females=1 ). It is revealed that women are relatively not that price-sensitive about the 10% higher general price level of the retailer( β =-1.375188+0.402619=0.972569) whereas men against more on it ( β =-1.375188, p<0.05 ***). Changing the format of price variation in percentage from part-worth to ideal point, the attribute itself became significant . Meanwhile, There is also a significant interaction effect between gender and price variation in percentage. It is revealed that gender moderates consumers’ preference for price variation. Men and women both prefer larger price variation but the extent of their preference is different. When the price changes, men prefer more on larger price variation than women. For example, when the price variation changes by 25%, the attribute utility for men is 0.41767 while the one for women is 0.19986 (Appendix 2). Compared the result to Model 1, men are also more sensitive to general price level of the retailer than the average, indicated by the increase in the absolute value of the general price level estimates from Model 1 to Model 2(higher: β = -1.1213615→ β =-1.375188; lower:β =0.9925146;β =1.100077). In conclusion, gender does have a significant influence on consumers’ preference of dynamic pricing, especially on price variation.
4.3 Segment level
Table 5. Model fit comparison of latent class analysis
Model 3 Model 4 Model 5
Number of classes 3 4 5
Log-likelihood -363.29 -347.24 -331.87
AIC 776.9 769 757.3
Adjusted r2 0.362 0.373 0.384 Classification error 0.09248 0.09387 0.07962
According to Table 6, The four segments can be identified mostly based on the preference extent of general price levels of the retailer. Respondents in Class 1 and 2 made their purchase decisions relying on either strongly positive or strongly negative preferences of the attribute. On the contrary, others in Class 3 and 4 shows their preferences on both general price level of the retailer and online retail brands within their purchase decisions. With the predicted probability of 43.1% to belong to, Class 2 seems to be the largest one. Respondents in this segment have the strongest preference on the general price level of the retailer that is 10 % lower than the average, indicated by the large and significant coefficient ( β = 5.013921, p<0.05 ). They also have an extreme negative and significant preference on the general price level that is 10% higher than the average( β =-5.896922, p<0.05). It is obvious that the general price levels are the only significant attributes for the respondents in Class 2. Similarly, the respondent in class 1 also made their purchase decisions mostly based on the general price level attributes while also having strong preference on the one with 10% lower than the average( β =2.053151,p<0.05) and a much more negative preference on the one with 10% higher( β =-2.082066,p<0.05). The predicted probability of an individual to belong in Class 1 is 19% which is much smaller than Class 2. The main difference between Class1 and Class 2 is the extent of their preference on the general price level of the retailer.
4 prefer Amazon the most (β =1.535389, p<0.05 ) while bol.com is negatively preferred ( β =-0.657,
p<0.05 *) and Mediamarkt is the least preferred one( β =-0.878182,p<0.05). Another interesting insight of Class 4 is that it is the only group of consumers who do not make their purchase decisions relying heavily on the general price levels of the retailers, indicated by the p-value of the attributes. Overall, the majority of the respondents show their corresponding preference on lower general price levels and care a lot about the attribute when purchasing consumer electronics. Nearly 37% (Class 3 and 4) of the respondents show their preference significantly on online retail brands. Last but not least, all of the respondents do not show significant preference on the attributes presenting the dynamicity of prices . Based on the analysis, the four segments can be labelled respectively as “ utilitarian”,“pro-utilitarian“, “inbetweeners“ and “Amazon advocates”.
4.4 Individual level
Based on the result of Model 2 and Model 4, it is shown that there consumers’ preference differs with a high degree of heterogeneity, indicated by the changes in estimates and significance between different levels of estimation. Thus, to understand more consumer heterogeneity, Hierarchical Bayes analysis is used to obtain individual parameter estimates.
Figure 2 shows a summary of the individual estimates by boxplots. With a wider spread of the boxplots, it is obvious that the preference of consumers in the Dutch market about general price level of the retailer differ the most. In addition, consumers also have an extreme preference in 10% higher general price level of the retailer and 10% lower general price level of the retailer, indicated by the different positions of the boxplots. Similar to the results from the estimation of the part-worth models, the general price level of the retailer that is 10% lower and the same as the average is more preferred than the one with the higher price level. On the other hand, it is also found that consumers’ preference of Amazon and bol.com relatively differs more than the one of Mediamarkt. According to the spread of the boxplots, consumers in the Dutch market relatively prefer bol.com more whereas they relatively do not prefer to purchase consumer electronics on Amazon. Among the three levels of price changing frequency, consumers show nearly no difference in the preference of changing price every 6 and 12 hours. The insight about price variation is that consumers have a wider spread of preference when price changes by ±25% per time.
Table 7. Selection of observations and their HB estimates
buyers’ interest from Ordóñez et al. (2000 ). Overall, Table 7 demonstrates how each respondent has various preferences and how they could actually be identified via the Hierarchical Bayes estimation. Conclusively, based on the analysis from three different levels, the result of hypothesis testing is listed below in Table 8.
Table 8. Hypothesis test overview
Hypothesis Supported
H1 Lower price changing frequency has a positive effect on consumers’ preference ✓
H2 Smaller size of price variation has a positive effect on consumers’ preference ✓/
H3 Online retailers has a significant effect on consumers’ preference ✓/
H4 Lower general price level has a positive effect on consumers’ preference ✓
5. Conclusion
Although dynamic pricing has been discussed for some time now and several field trials have been initiated, consumers’ preference on dynamic pricing is mostly measured only in electricity markets. Hardly any research has been done to analyze consumers’ preference on dynamic pricing in e-commerce– at least to the best knowledge of the author of this paper. This paper aimed to measure consumers’ preference of dynamic pricing in purchase scenarios in e-commerce. Dynamic pricing purchase scenarios are designed to mimic consumer purchase behaviors by using choice-based conjoint analysis. The analysis of conjoint data presented offers the hypothesized utilities of different dynamic pricing purchase scenarios at different estimation levels.
5.1 General Discussion
Preference on the pricing dynamics
The findings show that it is suggested for online retailers to estimate consumers’ preference at the segment level or individual level, indicated by the heterogeneity of consumers’ preferences and the extent of preference found across different levels of estimation. Similar to the result in electricity markets from Schlereth et al. (2018), respondents showed different preferences and the preference extent on pricing dynamics in the dynamic pricing purchase scenarios. It is also found that the result of this paper is
2017). Thus, consumers relatively have more different extents of preferences in the price variation of ±25%. Similarly, price unfairness perceptions or high perceived price complexity also possibly leads to the relatively larger heterogeneity found in the preference extent of the price changing frequency of every 2 hours. Compared with the two other price changing frequencies, consumers’ sensitivities of price change frequency differs more across consumers for the highest price changing frequency. On the other hand, gender was proved to be a moderator on consumers’ preference and the heterogeneity of preference extent on price variation was found between men and women. That is, men are more likely to prefer more when the price variation is larger, compared to women’s preference in the same situation. The result could be explained by the findings from Walsh et al. (2008) and Bae et al. (2011) that men are more willing to take risks than women. Conclusively, consumers’ preference on dynamic pricing in e-commerce is corresponding to the preference in electricity markets (Dütschke et al. 2013) that consumers are open to dynamic pricing but prefer simple ones, except for price variation of ±25%. At the same time, it is more possible to find a larger heterogeneity among consumers’ preference when the purchase scenarios are highly dynamic and gender do have a significant effect on consumers’ preference of dynamic pricing.
Preference on online retail brands and general price levels
The findings about the preference of price levels are consistent across different levels of estimation that the lower the price level is, the more consumers prefer whereas the extent of preference differs across consumers. In addition, it is found that the preference between price level that is 10% higher and that is 10% lower is asymmetric, indicated by the estimates’ absolute value of the general price level that is 10 % higher across models are always larger than the one from the general price level that is 10 % lower. That is, consumers react negatively more when the price level in the purchase scenarios is 10% higher than the average. A possible reason to explain the asymmetric preference between two levels could be that consumers felt the inequality is to their disadvantages when the price level is 10% higher than the average in the dynamic pricing purchase scenario, which causes a stronger feeling of unfairness ( Ordóñez et al. 2000).
on the 10% lower one. Apart from the asymmetric preference between two price levels, another possible reason could be that different levels of price consciousness triggered as one of the economic antecedents leads to the consistent preference on the lower or higher price levels but different extents of preferences across consumers(Schlereth et al. 2018). On the other hand, it is found that men are more sensitive to the price level than women do in general, indicated by mens’ preference on the 10% higher price level is more negative than womens’ and the aggregate one. The differences could be explained by the different ways of processing income data between men and women(Meyers-Levy et al. 2015). It is expected that general price levels triggers consumers’ price consciousness and their decision-making process based on simple heuristics due to the lower level of perceived price complexity. Thus, men are more likely to have a stronger emphasis on the price levels as they are selective data processors, relative to women.
main dimensions included in organizational associations, it is possible that the lower level of trust on the online retail brands that have been adopting dynamic pricing over time negatively affect consumers’ retail brand preferences(Garbarino et al. 2003; Sinha et al. 2008).
According to the results in this paper, the best combination of attributes at the aggregate level is any combination with the general price level of the retailer that is 10% lower than the average, regardless of online retailer brands, price changing frequency and price variation. The “utilitarian” and “pro-utilitarian” also share similar preferences on the combination of attributes as the one at the aggregate level. However, for the “inbetweeners”, the best combination is any combination with bol.com and general price level of the retailer that is 10% lower than the average whereas the best combination for the “Amazon advocates” is any combination with Amazon. Men are more sensitive to the general price level of the retailer and price variation in percentage in the case. The best combination of attributes for male consumers in the Dutch market is that price changes by 25% per time based on the general price level of the retailer that is 10% lower than the average from any online retailers.
The role of dynamic pricing relative to retailer characteristics
choice of purchase scenarios. Although price variation is actually the only attribute that directly influences the final purchase prices, making purchase decisions based on the general price levels is more intuitive and effortless for the consumers. Combining the findings, it is concluded that pricing dynamics, namely price changing frequency and price variation, are not critical to consumers’ purchase decisions in e-commerce.This finding is completely different from most of the findings from previous research about dynamic pricing based on price fairness perceptions in e-commerce, e.g., Haws et al. (2006). It raised a critical question- what are the reasons that the finding does not correspond to the others from previous research? One reason certainly is that the authors did not include economic antecedents and retailer characteristics in evaluating consumers’ preference of dynamic pricing. That is, they exclusively asked respondents about the dimension of price fairness and it is the essential difference between this paper and the others.
Future studies might deep-dive beyond our reasoning and investigate further the underlying psychological process that resulted in the emphasis on retailer characteristics in dynamic pricing purchase scenarios, not on the pricing dynamics. The theories proposed in this paper to explain the results are not empirically examined. Therefore, it is suggested for future research to propose the possible antecedents of preference for dynamic pricing purchase scenarios and link them as covariates to the preference measurement to find out the underlying reasons about consumers’ preference. Furthermore, it is also suggested for future research to measure the probabilities of choices on the dynamic pricing purchase scenarios to simulate the acceptability of dynamic pricing. On other hand, as the general price level of the retailer is proved to be a decisive attribute in this paper which triggers consumer price consciousness, it is also suggested to exclude the attribute to measure the respective influences of the 7 dimensions in retail brand equity proposed by Anselmsson et al. (2017) with the pricing dynamics on consumers’ choice of dynamic pricing purchase scenarios.
simple pricing policies more than the highly dynamic ones in general but dynamic pricing is not decisive to consumers’ purchase decisions in e-commerce. The findings imply the viability of dynamic pricing for online retailers.
5.2 Limitation
As with every research, certain limitations exist in this research paper. A potential issue is that transforming dynamic pricing scenarios as alternatives in choices set is rather an abstract concept which requires relatively more patience and cognitive efforts for participants to understand. They might feel information overload and then impatiently filling the survey which caused incorrect data. The issue in return might have an effect on the reported finding. Based on the high heterogeneity found in consumers’ preference, another potential issue is that the sample might not be representative enough of the Dutch market due to the limited sample size and the composition of respondents’ nationalities. An attempt to diversify the sample was done but it was not entirely successful in bringing an exact representation of the Dutch population. Thus, the results of consumers’ preference might not be representative enough of the Dutch market. Another potential issue is that only a small selection of possible dynamic pricing strategies is analyzed. When choosing the programs to be included, only the ones that are often being discussed and that are relatively easy to understand by the participants are chosen. One suggestion for future researchers is to measure consumers’ preference on different types of dynamic pricing strategies(e.g., customer-based, time-based) or considering other consumer characteristics(e.g, website usage behaviors). Last but not least, this study was conducted with a time and geographical limit, which limits the findings. Adaptability of this research paper to other countries needs further academic ongoing research. Future studies can focus on the patterns in consumer preference on dynamic pricing that can be generalized in other markets.
5.3 Managerial Implications
individually.Thereby, in case of causing price unfairness perception, online retailers are still suggested to conservatively apply simpler dynamic pricing policies.
For the retailer benefits, the majority of consumers take the general price level of the retailer into account from the choice of shopping platforms. Firms that use dynamic pricing and want to attract the most customers are suggested to optimize the price of all the products/ services provided to create a lower price level in general. However, it is a less positive solution due to the fact that online retailers would have to sacrifice a large part of profits and need to negotiate with all the suppliers. Therefore, a better solution for online retailers is to create the perception of low price level as an anchor. Studies by Adaval and Wyer (2011) shows that the deliberate consideration of price anchors that can play a key role in whether the effect of the anchors will generalize across product categories. Combining the study with the results, the perceived lower general price level is highly likely to have a positive effect on consumers’ purchase decisions in e-commerce. Online retailers are also suggested to increase their brand equity as nearly 40% of consumers have online retail brand preference. With a higher exposure, consumers would be more familiar with the online retail brand and more likely to commonly associate the brand during shopping. In addition, a positive brand image automatically activates the positive associations which spurs consumers to buy products (Fitzimons, Chartrand and Fitzsimons 2008). Furthermore, Consumers are proven to react more favorably to an element of the marketing mix for the brand with positive consumer-based brand equity (Kevin Keller, 2013). As a result, it is highly recommended for online retail brands using dynamic pricing to create more exposure and maintain a positive brand image.
Reference
Aaker, D.A. and Equity, M.B., 1991. The Free Press. New York, 206.
Aaker, D.A., 1996. Measuring brand equity across products and markets. California management review, 38(3). Ailawadi, K.L. and Keller, K.L., 2004. Understanding retail branding: conceptual insights and research priorities.
Journal of retailing, 80(4), pp.331-342.
Allenby, G.M. and Rossi, P.E., 1998. Marketing models of consumer heterogeneity. Journal of econometrics,
89(1-2), pp.57-78.
Anselmsson, J., Burt, S. and Tunca, B., 2017. An integrated retailer image and brand equity framework:
Re-examining, extending, and restructuring retailer brand equity. Journal of retailing and consumer services, 38, pp.194-203.
Angwin, J. and Mattioli, D., 2012. Coming soon: Toilet paper priced like airline tickets. The Wall Street Journal, 5. Aribarg, A., Arora, N. and Bodur, H.O., 2002. Understanding the role of preference revision and concession in group decisions. Journal of Marketing Research, 39(3), pp.336-349
Bae, S. and Lee, T., 2011. Gender differences in consumers’ perception of online consumer reviews. Electronic
Commerce Research, 11(2), pp.201-214.
Backhaus, K., Becker, J., Beverungen, D., Frohs, M., Müller, O., Weddeling, M., Knackstedt, R. and Steiner, M., 2010. Enabling individualized recommendations and dynamic pricing of value-added services through
willingness-to-pay data. Electronic Markets, 20(2), pp.131-146.
Bayaga, A., 2010. MULTINOMIAL LOGISTIC REGRESSION: USAGE AND APPLICATION IN RISK ANALYSIS. Journal of applied quantitative methods, 5(2).
Bechwati, N.N. and Morrin, M., 2003. Outraged consumers: Getting even at the expense of getting a good deal.
Journal of Consumer Psychology, 13(4), pp.440-453.
Bougie, R., Pieters, R. and Zeelenberg, M., 2003. Angry customers don't come back, they get back: The experience and behavioral implications of anger and dissatisfaction in services. Journal of the Academy of Marketing science,
31(4), pp.377-393.
Brynjolfsson, E. and Smith, M.D., 2000. Frictionless commerce? A comparison of Internet and conventional retailers. Management science, 46(4), pp.563-585.
Campbell, M.C., 1999. Perceptions of price unfairness: antecedents and consequences. Journal of marketing
research, 36(2), pp.187-199.
Chandrashekaran, R. and Grewal, D., 2006. Anchoring effects of advertised reference price and sale price: The moderating role of saving presentation format. Journal of Business Research, 59(10-11), pp.1063-1071.
de Bekker-Grob, E.W., Ryan, M. and Gerard, K., 2012. Discrete choice experiments in health economics: a review of the literature. Health economics, 21(2), pp.145-172.
Dittmar, H., Long, K. and Meek, R., 2004. Buying on the Internet: Gender differences in on-line and conventional buying motivations. Sex roles, 50(5-6), pp.423-444.
Dütschke, E. and Paetz, A.G., 2013. Dynamic electricity pricing—Which programs do consumers prefer?. Energy
Policy, 59, pp.226-234.
E-commerce news: Top 10 online stores in Europe (2019) https://ecommercenews.eu/top-10-online-stores-in-europe/
Eggers, F. and Sattler, H., 2011. Preference measurement with conjoint analysis. Overview of state-of-the-art approaches and recent developments. GfK Marketing Intelligence Review, 3(1), pp.36-47.
Eggers, F., Sattler, H., Teichert, T. and Völckner, F., 2018. Choice-Based Conjoint Analysis. Handbook of Market
Research, pp.1-39.
Eggers, F., Hauser, J.R. and Selove, M., 2016, September. The effects of incentive alignment, realistic images, video instructions, and ceteris paribus instructions on willingness to pay and price equilibria. In Proceedings of the
Sawtooth Software conference.
Elmaghraby, W. and Keskinocak, P., 2003. Dynamic pricing in the presence of inventory considerations: Research overview, current practices, and future directions. Management science, 49(10), pp.1287-1309.
Fang, J., Wen, C., George, B. and Prybutok, V.R., 2016. Consumer heterogeneity, perceived value, and repurchase decision-making in online shopping: The role of gender, age, and shopping motives. Journal of Electronic
Commerce Research, 17(2), p.116.
Farquhar, P.H., 1989. Managing brand equity. Marketing research, 1(3).
Finkel, N.J., 2001. Not fair!: The typology of commonsense unfairness. American Psychological Association. Fitzsimons, G.M., Chartrand, T.L. and Fitzsimons, G.J., 2008. Automatic effects of brand exposure on motivated behavior: how apple makes you “think different”. Journal of consumer research, 35(1), pp.21-35.
Folkes, V.S., Koletsky, S. and Graham, J.L., 1987. A field study of causal inferences and consumer reaction: the view from the airport. Journal of consumer research, 13(4), pp.534-539.
Folkes, V.S., 1990. Conflict in the marketplace: explaining why products fail. Attribution Theory: Applications to
Achievement, Mental Health, and Interpersonal Conflict, pp.143-160.
Furnham, A. and Boo, H.C., 2011. A literature review of the anchoring effect. The Journal of Socio-Economics,
40(1), pp.35-42.
Franciosi, R., Kujal, P., Michelitsch, R., Smith, V. and Deng, G., 1995. Fairness: effect on temporary and equilibrium prices in posted-offer markets. The Economic Journal, 105(431), pp.938-950.
Garbarino, E. and Lee, O.F., 2003. Dynamic pricing in internet retail: effects on consumer trust. Psychology &
Marketing, 20(6), pp.495-513.
Garbarino, E. and Maxwell, S., 2010. Consumer response to norm-breaking pricing events in e-commerce. Journal
